Ceramide-C16 Is a Versatile Modulator of Phosphatidylethanolamine Polymorphism

Abstract

Ceramide-C16 (CerC16) is a sphingolipid associated with several diseases like diabetes, obesity, Parkinson disease, and certain types of cancers. As a consequence, research efforts are devoted to identify the impact of CerC16 on the behavior of membranes, and to understand how it is involved in these diseases. In this work, we investigated the impacts of CerC16 (up to 20 mol %) on the lipid polymorphism of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE), using differential scanning calorimetry, and sequential ²H and ³¹P solid-state nuclear magnetic resonance spectroscopy. A partial phase diagram is proposed. The results indicate that the presence of CerC16 leads to an upshift of the temperature of the gel-to-liquid crystalline (L_β − L_α) phase transition, leading to a large L_β/L_α phase coexistence region where gel-phase domains contain ∼35 mol % CerC16. It also leads to a downshift of the temperature of the lamellar-to-inverted hexagonal (L − H_II) phase transition of POPE. The opposite influence on the two-phase transitions of POPE brings a three-phase coexistence line when the two transitions overlap. The resulting H_II phase can be ceramide enriched, coexisting with a L_α phase, or ceramide depleted, coexisting with a L_β phase, depending on the CerC16 proportions. The uncommon capability of CerC16 to modulate the membrane fluidity, its curvature propensity, and the membrane interface properties highlights its potential as a versatile messenger in cell membrane events.

Introduction

Ceramides are bioactive sphingolipids that play an important role in cellular signaling and mediate several biological processes (1, 2, 3, 4, 5, 6, 7, 8, 9). For example, they are shown to be involved in some cellular events, including apoptosis (10, 11, 12), and in biological pathways leading, for example, to insulin resistance, and obesity (13, 14). In mammalian cell membranes, the most abundant acyl chains borne by ceramides are saturated and contain 16–24 carbons (1, 15). The ceramide functions are intimately associated with their acyl chain length (10, 16). Among this family, ceramide with a palmitoyl chain (N-palmitoyl-D-erythro-sphingosine (CerC16)) has been specifically shown to have a central influence in some cellular events. Because of the biological key roles of these molecules, it is essential to gain a detailed understanding of the modulation of the properties of biological membranes by ceramides; in this work, we focused our effort on characterizing the impact of CerC16 on the membrane physical properties.

From a chemistry point of view, ceramides display a relatively small polar headgroup that has the capability of forming hydrogen bonds through both the hydroxyl groups, and the amide linkage. As a consequence, ceramides with saturated acyl chains exhibit a dense molecular packing and have relatively high gel (L_β)-to-fluid (L_α) phase transition temperatures (T_m) in comparison to phospholipids (17). For instance, CerC16 undergoes a chain-melting transition at ∼91°C (18, 19). The impacts of ceramides on phospholipid bilayer properties have been recently reviewed (20, 21). These impacts are highly dependent on the ceramide chain length (19, 22, 23, 24). In general, when mixed with phospholipids, long-chain ceramides lead to an increase of the L_β − L_α phase transition temperature; this shift was observed for different binary lipid mixtures including 1,2-dierucoyl-sn-glycero-3-phosphoethanolamine (DEPE) (19, 25), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) (24, 26, 27, 28), 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) (29), or 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) (30). Moreover, ceramides induce a considerable broadening of the L_β − L_α phase transition, a phenomenon observed for binary mixtures of phosphatidylcholines (PC) and of phosphatidylethanolamine (PE) (19, 24, 26). This effect leads to a large phase-coexistence region in the phase diagrams where ceramide-rich gel-phase domains are found with ceramide-depleted fluid domains (19, 24, 26). The strong interceramide interactions have been identified as the driving force leading to the formation of these gel-phase ceramide-enriched domains (26). Ceramides also lead to an increase in chain order of fluid bilayers as reported by studies using solid-state deuterium nuclear magnetic resonance (²H-NMR) (22, 26, 31), and diphenylhexatriene fluorescence depolarization techniques (27, 28, 32). It was proposed that phase separations could also be observed at low temperatures, leading to the formation of two coexisting gel phases containing different ceramide proportions (19, 23, 26).

