Raft domains of variable properties and compositions in plasma membrane vesicles

Abstract

Biological membranes are compartmentalized for functional diversity by a variety of specific protein–protein, protein–lipid, and lipid–lipid interactions. A subset of these are the preferential interactions between sterols, sphingolipids, and saturated aliphatic lipid tails responsible for liquid–liquid domain coexistence in eukaryotic membranes, which give rise to dynamic, nanoscopic assemblies whose coalescence is regulated by specific biochemical cues. Microscopic phase separation recently observed in isolated plasma membranes (giant plasma membrane vesicles and plasma membrane spheres) (i) confirms the capacity of compositionally complex membranes to phase separate, (ii) reflects the nanoscopic organization of live cell membranes, and (iii) provides a versatile platform for the investigation of the compositions and properties of the phases. Here, we show that the properties of coexisting phases in giant plasma membrane vesicles are dependent on isolation conditions—namely, the chemicals used to induce membrane blebbing. We observe strong correlations between the relative compositions and orders of the coexisting phases, and their resulting miscibility. Chemically unperturbed plasma membranes reflect these properties and validate the observations in chemically induced vesicles. Most importantly, we observe domains with a continuum of varying stabilities, orders, and compositions induced by relatively small differences in isolation conditions. These results show that, based on the principle of preferential association of raft lipids, domains of various properties can be produced in a membrane environment whose complexity is reflective of biological membranes.

The recent discovery of phase separation in plasma membrane (PM) vesicles isolated from mammalian cells (1–3) is the most convincing evidence of functionally relevant coexistence of liquid domains in biological membranes. The microscopic phases observed in these studies are likely the result of coalescence of nanoscopic assemblies (lipid rafts) present in cellular membranes under physiological conditions (4). This coalescence into a condensed “raft phase” (5) allows microscopic investigation of the composition (1, 6, 7) and physical properties (8, 9) of the underlying raft assemblies.

The current conception of lipid rafts in cell biology is of mesoscopic, isolated domains with defined properties and compositions. How lipid rafts in cells correspond to simple lipid model systems—which show complete, microscopic separation of a condensed, ordered phase (L_o) rich in saturated lipids and sterols from a disordered (L_d) phase depleted of these components and enriched in unsaturated glycerophospholipids (10, 11)—is yet to be determined. Large-scale membrane domains are not observable in intact cells without significant perturbation (e.g., by antibody cross-linking; ref. 12), so the submicroscopic membrane architecture has been inferred from biochemical fractionation of whole cells or spectroscopic techniques that probe the level of individual molecules. In contrast to biophysical results from simple model systems, such experiments have revealed the possibility of a variety of membrane domains whose biological relevance and physicochemical determinants remain undefined.

Phase separation in synthetic model systems has been recapitulated in isolated PMs, which contain the full complement of lipids and proteins of native membranes. Giant plasma membrane vesicles (GPMVs) (1) and plasma membrane spheres (PMS) (2) are distinct PM preparations that allow direct investigation of membrane properties divorced from cellular complexity (e.g., interactions with cytoskeletal elements, membrane trafficking, etc.). In both preparations, micrometer-scale, spherical PM protrusions are detached from the underlying cytoskeletal scaffold [which could inhibit (13), segregate (14), and/or nucleate (15) PM domains], thus allowing phase separation into microscopically observable liquid domains. These coexisting phases sort PM components based on their affinity for lipid rafts (2, 7, 16), thus making them ideal models for investigation of raft domains in live cells.

In a previous paper (7), we showed how isolation conditions affect protein palmitoylation and consequently the raft phase partitioning of transmembrane (TM) proteins in PMs isolated with chemical treatments (GPMVs). Here, we observe that the properties of the phases themselves are also dependent on the means of isolation, specifically the chemicals used to induce blebbing. Phase separation observed without chemical perturbation mirrored the behavior of a subset of chemically induced blebs, thus confirming the inherent phase separation potential of PMs. Finally, varying isolation conditions had a major effect on phase separation temperature, protein distribution among the phases, and the relative order difference between coexisting phases, suggesting that raft domains of continuously varying properties and compositions are achievable in membranes with biological complexity.

