Báo cáo khoa học: An X-ray diffraction study of model membrane raft structures doc

Three possible lamellar structures comprised of varying proportions of lipids are brainSM⁄ egg-PtdCho, brain-SM⁄ cholesterol or egg-PtdCho ⁄ cholesterol; a ternary complex of the three l

Peter J Quinn1and Claude Wolf2

1 Biochemistry Department, King’s College London, UK

2 ER7-Faculte´ de Me´decine-UPMC, APLIPID, Universite Paris 6, France

Introduction

Cell membranes, once regarded as uniform structures,

are now yielding up a complexity that is required to

explain the multiplicity of tasks they are reputed to

perform One particular function that demands a

highly specific assembly of membrane components is

the receipt and transmission of molecular signals from

one side of the membrane to the other Current

think-ing favours the, so-called, raft hypothesis, which postu-lates that the signalling elements are segregated and assembled in ordered lipid domains in the membrane [1–6] This membrane heterogeneity is rationalized on the basis that, for the efficient operation of a signalling system, the protein components must be closely associ-ated and organized in such a way that structural

Keywords

lipid rafts; liquid-ordered phase; membrane

rafts; sphingomyelin; X-ray diffraction

Correspondence

P J Quinn, Biochemistry Department,

King’s College London, 150 Stamford

Street, London SE1 9NH, UK

Fax: +442078484500

Tel: +442078484408

E-mail: p.quinn@kcl.ac.uk

(Received 6 July 2010, revised 1 September

2010, accepted 9 September 2010)

doi:10.1111/j.1742-4658.2010.07875.x

Protein sorting and assembly in membrane biogenesis and function involves the creation of ordered domains of lipids known as membrane rafts The rafts are comprised of all the major classes of lipids, including glycero-phospholipids, sphingolipids and sterol Cholesterol is known to interact with sphingomyelin to form a liquid-ordered bilayer phase Domains formed by sphingomyelin and cholesterol, however, represent relatively small proportions of the lipids found in membrane rafts and the properties

of other raft lipids are not well characterized We examined the structure

of lipid bilayers comprised of aqueous dispersions of ternary mixtures of phosphatidylcholines and sphingomyelins from tissue extracts and choles-terol using synchrotron X-ray powder diffraction methods Analysis of the Bragg reflections using peak-fitting methods enables the distinction of three coexisting bilayer structures: (a) a quasicrystalline structure comprised of equimolar proportions of phosphatidylcholine and sphingomyelin, (b) a liquid-ordered bilayer of phospholipid and cholesterol, and (c) fluid phos-pholipid bilayers The structures have been assigned on the basis of lamel-lar repeat spacings, relative scattering intensities and bilayer thickness of binary and ternary lipid mixtures of varying composition subjected to ther-mal scans between 20 and 50C The results suggest that the order created

by the quasicrystalline phase may provide an appropriate scaffold for the organization and assembly of raft proteins on both sides of the membrane Co-existing liquid-ordered structures comprised of phospholipid and cholesterol provides an additional membrane environment for assembly of different raft proteins

Abbreviations

brainSM, bovine brain sphimgomyelin; egg-PtdCho, hen egg-yolk phosphatidylcholine; GPI, glycerylphosphorylinsitol; SAXS, small-angle (1–14) X-ray scattering; WAXS, wide-angle (12–30) X-ray scattering.

changes accompanying the generation of a signal are

coupled to the transducing elements responsible for

execution of the response [7]