In addition to their effect on lateral mixing of lipids, ceramides have been shown to have an impact on the polymorphic propensities and curvature properties of bilayers; this phenomenon can be an alternative manner of modulating membrane properties. CerC16 (19), egg yolk, and brain ceramides (25), bearing mainly C16 and C18 saturated chains, respectively, cause a decrease in the L_α-to-inverted hexagonal (H_II) phase transition temperature (T_H) of DEPE. For example, the addition of 10 mol % CerC16 brings T_H down from 66°C for pure DEPE to 56°C (19). Differential scanning calorimetry (DSC) results showed that the endotherm peak associated with this transition becomes broader in the presence of CerC16. The promotion of the H_II phase by ceramide has been associated with its small headgroup, and with its extensive capability of forming H bonds (19). Because of their opposite effects on the two transitions (increasing T_m while decreasing T_H), ceramides can lead to an overlap of the L_β − L_α, and the L_α − H_II transitions (19). The presence of ceramides in bilayers formed by PE, PC, or a mixture of both, was shown to induce membrane fusion and leakage; these phenomena were associated with the fact that ceramides could facilitate the formation of nonlamellar phases (33, 34, 35, 36). It has also been proposed that ceramides promote the formation of nonlamellar intermediates because they display a rapid transbilayer motion (flip-flop) (37). Moreover, enzymatically produced ceramides cause the budding of giant unilamellar vesicles, and this formation of smaller vesicles was linked to the propensity of ceramides to acquire nonlamellar phases (38). Despite the fact that the ability of CerC16 to promote the H_II phase is established, there is a limited knowledge relative to the molecular details of inserted CerC16 in this lipid matrix.

In this work, we employed DSC, and sequentially acquired ²H and ³¹P solid-state NMR techniques to study the impacts of CerC16 on the polymorphism of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE). The CerC16 proportion was varied between 0 and 20 mol %; this range covers membrane concentrations observed in physiologically relevant conditions—for example, the ceramide content in mitochondria increases considerably during apoptosis and can reach 10% of total lipids (39, 40). Moreover, it has been reported that metastable phases sensitive to hydration and thermal history can be formed above 20 mol % (26). By using CerC16 bearing a fully deuterated palmitoyl chain (N-palmitoyl-d₃₁-D-erythro-sphingosine (CerC16-d₃₁)), ²H NMR spectra provided a quantitative characterization of the phases in which CerC16 was involved as well as a description of the ceramide chain order. In parallel, ³¹P chemical shielding anisotropy (CSA) of POPE provided a description of the phase behavior of the phospholipid. We carried out a sequential acquisition of the ²H, and ³¹P NMR spectra of POPE/CerC16-d₃₁ mixtures, in defined and controlled conditions, including temperature. These data allowed a detailed description of the mixture phase behavior. A mirror sample with POPE bearing a fully deuterated palmitoyl chain (POPE-d₃₁) was also examined, not only to validate the phase behavior of POPE in the mixtures, but also to compare the order of the POPE acyl chain with that of CerC16.

Materials and Methods

Materials

POPE, POPE-d₃₁, CerC16, and CerC16-d₃₁ were purchased from Avanti Polar Lipids (Alabaster, AL). All the lipids were >99% pure, and were utilized without further purification. 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES, >99%), ethylenediaminetetraacetic acid (EDTA, >99%), and deuterium-depleted water (≤1 ppm deuterium oxide) were purchased from Sigma-Aldrich (St. Louis, MO). NaCl (high purity grade) was supplied by AMRESCO (Solon, OH). Benzene (>99%) and methanol (HPLC grade) were acquired from EMD Millipore (Billerica, MA) and Fisher Scientific (Fair Lawn, NJ), respectively.

Lipid mixture dispersions

The lipid mixtures were prepared from organic solutions. For the DSC measurements, a solution of CerC16 with a known concentration was first prepared by dissolving a weighted quantity of CerC16 into an exact volume of a 90:10 (v/v) benzene/methanol mixture. Binary lipid mixtures with the desired composition were prepared by dissolving a weighted quantity of POPE into the appropriate volume of the CerC16 solution. The resulting solutions were freeze-dried for at least 24 h to ensure the complete solvent elimination. For the NMR samples, the binary lipid mixtures were prepared by dissolving the appropriate quantities of POPE, and CerC16 into the 90:10 (v/v) benzene/methanol mixture and by freeze-drying the solution for at least 24 h.

