The Two Faces of the Liquid Ordered Phase.

Coexisting liquid ordered (Lo) and liquid disordered (Ld) lipid phases in synthetic and plasma membrane-derived vesicles are commonly used to model the heterogeneity of biological membranes, including their putative ordered rafts. However, raft-associated proteins exclusively partition to the Ld and not the Lo phase in these model systems. We believe that the difference stems from the different microscopic structures of the lipid rafts at physiological temperature and the Lo phase studied at room temperature. To probe this structural diversity across temperatures, we performed atomistic molecular dynamics simulations, differential scanning calorimetry, and fluorescence spectroscopy on Lo phase membranes. Our results suggest that raft-associated proteins are excluded from the Lo phase at room temperature due to the presence of a stiff, hexagonally packed lipid structure. This structure melts upon heating, which could lead to the preferential solvation of proteins by order-preferring lipids. This structural transition is manifested as a subtle crossover in membrane properties; yet, both temperature regimes still fulfill the definition of the Lo phase. We postulate that in the compositionally complex plasma membrane and in vesicles derived therefrom, both molecular structures can be present depending on the local lipid composition. These structural differences must be taken into account when using synthetic or plasma membrane-derived vesicles as a model for cellular membrane heterogeneity below the physiological temperature.

When inserted into certain phospholipid membranes in sufficient amounts, cholesterol (CHOL) induces the liquid ordered (Lo) phase,[1] which is an intermediate between the fluid liquid disordered (Ld) and solid (gel) phases. As the name suggests, the Lo phase displays gel-like ordering of the lipid acyl chains, whereas the lipid rotational and translational dynamics are relatively fast akin to the Ld phase, highlighting the fluid character of the membrane. The addition of CHOL increases the order, thickness, and stiffness of the Ld phase, yet has the opposite effects on the gel phase.[2] As a result, CHOL renders these two phases more similar. Binary mixtures of CHOL and certain phospholipids have a threshold CHOL concentration, above which the first-order Ld–gel transition at Tm vanishes. For dipalmitoylphosphatidylcholine (DPPC), this happens at ∼25–30 mol % of CHOL, above which many membrane properties show continuous temperature dependence, indicating a uniform Lo phase. This behavior is reported by numerous experimental approaches[3−8] and simulations.[9−12] Still, some studies have reported nonidealities within the Lo phase: differential scanning calorimetry (DSC) and NMR detected a broad transition,[13,14] which was suggested to correspond to the transition between two distinct Lo phases differing in the lipid chain tilt.[14,15] NMR spectroscopy and X-ray scattering suggested that even at high CHOL concentrations, heating modifies the lipid structure at the glycerol level, repositions CHOL toward lipid headgroups, and disorders the acyl chain termini.[16,17]

Certain ternary mixtures of CHOL with low-Tm and high-Tm lipids display Lo/Ld phase coexistence[18] and are commonly used to model the putative nanoscale ordered lipid domains (“rafts”) in biomembranes.[19,20] Notably, this coexistence is detected in model systems only at temperatures below the Tm of the high-Tm lipid. Also, giant plasma membrane-derived vesicles (GPMVs) phase-separate well below the body temperature.[21] This agrees with the ability of CHOL to break the gel phase to an Lo one. Indeed, recent simulations have demonstrated that the Lo phase has an internal structure containing small hexagonally packed and CHOL-depleted regions,[22−24] despite CHOL’s preference for saturated lipid chains in fluid phases.[25] Experiments have struggled to provide a consistent molecular view of the Lo phase in binary mixtures, and the situation is at least equally complicated for ternary mixtures: NMR[26] and X-ray scattering[27] resolve signals originating from the two coexisting phases; yet, their compositions are temperature-dependent. Importantly, it is unclear whether the Lo phase observed in phase-separated vesicles differs structurally from that observed for the same mixture at physiological T and whether the former is a faithful model for plasma cell membrane heterogeneity.[28] Surprisingly, raft-associated proteins partition to the Ld phase in phase-separated synthetic vesicles,[29] although the same proteins can locate to the ordered phase in phase-separated GMPVs.[30] Moreover, there are even differences between the partitioning behavior of proteins in GMPVs[31] and plasma membrane spheres[32] since the latter can maintain phase separation at higher temperatures due to mechanisms not present in model systems.[32] Still, very few transmembrane domains are targeted to the Lo phase even in phase-separated GMPVs.[31,33] In general, the two coexisting phases in different model systems and in vivo have different levels of mutual similarity.[34−37] These findings indeed suggest that the putative rafts likely differ from the Lo phase observed in synthetic vesicles more than only by their size,[21,28] and some of these factors might also be temperature-dependent.[38]

