Lamellar Phases Composed of Phospholipid, Cholesterol, and Ceramide, as Studied by ²H NMR

Abstract

Sphingolipids constitute a significant fraction of cellular plasma membrane lipid content. Among sphingolipids, ceramide levels are usually very low. However, in some cell processes like apoptosis, cell membrane ceramide levels increase markedly because of the activation of enzymes like acid sphingomyelinase. This increase can change the physical state of the membrane by promoting molecular order and inducing solid-ordered (S_o) phase domains. This effect has been observed in a previous ²H NMR study on membranes consisting of palmitoyl sphingomyelin (PSM) and palmitoyl ceramide (PCer). Cholesterol (Chol), too, is present at high concentrations in mammalian plasma membranes and has a favorable interaction with sphingomyelin (SM), together forming domains in the liquid-ordered phase in model membranes. There are reports that Chol is able to displace ceramide (Cer) in SM bilayers and abolish the S_o phase domains formed by SM:Cer. This ability of Chol appears to be concentration dependent; in membranes with low Chol and high Cer contents, S_o phase domains rich in Cer coexist with the continuous fluid phase of the membrane. Here, we studied the effect of increasing PCer concentration in PSM:Chol bilayers, using ²H NMR. Chol:PCer mole ratios were 3:1, 3:2, and 3:3, at a fixed 7:3 phospholipid:cholesterol mol ratio. Both PSM and PCer were monitored in separate samples for changes in their physical state by introducing a perdeuterated palmitoyl chain in either molecule. Moreover, the effect of replacing PSM with DPPC was investigated to test the impact on membrane phase behavior of replacing the sphingosine with a palmitoylated glycerol backbone. We found that PCer can increase acyl chain order in both PSM:Chol and DPPC:Chol bilayers. Especially in bilayers with Chol:PCer 1:1 molar ratios, PCer induces highly stable S_o phase domains in both PSM and DPPC bilayers near 37°C. However, PCer has a more pronounced ordering effect on PSM compared to DPPC bilayers.

Significance

Using deuterium NMR, deuterated ceramide has for the first time, to our knowledge, been directly shown to induce the formation of ceramide-enriched “solid-ordered” domains in membranes composed of sphingomyelin and cholesterol. This adds weight to the hypothesis that such changes in cell membrane physical properties may occur because of sphingomyelinase action during apoptosis.

Introduction

Sphingolipids are a primary component of the outer leaflet of the plasma membrane (1). Ceramides (Cer) are sphingolipids that typically occur at very low membrane concentrations (less than 1 mol %), but under some cell processes like apoptosis, their levels can increase markedly (2). For instance, in apoptosis, acid sphingomyelinase cleaves the phosphocholine headgroup of sphingomyelins (SMs), converting them to ceramides (3). Because of their highly hydrophobic properties, even small concentrations of ceramide can induce the formation of solid-ordered (S_o) phase domains in the membrane (4, 5, 6). This ability has been observed in detail in model membranes containing palmitoyl ceramide (PCer) and palmitoyl SM (PSM) using a combination of ²H NMR and differential scanning calorimetry (DSC) (7). The authors found that even small amounts of PCer (2–3 mol %) were enough to induce S_o phase domains in PSM bilayers and that those domains were more stable than the S_o phase domains formed by pure PSM. The capability of ceramides to self-assemble into domains up to micrometer size has been observed in the so-called “ceramide-rich platforms.” These platforms reorganize the receptors and signaling molecules into clusters within the cell membrane to amplify transmembrane signaling processes (8). Increases in ceramide concentration can also lead to the formation of nonlamellar phases, like inverted hexagonal phases, and to enhanced membrane permeability (see a recent review article (9)). Among sphingolipids, both in SMs and Cer, slight changes in the structure can be crucial determinants of their biophysical properties (10, 11, 12).

Cholesterol (Chol) is the main sterol found in the plasma membrane of mammalian cells. Chol and sphingolipids tend to colocalize and form liquid-ordered (L_o) domains in lipid membranes (13, 14, 15). The Chol content of some cell membranes is relatively high, and in some cases, like the human erythrocyte, it can reach up to 48 mol % (16, 17). Therefore, it is important to study the effect of Chol on the physical behavior of lipids in the cell membrane. Notably, it is known from the partial phase diagram of Chol and PSM that even 10 mol % Chol is enough to induce S_o + L_o and liquid-disordered (L_d) + L_o domains at temperatures below and above the phase transition of PSM, respectively (15). In the same report, it was described that at Chol concentrations above ∼32 mol %, the phase of the PSM:Chol binary mixtures is L_o from 25 to 60°C. Sullan et al. (18) prepared ternary bilayers of DOPC:SM:Chol and, using atomic force microscopy (AFM), observed the lateral segregation of L_d and L_o domains. Very recently, Nyholm et al. (19) measured the partition of Chol between different phospholipid environments and concluded that the lower the affinity of Chol for the unsaturated phospholipids, the higher its capacity to stimulate lateral segregation in a bilayer composed of unsaturated phospholipid, SM, and Chol.

Because both Chol and ceramides are mainly hydrophobic molecules with relatively small polar headgroups, they need to be shielded against water by the polar headgroups of other lipids in the membrane (umbrella model) (20). Because SM has a large, strongly polar headgroup, in binary mixtures of SM:Chol and SM:Cer, both Chol and Cer show a high affinity toward SM and, at physiological temperatures, they form their own favored phases (L_o in the case of SM:Chol and S_o in the case of SM:Cer bilayers). Considering apoptosis, Cer is most likely generated in SM-rich areas, and because of highly favorable interactions between Chol and SM, the interactions among the three molecules are important in determining cell fate (21, 22).

