Impact of Acyl Chain Mismatch on the Formation and Properties of Sphingomyelin-Cholesterol Domains

Abstract

Lateral segregation and the formation of lateral domains are well-known phenomena in ternary lipid bilayers composed of an unsaturated (low gel-to-liquid phase transition temperature (T_m)) phospholipid, a saturated (high-T_m) phospholipid, and cholesterol. The formation of lateral domains has been shown to be influenced by differences in phospholipid acyl chain unsaturation and length. Recently, we also showed that differential interactions of cholesterol with low- and high-T_m phospholipids in the bilayer can facilitate phospholipid segregation. Now, we have investigated phospholipid-cholesterol interactions and their role in lateral segregation in ternary bilayers composed of different unsaturated phosphatidylcholines (PCs) with varying acyl chain lengths, N-palmitoyl-D-erythro-sphingomyelin (PSM), and cholesterol. Using deuterium NMR spectroscopy, we determined how PSM was influenced by the acyl chain composition in surrounding PC environments and correlated this with the affinity of cholestatrienol (a fluorescent cholesterol analog) for PSM in the different PC environments. Results from a combination of time-resolved fluorescence measurements of trans-parinaric acid and Förster resonance energy transfer experiments showed that the relative affinity of cholesterol for phospholipids determined the degree to which the sterol promoted domain formation. From Förster resonance energy transfer, deuterium NMR, and differential scanning calorimetry results, it was clear that cholesterol also influenced both the thermostability of the domains and the degree of order in and outside the PSM-rich domains. The results of this study have shown that the affinity of cholesterol for both low-T_m and high-T_m phospholipids and the effects of low- and high-T_m phospholipids on each other influence both lateral structure and domain properties in complex bilayers. We envision that similar effects also contribute to lateral heterogeneity in even more complex biological membranes.

Significance

Mammalian cell membranes are composed of a broad variety of structurally different lipids. The current view is that the role of these lipids is to create a versatile solvent to accommodate different membrane proteins and allow them to function properly. There are strong indications that several essential cellular processes are controlled by lipid-protein interactions. For instance, lipids can affect protein function by self-organization into lateral domains with different properties. We now report on how different structural features of phospholipids influence such lateral structuring of membranes through their different interactions with cholesterol. Detailed knowledge of these interactions will help us to predict how the lipids organize themselves in cell membranes.

Introduction

Mammalian membranes contain a large number of different lipids (1). We know from model membrane studies that lipids have an intrinsic tendency to segregate into lateral domains (or phases) based on their structure (1, 2). However, we still do not understand how and to what degree lateral domains exist in cell membranes. Further, we do not know all the forces that drive the formation of such. Using model membranes composed of three lipid components—an unsaturated low-T_m phospholipid, a saturated high-T_m phospholipid, and cholesterol—that very roughly mimic the composition of mammalian cell membranes has been a useful approach to examine these questions (2, 3).

Cholesterol is an essential lipid in mammalian cells, and because of its molecular features, cholesterol modulates acyl chain order in phospholipid bilayers. When present in phospholipid bilayers, cholesterol, together with the surrounding lipids, can form a liquid-ordered (L_o) phase in which the lipid acyl chains have a high degree of order, but the lateral diffusion rate remains relatively fast (4, 5). Cholesterol prefers to interact with saturated phospholipids over unsaturated phospholipids. Among the saturated phospholipids, cholesterol further seems to have an especially strong affinity for sphingomyelin (6, 7, 8). In bilayers composed of different phospholipids, the presence of cholesterol may influence the lateral organization in the membrane by interacting preferentially with saturated lipids such as sphingomyelin (9, 10, 11). This may lead to the formation of L_o domains. Such domains are believed to play important roles in many cellular processes such as endocytosis, protein and lipid sorting, and cell signal transduction (12, 13, 14).

It is well-known that phospholipids with different acyl chain length may segregate into lateral domains or phases in binary bilayers (15, 16). Similarly, in ternary bilayers, the degree of acyl chain length mismatch has also been shown to determine how fluid-fluid phase separation occurs (17). It has been proposed that changes in line tension at domain boundaries, resulting from increased acyl chain length mismatch between the two phospholipids, facilitate domain formation and lateral segregation (17, 18, 19). Other studies have further examined how line tension may influence lateral domain formation in a lipid composition-dependent manner (20, 21). In all likelihood, the lateral segregation process is modulated by more factors than line tension alone. One such factor may be the differential interactions between cholesterol and phospholipids. Because cholesterol has a higher affinity for PCs with longer acyl chains (22), it is plausible that shorter acyl chains in the low-T_m phospholipids facilitate lateral segregation as cholesterol’s relative affinity for high-T_m lipid increases.

