Induction of Ordered Lipid Raft Domain Formation by Loss of Lipid Asymmetry

Abstract

How lipid asymmetry impacts ordered lipid domain (raft) formation may yield important clues to how ordered domain formation is regulated in vivo. Under some conditions, a sphingomyelin (SM) and cholesterol-rich ordered domain in one leaflet induces ordered domain formation in the corresponding region of an opposite leaflet composed of unsaturated phosphatidylcholine (PC) and cholesterol. In other conditions, the formation of ordered domains in a SM and cholesterol-rich leaflet can be suppressed by an opposite leaflet containing unsaturated PC and cholesterol. To explore how PC unsaturation influences the balance between these behaviors, domain formation was studied in asymmetric and symmetric lipid vesicles composed of egg SM, cholesterol, and either unsaturated dioleoyl PC (DOPC) or 1-palmitoyl 2-oleoyl PC (POPC). The temperature dependence of ordered domain formation was measured using Förster resonance energy transfer, which detects nanodomains as well as large domains. In cholesterol-containing asymmetric SM+PC outside/PC inside vesicles, the PC-containing inner leaflet tended to destabilize ordered domain formation in the SM+PC-containing outer leaflet relative to ordered domain stability in cholesterol-containing symmetric SM/PC vesicles. Residual ordered domain formation was detected in cholesterol-containing asymmetric SM+DOPC outside/DOPC inside vesicles, but ordered domain formation was completely or almost completely suppressed by asymmetry in cholesterol-containing SM+POPC outside/POPC inside vesicles over the entire temperature range measured. Suppression of ordered domain formation in the latter vesicles was confirmed by fluorescence anisotropy measurements. Because mixtures of SM, POPC, and cholesterol form domains in symmetric vesicles, and this lipid composition mimics plasma membranes to a significant degree, it is possible that under some conditions in vivo the loss of lipid asymmetry could trigger ordered domain formation.

Significance

Lipid vesicles having lipid asymmetry mimicking that of biological membranes were used to investigate the principles governing ordered lipid domain formation by sphingolipids and cholesterol. Interactions between inner and outer lipid monolayers can modulate whether such domains do or do not form in lipid vesicles. In vesicles with lipid composition and asymmetry most analogous to that found in natural membranes (containing sphingomyelin, POPC, and cholesterol), sphingolipid-lacking inner leaflets suppressed ordered domain formation in sphingolipid-containing outer leaflets, although ordered domain formation was observed in the corresponding symmetric vesicles. Therefore, loss of asymmetry may have the potential to trigger ordered domain formation in mammalian cells under some conditions.

Introduction

Rules defining the conditions and lipid compositions in which ordered lipid domains (lipid rafts) form have been studied extensively in symmetric lipid vesicles, which are vesicles having the same lipid composition in each leaflet (1). This has contributed greatly to our understanding of how ordered lipid domains might form in the plasma membrane of eukaryotic cells. However, the lipid composition of mammalian plasma membranes is highly asymmetric, with ordered domain-forming sphingolipids largely restricted to the outer leaflet (2). This limits the applicability of studies using symmetric lipid vesicles. The recent development of methods to prepare asymmetric vesicles now allows more relevant studies to be carried out (3).

Asymmetry is particularly important for understanding how physical properties and domain formation in one leaflet of a lipid bilayer will be influenced (coupled) to those in the opposite leaflet. Interleaflet coupling has a variety of aspects. In this report, we investigate the extent of interleaflet coupling, defined here as the degree to which the physical properties of the lipids in one leaflet are altered as a consequence of interactions with lipids in the opposite leaflet. If the properties of one leaflet are altered so that they become almost the same as those of lipids in the other leaflet, there is a high extent of coupling. In addition, the question of coupling dominance is evaluated. This refers to which leaflet dominates the final physical state (ordered or disordered) when two leaflets have highly coupled behavior (3). For example, consider a lipid bilayer when one leaflet has a lipid composition that by itself would form an ordered state (e.g., liquid ordered, Lo) in a symmetric bilayer, whereas the other has a lipid composition that by itself would form a disordered state (Ld) in a symmetric bilayer. In the case of Lo dominance, contact between Lo-preferring and Ld-preferring leaflets would result in the Ld-preferring leaflet becoming ordered. In contrast, in the case of Ld dominance, contact between Lo-preferring and Ld-preferring leaflets results in a disordering of the Lo leaflet, so that it becomes more similar to the Ld leaflet. If neither leaflet is dominant but there is a high extent of coupling, the properties in both leaflets could assume an intermediate physical state. In contrast, in cases of no coupling, properties in the two leaflets would be unaltered relative to that in symmetric vesicles.

Other aspects of coupling can be evaluated experimentally. One is the range of temperatures over which physical properties in the two leaflets are coupled. Another is coupling energy, which can be defined as the energy needed to replace a leaflet in one physical state with a leaflet in a different physical state. Coupling energy is a measure of coupling strength that can be assayed for large ordered domains that remain intact when ordered domains in opposite leaflets are pulled apart (4).

All of the coupling parameters described above do not identify the physical origin of coupling, which involves defining which interactions result in coupling (e.g., interdigitation of acyl chains). An increase in the number or strength of specific interactions between lipids in opposing leaflets should be associated with an increase in both the extent and strength of coupling.