There have been reports on the concentration-dependent displacement effect of both ceramide and Chol (with respect to each other) in phospholipid bilayers (23, 24, 25, 26, 27) that could be important in various events of cell physiology (9, 13). These data suggest that the stability of the Chol-rich L_o phase and ceramide-rich S_o phase domains in the membrane depends on the relative concentration of the two lipids (9). Hence, there could be a critical balance between Chol and ceramide concentrations that regulates signaling processes in the cell membrane.

Busto et al. (28) described, perhaps for the first time, lamellar gel phases of ternary lipid composition containing saturated phospholipids, PCer, and Chol. In particular, the PSM:Chol:PCer (7:3:3 mol ratio) mixture was characterized by a variety of physical techniques and was found to present intermediate properties between the SM:Chol L_o and the SM:Cer S_o phases. In this report we intend, using ²H NMR, to address the question of the physical state of model membranes containing a fixed 7:3 mol ratio of PSM:Chol and increasing amounts of PCer. Some of the lipid compositions studied here were previously examined by other biophysical techniques (27, 28, 29, 30); thus, by comparing the data obtained here with previous results, a conclusion could be reached on their phase behavior regardless of the probes used and the types of bilayers (multilamellar vesicles (MLVs), giant unilamellar vesicles (GUVs), etc.). In addition, the effect of replacing PSM (or PSM-d31) with DPPC (or DPPC-d31) in the membranes studied was considered. The latter study was aimed at explaining the predominance of saturated SM over saturated PC in the plasma membrane.

In our first set of experiments, we monitored the spectral features of the perdeuterated N-linked palmitoyl chain of PSM-d31 or PCer-d31 separately in samples having the same molar ratios. SM’s N-linked palmitoyl chain was monitored in MLVs composed of PSM-d31:Chol:PCer 7:3:1, 7:3:2, or 7:3:3 while ceramide’s N-linked palmitoyl chain was monitored in MLVs having compositions PSM:Chol:PCer-d31 7:3:2 or 7:3:3. In the second set of data, the same molar ratios were used, but this time, DPPC-d31 and DPPC replaced PSM-d31 and PSM, respectively. These compositions were chosen to represent local domains in biological membranes rich in sphingolipid and Chol.

Materials and Methods

Sample preparation

N-palmitoyl-D-erythro-sphingosylphosphorylcholine (PSM), N-palmitoyl-d31-D-erythro-sphingosylphosphorylcholine (PSM-d31), 1-palmitoyl-2-palmitoyl-sn-glycero-phosphocholine (DPPC), 1-palmitoyl-2-palmitoyl-d31-sn-glycero-phosphocholine (DPPC-d31), and cholesterol were obtained from Avanti Polar Lipids (Alabaster, AL). N-palmitoyl-D-erythro-sphingosine (PCer) and N-palmitoyl-d31-D-erythro-sphingosine (PCer-d31) were obtained from Northern Lipids (Vancouver, BC). Deuterium-depleted water was obtained from Sigma-Aldrich Canada (Oakville, ON).

For each MLV preparation, the appropriate amounts of each lipid were mixed in either benzene:methanol 8:2 or chloroform:methanol 2:1 (v/v) to obtain a homogeneous mixture of the lipids. In the case of benzene:methanol 8:2, the samples were frozen to the liquid nitrogen temperature and lyophilized for 15 h to sublimate the solvent. In the case of chloroform:methanol 2:1, the solvent was dried under a stream of nitrogen gas, and then the samples were lyophilized for 3 h to remove the residual solvent. The samples containing Cer were hydrated with deuterium-depleted water and subjected to five or more cycles of freeze-thaw-vortex, performed between liquid nitrogen temperature and 95°C. (For pure phospholipids, the “thaw” temperature was ∼60°C, whereas for binary phospholipid:Chol samples. it was ∼75°C). Two sets of MLVs, depending on the purpose of the study, were prepared. The first set contained the following lipids at the indicated molar ratios: PSM-d31, PSM-d31:Chol:PCer 7:3:0, 7:3:1, 7:3:2, and 7:3:3; PSM:Chol:PCer-d31 7:3:2 and 7:3:3. The second set of MLVs contained the following lipid mixtures and molar ratios: DPPC-d31, DPPC-d31:Chol:PCer (7:3:0, 7:3:1, 7:3:2, and 7:3:3), DPPC:Chol:PCer-d31 (7:3:2 and 7:3:3). In both sets, after taking the spectra of each MLV preparation, the sample was recovered from the NMR tube and lyophilized to remove deuterium-depleted water from the sample. In the next step, an appropriate, additional amount of lipids in organic solvent was added to prepare the next MLVs as explained above. The initial MLVs in the first set were PSM-d31 and PSM:Chol:PCer-d31 7:3:2 and in the second set were DPPC-d31 and DPPC:Chol:PCer-d31 7:3:2. The total mass of lipids used varied between 30 and 135 mg, and 700–800 μL deuterium-depleted water were used for hydration. Most of the MLV compositions were checked by ¹H NMR (see Supporting Materials and Methods for details).

²H NMR spectroscopy: methods

²H NMR experiments were performed using a 7.0 T Oxford Magnet (Oxford Magnet Technology, Witney, Oxon, UK) and a Tecmag Scout spectrometer (Tecmag, Houston, Texas) at 46.8 MHz using the quadrupolar echo technique (31). The quadrupolar echo involves a two-pulse sequence in which both were 90° pulses and were 90° out of phase. The duration of each pulse was 3.95 μs with 40 μs interpulse spacing. The dwell time was 2 μs and the time between acquisitions was 300 ms. Quadrature detection with 8-CYCLOPS phase cycling was used for data acquisition. The number of scans varied between 10,000 and 80,000 scans depending on the amount of deuterated lipid in each sample (the lower the amount of deuterated lipid, the higher number of scans). The ²H NMR experiments were performed immediately after MLV preparation. For PSM-d31 MLVs, the starting temperature was 22°C, and data were collected at 1°C intervals from 21 to 38°C. To monitor the phase transition of PSM-d31 from S_o to L_d phase, spectra were collected at 0.1°C intervals from 38.0 to 42.0°C. The main transition temperature was found to be 40.0°C, in good agreement with previously published data (32). From 42 to 50°C, we continued to record data by increasing the temperature by 1°C intervals. For all other MLVs, data were collected at 1°C intervals from 21 to 45°C and at 5°C intervals from 45 to 70°C. The uncertainty in temperature measurements was 0.1°C. At least 20 min were allowed for equilibration between successive measurements. MLVs with higher PCer-d31 or PCer were checked for the solid phase (33) by increasing the repetition time between NMR scans to 3 s. No evidence of solid phase was observed for any of those MLVs.