In this project, our aim was to study how cholesterol interacts with unsaturated phosphatidylcholine (PC) and N-palmitoyl-D-erythro-sphingomyelin (PSM) in mixed bilayers and to show how such interactions could influence the formation of lateral domains. As unsaturated PC components of the model membranes, four lipids were chosen: di-14:1-PC, di-18:1-PC, di-20:1-PC, and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC). These unsaturated PCs were then mixed with PSM and cholesterol in different proportions, allowing insight into both the role of acyl chain mismatch and the degree of unsaturation in lateral segregation. The effect of cholesterol on lateral segregation was studied by comparing binary phospholipid bilayers with 0 and 20 mol% cholesterol. We used deuterium NMR spectroscopy (with either PSM-_d31 or PSM-_d2(9′9′)) to determine how the acyl chains of PSM were influenced by the surrounding PC environment, and we then compared these results with the measured sterol affinities for PSM in the different PC environments. From time-resolved fluorescence experiments with trans-parinaric acid (tPA) and Förster resonance energy transfer (FRET) experiments with different fluorophores, we examined how PC chain length and unsaturation influenced the formation of lateral domains enriched in PSM and cholesterol. To obtain information about the properties of the ordered domains that were formed in the different PC environments, the abovementioned techniques were used and were supplemented with differential scanning calorimetry. The combination of all these experimental approaches gave us information about the thermostability of the domains, as well as the degree of order in and outside the ordered domains. Our results showed that the unsaturated PCs present in the bilayers influenced cholesterol-PSM interactions and that both the formation and properties of ordered domains were clearly influenced by cholesterol.

Materials and Methods

Material

We obtained POPC, 1,2-oleoyl-sn-glycero-3-phosphocholine (di-18:1-PC), 1,2-dimyristelaidoyl-sn-glycero-3-phosphocholine (di-14:1-PC), 1,2-dieicosenoyl-sn-glycero-3-phosphocholine (di-20:1-PC), rhodamine-1,2-dioleoyl-sn-3-glycero-3-phosphoethanolamine (rhodamine-DOPE), 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (NBD-DPPE), and egg sphingomyelin (SM) from Avanti Polar Lipids (Alabaster, AL). PSM was purified from egg SM by reverse-phase high-performance liquid-chromatography (Supelco Discovery C18-column, Sigma-Aldrich, St. Louis, MO; dimensions 250 × 21.2 mm, 5 μm particle size) using methanol as an eluent. The purity and identity of the PSM was verified by electro-spray ionization mass spectrometry. PSM-d₃₁ and PSM-d_2(9′,9′) were prepared from sphingosylphosphorylcholine (Avanti Polar Lipids) and palmitic-d₃₁ acid or palmitic-d_2(9′,9′) acid (Cambridge Isotope Laboratories, Tewksbury, MA), respectively, as described previously (23). Cholesterol and mβCD were purchased from Sigma-Aldrich. Cholesta-5,7,9(11)-triene-3-β-ol (CTL) was prepared according to published procedures (24, 25). The tPA was synthesized as described in (26) and purified by crystallization from hexane (27). It was stored at −87°C and contained 1 mol% butylated hydroxytoluene to prevent oxidation. Methyl-β-cyclodextrin (mβCD) was obtained from Sigma-Aldrich. Stock solutions of diphenyl hexatriene (DPH; Molecular Probes, Eugene, OR), tPA, rhodamine-DOPE (Avanti Polar Lipids), and CTL were prepared in ethanol. Fluorophore concentrations were determined based on their respective molar extinction coefficients: CTL (11,250 cm⁻¹ M⁻¹), DPH (92,000 cm⁻¹ M⁻¹), rhodamine-DOPE (90,000 cm⁻¹ M⁻¹), NBD-DPPE (21,000 cm⁻¹ M⁻¹), and tPA (88,000 cm⁻¹ M⁻¹). The concentration of all phospholipid stock solutions was determined according to (28). The cholesterol concentration was determined using a surface barostat (29). All solutions were stored in the dark at −20°C and warmed to ambient temperature before use. The water used in all experiments was purified by reverse osmosis, followed by passage through a Millipore UF Plus water-purification system (Millipore, Billerica, MA), to yield a product resistivity of 18.2 mΩcm.

Deuterium NMR

The preparation of samples containing PSM-d₃₁ started with the mixing of unsaturated PC (di-14:1-PC, di-18:1-PC, di-20:1-PC, or POPC) and PSM-d₃₁ at a molar ratio of 80:20 in methanol. The solvent was evaporated under a constant flow of nitrogen at 40°C. The dry lipid films were re-dissolved in chloroform, and the solvent was removed at 40°C, after which the samples were kept in a vacuum for 20 h. The dry lipid films were hydrated in deuterium-depleted water at 65°C and vortexed, after which the suspensions were freeze-thawed 10 times, resulting in multilamellar vesicles (MLVs). The MLVs were lyophilized and then rehydrated in deuterium-depleted water until 50% hydration (w/w) and freeze-thawed 10 times. The samples were then transferred to 5-mm glass tubes (Wilmad, Vineland, NJ), and the tubes were sealed with epoxy glue. The samples containing PSM-d_2(9′,9′) were prepared similarly, but the sample composition was as follows: unsaturated PC (di-14:1-PC, di-18:1-PC, di-20:1-PC, or POPC)/PSM-d_2(9′,9′)/cholesterol (40:40:20 by mol). Deuterium NMR spectra were recorded on a 300-MHz CMX spectrometer (Chemagnetics, Agilent, Palo Alto, CA) with a 5-mm ²H static probe (Otsuka Electronics, Osaka, Japan). A quadrupole echo sequence was used, with a 90° pulse width of 2.5 μs, interpulse delay of 30 ms, and repetition rate of 0.6 s. The sweep width was 250 kHz, and the number of scans was ∼50,000. The data were analyzed with TopSpin 3.1 (Bruker, Berlin, Germany) and NMR DePaker 1.0rc1 (https://launchpad.net/nmrdepaker).

