Interaction of membrane-spanning with peripheral and -anchored membrane proteins: perspectives from ± lipid interactions (Review)

D. Marsh{*, L. I. HorvaÂth{, M. J. Swamy}, translational diffusion in fluid bilayers, they can be distin- S. Mantripragada} and J. H. Kleinschmidt# guished from the latter in the spin label ESR spectrum. This is possible whenever the lipid rotational mobility differs substantially in the two environments (see figure 1). Then the {Max-Planck-Institut fuÈr biophysikalische Chemie, Abt. stoichiometry and specificity of lipid ± protein interaction can Spektroskopie, 37070 GoÈttingen, Germany be determined by using difference spectroscopyto quantitate the `free’ and protein-interacting lipid populations (Marsh {Institute of Biophysics, Biological Research Centre, 6701 1989). Szeged, Hungary Analysis of the lipid stoichiometry (Nb) and specificity (Kr) from the two-component ESR spectra uses the equation for }School of Chemistry, University of Hyderabad, Hyderabad equilibrium exchange association. The fraction, f, of spin- 5000046, India labelled lipid that is motionally restricted at one of the Nb sites at the intramembranous surface of the protein is given by }Skye Pharma, San Diego, CA 92121, USA (see Marsh 1985):

#UniversitaÈt Konstanz, Fachbereich Biologie, 78457 f NbKr l nt Nb Kr 1 1 ˆ ‰ ‡ … ¡ †Š … † Konstanz, Germany

where Kr is the average association constant of the spin- Summary labelled lipid relative to the host lipid and nt is the total lipid/ Studies of lipid ± protein interactions in double-reconstituted protein ratio. ESR titrations performed by varying the lipid/ systems involving both integral and peripheral or lipid-anchored protein ratio confirm that a fixed stoichiometry, Nb, is proteins are reviewed. Membranes of dimyristoyl phosphatidyl- maintained independent of nt (Brotherus et al. 1981), and glycerol containing either proteolipid protein or cyto- yield values for both N and the mean relative association chrome c oxidase were studied. The partner peripheral proteins b bound to these membranes were or constant Kr. In general, Kr&1 for a particular spin-labelled cytochrome c, respectively. In addition, the interactions between lipid (e.g. phosphatidylcholine) relative to its unlabelled the myelin proteolipid protein and avidin that was membrane- parent host lipid. This allows the determination of stoichio- anchored by binding to N-biotinyl phosphatidylethanolamine metries from measurements at a single lipid/protein ratio. were studied in dimyristoyl phosphatidylcholine membranes. It is expected that the lipid stoichiometry at the intramem- Steric exclusion plays a significant role when sizes of the peripheral protein and of the integral branous perimeter is related directly to the transmembrane protein are comparable. Even so, the effects on avidin-linked structure and degree of oligomerization of the protein (Marsh are different from those induced by myelin basic protein on 1997). Figure 2 compares ESR-determined values of Nb with freely diffusible lipids, both interacting with the myelin proteo- predictions for simple, regular helix packing arrangements. lipid protein. Both the former and the cytochrome c/cytochrome oxidase couple evidence a propagation of lipid perturbation out For a single monomeric helix ( mutant), and from the intramembrane protein interface that could be a basis for a monomeric 7-helix sandwich (), the predic- for formation of microdomains. tions work well. For proteolipid hexamers (PLP and 16kD), the stoichiometries per monomer are predictably reduced. Keywords: Lipid ± protein interactions, spin labels, electron spin For large, highly polytopic proteins such as cytochrome resonance, myelin proteins, cytochrome oxidase, avidin-. oxidase (or ), with more complex and/ or less regular transmembrane arrangements, the simple models do not apply. Further discussion of this topic can be Introduction found in Marsh (1997) and Marsh and Horva th (1998). Electron spin resonance (ESR) with spin-labelled lipids at It is further expected that the selectivity patterns of lipid ± probe amounts has proved to be one of the most valuable protein interaction will reflect in detail the structure and spectroscopic methods for studying lipid ± protein interac- sequence of those sections of transmembrane protein tions in biological membranes (see e.g. Marsh 1985). This is segments that are located in the vicinity of the lipid because the dynamic sensitivity of spin-label ESR is headgroups. This expectation is borne out in practice by optimally matched to the timescale of the rotational motions the results of ESR studies. Figure 3 gives the values of Kr for of the lipid chains in biological membranes, which lies in the representative lipids interacting with a range of different ns range. Even if the exchange rate of the lipids at the membrane proteins. The lipid selectivity patterns differ for the intramembranous surface of an integral protein is as fast as different proteins. Whereas the highest selectivitiesare found for anionic lipids, they are not identical for lipids with the same formal charge, nor does a single lipid display the *To whom correspondence should be addressed. highest selectivity for all proteins. Cytochrome oxidase e-mail: [email protected] displays its highest selectivity for , a lipid unique

Molecular Membrane Biology ISSN 0968-7688 print/ISSN 1464-5203 online # 2002 Taylor & Francis Ltd http://www.tandf.co.uk/journals DOI: 10.1080/09687680210162419 248 D. Marsh et al.

