Detection of Sphingomyelin Clusters by Raman Spectroscopy

Abstract

Sphingomyelin (SM) is a major sphingolipid in mammalian cells that forms specific lipid domains in combination with cholesterol (Chol). Using molecular-dynamics simulation and density functional theory calculation, we identified a characteristic Raman band of SM at ∼1643 cm⁻¹ as amide I of the SM cluster. Experimental results indicate that this band is sensitive to the hydration of SM and the presence of Chol. We showed that this amide I Raman band can be utilized to examine the membrane distribution of SM. Similarly to SM, ceramide phosphoethanolamine (CerPE) exhibited an amide I Raman band in almost the same region, although CerPE lacks three methyl groups in the phosphocholine moiety of SM. In contrast to SM, the amide I band of CerPE was not affected by Chol, suggesting the importance of the methyl groups of SM in the SM-Chol interaction.

Introduction

Sphingomyelin (SM) is a major sphingolipid in mammalian cells that is involved in membrane signal transduction through the formation of a second messenger (ceramide), as well as through the formation of specific lipid domains (1). The formation of SM-rich domains is affected by the surrounding lipids. Naturally occurring SM is enriched with saturated fatty acids and thus exhibits a high liquid crystalline-gel phase transition temperature (Tc). In other words, SM is rigid at physiological temperature. When egg SM (Tc = ∼41°C) is mixed with a fluid lipid such as 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), whose Tc is −22°C (2), egg SM and DOPC are phase separated, resulting in the formation of SM-rich domains. In contrast, egg SM is well mixed with 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC) (Tc = 41°C). Fluorescence microscopy is often employed to study the formation of lipid domains in model membranes, using giant unilamellar vesicles labeled with a fluorescent marker that partitions to specific membrane domains (3). Fluorescence spectroscopy is also used to examine the partitioning of fluorescent dye (4).

Fluorescence-based techniques require exogenous fluorescent dye to be present in the membrane, and thus the effect of the dye on the membrane cannot be excluded. Label-free methods are also employed to study the formation of membrane domains. Differential scanning calorimetry is frequently used to measure phase separation (5). However, because this approach measures the difference in Tc of two lipids, it cannot be applied to study the miscibility of two lipids with the same Tc, such as egg SM and DPPC. Atomic force microscopy is also a powerful tool for studying the miscibility of lipids (6). Because atomic force microscopy measures the height difference of two lipids, it is difficult to use this method to analyze the miscibility of two lipids of similar sizes without a special setup (7). Infrared (IR) and Raman spectroscopy provides information about the structure and dynamics of lipid molecules. Recently, Raman spectroscopy was used to visualize lipid domains containing SM analogs (8). However, Raman spectroscopy has not been applied to study lipid domains composed of naturally occurring SM.

The structure of SM is similar to that of the mammalian major phospholipid, phosphatidylcholine (PC). The primary characteristics of SM are the interfacial hydroxyl and amide residues of the molecule. These residues are capable of donating and accepting hydrogen bonds, whereas the carbonyl group of the N-acyl chain and the headgroup phosphodiester moiety may act as a hydrogen bond acceptor. By contrast, PC has only hydrogen bond acceptors at the interface.

Based on a molecular-dynamics (MD) study, Mombelli et al. (9) claimed that the sphingosine OH group is mainly involved in intramolecular hydrogen bonds of SM, in contrast to the almost exclusive intermolecular hydrogen bonds formed by the amide NH moiety. Thus, the amide vibration mode can provide information to distinguish between intra- and intermolecular hydrogen bonds. Fourier transform IR spectroscopy indicates that the addition of cholesterol (Chol) has a strong effect on the amide I band of SM (10, 11). In dried SM, the amide I band is split into two peaks. In the presence of Chol, the amide I band is still asymmetric, although the splitting is unresolved (11). It is speculated that the addition of SM-SM intermolecular interactions and the SM-Chol interaction increases the heterogeneity of the environment surrounding the amide group.