Equilibrium partitioning of CTL between mβCD and LUVs

To determine the equilibrium partitioning of CTL between mβCD and large unilamellar vesicles (LUVs), we followed the protocol published previously (30). In brief, LUVs were made by extruding MLVs composed of different unsaturated PCs, with and without PSM, through 200 nm membranes (Whatman International, Maidstone, UK). These LUVs (final concentration 50 μM) were mixed with 0–1.0 mM mβCD in a total volume of 2.5 mL. The resulting samples (10 per PC) were incubated 2 h at 55°C after which the anisotropy of CTL was measured at the same temperature. The obtained anisotropy values were converted to the molar concentration of CTL in the membranes,$C_{CTL}^{LUV}$, according to

$$C_{CTL}^{LUV}=C_{CTL}\frac{\left(r_i-r_{CD}\right)}{\left(r_{LUV}-r_{CD}\right)},$$ (1)

where C_CTL is the total concentration of CTL in the samples, r_LUV is the anisotropy of CTL in the specific phospholipid (PL) bilayer, r_i is the CTL anisotropy in the sample, and r_CD is the anisotropy of CTL in the CTL-mβCD complex. The molar fraction partition coefficients (K_X), describing the equilibrium partitioning of CTL between the different PL bilayers and mβCD, was calculated by plotting the calculated molar concentrations of CTL in the LUV bilayers against the mβCD concentration and fitting the obtained curves with the following equation:

$$C_{CTL}^{LUV}=\frac{C_{L-}{C}_{CTL}+{\left(C_{CD}\right)}^n/K_X}{2}\times \left(\sqrt{1+4\frac{C_L{C}_{CTL}}{{\left[C_L-C_{CTL}+{\left(C_{CD}\right)}^n/K_X\right]}^2}}-1\right).$$ (2)

Here, C_L is the PL concentration, C_CD is the cyclodextrin concentration, and $C_{CTL}^{LUV}$ is the cholesterol concentration in lipid bilayers. The relative partitioning coefficient K_R was calculated by dividing the K_X obtained with different PC samples with the K_X obtained from PSM samples.

Time-resolved emission spectroscopy

The formation of ordered PSM-rich domains was detected from the average lifetime of tPA emission. The tPA excited state is very sensitive to the order of its local environment (27, 31, 32); hence, its emission lifetime varies in fluid and highly ordered membrane domains (33, 34). Fluorescence lifetimes of tPA (1 mol%) were measured in MLVs with a final lipid concentration of 0.1 mM. A FluoTime 100 spectrofluorometer with a TimeHarp260 pico time-correlated single-photon-counting module (PicoQuant, Berlin, Germany) was used for measurements. The tPA was excited with a 297 ± 10 nm LED laser source (PLS300; PicoQuant), and the emission was collected through a 435/40 nm single-band-pass filter. Fluorescence decays were recorded at 23°C (temperature controlled by water bath) with constant stirring during measurements. Data were analyzed using FluoFit Pro software obtained from PicoQuant. The decay was described by the sum of the exponentials, where α_i was the normalized pre-exponential and τ_i was the lifetime of decay of component i. The intensity-weighted average lifetime is given as follows:

$$\langle \tau \rangle =\frac{{\sum }_i{\alpha }_i{\tau }_i^2}{{\sum }_i{\alpha }_i{\tau }_i}.$$ (3)

FRET measurements

FRET was used both to detect the thermostability of ordered domains and as an alternative method to detect the formation of domains in ternary bilayers at constant temperature. For the thermostability experiments, the F₀ samples had the lipid composition of unsaturated PC/PSM/cholesterol (40:40:20) and contained 0.1 mol% DPH. The F samples had the same lipid composition but contained both 0.1 mol% DPH and 2 mol% rhodamine-DOPE. The total lipid concentration in both F and F₀ samples was 100 μM. The MLVs used in these measurements were prepared by mixing lipids and fluorophores in methanol and evaporating the solvent to form dry lipid films. These films were hydrated at 65°C in MilliQ water for 30 min, after which they were vortexed and bath-sonicated for 5 min at 65°C. Emission spectra were recorded on a PTI QuantaMaster-spectrofluorometer (Photon Technology International, Lawrenceville, NJ) at different temperatures. The excitation wavelength was set to 360 nm, and the emission was measured from 380 to 650 nm. The FRET efficiency was determined from the fluorescence intensity at 430 nm in F and F₀ samples. Samples used for the detection of domain formation at constant temperature had similar compositions as the samples for the time-resolved experiments with tPA but contained different fluorophores. For these experiments, the τ₀ samples contained 0.1 mol% DPH or 0.1 mol% NBD-DPPE, and the τ samples contained 0.5 mol% rhodamine-DOPE in addition to DPH or NDB-DPPE. The FRET efficiency was determined from time-resolved data recorded on the samples. For this, a FluoTime 200 spectrofluorometer with a PicoHarp300 time-correlated single-photon-counting module (PicoQuant) was used. The DPH was excited with a 378 ± 1 nm diode laser source and NBD-DPPE with a 460 ± 10 nm LED laser source (both from PicoQuant). The emission was measured at 430 nm (DPH) and 534 nm (NBD-DPPE) at 23°C. Data were analyzed using FluoFit Pro software obtained from PicoQuant, and the FRET efficiency was calculated from the average lifetimes of the fluorophores.

Differential scanning calorimetry

MLVs containing the indicated lipid compositions were prepared by hydration of dried lipid films in glass tubes. The lipids were hydrated for 0.5 h at 65°C in MilliQ water before being loaded into the sample cell of a MicroCal VP-DSC instrument (MicroCal, Northampton, MA). The final lipid concentration was 1.5 mM in all samples. The temperature ramp rate was 1°C/min. The data analysis was performed using Origin software (MicroCal).