Figure1. Schematic indication of the rotational mobility of the lipid andprotein components in membranes. The grey/ colourscale representsthe rotational correlation time of the different compo- nents.Motional restriction of the spin-labelled lipids in direct contact withthe transmembrane four-helix bundle results in a two-compo- nentESR spectrumfrom which the number of first-shell lipids is quantitated.The transmembrane topology of the myelin proteolipid Figure2. Dependence of the number of first-shell lipids, Nb per proteinis that determined in Weimbs and Stoffel (1992). The polar proteinmonomer, on the actual or predicted number of transmem- loopdeleted in theDM-20 isoform is also indicated. branehelices, na,forthe following integral proteins: M13, phage coat protein;PLB, L37A mutant of phospholamban; PLP, myelin proteolipidprotein; 16 kD, 16-kDa proteolipid from Nephrops; ADP, ADP-ATP carrier;Rho, rhodopsin; Ca, Ca-ATPase; NaK, Na,K- tomitochondria in eucaryotes, whereas rhodopsin displays ATPase;CR, cytochromereductase; AChR, acetylcholinereceptor; littleselectivity for any particular lipid species. This topic is CO, cytochromeoxidase. See Marsh (1997) for references. dealtwith further in Marsh (1995) and Marsh and Horva  th Predictionsfor helical sandwiches or regular polygonal arrange- (1998). ments(solid line) are given, where Da (=1 nm) and dch (=0.48 nm) Figures2 and3 showthat the database on lipid arethe diameters of an a-helixand a lipidchain, respectively. Dotted anddashed lines are corresponding predictions for protein dimers interactionswith single integral species andhexamers, respectively. isreasonably extensive. As already implied, this area has beenreviewed several times, one of themost recent being in Marshand Horva th(1998), which also includes consideration ofthe penetration of soluble proteins into membranes. A forbiotin, this system serves as a modelfor covalent lipid considerablebody of work also exists on lipid interactions anchors.Also, because phospholipases must bind specifi- withsingle peripheral proteins. Some of thiswas reviewed in callyto thelipid headgroups, the avidin/ biotinlipid interaction Sankaramand Marsh (1993) and (Marsh (1995). Far fewer maybe used to model certain features of their mode of biophysicalstudies have been undertaken on the mutual interaction. interactionsof integral and peripheral proteins in defined Theway in which the lipid chains anchor the water-soluble reconstitutedsystems. This review is devoted specifically to avidinprotein to membranes was studied by using biotin- thisaspect of protein ± lipidinteractions, including protein- phosphatidylethanolamines(biotin-PE) that are spin-labelled linkedlipid chains. The latter are considered first, in the in their sn-2chain (Swamy and Marsh 1997). These absenceof a secondprotein, by using avidin bound to N- biotinylatedlipids were incorporated in membranes of biotinylphosphatidylethanola mineas amodelsystem. Then, phosphatidylcholine,an inert host lipid that alone does not themyelin basic protein/ myelinproteolipid and cytochrome bindavidin. Figure 4 givesthe lipid chain flexibility profile as a c/cytochromeoxidase couples are treated, followed finally by functionof position, n,ofspinlabelling, and the modification lipid-linkedavidin interacting with the proteolipid protein. ofthis profile by binding avidin. A systematicdecrease in the

spectralouter hyperfine splitting, Amax,isfound as the spin- labelis stepped down the chain towards the centre of the Avidin/biotin-lipidmembrane anchoring membrane.This is a measureof the angular amplitude of Phosphatidylethanolaminesthatare N-derivatisedwith biotin motionalfreedom of the lipid chain segments. The char- canspecifically bind the protein avidin (see e.g. Swamy and acteristicflexibility gradient thus registered by the spin- Marsh2001). Because of the extremely high affinity of avidin labelledlipid is preserved on binding avidin to the biotin Protein± lipidinteractions 249

Figure3. Selectivity patterns for interaction of integral membrane proteinswith the following lipids: PA, phosphatidicacid; CL, cardiolipin;SA, stearic acid; PS, phosphatidylserine, PG, phospha- tidylglycerol;PE, phosphatidylethanolamine;PC, phosphatidylcho- line.Plotted is the association constant, Kr,ofthe various spin- labelledlipid species, relative to the host (usually PC). SeeMarsh and Horva Âth(1998)for references.