Raman spectroscopy has been applied to elucidate the membrane properties of SM (12, 13). However, the assignment of the Raman spectrum corresponding to intra- and intermolecular hydrogen bonds has not been established. In this study, we assigned the Raman peak responsible for the intermolecular hydrogen bonds of SM by means of Raman spectrum measurements and theoretical calculations. Our results indicate that the intermolecular hydrogen bonds of SM are sensitive to hydration, Chol, and the membrane distribution of SM. Our results also show that, in contrast to SM, the SM analog ceramide phosphoethanolamine (CerPE), which lacks three methyl groups in the phosphocholine moiety of SM, does not form a hydrogen bond with Chol, suggesting the importance of the methyl group of SM in the SM-Chol interaction.

Materials and Methods

Materials

We purchased SM (egg, chicken; egg SM; >80% of the amide-linked fatty acid of egg SM is palmitic acid according to the manufacturer), DPPC, DOPC, N-lauroyl-D-erythro-sphingosylphosphorylcholine (lauroyl SM), N-stearoyl-D-erythro-sphingosylphosphorylcholine (stearoyl SM), and N-palmitoyl-d31-D-erythro-sphingosylphosphorylcholine (d31 SM) from Avanti Polar Lipids (Alabaster, AL); Chol from Sigma (St. Louis, MO); and N-Acyl-sphingosylphosphorylethanolamine (CerPE) from Matreya LLC (Pleasant Gap, PA). 1′-¹³C-N-stearoyl-SM was prepared as described previously (14). Chloroform and methanol were purchased from Nacalai Tesque (Tokyo, Japan). Deuterium oxide (D₂O) was purchased from Acros Organics (Morris, NJ). Glass-bottomed dishes (3911-035) were obtained from AGC Techno Glass (Tokyo, Japan). Fig. 1 summarizes the structures of lipids employed in this study.

Figure 1.

Chemical structure of the lipids used in this study.

Preparation of lipid samples

Lipids were dissolved in chloroform, except that CerPE was dissolved in chloroform/methanol (2:1) and aliquoted to glass test tubes to prepare a lipid mixture. Solvent was evaporated under nitrogen flow and then in vacuo for 2 h. MilliQ water (Millipore, Billerica, MA) was added to the resultant lipid films (2 mM total lipids) and the test tubes were warmed up to 60°C and vigorously mixed until the solution turned turbid. The liposome solutions were centrifuged for 25 min at 16,000 × g. After removal of the supernatant, the concentrated suspensions were used for measurements. For the D₂O experiment, D₂O was used instead of H₂O. D₂O was added to the lipid film and the same procedure was followed. After centrifugation, the floated lipid suspensions were collected.

Fig. S1 in the Supporting Material shows a dark-field microscope image of the egg SM suspensions employed in this study. The suspensions exhibit different sizes and shapes, and seem to be multilamellar as deduced from the strong light scattering. Dynamic light scattering of the suspension solutions performed with a Zetasizer Nano S (Malvern Instruments, Malvern, UK) resulted in a high polydispersity (polydispersity index = 1–0.9) with a variety of liposome sizes, including a diameter greater than the detection limit of the apparatus (3 μm), consistent with the microscope image.

Raman data acquisition

Raman spectra were obtained with a laboratory-made laser Raman microscope based on an inverted microscope (IX71; Olympus, Tokyo, Japan). This microscope employed a 532 nm laser (Sapphire SF-532; Coherent, Santa Clara, CA) for excitation and an oil-immersion objective lens (UPLSAPO 60XO, NA = 1.35; Olympus). Through the objective lens, the excitation laser beam was focused on the sample. Then, the backscattered Raman radiation was collected with the same objective lens and passed through sharp-edge filters to block Rayleigh scattering and focused on the entrance slit of a spectrograph (SP2300; Princeton Instruments, Acton, MA). The signal light was dispersed by a 1200 lines/mm grating and finally detected by a thermoelectrically cooled charge-couple device (Pixsus 400BR-eXcelon; Princeton Instruments, Trenton, NJ) as a Raman spectrum. The sample suspension (2–10 μL) was deposited on a 35 mm glass-bottomed dish and placed on the sample stage of the microscope. All spectra were obtained at room temperature (23°C). The laser power at the sample and the exposure time were typically 10 mW and 60 s, respectively. All measurements were recorded with the charge-couple device software WinSpec (Princeton Instruments) and processed with MATLAB (The MathWorks, Natick, MA).