Various coupling behaviors have been observed experimentally. In some cases, leaflets with ordered domains have been found to be dominant (5, 6, 7, 8). In other cases, leaflets forming the Ld state are dominant (9,10). (However, a caveat for studies carried out by light microscopy is that the disappearance of large ordered domains does not distinguish the conversion of ordered domain into disordered domains from conversion of large ordered domains into submicroscopic ordered nanodomains). It has also been observed that under some conditions, ordered domain formation or properties in one leaflet appear to be uncoupled to domain formation in the opposite leaflet (5,6,9,11). In yet other cases, ordered domains have been found to be softened by contact with a disordered state leaflet (12). Finally, it has been found that interleaflet coupling can be controlled by the direction of curvature of the leaflet in which the ordered domain-favoring composition is located. For example, coupling has been found to differ when the most strongly ordered domain-favoring composition is in the inner or outer leaflet (13). Thus, many possible consequences of interleaflet coupling have been observed.

For these reasons, understanding how lipid structure can influence lipid domain formation behavior in cells requires investigating the role of lipid structure upon interleaflet coupling and ordered domain formation in more detail. Along these lines, we previously investigated cholesterol and phospholipid-containing asymmetric vesicles in which the structure of outer leaflet lipids with saturated acyl chains was varied (10). Lipids with saturated acyl chains tend to form the Lo state when cholesterol is present, but it was found that lipids with relatively short saturated acyl chains were unable to maintain ordered domain formation in cholesterol-containing asymmetric vesicles containing unsaturated phospholipids in the inner leaflet, even though ordered domain formation could be detected in the corresponding symmetric vesicles.

In this report, we extended this by investigating asymmetric vesicles in which the structure of the lipid with unsaturated acyl chains, which tend to promote formation of the Ld state, was varied. Specifically, the behavior of asymmetric vesicles containing sphingomyelin (SM), cholesterol, and either POPC or DOPC in their outer leaflets and a mixture of cholesterol and either POPC or DOPC in their inner leaflets was compared to that of symmetric vesicles containing the same lipids. As in previous studies, the temperature above which lipid domains disappear, i.e., the miscibility transition temperature, was used as a measure of ordered domain stability. It was found that asymmetry inhibited the formation of ordered domains. Specifically, in outer leaflets of asymmetric membranes containing SM and phosphatidylcholine (PC), there was less ordered domain formation relative to that in symmetric vesicles of a similar overall lipid composition. However, the inhibition of ordered domain formation by POPC, which is abundant in mammalian membranes, was much greater than inhibition by DOPC. The greater tendency to form ordered domains in symmetric vesicles than asymmetric ones for the compositions studied raises the possibility that ordered domains could form when asymmetry is lost. This may have implications for ordered domain formation in biological membranes in vivo. In addition, it shows that care must be taken when extrapolating behavior in artificial vesicles with PCs with unnatural acyl chains to biological behavior.

Materials and Methods

Materials

Egg SM, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), cholesterol, 1,2-dioleoylphosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (rhodamine-DOPE) were purchased from Avanti Polar Lipids (Alabaster, AL). 1,6-diphenyl-1,3,5-hexatriene (DPH) and 2-hydroxypropyl-α-cyclodextrin (HPαCD) were purchased from Sigma-Aldrich (St. Louis, MO). 1-(4-trimethylammoniumphenyl)-6-phenyl-1,3,5-hexatriene p-toluenesulfonate (TMADPH) was purchased from the Molecular Probes (Eugene, OR) division of Invitrogen (Carlsbad, CA). Lipids were dissolved in chloroform. DPH and TMADPH were dissolved in ethanol. Lipids and probes were stored at −20°C. The concentrations of lipids were determined by dry weight and that of fluorescent probes by absorbance using ε_{Rhodamine-DOPE} 88,000 M⁻¹cm⁻¹ at 560 nm, ε_DPH 84,800 M⁻¹cm⁻¹ at 352 nm, and ε_TMADPH 74,000 M⁻¹ at 350 nM. 1× phosphate-buffered saline (PBS) (10 mM Na phosphate and 150 mM NaCl (pH 7.8)) was obtained as a 10× solution from Bio-Rad Laboratories (Hercules, CA). Sucrose was obtained from Sigma-Aldrich, and a solution of 25.2% (w/w) was prepared in distilled water. High performance thin layer chromatography (TLC) plates (Silica Gel 60) were purchased from VWR International (Batavia, IL). Methyl-α cyclodextrin (MαCD) was purchased from AraChem Cyclodextrin-Shop (Tilburg, the Netherlands), dissolved in distilled water at close to 100 mM, and then filtered through a Sarstedt (Nümbrecht, Germany) 0.2-μm pore syringe filter. At the high MαCD concentrations used, the exact MαCD concentration was most conveniently assayed by comparing the refractive index of the solutions to that of a standard curve solution made with a known weight of cyclodextrin dissolved in a known final volume of solution. HPαCD was dissolved in 385 mM distilled water. Absorbance was measured using a Beckman 640 spectrophotometer (Beckman Instruments, Fullerton, CA) using quartz cuvettes.