The lipid mixtures were hydrated using a HEPES buffer (20 mM HEPES, 100 mM NaCl, and 0.05 mM EDTA, prepared in Milli-Q water; Millipore), pH 7.4. For the DSC experiments, the final lipid concentration was ∼20 mM, except for the pure POPE dispersion for which the concentration was 7.8 mM. For the NMR samples, ∼8 mg (∼14 μmol) of CerC16-d₃₁ were used for ensuring a good ²H-NMR signal. An ∼200 μL aliquot of buffer was used to hydrate each sample. Thus, the final lipid concentration varied from 703 mM for the POPE/CerC16-d₃₁ 95:5 mixture, to 350 mM for the molar ratio 80:20. In the case of the POPE-d₃₁/CerC16 90:10 dispersion, 12 mg (16 μmol) POPE-d₃₁, 38 mg (53 μmol) POPE, and 4 mg (8 μmol) CerC16 were used, and the final lipid concentration was 384 mM. The same hydration protocol was performed for both DSC and NMR techniques: two heating-cooling cycles between 45 and 0°C, and a third cycle between 95 and 0°C were imposed on each lipid sample.

DSC measurements

The DSC measurements were carried out on a VP-DSC MicroCalorimeter (MicroCal, Northampton, MA). Three consecutive heating scans between 5 and 85°C in the case of the pure POPE dispersion, and between 5 and 95°C for the binary lipid mixtures, were carried out at a heating rate of 30°C/h. The second and third scans provided very similar thermograms, and the second heating scans were selected for the analysis.

NMR spectroscopy

The spectra were recorded using an Avance II 400 WB spectrometer (Bruker, Billerica, MA) equipped with a 9.4 T magnet, leading to a resonance frequency of 61.43 and 162.03 MHz for ²H and ³¹P nucleus, respectively. A static probe (Bruker) with a 5 mm coil was used. In the case of ²H acquisitions, 5000 scans were recorded using the quadrupolar echo pulse sequence with a 90° pulse of 1.7 μs, an interpulse delay of 40 μs, and a recycle time of 300 ms. For the ³¹P nucleus, a single 90° pulse of 4.05 μs, a recycle delay of 1 s, and a “Waltz65” low-power proton decoupling were used for recording the spectra; typically, 1500 scans were recorded. The sequential acquisition of ²H and ³¹P NMR spectra was carried out as a function of temperature, between 0 and ∼65°C. After the temperature stabilization, the signal acquisition of either nucleus was initiated. After the acquisition of the first spectrum, the spectrometer was then tuned for the signal acquisition of the other nucleus. The probe that was used presented the advantage of being tuned for the different nuclei without having to be removed from the magnet; this aspect was essential in the sequential acquisition of the NMR spectra as it kept the samples in the very same conditions during the data collection. At the phase transition temperatures, the spectrum of the first recorded nucleus was duplicated after the signal acquisition of the second nucleus, to validate that no significant phase evolution could be observed.

Based on the ²H spectra, the phase distribution of CerC16 could be determined, as described in the Results and Discussion. In a similar way, the ³¹P NMR spectra provided a description of the phase distribution of POPE as discussed below. The smoothed order profiles of the lipid acyl chains were obtained from the dePaked (41) ²H-NMR spectra using the method previously described (42).

Results and Discussion

DSC measurements

The DSC thermograms of a pure POPE dispersion and of POPE/CerC16 dispersions of various compositions are shown in Fig. 1. POPE displayed a L_β − L_α phase transition at 25°C (T_m), whereas the L_α − H_II phase transition was observed at 72°C (T_H). CerC16 disturbed significantly the phase behavior of POPE, even at a molar fraction of 5 mol %. The presence of CerC16 resulted in two major impacts on the thermogram of a POPE dispersion: first, a broadening, and an upshift in temperature of the endothermic L_β − L_α transition peak; and second, a broadening and a downshift in temperature of the lamellar (L)−H_II transition peak. The molar enthalpies of both phase transitions of pure POPE and of the investigated POPE/CerC16 dispersions are reported in Table 1. On the one hand, CerC16 induced a decrease in the ΔH of the L_β − L_α transition, reducing it from 23.9 kJ/mol for pure POPE, which is in good agreement with Pozo Navas et al. (43), to ∼17.5 kJ/mol when the proportion of ceramide was between 5 and 20 mol %. On the other hand, CerC16 appeared to have no considerable influence on the ΔH of the L − H_II transition as the values obtained for the investigated ceramide contents remained at ∼2.2 kJ/mol, a value observed for pure POPE dispersions (44). These results were analogous to those reported for DEPE/CerC16 mixtures (19), as well as for DEPE/egg-ceramides mixtures (25).

Figure 1.

Given here are DSC thermograms of POPE/CerC16 dispersions: (A) 80:20, (B) 85:15, (C) 90:10, (D) 95:05, and (E) pure POPE dispersion. To enhance the clarity, the thermograms were offset by 5 kJ/mol/°C.

Table 1.