Simulation

Details of the computational procedure have been given elsewhere (15). An MD simulation of the SM bilayer was conducted using an all-atom force field for SM (16). After equilibration of a system consisting of 128 SM and 5012 water molecules, the trajectory was propagated for 150 ns with constant temperature (296 K) and pressure (1 bar). The temperature was controlled using the Langevin thermostat with a coupling time of 1 ps⁻¹. A semi-isotropic pressure coupling scheme was used, and the pressure was controlled using the Nosé-Hoover Langevin piston method (17, 18, 19) with the oscillation period time and the damping time set to 0.05 ps and 25 fs, respectively. A periodic boundary condition was used in all directions. The electrostatic interaction was treated using the smooth particle-mesh Ewald scheme (20, 21). The nonbonding interaction was decreased to zero between 10 Å and 12 Å by employing a switching function. The neighbor list was updated every 20 fs. The bond length involving hydrogen atoms was restrained by the SHAKE method (22) and the time step for integration was set to 2 fs. The MD calculation was performed using NAMD (23). We confirmed that the thickness and the area per lipid of the SM bilayer were kept stable during the simulation, and that the mean values (43.5 Å and 47.5 Å², respectively) were in good agreement with the experimental values (42.2 Å and 55 Å²) (24).

From the last 100 ns trajectory, we first searched for a cluster of SM and water molecules formed by hydrogen bonds. Then, we classified the clusters according to type (e.g., SM + water, SM + two waters, or SM dimer). For a cluster type with an existence weight greater than 1%, we selected a representative structure by using the k-means clustering algorithm, and performed a density function theory (DFT) calculation using the B3LYP exchange-correlation functionals (25, 26) and mixed basis sets of 6-31G(d,p) and 6-31++G(d,p) (27, 28). The latter basis set with diffuse functions was used for atoms involved in the hydrogen bonds (i.e., hydrogen, nitrogen, and oxygen atoms of the amino group and water). The combined basis set is denoted 6-31(++)G(d,p) in the following text. All of the DFT calculations were carried out using Gaussian 09 (29). Although the MD simulation based on the force field treated all atoms of the SM molecule explicitly, the DFT calculation was computationally much more intensive. Therefore, we used a fragment of the SM molecule in the DFT calculation, removing the headgroup and most of the acyl chains, and retaining only the atoms in the vicinity of the amide group and the C-C double bond that were essential for this study. The size of the fragment was carefully validated in a preliminary calculation (see Fig. S2 and Table S1, as well as data in (15)). The geometry was optimized starting from the MD snapshot of the clusters. Then, the harmonic frequency and the Raman activity were calculated for each cluster. The frequencies were corrected by

$${\omega }_{corr}=f\times \left(\omega +a\right),$$ (1)

where a = −16 and −3 cm⁻¹, and f = 0.976 and 0.981 for the C-C and C-O double-bond stretching modes, respectively. The shift parameter (a) accounts for the size of the fragment and basis set, and the scaling factor (f) incorporates the anharmonic effects. The scaling factors were derived from an anharmonic vibrational structure calculation on the SM monomer. The final Raman spectrum was obtained by a weight average of the spectrum of each cluster. The convoluted spectrum was obtained by using the Lorentz function with a full width at half-maximum of 15 cm⁻¹.