Preparation of symmetric large unilamellar vesicles

Lipids dissolved in chloroform were combined in glass tubes, dried under a warm nitrogen stream, and then subjected to high vacuum for 1 h. To prepare samples with Förster resonance energy transfer (FRET) acceptor, 1 mol % rhodamine-DOPE was included in the lipid mixture. The dried lipid mixtures were placed in a 70°C water bath and dispersed in 25.2% (w/w) sucrose to a final volume of 500 μL at 2 mM lipid. The samples were vortexed briefly and then agitated at 55°C for 15 min using a VWR Multi-Tube Vortexer placed within a convection oven (GCA, Precision Scientific, Chicago, IL). The lipid dispersions were then cooled to room temperature and subjected to seven cycles of freeze thaw in a liquid nitrogen bath, alternating with a 37°C water bath. To form large unilamellar vesicles (LUVs) of uniform vesicle size, the lipid mixtures were then extruded 11 times through 100-nm pore polycarbonate membranes (Avanti Polar Lipids). To wash away external sucrose, 333 μL aliquots of LUV formed in sucrose were mixed with 3.7 mL 1× PBS and pelleted by ultracentrifugation at 190,000 × g for 25 min at 23°C using a Beckman L8-80M ultracentrifuge with a SW-60 rotor. The LUV pellets were dispersed in 333 μL 1× or 2.5× PBS, covered with aluminum foil, and reserved for use. Unless otherwise noted, samples were used within 2 h of preparation. Lipid concentration was ∼2 mM (assuming no loss of lipid after ultracentrifugation). For FRET measurements, 50 μL aliquots of LUV lipid mixtures were dispersed in cuvettes containing 940 μL of either 1× PBS or 2.5× PBS to give a final lipid concentration of ∼100 μM.

Preparation of lipid-loaded MαCD for lipid exchange

Desired ratios of egg SM and DOPC or POPC were combined in glass tubes, dried under a warm nitrogen stream, and then subjected to high vacuum for 1 h. To prepare samples with FRET acceptor, an additional 13 or 15 mol % rhodamine-DOPE was added to the unlabeled lipids before drying. This high concentration of rhodamine-DOPE was needed because it does not transfer to acceptor vesicles efficiently and results in roughly 1 mol % rhodamine-DOPE in the final asymmetric LUVs (14). The dried lipids were placed in a 70°C water bath and dispersed with 205 μL of prewarmed (70°C) 1× PBS and 128 μL of prewarmed (70°C) 104 mM MαCD to give a final concentration of 40 mM MαCD and 16 mM lipid. The samples were vortexed briefly, agitated at 55°C as above for 15 min, cooled to room temperature, covered in foil, and reserved for further use.

Preparation of asymmetric LUVs

75 mol % DOPC (or POPC) and 25 mol % cholesterol dissolved in chloroform were combined in glass tubes. The mixtures were dried under a warm nitrogen stream and then subjected to high vacuum for 1 h. They were then placed in a 70°C water bath and dispersed to 8 mM lipid with 25.2% (w/w) sucrose in distilled water. The samples were vortexed briefly and agitated at 55°C for 15 min as above. LUVs were then prepared from the lipid mixtures as described above. After mixing 333 μL of these vesicles with 333 μL of the vesicle-MαCD mixtures described above, lipid exchange to prepare asymmetric LUVs was carried out for 30 min at room temperature (23°C), as described previously (14). Final lipid composition was determined by high performance TLC as described previously (14). Note that MαCD exchange does not alter cholesterol levels (7). Typical preparations of asymmetric LUVs contained roughly 2 mM lipid (as determined by TLC) suspended in 100–333 μL of 1× or 2.5× PBS. For fluorescence measurements, lipid was diluted to ∼100 μM with 1× or 2.5× PBS.

FRET measurements

Before DPH (FRET donor) was added, background (lipid-only) measurements of fluorescence intensity at 427 nm at room temperature were made on 990 μL aliquots of symmetric LUV or asymmetric LUVs prepared as described above using an excitation wavelength of 358 nm on a Horiba QuantaMaster Spectrofluorimeter (HORIBA Scientific, Edison, NJ) using quartz semimicro cuvettes (excitation path length 10 mm and emission path length 4 mm). The slits were set to 3-nm bandpass for excitation and 6-nm bandpass for emission. Next 10 μL of 10 μM DPH dissolved in ethanol was added to each cuvette to give a DPH concentration of 0.1 μM, and samples were vortexed vigorously in the cuvette several times during a 20-min incubation at room temperature. Samples were cooled to 15°C, and DPH fluorescence intensity was continuously measured as samples were heated at a rate of 0.8°C per minute, up to 64°C, unless otherwise noted. Sample temperature was controlled with a Quantum Northwest TC 1 Temperature Controller (Liberty Lake, Washington). Because of sample warming lag, corrected sample temperatures were calculated from a sample in which a probe thermometer (digital thermometer with a YSI microprobe; Thermo Fisher Scientific, Waltham, MA) had been inserted during a calibration sample heating.

