Cell-derived GPMVs (giant plasma-membrane vesicles) enable investigation of lipid phase separation in a system with appropriate biological complexity under physiological conditions, and in the present study were used to investigate the cholesterol-dependence of domain formation and stability. The cholesterol level is directly related to the abundance of the liquid-ordered phase fraction, which is the majority phase in vesicles from untreated cells. Miscibility transition temperature depends on cholesterol and correlates strongly with the presence of detergent-insoluble membrane in cell lysates. Fluorescence correlation spectroscopy reveals two distinct diffusing populations in phase-separated cell membrane-derived vesicles whose diffusivities correspond well to diffusivities in both model systems and live cells. The results of the present study extend previous observations in purified lipid systems to the complex environment of the plasma membrane and provide insight into the effect of cholesterol on lipid phase separation and abundance.
Several lines of evidence, including the identification  and characterization  of detergent-resistant membrane fractions, anomalous diffusion of membrane bound tracers  and nanoscale aggregation of fluorescent proteins and markers [4,5], have been used to develop the hypothesis of cholesterol- and sphingolipid-enriched membrane rafts in the plasma membrane . The current conception of rafts consists of transient nanoscopic domains within the bulk membrane which contain a variety of specific proteins and lipids. These domains have been proposed as platforms for the organization and concentration of signalling components . Model system experiments using mixtures of synthetic lipids in monolayers , supported bilayers  and giant vesicles [9,10] have reproduced and extensively characterized liquid-phase demixing in mixtures of cholesterol and various phospholipids . A consistent result across all model systems is that inclusion of cholesterol in lipid mixtures can result in liquid–liquid phase separation into an Lo (liquid-ordered) and an Ld (liquid-disordered) phase [11,12]. The Lo phase is characterized by conformational ordering resembling a crystalline/gel phase , but distinguished from it by a high degree of rotational and translation mobility, characteristic of the Ld phase . This Lo–Ld coexistence has been proposed to be the physicochemical basis for plasma-membrane rafts, which have been postulated to be ordered phase domains.
Although model system experiments have successfully recapitulated cholesterol-dependent liquid-phase coexistence, they have offered no conclusive evidence that Ld–Lo phase immiscibility is physiologically related to cell-membrane rafts. This limitation is due, in part, to the fact that model systems cannot replicate the tremendous complexity of the plasma membrane, both in the heterogeneity of lipid species and the inclusion of membrane-associated proteins that affect the thermodynamics of lipid-mediated demixing. However, recent experiments using GPMVs (giant plasma-membrane vesicles), cell-derived liposomes that maintain the lipid  and protein  diversity of the plasma bilayer, have shown temperature-dependent liquid–liquid phase separation, similar to that observed in model systems . This phase separation was found to segregate known protein and lipid markers of lipid rafts, providing a link between Ld–Lo phase separation in model systems and the lipid raft hypothesis in cellular plasma membranes. Recent experiments have suggested that GPMVs exist near miscibility critical points, and therefore that the behaviour of these complex lipid and protein mixtures can be understood by universally applicable scaling laws .
In the present study, we extend these studies by using fluorescence microscopy and FCS (fluorescence correlation spectroscopy) to characterize liquid–liquid phase coexistence in GPMVs and find good agreement between observations in GPMVs and purified lipid model systems with regard to the dependence of phase abundance and miscibility on the cholesterol fraction. This agreement implies that phase behaviour in simple lipid mixtures can be used to approximate domain formation in the complex environment of the plasma membrane. Additionally, we find a strong correlation between the temperature-dependent separation of a raft phase in GPMVs and the detergent-resistance of cellular membranes, suggesting that these two phenomena are related and that both are related to the existence of membrane domains in the native plasma membrane.
