The definition of lipid rafts has evolved considerably over the past 15 years. They now are recognized as dynamic “nanoscale assemblies of sphingolipids, cholesterol and proteins that can be stabilized into platforms” (1) and no longer viewed as static microdomains. Confusion occurred because of the different methods used to reveal and characterize the lipid rafts and in part because of the unfortunate concomitant revival of membrane physics, which was boosted by the possibility of observing phase separation in giant unilamellar vesicles, or GUVs, with confocal microscopy (see Bagatolli and Gratton and Baumgart et al).
Physics studies with reconstituted simple lipid mixtures generally were made at equilibrium, whereas cell lipid membranes are clearly not. Nevertheless, the interplay between cell biology and membrane physics inspired other physics studies, in particular investigations into the role that membrane curvature plays in sorting lipids. It was suggested, based on in vivo observations of fluorescent lipid homologs, that lipids could be redistributed upon budding due to the high curvature of the membrane (see Mukherjee and Maxfield and van Meer and Lisman).
The development of new in vitro systems was crucial to understand and quantify the corresponding sorting mechanisms. Using membrane nanotubes pulled from GUVs made of simple lipid mixtures and controlling their radius in the 10 to100 nanometer range by setting membrane tension, different groups of physicists quantitatively assessed the conditions and the efficiency of this lipid-sorting process (see Roux et al, Sorre et al and Tian and Baumgart). An enrichment in membrane nanotube in unsaturated phosphatidylcholine, or PC, lipids as compared to sphingomyelin was measured, and it was shown to result from the reduction of the energy used to bend the membrane.
Indeed, PC-rich membranes are more flexible than those enriched in sphingomyelin. However, this effect was not detected for arbitrary mixtures, and proximity to lipid demixing, and hence lipid-lipid interactions, were critical for observing sorting (see Roux et al, Sorre et al and Tian and Baumgart); otherwise, the mechanical gain is too small and completely dominated by the mixing entropy of the lipids.
Observation of macroscopic lipid domains on giant plasma membrane vesicles blebbing from cells suggests that this membrane could be, at equilibrium, close to demixing and that this sorting mechanism might be relevant at this level in cells. Nevertheless, interactions between lipids and proteins probably are more efficient for redistributing the lipids than the membrane shape only, as they can amplify (see Sorre et al and Tian and Baumgart) or completely reverse curvature-induced lipid sorting, depending on the affinity of the protein for curved membrane. (For reviews on these questions, see Callan-Jones et al 1, Callan-Jones et al 2, Baumgart et al and Römer et al)
Another interesting aspect revealed by membrane physicists, both theoretically and with their model systems, is that lipid domains in membrane accompany a constrictive force acting at the periphery of the domains, the line tension, resulting from the nonmiscibility of the different phases. The energy relative to line tension is proportional to the perimeter of the domain; thus, by reducing the domain contour length, line tension can induce the bending of the domains in moderately tensed GUVs and even squeeze the bud down to scission (unpublished data). Squeezing of membrane nanotubes also occurs when lipid domains are present and can lead to their spontaneous scission even in the absence of any protein.
A combination of in vitro and in vivo experiments now demonstrate that this squeezing effect can occur during membrane budding if lipid heterogeneities form induced by actin filaments. This might be the case also upon BAR-protein assembling at the neck of the bud (see Liu et al and Zhao et al) (16, 17).
In conclusion, lipids can give a hand to specialized proteins to produce forces that remodel membranes. But, conversely, when membranes get deformed by coats, cytoskeleton or molecular motors, lipids and proteins can be relocalized because of shape changes in the membrane matrix.