Critical tests of the hypothesis have largely been

aimed at characterizing the forces that govern the

crea-tion of lipid rafts rather than identifying the way in

which the signalling complexes are assembled within

the structure [8] The main obstacle to progress has

been the use of unreliable methods to isolate

mem-brane rafts The conventional protocol, irrespective of

the type of membrane, has been to recover a

mem-brane fraction that survives dissolution by Triton

X-100 treatment at 4C The integrity of this method

has recently been challenged [7] and alternative

meth-ods based on a milder detergent treatment that is more

compatible with physiological conditions have been

developed [9] The resulting membrane raft fraction

retains properties consistent with an arrangement of

constituents expected of its biological progenitor

Using such methods, it has been possible to

demon-strate that subpopulations of raft vesicles which

con-tain predominantly one surface antigen or another can

be separated by immunoadsorption Moreover,

analy-ses of the composition of these subpopulations show

that they contain different proportions of specific polar

lipids [10] The fatty acid substituents attached to

cere-brosides and sphingomyelins also differ and represent

products of different metabolic pools; they are

con-sequently remodelled via different pathways One

remarkable feature of the lipid analysis is the relatively

high proportion (20–30%) of monounsaturated polar

lipids Moreover, the proportion of polyunsaturated

molecular species of phospholipids, particularly

phos-phatidylinositols, increases following the activation of

raft proteins [11] These findings appear contrary to

the idea that cholesterol preferentially forms

liquid-ordered phases with saturated molecular species of

phospholipid [12] Taken together, these results

sup-port not only the notion that the rafts are truly

domains present in the parental membrane, but also

that the lipids are distinct in each raft population The

results also infer that membrane lipids may fulfil more

specific functions in the segregation and assembly of

protein components in the raft domains than hitherto

contemplated

Many attempts have been made to model membrane

lipid rafts, some of which are focused on gel-phase

separation of lipid mixtures comprised of molecular

species that differ in the temperature of their transition

between gel and liquid–crystal phases [13,14] The

relevance of these studies was underscored by the fact

that molecular species of sphingomyelin found in

membranes and enriched in membrane raft fractions

exhibited order–disorder transitions poised around physiological temperatures Attention switched to cho-lesterol when it was reported that the condensing effect

of sterols on phospholipids, particularly sphingomye-lins, created a bilayer phase that has properties inter-mediate between a gel and a liquid–crystal phase, referred to as a liquid-ordered phase [15,16] Choles-terol is known to be a prominent lipid component of membrane rafts irrespective of the isolation method [6,17]

A third type of lipid enriched in membrane rafts are the glycosphingolipids [18] Because the molecular species of sphingolipids are characterized by a high proportion of long N-acyl fatty acids (C-22 to C-26) it was suggested that these lipids may act to couple the two leaflets of the bilayer by interdigitation of the long chain fatty acid from one side to the other of the structure [19,20] Other suggested functions of these asymmetric lipids have been to stabilize highly curved membrane domains formed transiently in the process

of membrane budding and fusion during progress along the secretory pathway [21], or to increase hydro-carbon packing density to impede the permeability of small solute molecules [22] More recent molecular dynamics simulation studies are more equivocal on this point and although long hydrocarbon chains are able

to penetrate the opposing monolayers of fluid bilayers, the terminal region of the chain appears to be localized

in the centre of the bilayer [23] Other experimental and thermodynamic arguments have also cast doubt

on the action of long-chain molecular species of lipids

in coupling domains of bilayer structures [24,25] The role of these long-chain molecular species has now been reassessed in the light of the action of these asymmetric sphingolipids to form stoichiometric com-plexes with phospholipids that have the properties of a quasicrystalline phase [26,27]

We have undertaken an examination of the phase behaviour of ternary mixtures containing representa-tives of all the lipid classes identified in membrane raft preparations Phospholipids of biological extraction were used so that a range of molecular species of phos-phatidylcholines and sphingomyelins are present The thermotropic phase behaviour was examined in multil-amellar dispersions at temperatures spanning the phys-iological range to characterize the miscibility of the different lipids under conditions in which mammalian membrane rafts are likely to form The use of synchro-tron X-ray powder diffraction methods is able to pro-vide detailed information on phase coexistence in complex bilayers as well as on coupling of the two monolayers of the bilayer, an essential feature in the formation of a membrane raft

