Lipid Rafts and Membrane Microdomains

Picture a T-cell scanning for antigen, or a neuron deciding whether to fire. In both cases, the decisive event happens not in the cytoplasm but at the plane of the plasma membrane. There, a handful of receptor and kinase molecules must find each other fast. They must do this out of tens of thousands of competing lipid and protein neighbors. The cell solves this search problem the way a crowded party self-sorts into cliques. Certain lipids preferentially associate with each other and with certain proteins. This forms transient, nanoscale patches of distinct composition and physical order. These patches are lipid rafts. They are cholesterol- and sphingolipid-enriched, liquid-ordered microdomains. They float in a sea of more disordered, unsaturated glycerophospholipid. The raft concept matters because it reframes the plasma membrane. Instead of a uniform two-dimensional solvent (the classical fluid mosaic picture), it becomes something else. It becomes a laterally heterogeneous, self-organizing material. Its composition actively gates biochemistry. GPI-anchored proteins, Src-family kinases, and many GPCRs partition preferentially into ordered domains. Clathrin-independent endocytosis, viral budding, and immune synapse formation all exploit this partitioning. Diseases from Alzheimer's to HIV entry turn on getting this biophysics right. Amyloid-precursor-protein processing is raft-sensitive. HIV's gp41-mediated fusion occurs at raft-like domains. The big picture rests on a phase-behavior argument borrowed directly from physical chemistry. Mixtures of high-melting-temperature lipids (sphingomyelin, saturated phosphatidylcholines), low-melting unsaturated lipids, and cholesterol can separate into two coexisting liquid phases. These are liquid-ordered ( $L_o$ ) and liquid-disordered ( $L_d$ ). This separation occurs below a mixing critical point. Real membranes appear to sit close to this critical point. So thermal fluctuations nucleate transient $L_o$ -like nanodomains rather than macroscopic phase separation. This article works through the molecular basis of $L_o$ / $L_d$ coexistence. It covers the free-energy and line-tension arguments that set raft size. It reviews the biochemical evidence for their existence, including detergent resistance, FRET, and single-particle tracking. It also covers the sorting rules that route specific proteins into or out of these domains. By the end you should be able to do several things. You should write the condensed-complex model's free energy of mixing. You should derive why cholesterol's umbrella shape favors $L_o$ formation. You should explain quantitatively why line tension caps raft size. You should connect raft partitioning to the biology of GPI-anchor sorting and receptor clustering. This is the direct mechanistic precursor to Signal Transduction at the Membrane.

Key takeaways

Rafts are nanoscale, cholesterol/sphingolipid-rich patches in the liquid-ordered ( $L_o$ ) phase, dispersed in a liquid-disordered ( $L_d$ ) sea of unsaturated phospholipid.
Cholesterol's rigid, planar sterol ring packs against saturated acyl chains, raising conformational order without full crystallization — this is the physical origin of $L_o$ .
Line tension between $L_o$ and $L_d$ domains, opposed by the entropy of mixing and by membrane curvature/composition coupling, sets a finite equilibrium domain size rather than macroscopic phase separation.
Detergent (Triton X-100) insolubility at 4 °C historically defined "raft" biochemically, but it is now known to perturb domain size and composition — an operational, not structural, definition.
Protein partitioning is governed by transmembrane domain length/saturation, acylation (palmitoylation, GPI anchors), and cholesterol-binding motifs, not by charge or bulk hydrophobicity alone.
Rafts nucleate signaling (T-cell receptor clustering, Ras/Src activation) and are hijacked by pathogens (influenza HA/NA budding, HIV Gag assembly, cholera toxin GM1 binding).
Key idea: biological membranes sit near a liquid-ordered/liquid-disordered miscibility critical point, so cholesterol-sphingolipid rafts are thermally poised, finite-size composition fluctuations that the cell tunes to control local protein density and reaction rate.

Where this lives today

Raft biology is now inseparable from mainstream cell signaling and virology. Immune signaling: T-cell receptor engagement drives coalescence of small $L_o$ nanodomains containing Lck (a Src-family kinase anchored by dual myristoylation/palmitoylation) with the TCR/CD3 complex, concentrating kinase and substrate to trigger the immunological synapse; disrupting cholesterol (e.g., with methyl-β-cyclodextrin) blunts T-cell activation, a standard experimental raft-perturbation tool. Viral assembly and entry: influenza hemagglutinin and neuraminidase are palmitoylated and concentrate in raft-like budding sites at the apical plasma membrane; HIV-1 Gag polyprotein multimerizes at cholesterol-rich domains during particle assembly, and the gp41 fusion peptide engages ordered-domain lipids during membrane fusion — making cholesterol-depletion and raft-targeting compounds an active area of antiviral research. Toxin biology: cholera toxin B-subunit's pentavalent binding to GM1 ganglioside is the textbook raft-affinity probe, used routinely to visualize ordered domains by fluorescence microscopy. Neurodegeneration: amyloid precursor protein cleavage by β- and γ-secretase is enhanced when APP and the secretases co-localize in cholesterol-rich domains, linking membrane lipid composition to amyloid-β production in Alzheimer's disease, and motivating statin/cholesterol-lowering hypotheses (still debated clinically). Drug design and delivery: raft-targeting lipid conjugates and cholesterol-rich nanoparticle formulations are used to enhance receptor engagement or membrane fusion in lipid-nanoparticle (LNP) mRNA vaccine delivery, where the ionizable-lipid/cholesterol ratio modulates endosomal escape efficiency — a direct industrial descendant of raft biophysics. Instrumentation: super-resolution methods (STED, PALM/STORM), Förster resonance energy transfer (FRET) between raft-partitioning fluorophores, and single-particle tracking with quantum dots or gold nanoparticles remain the gold-standard tools for probing nanodomain size, lifetime, and composition in live cells, continually refining the quantitative line-tension and critical-point models developed above.