To shed light on the structure of ordered lipid phases across temperatures, we performed differential scanning calorimetry (DSC) and fluorescence measurements of binary and ternary Lo lipid mixtures, and provided a molecular picture of the observed phenomena using atomistic molecular dynamics simulations. We studied bilayers consisting of (1) a ternary mixture of 55 mol % DPPC, 15 mol % DOPC, and 30 mol % CHOL, and (2) a binary mixture of 70 mol % DPPC and 30 mol % CHOL. The composition of the ternary mixture reflects the Lo phase of a phase-separated 0.40/0.40/0.20 (DPPC/DOPC/CHOL) lipid bilayer at 298 K.[22,39] On the basis of the phase diagrams, both mixtures remain in the Lo phase at all studied temperatures, ranging from 293 K to 333 K.[8,39,40]

Our simulations revealed distinct changes in the bilayer structure depending on the temperature, as shown in Figure (see Figures S1 and S2 for more temperatures). Both in binary and ternary mixtures, DPPC chains mainly adapt an anti conformation at 293 K; yet, the amount of gauche isomers is significantly increased at 333 K. However, at 313 K the two mixtures differ as DPPC lipids remain substantially ordered in the binary mixture. This indicates that DPPC undergoes some kind of a transition, yet at different temperatures for the two mixtures. The question is whether this transition is manifested in membrane properties.

Figure 1

Final structures (after 1 μs of simulation) of ternary (left) and binary (right) mixtures at 293 K (top), 313 K (middle), and 333 K (bottom). The chains are mainly in an anti conformation at 293 K; yet, they melt at 333 K, resulting in an increase in the amount of chains in the gauche conformation. Here, DPPC is shown in green, DOPC in blue, and CHOL in yellow. Lipid hydrogens and ions are omitted. Water is shown in red and white.

As demonstrated in Figure , the answer is affirmative. The area per phospholipid (APPL, shown in Figure A) exhibits two regimes with different thermal expansion coefficients and a crossover at Tcoter ≈ 308 K (ternary mixture) or Tcobin ≈ 318 K (binary mixture). The coefficients differ between the binary and ternary mixtures in both temperature regimes. The ternary mixture contains fluid DOPC, which explains the slightly larger coefficient below Tco.[41]

Figure 2

Temperature dependence of membrane properties for the binary and ternary mixtures. The crosses show data extracted from additional simulations that were performed to ensure that hysteresis did not affect the results (see Methods in the Supporting Information). (A) Area per phospholipid (APPL) shows two different slopes, i.e., two different thermal expansion coefficients (solid and dashed lines) with a crossover at either Tcoter ≈ 308 K or Tcobin ≈ 318 K, highlighted by the vertical dotted lines. (B) The correlation time of the autocorrelation function of the orientation of the glycerol backbone. The autocorrelation data are shown in Figure S3. (C) The hexatic order parameter of the lipid chains. The parameter values range from 0 to 1. No change of slope in the ternary system is observed for DOPC.

DSC is the tool of choice to detect phase transitions or more subtle molecular rearrangements. We performed such measurements on large unilamellar vesicles for the binary and ternary mixtures (see Methods in the Supporting Information). As shown in Figure A, DSC detects broad peaks for the binary and ternary systems, and their maxima agree perfectly with the crossover temperatures found in our simulations (see labels in Figure A). The absence of a first-order phase transition suggests that a pure Lo phase is present at all temperatures. Nonetheless, it undergoes a minor structural transition. Measurements of pure DPPC reveal a well-defined sharp gel–liquid transition at Tm = 314 K, whereas binary and ternary mixtures with only 15% CHOL display Lo/gel coexistence at low T. This results in the presence of both a sharp and a broad peak (Figure S10), which are associated with the melting of CHOL-poor gel and CHOL-rich Lo phases.[13,15]

Figure 4

Results from differential scanning calorimetry and fluorescence spectroscopy. (A) Specific heat capacity. The profiles were fitted by two or three Gaussians, and the fitted data are shown here with a logarithmic y axis for easier comparison. The darker and lighter curves correspond to cooling and heating scans, respectively. The original data as well as the fits are shown in Figure S10, and the parameters obtained from the analysis of the DSC curves are available in Table S2. (B) Generalized polarization parameter with Patman. Two independent measurements were performed for each sample (except for DPPC), yet they overlap. The data for binary and ternary mixtures show a change of curvature (inflection point); yet POPC lacks it, as it is in the Ld phase at all measured temperatures. The spectra are shown in Figure S11, and the parameters extracted from fitting a modified Boltzmann curve are available in Table S3. (C) DPH anisotropy. Two samples were again measured for compositions other than DPPC, and they are almost identical. The spectra are shown in Figure S12 and the fitting parameters from a modified Boltzmann curve in Table S4. DPH anisotropy becomes wavelength-dependent at higher temperatures (see Figure S12), leading to the deviation of the measured data from the fit. Therefore, points measured above 333 K were excluded from the fitting process.

