Lipid Rafts As a Membrane-Organizing Principle

See allHide authors and affiliations

Science  01 Jan 2010:
Vol. 327, Issue 5961, pp. 46-50
DOI: 10.1126/science.1174621


Cell membranes display a tremendous complexity of lipids and proteins designed to perform the functions cells require. To coordinate these functions, the membrane is able to laterally segregate its constituents. This capability is based on dynamic liquid-liquid immiscibility and underlies the raft concept of membrane subcompartmentalization. Lipid rafts are fluctuating nanoscale assemblies of sphingolipid, cholesterol, and proteins that can be stabilized to coalesce, forming platforms that function in membrane signaling and trafficking. Here we review the evidence for how this principle combines the potential for sphingolipid-cholesterol self-assembly with protein specificity to selectively focus membrane bioactivity.

The lipid raft hypothesis proposes that the lipid bilayer is not a structurally passive solvent, but that the preferential association between sphingolipids, sterols, and specific proteins bestows cell membranes with lateral segregation potential. The concept has long suffered assessment by indirect means, leading to questions of fact or artifact (1). The resistance of sphingolipid, cholesterol, and a subclass of membrane proteins to cold detergent extraction (2) or mechanical disruption (3) has been widely used as an index for raft association with little or no regard for the artifacts induced by these methods. Though the acquisition of resistance to disruption may point to physiologically relevant biases in lateral composition (4), this disruptive measure tells us little about native membrane organization. Support from light microscopy was also missing because, with the exception of organization into specialized membrane domains such as caveolae or microvilli, putative raft components—specifically glycosylphosphatidylinositol (GPI)–anchored proteins, fluorescent lipid analogs, raft transmembrane (TM) domains, and acylated proteins—often show a homogeneous distribution at the cell surface (5). Moreover, early investigations into submicron membrane organization often yielded conflicting evidence regarding the distribution or motion of these constituents in the living cell (1). Today, however, the advancement of technology has produced compelling data that self-organization of lipids and proteins can induce subcompartmentalization to organize bioactivity of cell membranes.

Origins of the Lipid Raft Concept

Biochemically, it is clear that lipids are sorted within the cell (6). This is particularly notable in polarized epithelia where glycosphingolipids (GSLs) are enriched at the apical surface (7). Lipid rafts were originally proposed as an explanation: Self-associative properties unique to sphingolipid and cholesterol in vitro could facilitate selective lateral segregation in the membrane plane and serve as a basis for lipid sorting in vivo (7). This proposal for compartmentalization by lipid rafts suggested a nonrandom membrane architecture specifically geared to organize functionality within the bilayer. This function was initially thought to be membrane trafficking; however, rafts could influence organization of any membrane bioactivity (Fig. 1). Here, we highlight advances in technology that point to the existence of raft-based membrane heterogeneity in living cells and discuss the levels of preferential association underlying dynamic domain structure and biological function(s).

Fig. 1

Evolution of the raft concept for subcompartmentalization in cell membranes. A bold H indicates hydrogen bonding. VSV, vesicular stomatitus virus; DRMs, detergent-resistant membranes; GUV, giant unilamellar vesicle; m/z, mass/charge ratio; SPT, single-particle tracking; FCS, fluorescence correlation spectroscopy; FRET, fluorescence resonance energy transfer; STED, stimulated emission depletion; FPALM, fluorescence photoactivation localization microscopy.