Figure4. Spin-labelled N-biotinylphosphatidylethanolamine, with (filledcircles) and without (open circles) bound avidin, in dimyristoyl headgroup,but the overall size of the outer hyperfine phosphatidylcholinemembranes. Spin-label outer hyperfine splitting splittingsincreases dramatically. This direct and highly constant, Amax,isplotted against spin-label position, n, in the sn-2 specificeffect of the protein ± lipidinteraction corresponds chain.Open squares are for spin-labelled phosphatidylcholine. (Data reproducedwith permission from Swamy and Marsh (1997)). The toan upward shift in the flexibilityprofile by *7 CH2 units. As structureof biotin-bound is from PDB:1STP (Weber et indicatedin figure 4, this translates to anupwardmovement al.1989).The location of the lipid surface is approximately that ofthe biotin lipid, relative to the phosphatidylcholine determinedin Darst et al. (1991). membrane,by 0.7 ± 0.8nm. Electron crystallographic studies ontwo-dimensional arrays of streptavidinbound to biotinlipid haveshown that the carboxyl group of the bound biotin is hydrolysis.The use of the biotin-PE/ avidincouple as a positioned *0.8nm from the surface of the supporting lipid modelfor a lipid-anchoredprotein will bereturned to in a later layer(Darst et al.1991).This is consistent with the vertical section. movementsuggested by the spin-label ESR studies. Directevidence for an upward vertical movement of the Myelinbasic protein/ myelinproteolipid protein biotinlipid on complexation with avidin comes from ESR interactions measurementsof spinlabel relaxation enhancement induced byparamagnetic species (Arora and Marsh 1998). The Myelinproteolipid (PLP) is the major transmembrane protein relaxationenhancement of biotin-PEspin labels by aqueous andmyelin basic protein (MBP) isthe major peripheral 2+ Ni isincreased, and that by lipid-soluble O 2 isdecreased, proteinin the membranes of the myelin sheath. Together, 2+ onbinding avidin. The ratio of the Ni /O2 enhancements they make up *80%ofthe protein of central nervous system increasesfrom 0.5 to 2.6 for a spinlabel on the C-8 atom of myelin,in which they are present in roughly equimolar biotin-PE,on binding avidin. amounts.The proteolipid protein has a molecularmass of Thistype of vertical displacement is likely to bea general *25kDa and is thought to be composed of a four-helix featureof the interaction with proteins, e.g. cholera toxin or bundle,with the principal polar loop connecting helices II and anti-cardiolipinantibodies, that bind to lipidheadgroups. The III(Weimbsand Stoffel 1992). There are three enzymesof hydrolysis are a furtherclass of sitesin this loop, in addition to three further sites in the N- examples.In phospholipase A 2,theactive site is located terminalsection. A stretchof 34 contiguous residues *1.5nm from the protein surface (Scott et al. 1990). A (Val116-Lys150)in the II ±IIIloopcontains six basic and movementof the phospholipid substrate out of the mem- twoacidic residues, together with four histidines; this is brane,through a hydrophobicchannel that leads to the deletedin the DM-20 isoform of theproteolipid (see figure 1). catalyticsite, is, therefore, required for the enzymatic Thebasic protein is a water-solubleprotein of molecular 250 D. Marsh et al. mass *18.4kDa, which has little tertiary structure in solution Doublereconstitutions, involving the binding of MBP to but adopts a-helicaland b-sheetstructure on binding to dimyristoylphosphatidylglycerol (DMPG) membranesthat membranes(Surewicz et al.1987).MBP contains31 containPLP, were used to study the mutual protein ± lipid positivelycharged residues that are distributed roughly interactionsof integral and peripheral proteins (Sankaram et evenlythroughout the sequence, and has a pIof 10.1. In al.1991).Figure 5 givesthe saturation binding stoichiometry addition,there is a lipidattachment site for phosphatidyl ofMBP (perlipid) as a functionof PLP contentin the inositol4,5-biphosphate at residue Ser54, which is also a membrane.Up toa criticalPLP/ DMPG ratio,the binding of potentialphosphorylation site (Chang et al. 1986). MBP isessentially undisturbed by the presence of PLP. PLP proteolipidreconstituted in phospholipid bilayer Beyondthis critical content, the saturation binding of MBP membranesby dialysis from 2-chloroethanol motionally decreasesroughly linearly with PLP content,reaching zero at restricts11 lipids per monomer and displays a pronounced aDMPG/PLP ratioof N1&11.The latter corresponds also selectivityfor various negatively charged withthe number of lipids Nb&11thatare motionally restricted (Brophy et al.1984).A largepart of this lipid selectivity byPLP alone(see above). Apparently, this first boundary arisesfrom the polar sequence that is deleted in DM-20 shellof lipidsis unavailablefor bindingto MBP, oncethe PLP (Horva th et al.1990).The basic protein (MBP) bindsto densityin themembrane exceeds a criticalvalue. The critical bilayermembranes that are negatively charged, with a PLP contentcorresponds to a DMPG/PLP ratioof Nc&37. bindingstoichiometry of 36 lipids/ proteinat saturation Thisalso corresponds to thenumber of lipids interacting with (Sankaram et al.1989a).The surface binding of MBP MBP atsurface saturation (see above). Therefore, the preservesthe characteristic flexibility gradient of spin- minimumcritical area of lipid surface between PLP assem- labelledlipid chains. Mobility is decreased uniformly through- bliesthat is required for undisturbed binding of MBP is outthe length of the chain as a resultof the increased lipid comparableto the size of thebasic protein itself. This simple packingdensity induced by surfaceassociation of MBP. The situationof mutual steric exclusion between the two proteins resultingincrease in ESR spectralanisotropy, DAmax, can be isillustrated schematically in figure 5. A phenomenological usedas a measureof bindingextent (Sankaram et al. 1989b) descriptionof the decrease in MBP bindingstoichiometry is andselectivity of lipid interaction (Sankaram et al. 1989c). possible,if proportionality to the total fraction of available