Results

GPMVs isolated from a rat basophilic leukemia (RBL) cell line using the typical chemical mixture of 25 mM paraformaldehyde (PFA) and 2 mM DTT (PFA + DTT) separated into two liquid phases observable by prestaining the cells with a membrane marker with nonuniform phase partitioning (FAST-DiO; Fig. 1A). In this preparation, coexisting phases typically coalesced into two large domains that remained separated close to room temperature (T_misc = 17.8 ± 1.3 °C), as previously observed (1, 7–9) (Fig. 1 A–C). In contrast, GPMVs isolated using 2 mM N-ethylmaleimide (NEM) (7), a non-cross-linking, nonreducing sulfhydryl-reactive chemical, had domains that were smaller, more dispersed (Fig. 1A), and which were only observable at relatively low temperatures (below 10 °C) (Fig. 1 A–C).

Fig. 1.

GPMVs derived with a combination of PFA and DTT remain phase separated at higher temperatures than other sets of isolation conditions. (A and B) Representative images and quantification of GPMVs derived with either 2 mM NEM or a combination of 25 mM PFA and 2 mM DTT (curves are sigmoidal fits to data). Microscopically large round coalesced domains persist in PFA + DTT GPMVs to approximately 20 °C. In contrast, domains in NEM-derived vesicles are smaller, more dispersed, and not observable above 10 °C. (C) Miscibility transition temperature (T_misc, defined as the temperature at which 50% of vesicles are phase separated) in GPMVs as a function of isolation agent(s). Isolating GPMVs with 2 mM NEM alone, with a combination of 2 mM NEM and 25 mM PFA, or with PFA + NEM supplemented with either a nonsulfhydryl reducing agent (TCEP) or a palmitoylation inhibitor (2BP) does not recapitulate the PFA + DTT phenotype.

To test the hypothesis that the striking difference in phase behavior between GPMVs produced using PFA + DTT versus NEM was due to the efficient and nonspecific cross-linking of amine-containing compounds by formaldehyde, GPMVs were produced with 2 mM NEM supplemented with 25 mM PFA. Phase separation was observed, but persisted only to approximately 5 °C (i.e., similar to vesicles produced with NEM alone) (Fig. 1C). To test whether the reducing behavior of DTT was responsible for the enhanced domain stability in the PFA + DTT GPMVs, the NEM preparation was supplemented with 2 mM tris(2-carboxyethyl)phosphine (TCEP) a potent, nonsulfhydryl reducing agent (we used TCEP instead of DTT because NEM contains a sulfhydryl-reactive maleimide group, thus a mixture of NEM and DTT would inactivate one or both of the compounds, consistent with the lack of blebbing in this preparation.) Again in this case, phase-separated vesicles were observed whose miscibility transition temperature (T_misc) was statistically indistinguishable from NEM GPMVs (Fig. 1C). Finally, inhibition of palmitoylation (by 2-bromopalmitate, 2BP) prior to vesicle isolation did not affect T_misc (Fig. 1C), suggesting that DTT-dependent protein depalmitoylation (7) was not responsible for the observed effects. Thus, neither nonspecific cross-linking, nor disulfide reduction, nor depalmitoylation was solely responsible for the phase separation observed in PFA + DTT GPMVs.

To control for possible cross-reaction of the isolation agents, as well as unpredictable cellular effects of these chemicals, GPMVs were isolated with NEM, PFA + DTT, or PFA + NEM, dialyzed to remove the isolation agents, then treated after isolation. Observing the isolated, dialyzed, and posttreated vesicles at 10 °C, the distinct phase behaviors observed above could be readily reproduced. NEM treatment of PFA + DTT vesicles had no observable effect (Fig. 2A), whereas PFA + DTT treatment of NEM GPMVs reproduced the PFA + DTT behavior (Fig. 2B). Thus, because the PFA + DTT behavior could be reproduced by posttreatment of isolated vesicles, the observed phase separation “phenotypes” are not a function of cellular processes (e.g., phospholipase activity, protein degradation, etc.) that could be affected by GPMV isolation conditions.

Fig. 2.

Postisolation treatment reproduces PFA/DTT phenotype only when both PFA and DTT are present. (A–F) Representative images (all taken at 10 °C) of GPMVs isolated with the conditions shown on the left of the images, then dialyzed to remove isolation chemicals, then treated with the chemicals indicated above the arrows. Phase separation at this temperature is observed only when PFA and DTT are present, either as the isolation condition or the postisolation treatment. (G) Percentage of phase-separated GPMVs as a function of temperature for different isolation and treatment conditions. Each point represents 50–100 vesicles; curves are sigmoidal fits to data. Results representative of three independent experiments.