Lipid Interactions in Model Membranes

Assembly into two-dimensional liquid crystalline biomembranes is a fascinating property characteristic of lipids. Long thought to be incapable of coherent lateral structure (8), it is now apparent that principles of lipid self-association can also confer organization beyond nonspecific measures of fluidity. An important advance in model-membrane systems was the discovery of phase separation in wholly liquid bilayers (9, 10). It is a cholesterol-dependent lateral segregation, wherein the planarity (molecular flatness) of the rigid sterol ring favors interaction with straighter, stiffer hydrocarbon chains of saturated lipids and disfavors interaction with the more bulky unsaturated lipid species (11). Interaction with cholesterol also forces neighboring hydrocarbon chains into more extended conformations, increasing membrane thickness and promoting segregation further through hydrophobic mismatch (12). In purified lipid systems, the combined effect is a physical segregation in the membrane plane: A thicker, liquid-ordered, Lo phase coexists with a thinner, liquid-disordered, Ld phase (13). Sphingolipids tend to display longer and more saturated hydrocarbon chains, thus potentiating interdigitation between leaflets (14) and favoring interaction with cholesterol. Moreover, unlike glycerophospholipids, the region of chemical linkage between the head group and sphingosine base contains both acceptors and donors of hydrogen bonds, thus increasing associative potential, both with cholesterol and other sphingolipids (11). Other explanations for cholesterol selectivity include the proposed umbrella effect, in which cholesterol hydrophobicity is preferentially shielded by the strongly hydrated head groups of sphingolipid (15) or stoichiometric, but reversible, complex formation between cholesterol and sphingolipid or saturated glycerophospholipid (16).

Immiscible liquid phase coexistence in vitro has been suggested as the physical principle underlying rafts in vivo (17). Of central importance is the demonstration of selectivity in association between certain lipids. However, phase separation in simple systems at thermodynamic equilibrium in vitro cannot be translated into proof for membrane domain formation in living cells (1). Instead, model-membrane work emphasizes the fact that certain lipids exhibit preferential association and provides a framework for understanding how heterogeneity in cell membranes may arise (18). In this respect, the terms Lo and Ld should not be applied to the living cell, as they refer only to the liquid-ordered and liquid-disordered phases of model-membrane systems where parameters relating to translational order (lateral diffusion) and conformational order (trans/gauche ratio in the acyl chains) can be accurately measured (11).

Glimpses of Nano-Assemblies in Living Cells

Currently, lipid rafts are viewed as dynamic nanoscale assemblies enriched in sphingolipid, cholesterol, and GPI-anchored proteins (19) (Fig. 2A). To reach this viewpoint, membrane research has had to contend with the observer’s effect, akin to Heisenberg’s uncertainty principle: We can change and/or induce heterogeneity in membranes simply by trying to observe it. Initially, this required moving away from detergent extraction as a means to infer native organization. In a first step, detergent-free, chemical cross-linking of GPI-anchored proteins at the plasma membrane suggested that the intrinsic heterogeneity by rafts was present in nanoscale complexes below the optical resolution limit set by the diffraction of light (19). This nanometer-size scale was later supported by viscous drag measures of antibody-bound raft proteins (21) and electron microscopic observation of immunogold-labeled raft antigens (20). Indeed, recent near-field scanning optical microscopy has confirmed a nanoscale bias in the distribution of raft-associating proteins in fixed cells (22). Less perturbing measures of spatial and temporal dynamics in living cells have also provided correlating data. For example, single-particle tracking of colloidal gold–labeled GPI-anchored receptors reveals “stimulation-induced temporary arrest of lateral diffusion,” or STALL, in short-lived (~0.5-s) 50-nm areas as a bioactive feature of receptor function (23). Parallel advances in microscopy and spectroscopy have revealed similar heterogeneity for raft molecules in uncross-linked, “resting” conditions. For GPI-anchored proteins, variable waist fluorescence correlation spectroscopy points to <120-nm assemblages that fluctuate on a subsecond time scale (24). High spatial and temporal resolution fluorescence resonance energy transfer (25) has generated a more conservative size estimate with GPI-anchored receptors residing in more temporally stable clusters of ~10 nm. Assembly formation is always cholesterol-dependent, and, in some cases, an actin requirement has also been seen (23, 25). However, other techniques have indicated that nanoheterogeneity is actin-independent (26). The situation for TM proteins is not yet clear. However, fluorescence photoactivation localization microscopy has revealed a dynamically clustered nanoscale distribution of hemagglutinin (27), a TM protein previously described as raft-associating (21). The role of the association between cholesterol and sphingolipids in assembly formation has been analyzed recently by stimulated emission depletion microscopy. This study revealed that, unlike glycerophospholipids, plasma-membrane sphingolipids display transient cholesterol-dependent confinement in areas of <20 nm (28). In this case, differences in diffusion were attributed to differential hydrogen-bonding capacities of glycerol- versus sphingosine-based lipids. However, spin-labeled lipid probes at the cell surface have also revealed heterogeneity in membrane order on an electron spin resonance time scale (29).