Ask the tutor: How does raft partitioning of Ras isoforms (palmitoylated vs. non-palmitoylated) quantitatively change the local density and activation kinetics of downstream MAPK signaling?

The physical picture

Start from the ternary phase diagram that underlies everything here: a mixture of a high- $T_m$ lipid (e.g., palmitoyl-sphingomyelin, PSM, $T_m \approx 41\,^{\circ}\mathrm{C}$ for the pure lipid), a low- $T_m$ unsaturated lipid (e.g., dioleoyl-phosphatidylcholine, DOPC, $T_m \approx -20\,^{\circ}\mathrm{C}$ ), and cholesterol. At physiological temperature and typical mole fractions (roughly 30–40 mol % cholesterol, comparable sphingomyelin and unsaturated PC), giant unilamellar vesicles made from this mixture visibly separate into two coexisting liquid phases under fluorescence microscopy: a liquid-ordered phase enriched in sphingomyelin and cholesterol, and a liquid-disordered phase enriched in the unsaturated lipid. Both phases are fluid — lipids diffuse laterally in both — but they differ in acyl-chain conformational order, thickness, and packing density. The molecular reason cholesterol drives this separation is its rigid, near-planar four-ring sterane nucleus fused to a short iso-branched tail and a single 3β-hydroxyl. The hydroxyl sits at the membrane–water interface (the "umbrella" polar head, too small to shield the whole sterol from water on its own — hence "umbrella model": phospholipid head groups tilt to cover the cholesterol's hydrophobic bulk). The flat ring system intercalates between saturated acyl chains and forces them into extended, nearly all-trans conformations — the same conformational ordering that occurs in a gel phase, but without the constraint of a fixed lateral lattice, so chains still diffuse freely. Unsaturated chains, by contrast, carry a rigid cis kink at each double bond that cholesterol's flat face cannot accommodate efficiently; cholesterol therefore partitions away from polyunsaturated chains and toward the straight, saturated chains typical of sphingolipids. This differential affinity is the microscopic root of $L_o$ / $L_d$ coexistence. Sphingomyelin contributes two more chemical features beyond saturation. Its amide-linked chain (from the sphingoid base) can donate a hydrogen bond from the amide N–H and its 3-hydroxyl, while the ester carbonyl of glycerophospholipids is a weaker acceptor only — sphingomyelin can hydrogen-bond to cholesterol's 3β-OH and to neighboring sphingolipid headgroups, reinforcing lateral cohesion within the ordered phase. Glycosphingolipids (e.g., GM1 gangliosides) add bulky, sugar-rich headgroups and very long, fully saturated chains, further stabilizing $L_o$ and providing specific receptor sites (GM1 binds cholera toxin B-subunit with nanomolar affinity, a classic raft marker). ```svg ``` The second key physical picture is the phase diagram itself, drawn as a function of cholesterol mole fraction $x_{chol}$ and temperature, with a miscibility critical point $T_c$ above which the two-phase region vanishes into a single homogeneous liquid. ```svg ``` Note the green point marking a plasma membrane at 37 °C: it sits close to, but just outside, the two-phase dome — consistent with experiments (Veatch & Keller, and later live-cell measurements by Levental/Baumgart) showing that isolated plasma-membrane vesicles from mammalian cells display macroscopic $L_o/L_d$ phase separation only when cooled several degrees below physiological temperature, implying the live membrane sits near — not inside — the miscibility critical point.

The core ideas

1. Regular-solution free energy of mixing (why demixing occurs at all).

\Delta G_{mix} = \Delta H_{mix} - T\Delta S_{mix}

where: $\Delta H_{mix}$ is the enthalpy of mixing two lipid species (unfavorable when saturated and unsaturated chains pack poorly together), $\Delta S_{mix}$ is the ideal entropy of mixing (always favors mixing), $T$ is absolute temperature. Reading: demixing into $L_o$ and $L_d$ phases occurs whenever unfavorable packing enthalpy between mismatched chains outweighs the entropic cost of segregating — i.e., when $\Delta H_{mix}$ is large and positive relative to $T\Delta S_{mix}$ . Raising $T$ always favors mixing (single phase); this is why the coexistence region has an upper critical solution temperature $T_c$ .