Results and Discussion

Hydration-dependent alteration of the Raman spectra of SM

Fig. 2 A shows the Raman spectra of egg SM suspensions at 23°C. The bottom and top spectra were measured 1 h and 25 h, respectively, after sample deposition on a glass-bottomed dish. The two spectra are almost identical except for the peak at 1643 cm⁻¹ (arrow). A spectrum similar to that of the 25 h sample was obtained previously (30). However, the origin of the 1643 cm⁻¹ band has not been well characterized. Tentative assignments of the main Raman bands of egg SM (after 1 h) are listed in Table 1. Fig. 2 B shows the changes in the Raman spectra of egg SM in the region of 1600–1720 cm⁻¹ from 1 to 48 h after sample deposition. A shoulder that was seen in the spectrum 1 h after sample deposition gradually evolved into a sharp band at ∼1643 cm⁻¹. Because the sample was left on the glass-bottomed dish throughout the measurement process, this change is most likely due to water evaporation. To confirm the influence of water, we added 2 μL of water to the same volume of 1-day-dried suspension (100 mM egg SM) and recorded the temporal changes in the Raman spectrum (Fig. 2 C). The band at 1643 cm⁻¹ that existed in the 1-day-dried sample became the shoulder within 15 min after the addition of water and reappeared as the band within 30 min as the water evaporated. Water evaporation in our experiment was confirmed by the OH stretching band at ∼3400 cm⁻¹ (Fig. S3). The band at ∼1643 cm⁻¹ did not depend on the excitation laser intensity, as shown in Fig. S4. Thus, the Raman analysis of egg SM revealed that its spectrum has a hydration-sensitive band at ∼1643 cm⁻¹. In addition to the 1643 cm⁻¹ band, the 830 and 850 cm⁻¹ bands also correlated with the water content, as shown in Fig. S5. The molecular origin of these bands has not yet been characterized. In this study, we primarily examined the 1643 cm⁻¹ band.

Figure 2.

Time variations of the Raman spectrum of egg SM suspension and a spectral comparison of different SM suspensions. All samples were deposited on the glass-bottomed dish and measured at 23°C. (A) Raman spectra collected 1 h (bottom) and 25 h after sample deposition (top). (B) Changes in Raman spectra of egg SM in the region between 1600 and 1720 cm⁻¹. Lipid suspensions were deposited on a glass-bottomed dish from 1 h to 48 h at 23°C (blue, 1 h; red, 6 h; green, 25 h; black, 48 h). Measurements were made on 10 μL of a 40 mM lipid suspension. (C) Changes in Raman spectra in the region between 1600 and 1720 cm⁻¹ after addition of water (2 μL) to a 1-day-dried lipid suspension: before addition of water (red), elapsed time t = 0 (blue), t = 15 min (green), and t = 30 min (black). Measurements were made on 2 μL of a 100 mM lipid suspension. (D) Raman spectra of (a) lauroyl SM, (b) egg SM prepared with D₂O, (c) d31 palmitoyl SM, (d) ¹³C stearoyl SM, (e) stearoyl SM, and (f) egg SM. (E) Raman spectra of the 1560–1720 cm⁻¹ region for (a) lauroyl SM, (b) egg SM with D₂O, (c) d31 palmitoyl SM, (d) ¹³C stearoyl SM, (e) stearoyl SM, and (f) egg SM. All spectra in (E) and (F) were recorded after the same time interval from sample deposition.

Table 1.

Assignment of the Main Raman Bands for Egg SM Shown in Fig. 2A

Raman shift (cm−1)	Assignments
1676	C=C stretching (12, 13)
1643	Amide I band (12)
1445	CH2 deformation (12, 30, 43)
1301	CH2 twist (13, 30)
1135	C-C stretching (trans) (30, 44, 45)
1097	C-C stretching (gauche) (30, 44, 45)
1067	C-C stretching (trans) (30, 44, 45)
961	CN asymetric stretching (46)
896	acyl C1-C2 streching (trans) (47)
881	acyl C1-C2 streching (gauche) + choline deformation (47)
769	O-P-O symmetric streching (45)
721	C-N symmetric stretching of the choline group (46, 48)