Results

Effect of co-lipid structure on PSM properties observed with deuterium NMR

To gain an insight into how an SM molecule with saturated acyl chains is influenced by the structure of the surrounding unsaturated phospholipids, we measured deuterium NMR spectra of PSM-d₃₁ in various phospholipid environments (Fig. 1). Our major focus was on the acyl chain mismatch between PSM and the surrounding PCs. Hence, the interactions of PSM with di-14:1-PC, di-18:1-PC, and di-20:1-PC were investigated. For comparison, POPC was also included as a co-lipid. Deuterium NMR spectra of unsaturated PC/PSM (80:20) bilayers were recorded at 23, 30, and 40°C. From the spectra shown in Fig. 1, it is clear that, in all studied systems, the spectra from PSM-d₃₁ got narrower when the temperature and acyl chain dynamics were increased. The acyl chain composition in the PC molecules interacting with PSM also influenced the shape of the recorded spectra, especially in the bilayers containing di-14:1-PC. These spectra had an increased intensity in their central parts as compared to those observed with the other PCs. Similar observations have been made in binary mixtures of PCs with different acyl chains (35). However, all spectra looked like typical spectra of phospholipids with perdeuterated acyl chains (36, 37, 38).

Figure 1.

Representative deuterium NMR spectra of PSM-d₃₁ in different unsaturated PC bilayers. The spectra were recorded at different temperatures for samples having a PC/SM molar ratio of 80:20. The PCs that were included were (A) di-14:1-PC, (B) di-18:1-PC, (C) di-20:1-PC, and (D) POPC. The dashed lines mark the outer most peaks in the spectra at 23°C.

For further analysis, the recorded spectra were dePaked (Fig. S1), and order parameter profiles for PSM-d₃₁ in the different lipid systems at different temperatures were prepared (Figs. 2 and S2). The assignment of the Pake pairs to different positions along the carbon chains was done in accordance with published data on deuterated SM (6, 39, 40, 41), and the resulting order parameter profiles were in good agreement with the published data. At 23°C, the order parameter profiles of the acyl chain of PSM in the thickest bilayers (POPC and di-20:1-PC) resembled those of PSM in pure PSM bilayers at 50°, both having a plateau region with high order until about carbon 12, after which the order of the chain decreased markedly. In the di-18:1-PC bilayers, the order of the PSM acyl chain decreased closer to the membrane water interface, and even more so in di-14:1-PC bilayers. At higher temperatures, the order parameters of PSM in di-18:1-PC, di-20:1-PC, and POPC were similar to one another, but in di-14:1-PC, the order parameter profile was still markedly different from the rest. Clearly, the acyl chains of PSM were increasingly disordered when the surrounding PCs had shorter acyl chains.

Figure 2.

Order parameter profiles of PSM-d₃₁ in different PC environments. The order parameter profiles were calculated from dePaked deuterium spectra measured at (A) 23°C and (B) 40°C in bilayers with a PC/SM molar ratio of 80:20. To see this figure in color, go online.

Equilibrium partitioning of CTL between mβCD and LUVs

Next, we wanted to test how the unsaturated PC environment influenced the interactions between PSM and cholesterol. For this, we used an established experimental setup in which the equilibrium partitioning of the fluorescent cholesterol analog CTL between mβCD and LUVs was determined based on the anisotropy of the fluorophore (30). When the molar ratio partitioning coefficients (K_X) were determined at 23°C with different symmetric PC bilayers, it was found that K_X was lowest in di-14:1-PC, and when the acyl chain lengths increased, the affinity of the sterol for the bilayers also increased (Fig. 3). In POPC bilayers, K_X was markedly higher than in all the symmetric PC bilayers, as may be expected based on the inclusion of a saturated acyl chain in the sn-1 position. Inclusion of 20 mol% PSM in the LUVs led to higher K_X for all bilayer compositions (Fig. 3); however, the effect was clearly greatest with the POPC/PSM vesicles. In agreement with published results, the effect of adding PSM to POPC bilayers had an additive effect on K_X (10), that is, K_X increased linearly with PSM concentration (Fig. S3). This was not the case with the other systems. In fact, with the symmetric unsaturated PCs, the measured K_X deviated more from the linear dependence on PSM concentration as the length of the acyl chains decreased. We assume that this effect resulted from the influence the bulk lipids had on the PSM properties.

Figure 3.

Equilibrium partition of CTL between mβCD and different phospholipid vesicles. The equilibrium partition of CTL between mβCD and pure PC vesicles (open circles) or PC/PSM (80:20) vesicles (filled squares) was measured at 23°C. The values are averages of ≥3 experiments ±SD.