Fig. 2

Hierarchy of raft-based heterogeneity in cell membranes. (A) Fluctuating nanoscale assemblies of sterol- and sphingolipid-related biases in lateral composition. This sphingolipid/sterol assemblage potential can be accessed and/or modulated by GPI-anchored proteins, certain TM proteins, acylated cytosolic effectors, and cortical actin. Gray proteins do not possess the chemical or physical specificity to associate with this membrane connectivity and are considered non-raft. GPL, glycerophospholipid; SM, sphingomyelin. (B) Nanoscale heterogeneity is functionalized to larger levels by lipid- and/or protein-mediated activation events (e.g., multivalent ligand binding, synapse formation, protein oligomerization) that trigger the coalescence of membrane order–forming lipids with their accompanying selective chemical and physical specificities for protein. This level of lateral sorting can also be buttressed by cortical actin. (C) The membrane basis for heterogeneity as revealed by the activation of raft phase coalescence at equilibrium in plasma-membrane spheres. Separated from the influence of cortical actin and in the absence of membrane traffic, multivalent clustering of raft lipids can amplify the functional level to a microscopic membrane phase. Membrane constituents are laterally sorted according to preferences for membrane order and chemical interactions.

Different techniques are yielding a range of values for different molecular constituents in diverse cell types. However, these methods all point to the existence of small, dynamic and selective cholesterol-related heterogeneity in the plasma membranes of living cells. Recent data point to critical behavior as a potential physical basis for the existence of fluctuating nanoscale assemblies in plasma membranes (30).

Functionalization of Nanoscale Heterogeneity

Antibody cross-linking at the cell surface causes raft proteins and lipids to co-patch and exclude non-raft proteins (31). This selectivity in patching is cholesterol-dependent and can be transmitted across the plasma-membrane leaflets (32). The nonrandom coalescence behavior observed in these artifactual cross-linking studies suggests how raft organizing potential may be functionalized to larger, more physiologically relevant temporal and spatial scales (Fig. 2B). A contention of the lipid raft hypothesis is that dynamic nanoscale heterogeneity can be stabilized to coalesce into larger raft domains by specific lipid-lipid, protein-lipid, and protein-protein interactions (20). In this sense, cell membranes would possess an underlying sphingolipid/cholesterol-based connectivity that can be activated to cluster membrane bioactivity with little energy cost. Indeed, multimerization promotes the sorting of GPI-anchored proteins into sphingolipid/cholesterol-enriched carriers during clathrin-independent endocytosis (33). Along similar lines, clustering of cell surface Gb3 or GM1 (both GSLs) by their respective ligands Shiga toxin and cholera toxin induces energy-independent tubular invaginations of sphingolipid-biased membrane composition (34, 35). Similar behavior has also been reported during the multivalent binding of SV40 virus to its GM1 receptor (35). Invagination from the plasma membrane was dependent on having longer receptor hydrocarbon chains, which are common to sphingolipid, and suggests that the effect is mediated by line tension arising between membrane domains of different compositional order (35). Coalescence of dynamic heterogeneity also occurs during signaling, for example, during the formation of B cell receptor (BCR) or T cell receptor (TCR) foci. Antigen binding induces the dynamic association of BCR to its signaling effector Lyn kinase and leads to the formation of an immune synpase. The interaction is dependent on the nature of Lyn lipid anchorage, with membrane order–disrupting bulky hydrocarbon chains preventing association with the BCR (36). During TCR activation, raft components of this receptor complex (e.g., GPI- anchored proteins) become selectively immobilized in nanoscale clusters (37), seeding the accumulation of cholesterol, sphingomyelin, and saturated and long-chain phosphatidyl choline into the synapse, effectively sorting proteins according to their affinity for raft domains (38). Rafts in this “activated” or coalesced condition constitute a more ordered assemblage: a fluid membrane environment in which proteins can be modulated specifically (39), yet that exists separately from the surrounding membrane rich in unsaturated glycerophospholipid. Raft activation is often stabilized or nucleated by scaffolding elements such as cortical actin (40) and may become dominating when the mole fraction of sphingolipids and cholesterol increases, as is the case in the apical membrane of epithelial cells (41).