²H NMR spectroscopy: moment analysis and order parameter profile

Moment analysis is very useful in exploring the phase transition of the deuterated lipids in MLVs. For the moment analysis of the ²H NMR spectra, the first moment, or average spectral width, M₁, was calculated using the following equation:

$$M_1=\frac{{\sum }_{\omega =-x}^x\left|\omega \right|f\left(\omega \right)}{{\sum }_{\omega =-x}^{x}f\left(\omega \right)},$$ (1)

where ω is the offset from the Larmor frequency, f(ω) is the signal intensity at frequency ω, and x was chosen so that the full spectrum resides in the frequency range of −x to x.

For Pake doublets, there is a linear relationship between the M₁ and the average carbon-deuteron (CD) bond order parameter (34). The order parameter, S_CD, is defined in the following equation:

$$S_{\text{CD}}=\langle 3{\cos }^2\theta -1\rangle /2,$$ (2)

where θ is the angle between the CD bond and the bilayer normal. The smoothed order parameter profile of the deuterons on the palmitoyl chain of the deuterated lipid in each sample was calculated from the dePaked spectrum (see Supporting Materials and Methods). S_CD can be derived from the dePaked spectrum using the following equation:

$$\mathit{\Delta }{\nu }_{\text{Q}}=\frac{3}{2}\frac{e^{2}qQ}{h}\left|S_{\text{CD}}\right|,$$ (3)

where $\frac{e^{2}qQ}{h}$ is the static quadrupolar coupling constant, 168 kHz, and Δv_Q is the quadrupolar splitting of the deuterons measured from the dePaked spectrum.

Results

²H NMR spectroscopy: line-shape analysis

The main advantage of using ²H NMR to investigate the phase of a specific phospholipid in MLVs is that the phase of the lipid can be identified from line shapes of the obtained spectra. In Fig. 1, typical spectra for the S_o, L_o, and L_d phases are shown. In the L_d spectrum, the gray arrows show the “shoulder,” and the plain black arrows show the “edge” regions of the L_d spectrum. The L_o spectrum is much wider than the L_d spectrum, with a broader shoulder. Also, by comparing the middle (the region between the two edges) of the L_d and L_o spectra, it is evident that more peaks are resolved in the L_d spectrum. For a membrane in the S_o phase, the spectrum features a sloping shoulder as shown in the lowest spectrum in Fig. 1 (bulleted arrows). When there is S_o + L_o phase coexistence, both the shoulder and the edge of the spectrum are more sloped compared to what is seen in a pure L_o spectrum. In addition, S_o + L_o phase coexistence may be ascertained from increased intensity in the methyl group shoulder region of the spectrum. Note that because the ²H NMR spectra of lipids in the ripple phase and gel phase are difficult to distinguish, the term S_o used here encompasses both of these ordered phases.

Figure 1.

Different ²H NMR line shapes for different lipid phases. Typical spectra for liquid disordered (L_d), liquid ordered (L_o), and solid ordered (S_o) phases are shown. The S_o and L_d spectra were obtained from pure PSM-d31 MLVs at 22 and 50°C. The L_o spectrum was obtained from the PSM-d31:Chol 7:3 MLVs at 35°C. For more information on the arrows, please refer to the text.

In Fig. 2, the spectra of all 14 MLVs at 22°C are shown. The spectra for pure phospholipids, PSM-d31 and DPPC-d31, show that these lipids are in the S_o phase at 22°C. By examining the spectral line shapes in the presence of 30 mol % cholesterol, it is apparent that both PSM-d31 and DPPC-d31 are predominantly in the L_o phase. However, the slightly sloping shoulders (between ∼30 and 60 kHz) and edges (between ∼26 and 30 kHz) of these spectra indicate that small fractions of the phospholipids are in the S_o phase. This S_o + L_o phase coexistence agrees with the phase diagrams of PSM-d31:Chol and DPPC-d31:Chol (32).

Figure 2.

²H NMR spectra of MLVs at 22°C. (A) From bottom to top are the spectra of pure PSM-d31 and PSM-d31:Chol:PCer MLVs with molar ratios of 7:3:0, 7:3:1, 7:3:2, and 7:3:3 (red spectra). The blue spectra of PSM:Chol:PCer-d31 MLVs with molar ratios of 7:3:2 and 7:3:3 are overlaid on the top two red spectra. The signal at 0 kHz for PSM:Chol:PCer-d31 7:3:2 and 7:3:3 spectra (blue spectra) arises from DOH in the sample. (B) From bottom to top are the spectra for pure DPPC-d31 and DPPC-d31:Chol:PCer MLVs with molar ratios of 7:3:0, 7:3:1, 7:3:2, and 7:3:3 (green spectra). The purple spectra of DPPC:Chol:PCer-d31 MLVs with molar ratios of 7:3:2 and 7:3:3 are overlaid on the top two green spectra. To see this figure in color, go online.