We then verified whether lipids really exhibit Lo-like fluidity and order both below and above Tco. The characteristic rotational times of the DPPC glycerol backbone in the binary and ternary systems are shown in Figure B. The values are indeed considerably smaller than those of the gel phase.[42] The rotational times decrease continuously upon heating in the studied temperature range by two orders of magnitude both for binary and ternary mixtures. Our values agree reasonably well with those measured for the binary mixture at 285 K.[43] The fluid character of the lipid bilayer can also be concluded by comparing lateral diffusion coefficients of lipids, shown in Figure S4, with experimental data in different phases.[3] Even though the membranes can be considered fluid across all studied temperatures, the dynamics still speed up significantly above Tco. Additionally, we verified that the mean deuterium order parameters show high Lo-like ordering at all studied temperatures (see Figure S5).

Next, we looked into lipid packing in the bilayers in detail. We used the in-plane hexatic order parameter Shex of chosen atoms in the acyl chain region to characterize the degree of hexagonal arrangement of lipid molecules.[22−24] As shown in Figure C, DPPC shows relatively high Shex values for both the binary and ternary mixtures below Tco. However, after the crossover they converge rapidly toward values extracted for DOPC in the ternary mixture. The Shex of DPPC is slightly higher in the binary mixture due to the lack of packing perturbations by DOPC. The absence of any crossovers in the DOPC curve indicates that this lipid is excluded from the hexagonally packed clusters. These findings agree with our movies showing the melting of the clusters in the ternary mixture at Tcoter, available at DOI: 10.6084/m9.figshare.13176167. Lipid exchange rates in these clusters, shown in Figure S6, also show a crossover at Tco.

We next evaluated whether structural changes, suggested previously to take place within the Lo phase,[13−17] are present in our simulations. As shown in Figure A, the mean tilt angle of DPPC acyl chains shows two different slopes with crossovers at Tco, in line with the predicted variation in chain tilting within the Lo region.[15] Clarke et al. and Reinl et al. suggested that the temperature increase within the Lo phase shifts CHOL toward the headgroup region, leading to the disordering of the lipid chain terminal carbons.[16,17] This repositioning of CHOL toward the headgroup region indeed takes place in our simulations (Figure S7). This process further decreases the acyl chain ordering in the membrane core. Indeed, as shown in Figure B, the mean deuterium order parameter of the last nine carbons of the DPPC sn-2 chain displays a clear decrease with crossovers at Tco. The glycerol region melting, suggested to occur within the Lo phase by Clarke et al.,[16] also takes place as indicated by the significant increase in glycerol dynamics in Figure B, albeit with a less pronounced crossover.

Figure 3

Membrane properties which were suggested to differ across temperature in the Lo regime by earlier experimental studies. The crosses correspond to additional simulations that demonstrate that hysteresis does not affect the obtained results (see Methods in the Supporting Information). (A) DPPC chain tilt shows a change in slope at the crossover temperatures of Tcoter ≈ 308 K and Tcobin ≈ 318 K. (B) The mean order parameter of all DPPC sn-2 chains, starting from the eighth carbon, i.e., the middle of the chain. The carbon-wise order parameter plots are shown in Figure S5. (C) Mean order parameter of DPPC chains as a function of cluster identity (bars) and the fractions of the different cluster identities (lines). “Core” chains are surrounded by hexagonally packed chains, “free” chains are not part of a hexagonally packed cluster, and “edge” points roughly correspond to the cluster edges.

DPPC chains can be categorized based on their local structure. Our algorithm (see Methods in the Supporting Information) distinguishes “core” chains encapsulated in the tightly-packed clusters, “edge” chains mainly at cluster edges, and “free” chains that are excluded from the clusters (see Figure S8 for demonstration). The mean deuterium order parameter , shown in Figure C, demonstrates that “core” chains are the most ordered at all temperatures. Notably, they are also the furthest (0.90 nm on average) away from CHOL molecules. The “free” chains are the least ordered and closest to a CHOL (0.62 nm), whereas “edge” chains show intermediate behavior. This agrees with the exclusion of CHOL from the tightly packed clusters in ref (22). As shown in Figure S9, the “core” chains are also the least tilted, followed by the “edge” chains, and finally the “free” chains. The temperature dependencies of both and chain tilt consist of two factors, as demonstrated in Figures C and S9: The “core”, “edge”, and “free” chain populations depend on the temperature with the “core” fraction declining rapidly after Tco, corresponding to the melting of the clusters. Moreover, while “free” chains show a continuous change in ordering and tilt, there is a crossover in the values of the “core” and “edge” chains at Tco. These two factors contribute to the striking change of slope in the temperature dependencies of these properties in Figure .