2. Regular solution model for a binary (simplified two-component) mixture:

\Delta G_{mix} = RT\left[x_A \ln x_A + x_B \ln x_B\right] + \Omega\, x_A x_B

where: $x_A, x_B = 1-x_A$ are mole fractions of the two lipid species, $\Omega$ is the interaction parameter (energy cost per mole of unlike neighbor contacts, in J/mol), $R$ the gas constant, $T$ temperature. Reading: the first (entropic) term is always negative and mixing-favoring; the $\Omega x_A x_B$ term is positive when unlike neighbors are unfavorable ( $\Omega>0$ , as for saturated/unsaturated chain mismatch) and drives phase separation once $\Omega$ exceeds a critical value relative to $RT$ .

3. Critical temperature of the regular solution model:

T_c = \frac{\Omega}{2R}

Reading: this comes from the condition that the second derivative $\partial^2 \Delta G_{mix}/\partial x_A^2 = 0$ at $x_A = 1/2$ (the spinodal condition at the symmetric composition); below $T_c$ the free-energy curve develops two minima (two coexisting phases), above $T_c$ a single minimum (fully mixed).

4. Line tension and the equilibrium raft radius. For a two-dimensional circular domain of radius $r$ embedded in a differing lipid phase, the free energy has a boundary (line tension) cost and a bulk (composition) term:

F(r) = 2\pi r \,\gamma - \pi r^2 \Delta\mu\, c + k_B T \ln\Omega(r)

where: $\gamma$ is the line tension (energy per unit length of $L_o$ – $L_d$ boundary, typically $0.1$ – $2\ \mathrm{pN}$ measured on GUVs), $\Delta\mu\,c$ is a bulk driving-force term (chemical potential difference times concentration favoring domain growth), and the logarithmic term accounts for the mixing entropy of many small domains versus one large one ( $\Omega(r)$ the number of configurational microstates). Reading: line tension alone (first term, linear in $r$ ) always favors coalescence into one large domain (macroscopic phase separation, minimizing total perimeter). The entropic term favors many small domains (more ways to arrange them, more translational entropy). Balancing these two opposing effects — rather than balancing line tension against a bulk area term alone — is what produces a finite, thermally stable domain size instead of either full de-mixing or no domains at all.

Derivations

Derivation ACritical solution temperature from the regular-solution model

Goal: show explicitly why cholesterol/sphingolipid-rich domains form below a critical temperature $T_c$ , and derive $T_c$ in terms of the microscopic interaction parameter $\Omega$ , giving the physical picture for why real membranes near 37 °C sit close to (not deep within) the two-phase region.

Step 1 — Free energy of a binary regular solution. Treat the bilayer leaflet as a two-dimensional lattice of $N$ lipid sites, each occupied by species A (ordered, e.g. sphingomyelin/cholesterol complex) or B (disordered, e.g. DOPC), with mole fractions $x_A$ and $x_B = 1-x_A$ . The mixing free energy per mole is

\Delta G_{mix}(x_A) = RT\left[x_A\ln x_A + (1-x_A)\ln(1-x_A)\right] + \Omega\, x_A(1-x_A)

The entropic term is the standard ideal mixing entropy (always $\le 0$ , minimized/most negative at $x_A = 0.5$ ). The $\Omega$ term captures the excess enthalpy of unlike A–B contacts relative to the average of A–A and B–B contacts; for saturated/unsaturated lipid pairs this is positive (unfavorable unlike contacts, poor chain packing).

Step 2 — Condition for phase splitting: convexity of $\Delta G_{mix}$ . A single mixed phase is stable only if $\Delta G_{mix}(x_A)$ is convex everywhere ( $\partial^2\Delta G_{mix}/\partial x_A^2 > 0$ for all $x_A$ ). If the curve develops a region of negative curvature, the system lowers its total free energy by splitting into two phases of different composition connected by a common tangent line (the coexistence, or "binodal," construction) — this is the same convexity argument used for any regular solution or van der Waals gas. Differentiate twice:

\frac{\partial \Delta G_{mix}}{\partial x_A} = RT\left[\ln x_A - \ln(1-x_A)\right] + \Omega(1-2x_A)

\frac{\partial^2 \Delta G_{mix}}{\partial x_A^2} = RT\left[\frac{1}{x_A} + \frac{1}{1-x_A}\right] - 2\Omega

Step 3 — Evaluate at the symmetric composition $x_A = 1/2$ (where instability first appears as $T$ is lowered). At $x_A = 1/2$ :

\frac{1}{x_A}+\frac{1}{1-x_A} = 2+2 = 4

\frac{\partial^2 \Delta G_{mix}}{\partial x_A^2}\bigg|_{x_A=1/2} = 4RT - 2\Omega

Setting this to zero defines the critical temperature, the boundary between convex (single phase, stable mixing) and non-convex (unstable, phase-separating) behavior:

4RT_c - 2\Omega = 0

T_c = \frac{\Omega}{2R}

Step 4 — Physical interpretation and scale check. Experimentally, model raft-forming ternary mixtures (DOPC/PSM/cholesterol) show miscibility critical points in giant unilamellar vesicles around $T_c \approx 20$ – $25\,^{\circ}\mathrm{C}$ (293–298 K) for compositions close to physiological ratios. Using $T_c = \Omega/2R$ with $R = 8.314\ \mathrm{J\,mol^{-1}K^{-1}}$ :

\Omega = 2RT_c = 2 \times 8.314 \times 295 \approx 4.9\times 10^{3}\ \mathrm{J/mol} \approx 4.9\ \mathrm{kJ/mol}

This is a modest interaction energy — a few kJ/mol, comparable to a single weak hydrogen bond or van der Waals contact — consistent with the fact that unlike-chain packing mismatch (not a strong chemical bond) drives the separation. Because plasma membranes contain many more lipid species (with a distribution of chain lengths/saturations, not just two idealized species) and are asymmetric across the bilayer leaflets, their effective $T_c$ is depressed and broadened relative to the clean three-component GUV systems, placing physiological 37 °C just outside (or very near) the two-phase envelope — exactly what live-cell vesicle studies observe as a sub-critical, fluctuation-dominated regime rather than stable macroscopic domains.

Limiting check: as $\Omega \to 0$ (no enthalpic penalty for unlike contacts — e.g., mixing two chemically similar unsaturated PCs), $T_c \to 0$ and the mixture stays single-phase at all accessible temperatures, correctly predicting that non-raft-forming lipid mixtures do not phase-separate.

Go deeper · Derivation B · 3 minLine tension, domain coalescence, and the finite-size raft argument

Derivation BLine tension, domain coalescence, and the finite-size raft argument

Goal: derive why $L_o$ / $L_d$ line tension, unopposed, would drive complete macroscopic phase separation (a single giant raft), and identify the physical terms that instead cap raft size at the tens-of-nanometers scale actually observed in live cells.

Step 1 — Energy of a single circular domain. Consider a circular $L_o$ domain of radius $r$ in an $L_d$ background. Its excess free energy relative to a homogeneous membrane has a perimeter (line-tension) cost:

F_{line}(r) = 2\pi r\gamma

where $\gamma$ is the line tension (SI units J/m, commonly reported in pN since 1 pN = $10^{-12}$ J/m). Measured values for $L_o/L_d$ boundaries in GUVs range from roughly $0.5$ – $3$ pN depending on composition and temperature (Baumgart, Feigenson, and coworkers).

Step 2 — Why line tension alone favors one large domain. Suppose the total ordered-phase area $A_{tot}$ is fixed (fixed total sphingolipid/cholesterol content) but distributed among $N$ identical circular domains each of area $a = A_{tot}/N$ , hence radius $r = \sqrt{A_{tot}/(\pi N)}$ . Total line-tension energy:

F_{line}^{total} = N \times 2\pi r \gamma = N\times 2\pi\gamma\sqrt{\frac{A_{tot}}{\pi N}} = 2\gamma\sqrt{\pi A_{tot} N}

This increases with $N$ (more, smaller domains have more total perimeter for fixed area — think of cutting a pizza into more slices, more crust-to-area ratio). So minimizing line tension alone always favors $N=1$ : one large domain, i.e., macroscopic phase separation. This is exactly what is observed in slowly-cooled model GUVs, but it is not what is seen in live plasma membranes, which show many small (10–200 nm), highly dynamic domains. Something must oppose coalescence.

Step 3 — Opposing term: translational/mixing entropy of many domains. Treat the $N$ domains as a 2D gas of dilute particles confined to membrane area $A_{mem}$ . Their translational entropy is

S_{trans} = k_B\ln\binom{A_{mem}/a_0}{N} \approx -Nk_B\left[\ln\left(\frac{N a_0}{A_{mem}}\right) - 1\right]

(using Stirling's approximation, with $a_0$ a molecular area unit). This entropy increases (becomes less negative) as $N$ increases at fixed total ordered area — many small domains explore more configurations than one big blob, exactly analogous to the entropy gain of dispersing a solute into many small droplets versus one large drop (opposed, in real emulsions, by surface tension in the same way).

Step 4 — Balance and scale estimate. Minimizing $F_{total}(N) = F_{line}^{total}(N) - T S_{trans}(N)$ with respect to $N$ gives an optimal domain number (and hence radius) set by the competition

\frac{dF_{line}^{total}}{dN} \sim \gamma\sqrt{\frac{A_{tot}}{N}} \quad\text{balances}\quad k_BT\ln\!\left(\frac{A_{mem}}{Na_0}\right)

Rather than solving this transcendental balance exactly, take the dimensional/scaling estimate used in the raft literature: a domain remains thermodynamically stable against unlimited growth up to a size where the line-tension energy of adding more perimeter is comparable to $k_BT$ of translational/entropic stabilization, giving a characteristic radius

r^* \sim \frac{k_BT}{\gamma}

Step 5 — Numerical estimate. At $T=310\ \mathrm{K}$ (37 °C), $k_BT = 1.38\times10^{-23}\times310 = 4.28\times10^{-21}\ \mathrm{J}$ . Taking a modest line tension $\gamma = 1\ \mathrm{pN} = 1\times10^{-12}\ \mathrm{J/m}$ (near the low end of measured values, appropriate for compositions close to the critical point where $\gamma\to 0$ ):

r^* \sim \frac{4.28\times10^{-21}\ \mathrm{J}}{1\times10^{-12}\ \mathrm{J/m}} = 4.3\times10^{-9}\ \mathrm{m} = 4.3\ \mathrm{nm}