Low concentrations of PCer do not markedly affect the membranes’ spectra at 22°C. Qualitatively, the spectra of PSM-d31:Chol:PCer and DPPC-d31:Chol:PCer 7:3:1 MLVs are very similar to those from the binary mixtures of phospholipid:Chol 7:3. However, the addition of 20 or 30 mol % PCer to the 7:3:0 MLVs makes the shoulder and edges of the spectra significantly more sloped compared to the spectra of 7:3:0 and 7:3:1 MLVs. This implies that in the presence of 20 or 30 mol % added PCer, more PSM-d31 and DPPC-d31 are in the S_o phase than in the 7:3:0 and 7:3:1 MLVs.

Finally, the overlaid blue and purple spectra in Fig. 2 clearly show S_o + L_o phase coexistence. These spectra represent PCer-d31 in PSM:Chol:PCer-d31 and DPPC:Chol:PCer-d31 7:3:2 and 7:3:3 MLVs. Although the spectra of PSM-d31 and DPPC-d31 in 7:3:2 and 7:3:3 MLVs also show S_o + L_o phase coexistence, the S_o phase features are more prominent in all MLVs containing PCer-d31. Examining the signal between ∼26 and 30 kHz, we expect to see an approximately vertical edge in a typical L_o spectrum and a sloping shoulder for a S_o phase spectrum. When comparing the edges of the spectra for PCer-d31 in both 7:3:3 MLVs (Fig. 2, top spectra), it is apparent that the spectrum of MLVs containing PSM (top left spectrum) has less edge contribution than the spectrum of the corresponding MLVs containing DPPC (top right spectrum). Thus, a greater proportion of PCer-d31 is in the S_o phase in MLVs containing PSM relative to those containing DPPC. Quantitative fractions of the perdeuterated lipid present in either S_o or L_o phases could not be measured. This is because S_o and L_o spectra are similar in width and have similar relaxation times and therefore cannot easily be separated. Also, lipid exchange may complicate the quantitative determination of the amounts of S_o and L_o phases present when the S_o fraction is small. The spectral subtraction technique (35) cannot be used here because it would require subtracting two spectra along a given tie line in a known phase diagram. The phase diagram of the ternary MLVs studied here has not yet been determined.

In Fig. 3, the spectra for all the MLVs at 40°C are shown. PSM-d31 is in the transition from S_o to L_d phase, and its spectrum is a clear superposition of spectra from the S_o and L_d phases. The spectrum of DPPC-d31 is mainly in the S_o phase, with only a small amount of evidence of L_d phase (a more vertical shoulder compared to its 22°C spectrum; a liquid crystalline methyl group signal at about ±4 kHz). Adding 30 mol % cholesterol creates the L_o phase in MLVs of both PSM-d31 and DPPC-d31, and adding 10 mol % PCer does not change this. However, upon the addition of 20 or 30 mol % PCer to either 7:3:0 MLV system, subtle changes in the spectral edges and shoulders suggest that a fraction of either DPPC-d31 or PSM-d31 is in the S_o phase. This effect is slightly more pronounced for PSM-d31:Chol than for DPPC-d31:Chol, indicating that PCer has a larger ordering effect on PSM-d31 than on DPPC-d31. The shoulders and the edges of the spectra of PCer-d31 in both 7:3:2 and 7:3:3 MLVs are somewhat more sloped compared to DPPC-d31 and PSM-d31 in the analogous MLVs. Hence, close to physiological temperatures, a fraction of PCer-d31 is still in the S_o phase. As was the case at 22°C, the PSM:Chol:PCer-d31 7:3:3 spectrum has stronger S_o phase features than the DPPC:Chol:PCer-d31 spectrum.

Figure 3.

²H NMR spectra at 40°C of (A) PSM-d31:Chol:PCer MLVs (red) and PSM:Chol:PCer-d31 MLVs (blue); (B) DPPC-d31:Chol:PCer MLVs (green) and DPPC:Chol:PCer-d31 (purple). A second methyl signal, visible most clearly in the DPPC-d31 7:3:2 and 7:3:3 spectra, is due to a small proportion of sn-1 chain perdeuterated DPPC-d31 that results from chain transmigration during synthesis. To see this figure in color, go online.

Fig. 4 shows that PSM-d31 and DPPC-d31 MLVs are in the L_d phase at 50°C, meaning that the transition from S_o phase to L_d phase is completed. All the binary and ternary MLVs display L_o spectra. Upon addition of PCer (or PCer-d31) to the binary MLVs, a slight increase in the width of the spectra, in both Fig. 4, A and B, is observed. This increase will be illustrated quantitatively next.

Figure 4.

²H NMR spectra at 50°C of (A) PSM-d31:Chol:PCer MLVs (red) and PSM:Chol:PCer-d31 MLVs (blue) and (B) DPPC-d31:Chol:PCer MLVs (green) and DPPC:Chol:PCer-d31 (purple). To see this figure in color, go online.

Further insight into the effect of ceramide on PSM:Chol or DPPC:Chol 7:3 MLVs can be gained from examining the temperature dependence of the average width of the spectrum (M₁) in more detail (Fig. 5). In the absence of ceramide, both of these binary MLV systems display L_o phase spectra that gradually decrease in width as the sample is heated.

Figure 5.

M₁ versus temperature for MLVs containing the indicated molar ratios of (A) PSM-d31:Chol:PCer (open symbols) and PSM:Chol:PCer-d31 (solid symbols) and (B) DPPC-d31:Chol:PCer MLVs (open symbols) and DPPC:Chol:PCer-d31 (solid symbols). These data are compared to values for pure PSM-d31 and pure DPPC-d31 in Fig. S2.