FRET is expressed in terms of the fraction of donor fluorescence unquenched by acceptor (F/Fo). (FRET efficiency is 1 − F/Fo.) The effective Ro for the DPH-rhodamine-DOPE pair in membranes is ∼3.6 nm (15). Background fluorescence intensity, generally <1% of sample, was subtracted before computing F/Fo values. F/Fo ratios were normalized by dividing the F/Fo ratio calculated at each temperature to the F/Fo ratio at 64°C. Fluorescence from three different F samples from three different exchange preparations and from one Fo sample prepared by combining three different Fo preparations without rhodamine-DOPE were used. Normalized F/Fo results were analyzed to calculate Lo transition/melting midpoints (T_mid) and endpoints (T_end), the latter defined by the minimum of F/Fo vs. temperature, as described previously (14).

Heat scrambling of lipids

To scramble the asymmetric distribution of lipids, aliquots of asymmetric LUVs dispersed in PBS were sealed in glass tubes, incubated in a heat block for 30 min at almost 100°C, and after cooling to room temperature, used for FRET measurements.

Measurement of fluorescence anisotropy

Anisotropy measurements were performed at 16 and 23°C using a Spex automated Glan-Thompson polarizer accessory (HORIBA Scientific). Symmetric LUVs and asymmetric LUVs composed of SM, POPC, and cholesterol were prepared. Asymmetric LUVs were prepared using the HPαCD protocol previously published (10). Samples were prepared in 1× PBS at an approximate lipid concentration of 200 μM. TMADPH and DPH fluorescence intensity was measured at λ_ex of 364 nm, λ_em of 426 nm, or λ_ex of 358 nm, and λ_em of 427 nm, respectively, and slits bandpass were set 3 nm for excitation and 3 nm for emission. Background fluorescence was measured before fluorescent probe was added. Then, ∼0.1 mol % (relative to lipids) of DPH or TMADPH was added from concentrated (25 and 200 μM, respectively) stock solutions in ethanol and incubated for 10 min in the dark. Then, fluorescence was remeasured. Anisotropy was calculated using the equation: A = [(Ivv × Ihh)/(Ivh × Ihv)−1]/[(Ivv × Ihh)/Ivh × Ihv) + 2], where A is anisotropy, and I values are fluorescence intensities for vertical (v) and horizontal (h) orientations of excitation and emission polarization filters, respectively. The background values were not subtracted because of their low fluorescence intensities, <1% of that in samples with fluorescent probe.

Results

FRET measurement of ordered domain thermal stability

Ordered domains (liquid ordered (Lo), except at low cholesterol concentrations where gel domains can form (16)), and disordered (Ld) domains can co-exist in membranes containing ternary mixtures of 1) cholesterol, 2) a lipid with a high gel/Ld melting temperature when in a pure bilayer (e.g., SM), and 3) a lipid with cis unsaturated acyl chains and thus a low gel/Ld melting temperature in a pure bilayer (here, DOPC or POPC (17)). The thermal stability of ordered domains in lipid vesicles containing such mixtures was evaluated by FRET. FRET was assayed by measuring F/Fo, the fraction of fluorescence of the FRET donor (DPH) unquenched by the presence of the FRET acceptor (rhodamine-DOPE) (10,14). FRET efficiency equals 1 − F/Fo. Rhodamine-DOPE preferentially partitions into disordered domains (7), whereas DPH generally partitions almost equally between ordered and disordered domains (18). A consequence of this is that FRET is weaker (F/Fo is higher) in domain containing bilayers than in homogenous bilayers (Fig. 1 A). The reason is that upon separation into ordered and disordered domains, there are two populations of DPH, the population in the ordered domains (which are depleted in FRET acceptor) and that in the Ld domains (which are enriched in FRET acceptor). Because of the fact that FRET is not linear with regard to FRET acceptor concentration, the lesser FRET (higher F/Fo) from the DPH population in ordered domains, which is partly segregated from acceptor, more than compensates for the stronger FRET (lower F/Fo) arising from DPH in Ld domains. At higher temperatures, at which ordered domains “melt”/become miscible with Ld lipids, there is a decrease in the average distance between DPH molecules that were in the ordered domains and rhodamine-DOPE, and overall FRET increases (F/Fo decreases).

Figure 1.

Simplified schematic illustration of FRET experiment. (A) At lower temperatures, the FRET acceptor (A) preferentially partitions into disordered domains and is partly segregated from the subpopulation of FRET donor (D) in the ordered domains. Upon heating, melting/mixing of ordered and disordered domains decreases the overall fraction of FRET donor fluorescence unquenched by FRET acceptor (F/Fo) (see main text for details). (B) Schematic illustration of melting/mixing parameters. F/Fo versus temperature is shown for two samples containing ordered domains with different melting/mixing temperatures. The sample denoted by the solid curve has a higher melting temperature than the sample represented by the dashed line. The definitions of T_mid (apparent midpoint of the melting transition) and T_end (the approximate end of the melting transition) are schematically illustrated. Notice that F/Fo can be weakly temperature dependent for a sample with no ordered domains (straight line) (see main text for details).