Thermotropic phase behaviour of ternary

mixtures

To characterize the thermotropic phase behaviour of

ternary mixtures of egg-phosphatidylcholine

(Ptd-Cho)⁄ brain sphingomyelin (SM) ⁄ cholesterol, aqueous

dispersions equilibrated at 20C were subjected to

ini-tial heating scans to 50C and subsequent cooling

scans to 20C at 2Æmin)1 The intensity of scattered

X-rays was recorded simultaneously in the small-angle

(SAXS = 1–14) and wide-angle (WAXS = 12–30)

scattering regions during the scans The results

obtained from an initial heating scan of a ternary

mix-ture comprised of egg-PtdCho⁄ brainSM ⁄ cholesterol in

molar proportions 80 : 10 : 10 are presented in

Fig 1A Two series of reflections in the SAXS region

can be detected and they are in the order 1 : 1⁄ 2 : 1 ⁄ 3

(only the first two-orders are shown), indicating that

all structures are lamellar Within each order of Bragg

reflection more than one lamellar phase is present; this

is particularly evident from the second-order reflections

in which overlapping peaks are obvious The absence

of a sharp WAXS peak indicates that no gel or crystal phases are present in the mixture [28] The scattering intensity profiles were subject to a peak fitting analysis

to characterize the coexisting lamellar phases The SAXS data were best fitted by three Gaussian + Lorentzian curves as seen in Fig S1C,D The fit of two peaks to the Bragg peak is shown for comparison

in Fig S1A,B The relationship between d-spacings of the three individual peaks and temperature is plotted

in Fig 1B The fact that discrete lamellar reflections can be deconvolved from the scattering bands means that the two leaflets of each of the respective bilayer structures are coupled

An analysis of the scattering intensity profiles recorded during a subsequent cooling scan (see Figs S2 and S3) indicates that the changes observed in lamellar d-spacings (Fig 1B) during the heating scans are completely reversible with no significant temperature hysteresis This is consistent with the absence of any structural alteration in the bilayer or thickness of the

Fig 1 Characterization of

egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol; 80 : 10 : 10 An

overview of small- and wide-angle X-ray

scattering intensity profiles recorded from

an aqueous dispersion of

egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol in molar

propor-tions 80 : 10 : 10, recorded during a heating

scan at 2Æmin)1between 20 and 50 C, is

shown in (A) as the scattering intensity

profiles from the first two orders of lamellar

repeat structures and wide-angle scattering

profiles (B) Lamellar d-spacings (C)

Scatter-ing intensities (D) Peak shape (scatterScatter-ing

amplitude ⁄ full-width at half maximum

inten-sity) of the first-order lamellar structures.

(E) WAXS d-spacings (F) WAXS scattering

intensities.

hydration layer characterizing the dimensions of the

lamellar unit cell The scattering intensities of the three

peaks and an index of the peak sharpness (peak

ampli-tude⁄ full width at half maximum intensity) are

presented in Fig 1C,D, respectively Unlike lamellar

d-spacings, the decrease in scattering intensity observed

during the initial heating scan is not reversed during

the subsequent cooling scan (Fig S1B,C) Likewise,

the simultaneous broadening of these peaks is not

reversed on cooling This suggests that the size, but

not the structure as judged by lamellar d-spacing, of

the scattering arrays decreased during heating from the

equilibration temperature to 35 C as a consequence

of the fragmentation of the scattering units into

smal-ler, possibly less well-ordered, arrays Heating to

higher temperatures appears to have no additional

effect on the arrangement of the scattering units,

there-fore, a reliable indication of the relative amounts of

lamellar structure in the deconvolved peaks

contribut-ing to the overall scattercontribut-ing intensity can be obtained

at 38C It is noteworthy that the parameters of the

peak of greatest d-spacing, which contributes least to

the total scattering intensity, are relatively constant

during the temperature scans This may indicate that

the arrangement and presentation of the scattering units in this lamellar structure do not change signifi-cantly with temperature

A peak-fitting analysis of the WAXS intensity pro-files was undertaken and the results are presented in Fig 1E,F There is no evidence of a sharp peak at

 0.42 nm to indicate the presence of a gel phase A minor peak located at a d-spacing of 0.45 nm can be deconvolved from the scattering profiles recorded at temperatures < 30C during the initial heating scan, but this peak becomes indistinguishable from a broad scattering band at  0.463 nm typical of disordered hydrocarbons at higher temperatures