The L_β − L_α, and L − H_II Phase Transition Temperatures and Enthalpies of a Pure POPE Dispersion and POPE/CerC16 Dispersions of Various Molar Fractions

POPE/CerC16 (mol ratio)	Lβ − Lα Phase Transition Onset/End (°C)	ΔH (Lβ − Lα) (kJ/mol)	L − HII Phase Transition Onset/End (°C)	ΔH (L − HII) (kJ/mol)
100:00	23.6:27.1	23.9	71.7:73.8	2.2
95:05	23.0:34.6	17.9	60.5:68.4	1.6
90:10	23.0:40.9	17.5	56.7:66.2	1.8
85:15	23.1:—	18.3	—:57.8	2.4
80:20	23.1:—	16.9	—:56.2	2.4

The ΔH values are reported per mole of lipids, i.e., POPE + CerC16.

NMR spectroscopy

The sequential ²H and ³¹P NMR spectra of POPE/CerC16-d₃₁ dispersions of various compositions are shown in Figs. 2 and S1. As mentioned above, ²H and ³¹P NMR spectra provided independent information about the phase characteristics of CerC16-d₃₁ and POPE, respectively. For ³¹P NMR, the CSA is representative of the different phospholipid phases. For instance, at 10°C, the ³¹P spectra of the lipid dispersions showed a CSA of ∼69 ppm, a value characteristic of a gel lamellar phase. At 42°C, the CSA was reduced to ∼44 ppm and the spectra were typical of fluid systems (45, 46). The spectra indicated the shift of the L_β − L_α phase transition toward high temperatures as the ³¹P-NMR spectra recorded in the L_β/L_α coexistence region displayed CSA values intermediate between those observed in the pure L_β, and the pure L_α phases; for example, at 42°C, the spectrum of 20 mol % CerC16 mixture (CSA of 47 ppm) was broader than that of 10 mol % CerC16 dispersion (CSA of 45 ppm). ³¹P-NMR spectra of phospholipids in the H_II phase display a CSA reduction by a factor of ∼2, and an inversion of their line shape because of the rapid diffusion of the lipid molecules around the cylinders (46, 47). The shift of the L_α − H_II phase transition toward low temperatures in the presence of CerC16 was also observed by ³¹P-NMR spectroscopy. For example, the spectrum recorded at 58°C for the mixture containing 10 mol % CerC16 displayed a coexistence of the L_α, and H_II patterns, whereas that at 57°C for the mixture containing 15 mol % CerC16 was essentially representative of the H_II phase. For the samples containing 15 and 20 mol % CerC16, a small narrow peak at 0 ppm was observed at high temperatures; it could correspond to the formation of a cubic phase, or small lipid assemblies. This narrow component always corresponded to <10% of the ³¹P spectra area. The formation of these structures was reversible as no such narrow component was observed in the spectra recorded at 25°C after the heating run.

Figure 2.

Shown here are the ²H (left column) and ³¹P (right column) NMR spectra of POPE/CerC16-d₃₁ dispersions with various molar ratios: (A) 90:10, (B) 85:15, and (C) 80:20. The acquisition temperature is indicated on the right. The narrow peaks at 0 kHz in the ²H-NMR spectra at high temperature were topped to allow a good representation of the H_II signal.

In the case of ²H-NMR, a broad featureless spectrum characteristic of a gel phase (48) was obtained at 10°C for all the lipid dispersions. Upon heating, there was the apparition of a pattern representative of the L_α phase. This pattern was composed of several overlapping powder patterns with different quadrupolar splittings associated with the orientational order gradient existing along the lipid perdeuterated acyl chain (42, 48, 49). The ²H-NMR spectra indicated a shift toward higher temperatures of the L_β − L_α phase transition experienced by CerC16-d₃₁. Several spectra were a superposition of L_β-, and L_α-phase components and it was observed from those recorded at 42–47°C that the proportion of the L_β-phase component increased with increasing CerC16 content. Upon further heating, a profile associated with the H_II phase was observed. This pattern displayed quadrupolar splittings reduced by a factor of >2 compared to that of L_α phase (50). The overall shape was also different compared to that of the L_α-phase component because the symmetry of the H_II phase led to a more linear decrease of orientational order along the lipid chain. The spectra of all the dispersions at ∼65–68°C corresponded to the H_II pattern. These spectra included, as for the ³¹P-NMR spectra, a small narrow peak centered at 0 kHz, representative of ceramides experiencing isotropic motions on the NMR timescale. This component represented, at its maximum, 5% of the area. Three observations must be highlighted. First, the spectra recorded at 54–55°C for the POPE/CerC16-d₃₁ mixtures with 15 and 20 mol % CerC16 were a superposition of components typical of the L_β, L_α, and H_II phases, revealing the coexistence of these three phases. Second, the spectra at 58°C of the POPE/CerC16-d₃₁ mixtures with a molar ratio of 95:5 and 90:10 included L_α-, and H_II-phase components whereas those of the 85:15 and 80:20 POPE/CerC16-d₃₁ mixtures at 57°C were composed of the L_β- and H_II-phase components. Third, all the CerC16 molecules were solubilized in the H_II phase upon heating, as inferred from the ²H-NMR spectra showing a single profile characteristic of the H_II phase.