Remarkably, neither posttreatment with DTT alone (Fig. 2C) nor PFA alone (Fig. 2D) had any observable effect on phase separation; however, treating NEM + PFA vesicles with DTT reproduced the PFA + DTT phenotype (Fig. 2E). As in vesicles isolated with TCEP, reduction of dialyzed PFA + NEM vesicles with TCEP did not reproduce the PFA + DTT behavior (Fig. 2F). Quantification of these effects suggested that the observed behaviors could clearly be distinguished into two phenotypes—i.e., those preparations that contained both PFA and DTT (characterized by coalesced domains persisting above 15 °C) and those that did not (phase separation observed only below 10 °C). These results reveal a surprising synergy between PFA and DTT in producing the stable domains observed in PFA + DTT GPMVs.

To determine which of the phenotypes observed corresponded to phase separation in chemically unperturbed PMs, we used the GPMV imaging methods (i.e., staining with a membrane dye followed by cooling below 10 °C) to observe isolated PMS induced by swelling A431 cells (2). We were able to directly observe liquid–liquid phase separation of PMs without chemical perturbation or cross-linking with exogenous agents (Fig. 3A). Domains were typically small and dispersed, and only observable below 5 °C, thus suggesting that the phase separation phenotype observed in NEM GPMVs is reflective of chemically unperturbed PMs. Similar to GPMVs, the PFA + DTT phenotype could be reproduced in these chemically unperturbed membranes, but only by treatment with both PFA and DTT (Fig. 3B).

Fig. 3.

Phase separation in PMS. (A) PMS isolated by cell swelling without additional chemicals show clear liquid–liquid phase separation below 5 °C (staining with 1 μg/mL FAST-DiO, disordered phase marker). (B) Treatment of PMS with PFA + DTT reproduces the PFA + DTT phenotype in GPMVs, whereas treatment with PFA or DTT alone does not have a significant effect on phase behavior.

The transition from the NEM to the PFA + DTT phenotype was dependent on the concentration of DTT (Fig. 4). Increasing [DTT] from 0.5 to 5 mM at constant [PFA] (25 mM) produced GPMVs that phase separated at progressively higher temperatures, from below 10 to above 15 °C (Fig. 4A). Combining data from many experiments showed a sharp transition in phase behavior between 0.2 and 1 mM DTT (Fig. 4B). In contrast, varying [PFA] through the range of concentrations that yielded measurable vesicles (5–250 mM) did not have an effect on the miscibility transition temperature (Fig. 4C).

Fig. 4.

[DTT]-dependent phase separation. (A) Phase separation could be induced at progressively higher temperatures by increasing [DTT] in the GPMV isolation buffer at constant [PFA] (25 mM). Each point is representative of 75–100 vesicles per condition; curves are sigmoidal fits to data. Results representative of three independent experiments. (B) T_misc increases from below 5 °C to above 15 °C between [DTT] = 0.2 and 2 mM. Data at 0 mM DTT includes vesicles produced with NEM alone, NEM + PFA, and PFA alone. (C) No variation in T_misc observed with varying [PFA] at constant [DTT] (2 mM) through the range that yields GPMVs. Points in B and C are average ± SD from three to seven independent trials.

This striking effect of DTT could also be observed in the difference in order between the coexisting phases in the various preparations (Fig. 5). GPMVs produced with varying [DTT] were labeled with C-Laurdan and imaged under phase-separated conditions (5 °C for all preparations). C-Laurdan is a fluorescent membrane dye whose emission depends on the relative packing of the membrane (17), which has been previously used to characterize isolated PMs (18). Quantification of red-shifted fluorescence, characteristic of disordered membranes, relative to fluorescence from ordered membranes yields the generalized polarization (GP), a relative index where higher GP is indicative of a more ordered membrane environment (Fig. 5A). Domains with varying order could be observed in all isolation conditions (Fig. 5A), although neither the absolute order of the domains nor the order difference between coexisting phases remained constant (Fig. 5B). The inclusion of PFA in the isolation buffer did not increase membrane order, as might be expected from cross-linking of membrane components. In contrast, increasing [DTT] strikingly increased the order difference between the phases from 0.13 ± 0.04 with no DTT to 0.21 ± 0.04 at 2 mM (Fig. 5 B and C). The variation in ΔGP (GP_ordered - GP_disordered) was observed between 0.2 and 2 mM DTT, mirroring the variation in T_misc shown in Fig. 4. Thus, increasing order difference between the two coexisting phases correlates strongly with increased miscibility transition temperature.