The addition of 10 mol % PCer to PSM-d31:Chol 7:3 results in increased M₁ between 34 and 65°C but does not affect M₁ between 21 and 33°C. In contrast, the addition of 20 mol % PCer to the PSM-d31:Chol 7:3 causes a significant increase in M₁ values below 35°C because of an increased amount of S_o phase (see Fig. 2). At 35°C and above, the increase in M₁ is similar to that found for the addition of 10 mol % PCer. Increasing the PCer content further, to PSM-d31:Chol:PCer 7:3:3, results in slightly larger M₁ values below 35°C and significantly larger M₁ values above 35°C. The increased M₁ values below 35°C are caused by S_o phase domains found in PSM-d31:Chol:PCer 7:3:2 and 7:3:3, as discussed earlier. Above 35°C, the progressively larger M₁ values observed as PCer is added to PSM-d31:Chol 7:3 are due to enhanced chain order of L_o phase PSM-d31 molecules.

The temperature dependence of M₁ depends on which lipid is deuterium labeled, reflecting differences in phase behavior. For example, below 40°C, the M₁ values of PCer-d31 in PSM:Chol:PCer-d31 7:3:2 and 7:3:3 MLVs (Fig. 5 A) are substantially higher than the analogous MLVs containing PSM-d31. This is because of the presence of a higher proportion of PCer-d31 than of PSM-d31 in the S_o phase (see Fig. 2). When PCer-d31 is in the L_o phase, above 50°C, the M₁ values of MLVs containing 20 or 30 mol % added PCer-d31 are more similar to those of PSM-d31:Chol:PCer 7:3:2 and 7:3:3, indicating that the lipids are well mixed.

How does added PCer affect the spectra of DPPC-d31:Chol 7:3 MLVs (Fig. 5 B)? Above ∼35°C, the temperature dependence of M₁ for DPPC-d31:Chol 7:3 is very similar to that of PSM-d31:Chol 7:3 as PCer is incorporated. On the other hand, below 35°C, PCer does not appreciably change M₁ of DPPC-d31:Chol 7:3. Assessing instead the effect of PCer-d31 on DPPC:Chol 7:3, above 35°C, 30 mol % PCer-d31 added to DPPC:Chol 7:3 increases M₁ by ∼3% compared to 20 mol % added PCer-d31. Below 35°C, adding 20 mol% PCer-d31 to DPPC:Chol 7:3 MLVs results in spectra having nearly identical M₁ values to those observed when 20 mol % PCer is added to DPPC-d31:Chol 7:3 MLVs. However, below ∼30°C, the addition of 30 mol % PCer-d31 to DPPC:Chol 7:3 leads to spectra with a significant S_o component and consequently markedly higher M₁ values.

Generally, above 35°C, PCer addition to either DPPC-d31:Chol 7:3 or PSM-d31:Chol 7:3 increases M₁ in a similar way (Fig. 5, A and B). This is also true for PCer-d31 addition to either DPPC:Chol 7:3 or PSM:Chol 7:3. For example, above 35°C, the maximal difference between the M₁ values of DPPC:Chol:PCer-d31 7:3:3 and PSM:Chol:PCer-d31 7:3:3 is 5% (at 65°C and 70°C). But below 35°C, the addition of either PCer or PCer-d31 to the phospholipid:cholesterol (7:3) binary mixtures has a significantly different effect on the temperature dependence of M₁. Adding 20 or 30 mol % PCer dramatically increases the M₁ of the spectra of PSM-d31:Chol 7:3 MLVs but has nearly no effect on those of DPPC-d31:Chol 7:3 MLVs. Adding 20 or 30 mol % PCer-d31 to either PSM:Chol 7:3 or DPPC:Chol 7:3 results in clearly distinct changes to the temperature dependence of M₁. At 27°C and below, the spectra of PSM:Chol:PCer-d31 7:3:2 and 7:3:3 have much larger M₁ values than those of the spectra of DPPC:Chol:PCer-d31 7:3:2 and 7:3:3. This reflects the fact that a substantially higher proportion of PCer-d31 is in the S_o phase in MLVs containing PSM.

At 50°C, the spectra of all the MLVs with 20 or 30 mol % added PCer or PCer-d31 have very comparable M₁ values (Fig. 5). A detailed comparison of these spectra follows. In Fig. 6, the dePaked spectra of all the MLVs at 50°C are shown. The addition of PCer and Chol to PSM-d31 (Fig. 6 A) makes quadrupolar splittings larger, indicating that the palmitoyl chains are more ordered. An exception to this general observation is that the splitting of the 19 kHz peak on the PSM-d31 spectrum, which is assigned to one of the C2 deuterons (36), gets narrower upon addition of Chol and PCer. This effect has been observed before for PSM-d31:Chol binary mixtures (32). By comparing the dePaked spectra of PSM-d31:Chol:PCer and PSM:Chol:PCer-d31 7:3:2 (and 7:3:3), we observe that the width of the corresponding spectra and the positions of the methyl group signals are remarkably similar. However, there are some subtle differences that become apparent when order parameter profiles are calculated (see below).

Figure 6.

²H NMR dePaked spectra for (A) PSM-d31:Chol:PCer MLVs and PSM:Chol:PCer-d31 MLVs and (B) DPPC-d31:Chol:PCer MLVs and DPPC:Chol:PCer-d31 at 50°C. For each spectrum, the underlined digit in the composition indicates which lipid is deuterated. The plain arrows represent the signal from the deuterons on C2 of the deuterated palmitoyl chains in PSM-d31 and DPPC-d31 MLVs (36, 40, 45, 46). The bulleted arrow indicates the peak, representing one of the deuterons on C3 of the deuterated palmitoyl chain in PSM-d31 MLVs (45). The latter peak is not resolved in the DPPC-d31 spectrum. The reason for the different behavior of the deuterons attached to C2 and C3 is the kink in the palmitoyl chain at these carbons (47).