Lateral segregation in lipid bilayers with and without cholesterol

We have previously observed that cholesterol can facilitate lateral segregation of membrane phospholipids through its differential affinity for different phospholipid components in mixed bilayers (9, 10). The results from these studies showed that the number of double bonds and the headgroup type in the unsaturated phospholipids influenced the degree with which cholesterol promoted segregation. Now, we have tested how cholesterol affects the lateral segregation in bilayers when only the acyl chain length of the unsaturated phospholipids was varied. The formation of lateral ordered domains (gel or L_o) was determined from measured fluorescence lifetimes of tPA in bilayers with 0 or 20 mol% cholesterol and varying proportions of unsaturated PC and PSM. Without cholesterol, PSM formed gel domains (33, 42), and in the presence of 20 mol% cholesterol, PSM and cholesterol should form L_o domains (33, 43, 44, 45). To quantify the solubility limit of PSM in the liquid disordered (L_d) bilayers (or the PSM concentration at which ordered domains started to form), the obtained average lifetimes of tPA were plotted against the PSM concentration as shown in Fig. 4. In these plots, the average lifetime gradually increased with PSM content (in the di-20:1-PC bilayers) in a rather linear fashion up to a point at which the lifetimes started to increase more dramatically. The more steeply increasing part of the lifetime plot initially increased linearly. To determine the PSM solubility limit (at which ordered domains started to form), we made two linear fits: one to the initial slope and one to the initial part of the function with a steeper slope. We defined the PSM solubility limit as the intercept of these two linear functions (Fig. 4; (9, 10)). To verify that the results obtained with the single-probe method reported lateral segregation, we performed FRET experiments with the same lipid systems (Fig. S4). Because the FRET efficiency decreased dramatically above the same concentrations as the tPA reported the formation of PSM-enriched domains, these results supported the tPA results.

Figure 4.

Determination of ordered domain formation based on tPA fluorescence lifetime. Representative data from two experiments with di-20:1-PC bilayers (0 and 20 mol% cholesterol) are shown. To determine the PSM amount needed to form ordered domains, the data were fitted with straight lines: one to the slowly rising slope at low PSM content and one to the initial linear part of the steeply rising part. The crossing point of the two lines was defined as the PSM concentration at which ordered domains started to form.

Fig. 5 A shows the PSM solubility limits measured in different lipid compositions at 23°C. In the binary phospholipid bilayers, PSM-enriched gel domains started to form at rather similar PSM content levels (from ∼27 to 33 mol% PSM) in all studied systems. The lowest solubility of PSM in the L_d phase was observed in the di-14:1-PC bilayers and the highest in di-18:1-PC bilayers. Why PSM was more soluble in the di-18:1-PC L_d phase than in both di-14:1-PC and di-20:1-PC is unclear. However, it is possible that the position of the double bond has some impact because it has been observed to affect the molecular area of symmetric unsaturated PCs (46). With 20 mol% cholesterol in the bilayers, a larger difference between the different PCs was observed because ∼6 to ∼20 mol% PSM could be dissolved in the L_d phase made up of the different PCs before L_o domains formed. Di-14:1-PC bilayers could dissolve the smallest amount of PSM (∼6 mol%), and the solubility was higher in di-18:1-PC and di-20:1-PC bilayers. The L_d phase in POPC bilayers could dissolve the most PSM (∼20 mol%), likely because of the presence of the palmitoyl chain.

Figure 5.

The effect of cholesterol on the solubility limit of PSM in the L_d phase. (A) The solubility of PSM in the L_d phase was measured at 23°C in different PC bilayers containing 0 or 20 mol% cholesterol. Above these concentrations, PSM formed ordered domains. The values are averages of ≥3 experiments ±SD. (B) The change in PSM solubility due to cholesterol addition was plotted against the measured relative partitioning coefficients of CTL for the different phospholipid in the different systems.

The way cholesterol influenced lateral segregation in the bilayers was assessed by comparing the solubility limit of PSM in the L_d phase with and without cholesterol. In Fig. 5 B, the change in PSM solubility is shown as the percentage of change induced by inclusion of 20 mol% cholesterol. Clearly, the influence of cholesterol was greatest in the di-14:1-PC and di-18:1-PC bilayers, and the least influence was observed in POPC bilayers. When the change in PSM solubility was plotted against the K_R (describing the relative partitioning between PSM and the different PCs), a clear correlation between K_R and cholesterol’s promotion of lateral domain formation could be seen. In agreement with our previous data (9, 10), the result suggests that the more strongly cholesterol favored interaction with PSM over the fluid PC, the greater the influence of cholesterol on the lateral segregation.

Lateral domain properties as reported by deuterium NMR

The domain properties in ternary bilayers (unsaturated PC/PSM/cholesterol, 40:40:20) were investigated with NMR. Using PSM-d_2(9′,9′), SM labeled in a single carbon position in the acyl chain, we could obtain ²H-NMR spectra that showed Pake doublets both from the disordered and the ordered domains (Figs. 6 A and S5). As we performed the measurements at a series of temperatures, the melting of the L_o domains could also be observed. At 23°C, the Pake doublets arising from PSM in PSM-d_2(9′,9′) ordered domains (larger quadrupolar splittings), as well as the Pake doublets arising from PSM-d_2(9′,9′) in the disordered domains (narrower quadrupolar splittings), could be seen in all three lipid mixtures (Fig. 6). As the temperature increased, the outer Pake doublet (from ordered domains) decreased in size, indicating that the fraction of ordered domains decreased. The quadrupolar splittings from PSM-d_2(9′,9′) in the ordered and disordered domains are shown in Fig. 6. In the ordered phase, the quadrupolar splittings were very similar in all three systems, and at 23°C, they were similar to the quadrupolar splittings from PSM/cholesterol (1:1) at the same temperature. This suggests that the lipid composition in the ordered domains was similar in all three systems and that these domains contained only small amounts of unsaturated PC, irrespective of acyl chain length. The order in the disordered domains (Fig. 6 B) deviated more in the three systems, as can be seen from the quadrupolar splittings. In these disordered domains, the acyl chain length of the unsaturated PCs clearly affected the order in the bilayers (lowest order in di-14:1-PC and highest in di-20:1-PC). The quadrupolar splittings in the disordered domains also increased at higher temperatures. This increase in order occurred at the same temperature that the ordered domains remixed with the rest of the lipids. Hence, the increased order was likely due to an increased amount of PSM and cholesterol because of the loss of lateral segregation. Similar observations have been reported by Bartels and co-workers (39).