Fig. 5.

Order difference between coexisting phases is dependent on [DTT]. (A) Exemplary GP images of coexisting domains in GPMVs (all imaged at 5 °C). GP is a relative indicator of the membrane order (higher GP equals more ordered membranes). (B and C) Although the inclusion of PFA did not have a major effect on order (comparing NEM versus 25 mM PFA + 0.2 mM DTT), increasing [DTT] reduced the overall order and increased the relative order difference between the coexisting phases. Average ± SD from 10 vesicles per condition and representative of three experiments.

The correlation between order difference and [DTT] could be attributed directly to the progressive loss of protein from the raft phase (Fig. 6). The relative concentration of protein in the coexisting phases was imaged by nonspecifically labeling extracellularly accessible proteins with Sulfo-N-hydroxysulfosuccinimide (NHS)-biotin and producing GPMVs that were then labeled with fluorescently modified antibiotin Fab’ (as in ref. 7). Quantification of relative fluorescence intensity between coexisting phases is thus a reporter of the relative concentration of protein in the two phases (Fig. 6A). Using this method, it was clearly observed that increasing [DTT] in the GPMV isolation medium was responsible for a loss of protein from the raft phase, with the relative raft/nonraft protein concentration decreasing from 0.52 in NEM GPMVs (no DTT) to 0.07 in vesicles isolated with 2 mM DTT (Fig. 6B).

Fig. 6.

Relative protein concentration in coexisting phases is dependent on [DTT]. (A) Relative protein concentration in coexisting phases of GPMVs can be quantified by fluorescent imaging of antibiotin Fab’ after nonspecific protein labeling with Sulfo-NHS-biotin. (B) K_p,raft (ratio of intensity in the raft versus nonraft phase) of extracellularly accessible proteins decreases with [DTT] from a high concentration of raft proteins in NEM GPMVs to barely observable raft protein signal in 2 mM DTT. Points are average ± SD of > 15vesicles per condition, representative of three independent experiments.

Protein removal from the ordered phase could be attributed to the chemical coupling of membrane proteins to the primary amine-containing lipid, phosphatidylethanolamine (PE) by aldehyde-mediated cross-linking (Fig. 7). Thin-layer chromatography (TLC) of lipids extracted from GPMVs showed a quantitative depletion of PE from the organic lipid extract [relative to phosphatidylcholine (PC) which was unaffected] when GPMVs were isolated with PFA and DTT, most likely due to the covalent attachment of this lipid to neighboring protein moieties. The [DTT] dependence of this effect correlated strongly with the [DTT] dependence observed for the miscibility transition temperature (Fig. 4), order (Fig. 5), and protein partitioning (Fig. 6). To verify PE cross-linking to peptides by PFA + DTT in a controlled system, PE-containing liposomes (80/20 PC/PE) were treated with PFA + DTT in the presence or absence of peptides (Fig. 7C). PFA + DTT treatment of pure lipid liposomes had no effect on the extracted lipids, whereas inclusion of a TM peptide (3 mol %) led to a near total depletion of PE from the organic extract. The same effect, though to a much smaller magnitude, was observed with a soluble protein (γ-globulin). Our results thus indicate that raft phase order and coalescence is strongly dependent on membrane protein partitioning, which can in turn be modulated by the degree of coupling between membrane proteins and unsaturated phospholipids.

Fig. 7.

GPMV isolation with PFA and DTT leads to reduction of extracted PE. (A) TLC of lipids extracted from GPMVs as a function of isolation conditions. The PE band decreases in intensity relative to PC with increased [DTT]. (B) Average ± SD of background subtracted peak intensities of PE relative to PC. Results representative of three independent experiments. (C) TLC of lipids extracted from 80/20 PC/PE liposomes. PFA + DTT treatment does not reduce extractable PE unless either a TM peptide (3 mol % LW peptide) or a soluble protein (3 mol % γ-globulin) is included in the preparation.