Phase Separation in Cell Membranes

Despite their selective co-patching with raft markers at the cell surface, raft TM proteins are depleted from the tightly packed Lo phase when reconstituted in model systems (42, 43). Thus, the Lo phase as it exists in simple model membranes is unlikely to be identical to raft-based heterogeneity in plasma membranes that selectively includes TM proteins (44, 45). Giant plasma-membrane vesicles isolated by a chemical membrane blebbing procedure can be cooled to phase separate into Lo- and Ld-like phases (46), and here also, raft TM proteins are typically excluded from the ordered membrane phase (47). Remarkably, this phase coexistence indicates that after chemical modification of protein (e.g., formaldehyde cross-linking, thiol treatment), the capacity for physical or lipid-based liquid-liquid phase separation can be manifested by the plasma membrane, despite its compositional complexity. Now the question is, how might phase length–scale separation take place in plasma membranes at physiological temperatures?

Some insight has come from a cell-swelling procedure to separate plasma-membrane spheres from the influence of cytoskeletal, endocytic, or exocytic processes in a cell line enriched in the raft ganglioside GM1. Pentavalent clustering by cholera toxin resulted in sterol-dependent coalescence of a micron-scale raft “phase” at 37°C, selectively reorganizing the lateral distribution of proteins and lipids according to their predicted affinity for raft domains (44). In this case, selective incorporation of TM proteins was achieved at a lipid-ordering level far below that observed in model-membrane Lo phases (45). Thus, whereas preferential lipid-lipid associations do underlie raft clustering, at physiological temperature they do not form a Lo phase when raft proteins become included, and specific lipid-protein interactions must come into play to modify organization (Fig. 2C). Along these lines, a comparison of lipid-packing in vesicles formed from lipids of the plasma membrane versus the plasma membrane itself reveals that lipid domain–forming order is tightly regulated by the presence of protein (48).

Rafts as Entities of Physical and Chemical Specificities

Selective coordination of TM protein organization suggests that cells functionalize lipid order–based sorting by including another specificity, most likely interactions involving proteins (45). Cell membranes are crowded with membrane proteins and their associated biases in regional composition (49). Proteins can specifically organize the distribution of lipids, a property that combines with sphingolipid-cholesterol assemblage potential to produce raft-based membrane heterogeneity.

During vertical distortion of the bilayer, certain lipids of varying chain length are perturbed by the protein surface to different extents via the hydrophobic matching condition. Generally, it is assumed that the hydrophobic membrane–spanning part of the protein is stiff with no appreciable internal flexibility (50). However, by distorting lipids in the vertical direction, it would be possible to counter mismatch. The lipid species best adapted to the matching condition will have an increased probability of being close to the protein-lipid interface (50). Defined as “wetting” (51), the membrane protein surface is proposed to stabilize a sterically favored lipid environment. Electron spin resonance has identified a highly dynamic selection of boundary lipids for a number of proteins (52). However, some membrane proteins retain tightly bound lipids, even in the detergent solution present after purification (53). In such cases, lipids may have defined binding sites, where specific intercalation into protein structure is achieved (54).

Because lipids must vertically complement the rigid hydrophobic surface of the membrane domain of integral membrane proteins, variation in the protein boundary also has direct consequences for the thickness and conformational order of the bilayer (50). In model membranes, long amphiphilic peptides order and thicken the bilayer in the absence of cholesterol, whereas shorter peptides offset the membrane order and thickness induced by the presence of cholesterol (55). It has previously been argued that cholesterol-based increases in membrane thickness influence the subcellular distribution of membrane proteins in accordance with the length of their TM domain (56). Conversely, changes in protein TM length itself have been argued to be the thickness-determining factor (57).