A ternary mixture containing higher proportions of brainSM and cholesterol was then examined and the results are presented in Fig 2 The scattering intensity patterns recorded in the SAXS and WAXS regions during the initial heating scan from 20 to 50C from the ternary mixture comprised of egg-Ptd-Cho⁄ brainSM ⁄ cholesterol, 60 : 20 : 20, are presented

in Fig 2A The SAXS intensity peaks in this mixture are best fit by only two Gaussian + Lorentzian curves,

in contrast to the mixture shown in Fig 1 Lamellar d-spacings, scattering intensity profiles and peak

Fig 2 Characterization of egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol; 60 : 20 : 20 An overview of small- and wide-angle X-ray scattering intensity profiles recorded from

an aqueous dispersion of egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol in molar propor-tions 60 : 20 : 20, recorded during a heating scan at 2Æmin)1between 20 and 50 C, is shown in (A) as the scattering intensity pro-files from the first two orders of lamellar repeat structures and wide-angle scattering profiles (B) Lamellar d-spacings (C) Scatter-ing intensities (D) Peak shape of the second-order of the lamellar structures (E) WAXS d-spacings (F) WAXS scattering intensities.

shapes derived from analysis of the thermal scans are

shown to be distinct in Fig 2B,C,D, respectively It

can be seen that the peak of shortest d-spacing

observed in the mixture comprised of 80 : 10 : 10

(Fig 1B) is absent from this ternary mixture

More-over, the remaining two lamellar phases have

corre-spondingly greater lamellar d-spacings than those

observed in the mixture shown in Fig 1 The

tempera-ture-dependent change in scattering intensity is

consid-erably less marked, suggesting that the scattering units

are more stable when the proportions of brainSM and

cholesterol in the mixture are increased relative to

egg-PtdCho The Bragg peaks also tend to be sharper The

dominant scattering peak in the WAXS region is

shifted to shorter d-spacings indicating that increased

proportions of brainSM and cholesterol bring about

a closer packing in the hydrocarbon region of the

bilayers

The effect of increasing only the proportion of

brainSM in the ternary mixture is exemplified by the

behaviour of a mixture comprised of egg-PtdCho⁄

brainSM⁄ cholesterol, 10 : 80 : 10 shown in Fig 3 The

scattering intensity profiles in the SAXS region show

the first two orders of reflection of lamellar phases and

the presence of a relatively sharp WAXS peak at

0.42 nm, indicating that a gel phase forms on equili-bration at 20C This WAXS peak coexists with a rel-atively weak scattering band at  0.47 nm which can

no longer be distinguished from the main peak at a d-spacing at 0.44–0.45 nm upon heating above  32 C The changes observed in the SAXS⁄ WAXS profiles are consistent with a progressive replacement of a gel phase of brainSM and the disappearance of a small proportion of a coexisting highly disordered lamellar phase with a homogeneous liquiordered phase of d-spacing 0.44 nm during heating to 32C At higher temperatures, a new lamellar phase of greater d-spac-ing appears but represents only a relatively minor com-ponent of the overall scattering intensity It can be concluded from analysis of the behaviour of this ter-nary mixture that the properties of the major constitu-ent of the mixture, namely, long N-acyl chain molecular species of sphingomyelin, tend to dominate the temperature-dependent structural parameters of the bilayers

Assignment of lamellar structures The next task was to establish the identity of the coexisting lamellar phases in ternary mixtures

Fig 3 Characterization of

egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol; 10 : 80 : 10 An

overview of small- and wide-angle X-ray

scattering intensity profiles recorded from

an aqueous dispersion of

egg-Ptd-Cho ⁄ brainSM ⁄ cholesterol in molar

propor-tions 10 : 80 : 10, recorded during a heating

scan at 2Æmin)1between 20 and 50 C, is

shown in (A) as the scattering intensity

pro-files from the first two orders of lamellar

repeat structures and wide-angle scattering

profiles (B) Lamellar d-spacings (C)

Scatter-ing intensities (D) Peak shape of the

first-order lamellar structures (E) WAXS

d-spacings (F) WAXS scattering intensities.