Discussion

Lipid-mediated phase separation in PMs can be observed across various preparations, independent of the nature (Figs. 1 and 2) or presence (Fig. 3) of chemicals used for isolation. Microscopically observable phases in these isolated membranes are reflective of lipid–lipid and lipid–protein interactions that give rise to nanoscopic domains in live cells (5), whose coalescence can be induced by biochemical (2, 3) or physical means (e.g., lowering temperature). The most striking result of the above studies is that, although phase separation is independent of the protein content of the raft phase, the properties of raft domains in GPMVs are variable across a continuum of stabilities (Fig. 4), orders (Fig. 5), and compositions (Fig. 6), and that these are modulated by small changes in preparation conditions. These properties in GPMVs derived without nonspecific cross-linking of protein to lipids (i.e., NEM GPMVs) mirrored those of unmodified PMS (Fig. 3), suggesting that these may be most reflective of the native PM.

We observe strong correlations between the protein concentration of the raft phase, the order difference of coexisting, and phase separation temperature. These correlations are likely causative: Exclusion of integral proteins would be expected to increase lipid packing of the raft phase, thus enhancing its order relative to the nonraft phase, and promoting phase separation. These data are conceptually summarized in Fig. 8. The very low protein concentration of the raft phase in high DTT preparations (Fig. 6) suggests that it is comprised mostly of lipids and lipid-anchored proteins (GPI-anchored proteins are consistently enriched in raft phases regardless of preparation; refs. 1, 6, 7) separated from the protein-rich, highly disordered phase (Fig. 8, Top). In contrast, preparations with low (or no) DTT produced coexisting domains with similar protein concentrations—though not compositions (7)—and ordering (Fig. 8, Middle and Bottom). The relative difference between coexisting domains is likely to influence their separation, as suggested by the T_misc data in Fig. 4.

Fig. 8.

Domains of continuously varying orders and compositions observable in PM vesicles isolated with PFA and varying [DTT].

The possible relationship of these results to phase separation at physiological temperatures can be inferred from the recent observation of critical behavior in GPMVs. Veatch et al. showed that line tensions and compositional fluctuations in phase-separated GPMVs could be fitted to universal scaling factors for systems near critical points (8). Extrapolating these universal behaviors, they predicted compositional fluctuations of tens of nanometers at 37 °C. Extending these findings to our experiments, a change of T_crit from 17 to 5 °C (the difference in T_misc between high and low DTT preparations) would equate to an approximately twofold change in domain size (and corresponding fourfold change in domain lifetime) at 37 °C. However, two major caveats must be emphasized: (i) Critical behavior has so far been observed only in 2 mM DTT GPMVs and should therefore not be automatically inferred for all other isolation conditions; and (ii) even if critical behavior allows extrapolation of domain properties to physiological temperature in GPMVs, these are still isolated membranes derived with chemical treatment and divorced of the inherent complexity of living cells, thus making direct comparison between their properties and the structure/function of living membranes inherently speculative.

Nevertheless, our observations that biological membranes are capable of forming ordered liquid domains of various properties may explain the bewildering variation in raft composition and size measured in live cells: Although Pralle et al. derived a raft diameter of approximately 50 nm consistent across two different cells types (19), other estimates have placed the value anywhere from a few nanometers (20) up to several hundred (21). Our observations suggest that differences in cell type, state of cells, or even local intracellular environments could account for raft domains of different sizes and compositions all governed by the same organizing principle. For example, although the apical membrane of a polarized epithelial cell can be conceptualized as a large, continuous raft domain (22) (presumably due to the enrichment of raft lipids at the apical PM; ref. 23), the PMs of many nonpolarized cells may be expected to have small, short-lived domains that are only observable by indirect methods (20). Similarly, the relatively disordered raft domains observed here (e.g., in NEM GPMVs) could be the physicochemical analogs of various “nonraft” PM domains inferred from biochemical and nanoscopic observations, including GPI- anchored protein (AP) domains (24), heavy rafts (25), Lubrol rafts (26), Hras domains (27), T-cell receptor islands (28), and tetraspanin webs (29). Thus, the raft principle—defined as preferential interactions between sterols, saturated lipids, glycosphingolipids, and certain proteins—could be employed by the cell to produce highly ordered, stable domains containing only lipids and lipid-anchored proteins (like GPI-AP domains inferred from FRET experiments or Ras domains observed by EM) as well as more disordered, TM protein-inclusive domains that need not necessarily be insoluble in nonionic detergents.