Figure5. Steric exclusion between the integral proteolipid protein (PLP) and peripheral basic protein (MBP) in dimyristoyl phosphatidylglycerol (DMPG) membranes.A criticalnumber of lipids, Nc perPLP, is requiredfor undisturbed surface binding of MBP. For total number of lipids nt 5Nc, MBPbindingdecreases approximately linearly with 1/ nt,reachingzero at nt=N1.Thephenomenological equation given in the inset assumes that Protein± lipidinteractions 251

Table1. Increase in maximum hyperfine splitting, A , of lipids,viz., ( nt N1)/nt,isassumed. This leads to theequation D max givenin the inset to figure 5 andresults in the functional phosphatidylglycerolspin-labelled at the 5-C atomof the sn-2 chain, onsaturation binding of myelin basic protein to membranes of dependenceshown by the solid lines. dimyristoylphosphatidylglycerol (DMPG) containingthe myelin Accompanyingthe steric exclusion is a reductionby 30 ± proteolipidprotein (PLP). 40%in the stoichiometry of lipidsinteracting with PLP thatis A (gauss) causedby binding of MBP atsaturation. This suggests that D max MBP disturbsthe interaction of lipids with the intramembra- DMPG/PLP(mol/mol)10 m M NaCl 100 mM NaCl nousportion of PLP, quite possibly by means of the short 1 : 0 3.6 3.3 membrane-penetrantportions of the peripheral protein. 25 : 1 1.0 0.7 Unlikecytochrome c,MBP interactsdirectly with the lipid 17 : 1 2.3 1.2 chainsto a limitedextent, in addition to the surface Valuesare given for different ionic strengths (NaCl) ofthe electrostaticassociation (see Sankaram et al. 1989a). In suspendingmedium and T=308C.Datafrom Sankaram et al. (1991). thisrespect, MBP resemblesthe precursor protein apocy- tochrome c (GoÈrrissen et al. 1986). Table2. Relative association constants, Kr,ofspin-labelled lipids Inaddition to steric exclusion between the two proteins, withmyelin proteolipid protein (PLP) in dimyristoyl phosphatidylgly- thereare also other mutual influences on the lipid ± protein cerolmembranes, with and without myelin basic protein (MBP). interactions.Table 1 givesthe increase in spectral aniso- PC Kr /Kr tropy, DAmax,ofspin-labelledphosphatidylglycero lonsatura- DDGL tionbinding of MBP. Thiseffect on lipid mobility is Lipidspin label PLPPLP+MBP(kJ/ mol) substantiallyattenuated in the presence of the proteolipid stearicacid 2.9 2.3 0.6 protein.At a DMPG/PLP ratioof 25:1, the weakening of the cardiolipin 3.0 2.7 0.3 MBP-DMPG interactionis mostly a consequenceof the phosphatidicacid 2.4 1.7 0.8 reducedbinding of MBP (cf.figure 5). At the lower DMPG/ phosphatidylglycerol 2.0 1.8 0.3 phosphatidylserine 1.4 1.1 0.6 PLP ratio,the lipid perturbation is reduced less, in spite of phosphatidylethanolamine1.7 1.2 0.9 furtherdecrease in MBP binding.This undoubtedly results K isnormalized to the corresponding value, K PC,forspin-labelled froma mutualreinforcing of the lipid perturbation by both r r phosphatidylcholine. DDGL isthe increase in free energy of lipid PLP andMBP, athigh protein/ lipidratios. associationon bindingMBP. Data from Sankaram et al. (1991). Table2 givesthe effect of MBP bindingon the selectivity PC oflipid interactions with PLP. Selectivities (i.e. Kr/Kr ) of all lipidsfor PLP arereduced on binding MBP, particularlythe Table3. Fraction, f,ofspin-labelledlipid, 14-BPESL (biotin-PE) or 14-PCSL(PC), thatis motionally restricted by interacting with PLP negativelycharged ones. This weakening of the selectivity (orDM-20) in DMPC membranes(lipid/ protein=37: 1mol/mol),in the andmodification of the selectivity ranking can be understood presenceand absence of excess avidin. asa directcompetition for the lipids by the two proteins. The f relativeassociation constants, Kr,forinteraction of lipidswith PLP thatare given in table 2 aredefined relative to DMPG as Protein Spin label 7 avidin + avidin standardstate in one case, but relative to DMPG withMBP PLP 14-BPESL 0.51 0.80 boundin the other. 14-PCSL 0.29 ± a It,therefore, seems that MBP andPLP probablyinteract DM-20 14-BPESL 0.57 0.79 independently,rather than synergistically, in their role in the 14-PCSL 0.30 ± a compactionof nerve myelin. In particular, the double aAvidinhas no effect on 14-PCSLin DMPC membranes.Data from reconstitutionsoffer little evidence for the direct binding of Swamy et al. (1999). MBP toPLP thathas been suggested in some models.