ing relatively high proportions of the fluid host

phospholipid, egg-PtdCho, which are representative of

the lipid composition of mammalian membrane

extra-cellular leaflet embedding the raft microdomains Three

possible lamellar structures comprised of varying

proportions of lipids are brainSM⁄ egg-PtdCho,

brain-SM⁄ cholesterol or egg-PtdCho ⁄ cholesterol; a ternary

complex of the three lipids is excluded as implausible in

this analysis because most ternary mixtures are

com-prised of more than one bilayer component Figure 4

shows a method of assigning the composition of the

dif-ferent lamellar phases on the basis of the relationship

between lamellar d-spacing and temperature The result

of a peak-fitting analysis of the SAXS intensity profile

recorded from the ternary mixture comprised of

egg-PtdCho⁄ brainSM ⁄ cholesterol, 80 : 10 : 10, at 38 C

is presented in Fig 4A The profile can be seen to be

best fit by three Gaussian + Lorentzian peaks which

are shown in Fig 4B together with the difference

between the observed and calculated fit to the data

(Fig 4C) (see Fig S1)

Peak 1 represents  10% of the total scattering

from the first-order Bragg reflections The d-spacings

of this peak coincide closely with d-spacing recorded

from a binary mixture of egg-PtdCho⁄ brainSM in

equimolar proportions recorded under the same

condi-tions (Fig 4D) It is known that gel-phase separation

occurs in this binary mixture when equilibrated at

20C [29], however, there is no evidence that gel-phase

separation occurs in this ternary mixture (Fig 1E)

The presence of 10 mole% cholesterol in the ternary

mixture apparently hinders formation of a gel phase

by brainSM in this mixture Assignment of peak 1 to a

structure of pure brainSM can also be excluded on this

evidence The fit of peak 1 to brainSM⁄ cholesterol

mixtures was considered from the respective

dimen-sions of the unit cell (d-spacings) and peak shape

parameter representing the order of the diffracting

units The effect of varying proportions of cholesterol

in binary mixtures with brainSM is presented in

Fig 5A It can be seen that the d-spacing of brainSM

bilayers at 38C is 8.3 nm and this is progressively

reduced by increasing the proportions of cholesterol

(Fig 5C) An equimolar proportion of cholesterol

would be required to reduce the d-spacing to that

observed for peak 1 (6.8 nm) in Fig 4C The

assign-ment of peak 1 as comprised of egg-PtdCho and

 20 mole% cholesterol (Fig 5D) cannot be excluded

on this criterion Other evidence presented below,

how-ever, indicates that cholesterol is not a constituent of

peak 1

The scattering intensity of peak 2 contributes

 30% to the total intensity of the first-order Bragg

peaks of the mixture shown in Fig 4 The d-spacing of peak 2 in Fig 4B coincides closely with bilayers formed from a binary mixture of egg-PtdCho and pro-portions of cholesterol of  25 mole% (Fig 5D) The effect of cholesterol on d-spacings of egg-PtdCho bilayers is complex The presence of only 10 mole%

A

B

C

D

Fig 4 Assignment of lamellar structures An analysis of the ternary mixture of egg-PtdCho ⁄ brainSM ⁄ cholesterol; 80 : 10 : 10 recorded

at 38 C (A) Fit of scattering intensity from the first-order Bragg reflection (d) to three Gaussian + Lorentzian area curves (s) (B) Peak deconvolution from the scattering intensity profile shown in (A) (C) Difference between observed and calculated fits to the data (D) Lamellar d-spacings as a function of temperature recorded dur-ing heatdur-ing scans at 2Æmin)1 d, peak 1; s, egg-PtdCho ⁄ brainSM,

50 : 50; j, peak 2; h, egg-PtdCho ⁄ cholesterol, 70 : 30; m, peak 3;

D, egg-PtdCho.

cholesterol causes a dramatic increase in d-spacing

because of the full extension and vertical orientation of

the hydrocarbon chains of the phospholipid in the

bilayer Increasing proportions of cholesterol up to

30 mole% result in a progressive decrease in repeat

spacing because of hydration effects at the bilayer–

water interface An assignment of peak 2 to a binary

mixture of brainSM and cholesterol can be excluded

on the basis of lamellar d-spacings > 7 nm at 38C

[30] Assuming peak 2 is comprised of phospholipid

and 25 mole% cholesterol, the contribution of the

peak to the total scattering from the ternary mixture is

calculated to be 25% This is close to the observed

proportion of the total scattering from peak 2 in the

ternary mixture

Assignment of the dominant peak (peak 3)