To provide further experimental evidence for the described changes in the molecular structure, we performed two complementary fluorescence experiments on large unilamellar vesicles. First, we utilized Patman—a fluorescent polarity probe stably located at the lipid carbonyl region of all membrane phases.[44,45] Upon electronic excitation, a charge transfer over the naphthalene ring provides a major change in the dipole moment of Patman. Subsequently, its Stokes shift provides information on the hydration and mobility of the carbonyl region—both highly sensitive to the lipid phase state.[46] Patman generalized polarization (GP) was calculated based on the probe emission at 420 nm (typical for gel and Lo phases) and 495 nm (characteristic for Ld). Thus, GP contains information on both lipid fluidity and hydration. High GP values associated with gel and Lo phases, and low ones with the Ld phase. For details, see the Methods in the Supporting Information.

GP results of the binary and ternary mixtures, as well as DPPC (gel–LdTm = 314 K) and POPC (Ld across all measured temperatures), are shown in Figure B. In all systems, the emitted wavelengths red-shift gradually when the temperature increases, reflecting an increased mobility of the lipid carbonyls (Figure S11). For POPC, GP reflects increasing lipid mobility in the Ld phase with increasing temperature. This change is continuous and saturates at higher temperatures. On the other hand, GP measured for DPPC displays an abrupt jump at Tm as the lipid mobility changes drastically. Please note that below and above Tm, GP is still decreasing with increasing temperature. The GP curves for the binary and ternary mixtures present character intermediate to that of POPC and DPPC. The changes are much smoother than for DPPC, but the fits by modified Boltzmann sigmoidal functions (see the Supporting Information) still display clear inflection points that correspond to the Tco extracted from our simulations and DSC (see Table S3 for all the fitting parameters).

The second fluorescence technique we used is based on a different principle and provides complementary information about the lipid hydrocarbon chain region. Diphenylhexatriene (DPH) is a rod-like hydrophobic fluorescent probe with simple geometry and directional emission of polarized light. It orients along the lipid hydrocarbon chains and measures their order and dynamics—both indicative of the membrane phase.[47] Our previous studies using time-resolved anisotropy measurements demonstrated great sensitivity of DPH to the ordering effect of different amounts and different species of sterols in the lipid bilayer.[48,49]

The temperature dependence of the anisotropy emission spectra of DPH is shown in Figure S12. Similarly to GP discussed above, the average anisotropy values (Figure C) show a gradual decrease in anisotropy upon a temperature increase for both the ternary and binary mixtures. This change reflects a decreased order and increased dynamics of the lipid acyl chains. Noteworthy, the qualitative similarity of the results obtained using Patman GP and DPH anisotropy reassures us that the observed changes can be detected both in the bilayer core, but also closer to the headgroup region, and are not limited to a single fluorescent probe or detection method used. The data can again be fitted by the modified Boltzmann sigmoidal functions (Figure B). Interestingly, Patman reports the changes to take place at a slightly lower temperature—particularly for the binary system—indicating that the melting of the hexagonal packing of DPPC chains initiates from the carbonyl region probed by Patman, before it proceeds to the chain region probed by DPH. This finding is in agreement with the previously observed initial melting of the glycerol backbone.[16,17] All of the fitting parameters are provided in Table S4. Please note that for higher temperatures (>340 K), DPH does not provide trustworthy data for DPPC and the binary mixture. This is likely caused by the probe relocation and can be detected by the wavelength-dependent anisotropy data (Figure S12C,E; see also ref (50)). This is not observed in POPC nor in the ternary mixture.

Summarizing, we have provided a molecular-level view into the structures of the Lo phase across a range of temperatures in a binary DPPC/CHOL mixture and a ternary DPPC/DOPC/CHOL mixtures. Both mixtures are expected to be in the Lo phase across the studied temperature range from 293 to 333 K. Still, we discovered two different regimes separated by a crossover at either Tcoter ≈ 308 K or Tcobin ≈ 318 K; yet, no lipid demixing was observed.