ablepenetration of cytochrome c intothe hydrophobic core Cytochrome c/cytochromeoxidase interactions (Kostrzewa et al. 2000). Cytochrome c andcytochrome oxidase form the terminal Themutual interactions of this integral/ peripheralprotein redoxcouple of the mitochondrial , couplewere studied in double reconstitutions, in which endingin the reduction of oxygen to water. Eucaryotic cytochrome c isbound to dimyristoyl phosphatidylglycero l cytochromeoxidases are large ( Mr *200kDa) integral membranescontaining cytochrome oxidase (Kleinschmidt et proteinscomposed of 13 sub-units and containing 28 al.1998).The binding of cytochrome c occurs in an transmembranehelices. Cytochrome c isa small,water- unrestrictedfashion with a saturationstoichiometry of 9 soluble,basic of molecular mass DMPG percytochrome c forall realizable cytochrome

*11.7kDa. Cytochrome oxidase motionally restricts Nb=55 oxidasecontents. This is unlike the situation with the myelin lipidsper 200 kDa protein, at its intramembranous surface basicprotein/ proteolipidprotein couple already described, (Knowles et al.1979).It displays a pronouncedselectivity for because,in this case, the peripheral protein is muchsmaller anioniclipids, particularly cardiolipin (Knowles et al. 1981). thanthe integral protein (see figure 6). The ratio of the Cytochrome c bindsto negatively charged lipid membranes numberof lipidsper MBP atsaturationbinding to thenumber witha stoichiometryat saturation of nine lipids per bound offirst-shell lipids motionally restricted by PLP is36/ 11 *3, protein(Go Èrrissen et al.1986,Sankaram and Marsh 1993). whereasthe corresponding ratio for cytochrome c and Bindingis to the surface of the membrane without appreci- cytochromeoxidase is 9/ 55 *0.2.With decreasing lipid 252 D. Marsh et al. content,steric interactions between cytochrome oxidase Figure6 givesthe population of motionallyrestricted lipid, moleculesbecome limiting before these can affect the normalizedto cytochrome oxidase content, as a functionof bindingof cytochrome c appreciably.Indeed, for efficient totallipid (DMPG)/ cytochromeoxidase ratio, nt.Theordinate electrontransfer, cytochrome oxidase must be fully acces- isthe fraction, f,ofspin-labelled phosphatidylglycerol that is sibleto cytochrome c atthe rather high protein packing motionallyrestricted, multiplied by nt.Inthe absence of densitiesof the mitochondrial inner membrane. cytochrome c,thissimply gives the number of lipid sites in

Figure6. Number of motionally restricted lipids, Nb percytochrome oxidase, in the absence (open circles) and presence (filled circles) of cytochrome c boundat saturationto reconstituted membranes with dimyristoyl phosphatidylglycerol/ cytochromeoxidase ratio, nt. Assuming Kr=1 forspin-labelled phosphatidylglycerol, Nb=f6nt, where f isthe fraction of motionally restricted spin-labelled lipids obtained from the ESR spectrum(data reproduced with permission from Kleinschmidt et al.(1998)).Structure of thetransmembrane domain of the cytochrome oxidase dimeris from PDB:1OCC (Tsukihara et al.1996),and of cytochrome c isfrom PDB:1AKK (Banci et al.1997).Orientation of cytochrome c at the membranesurface is that determined in Kostrzewa et al. (2000). Protein± lipidinteractions 253 thefirst shell at the intramembranous perimeter of cyto- Datagiven in table 3 showthat spin-labelled biotin-PE chromeoxidase. A constantvalue of Nb8=55lipids/ cyto- alonedisplays a selectivityover phosphatidylcholine for chromeoxidase, independent of nt,isobtained from the data interactionwith the proteolipid protein. In part, this may be infigure 6. This confirms that Kr=1 forspin-labelled becausebiotin-PE is a negativelycharged lipid. Saturation phosphatidylglycerolrelative to DMPG, inaccordance with bindingof avidin to the biotin-PE headgroups has rather equation(1), and is in quantitative agreement with similar dramaticeffects that at first sight are perhaps somewhat measurementson phosphatidylcholine membranes surprising.It leads to a verysubstantial increase in the (Knowles et al. 1979). populationof motionally restricted biotin-PE chains (see table Incytochromeoxidase-DMPG membranes saturated with 3).Approximately 80% of the avidin-linked chains are cytochrome c,thenumber of motionally restricted lipids restrictedin membranes with a DMPC/PLP molarratio of increasesprogressively with the DMPG/ cytochromeoxidase nt =37:1.This increased population, however, exhibits a ratio(see figure 6). The population of motionally restricted lesserdegree of chain motional restriction than is typical for lipidis greater, the greater is the cytochrome c/cytochrome first-shellboundary lipids, because it displays a steeper oxidaseratioÐ at least up to the highest value of nt given in temperaturedependence (Swamy et al. 1999). figure6. Again, this situation is totally unlike that which Therelatively high effective stoichiometry of interaction obtainswith the MBP-PLP coupledescribed earlier. In the canbe explained when allowance is made for the closest lattercase, the number of lipidsthat are motionally restricted interactiondistance between the lipid-anchored avidin byPLP isactually reduced by the binding of MBP. The tetramerand the transmembrane proteolipid hexamers, synergisticeffect of cytochrome c canbe rather large; the withoutany specific interaction between the two types of motionallyrestricted lipid population (per cytochrome oxi- membrane-associatedproteins. As figure 7 indicates,the dase)is more than doubled at thehighest lipid/ proteinratios, PLPhexamerand avidin tetramer are of comparable nt.Therefore,it is most likely that surface-bound cytochrome c propagatesthe chain restriction induced by cytochrome oxidaseto the second and possibly even third shell of lipids surroundingthe protein. As already mentioned, cytochrome c bindsto the first shell of lipids directly in contact with cytochromeoxidase. Because it binds to about nine DMPG molecules,it could, therefore, form a surface-bridgebetween lipidsin the first and adjacent boundary shells. Such a propagationmechanism is possible because, although cytochrome c alonedoes not create a specificpopulation ofmotionally restricted lipids, it does induce a generalized motionalrestriction of alllipids to whichit binds at saturation (GoÈ rrissen et al. 1986). Cytochrome c,therefore,can induce formation of micro- domainscontaining cytochrome oxidase in which the mobility ofthe lipid chains is appreciably restricted relative to that in fluidmembranes. Comparison with the MBP-PLP couple suggeststhat pre-conditions for this mechanism are that the inducingprotein binds the lipids directly contacting the integralprotein, and possibly also that it does not itself penetratethe membrane. A qualitativelysimilar phenomenon wasobserved in the interaction of melittin with the Ca- ATPaseand associated lipids in sarcoplasmic reticulum membranes(Mahaney et al. 1992).