GPMVs were isolated after chemically inducing cell blebbing with 25 mM paraformaldehyde and 2 mM DTT (dithiothreitol) in calcium-containing buffer [150 mM NaCl, 10 mM Hepes and 2 mM CaCl2 (pH 7.4)] for 1 h at 37 °C as previously described , and labelled with rhoPE (rhodamine 1-stearoyl 2-oleoyl phosphatidylethanolamine; Avanti) and/or nap (naphthopyrene; Sigma) by incubation at room temperature (23 °C) for 15 min with 2.5 μg/ml rhoPE and/or 10 μg/ml nap. A chamber was created by making a square of silicon sealant (Dow Corning) on a BSA-coated coverslip, into the middle of which 20 μl of labelled GPMV suspension was deposited, followed by sealing of the chamber with another coverslip. The fluorescence of the vesicles was visualized using an inverted microscope (Leica) equipped with appropriate filter sets. The temperature was controlled using a Peltier temperature control stage (TS-4; Physitemp) at 10 °C, unless otherwise stated. The percentage surface area covered by the Ld phase was quantified by calculating the surface area of the spherical cap (SAcap) covered by the rhoPE-rich (bright) phase using the relationship (eqn 1):
where hcap is the height of the spherical cap and equivalent to (eqn 2):
if Ld is the minority phase and (eqn 3),
if Ld is the majority phase.
rvesicle and rcap are the radii of the vesicle and rhoPE-rich cap respectively, and were measured using ImageJ software.
FCS was performed with a confocal microscope on GPMVs as previously described for GUVs (giant unilamellar vesicles) [19,20]. Detergent-resistant membranes were isolated on a discontinuous sucrose gradient as described previously . Purification and quantification of membrane lipids were performed in accordance with previously published protocols [22,23]. Further experimental details are available in the Supplementary material (at http://www.biochemj.org/bj/424/bj4240163add.htm).
RESULTS AND DISCUSSION
GPMVs derived from NIH 3T3 fibroblasts were stained with rhoPE (disordered-phase tracer), and observed by fluorescence microscopy to quantify the relative abundance of Ld and Lo phases (the Lo phase is also referred to as the ‘raft phase’ because it enriches for raft components ). The results in Figure 1 show that more than 70% of the surface area of vesicles derived from untreated cells is composed of the raft phase, suggesting a continuous ordered bilayer in which disordered domains exist as inclusions [imaging performed at 10 °C unless otherwise stated, although phase abundances were temperature-independent over a wide range (10–24 °C; Supplementary Figure S1 at http://www.biochemj.org/bj/424/bj4240163add.htm)]. This finding is inconsistent with the concept of liquid-ordered domains as isolated and scarce lipid rafts, instead suggesting the possibility of a plasma membrane existing as a majority liquid-ordered continuum interrupted by disordered domains. The idea of a percolating raft phase is consistent with previous measurements of diffusivity of raft and non-raft protein and lipid markers , electron-spin resonance in live cells  and single-cell detergent extractions .
Fluorescence images and quantification of the Ld (non-raft) phase fraction in GPMVs (stained with rhoPE to label the Ld phase) as a function of cholesterol modulation
Detergent-resistant membrane fractions, which formed the original basis for the raft hypothesis, are enriched in cholesterol, suggesting that lipid rafts are liquid-ordered membrane structures enriched in, and possibly dependent on, the presence of cholesterol. We modulated cholesterol levels in cells prior to GPMV isolation, and in GPMVs isolated from untreated cells, to determine whether cholesterol affects the abundance and coexistence of the two liquid phases. Depletion of cellular cholesterol by treatment with 5 mM MBCD (methyl-β cyclodextrin) decreased the cholesterol molar fraction by more than 20% in the derived GPMVs, and resulted in significant changes to the relative abundance of the phases, more than doubling the relative abundance of the disordered, non-raft, phase from 28% to more than 65% of the surface area (Figure 1). Inversely, loading cells with cholesterol by treatment with cholesterol-saturated MBCD (decreasing the GPMV phospholipid/cholesterol ratio from 1:1 to 1:0.8) led to a near disappearance of the Ld phase, from 28% to 5% surface area, causing these GPMVs to appear nearly dark with very small bright disordered patches. Similar results were observed when GPMVs were cholesterol-loaded after isolation from untreated cells.