The thermal stability of ordered domains can be defined by T_mid, the apparent midpoint temperature (inflection point) in the F/Fo versus temperature curve (14). When ordered domains are too thermally unstable to measure T_mid, one can estimate T_end, the approximate endpoint temperature for the miscibility transition (14). Fig. 1 B schematically illustrates these parameters. It should also be noted that because of the variability in delivery of rhodamine-DOPE into the outer leaflet of asymmetric LUVs, absolute F/Fo values can vary, so F/Fo values were normalized to that at high temperature, as previously (14). Absolute F/Fo values at high temperature, from which absolute F/Fo values at all temperatures can be calculated, are given in Table S1.

There also can be a weak dependence of FRET upon temperature in the absence of domains (10,19). One reason for this is the competition between FRET and other processes that destroy the donor excited state. As the fluorescence lifetime generally decreases with increasing temperature, i.e., the length of time during which FRET can rob the donor of excited state energy and thus prevent donor fluorescence decreases, so the vulnerability of the donor to FRET decreases. A second issue affecting the temperature dependence of FRET is that it is orientation dependent, and relative donor and acceptor orientations and motions (which can average out different orientations) will likely be temperature dependent (15).

Ordered domain stability in symmetric LUVs

We used FRET to define how low Tm lipid unsaturation influences ordered domain formation by comparing the behavior of symmetric LUVs to that of asymmetric LUVs containing mixtures of SM, cholesterol, and PC (DOPC and/or POPC). Ordered domain stability was first measured in symmetric vesicles composed of 25 mol % cholesterol and 1:1 (mol:mol) SM:POPC, 2:1:1 SM:POPC:DOPC, or 1:1 SM:DOPC. Consistent with previous studies (20), relatively stable domains were detected by the sigmoidal decrease in F/Fo versus temperature as temperature increased (Fig. 2). T_mid was ∼48°C and T_end was 62°C for SM:DOPC:cholesterol vesicles and slightly lower for SM:POPC:cholesterol vesicles, 35 and 57°C, respectively (Table 1).

Figure 2.

Domain melting/mixing curves in symmetric vesicles assayed by FRET. All samples contained 100 μM lipid and 25 mol % cholesterol. Symmetric vesicles contained (A) 1:1 SM:POPC, (B) 2:1:1 SM:POPC:DOPC, and (C) 1:1 SM:DOPC (filled triangles). Control vesicles contained (A) 3:1 (mol:mol) POPC:cholesterol or (C) 3:1 (mol/mol) DOPC:cholesterol (open circles). The fraction of DPH fluorescence intensity unquenched by rhodamine-DOPE (F/Fo) versus temperature normalized to F/Fo 64°C is shown. The average (mean) F/Fo value and SD is shown for three experiments. Note that error bars were too small to illustrate in many cases.

Table 1.

T_mid and/or T_end Values in Symmetric LUVs and Asymmetric LUVs with Different Lipid Compositions

Samples (All with 25 mol % Cholesterol)	Transition Temperature°C
Symmetric Vesicles	Tmid	Tend
1:1 SM:POPC	34.9 ± 0.4	57.3 ± 0.1
2:1:1 SM:POPC:DOPC	36.2 ± 0.4	53.7 ± 0.8
1:1 SM:DOPC	48.4 ± 0.2	62.0 ± 0.6
SM:POPCout/POPCin Heat Scrambled	32.1 ± 3.9	54.5 ± 4.4
Asymmetric vesicles	–	–
SM:POPCout/DOPCin	<16	34.4 ± 3.0
SM:DOPCout/POPCin	<16	36.5 ± 1.7
SM:DOPCout/DOPCin	<16	40.7 ± 0.4
SM:POPCout/POPCin	not detected	not detected

T_mid and T_end were calculated as described in Materials and Methods. Mean and SD from three experiments is shown (Figs. 2and3).

Vesicles with SM, cholesterol, and a 1:1 mixture of POPC and DOPC had domain-forming behavior similar to that of SM:POPC:cholesterol (compare Fig. 2 B with Fig. 2 A). In contrast to vesicles containing SM, control samples composed of POPC:cholesterol or DOPC:cholesterol showed only a weak temperature dependence of F/Fo, as expected. (However, a small amount of ordered domain formation in POPC/cholesterol at very low temperatures cannot be ruled out).

It should be noted that to facilitate comparison to asymmetric LUVs (see below), the symmetric LUVs contained trapped sucrose and were prepared with 2.5× PBS in the external solution, which avoids a significant osmotic pressure across the bilayer. Domain-forming behavior when symmetric LUVs were dispersed in 1× PBS was very similar to that in 2.5× PBS (see Fig. S1).

Ordered domain stability in asymmetric LUVs

Next, ordered domain stability was measured in asymmetric LUVs formed by exchanging a mixture of SM and PC with rhodamine-DOPE into LUVs composed of 3:1 mol:mol DOPC:cholesterol or 3:1 mol:mol POPC:cholesterol. SM replaced ∼30–45% of the PC in these vesicles, depending on the PC used (Table S2). This corresponds to ∼60–85% of the outer leaflet PC (Table S2). Previous studies have shown that after exchange asymmetry is stable for days, and SM and rhodamine-DOPE reside in the outer leaflet of the asymmetric LUVs (10,21). The previously introduced nomenclature for asymmetric vesicles is used. In this nomenclature, the lipids in the outer leaflet are stated first, then the lipids in the inner leaflet, and finally, if present, cholesterol, which is in both leaflets (7,10,14,22). The subscripts “out” and “in” are used to refer to the lipids of the outside and inside leaflets.