The origin of the Raman band of SM at ∼1643 cm⁻¹ is not established. Levin et al. (12) pointed out that the broad feature centered at ∼1644 cm⁻¹ reflects the amide I band associated with amide linkage between the acyl chain and amino group of sphingosine base. In contrast, Lamba et al. (13) claimed that a weak feature at ∼1644 cm⁻¹ was the HOH deformation mode of water molecules involved in the bilayer hydrogen-bonding network. In Fig. 2, D and E, the Raman spectra of different SM samples are compared. All of these spectra were recorded after the same time interval from sample deposition. Measuring the Raman spectrum of egg SM in D₂O instead of H₂O did not alter the 1643 cm⁻¹ band, indicating that the HOH deformation mode of water is not responsible for this band. SMs with different fatty acid compositions (lauroyl (C12:0) SM, stearoyl (C18:0) SM, and egg SM (mainly C16:0)) exhibited the 1643 cm⁻¹ band. This observation supports the idea that the acyl chain length does not affect the orientation and position of NH- and OH- groups at the lipid-water interface, and consequently the formation of intermolecular hydrogen bonds and the assignment of the 1643 cm⁻¹ band. D₃₁-palmitoyl SM did not alter the 1643 cm⁻¹ band, indicating that the fatty acid residue of SM is not involved in this band. Raman spectra were measured at 23°C. At this temperature, egg SM and stearoyl SM are in the gel phase, whereas lauroyl SM is in the liquid crystalline phase (31). Fig. 2 E indicates that the 1643 cm⁻¹ band was not affected by the phase state of SM. In contrast to stearoyl SM, 1′-¹³C-N-stearoyl-SM exhibited a Raman shift of the 1643 cm⁻¹ band to 1603 cm⁻¹. The wavenumber ratio of the ¹³C=O and ¹²C=O bands (1603/1643 = 0.976) is consistent with the value calculated from reduced masses (0.977), and the isotope shift was reported as ∼40 cm⁻¹ (32). This strongly supports the idea that 1643 cm⁻¹ band reflects the amide I band associated with amide linkage between the acyl chain and amino group of sphingosine base.

Simulation of the Raman band of SM at ∼1643 cm⁻¹

To further investigate the physical origin of this band, we conducted MD simulations and DFT calculations for the SM bilayer to obtain the weight of SM clusters in an isothermal-isobaric (NPT) ensemble formed by the intermolecular hydrogen bonds of the amide group. We then calculated the harmonic frequency and Raman activity for each cluster at the level of B3LYP/6-31(++)G(d,p) and corrected the frequencies according to Eq. 1. We note that both the level of the DFT calculation and the correction scheme were carefully examined to obtain the frequency with quantitative accuracy. For example, the C=O stretching frequency was obtained as 1774 cm⁻¹ for a single SM fragment at the B3LYP/6-31G(d,p) level. Adding the diffuse functions lowered the frequency by 30 cm⁻¹, and the correction mainly for the anharmonicity further lowered the frequency by 40 cm⁻¹. Consequently, the scheme used here yielded a C=O stretching frequency of 1706 cm⁻¹ for a single SM fragment. See Table S1 and (15) for more details.

Fig. 3 A plots the calculated Raman shift of the C=O stretching mode for isolated SM and SM clusters of the amide groups terminated by water molecules (denoted w(SM)_nw). The C=O stretching frequency is strongly dependent on the cluster type. The formation of hydrogen bonds between the amide group and water molecules induces a sizable red shift of the C=O stretching by 36 cm⁻¹. Furthermore, the extension of the hydrogen bond network (i.e., the amide chain) exhibits a red shift of 9.6 and 5.3 cm⁻¹ when the value of n is varied from 1 to 2 and from 2 to 3. The hydrogen bond of the amide groups gives rise to a band at ∼1650 cm⁻¹, as shown in Fig. 3 B. The band is mainly composed of the C=O stretching modes of wSMw and w(SM)₂w, although that of the larger cluster is hardly visible due to its small probability. The C=C stretching mode gives a rather strong band at 1685 cm⁻¹. The calculated spectrum is in good agreement with the observed spectrum 1 h after sample deposition (Fig. 2 B, 1 h, blue). Therefore, the two bands observed at 1672 and 1642 cm⁻¹ are assigned to the C=C stretching modes and the C=O stretching modes of the hydrogen-bonded amide group, respectively. The C=O stretching mode in the amide group predominantly contributes to the amide I band (33). Consequently, these results support the findings obtained using Raman spectroscopy. In essence, the characteristic shift in the position of the amide I band observed in the Raman spectra of SM reflects the degree of intermolecular hydrogen bonding of SM.