repre-senting  60% of total scattering at 38 C in the

ternary mixture egg-PtdCho⁄ brainSM ⁄ cholesterol,

80 : 10 : 10, was made by comparison with bilayers of

pure egg-PtdCho, the most abundant phospholipid in

the mixture As can be seen from Fig 4C, the

d-spac-ing of the pure phospholipid is 0.5 nm less at

equiv-alent temperatures than observed for peak 3 Because

the d-spacing is less than binary mixtures containing

high proportions of cholesterol in egg-PtdCho, it

fol-lows that the increase in d-spacing of peak 3 must be

caused by the presence of proportions of cholesterol

< 10 mole% That peak 3 is comprised predominantly

of egg-PtdCho is also evident from the absence of this

peak in ternary mixtures containing lower proportions

of egg-PtdCho, as demonstrated in the ternary mix-ture consisting of egg-PtdCho⁄ brainSM ⁄ cholesterol,

60 : 20 : 20 examined in Fig 2 Thus, assignment of peak 3, as judged by d-spacing, can be made as egg-PtdCho containing a relatively small proportion of cholesterol

Further evidence consistent with assignment of peak 1 to a liquid-ordered lamellar phase comprised of

an equimolar proportion of egg-PtdCho and brainSM was obtained by relating the relative mass of brainSM

in ternary mixtures of egg-PtdCho⁄ brainSM ⁄ choles-terol to the scattering contribution from peak 1 to the total scattering intensity recorded at 38C A peak of lamellar repeat of 6.7–6.8 nm could be deconvolved from first-order Bragg reflections in 12 ternary mix-tures examined; no peak at this position was observed

in mixtures with proportions of cholesterol exceeding

50 mole% There was no correlation between the scat-tering intensity of this peak and the relative mass of egg-PtdCho, brainSM, cholesterol or any binary com-binations of the lipids in the mixtures (Fig S4) How-ever, if the contribution to the scattering of the peak was limited to a mass of brainSM equivalent to the proportion of egg-PtdCho in those ternary mixtures where the mol% of brainSM exceeds that of egg-Ptd-Cho, a correlation is obtained A plot of the relation-ship between the scattering intensity of the peak and the mass of equimolar proportions of brainSM⁄

Fig 5 Binary mixtures of phospholipid and

cholesterol Small-angle X-ray scattering

intensity patterns of binary mixtures of (A)

brainSM and (B) egg-PtdCho with the

indicated mol% cholesterol at 37 C (C)

and (D) show the respective relationships

between lamellar d-spacing and mol%

cholesterol.

egg-PtdCho in different ternary mixtures is shown in

Fig 6

A third method to investigate the assignments of

composition of peaks 1 and 2 was to compare relative

electron densities through the bilayer repeat structures

The results of such calculations are summarized in

Fig 7 It can be seen that relative electron density

dis-tributions across the bilayer repeats calculated at

38C for peaks 1 and 2 are different By contrast, the

thickness of the bilayers and the water layers of peak 1

are almost identical to those calculated for bilayers

consisting of an equimolar mixture of egg-PtdCho and

brainSM The bilayer thickness of peak 2 is

signifi-cantly greater, and the water layer signifisignifi-cantly less,

than calculated for peak 1 Resolution of the

electron-density calculation for peak 2 was relatively low

because only three orders of reflection were detected in

this mixture Nevertheless, the thickness of the bilayer

and water layer are almost identical to the parameters

calculated from a binary mixture comprised of

egg-PtdCho and 30 mole% cholesterol, a characteristic Lo

phase [31] where three orders of reflection were used in

the calculation

Discussion

The lipids identified in rafts isolated from biological

membranes differ from those of the parent membrane,

but such results need to be regarded with some cau-tion at this stage The reason is that reliable methods

of isolating rafts have not generally been employed The size of domains in living cell membranes is defined as between 10 and 200 nm [32], and vesicles derived from rafts occupying areas of the parent membrane of this order would be between 5 and