Avidin-linkedchains/ proteolipidinteractions Theuse of avidin/ biotin-PEconjugates as a modelfor proteinsthat are anchored to membranes by covalently linkedchains, e.g. those of glycosyl phosphatidyl inositol (GPI),was described already. Characterization of themyelin proteolipidas an integral transmembrane protein has also beengiven. These two different protein classes were used to investigatethe interactions of protein-linkedlipid chains with Figure7. Steric exclusion between the myelin proteolipid hexamer transmembraneproteins by using spin-label ESR spectro- (PLP)and lipid-linked avidin tetramer. The avidin-linked chains ( *) scopy.Proteolipid (PLP or DM-20) was reconstituted in canapproach only to within distance rp (& 2.5nm) of the PLP dimyristoylphosphatidylcholine (DMPC) membranesin perimeter.The number of lipids, Nb,thatcan be accommodated at thisdistance is given in the figure. nagg=6is the aggregation number which *1mol%biotin-PE spin label was incorporated and ofPLP and other symbols are defined in the legend to figure 2 boundby saturatingquantities of avidin (Swamy et al. 1999). (adaptedwith permission from Swamy et al. (1999)). 254 D. Marsh et al. cross-sectionalarea, and the biotin-lipid binding sites are Acknowledgements offsetfrom the PLP-lipid interface by *2.5nm (Hendrick- Wethank Frau B. Angersteinfor the synthesis of spin-labelled son et al.1989).At this distance, 15 lipids per PLP phospholipidsused in our studies of lipid ± proteininteractions. We monomercan be accommodated around the perimeter of aregrateful to Dr T.PaÂlifor help with figure 6 andFrau I. Dregerfor the thehexamer, compared with *10at the true boundary preparationof the figures. Our studieson lipid interactions with V- layer.In addition, the region of that lies within ATPasesand the 16-kDa Nephrops proteinare supported by the EuropeanUnion (Contract No. QLGI-CT 2000-01801).DM andJHK thisextended perimeter is inaccessible to the avidin-linked aremembers of the European COST D22Action. chains.This region corresponds to approximately half the cross-sectionalarea of avidin, i.e. *25lipids per avidin tetramer,which translates into 17lipids per PLP * References monomer(Swamy et al.1999,Swamy andMarsh 2001). Theeffective lipid/ proteinratio `seen’ by the avidin-linked Arora,A. andMarsh, D., 1998, Protein-induced vertical lipid dislocationin a modelmembrane system: spin-label relaxation chainsis reduced by this amount and, therefore, the studieson avidin-biotinylphosphatidylethanol amineinteractions. fractionof perimeter chains is given by f=15/(nt717)=0.75. BiophysicalJournal , 75,2915± 2922. Thisis close to the experimental values of f for14-BPESL Banci,L., Bertini, I., Gray,H. B.,Luchinat, C., Reddig, T., Rosato,A. thatare given in table 3. It is, therefore, concluded that, andTurano, P., 1997,Solution structure of oxidizedhorse heart afterthe geometric extent of the lipid-anchored avidin is cytochrome c. Biochemistry , 36,9867± 9877. Brophy,P. J.,HorvaÂth,L. I.andMarsh, D., 1984, Stoichiometry and takeninto account, it displays little selective interaction specificityof lipid-protein interaction with myelin proteolipid withthe transmembrane proteolipid. proteinstudied by spin-labelelectron spin resonance. Biochem- Amajorconsequence of the above interpretation of the istry, 23, 860 ± 865. increasedstoichiometry is that the avidin-linked chains are Brotherus,J. R.,Griffith,O. H.,Brotherus, M. O.,Jost, P. C.,Silvius, J.R.andHokin, L. E.,1981,Lipid-protein multiple binding somewhatremoved from the innermost protein ± lipidinter- equilibriain membranes. Biochemistry , 20,5261± 5267. face.Nevertheless, their mobility is very markedly reduced Chang,P. C.,Yang, J. C.,Fujitaki, J. M.,Chiu,K. C. andSmith, R. relativeto that in lipid membranes without transmembrane A.,1986,Covalent linkage of phospholipid to myelin basic- protein,although to a lesserextent than for diffusible first- protein:identification of serine-54 as the site of attachment. shelllipids. The effects on the lipid chains of anchoring to Biochemistry , 25,2682± 