This is of the right order of magnitude (single-digit to tens of nm) compared with the 10–200 nm domain sizes inferred from FRET, super-resolution (STED/PALM), and single-particle tracking studies of live plasma membranes — remarkably, given how simple the scaling argument is. If instead the membrane were deep in the two-phase region with a larger line tension, say $\gamma=5$ pN, the same formula gives $r^*\approx 0.9$ nm, i.e., domains shrink toward the resolution limit — consistent with the observation that domains are smallest, most numerous, and hardest to detect far from $T_c$ being approached from the low-tension side, and grow (coalesce) as $\gamma$ increases upon cooling toward and below a true phase transition.

Limiting check: as $\gamma \to 0$ (approaching the critical point, where the free-energy cost of an $L_o$ / $L_d$ interface vanishes), $r^* \to \infty$ — domains should grow without bound and fluctuate over all length scales, which is precisely the critical fluctuation behavior (diverging correlation length) observed experimentally in plasma-membrane vesicles cooled to just above their measured miscibility transition temperature. This is the direct link between the raft hypothesis and the "membranes poised near a critical point" model of Veatch, Keller, and Baumgart.

Worked examples

Example 1 (foundational): Does DPPC/DOPC/cholesterol at 1:1:0.6 sit inside the two-phase region? Given the regular-solution estimate $T_c \approx \Omega/2R$ and a literature value $\Omega \approx 5.5\ \mathrm{kJ/mol}$ for a DPPC(dipalmitoyl-PC)/DOPC/cholesterol mixture near this composition (values in this range are reported from GUV phase-diagram fits), compute $T_c$ and compare to physiological temperature.

T_c = \frac{5500\ \mathrm{J/mol}}{2\times 8.314\ \mathrm{J\,mol^{-1}K^{-1}}} = \frac{5500}{16.63} = 330.7\ \mathrm{K} = 57.5\,^{\circ}\mathrm{C}

At 37 °C (310 K), $T < T_c$ , so this composition is predicted to lie inside the two-phase coexistence region — consistent with experiment: DPPC/DOPC/cholesterol GUVs at these ratios do show visible $L_o$ / $L_d$ domain separation well above physiological temperature, which is exactly why this ternary system is the classic model membrane for studying rafts in the first place (its $T_c$ is deliberately higher than that of the live plasma membrane, making domains large enough to see by light microscopy). Limiting-case check: if instead we lowered cholesterol content toward $x_{chol}\to 0$ , the ordered phase loses its stabilizing sterol and $\Omega$ (and hence $T_c$ ) collapses toward the pure DPPC/DOPC gel–fluid value, which for a saturated/unsaturated PC pair without cholesterol is much lower — correctly predicting that cholesterol is required to produce a robust $L_o$ phase at physiological temperature, not just any saturated/unsaturated mixture.

Example 2 (applied): Estimating the fraction of a GPI-anchored protein population found in ordered domains, from a simple partition coefficient. Suppose a GPI-anchored protein has partition coefficient $K_p = [\mathrm{protein}]_{L_o}/[\mathrm{protein}]_{L_d} = 8.5$ (typical order-of-magnitude value inferred from FRET/detergent-partition assays for well-raft-associated GPI-anchored proteins), and the plasma membrane is $12\%$ ordered-phase area by fraction $\phi_{L_o} = 0.12$ (small, nanoscale raft coverage, not macroscopic phase separation). What fraction of total protein resides in $L_o$ domains at equilibrium? Let $n_o, n_d$ be the moles of protein in each phase, with area fractions $\phi_{L_o}=0.12,\ \phi_{L_d}=0.88$ . Partitioning at equilibrium means the concentrations (moles per unit area) satisfy $K_p = c_o/c_d$ , where $c_o = n_o/(\phi_{L_o}A)$ and $c_d = n_d/(\phi_{L_d}A)$ for total membrane area $A$ . So:

\frac{n_o/\phi_{L_o}}{n_d/\phi_{L_d}} = K_p \quad\Rightarrow\quad \frac{n_o}{n_d} = K_p\frac{\phi_{L_o}}{\phi_{L_d}}

\frac{n_o}{n_d} = 8.5 \times \frac{0.12}{0.88} = 8.5\times 0.1364 = 1.159

The fraction in ordered domains is

f_o = \frac{n_o}{n_o+n_d} = \frac{n_o/n_d}{1+n_o/n_d} = \frac{1.159}{2.159} = 0.537

So despite ordered domains covering only 12% of the membrane area, they contain about 54% of this GPI-anchored protein population — a striking, testable prediction: strong partitioning ( $K_p \gg 1$ ) means a minority-area phase can host a majority of a raft-philic protein, exactly the enrichment measured for GPI-anchored placental alkaline phosphatase and similar markers in detergent-resistant membrane assays. Limiting-case check: if $K_p \to 1$ (no preference), $f_o \to \phi_{L_o} = 0.12$ , i.e., the protein simply distributes according to area fraction with no enrichment — the correct null result.