Figure 3.

(A) The Raman shift of the C=O stretching modes of w(SM)_nw (n = 1–3) clusters and isolated SM was calculated at the level of B3LYP with mixed 6-31G(d,p) and 6-31++G(d,p) basis sets, using the correction scheme in Eq. 1. The Raman band that exhibits the highest Raman activity is shown for n = 2 and 3. (B) Comparison of theoretical and experimental Raman spectra. The experimental spectrum is the same curve shown in Fig. 2B (1 h, blue).

Chol inhibits hydrogen bonding between SM molecules

A preferential interaction between SM and Chol was reported previously (34). This preference is caused not only by the interaction between the steroid ring of Chol and the saturated acyl chains of the SM but also by the hydrogen bond between the 3-hydroxyl group of Chol and the amide group of SM (34). A previous study using Fourier transform IR showed that the amide I band at ∼1650 cm⁻¹ in dried egg SM was altered by the presence of Chol (35).

We measured the effect of adding Chol to egg SM on the Raman band at ∼1643 cm⁻¹ (Fig. 4). The presence of Chol abolished the band at 1643 cm⁻¹ even in the mixture of SM/Chol = 1:0.2. It is noteworthy that above this concentration of Chol, the sharp gel-to-liquid crystalline transition of SM is no longer detectable in differential scanning calorimetry (36). Through Monte Carlo calculations using interaction energies deduced from calorimetric results, Snyder and Freire (37) previously showed that at ∼20% Chol, the Chol-rich area suddenly forms a network that extends over the entire SM bilayer. These results suggest that ∼20% Chol is sufficient to interact with whole SM molecules in the membrane, and interaction with Chol disrupts the intermolecular hydrogen bonding of SM molecules.

Figure 4.

Raman spectra in the region of 1600–1720 cm⁻¹ for egg SM (red) and four mixtures of egg SM/Chol (1:0.1 (blue), 1:0.2 (black dotted), 1:0.5 (black dashed), and 1:1 (black solid)). The Raman spectrum was collected 25 h after sample deposition at 23°C.

The amide I Raman band reflects the presence of SM clusters in the membrane

We next asked whether the characteristic 1643 cm⁻¹ band of SM derived from intermolecular hydrogen bonding could be utilized to study the membrane distribution of SM. Egg SM has a gel-to-liquid crystalline phase transition temperature (Tc) at ∼41°C and mixes well with DPPC (Tc = 41°C), but not with DOPC (Tc = −22°C) (2). Thus, in an egg SM/DOPC mixture, the lipids are phase separated and SM is present in the form of clusters (38). By contrast, palmitoylsphingomyelin is dispersed in DPPC (39).

Fig. 5 A shows the Raman spectra in the 600–1800 cm⁻¹ region for egg SM, egg SM/DOPC (1:1), and egg SM/DPPC, DOPC, and DPPC. The 1741 cm⁻¹ band in the DOPC and DPPC spectra was assigned to the C=O stretching mode of the saturated aliphatic ester group, and the 1659 cm⁻¹ band in the DOPC was assigned to the cis C=C stretching mode in the hydrocarbon chains. The amide I region of egg SM and the egg SM/DOPC (1:1) mixture is enlarged in Fig. 5 B, whereas its existence is unclear in the egg SM/DPPC (1:1) mixture. As shown in Fig. 2 B, egg SM shows the amide I (C=O stretching) band at 1644 cm⁻¹. The shoulder of amide I clearly exists in the egg SM/DOPC mixture, whereas its existence is unclear in the egg SM/DPPC mixture. The relative intensities of the egg SM trans C=C stretching band at ∼1675 cm⁻¹ and the amide I band at ∼1643 cm⁻¹ for egg SM/DOPC and egg SM/DPPC with different SM/PC ratios are shown in Fig. 5, C and D, respectively. Whereas the intensity of the trans C=C stretching band decreased linearly with a decreasing SM/PC ratio in both egg SM/DOPC and egg SM/DPPC suspensions, the decrease of the amide I band was much less sensitive to the SM concentration in egg SM/DOPC suspensions. By contrast, the addition of DPPC dramatically reduced the intensity of the amide I band of egg SM. These results indicate that the characteristic amide I Raman band reflects the clustering of SM.