The hexagonally packed clusters of DPPC chains that are present at low temperatures melt upon heating at Tco, which affects many structural and dynamic membrane properties. Interestingly, the temperature dependencies of these properties were different not only below and above the crossover temperature but also between the binary and ternary mixtures. Still, the properties of the two mixtures converged toward each other at low and high temperatures. The properties of the binary mixture were more different on the two sides of the crossover temperature, highlighting that the presence of DOPC is able to buffer many properties in the ternary mixture. Still, all studied properties change continuously with temperature, and no radical changes akin to a first-order phase transition were observed.

On the experimental side, both the DSC scans and the fluorescence measurements suggest that a structural transition takes place at temperatures close to Tco obtained from simulations. Moreover, both experimental approaches indicate that this transition is more subtle than a proper first-order phase transition. Despite the use of probes in fluorescence experiments, the limitations of the simulation models, and the possible differences in the vesicles generated in two laboratories, the estimated transition temperatures differ by a mere few K. Moreover, all approaches systematically suggest that the Tco values differ by ∼10 K between the ternary and binary mixtures. Despite this difference, the behavior of the ternary and binary mixtures is strikingly similar.

Curiously, the two very different molecular structures at different temperatures—one with gel-like hexagonally packed DPPC clusters among more fluid regions and the other consisting of a fully fluid structure with only slightly preferred interactions among the components—are both fluid: The characteristic time scales of lipid rotational and translational motions differ by only two orders of magnitude and thus all fall within the Lo-like regime between the Ld and gel phases. The Lo-like high values for the deuterium order parameters are also reproduced by our simulations at all the studied temperatures, i.e., by structures where CHOL either is in direct contact with the chains or excluded from the hexagonally packed clusters of DPPC chains. Therefore, it seems that the two distinct molecular structures where CHOL has two very different roles—either ordering nearby-residing DPPC chains (above Tco) or lubricating the space between the gel-like DPPC clusters (below Tco)—lead to similar average behaviors that are indistinguishable by many experiments.

CHOL induces the Lo/Ld coexistence in mixtures with a low-Tm lipid and a high-Tm lipid by melting the gel phase (formed mainly by the high-Tm lipid) that would exist in the mixture in the absence of CHOL. However, this “melting” only corresponds to breaking the hexagonally packed lipids into smaller and mobile clusters. While this leads to Lo-like behavior, a large fraction of the lipid chains are still in the gel-like highly ordered yet untilted state. Only above Tco do the hexagonally packed clusters really melt, leading to proper lipid mixing. Curiously, coexisting Lo and Ld phases in phase-separated DPPC/DOPC/CHOL liposomes mix above 308 K (= Tcoter).[27] Thus, it seems likely that the hexagonally packed clusters are required to sustain the Lo/Ld coexistence. Indeed, CHOL cannot induce this coexistence from any uniform Ld-phase mixture above Tm of the high-Tm lipid, supporting this idea.

The differences between the two molecular-level structures of the Lo phase in these two temperature regimes might play key roles in biomembranes. Importantly, on the basis of our findings, the use of Lo/Ld coexistence observed in synthetic or giant plasma membrane vesicles at relatively low temperatures as a model for lipid rafts existing at body temperature should be questioned, as the structures of the ordered domains in these systems might be very different. This might explain the discrepancies in protein partitioning between in vitro studies exploiting phase-separated vesicles and indirect in vivo data. Synthetic and giant plasma membrane vesicles display Lo/Ld coexistence only at room temperature or lower. In this temperature range, the Lo phase possibly contains clusters of very tightly packed lipid chains—a structure which is likely unable to solvate large transmembrane protein segments. Indeed, raft-associated proteins were found in the Ld phase in synthetic lipid mixtures containing sphingomyelin[29] for which similar tight lipid chain packing has been observed.[51] The effect of temperature is manifested by an increased tendency of peptides to partition to the ordered detergent-resistant membranes at 310 K as compared to 277 K.[38] While this finding was originally associated with hydrophobic mismatch,[38] it could as well result from changes in the molecular structure of these membranes.

Coexisting phases in plasma membrane-derived vesicles are compositionally and thus also structurally more similar than in their synthetic counterparts,[28,34,35,37] suggesting that their Lo phases might not be as rigid as those in synthetic mixtures. Since the plasma membrane is compositionally complex,[52] it is possible that critical fluctuations[39] can assemble domains with varying lifetimes and compositions.[37] In this picture, domains containing very long chain sphingolipids[52] could still have a Tco that is higher than the physiological temperature, suggesting that both of the Lo structures described in this study could coexist in the plasma membrane.