Example 3 (exam/competition level): Combining line tension and partitioning to predict raft-mediated clustering of a receptor upon ligand binding. A receptor R exists as a monomer with weak raft partition coefficient $K_p^{mono}=2$ at $\phi_{L_o}=0.10$ . Upon ligand-induced dimerization, avidity for the ordered phase increases (two acylated cytoplasmic tails now cooperate), giving $K_p^{dimer} = (K_p^{mono})^2 = 4$ for the dimer's partitioning (a simple cooperative model: each raft-favoring anchor contributes an independent free-energy increment, so free energies of partitioning add and partition coefficients multiply). (a) Compute the free-energy of raft partitioning per monomer, $\Delta G_p = -RT\ln K_p^{mono}$ , at $T=310$ K.

\Delta G_p^{mono} = -RT\ln K_p^{mono} = -(8.314)(310)\ln(2) = -2577\times 0.6931 = -1786\ \mathrm{J/mol} \approx -1.79\ \mathrm{kJ/mol}

(b) Confirm the dimer's free energy is twice this (additivity check):

\Delta G_p^{dimer} = -RT\ln K_p^{dimer} = -(2577)\ln(4) = -2577\times1.386 = -3572\ \mathrm{J/mol} \approx -3.57\ \mathrm{kJ/mol}

Indeed $-3.57 \approx 2\times(-1.79) = -3.57$ kJ/mol — additivity holds exactly by construction, confirming the multiplicative partition-coefficient model is thermodynamically self-consistent (each independent raft-favoring anchor contributes an additive free-energy increment, and $\Delta G = -RT\ln K$ correctly turns products of $K$ into sums of $\Delta G$ ). (c) Using the Example 2 formula with $\phi_{L_o}=0.10,\ \phi_{L_d}=0.90$ , find $f_o$ for monomer vs dimer. Monomer:

\frac{n_o}{n_d} = 2\times\frac{0.10}{0.90} = 0.222,\qquad f_o^{mono} = \frac{0.222}{1.222} = 0.182

Dimer:

\frac{n_o}{n_d} = 4\times\frac{0.10}{0.90} = 0.444,\qquad f_o^{dimer} = \frac{0.444}{1.444} = 0.308

So ligand-induced dimerization roughly doubles raft occupancy (18% → 31%), concentrating the activated receptor into the small ordered-domain area where downstream partners (Src-family kinases, similarly raft-enriched via N-terminal myristoylation/palmitoylation) are co-localized — a quantitative, mechanistic version of the long-standing qualitative claim that "receptor clustering nucleates raft coalescence and drives signaling," directly setting up the kinetics treated under Signal Transduction at the Membrane. Limiting-case check: as ligand-induced avidity increases further (e.g., a cross-linked oligomer with $n$ cooperating anchors, $K_p^{(n)} = (K_p^{mono})^n$ ), $f_o \to 1$ as $n\to\infty$ — full raft sequestration — matching the well-documented experimental observation that antibody or toxin cross-linking of GPI-anchored proteins (e.g., anti-GPI IgM patching) drives near-complete raft coalescence visible as micron-scale patches, the classic immunological demonstration of raft-mediated clustering.

Limiting cases, scope & common traps

Regime / claim	What actually holds	Common error
"Rafts are stable, fixed structures"	Rafts are dynamic, sub-100 ms to seconds, nanoscale (~10–200 nm) composition fluctuations, not stable platforms	Treating rafts as permanent organelle-like structures visible by conventional light microscopy
Detergent resistance (DRM) assay	Triton X-100-insoluble membrane at 4 °C is operationally enriched in $L_o$ -favoring lipids/proteins	Equating "detergent-resistant membrane" with "raft" — cold detergent itself induces/enlarges ordered domains and can create artifacts
High cholesterol always stabilizes $L_o$	True only with concomitant saturated/sphingolipid chains present; cholesterol alone in a fully unsaturated bilayer can fluidize (disorder) it	Assuming cholesterol universally "orders" all membranes regardless of chain composition
GPI-anchored protein → automatically raft-associated	Depends on both anchor lipid saturation and specific protein sequence/glycosylation; some GPI proteins are largely $L_d$	Assuming any GPI or palmitoylation modification guarantees strong raft partitioning
Membrane at $T\to\infty$	Approaches ideal mixing, $\Delta S_{mix}$ dominates, single phase, no domains — as required by $T_c = \Omega/2R$	—
Equal composition $x_A=x_B=0.5$ , $\Omega \to 0$	No phase separation at any $T$ ; regular-solution model reduces to ideal solution	Forgetting that $\Omega=0$ removes the driving force even at the symmetric composition where phase separation is otherwise easiest to trigger
Transmembrane protein sorting	Longer, straighter (more saturated-mimicking) TM helices partition into thicker $L_o$ domains (hydrophobic matching); short/kinked TM helices favor thinner $L_d$	Assuming sequence alone (not TM length/hydrophobic mismatch) governs raft partitioning