As shown in Fig. 6 B, the DPPC-d31 quadrupolar splittings also get larger upon addition of cholesterol and PCer, as was observed for PSM-d31 in Fig. 6 A. When comparing the spectrum of DPPC-d31:Chol:PCer and DPPC:Chol:PCer-d31 in 7:3:2 (and 7:3:3) MLVs in Fig. 6 B, we find that they are very similar to the analogous spectra (from MLVs containing PSM or PSM-d31) in Fig. 6 A. Quantitative comparisons of dePaked spectra can be accomplished by examining the order parameter profiles (Fig. 7).

Figure 7.

Order parameter profiles, $\left|{\mathbf{S}}_{\mathbf{C}\mathbf{D}}\right|$ values for each carbon’s deuterons, at 50°C obtained from the dePaked spectra of MLVs containing (A) PSM-d31:Chol:PCer (open symbols) and PSM:Chol:PCer-d31 (solid symbols); (B) DPPC-d31:Chol:PCer (open symbols) and DPPC:Chol:PCer-d31 (solid symbols). Please refer to Figs. S1 and S2 for more details on the peak assignments of the dePaked spectra used to calculate $\left|{\mathbf{S}}_{\mathbf{C}\mathbf{D}}\right|$.

In Fig. 7, the smoothed order parameter profiles at 50°C for the carbons along the palmitoyl chain of the deuterated lipid in MLVs having each studied composition are shown. The peaks were assigned as explained in the Supporting Materials and Methods. In Fig. 7, we observe that Chol orders all PSM-d31 and DPPC-d31 deuterons with the exception of one of those attached to C2. Upon addition of 10–30 mol % PCer to PSM-d31:Chol 7:3, the order parameter of carbons in the palmitoyl chain of PSM-d31 increases slightly. Examining the effect of addition of 30 mol % Chol to either of the pure phospholipid MLVs using the order parameter profiles in Fig. 7, A and B shows that Chol increases the order parameter of PSM-d31 and DPPC-d31 by approximately the same amount, especially in carbons close to the headgroup. For example, Chol increases the order parameter at C4 in both DPPC-d31 and PSM-d31 by ∼0.16, but it increases the order at C15 by 0.12 in DPPC-d31 and by 0.15 in PSM-d31. Generally, this confirms the results obtained from M₁ (which is directly proportional to the average order parameter) analyses. There, it was shown that the differences in the M₁ of the spectra at 50°C, upon addition of 30 mol % Chol, are very similar. Comparing the order parameter profiles of MLVs containing PCer (or PCer-d31) in Fig. 7 B with those in Fig. 7 A reveals that PCer is more effective at ordering DPPC-d31 than PSM-d31. The order parameter profile data show that DPPC-d31 is consistently less ordered in DPPC-d31:Chol:PCer 7:3:0 and 7:3:1 compared to 7:3:2 and 7:3:3. The increase in order for a given chain position in DPPC-d31:Chol 7:3 is approximately linear up to 20 mol % added PCer, but no significant further increase is observed when 30 mol % PCer is added. In general, based on the results of order parameter analysis, it can be stated that the palmitoyl chain in PSM-d31 is more ordered than the sn-2 palmitoyl chain of DPPC-d31 in pure phospholipid, 7:3:0, and 7:3:1 MLVs. When PCer is present in higher concentrations, the order parameters of the deuterated lipids are very similar. The order parameter of 20 or 30 mol % PCer-d31 added to either PSM:Chol 7:3 or DPPC:Chol 7:3 are also very similar, both to each other and to the order parameters of PSM-d31 or DPPC-d31 in the analogous MLVs. So, at 50°C, the order parameter profiles of all the MLVs containing 20 or 30 mol % added PCer (or PCer-d31) are nearly indistinguishable.

In Figs. S3–S5, pairwise comparisons of the dePaked spectra of MLVs containing 30 mol % added PCer or PCer-d31 are plotted at 50, 55, 60, 65, and 70°C. The dePaked spectra of MLVs composed of PSM-d31:Chol:PCer and PSM:Chol:PCer-d31 are compared in Fig. S3. Both sphingolipids display essentially the same spectra from 50 to 70°C, indicating that they are well mixed in the presence of Chol. The same comparison was done for DPPC-d31:Chol:PCer versus DPPC:Chol:PCer-d31 in Fig. S4. At 55°C and above, DPPC-d31 is less ordered than PCer-d31. However, this is not evidence of inhomogeneous mixing because the lipid backbone is known to affect chain order (37). In a binary mixture of a sphingolipid and a glycerophospholipid with a known phase diagram, such as POPC:PCer, the average order in the plateau region (C4–C6) of the palmitoyl chain can differ by more than 12% even in a homogeneous liquid crystalline region of the phase diagram (38). The maximal difference in plateau order observed in Fig. S4 is 11.4%, so we propose that the lipids are likely to be well mixed.

Finally, the dePaked spectra of PSM:Chol:PCer-d31 and DPPC:Chol:PCer-d31 7:3:3 MLVs are plotted in Fig. S5. From this figure, it is apparent that the spectra at 50°C are almost identical, but as the temperature increases, PCer-d31 is less disordered in PSM-containing MLVs than in those containing DPPC. A similar conclusion can be reached by comparing the slopes of the high-temperature region of Fig. 5: it is apparent that the sphingolipid membrane is relatively insensitive to increasing temperature compared to the analogous DPPC-containing membrane.

Discussion

We have thoroughly examined the influence of increasing ceramide concentrations on membranes composed of either PSM:Chol or DPPC:Chol. In the absence of Chol and ceramide, both DPPC-d31 and PSM-d31 undergo a sharp transition from S_o to L_d phases near 40°C. We confirmed that the addition of 30 mol % Chol to either DPPC-d31 or PSM-d31 abolishes the phase transition from S_o to L_d phases. We also found that the addition of 30 mol % Chol to PSM-d31:PCer and PSM:PCer-d31 7:3 abolishes the broad S_o to L_d phase transition, complete by 70°C, which was previously observed using ²H NMR and DSC (7).