Figure 6.

Quadrupolar splittings from PSM-d_2(9′,9′) in and outside the ordered domains. The samples were composed of unsaturated PC/PSM-d_2(9′,9′)/cholesterol (40:40:20). (A) The quadrupolar splitting from the deuterated PSM in the ordered domains is shown. (B) The splittings from the disordered domains are shown. For comparison, the quadrupolar splittings from a binary PSM-d_2(9′,9′)/cholesterol bilayer have been included in (A). The insert in (A) shows a representative spectrum from these samples. To see this figure in color, go online.

From the ²H spectra in Fig. S4, it is also clear that the thermostability of the SM-enriched domains were affected by the acyl chain length of the unsaturated PC. The thermostability of the ordered domains was determined by fitting the spectra with two Pake doublets. At 23°C, most of the deuterium labeled PSM (70–80%) was in the ordered domains in all three studied systems (Fig. 7). As the temperature was increased, the amount of PSM in the ordered domains decreased until a critical temperature at which only disordered domains remained. In the di-20:1-PC bilayers, the ordered domains disappeared at 45°C, with those in di-18:1-PC disappearing at 55°C and with those in di-14:1-PC disappearing at 65°C. Because the acyl chain order in the SM-enriched domains was similar in all three bilayers, the difference in thermostability may be explained by an increased line tension with shorter acyl chains in the PC molecules.

Figure 7.

Thermostability of the ordered PSM-enriched domains as reported by deuterium NMR. The fraction of PSM in the ordered domains was determined by fitting the data shown in Fig. 7. The samples were composed of unsaturated PC/PSM-d_2(9′,9′)/cholesterol (40:40:20). To see this figure in color, go online.

Thermostability of SM-rich domains determined with FRET

The thermostability of the ordered domains in ternary bilayers (unsaturated PC/PSM/cholesterol, 40:40:20) was also determined with FRET. For this, F samples, containing 0.1 mol% DPH and 2 mol% rhodamine-DOPE, and F₀ samples, containing 0.1 mol% DPH, were prepared. The emission spectra of DPH were recorded at different temperatures for both F and F₀ samples, and the FRET efficiency as a function of temperature was calculated (Fig. 8). In all samples with symmetric unsaturated PCs, the FRET efficiency was ∼0.3 at 10°C, whereas in the POPC-containing bilayers, a FRET efficiency of ∼0.7 was determined at the same temperature. Assuming that the probe partitioning between ordered and disordered domains was not markedly different in bilayers with different lipid compositions, this suggests that the ordered domains in di-14:1-PC, di-18:1-PC, and di-20:1-PC bilayers were markedly larger than the domains formed in POPC bilayers. Based on published data, this could also be expected (47, 48). When the temperature was increased, the FRET efficiency was relatively unchanged at first. At specific temperatures, the SM-enriched domains melted, resulting in a marked increase in the FRET efficiency. Finally, when the SM-enriched domains had remixed completely with the disordered lipids, the FRET efficiency was not affected any further by increased temperature.

Figure 8.

Thermostability of the ordered PSM-enriched domains as reported by FRET. Samples composed of unsaturated PC/PSM/cholesterol (40:40:20) were prepared. The F₀ samples contained 0.5 mol% DPH, and the F samples contained 0.1 mol% DPH and 2 mol% rhodamine-DOPE. The fluorescence intensity of the samples was measured with 5° temperature intervals. The data shown are representative data sets.

From the results, it was clear that the acyl chain length mismatch between PSM and the different PCs affected the thermotropic behavior of the ordered domains. The L_o domains were the least thermostable in the POPC-containing bilayers. In bilayers containing symmetrical unsaturated PCs, the thermostability of the L_o domains increased with decreasing acyl chain length.

Thermostability of binary phospholipid bilayers

To be able to evaluate the role of cholesterol-phospholipid interactions as a determinant of the thermostability of the L_o domains, we had to know the thermotropic behavior of the cholesterol-free, binary phospholipid bilayers. Therefore, differential scanning calorimetry was used to study 1:1 mixtures of PSM and different PCs. The thermograms from these measurements are shown in Fig. S5, and as is evident, the cholesterol-free bilayers showed rather similar phase behavior. With all lipid compositions, the melting endotherms were broad and rather complex (more than one melting component). Interestingly, all lipid compositions had very similar end-melting temperatures (∼32°C), whereas the remixing temperatures of the cholesterol-containing samples were clearly different (Figs. 7 and 8). However, the midmelting temperature of the main transition showed similar composition dependency as observed in cholesterol-containing samples. That is, di-14:1-PC-containing bilayers had the highest T_m (∼26.5°C), di-18:1-PC had a slightly lower T_m (∼24.7°C), and di-20:1-PC and POPC had the lowest T_m-values (∼23.9 and ∼23.3°C, respectively).

Although the T_m-values in the cholesterol-free bilayers had a similar dependence on the lipid structure as in the cholesterol-containing bilayers, the addition of cholesterol enhanced the difference between the different samples. This was most likely the result of the presence of cholesterol, which enhanced the segregation of the PC and SM molecules.