Heterogeneity at the protein boundary is intensified by the fact that most membrane proteins are oligomeric, acting in specific macromolecular complexes to organize function (49). Superficially, these complexes are a source of steric restrictions and molecular crowding (49), but they can also incorporate specific lipids as integral features of their quaternary structure, thus functionally uniting protein-protein and lipid-protein interactions (54). Lipid incorporation is a function of specific polar–head group interactions and hydrocarbon-chain space-filling functions within the oligomeric complex (54). Many of these oligomeric structures are also formed by strong associations resistant to detergents, with the binding of cholesterol to oligomers of caveolin-1 being a prominent example (58).

In the raft field, we should be asking what it means for proteins to be wetted or, as we define the term in this context, “lubricated” (59), by specific lipids or lipid environments, particularly when it involves constituents that are important components of heterogeneity by rafts. A number of sphingolipid binding motifs have been described for membrane proteins (60). We propose that specific protein interaction with membrane-ordering “raft lipids” provides a functionalizing connection to the sphingolipid-cholesterol basis for raft assembly (Fig. 3). Interestingly, cholera toxin–cross-linking of GM1 was found to increase the partitioning of the raft enzyme beta-secretase to the Lo phase in giant unilamellar vesicles (43). As pointed out earlier, the Lo phase does not reproduce the modestly ordered, TM protein–inclusive raft structure of cell membranes. Thus, the fact that an undefined specificity for GSLs overcomes this stringent lateral sorting condition suggests that the specific lubrication of protein by lipid is likely to couple proteins to raft-based heterogeneity. Under this scheme, the functionalization of this heterogeneity depends on both lipid physical parameters and specific interactions that may include or even require proteins. For example, the TM protein LAT is an obligate component of raft-based accumulation of membrane-ordering lipids during the formation of the immunological synapse (38). As previously mentioned, membrane proteins work in functional complexes, so it is not surprising that evolution has crafted additional specificity to a lipid-based connectivity, effectively reducing lateral space with minimal energy input. A cholesterol-binding pocket, as well as six palmitate residues, has recently been identified in the crystal structure of the β2-adrenergic receptor dimer interface (61). Palmitoylation of some membrane proteins has been shown to enhance association with raft nano-assemblies (62). Perhaps in the context of forming functional protein oligomers, the propensity of palmitate for raft association is augmented by combination with cholesterol. However, whether the β2-adrenergic receptor harnesses this lipid to connect with raft-based heterogeneity is not yet clear.

Fig. 3

The lubrication of a raft TM protein by lipid. Membrane proteins bind and/or enrich certain lipids through chemical and physical specificities. These lipids may themselves exhibit sphingolipid/sterol assemblage potential. In this scheme, a TM raft protein (light blue) specifically interacts with sterol and GSL, an interaction that lubricates its inclusion to and the assembly of functionalized (coalesced) raft membrane.

Given the contribution of both physical and chemical specificities to lateral selectivity in the bilayer, lipid rafts are probably functionalized by both lipid-lipid and specific protein-lipid interactions. These lateral associations are governed by both physical and chemical specificity. Lipid-protein interactions alone cannot describe lipid rafts (63), because these do not account for the preferentially connecting lipid-lipid interactions that have so convincingly been demonstrated in model lipid membranes. Rather, we assert that sphingolipid-cholesterol assemblage potential forms a core raft connectivity that can be precisely modulated by protein specificity. In this view, raft-based membrane heterogeneity couples specific chemistries of association to the physical order preferences of lipids and proteins. Moreover, the assembly of proteins into rafts may be accompanied by conformational changes that modify protein activity.

Rafts Inside the Cell

The propensity to form heterogeneity by rafts is positively correlated with sterol content, which is maximized in the plasma membrane (6, 64), where the actin cortex also plays a central role in influencing or organizing sphingolipid-cholesterol assemblage potential (23, 25). However, for intracellular membranes the situation is less clear. But the emerging field of lipidomics is proving an important tool in evaluating both surface and intracellular membrane heterogeneity (38, 65). The concentration of sphingolipids and sterols increases along the biosynthetic pathway from the endoplasmic reticulum to the trans-Golgi network (TGN) (65). Recently, the comprehensive lipidome of immunoisolated post-Golgi carriers has revealed that order-forming sphingolipids, long and saturated glycerophospholipids, and sterols are selectively enriched in raft protein carriers bound from the TGN to the plasma membrane (66).