Our results show that ceramide is able to promote the formation of S_o phases in DPPC and PSM bilayers in the presence of Chol at low temperatures. Upon adding PCer (or PCer-d31) to the 7:3 binary mixture of phospholipid:Chol, S_o phase formation was observed in both DPPC (or DPPC-d31) and PSM (or PSM-d31) MLVs. The reason that PCer-d31 “prefers” the S_o phase at lower temperatures could be related to its having a smaller ratio of headgroup size to chain cross-sectional area compared to PSM. At lower temperatures, PSM chains have more freedom compared to those of PCer, therefore making the S_o phase less favorable for PSM. At higher temperatures, PCer chains have enough thermal energy to compensate for the fact that its small headgroup does not provide space for chain motion and the S_o phase is no longer observed. Below 35°C, all ternary mixtures show a superposition of S_o and L_o phase spectra. Thus, the S_o phases induced by addition of PCer (or PCer-d31) coexist with the L_o phase induced by the addition of Chol to either phospholipid. As shown in Fig. 2, the spectra of PCer-d31 in PSM:Chol:PCer-d31 and DPPC:Chol:PCer-d31 7:3:3 MLVs at 22°C show line shapes dominated by a S_o phase spectrum. At the same temperature, PSM-d31 and DPPC-d31 in both PSM-d31:Chol:PCer and DPPC-d31:Chol:PCer 7:3:3 are primarily in the L_o phase. Because the predominant lipid in all ternary MLVs is the phospholipid, the surface area of L_o phase domains should be greater than that of the S_o domains.

One reason for an observation of higher contributions from S_o phases at higher concentrations of PCer (or PCer-d31) is the dilution of Chol. For example, 7:3:1, 7:3:2, and 7:3:3 MLVs contain 27, 25, and 23 mol % Chol respectively. It should also be noted that a substantial amount of S_o phase is present in the binary mixture of either PSM-d31:Chol and DPPC-d31:Chol containing 23 mol % Chol (32). Our results are consistent with previously published data on lipid mixtures with similar molar ratios. For example, confocal microscopy of GUVs containing PSM:Chol:PCer and DPPC:Chol:PCer 7:3:3 (molar ratios) that are stained with naphtho[2,3-a]pyrene (NAP) clearly show two phases at 22°C (28). Because there are six fused rings in the molecular structure of NAP, it tends to accumulate in the Chol-rich domains in the bilayers (28). Thus, combining the results of NAP-stained GUVs and ²H NMR at 22°C, it can be concluded that the probe partitions into domains that are rich in PSM and Chol (i.e., the L_o phase). A comparison of the ²H NMR and GUV fluorescence data yields some information about the composition of the observed S_o phases. From ²H NMR, we know that they are rich in ceramide, whereas their dark nature in NAP-labeled GUVs indicates that they are Chol depleted. In the same article, both GUVs (PSM:Chol:PCer and DPPC:Chol:PCer 7:3:3) were examined with another probe, DiIC-18. The results for both DiIC-18-stained GUVs show an almost homogenous single-phase GUV (28). This is because DiIC-18 partitions into ordered domains and therefore partitions nearly equally in the L_o and S_o phase domains (28). Thus, the combined results from NAP- and DiIC-18-stained GUVs are consistent with the results obtained from ²H NMR experiments at 22°C shown here.

DSC experiments have also been performed on PSM:Chol:PCer 7:3:0, 7:3:1, 7:3:2, and 7:3:3 MLVs (27). The DSC thermogram of PSM:Chol 7:3 consists of a very wide, nearly flat peak, indicating that there is no abrupt phase transition between 10 and 90°C. As PCer is incrementally added to the MLVs, this endotherm becomes somewhat sharper and the center of the peak shifts to higher temperatures (58°C for PSM:Chol:PCer 7:3:3). Our ²H NMR results suggest that this peak becomes sharper because, at low temperature, some S_o phase forms in the MLVs upon addition of PCer. However, when comparing DSC and ²H NMR data, it should be noted that each spectrum in ²H NMR experiments is taken when the system is at thermal equilibrium, which is not the case in DSC measurements. This could be why a transition is seen in DSC thermograms but not in M₁ versus temperature graphs calculated from ²H NMR spectra (Fig. 5).

In an AFM imaging study, the contact mode height images of the supported planar bilayers (SPBs) of DPPC and PSM at 23°C showed almost the same height for each SPB (29). Similar increases in height resulted from the addition of Chol to both PSM and DPPC SPBs. This corresponds, probably, to the formation of L_o phases in both kinds of binary MLVs observed in our ²H NMR. In the presence of PCer in PSM:Chol:PCer and DPPC:Chol:PCer 7:3:3, domains could not be detected by AFM either in imaging or in force spectroscopy. It is possible that the L_o and S_o domains in the ternary mixture have similar heights that cannot be resolved by AFM. This would still be compatible with the S_o + L_o phase coexistence for PSM/DPPC-d31:Chol:PCer and PSM/DPPC:Chol:PCer-d31 7:3:3 MLVs found in this report.

Highly ordered ceramide-enriched domains have been observed in quaternary mixtures of SPBs containing stearoyl SM (SSM), Chol, stearoyl Cer (SCer) with dioleoyl-phosphatidylcholine (DOPC) using AFM and fluorescence correlation spectroscopy (39). In SPBs with high Chol and low Cer contents, DOPC:SSM:Chol:SCer 1:1:0.88:0.12 (4% SCer), Chol-driven domains in the L_o phase were observed. Upon increasing ceramide and decreasing Chol contents in DOPC:SSM:Chol:SCer 1:1:0.64:0.36 (12% SCer) SPBs, a decrease in the surface area and increase in the height of the domains were observed using both techniques. Integrating these observations with ours, one can conclude that the formation of the ceramide-enriched S_o domains depends on the relative concentration of Chol and ceramide. So, if ceramide and Chol are present in comparable concentrations, ceramide can partially disassemble the Chol-driven L_o phase domains, as we observed in 7:3:2 and 7:3:3 MLVs at temperatures below 35°C. As proposed earlier, cells could employ this critical balance between ceramides and Chol to regulate membrane fluidity and control the size of highly ordered ceramide-enriched domains in processes like signaling or apoptosis (see review in (9)).