Discussion

SM-cholesterol interactions in different lipid environments

Commonly, the characterization of membrane lipids and their interactions has been performed in simple lipid systems based on a single lipid or binary lipid mixtures (15, 49, 50, 51, 52). The advantage of this approach is, of course, that the experimental results are easily connected to the molecular features in the lipids of interest. However, it is also clear that membrane properties of lipids can be modulated by interactions with other membrane lipids. For example, molecular dynamics simulations suggest that the headgroup conformations of SM and ganglioside M₁ are altered through interactions with cholesterol (53, 54). In deuterium NMR experiments with binary phospholipid mixtures, it has been observed that deuterated acyl chains adapt to the lipid environment (35, 55); for example, the acyl chains may be more or less extended because of acyl chain composition of the other lipids in the bilayer. Further, it may be expected that the interactions—for example, between SM and cholesterol—can be affected by the interaction between SM and other phospholipids present in the bilayer. However, presently, we have limited insight into this.

By measuring deuterium NMR spectra of PSM with perdeuterated acyl chains in fluid bilayers composed of PSM and different unsaturated PCs, we obtained information about how the conformation of the PSM was altered through its interactions with the neighboring PC molecules (Fig. 1). An initial visual analysis of the deuterium spectra indicated that the PSM acyl chain order was altered through the interactions with the different PCs; however, to be able to make comparisons in more detail, order parameter profiles were prepared from the measured spectra (Figs. 2 and S1). At 23°C, the order parameter profiles of PSM in POPC and di-20:1-PC were rather similar both to each other and to that of PSM in a pure fluid PSM bilayer (at 50°C). This suggests that at this temperature, the PSM acyl chain mobility was similar in these three systems. In di-18:1-PC bilayers and especially in di-14:1-PC bilayers, on the other hand, the order parameter profiles deviated more from that of the pure PSM bilayer, indicating marked changes in the PSM acyl chain mobility. Increased temperature led to order parameter profiles that deviated more from that of pure PSM and became more similar to one another. At 40°C, all order parameter profiles (from binary systems) were more or less similar, except that of the di-14:1-PC, which still clearly reported lower chain order than the rest. In summary, the NMR data indicated that the hydrocarbon chains of PSM adapted to the acyl chain order and acyl chain length of the neighboring PC molecules, much as was expected based on previous observations (35, 55).

At the used PC/SM ratio, all studied systems were totally fluid at 23°C. In the CTL equilibrium partition experiments, we included 2 mol% CTL in the membranes. Because these samples did not form ordered domains, we assumed that the sterol had not markedly altered the PSM acyl chain order in the bilayers. Hence, we further assumed that the affinity of cholesterol for PSM may have reflected the conformation of the SM molecules in the different PC environments. In the CTL partitioning experiments, we observed that the affinity of the sterol for the bilayers increased with both acyl chain length and saturation (Fig. 2), in perfect agreement with published data (8, 10, 22, 30, 56, 57, 58). In all studied membranes, the addition of PSM led to a significantly increased affinity of CTL for the bilayers. Clearly, the effect of PSM inclusion was greatest in the POPC bilayers. For bilayers composed of POPC and PSM, we previously reported that the sterol affinity had a linear dependence on the PSM (10), which also seemed to be the case in this work (Fig. S2), although the pure PSM membrane was determined at a different temperature (50°C). However, with the other PCs, this was not the case. Rather, as the dependence on PSM concentration became less linear, the shorter the acyl chains in the PC were (Fig. S2). What was the reason for this difference in partition behavior? We propose that the difference is, in part, due to the different conformations of PSM in the different PC environments. In the POPC, the perdeuterated chains of PSM were found to have similar order parameter profiles as in pure fluid PSM, which allowed similar sterol-SM interactions in these systems. Similar interactions resulted in a more or less linear dependence on PSM content. The more the acyl chain order profile of PSM deviated from that in the pure fluid PSM, the more nonlinear the dependence on PSM concentration became.

The affinity of cholesterol for PSM-containing bilayers was likely in part dependent on the degree of order in the PSM hydrocarbon chains (as seen with ²H NMR), but it is possible that also other parameters were involved. For example, the presence of saturated acyl chains in the PC could be expected to have some influence. This may be the reason why cholesterol bound more poorly to di-20:1-PC than to POPC, although the effective acyl chain length in the bilayers should be about the same.

Lateral segregation

Segregation of lipid into lateral domains is a process that seems to be driven by many different forces (10, 17, 18, 59). One factor that has been observed to influence the formation of ordered domains in the membrane plane is the interactions between cholesterol and the different phospholipids present (9, 10). An increased relative affinity of cholesterol for the saturated phospholipids, as compared to the unsaturated phospholipids, has been observed to decrease the solubility of PSM in the L_d phase and thereby promote the formation of ordered domains. We have previously observed that both the degree of acyl chain unsaturation and the lipid headgroup structure affected the affinity of cholesterol for the phospholipids and the susceptibility of the lipids to form a lateral domain. Now, we investigated how cholesterol facilitated lateral segregation when the acyl chain length mismatch between unsaturated and saturated phospholipids was varied. The results showed that, without cholesterol, the gel domain formation started with ∼30 mol% PSM in all PC bilayers, but when 20 mol% cholesterol was added, the trend was that the longer the acyl chains in the PC, the higher the solubility of PSM in the PC L_d phase (Fig. 4). This indicated that the system most prone to form L_o domains, at least with 20 mol% cholesterol, was the di-14:1-PC/PSM/cholesterol bilayers, closely followed by the di-18:1-PC/PSM/cholesterol bilayers. When the change in PSM solubility in the L_d phase was correlated with the relative partitioning coefficient for CTL between PSM and the different PCs (Fig. 4 B), a clear correlation between the two parameters was observed, in line with previous reports (9, 10).