Fig. 3 A shows a comparison of SM:DOPC_out/DOPC_in/cholesterol asymmetric LUV to control DOPC:cholesterol vesicles that do not form ordered domains. (Control samples were prepared by introducing rhodamine-DOPE into the outer leaflet of DOPC:cholesterol vesicles via lipid exchange using a donor mixture of DOPC and rhodamine-DOPE mixed with MαCD). The asymmetric LUV showed ordered domain formation in which T_mid was below the temperature range studied (i.e.,<16°C) and T_end ∼41°C (Table 1). These values are lower than in the corresponding symmetric SM:DOPC:cholesterol LUVs (Fig. 2 C), showing that the ordered domains in these asymmetric LUVs were less thermally stable than in the corresponding symmetric LUVs, even though the asymmetric LUVs contained more SM in their outer leaflet than that in the outer leaflet of the corresponding symmetric LUVs.

Figure 3.

Effect of phospholipid acyl chain structure upon outer leaflet ordered domain thermal stability in asymmetric LUVs. All asymmetric LUVs and control LUVs contained 25% cholesterol. In (A) and (B), vesicles contained the same PC species in the outer and inner leaflets. (A) SM:DOPC_out/DOPC_in/cholesterol. (B) SM:POPC_out/POPC_in/cholesterol. In (C and D), vesicles contained different PC species in the outer and inner leaflets. (C) SM:DOPC_out/POPC_in/cholesterol. (D) SM:POPC_out/DOPC_in/cholesterol. Control samples for (A and D) were DOPC:cholesterol and for (B and C) were POPC:cholesterol. Domain formation in (filled circles) asymmetric LUVs and the (open circles) controls are shown. The fraction of DPH fluorescence intensity unquenched by rhodamine lipid (F/Fo) versus temperature normalized to F/Fo 64°C is shown. The average (mean) F/Fo value and SD from three experiments are shown.

Even more strikingly, Fig. 3 B shows that in asymmetric SM:POPC_out/POPC_in/cholesterol LUVs, there was little ordered domain formation, with the F/Fo versus temperature curve being very similar to that for control POPC:cholesterol vesicles (prepared by exchanging POPC:rhodamine-DOPE into the outer leaflet of POPC:cholesterol LUVs). This is in contrast to the behavior of symmetric SM:POPC:cholesterol vesicles, which form ordered domains that are almost as thermally stable as those in symmetric SM:DOPC:cholesterol. It should be noted that the slightly lesser amount of SM in the asymmetric SM:POPC_out/POPC_in/cholesterol LUV relative to the asymmetric SM:DOPC_out/DOPC_in/cholesterol LUV does not explain the loss of ordered domains in the former case, as shown by anisotropy experiments (see below).

Additional experiments were carried out with asymmetric LUVs containing different PC species in their outer and inner leaflets (Fig. 3, C and D). For both SM:POPC _out/DOPC_in/cholesterol and SM:DOPC_out/POPC_in/cholesterol asymmetric LUVs, very weak domain formation was observed, with T_mid below the temperature range studied and T_end ∼34 and 37°C, respectively. Thus, the behavior of these asymmetric LUVs fell between that of the SM:DOPC_out/DOPC_in/cholesterol asymmetric LUVs and SM:POPC_out/POPC_in/cholesterol asymmetric LUVs.

As in case of the symmetric LUVs, the PBS concentration in the external solution did not greatly affect ordered domain thermal stability in the asymmetric LUVs (Fig. S1). This was also true for the control vesicles (data not shown).

The difference between the stability of ordered domains in symmetric and asymmetric LUVs raises the possibility that loss of asymmetry might induce domain formation. To confirm this, the domain-forming behavior of asymmetric SM:POPC_out/POPC_in/cholesterol LUVs was compared to that of symmetric LUVs with the identical overall lipid composition. As shown in Fig. 4, the symmetric vesicles with this lipid composition form relatively stable ordered domains even though the asymmetric LUVs do not. To further confirm this, when samples of the asymmetric LUVs were preheated to near 100°C to increase the rate of lipid flip and degrade asymmetry, the resulting samples formed ordered domains with T_mid ∼32°C and T_end 54°C (Fig. 4; Table 1), indicating that there had been a significant, although not necessarily complete, loss of asymmetry. Preheating samples to just 64°C, as occurs during the temperature scans for measuring the temperature dependence of FRET, had no effect on FRET, indicating a negligible degradation of asymmetry upon heating to 64°C (Fig. S2). The difference between heating to 64°C and near 100°C is not surprising because the lipid flip-flop is very temperature dependent, increasing greatly as temperature is increased, but can still take hours at 65°C (23).

Figure 4.