Our order parameter analyses at 50°C show that whether Chol is present, PSM-d31 is more ordered than DPPC-d31 in MLVs because of the local ordering effect of the sphingosine backbone (40). Notably, we see this in the PSM-d31:Chol and DPPC-d31:Chol 7:3 MLVs for which the order parameters of PSM-d31 are higher than those of DPPC-d31. However, Chol increases the order of PSM-d31 and DPPC-d31 to the same degree, as was shown by both M₁ and order parameter profiles analyses (Figs. 5 and 7). Adding 10 mol % PCer similarly increases the order of both PSM-d31:Chol and DPPC-d31:Chol 7:3 MLVs. The properties of our saturated phospholipid:Cer:Chol systems at high temperatures should be comparable to those of the DOPC- or POPC:Cer:Chol studied by Slotte et al. (41). In fact, those authors observed, using ²H NMR, a marked ordering effect of the sterol on the ceramide acyl chain. Higher amounts of added PCer increase the order of DPPC-d31:Chol more than PSM-d31:Chol, resulting in very similar order parameters for the PSM-d31:Chol:PCer and DPPC-d31:Chol:PCer 7:3:3 MLVs. Our results provide a physical basis for the observations by Silva et al. (26) who treated SM:Chol bilayers with a sphingomyelinase, under conditions such that part of the SM would be converted to Cer. These authors found that at low Chol concentrations, Cer would segregate laterally with SM into small S_o gel nanodomains, whereas at higher Chol concentrations, a Chol-rich L_o phase would predominate.

The observation that SM is not essential in the formation of ordered domains, as long as saturated PC is present, confirms previous work by Sullan et al. (18) and Nyholm et al. (19), and it might open the applicability of our studies to nonsphingolipid-containing membranes (e.g., in bacteria). In relation to this matter, Huang et al. (42) demonstrated, in the plasma membrane of the sterol-containing but sphingolipid-lacking bacterium Borrelia burgdorferi, the formation of L_o domains rich in sterol and saturated PC.

Conclusions

Here, we have investigated the changes in the physical behavior of two major lipids involved in apoptosis. Two important sphingolipids, PSM, and PCer were studied in MLVs containing Chol. The lipid molar ratios were chosen to determine the effect of PCer addition to the PSM:Chol 7:3 membrane, monitored using ²H NMR. We also tried to shine a light on the question of why nature chose saturated sphingolipids over saturated glycerophospholipids in the plasma membrane, even though DPPC shares a very similar molecular structure and main transition temperature (i.e., physical behavior) with PSM. To this end, the effect of replacing PSM (or PSM-d31) with DPPC (or DPPC-d31) in all MLVs was studied as well.

We have directly shown that PCer induces formation of S_o phase domains in MLVs containing PSM and Chol at temperatures below 40°C. The observed S_o domains are enriched in PCer, whereas most of the PSM remains in the L_o phase. This was proven by placing the probe, that is, the perdeuterated palmitoyl chain, on either sphingolipid in two different MLV preparations having the same molar ratios of lipids and performing ²H NMR spectroscopy. Generally, our results are in agreement with what has been observed in similar lipid mixtures using other biophysical techniques such as AFM, DSC, and fluorescence microscopy. By combining these observations, we propose that near physiological temperature, S_o domains in SM form in the presence of Chol only when the concentration of PCer is comparable to that of Chol. The potential of Chol and ceramide together to affect the lipid packing in membranes could have significant biological importance. Cells could utilize this potential by maintaining a critical balance between the local membrane concentrations of ceramide and Chol to regulate the membrane fluidity in signaling processes such as apoptosis.

Comparing the effect, at 35°C and below, of increasing ceramide content in phospholipid:cholesterol 7:3 MLVs containing DPPC (or DPPC-d31) instead of PSM (or PSM-d31), we observed that ceramide is much more capable of inducing S_o phase domains in MLVs containing PSM (or PSM-d31). From 35 to 50°C, however, no significant differences were found upon an increase in ceramide content in MLVs containing either PSM (or PSM-d31) or DPPC (or DPPC-d31). We speculate that at low temperatures, the interaction between PSM and PCer is stronger than of that between DPPC and PCer so that cholesterol is excluded from PSM:PCer domains to a greater extent, allowing a higher proportion of S_o phase to form. Presumably, above 35°C, thermal motion negates the increased stability of the PSM:PCer interaction. It cannot be determined here whether the behavior below 35°C or between 35 and 50°C is more biologically relevant. Therefore, the reason why nature chose saturated SM over saturated PC in biological membranes remains elusive.

One important question that could be addressed in future studies is whether in the presence of ceramide, Chol partitions into both L_o and S_o phases or is concentrated mainly in the L_o phase. Selectively labeled deuterated Chol, like Chol-d1 (43), could be used to determine the phase of Chol in MLVs. A second important question to pursue is the effect of unsaturated PC, for example POPC, on the observed phase behavior of sphingomyelin:Chol:ceramide. POPC, like SM, is a significant component of the outer leaflet of cell plasma membranes (1, 44). It is important to see to what extent the S_o phases formed at high PCer concentration can still exist in MLVs with compositions that are better mimics of the outer leaflet of cell membranes.

Author Contributions

R.S. designed research, performed research, analyzed and interpreted data, and wrote the article. T.P. performed research and wrote the article. S.S.W.L., A.A., and F.M.G. designed research and wrote the article. J.L.T. designed research, interpreted data, and wrote the article.

Editor: Michael Brown.