We proceeded to a quantitative analysis of the spectra to determine the composition of the various phases. Because the lamellar, and H_II phases lead to distinct ²H- and ³¹P-NMR signals, the proportion of CerC16 (bearing the deuterated chain) and POPE (bearing the phosphate-containing headgroup) in each phase was inferred from the areas of the two components that could reproduce the experimental spectra by a linear combination. The components of the pure phases corresponded to experimental spectra acquired under different conditions (generally at a slightly different temperature and/or with a different CerC16 content). The area of the narrow peaks was determined directly on the spectra. In the case of ²H-NMR spectra, the L_β, and L_α phases also led to two different profiles that could be resolved. Therefore, the distribution of CerC16-d₃₁ between the different lamellar phases was inferred from the relative area of each component whose linear combination led to the best fit of the experimental spectra. The spectra typical of the pure phases were again obtained experimentally under different conditions leading to the presence of a single phase. The phase composition graphs of the binary lipid dispersions obtained from this quantitative analysis are presented in Fig. S2.

Sequential ²H- and ³¹P-NMR was also used to determine the phase behavior of a POPE-d₃₁/C16-Cer 90:10 dispersion (Fig. 3). The phase composition analysis from the ²H- and ³¹P-NMR spectra was carried out using the approach described above; in this case, both nuclei probed the phase behavior of POPE in the mixture. For the L_α and H_II phases, the ²H- and ³¹P-NMR results provided a similar phase description, with an average difference of 5%. These were also consistent with those provided by the ³¹P-NMR spectra of the POPE/CerC16-d₃₁ mirror sample. The inferred phase distribution at 57–58°C displayed a difference of 20% between the mirror samples, illustrating that the L_α-H_II phase transition was very sensitive to the experimental conditions. This susceptibility also confirms that sequential NMR acquisition, as carried out in this work, is essential to accurately describe the behavior of the labeled species. The fractions of POPE in the L_β, and L_α phases were obtained from the ²H NMR spectra of POPE-d₃₁, as described above (Fig. S3).

Figure 3.

Given here are the ²H (left) and ³¹P (right) NMR spectra of a POPE-d₃₁/CerC16 90:10 dispersion.

The combination of the NMR and DSC data led to the construction of the partial phase diagram of POPE/CerC16 system (Fig. 4). This diagram does not take into account the contribution of the narrow line in the NMR spectra as this was always a small component and its assignment could not be clearly established. The onset of the L_β − L_α phase transition of the investigated POPE/CerC16 mixtures was very similar for the different proportions of CerC16, suggesting the existence of a three-phase line and, consequently, a gel/gel phase coexistence region (indicated as L_β1 + L_β2 in the phase diagram). The existence of ceramide-rich and ceramide-depleted domains is also reflected by the solidus line that is considerably different compared to the one predicted from regular solution theory (51) (see Fig. S4). It is noteworthy that the gel-phase NMR spectra of POPE-d₃₁/CerC16 mixtures at 10°C displayed slightly more pronounced components at ±20 kHz than that of POPE/CerC16-d₃₁ mixtures (Figs. 2 and 3), suggesting a slightly faster rotational diffusion (52). This observation would be compatible with POPE involved in a more dynamic gel phase than CerC16. The two spectra showed, however, a single and similar component for the CD₃ signal; therefore, the methyl dynamics appeared to be not sufficiently different to be resolved in the spectra. A gel/gel phase coexistence region was also observed for DEPE/CerC16 (19), POPC/CerC16 (26), and N-palmitoyl-sphingomyelin (PSM)/CerC16 (53) phase diagrams. AFM experiments revealed the existence of two gel phases with distinct nanomechanical properties in PSM/CerC16 (53), and DPPC/CerC16 mixtures (54). Moreover, 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate, a lipophilic fluorescent probe, was found to partition differently between two gel phases.

Figure 4.