Common trapMisconception: "raft" = "detergent-resistant membrane" = the same physical object

Detergent-resistant membranes (DRMs) were historically the operational definition of rafts: extract cells with cold (4 °C) Triton X-100, pellet by density gradient, and whatever floats to low density is "the raft fraction," enriched in cholesterol, sphingomyelin, GPI-anchored proteins, and Src-family kinases. This assay was foundational but is now understood to be a poor proxy for the in vivo nanodomain: cold temperature itself can drive lipids that are miscible at 37 °C into $L_o$ / $L_d$ coexistence, and the detergent perturbs lipid-protein interactions during extraction. Modern evidence for rafts instead comes from FRET between raft markers, single-particle tracking showing transient co-confinement, super-resolution imaging (STED, PALM/STORM) resolving sub-100 nm clusters, and NMR/EPR order-parameter measurements — all performed on intact, unperturbed membranes. Treat DRM composition as a rough compositional hint, never as direct structural proof of a pre-existing raft.

Common trapMisconception: rafts undergo the same phase transition as pure-lipid model membranes

Giant unilamellar vesicles made of DPPC/DOPC/cholesterol show macroscopic, micron-scale $L_o$ / $L_d$ domains visible by ordinary fluorescence microscopy — a true first-order phase transition below a well-defined $T_c$ . Plasma membranes are compositionally far more complex (dozens of lipid species, asymmetric leaflets, an actin cortex, and thousands of membrane proteins), and live-cell studies show they sit close to, not deep within, a miscibility critical point at physiological temperature. The correct picture is critical-point-proximal fluctuations (correlation lengths of tens of nanometers, lifetimes of milliseconds) rather than stable macroscopic phase coexistence; scaling naive GUV phase diagrams directly onto living cells overstates both the size and the stability of real rafts.

The chemist behind it

Origins · 2 minKai Simons and how this was discovered

Kai Simons

1938–2022 · Cell biologist, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden

Kai Simons trained originally in medicine and biochemistry in Helsinki before moving into membrane biology at the European Molecular Biology Laboratory (EMBL) in Heidelberg, where his work on epithelial cell polarity in the 1980s led him to a puzzle: apical and basolateral plasma membrane domains of polarized epithelial cells have strikingly different protein and lipid compositions, yet both are made by the same biosynthetic (Golgi) machinery and separated by no physical barrier other than tight junctions. Simons and collaborators (notably Gerrit van Meer) proposed that sphingolipids and cholesterol could self-associate into ordered "rafts" in the Golgi membrane, acting as sorting platforms that concentrate apically-destined GPI-anchored and other lipid-modified proteins into transport vesicles — the 1988 van Meer–Simons sorting hypothesis. Simons, with Elina Ikonen, formalized and popularized the general "lipid raft" concept in a landmark 1997 Nature review, connecting the sorting hypothesis to detergent-insolubility biochemistry and to signal-transduction biology (T-cell receptor and Fc-receptor clustering). He went on to found the Max Planck Institute of Molecular Cell Biology and Genetics in Dresden and remained a central, sometimes deliberately provocative, advocate for raft biology through decades in which the field's methods (fluorescence correlation spectroscopy, super-resolution imaging, single-molecule tracking) matured to test — and substantially revise — his original static-platform picture into the modern dynamic, critical-fluctuation model.

History

- 1972 — Singer and Nicolson propose the fluid mosaic model, the baseline "uniform 2D fluid" picture that rafts later revise.

- 1988 — van Meer and Simons propose sphingolipid/cholesterol-based sorting of apical membrane proteins in polarized epithelial cells.

- 1992 — Brown and Rose isolate detergent-resistant, glycolipid-enriched membrane fractions (DRMs), giving the raft hypothesis its first widely used biochemical assay.

- 1997 — Simons and Ikonen publish the influential Nature review coining and popularizing "lipid rafts" as a general membrane-organizing principle.

- 2003–2008 — Veatch, Keller, and Baumgart demonstrate micron-scale

L_o

L_d

phase coexistence directly in giant unilamellar and plasma-membrane-derived vesicles, quantifying line tension and phase diagrams.

- 2008–2016 — Super-resolution (STED) and single-particle tracking (Eggeling, Kusumi) resolve nanoscale (<100 nm), transient raft-like clusters in live cell membranes, refining the static-platform model into a dynamic, critical-fluctuation model.

Are lipid rafts a proven structure, or still a contested hypothesis?