Effect of loss of lipid asymmetry on ordered domain thermal stability. Behavior of asymmetric SM:POPC_out/POPC_in/cholesterol LUVs with 25 mol % cholesterol is shown before (filled circles) and after they were subjected to lipid scrambling by preheating at almost 100°C for 30 min (open circles). Ordered domain thermal stability is also shown for symmetric LUVs (triangles) composed of 38:62 (mol:mol) SM:POPC and 25 mol % cholesterol, a lipid composition equivalent to the lipid composition of the asymmetric LUVs. The fraction of DPH fluorescence intensity unquenched by rhodamine lipid (F/Fo) versus temperature is shown normalized to F/Fo at 64°C. The average (mean) F/Fo and SD is shown from three experiments.

An additional control showed preheating symmetric SM:POPC:cholesterol vesicles to near 100°C and then cooling did not alter subsequent measurement of F/Fo versus temperature when compared to the sample before heating to 100°C (Fig. S3). Also, TLC of samples heated to 100°C showed no changes (data not shown). These results indicate that there were no significant changes to the lipids themselves upon heating.

Fluorescence anisotropy measurements confirm inhibition of ordered domain formation in asymmetric LUVs

Fluorescence anisotropy experiments were carried out to confirm that formation of ordered domains does not occur in asymmetric LUVs composed of SM:POPC_out/POPC_in/cholesterol. Fig. 5 shows the anisotropy at 23°C of TMADPH (Fig. 5 A) and of DPH (Fig. 5 B) added to symmetric and asymmetric LUVs. Fluorescence anisotropy measures membrane order, with TMADPH reporting order in the outer leaflet and DPH reporting the average for both leaflets. The standard curve for symmetric vesicles (filled circles) shows the expected increase in anisotropy with increasing amounts of SM, reflecting an increase in order, which mainly reflects the increasing formation of ordered domains as SM levels are increased. Despite the presence of SM, at either ∼60 or ∼80% of the outer leaflet, the anisotropy in the asymmetric LUVs (open symbols) was close to that of vesicles lacking SM, indicating that there was no or little formation of ordered domains. This was true for the outer leaflet, as shown by TMADPH anisotropy, and the overall bilayer, as shown by DPH anisotropy. These results are indicative of suppression of outer leaflet ordered domain formation due to asymmetry, i.e., the inner leaflet dominating physical properties of the asymmetric LUVs. Similar anisotropy results were obtained when anisotropy was measured at 16°C (Fig. S4).

Figure 5.

Effect of SM content in symmetric and asymmetric LUVs upon fluorescence anisotropy at 23°C. (A) TMADPH anisotropy. (B) DPH anisotropy. (Filled circles) Anisotropy of symmetric LUVs with different SM/POPC ratios and 25 mol % cholesterol. Vesicles contained trapped 25% w/w sucrose and were dispersed in 1× PBS. (Open symbols) Anisotropy of asymmetric LUVs prepared with two different ratios of SM to POPC. Mean anisotropy and SD from three separate experiments are shown. The overall percentage of SM in total phospholipid (SM/(SM+POPC) × 100%) in the two versions of asymmetric LUV preparations was 30 and 41%. The percentage of outer leaflet SM was calculated from this assuming that SM transferred into the asymmetric LUVs resides in the outer leaflet and that the outer leaflet contains 52% of the total lipid in the LUVs. Inner leaflet composition in the asymmetric LUVs is ∼75% POPC/∼25% cholesterol. The SD for the percentage of SM in the asymmetric LUVs (SM/(POPC+SM) × 100%) was ∼3% of total SM+POPC content. Note that the anisotropy error bars were too small to illustrate in most cases.

It should be noted that if the outer leaflet had induced ordered domain formation in the inner leaflet, the predicted anisotropy values would be close to that seen in symmetric vesicles having the same amount of SM as in the outer leaflet of the asymmetric LUVs. Such anisotropy behavior, indicative of the outer leaflet dominating asymmetric vesicle physical properties, has been observed with other lipid compositions (10).

Discussion

Differences between ordered domain formation in asymmetric and symmetric LUVs

In this report, we used asymmetric LUVs to study how phospholipids that tend to exist in the disordered state alter the behavior of the lipids in a leaflet containing phospholipids that tend to form ordered domains. In particular, we compared the effect of the often studied but less biologically relevant DOPC to the more biologically abundant lipid POPC. It should be noted that the methods used do not distinguish between ordered domains in the gel state and Lo state. However, for the lipid mixtures used, prior studies have shown that the Lo state should be present for the lipid compositions studied here (16).

The results showed that there was a high extent of interleaflet coupling (see Introduction) such that the inner leaflet, containing only unsaturated lipid and cholesterol, dominated physical properties, inhibiting ordered domain formation in an outer leaflet containing SM over a wide temperature range. This was especially true for asymmetric LUVs containing POPC. Anisotropy showed that there was lesser ordered domain formation in asymmetric LUVs relative to symmetric ones, even when the asymmetric vesicles had a larger fraction of SM in their outer leaflet than that in the outer leaflet of the symmetric LUVs. The reason that POPC inhibits outer leaflet ordered domain formation to a greater extent than DOPC is unclear. It is possible that part of the reason is that inner leaflet POPC interferes with tight lipid packing in the outer leaflet via a greater degree of acyl chain interdigitation between the leaflets relative to that with DOPC (6).