Given here is the proposed partial phase diagram for POPE/CerC16 system. The brown squares define both the onsets and ends of the endothermic transitions obtained by DSC. The plus signs indicate the solidus and the liquidus lines inferred from the spectral subtraction method introduced by Vist and Davis (55) The blue triangles indicate the solidus and liquidus lines derived from the lipid distributions obtained from the ²H-NMR spectral simulation of the POPE/CerC16 90:10 mirror samples. The blue diamonds define the frontiers of the L_α/H_II coexistence region inferred from the lipid distributions obtained from the sequential ²H-, and ³¹P-NMR spectra of the POPE/CerC16-d₃₁ 90:10 mixture. The blue circles define the frontiers of the L_β/H_II coexistence region inferred from the lipid distributions obtained from the sequential ²H-, and ³¹P-NMR spectra of POPE/CerC16-d₃₁ 85:15 and 80:20 mixtures. The phase frontiers inferred from the data, in the CerC16 proportion region that is not directly explored, are indicated by dotted lines. To see this figure in color, go online.

The DSC results indicated an upshift of the L_β − L_α transition temperature as a function of the CerC16 proportion in the lipid mixture, and the NMR results showed that both lipids experienced this upshift. The coexistence phase regions reflect well the asymmetrical shape of the endotherms in the DSC thermograms. The solidus (the frontier between the L_β, and the L_β/L_α regions), and the liquidus (the frontier between the L_β/L_α, and the L_α regions) lines could be determined from the ²H-NMR spectra, using the method introduced by Vist and Davis (55) and successfully applied to PSM/CerC16 mixtures (53). This subtraction method was applied using the ²H-NMR spectra obtained from the mixtures containing 10 and 20 mol % CerC16-d₃₁. The results indicated that between 30 and 50°C, L_α-phase domains containing ∼5 mol % CerC16 coexisted with L_β-phase domains that contained ∼30 mol % CerC16. The L_β/L_α phase distributions of POPE, and of CerC16 obtained from the two series of mirror-sample ²H-NMR spectra (Figs. 1 and 2; Figs. S1–S3) were put together to infer the composition of each phase, providing an alternative method to determine the solidus and the liquidus boundaries from a different data set. The agreement between the liquidus lines obtained from the spectral subtraction method (55) and from the phase distributions in the mirror samples was very good. The latter method determined that the gel-phase domains included ∼40 mol % CerC16, a value slightly higher than that inferred from the spectral subtraction method. This difference is likely within the experimental error associated with the uncertainties in the subtraction factors, the area determinations, and the sample compositions.

At 54–55°C, the coexistence of three phases (L_β/L_α/H_II) could be assessed from the NMR spectra. The L_β phase appeared to involve exclusively CerC16, whereas the L_α, and H_II phases were a mixture of the two lipids—it should be pointed out that the ²H-NMR spectra of hydrated CerC16-d₃₁ recorded above 50°C included a considerable narrow line (data not shown), suggesting that the gel phase in the POPE/CerC16 mixtures was somehow stabilized—perhaps by a small amount of POPE. Because the investigated lipid mixtures included two components (water was in excess in all the dispersions), a three-phase horizontal line must be included in the phase diagram. This line should meet with the solidus border of the L_β/L_α coexistence region but we do not have sufficient information to define precisely the meeting point. Moreover, the phase behavior of phospholipid bilayers containing higher CerC16 proportions is somehow controversial because of the presence of putative metastable phases (18, 26), adding to this uncertainty. The NMR results revealed that the nature of transition toward the H_II phase was deeply affected by the CerC16 content. The lipid mixtures underwent a transition from L_α phase to H_II phase when their CerC16 content was <∼10 mol %, whereas a L_β phase coexisting with the H_II phase was observed when the CerC16 content was 15 and 20 mol %. This finding led to the inclusion of an eutectic point where L_β, L_α, and H_II phases should coexist.

The compositions determined from the ²H- and ³¹P-NMR spectra (Figs. S2 and S3) defined the boundaries of the L_α/H_II, and L_β/H_II phase coexistence regions. There is a general good agreement between the DSC and NMR results. Small differences in transition temperatures were observed, a phenomenon that could be associated with the use of deuterated CerC16 in the case of the NMR samples (53, 55), with the fact that NMR data were (assumed to be) acquired at equilibrium, with stepwise temperature increments whereas DSC data were collected during continuous heating, and that the sample thermal history could not be identical. The main discrepancy between the results obtained by the two techniques was related to the L − H_II phase transition of the mixture with 20 mol % CerC16, where ²H-NMR results indicate a coexistence of the L_β and H_II phases over a wider temperature range. As previously pointed out, the NMR, and DSC techniques do not reproduce the very same conditions and the L_α–H_II phase transition is particularly sensitive to such variations.