30 nm in diameter This is at odds with the size of vesicles isolated as detergent-resistant membranes Estimates of the size of detergent-resistant membrane vesicles prepared from rat brain indicate a relatively homogeneous population of unilamellar vesicles of diameter ranging from 130 to 240 nm [33] This corre-sponds to an average domain diameter in the parent membrane in the order of 600 nm, somewhat larger than areas envisaged for membrane raft domains Subpopulations of these vesicles can be separated by immunoadsorption methods containing different sur-face antigens, which argues against the fusion and amalgamation of domains in the parent membrane Electron microscopy examination of these vesicles indicates that the prion protein (PrPc) and thymus-derived antigen 1 (Thy-1) associated with these raft preparations are generally clustered together and occupy only a small fraction of the vesicle membrane [9] This suggests that the rafts are not homogeneous

Fig 6 Relationship between brainSM and scattering intensity

Cor-relation between relative scattering intensity of peak 1 in Fig 4

from the first-order Bragg reflection (d-spacing 6.7–6.8 nm) and

mass of brainSM + equimolar egg-PtdCho in different ternary

mix-tures of the two phospholipids and cholesterol recorded at 38 C.

Mixtures with proportions of brainSM greater than egg-PtdCho

were taken as equimolar to the proportion of egg-PtdCho in the

mixture.

Fig 7 Electron-density calculations Relative electron density pro-files were calculated through the lamellar repeats of peak 1 and peak 2 recorded at 38 C from the data in Fig 1A Relative electron densities calculated from binary mixtures of brainSM ⁄ egg-PtdCho

in equimolar proportions and egg-PtdCho ⁄ 30 mole% cholesterol at

38 C are shown for comparison.

bilayers of lipids in liquid-ordered phase but that an

organization is imposed on the proteins that causes

their association within the structure

The question of whether membrane proteins or

lip-ids alone or together play a part to bring about the

segregation of raft components is a moot point It is

known that successful delivery of plasma membrane

raft proteins from the Golgi in yeast depends on the

biosynthesis of ergosterol and sphingolipids [34]

Genetic screening of defective mutants indicated a lack

of a functional fatty acid elongating system for

synthe-sis of long-chain molecular species of sphingolipids

[35], or a defect of dihydrosphingosine C4 hydroxylase

in the biosynthesis of phytosphingosine [36] may be

responsible One possible mechanism for organizing

proteins in the liquid-ordered phase is by homotypic

interactions between the proteins themselves or

interac-tions mediated by intermediary proteins An example

of the latter is the clustering of Pma1p, the plasma

membrane H+-ATPase of yeast, in raft lipid domains

This has been shown to require a peripheral membrane

protein, Ast1, in the endoplasmic reticulum, a process

that is an essential step in the transfer of the raft

pro-tein to the plasma membrane [37] In the case of the

N+⁄ H+ antiporter in yeast (Nha1p), however, the

sorting signal apparently resides in the hydrophobic

domain of the membrane [38] and sphingolipid is

essential for retention of the protein in the plasma

membrane [39] There is also a strong possibility that

the lipid anchors that tether proteins to membrane

rafts may interact in a specific way with the lipids

forming the raft

Clearly, there is scope for different methods of

orga-nizing proteins within membrane rafts, but it is not

easy to envisage how the specificity required to bring

about clustering of one type of receptor protein on

one side of the membrane, and co-localizing this with

specific lipid-anchored proteins on the opposite side of

the membrane, can occur simply within a

liquid-ordered phase of polar lipid and cholesterol On the

basis of the evidence obtained in this study, such

speci-ficity can be proposed The structures formed by

binary mixtures of long N-acyl molecular species of

sphingolipids and phospholipids consist of

stoichiome-tric complexes of 1 : 2 phosphatidylethanolamines [27]