Rafts are not directly "seen" as static structures the way, say, a ribosome is seen by cryo-EM; they are inferred from converging biophysical evidence — detergent resistance, FRET between raft markers, single-particle tracking showing transient co-confinement, and super-resolution imaging resolving sub-100 nm, sub-second clusters. The modern consensus (supported by Veatch/Keller/Baumgart-style phase-diagram work and live-cell critical-fluctuation studies) is that rafts are real but dynamic: nanoscale, short-lived composition fluctuations near a liquid-ordered/disordered miscibility critical point, actively stabilized and enlarged by protein clustering, rather than pre-formed static platforms. The debate today is mostly about size, lifetime, and the degree of active (versus purely thermodynamic) control, not about whether cholesterol-sphingolipid ordered domains exist at all.

Why does the plasma membrane sit so close to a phase-separation critical point instead of safely in one regime or the other?

Sitting near a critical point maximizes the membrane's susceptibility to small perturbations — composition fluctuations, temperature changes, or protein binding can nucleate or dissolve ordered domains with very little energetic cost, because the free-energy cost of forming an $L_o$ / $L_d$ interface (the line tension $\gamma$ ) approaches zero near $T_c$ . This gives the cell a cheap, fast, and reversible way to modulate local membrane order in response to signaling events (e.g., receptor clustering) without paying the large enthalpic cost of nucleating a domain far from criticality. It is essentially the same logic organisms use elsewhere: poising a system near a critical or bifurcation point buys sensitivity and tunability at low energetic cost, at the price of some inherent noise.

How do transmembrane protein sequences determine raft partitioning?

Hydrophobic matching is central: the liquid-ordered phase is thicker (more extended, all-trans acyl chains) than the liquid-disordered phase, so transmembrane helices with longer, straighter hydrophobic stretches match $L_o$ thickness better and partition there, while shorter or kinked/bulky-residue-containing helices mismatch $L_o$ and prefer the thinner $L_d$ phase. Palmitoylation (a 16-carbon saturated thioester-linked chain on cytoplasmic cysteines) is a major additional driver, since the saturated acyl chain packs favorably with sphingolipid/cholesterol-rich $L_o$ domains — this is why Src-family kinases (myristoylated + palmitoylated) and Ras isoforms (palmitoylated H-Ras/N-Ras vs. non-palmitoylated K-Ras) show markedly different raft affinities and correspondingly different signaling kinetics.

What experimental methods actually measure raft size and lifetime in live cells?

Fluorescence correlation spectroscopy (FCS) and its super-resolution variant (STED-FCS) measure how diffusion coefficients depend on the observation area, revealing transient trapping in nanodomains from characteristic deviations from simple Brownian diffusion. Single-particle tracking (quantum dots, gold nanoparticles, or fluorescent proteins tracked at high frame rate) directly visualizes brief co-confinement events, typically tens of milliseconds. FRET between two differently labeled raft markers reports on their average nanoscale proximity. Super-resolution imaging (PALM/STORM, STED) resolves spatial clustering below the diffraction limit, typically finding raft-like clusters in the 10–200 nm range with lifetimes from milliseconds to a few seconds — far smaller and shorter-lived than the micron-scale, minutes-to-hours-stable domains seen in slowly-cooled model GUVs.

Is cholesterol depletion (e.g., with methyl-β-cyclodextrin) a clean way to test for raft involvement?

It is the standard tool, but not a perfectly clean one. Methyl-β-cyclodextrin (MβCD) extracts cholesterol from the outer leaflet by forming an inclusion complex within its hydrophobic cavity, rapidly disordering $L_o$ -like domains and commonly used to test whether a process (e.g., a signaling pathway or viral entry step) is "raft-dependent." However, cholesterol also has independent roles in membrane fluidity, permeability, and direct binding to specific proteins (cholesterol-recognition amino acid consensus, CRAC, motifs), so a functional effect of MβCD treatment does not uniquely prove raft/domain involvement — it must be combined with other evidence (partitioning assays, imaging) for a rigorous conclusion.

How does the raft concept connect to sphingolipid and glycolipid chemistry specifically?

Sphingolipids contribute the saturated, often very long (C22–C24) acyl chains and the amide-linked hydrogen-bonding backbone that stabilizes tight lateral packing in the ordered phase, distinguishing them chemically from the more commonly unsaturated glycerophospholipids that dominate the disordered phase. Glycosphingolipids add bulky, hydrophilic sugar headgroups (e.g., the GM1 ganglioside pentasaccharide) that can mediate specific carbohydrate-lectin recognition — cholera toxin, Shiga toxin, and several viral capsid proteins bind glycosphingolipid headgroups displayed at the surface of raft-like domains, making sphingolipid identity both a physical determinant of domain formation and a direct biochemical recognition code layered on top of it.

Do rafts exist only in the plasma membrane, or in internal organelle membranes too?

The original hypothesis (van Meer/Simons) actually began in the Golgi, proposing that sphingolipid/cholesterol-based sorting occurs during the trans-Golgi network's segregation of apical versus basolateral cargo in polarized epithelial cells. Cholesterol and sphingolipid concentrations increase progressively along the secretory pathway (lowest in the ER, highest in the plasma membrane), and evidence for ordered-domain-like sorting has been reported in the trans-Golgi network and in endosomal membranes, though it is best characterized at the plasma membrane where imaging and biochemical tools are most mature.