This is, to the best of our knowledge, the first proposal of a partial phase diagram for POPE/CerC16 system. Some features should be compared with those previously proposed for systems including CerC16, and a phospholipid. The POPE/CerC16 phase diagram includes a relatively large L_β/L_α phase coexistence region. The fluid domains include ∼5 mol % CerC16, whereas the L_β-phase domains contain ∼35 mol %. The compositions of the gel-phase domains are consistent with those observed for the PSM/CerC16 system (53, 56). This similar behavior adds to the understanding of the driving force leading to the phase separation. It has been proposed that the capacity of ceramide headgroup to make H bonds, both as an acceptor and a donor, leads to strong interceramide interactions and promotes phase separation (53, 56, 57). POPE headgroup has also an extensive H-bond capability as a donor and an acceptor because of its quaternary ammonium. Despite the H-bond capacity of POPE, a phase separation was still observed in the POPE/CerC16 system. It is proposed that the absence of a phosphate group, reducing considerably the effective size of the headgroup (POPE molecular area: 0.74 nm² (58), CerC16 molecular area: 0.40 nm² (30, 59)), plays a major role in the phase separation, leading to a closer proximity of the molecules and therefore, stronger intermolecular interactions.

A distinctive part of the POPE/CerC16 phase diagram is the three-phase line that leads to L_β/H_II, and L_α/H_II phase coexistence regions. This feature is a consequence of the opposite effect of CerC16 on the transitions: it leads to the existence of the L_β phase at higher temperatures, and, at the same time, to the formation of H_II phase at lower temperatures. The overlap of the L_β − L_α and L_α − H_II phase transitions was previously observed for the DEPE/CerC16 system at a molar proportion between 10 and 20 (mol %) (19). It should be pointed out that for 10 mol % CerC16 and below, the H_II phase, coexisting with the L_α phase, was enriched in CerC16. It is inferred that its relatively small headgroup favored the formation of an inverted nonlamellar phase. Conversely, for the mixtures with >15 mol % CerC16, the H_II cylinders were depleted in CerC16 relative to the overall content. This behavior reflected the ordering effect of this lipid and its propensities to form the gel phase. This work clearly establishes the L_β − H_II phase transition for mixtures containing ≥15 mol % CerC16. Such a transition was mentioned for the DEPE/CerC16 system (19), and for partly dehydrated egg-PE (60). Recently it was shown that egg ceramide (40–60 mol %) in egg sphingomyelin would also lead to a L_β − H_II phase transition observed at high temperature (∼70°C) (61). In the POPE/CerC16 system, this transition could be associated with the fact that the increase in temperature leads to the disordering of CerC16 acyl chains, promoting the cone shape of the lipid, and to an enhanced solubilization of CerC16 in the existing H_II phase mainly formed by POPE.

The ²H-NMR spectroscopy of the mirror samples also provided the characterization of the dynamics of the lipid acyl chains in the POPE/CerC16 90:10 mixture. The smoothed order profiles of ceramide (derived from the mixture containing CerC16-d₃₁) as well as of POPE (derived from the mixture containing POPE-d₃₁) acyl chains are presented in Fig. 5 for the L_α and H_II phases.

Figure 5.

Shown here are the smoothed order profiles of POPE-d₃₁ and CerC16-d₃₁ in the L_α (left, at 42°C) and the H_II (right, at 65°C) phases. (Left) Here we show POPE/CerC16-d₃₁ 90:10 (blue triangles), POPE-d₃₁/CerC16 90:10 (red circles), and pure POPE-d₃₁ (black squares). (Right) Here we show POPE/CerC16-d₃₁ 80:20 (black stars), POPE/CerC16-d₃₁ 80:10 (blue triangles), and POPE-d₃₁/CerC16 90:10 (red circles). To see this figure in color, go online.

The L_α order profiles of both CerC16 and POPE, typical of the liquid-lamellar phase, consisted of a plateau associated with the carbons near to the lipid headgroup, followed by a sharp decrease of order toward the end of the chain. The profiles revealed that in the POPE/CerC16 90:10 mixture, the chain order parameters of CerC16 were noticeably higher than those of POPE. This observation is similar to the results obtained for POPC/CerC16 mixtures (26), where the palmitoyl chain of POPC was found less ordered than that of CerC16. An orientational order of the ceramide palmitoyl chain greater than that of PC was also deduced from a molecular dynamics simulation of the DMPC/CerC16 system (62). Therefore, even though the binary mixtures were likely homogeneous in the L_α phase, the palmitoyl chain of CerC16 and that of the phospholipid did not have the same orientational order. It has been proposed that the limited molecular area associated with CerC16 would leave a limited space for chain conformational disordering (26). Alternatively, it may be associated with the position of the headgroup relative to the bilayer normal. It has been recently shown (63) that the orientational order parameters of fatty acid chains in bilayers are modulated by the protonation state of the carboxylic group; deprotonated acid groups are more exposed to the aqueous environment and this location leads to higher chain order parameters. The small headgroup of CerC16 combined with its capability of making H bonds may position this lipid headgroup at a higher level than PE headgroups. This organization would lead, similar to the unprotonated fatty acid (63), to a larger overall orientational order.