Although MBCD depletion of plasma membrane cholesterol strongly affected the phase abundance of derived GPMVs, this technique was unable to deplete cholesterol below ~ 35 mol% (Supplementary Figure S2 at http://www.BiochemJ.org/bj/424/bj4240163add.htm). This limitation was overcome by direct MBCD treatment of vesicles following their isolation from cells, which resulted in non-circular, jagged and ribbon-like domains (Figure 1) similar in morphology to gel-phase domains observed in cholesterol-free GUVs where demixing was the consequence of acyl chain length differences between the component phospholipids .
These cholesterol modulation data are consistent with purified lipid experiments that have shown cholesterol-dependent formation of a Lo phase, and the abolition or reduction of that phase when cholesterol was depleted [10,27]. The induction of a non-fluid gel phase by wholesale depletion of cholesterol is consistent with liquid/gel phase separation in GUVs lacking cholesterol , as well as measurements in cholesterol-depleted live cells  and observations of viral lipid extracts .
In addition to phase abundance, the cholesterol molar fraction affects phase separation in numerous simplified lipid model systems , prompting the hypothesis that the same effect might be observed in the complex lipid and protein mixture of GPMVs. Cellular cholesterol levels were manipulated as above, and the temperature-dependent phase separation of GPMVs derived from those cells was measured. Loading cells with cholesterol (~ 20% decrease in [phospholipid]/[cholesterol]) decreased the average miscibility transition temperature (Tmisc) from 24 to 21 °C, whereas depleting cellular cholesterol increased Tmisc to 32 °C. Cholesterol depletion also produced a significant fraction (15%) of microscopically phase-separated vesicles at 37 °C, suggesting the possibility of cholesterol-dependent phase separation in complex membranes at physiological conditions (Figure 2). This observation corresponds well to phase separation induced by cholesterol depletion of live cells at physiological temperatures . Additionally, the cholesterol molar fraction dependence of Tmisc (Figure 2) closely resembles the same dependence measured in model liposomes . This agreement is particularly striking since not only the trends, but also the absolute values of the transition temperatures, are similar for three-component GUVs and the much more complex cell-derived vesicles considered here, suggesting that findings in purified lipid systems can be extended to multi-component, cell-derived protein–lipid mixtures.
Temperature-dependence of phase separation in GPMVs isolated from untreated cells (black), cholesterol-depleted cells (red circles) and cholesterol-loaded cells (blue squares)
The findings presented above show that phase abundance and miscibility as a function of cholesterol in GPMVs correspond well to the same properties observed in simple systems. At the opposite extreme of complexity from purified lipid models of membrane rafts are the detergent-resistant membrane fractions that suggest the existence of biochemically distinct membrane domains in whole-cell lysates. To determine the relationship between fluid-phase coexistence in GPMVs and the presence of a low-density membrane fraction in detergent-lysed cells, the temperature-dependence of these two distinct membrane phenomena was investigated by also performing sucrose-gradient fractionation of cell membranes at different temperatures. The temperature profile of microscopically observable phase coexistence in GPMVs from untreated cells followed a relatively abrupt transition from entirely phase-separated to microscopically uniform vesicles between 20 and 25 °C. The temperature-dependent abundance of the mass of cholesterol in detergent-resistant membrane fractions yielded a very similar temperature profile, with the detergent-resistant fractions making up 20–25% (w/w) of the total cholesterol below the phase-transition temperature of the GPMVs, but <10% above (Figure 3 and Supplementary Figure S3 at http://www.BiochemJ.org/bj/424/bj4240163add.htm). The temperature-dependent presence of a detergent-resistant component was confirmed by quantifying the sucrose-gradient distribution of a protein reported to partition to detergent-insoluble rafts. Below 20 °C, >50% of Alexa Fluor® 488-labelled CTB (fluorescently labelled cholera toxin B) was recovered in low-density fractions, whereas <15% was detergent-resistant above this average GPMV phase-separation temperature (Supplementary Figure S4 at http://www.BiochemJ.org/bj/424/bj4240163add.htm).