Thus, functional raft clustering probably underlies the lateral sorting of cell surface–destined constituents within the TGN, in keeping with the hypothesis of raft “phase” segregation principles as a means to selective carrier formation.

Compositional Evolution of the Cell Membrane

The cellular lipodome is theoretically made up of 9600 species of glycerophospholipids; more than 100,000 species of sphingolipids; thousands of mono-, di-, or triglycerides variants; and numerous fatty-acid and sterol-based structures (67). This amounts to an abundance of composition that might seem geared to dampen collective behavior in the membrane. However, this is not the case. Activating the sphingolipid-cholesterol assemblage potential in thermodynamically equilibrated plasma-membrane spheres leads to demixing of only two “phases” (P) as presently observed (44). The Gibbs phase rule states that the number of de-mixed entities (P) for a system at equilibrium is strictly correlated with the number of chemically independent components (C) by the expression P = CF + 2, where F is the number of independently variable intensive properties (i.e., pressure, temperature, and mole fractions of phase components). Thus, one could venture to postulate that these physical segregation principles have guided the coevolution of both membrane lipid and protein species, such that instead of having the vast P complexity possible from the phase rule, very few different P have survived. This could be explained by the fact that many components of the plasma membrane are not chemically independent, often forming specific complexes to reduce the lateral dimensionality of function. How then can long-range collective behavior arise from a chemically cross-talking plasma membrane? The answer is physicochemical teamwork. Activating the sphingolipid-cholesterol assemblage potential does not involve a purely physical phase transition with defined melting points and the like, but rather the coalescence of raft membrane arises through the functionally relevant cooperation of physical order (from lipid hydrocarbon chains, sterols, and the protein boundary) with specific chemical interactions (between proteins and lipids). In this sense, the cell appears to have designed a membrane composition that manipulates the physically selective behavior of lipids in a chemically specific manner, enabling organized heterogeneity to occur in the living condition.

The introduction of membrane-organizing cholesterol seems to have coincided with the evolution of multicellular complexity after the oxidation of our atmosphere (68). This correlation may imply that, in the pre-sterol era, other chemical means of reducing lateral dimensionality could have evolved. Interestingly, Caenorhabditis elegans does not use sterol as a structural element in its membranes (69). Principles of organized heterogeneity in such organisms are unknown but when revealed will potentially unravel a new chemical toolkit for membrane subcompartmentalization.


Cell membranes are complicated in composition but precise in purpose: to selectively compartmentalize the constituents of life away from environmental lifelessness. Thus, it is not surprising that membranes have innovated a means to laterally organize gatekeepers of this task. In living cells, there is strong evidence for dynamic raft-based membrane heterogeneity at the nanoscale, which can be functionally coalesced to more stable membrane-ordered assemblies. At its core, sphingolipid-cholesterol assemblage potential supplies membranes with a subcompartmentalization propensity that can be accessed or organized by proteinaceous input at little energetic cost. Raft proteins are envisioned as being equipped with a dynamic sterol-sphingolipid–dependent bias in composition at the nanoscale, allowing for the partitioning to and assembly of more stable raft platforms in the functionalized state. During raft activation, protein-lipid interactions are coupled to lipid-order–based sorting, generating heterogeneity serving to functionalize, focus, and coordinate the bioactivity of membrane constituents.

References and Notes

  1. We thank F. Jülicher, G. Feigenson, J. Howard, P. Schwille, C. Brangwynne, H.-T. He, G. van Meer, and members of the Simons lab for critical reading of this paper. A special thank you goes to M. Surma, M. Gerl, and I. Levental for their construction of and contribution to the figures. This work was supported by European Union FP6 Lipid PRISM grant no. LSHB-CT2007-037740, Deutsche Forschungsgemeinschaft Schwerpunktprogramm 1175, and Bundesministerium für Bildung und Forschung BioChance Plus grant no. 0313827.

Stay Connected to Science

Navigate This Article