It should be noted that the ability to suppress ordered domain formation in asymmetric LUVs is not unique to POPC. In cholesterol-containing asymmetric LUVs in which a saturated lipid (di C14:0 PC) with a more modest ability to form ordered domains than SM was studied, even DOPC in the inner leaflet was able to completely suppress outer leaflet ordered domain formation (10). Suppression of large-scale phase separation by a saturated PC by diphytanoyl PC has also been reported (9). However, it is not clear in that case if ordered nanodomains too small to detect by microscopy remained after the large-scale ordered phase disappeared.

It should also be noted that the cholesterol-containing mixtures of SM and POPC studied here come closer to that of mammalian membranes than vesicles containing DOPC, or diphytanoyl PC. Even more natural lipid mixtures and the effect of proteins could be studied in the future, especially as the composition and asymmetry of natural mammalian plasma membranes becomes better defined.

It is important to point out that it is certainly not the case that outer leaflet saturated lipids can never dominate the inner leaflet lipid to induce ordered domains in the inner leaflet. For example, in asymmetric giant unilamellar vesicles with a SM:DOPC_out/DOPC_in/cholesterol composition, we found previously that the SM-containing outer leaflet dominated coupling and induced (large) inner leaflet ordered domains up to 30–33°C (7), about the same temperature range over which outer leaflet domains were observed in this study in asymmetric LUVs with a similar lipid composition. Other studies have shown that the outer leaflet distearoyl PC can dominate and induce ordered domain formation in inner leaflets containing DOPC (8). In addition, in some prior studies with asymmetric vesicles, we found POPC-containing inner leaflets did not always dominate the physical behavior of a SM-containing outer leaflet (11). Those studies used small unilamellar vesicles, which have a higher outer to inner leaflet lipid ratio than larger vesicles and vesicles that lacked cholesterol, plus used brain SM, which has a different acyl chain composition than the egg SM used in this study. One or more of these variables may strongly influence interleaflet coupling.

Detection of domain formation by FRET

It is important to consider under what conditions FRET can and cannot detect ordered domain formation. When using FRET probes that partition differently into Lo and Ld domains to investigate lipid organization in a lipid vesicle, FRET levels similar to those in homogeneous vesicles most often indicate that domains are absent. An alternative is that the domains are too transient to detect by FRET. If domains form and break down too rapidly for FRET probes to equilibrate their location and segregate into different domains, there will be little or no effect of domain formation upon FRET. Of course, lipid domain formation itself should require times long enough for lipids to segregate from each other into Lo and Ld domains with different lipid compositions. If there is enough time for domains with distinct lipid compositions to form, there should be enough time for fluorescent probes to segregate in accordance with their Lo/Ld partition coefficients.

A different issue is that domains smaller in size than the FRET interaction distance, Ro (i.e., with a radius or width less than Ro), are difficult to detect by FRET. The reason is that FRET acceptors just outside a very small domain would be close enough to FRET donors inside the domain to quench donor fluorescence by FRET (20). Such ultra-nanodomains, too small to detect with FRET donor-acceptor pairs with large Ro values, have been detected in samples with brain SM, POPC, and equal to or greater than 33 mol % cholesterol (15,20). However, domains formed by egg SM, used in this report, are much larger than those formed by brain SM (7,15), and domain size increases when cholesterol concentrations are below 33 mol % (20), as in this report (25 mol %). Therefore, it is unlikely that domains too small to detect by FRET were present in our experiments. This is confirmed by agreement of the anisotropy experiments, which would not be sensitive to domain size, with the conclusions from FRET measurements.

It should also be noted that the FRET measurements in asymmetric LUVs carried out here only measured outer leaflet lipid behavior and thus did not investigate whether outer leaflet ordered domains induced ordered domains in the inner leaflet. However, we believe it is likely that under the conditions in which the outer leaflet formed ordered domains, the inner leaflet also formed ordered domains because inner leaflet ordered domain formation under such conditions has been observed in giant asymmetric egg SM:DOPC_out/DOPC_in/cholesterol vesicles, in which domains are large enough to detect by microscopy (7).

Potential biological implications of loss of asymmetry

It is possible that (at least in some cases) a loss of lipid asymmetry is the trigger for ordered domain formation in the plasma membrane of living mammalian cells, i.e., that there is little or no segregation into well-defined Lo and Ld domains under ordinary conditions in which the plasma membrane is asymmetric. A transient loss of lipid asymmetry can occur upon Ca²⁺ influx into the cytoplasm, a process which can be triggered by biological events associated with signal transduction and ordered membrane domain formation (24, 25, 26). This loss of asymmetry is likely due to the Ca²⁺-dependent scramblase activity of the TMEM16F protein, which appears to be able to scramble many different types of lipids (27). In addition, it should be noted that at least a partial loss of lipid asymmetry has been observed in giant plasma membrane vesicles, in which membrane domains can be more readily detected than in intact cells (28). Thus, it is possible that loss of asymmetry is one reason domains are more easily detected in giant plasma membrane vesicles than in intact cells.

Author Contributions

E.L. conceptualized the project and was responsible for funding acquisition and project administration. J.W.S.C., and S.K. carried out the experimental studies. J.W.S.C., S.K., and E.L. designed the experiments, carried out data analysis, and wrote and revised the manuscript.

Editor: Sarah Veatch.