2686. Darst,S. A.,Ahlers, M., Meller,P. H.,Kubalek, E. W.,Blankenburg, avidin(that were discussed in an earlier section) render R.,Ribi,H. O.,Ringsdorf, H. andKornberg, R. D.,1991, Two- themfar more susceptible to interactions and perturbations dimensionalcrystals of streptavidin on biotinylated lipid layers withinthe lipid matrix. This is demonstrated here by the andtheir interactions with biotinylated macromolecules. Bio- interactionswith an integral protein, but may conceivably physicalJournal , 59, 387 ± 396. GoÈrrissen,H., Marsh, D., Rietveld, A. andDe Kruijff,B., 1986, alsoextend to interactions with membrane lipids, in Apocytochrome c bindingto negatively charged lipid disper- particularwith the putative raft components sionsstudied by spin-label electron spin resonance. Biochem- andcholesterol. istry, 25,2904± 2910. Hendrickson,W. A.,Pahler, A., Smith, J. L.,Satow, Y., Merritt,E. A. andPhizackerley, R. P.,1989, Crystal-structure of core Conclusions streptavidindetermined from multiwavelength anomalous dif- fractionof synchrotron radiation. Proceedingsof the National Theinteractions between the two peripheral/ integralprotein Academyof Sciences(USA) , 86,2190± 2194. couplesconsidered here differ greatly, being determined by HorvaÂth,L. I.,Brophy,P. J.andMarsh, D., 1990, Influence of polar therelative size of the two protein partners and possibly also residuedeletions on lipid ± proteininteractions with the myelin proteolipidprotein. Spin-label ESR studieswith DM-20/ lipid bytheirfunctional roles. Myelin basic protein and proteolipid recombinants. Biochemistry , 29,2635± 2638. proteinare comparable in size and have a structuralrole in Kleinschmidt,J. H.,Powell, G. L.andMarsh, D., 1998, Cytochrome thecompaction of nerve myelin. The interaction between c-inducedincrease of motionally restricted lipid in reconstituted thesetwo membrane proteins is principally one of mutual cytochrome c oxidasemembranes, revealed by spin-label ESR stericexclusion. Cytochrome c ismuch smaller than spectroscopy. Biochemistry , 37,11579± 11585. Knowles,P. F.,Watts, A. andMarsh, D., 1979, Spin label studies of cytochromeoxidase and acts as a substrateof this multi- lipidimmobilization in dimyristoylphosphatidylcholine-s ubsti- sub-unitredox- complex. The cytochrome associates tutedcytochrome oxidase. Biochemistry , 18,4480± 4487. intimatelywith its oxidase partner and both reinforces and Knowles,P. F.,Watts, A. andMarsh, D., 1981, Spin label studies of propagatesthe influence of the integral protein on thechains headgroupspecificity in the interaction of phospholipids with yeastcytochrome oxidase. Biochemistry , 20,5888± 5894. ofthe surrounding lipids. Although the interaction of avidin Kostrzewa,A., Pa Âli,T., Froncisz, W. andMarsh, D., 2000, Membrane withthe myelin proteolipid protein appears to be one of steric locationof spin-labelled cytochrome c determinedby paramag- exclusion,the anchoring chains attached to avidin strongly neticrelaxation agents. Biochemistry , 39,6066± 6074. sensethe influence of the somewhat remote transmembrane Mahaney,J. E.,Kleinschmidt, J., Marsh, D. andThomas, D. D.,1992, sectorsof the proteolipid protein. Thus, one has two rather Effectsof melittin on lipid-protein interactions in sarcoplasmic reticulummembranes. BiophysicalJournal , 63,1513± 1522. differentmechanisms for propagating lipid interactions Marsh,D., 1985, ESR spinlabel studies of lipid ± proteininteractions. inducedby transmembraneproteins. Both are protein-driven In Progressin Protein ± LipidInteractions ,Vol.1, A. Wattsand andinvolve neither specifically cholesterol nor gel-phase lipid J.J.H.H.M.dePont, eds. (Amsterdam: Elsevier), pp. 143 ± (Schroeder et al.1994).Nonetheless, both potentially could 172. Marsh,D., 1989, Experimental methods in spin-label spectral stabilizemembrane microdomains and, hence, be operative analysis.In BiologicalMagnetic Resonance ,Vol.8, L. J. inthe formation of putative membrane rafts Berlinerand J. Reuben,eds. (New York:Plenum Publishing (Simonsand Ikonen 1997). Corp.),pp. 255 ± 303. Protein± lipidinteractions 255