The presence of ceramide appears to have a limited impact on the orientational order of the POPE acyl chain (Fig. 5), in contrast with its significant ordering effect on the POPC acyl chain (26). It should be highlighted that pure POPE L_α-phase bilayers exhibit already a relatively high chain order because of the limited headgroup size as well as their capacity to make H bonds between ethanolamine groups (49). The ordering effect of CerC16 on fluid phospholipid bilayers is likely limited when the bilayer existing order is considerable. In fact, it has been shown that cholesterol has a less pronounced ordering effect on POPE than on POPC fluid bilayers (64); it appears that CerC16 behaves similarly.

The acyl chain order profiles in the H_II phase have also been determined (Fig. 5 B). As expected (50), the order decrease along the chain was more linear than in the case of the L_α phase; this was observed for the chains of both POPE and CerC16. These ²H-NMR profiles are solid evidence that the ceramide molecules were inserted in the H_II cylinders. The ratios S(L_α)/S(H_II) were 2.7 close to the headgroup and were 4.0 near the end of the palmitoyl chain; these values were similar for the palmitoyl chain of POPE and of CerC16. These ratios were also consistent with those reported for pure POPE (50). This reduction is associated with the diffusion of the lipid molecules around the H_II cylinders, a motion that causes additional averaging of the quadrupolar interactions, as well as with the different lipid phase symmetry. Similar to the L_α phase, CerC16 displayed higher chain order parameters compared to those of POPE in the H_II phase formed by the POPE/CerC16 90:10 mixture. It is unlikely that the difference in chain order between POPE and CerC16 could be due to a phase separation within the H_II cylinders. A position of CerC16 headgroup closer to the H_II cylinder center than that of POPE headgroup could be the origin of the orientational order difference.

The proposed partial phase diagram of the POPE/CerC16 system shows that CerC16 has a compound impact: it can act as a versatile modulator, being able to promote L_β or H_II phase, and to lead to ceramide-enriched (coexisting with L_α phase) or ceramide-depleted (coexisting with L_β phase) H_II cylinders. Under some specific conditions (in the case of POPE/CerC16 mixture, at ∼10 mol % CerC16 and 54°C), these propensities are regulated by small changes in temperature and/or CerC16 proportion. It should be pointed out that these effects are reported to be buffered by cholesterol. For example, the formation of ceramide-rich gel-like domains was hampered in the presence of cholesterol (20, 65, 66). Similarly the gel-phase domains existing in phospholipid bilayers at low CerC16 contents (e.g., 5 mol % in this work) were not observed when 10 mol % CerC16 was added to red blood cell lipid extract. However, phase separations were observed when the CerC16 content was increased to 30 mol % or when the red blood cell lipid extract was cholesterol depleted (67). This being said, local fluctuations of the lipid compositions in biological membranes could lead to gel-phase domains such as those reported in this work. Both CerC16, and cholesterol (up to 30 mol %), (64) promote the formation of the H_II phase. At this point, their combined effect is not established but local variations of these species are bound to modulate the local curvature of membranes. It is well established that the order of a bilayer core and the polymorphic propensities of membranes have a pivotal role in controlling protein activity and consequently, many cellular processes (68, 69). As an example, it has been shown that Bax proteins are involved in the perturbation of the permeability of outer mitochondrial membranes, a phenomenon associated with apoptosis. The pore formation associated with these proteins is modulated by the curvature properties and the fluidity of the membrane (70, 71, 72). The variation of CerC16 content in membranes can be a way to control the membrane association, and the aggregation of Bax proteins and therefore could play a pivotal role in the modulation of the membrane permeability and apoptosis. The fact that CerC16 can modulate bilayer order, curvature propensities, domain formation, and the chemistry of the interface, suggests that it could act as a versatile messenger in cellular processes.

Author Contributions

M.L. designed the research, contributed to the data analysis, and to the redaction of the article. M.D. carried out the experiments, contributed to the data analysis, and to the redaction of the article.

Editor: Klaus Gawrisch.