Figure 5.

Raman spectra of various lipid suspensions. (A) Raman spectra of egg SM, egg SM/DOPC (1:1), egg SM/DPPC (1:1), DOPC, and DPPC. (B) Raman spectra of the 1620–1720 cm⁻¹ region for egg SM (red), egg SM/DOPC (1:1) (blue), and egg SM/DPPC (1:1) (black). (C) Concentration dependence of the relative intensities of the trans C=C stretching band (circle) and the amide I band (square) for a 1:1 mixture of egg SM and DOPC. (D) Concentration dependence of the relative intensities of the trans C=C stretching band (circle) and the amide I band (square) for a 1:1 mixture of egg SM and DPPC. The Raman spectrum was collected 25 h after sample deposition at 23°C.

The amide I band of CerPE is insensitive to Chol

The sphingolipid composition differs among different species. Although SM is the major sphingolipid in mammalian cells, the major sphingolipid in Drosophila is CerPE. In CerPE, the phosphocholine residue of SM is replaced with phosphoethanolamine, resulting in a stronger intermolecular interaction compared with SM, as indicated by its higher phase transition temperature (40).

CerPE shows almost the same Raman spectrum in the region of 1620–1700 cm⁻¹ as SM, as shown in Fig. 6. We next asked whether the amide I band of CerPE at ∼1641 cm⁻¹ is sensitive to the addition of other lipids. Similarly to SM, a CerPE/DOPC (1:1) mixture exhibited a shoulder at ∼1641 cm⁻¹, which was abolished in CerPE/DPPC (1:1). These results indicate that CerPE forms clusters in DOPC and is dispersed in DPPC. However, in contrast to egg SM/Chol, the CerPE/Chol (1:1) membrane clearly shows an amide I band. Fig. 6 shows the immiscibility of CerPE and Chol. This result is consistent with a previous observation that CerPE failed to form sterol-rich domains (40). A previous MD simulation (41) indicated that CerPE has more hydrogen bonds with Chol than with SM. However, the overall decrease in the number of hydrogen bonds is compensated for by an increasing number of charge pairs between SM and Chol, which leads to an increase in the total number of intermolecular interactions. In addition, the small headgroup of CerPE results in a strong CerPE-CerPE interaction that excludes Chol. Furthermore, the small headgroup of CerPE results in less effective shielding of the sterol molecule from unfavorable interactions with water.

Figure 6.

Raman spectra in the region of 1620–1700 cm⁻¹ for CerPE (red) and three mixtures of CerPE/DOPC (1:1) (blue), CerPE /DPPC (1:1) (green), and CerPE/Chol (1:1) (black). The Raman spectrum was collected 25 h after sample deposition at 23°C.

Drosophila contains very low levels of sterols compared with mammalian cells (42). Our results suggest that the sphingolipid domains in Drosophila and mammalian cells have different lipid compositions and physical properties.

In conclusion, we identified the Raman amide I band at ∼1643 cm⁻¹ of SM, which reflects the intermolecular hydrogen bond of SM clusters. The amide I band is sensitive to the hydration of lipids and the presence of Chol. We showed that this Raman band can be utilized as a simple marker to examine the membrane distribution of SM in model systems. However, the application of this method in cell systems may be hindered by the amide I band of proteins. We also showed that the interaction of SM and Chol is dependent on the methyl groups of the phosphocholine moiety of the molecule.

Author Contributions

T.K. conceived and coordinated the project. K.S. performed Raman spectroscopy and data analysis. T.I. prepared lipids for measurement. K.Y., P-C.L., and Y.S. performed the simulation. M.M. provided the 1′-¹³C-N-stearoyl-SM. K.S., K.Y., Y.S., and T.K. wrote the manuscript.

Editor: Arne Gericke.