Marsh,D., 1995, Specificity of lipid-proteininteractions. In Biomem- Scott,D. L.,White,S. P.,Otwinowski, Z., Yuan,W., Gelb,M. H.and branes,A.G.Lee,ed. (Greenwich, CT: JAI Press),pp. 137 ± Sigler,P. B.,1990, Interfacial catalysis: the mechanism of 186. phospholipaseA 2. Science, 250,1541± 1546. Marsh,D., 1997, Stoichiometry of lipid ± proteininteraction and Simons,K. andIkonen, E., 1997,Functional rafts in cell membranes. integralmembrane protein structure. EuropeanBiophysics Nature, 387, 569 ± 572. Journal, 26, 203 ± 208. Surewicz,W. A.,Moscarello, M. A.andMantsch, H. H.,1987, Marsh,D. andHorva Âth,L. I.,1998,Structure, dynamics and Fouriertransform infrared spectroscopic investigation of the compositionof the lipid ± proteininterface. Perspectives from interactionbetween myelin basic protein and dimyristoyl spin-labelling. Biochimicaet Biophysica Acta , 1376, 267 ± 296. phosphatidylglycerolbilayers. Biochemistry , 26,3881± 3886. Sankaram,M. B.andMarsh, D., 1993,Protein-lipid interactions with Swamy,M. J.andMarsh, D., 1997, Spin-label studies on the peripheralmembrane proteins. In New ComprehensiveBio- anchoringand lipid ± proteininteractions of avidin with N- chemistry,Vol. 25. Protein-Lipid Interactions ,A.Watts,ed. biotinylphosphatidylethanolaminesin lipid bilayer membranes. (Amsterdam:Elsevier), pp. 127 ± 162. Biochemistry , 36,7403± 7407. Sankaram,M. B.,Brophy, P. J.andMarsh, D., 1989b, Interaction of Swamy,M. J.andMarsh, D., 2001, Specific surface association of twocomplementary fragments of thebovine spinal cord myelin avidin with N-biotinylphosphatidylethanolaminemembrane as- basicprotein with phospholipid bilayers. An ESR spinlabel semblies:effect on lipid behavior and acyl-chain dynamics. study. Biochemistry , 28,9692± 9698. Biochemistry , 40,14869± 14877. Sankaram,M. B.,Brophy, P. J.andMarsh, D., 1989c, Selectivity of Swamy,M. J.,HorvaÂth,L. I.,Brophy,P. J.andMarsh, D., 1999, interactionof phospholipids with bovine spinal cord myelin basic Interactionsbetween lipid-anchored and transmembrane pro- proteinstudied by spin-label electron spin resonance. Biochem- teins.Spin-label ESR studieson avidin-biotinyl phosphatidyl- istry, 28,9699± 9707. ethanolaminein membrane recombinants with myelin proteoli- Sankaram,M. B.,Brophy,P. J.andMarsh, D., 1989a, Spin label pidprotein. Biochemistry , 38,16333± 16339. ESR studieson the interaction of bovine spinal cord myelin Tsukihara,T., Aoyama,H., Yamashita, E., Tomizaki,T., Yamaguchi, basicprotein with dimyristoylphosphatidylglycerol dispersions. H.,Shinzawa-Itoh, K., Nakashima, R., Yaono, R. andYoshika- Biochemistry , 28,9685± 9691. wa,S., 1996,The whole structure of the 13-subunit oxidized Sankaram,M. B.,Brophy, P. J.andMarsh, D., 1991, Lipid ± protein cytochrome c oxidaseat 2.8 A Ê . Science, 272,1136± 1144. andprotein ± proteininteractions in double recombinants of Weber,P. C.,Ohlendorf, D. H.,Wendoloski,J. J.andSalemme, F. myelinproteolipid apoprotein and myelin basic protein with R.,1989, Structural origins of high-affinity biotin binding to dimyristoylphosphatidylglycerol. Biochemistry , 30,5866± 5873. streptavidin. Science, 243, 85 ± 88. Schroeder,R., London, E. andBrown, D., 1994, Interactions Weimbs,T. andStoffel, W., 1992, Proteolipid protein (PLP) of CNS betweensaturated acyl chains confer resistance on myelin:positions of free, -bonded, and lipidsand glycosylphosphatidylinositol (GPI)-anchored proteins: thioester-linkedcysteine residues and implications for the GPI-anchoredproteins in liposomes and cells show similar membranetopology of PLP. Biochemistry , 31,12289± 12296. behavior. Proceedingsof the National Academy of Sciences (USA), 91,12130± 12134. Received27 March 2002, and in revised form 5 June2002.