Obviously, the perturbations observed here are artifactual consequences of treatment of membranes with exogenous chemical agents. We have previously shown that DTT can cause protein removal from the raft phase by depalmitoylation (7), and this effect may be partially responsible for the perturbations observed here. However, the concentration range at which the DTT-dependent effects are observed here (0.2–1 mM) is significantly lower than that necessary to induce depalmitoylation (at least 2 mM DTT) (7). Additionally, pretreatment of cells with the palmitoylation inhibitor 2BP did not have an effect on phase separation (Fig. 1C), confirming the limited role of palmitoylation in the observed effects. Rather, the activity that seems to be the biggest driver of the observed effects is the coupling of PE to membrane proteins mediated by PFA and DTT (Fig. 7). This activity would be expected to deplete the raft phase of TM proteins because the acyl chains of PE are nearly always unsaturated (30) and therefore a covalent protein–PE complex would be nonraft preferring. A testable hypothesis that follows from this proposal is that protein compositions and properties of raft domains could be regulated by a switchable affinity of membrane proteins for certain membrane phospholipids. Although several specific interactions between integral proteins and membrane lipids have been observed (e.g., for the epidermal growth factor receptor; ref. 31 and others reviewed in ref. 32), there has been no systematic analysis to determine how widespread this phenomenon is, nor whether it can be reversibly regulated. This field represents a neglected area of membrane biology that is ripe for further investigation.

Conclusion

The observation of phases of various orders and compositions in PM vesicles suggests that biological membranes have the potential to form domains of continuously variable compositions and stabilities. The functional consequences of these findings remain to be investigated; however, the notion of raft domains of varying sizes, properties, and compositions presents an evolution of our understanding of lipid rafts from a confusing assortment of difficult to define domains to a physicochemical membrane organizing principle for the functional compartmentalization of biological membranes.

Materials and Methods

Cell Culture and Labeling.

RBL cells were cultured in medium containing 60% MEM, 30% RPMI medium 1640, 10% FCS, 2 mM glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin at 37 °C in humidified 5% CO₂. A431 cells were grown under similar conditions in 90% MEM + 10% FCS. Prior to GPMV/PMS isolation, cell membranes were stained with 5 μg/mL of FAST-DiO (Invitrogen), a lipidic dye that strongly partitions to disordered phases due to double bonds in its fatty anchors.

GPMV and PMS Isolation, Labeling, and Treatment.

GPMVs were isolated (1) and imaged (9) under temperature-controlled conditions as previously described. For postisolation treatments, isolation chemicals were removed by dialysis and GPMVs were treated for 1 h at 37 °C. PMS were isolated as described (2). Relative order magnitudes were measured by two-photon microscopy of C-Laurdan in phase-separated GPMVs at 5 °C as previously described (18).

Labeling of Cell Surface Proteins.

Surface-exposed proteins were biotinylated using a membrane-impermeable, amine-reactive biotin reagent, as described (7). Briefly, cells were incubated on ice for 30 min in the presence of Sulfo-NHS-biotin (1 mg/mL; Sigma), followed by GPMV isolation and staining with a monomeric Fab’ fragment of goat antibiotin coupled to Texas red (10 μg/mL; Rockland Immunochemicals). Images of the Texas red signal were quantified as described (7) to derive K_p,raft, the relative concentration of surface protein in the raft phase.

Thin-Layer Chromatography and Model Liposomes.

Lipids were extracted using two-step extraction protocol (chloroform∶methanol 10∶1 followed by 2∶1) (33). The lipid-containing organic phases from both steps were pooled and dried under a nitrogen stream. The lipids were redissolved in a small volume of chloroform/methanol (3∶1) and applied to a TLC plate. For the TLC analysis, a 9∶7∶2 mixture of chloroform∶methanol∶ammonia (25% solution) was used. The lipids were detected after spraying the plate with 20% H₂SO₄ solution and heating (200 °C, 3–5 min). For quantification, TLC plates were scanned and intensity profiles of each lane were generated using ImageJ.

Model liposomes were prepared from 80 mol % PC (l-α-phosphatidylcholine; Avanti) and 20 mol % PE (l-α-phosphatidylethanolamine; Avanti) as described (31). LW peptide [acetyl-KKWWLLLLLLLLALLLLLLLLWWKK-amide; kind gift from E. London (Stony Brook University, NY)] and γ-globulin (Rockland) were included at 3 mol %.