Correlation between the temperature-dependence of phase separation of GPMVs (diamond points, line is a sigmoidal fit) and the abundance of detergent-resistant membranes (detergent solubilization performed at the indicated temperatures) as quantified by the percentage of cholesterol in the detergent-resistant fractions (striped bars)
Although the requirement for low temperature and detergent treatment has led to intense scrutiny regarding the physiological relevance of detergent-resistant membranes , we find a strong correlation between detergent-resistance and GPMV phase separation. This result suggests that detergent-insolubility and the existence of a separated ordered phase in complex mixtures are related, and that both are related to membrane rafts. These results confirm findings from detergent extractions of mixed lipid vesicles [9,31] and GPMVs , and argue against the artefactual induction of liquid-phase coexistence by detergent as the sole mechanism for raft formation. Although neither microscopically observable phase separation nor detergent-resistance (without cyclodextrin treatment) was observed in our experiments at physiological temperature (37 °C), the phase separation in GPMVs and detergent-resistance of certain membrane fractions might occur due to the coalescence of underlying nanoscopic assemblies of raft lipids and proteins, recently proposed thermodynamically  and experimentally , that are neither microscopically observable nor detergent-insoluble at 37 °C.
One of the distinguishing characteristics of ordered compared with disordered fluid phases in model lipid systems is a difference in translational and rotational diffusivity ; lipid and protein diffusivity are major determinants of the cellular distribution and corresponding functions of plasma-membrane components. Lipid diffusivity was quantified in both phase-separated and uniform GPMVs by FCS on fluorescent tracer lipids incorporated into the vesicles. At a temperature at which GPMVs separate into two liquid phases (10 °C), we observed two distinct populations of diffusion coefficients obtained from fits to autocorrelation data (Figure 4). The histogram fitted to normal distribution yield average diffusivities of 1.8 and 5.6 μm2/s. The correlation data at 37 °C suggest a single population of diffusion coefficients with a mean value approximately equivalent to that of the faster diffusing component at 10 °C (Figure 4b). These results agree well with previous measurements in both cells and model systems. The 3-fold difference in lateral mobility between the two phases is almost exactly the same as was measured by fluorescence recovery in DMPC (dimyristoyl phosphatidylcholine)-cholesterol bilayers at physiological temperature . The diffusion coefficients measured in Lo and Ld phases are very close to the diffusion coefficients measured here, strongly suggesting that the 1.8 μm2/s component corresponds to the Lo (raft) phase of GPMVs, whereas the faster component is likely to be the disordered phase. The diffusivity differences and magnitudes measured here correspond very well to those measured by FCS for Ld and Lo phase markers in raft-composition GUVs , as well as to small-scale diffusivities previously measured in live cells by optical tweezers , underlining the agreement not just in phase separation, but also in the properties of those phases, among live cells, GPMVs and purified lipid systems.
Histograms of diffusion coefficients obtained by FCS of rhoPE diffusing in (A) phase-separated vesicles at 10 °C and (B) microscopically uniform vesicles at 37 °C
The results of the present study demonstrate that complex multi-component lipid and protein membranes can exhibit similar phase behaviour as simple three-component model membranes with respect to abundance, diffusivity and phase separation. Our findings also suggest that this behaviour is related to the detergent-resistance of biological membranes, thereby relating biochemical and biophysical descriptions of membrane rafts and suggesting an important role for GPMVs as an intermediate lipid raft model system combining the complexity of the biological membrane with the observable phase separation and experimental simplicity of purified lipid mixtures.
fluorescence correlation spectroscopy
giant plasma-membrane vesicle
giant unilamellar vesicle
- Lo phase
rhodamine 1-stearoyl 2-oleoyl phosphatidylethanolamine
miscibility transition temperature
Ilya Levental designed and performed experiments and wrote the paper. Fitzroy Byfield and Pramit Chowdhury performed experiments. Feng Gai and Tobias Baumgart designed experiments. Paul Janmey designed experiments and wrote the paper.
This work was supported by the National Science Foundation [grant number MRSEC 05-20020]; and the National Institutes of Health [grant numbers AR38910, R21AI073409].