The Lipid Raft Debate: What Ternary Model Membranes Settled, and What They Left Open — Epoche C1
A twenty-year argument about an object nobody could see A lipid raft is a patch of cell membrane, perhaps twenty nanometres across, that is supposed to be enriched in cholesterol and to hold together for long enough to gather particular proteins into it — and for two decades nobody could establish whether such patches exist. The reason was not a shortage of experiments but a mismatch between the object and the instruments: every technique available either could not resolve a raft or could manufacture one. This review follows the numbers that make that mismatch quantitative, shows how artificial membranes broke the deadlock by separating a question of physics from a question of cell biology, and identifies precisely which question they left open. The background is the picture rafts were proposed against. Singer and Nicolson's fluid mosaic model of 1972 described a membrane as a two-dimensional fluid — a bilayer of lipid molecules, each with a water-loving head and two water-avoiding hydrocarbon tails, in which proteins float and everything diffuses freely and evenly. In 1997 Simons and Ikonen proposed the opposite: that the fluid is not uniform, and contains rafts rich in cholesterol and in sphingolipids, a class of lipid whose long, mostly saturated tails — saturated meaning without double bonds, hence straight and able to pack closely — behave differently from the kinked, unsaturated tails of ordinary phospholipids. The proposal mattered because a raft would be a mechanism: a way for a cell to concentrate signalling proteins in one place without building a compartment. The problem: an object below the resolution of the evidence for it This section sets out why the direct approach, looking at a membrane through a microscope, could not decide the question either way. Rafts were said to be tens of nanometres across. Abbe's diffraction limit gives the smallest separation two points can have and still be told apart, $d = \lambda/(2\,\mathrm{NA})$, where $\lambda$ is the wavelength of the light and NA the numerical aperture of the lens — a measure of how wide a cone of light it collects, equal to the refractive index of the medium times the sine of the half-angle of that cone. For green light, $\lambda = 500$ nm, and a good oil-immersion objective, $\mathrm{NA} = 1.4$, this gives $d = 179$ nm. A 20 nm raft is nine times smaller: just under one order of magnitude below the resolution. The timescale was worse, and it is the more decisive of the two. A lipid in a fluid membrane has a diffusion coefficient of about $D = 1\ \mu\mathrm{m}^2\,\mathrm{s}^{-1}$. For a random walk the mean squared displacement grows as $\langle r^2\rangle = 2\,d_{\mathrm{s}}Dt$, where $d_{\mathrm{s}}$ is the number of spatial dimensions, because each independent direction contributes $2Dt$; a membrane is two-dimensional, so $\langle r^2\rangle = 4Dt$. Crossing a domain of radius $R = 10$ nm therefore takes $t_{\mathrm{cross}} = R^2/(4D) = 25\ \mu\mathrm{s}$. A video frame at 30 Hz lasts 33 ms, 1320 times longer — more than three orders of magnitude. Whatever a lipid does inside such a domain, a microscope records only the average over more than a thousand entries and exits. The proxy that failed, and why it could not have succeeded Lacking an image, the field used a biochemical proxy, and the logic of its failure is the methodological lesson of the whole episode. Brown and Rose showed in 1992 that treating cells with cold Triton X-100, a detergent, leaves an insoluble fraction rich in cholesterol, sphingolipids and proteins attached to the membrane by a glycolipid anchor rather than by a membrane-spanning helix. These detergent-resistant membranes, or DRMs, were read as rafts that had survived extraction: the detergent dissolved the disordered bulk and left the ordered patches behind. The objection, set out by Munro in 2003, was hard to answer. The extraction is done at 4 °C, well below the temperature at which such lipid mixtures order, and Heerklotz showed in 2002 by calorimetry that Triton does not merely reveal ordered domains but promotes them, shifting the mixture towards phase separation as it partitions into the bilayer. So the proxy could not separate the accounts. If rafts pre-exist, DRMs are their residue; if the membrane is uniform, the cold detergent creates the order it then isolates. Both hypotheses predict the same recovered fraction with the same composition, which means the measurement carries no information about which is true — a property no amount of reproducibility or refinement can repair, since the defect is in the inference and not in the technique. The step that broke the deadlock The way out was to stop asking about cells. Veatch and Keller (2003) built giant unilamellar vesicles — closed bags of a single lipid bilayer, tens of micrometres across, large enough to see — from a ternary mixture of a saturated phospholipid, an unsaturated phospholipid and cholesterol, with a trace of fluorescent dye. Under an ordinary microscope these show two coexisting phases, marked out by which one the dye prefers. Two observations identify them as liquids: the domains are circular, because a liquid boundary minimises its length under an isotropic boundary energy, whereas a crystalline domain takes faceted shapes reflecting its lattice; and domains merge on contact and relax back to circles, which liquids do and solids cannot. The two phases are named for what distinguishes them. In the liquid-ordered phase, written $l_o$, the hydrocarbon chains are extended and tightly packed, yet the molecules still diffuse laterally; in the liquid-disordered phase, $l_d$, the chains are loose and disordered as in an ordinary fluid membrane. The combination of chain order with lateral fluidity is what cholesterol supplies. Its rigid, flat, fused-ring body inserts alongside the saturated chains and prevents them from kinking, which raises their order, while its bulk and shape prevent the close crystalline registry that would