Another commonly used parameter to describe the lateral segregation propensity of mixed lipid bilayers is the thermostability of the formed ordered domains (17, 20, 48, 60). We used both deuterium NMR and FRET to study the influence of the acyl chain length in the PC molecules on the thermostability of the PSM-enriched ordered domains (Figs. 7 and 8). In agreement with observations made by García-Sáez and co-workers (17), both methods showed that a larger mismatch in acyl chain length between PSM and the PCs resulted in more thermostable ordered domains. When the thermostability of L_o domains is compared to that of gel domains (see Fig. S5), the importance of the cholesterol-PL interactions as determinants of thermal stability becomes clear.

The fact that cholesterol can affect both the thermostability of ordered domains and the amount of PSM needed to start forming these domains is linked to how the sterol interacts with saturated and unsaturated acyl chains (61). Based on molecular dynamics simulations, Bennett and co-workers suggested that the enthalpic gain from the interactions between cholesterol and saturated acyl chains drives domain formation (59). Similar conclusions have also been made based on nearest-neighbor recognition methodology and Monte Carlo simulations (62), from which the authors concluded that the unfavorable interaction enthalpies between cholesterol and the unsaturated acyl chains pushes cholesterol out of the L_d and into the L_o domains. Our results agree with the importance of this push effect of unsaturated chains. However, based on previous work, it seems that an attractive force that pulls cholesterol into L_o domains is also needed (10).

Using site-specific deuterated PSM, we could measure the acyl chain order both inside and outside the PSM-enriched domains (Fig. 6). The results suggest that the acyl chain order within the ordered domains was similarly independent of the PC acyl chain length. At 23°C, the chain order was also similar to that of binary PSM-cholesterol bilayers. Our interpretation of this is that the ordered domains almost exclusively contained PSM and cholesterol (at least at lower temperatures). The acyl chain of PSM molecules in the L_d phase was, however, clearly affected by the acyl chain composition in the PCs (Fig. 6). The highest order in the L_d phase was composed of di-20:1-PC, and the lowest in the L_d phase was composed of di-14:1-PC. Interestingly, the acyl chain order in the L_d phase increased at higher temperatures. Because this occurred at the same temperatures that the ordered domains no longer were detected, we concluded that the increased order was the result of an enhanced PSM and cholesterol concentration in the L_d phase due to the remixing of the ordered domains lipids with the rest of the lipids, in agreement with published data (39).

The fact that the acyl chain order was similar inside the ordered domains and different in the L_d phase suggests that the line tension at the domain boundaries increased as the PC acyl chain length was decreased. This, of course, could be expected based on the lipid composition. Line tension, due to height mismatch between disordered and ordered domains, has been shown to modulate both the size and the thermal stability of ordered domains (17, 18, 19, 63). However, the generalization of the role of line tension in controlling lateral membrane structure remains debated (20, 21). In this work, we observed that the thermostability of the L_o domains increased with assumed line tension (Figs. 7 and 8). However, the FRET efficiency did not indicate any significant size difference among the three lipid systems based on symmetric unsaturated PCs, although possible differences in domain size may not have been detected as a result of the domains being much larger than the Förster distance of the probe pair (36 Å according to (64)). On the other hand, the domains in the POPC-based system were markedly smaller according to the measured FRET efficiencies (Fig. 8). This observation is in line with the proposition that hybrid lipids (such as POPC), may act as linactants, thereby reducing the line tension and the domain size (65). However, the presence of the palmitoyl chain in POPC may also have led to increased partitioning of POPC into the L_o domains. The fact that the L_o domains were dramatically more thermostable than the gel domains (without cholesterol) could indicate that the presence of cholesterol increased the line tension. Because it has been proposed that nucleation of domains may depend on line tension (66, 67, 68), it is possible that the solubility of PSM in the L_d phase was also, in part, driven by increased line tension arising from cholesterol-SM interactions. However, it has been reported that increased cholesterol content can lower the line tension in ternary lipid systems (69, 70).

Conclusions

We studied how the acyl chain length of unsaturated PCs affects the conformation of PSM in mixed bilayers, and we observed marked effects. These effects correlated rather well with the affinity of CTL for the different lipid systems. Hence, our findings suggest that interactions between cholesterol and PSM were modulated by the influence of the surrounding lipid environment. In bilayers composed of unsaturated PCs and PSM, an addition of 20 mol% cholesterol promoted the formation of lateral domains enriched in PSM and cholesterol. The degree to which cholesterol promoted lateral segregation seemed to depend on cholesterol’s relative affinity for the two phospholipids in the bilayer. A more significant affinity difference lead to a stronger promotion of lateral segregation, in line with published data (9, 10). Similarly, the thermal stability of the domains also showed a similar dependence, although changes in line tension may have played a role in these processes. Overall, the results of this work further underline the importance of understanding the overall interactions between phospholipids and cholesterol in the pursuit of understanding the structure and function of mammalian membranes.

Author Contributions

T.K.M.N., O.E., M.M., and J.P.S. planned the research. O.E., V.H., K.-L.L., and H.T. performed the experiments. All authors analyzed the data. T.K.M.N. wrote the article with contributions from the other authors.

Editor: Georg Pabst.