and 1 : 1 phosphatidylcholine [26] A phase with

hydrocarbon chain spacings consistent with a

liquid-ordered quasicrystalline phase, formed from equimolar

proportions of egg-PtdCho and brainSM [29], has

been identified in this study in ternary mixtures with

cholesterol

We propose that the structure formed by long

N-acyl fatty acid molecular species of sphingolipids

and phospholipids creates a matrix that is coupled across the raft membrane According to such a model, glycosphingolipids based on molecular species of galactosylceramides with long N-acyl fatty acids and phosphatidylcholines would reside in the outer mono-layer in mammalian plasma membrane These domains are coupled with glucosylceramides and phosphatidy-lethanolamines in the cytoplasmic leaflet These struc-tures form a matrix into which GPI-anchored receptor proteins are interpolated on the outer surface and are coupled with corresponding lipid-anchored effecter proteins located in the cytoplasmic leaflet The specific-ity of these interactions may involve the sugar residues

of the glycosphingolipids and the domains of the pro-teins, or the configuration of the lipid anchor, or both That activity-associated remodelling of lipid anchors is

a recognized process in raft function [40,41] suggests that the configuration of the raft anchors is a likely candidate

The lipid matrix model of the membrane raft struc-ture can be formulated according to a two-stage pro-cess of molecular assembly Cartoons of the structures comprising the model and their relationship to bilayer spacings are presented in Fig 8 Liquid-ordered domains comprised of more saturated molecular spe-cies of phospholipid and cholesterol serve to exclude most membrane proteins and accommodate those proteins required for raft function Glycosphingolipids with long N-acyl fatty acid chains and phospholipids form a quasicrystalline matrix acting to concentrate and organize the protein components into a functional complex in the raft It is reported that cholesterol is excluded from such phases [42] The remodelling of lipid anchors takes place in the liquid-ordered domain

in a manner that allows them to interpolate into the matrix component of the raft membrane The intimate association between receptors and effectors brought about by their integration into the matrix is an essen-tial feature designed to facilitate the transmission of molecular signals generated on one side of the matrix

to the other

The model of raft structure we propose fits current knowledge of the lipid composition of membrane rafts obtained without detergent treatment Eighty-three molecular species of membrane lipid have been identi-fied and quantiidenti-fied in highly puriidenti-fied raft preparations from yeast [43] Glycosphingolipids with almost exclu-sively long-chain hydroxylated N-acyl substituents [C26:0(OH)] are present in equimolar proportions with di-unsaturated molecular species of phosphatidylcho-line and would be expected to form a quasicrystalphosphatidylcho-line bilayer structure The remaining phospholipids are dominated by phosphatidylinositol with a saturated

fatty acid acylated to the C-1 position of the glycerol.

This would form a liquid-ordered phase with a

compo-sition comprised of 38 mole% ergosterol The order of

the lipids in the membrane rafts as measured by

spectroscopic studies using C-Laudan is consistent with

the tight packing of acyl chains in the raft model

membrane

Materials and methods

Lipids

Egg-yolk phosphatidylcholine (egg-PtdCho, 715 Da),

bovine brain sphingomyelin (brainSM, 788 Da) and

choles-terol (387 Da) were purchased from Sigma (Sigma-Aldrich,

St Quentin-Fallavier, France) A complete lipid analysis of

each phospholipid was performed by ESI-tandem MS [29]

and the data are presented in Table S1

Sample preparation

Samples for X-ray diffraction examination (Table S2) were prepared by dissolving lipids in warm (45C) chloro-form⁄ methanol (2 : 1, v ⁄ v) and mixing them in the desired proportions (denoted as molar ratios in binary mixtures) The organic solvent was subsequently evaporated under a stream of oxygen-free dry nitrogen at 45C and any remaining traces of solvent were removed by storage under high vacuum for 2 days at 20C Dry lipids were hydrated with an equal mass of water This was sufficient to fully hydrate egg-PtdCho [44] and brainSM [45], respectively The lipids were stirred thoroughly with a thin needle, sealed under argon, and annealed by 50 thermal cycles between 20 and 65C, ensuring a complete mixing of phospholipids Samples were stored under argon at a temperature not below 4C X-Ray diffraction examination was performed after 5 h sample equilibration at 20C and after careful stir-ring before transfer into the sample cell In order to check

Fig 8 Cartoons of the molecular composition of the different structures proposed for the lipid matrix model of membrane raft struc-ture and their relationship to the Bragg spacings of ternary mixstruc-tures of egg-PtdCho ⁄ brainSM ⁄ cholesterol Other evidence is reviewed in Quinn [6].