Enhanced Ordering in Monolayers Containing Glycosphingolipids: Impact of Carbohydrate Structure

Abstract

The influence of carbohydrate structure on the ordering of glycosphingolipids (GSLs) and surrounding phospholipids was investigated in monolayers at the air-water interface. Binary mixtures composed of GSLs, chosen to span a range of carbohydrate complexity, and zwitterionic dipalmitoylphosphatidylcholine phospholipid, were studied. X-ray reflectivity was used to measure the out-of-plane structure of the monolayers and characterize the extension and conformation of the GSL carbohydrates. Using synchrotron grazing incidence x-ray diffraction, the in-plane packing of the lipid acyl chains and the area per molecule within ordered domains were characterized at different mole ratios of the two components. Our findings indicate that GSL-containing mixtures, regardless of the carbohydrate size, enhance the ordering of the surrounding lipids, resulting in a larger fraction of ordered phase of the monolayer and greater dimensions of the ordered domains. Reduction of the averaged area per molecule within the ordered domains was also observed but only in the cases where there was a size mismatch between the phospholipid headgroups and GSL components, suggesting that the condensation mechanism involves the relief of steric interactions between headgroups in mixtures.

Introduction

Glycosphingolipids (GSLs) are present in the extracellular leaflet of most animal cellular plasma membranes and are thought to regulate various physiological events at the cell surface. Gangliosides, glycosphingolipids that contain one or more negatively charged sialic acid groups attached to hydrophilic sugar moieties, are the most structurally complex form of glycolipids. The oligosaccharide chain of glycosphingolipids is variable because of the sugar content, sequence, and connections leading to a large amount of variability in GSL structure. Although a minor component in most cells, GSL constitute 5–10% of the total lipid mass in nerve cells (1), and because they reside primarily on the outer leaflet of the cell membrane, the external surfaces of certain cells contain 10–20 mol % GSL. Glycosphingolipids are thought to play physiological roles in a number of cellular functions that can be linked to assembling properties of GSL on cell surfaces, including mediators of adhesion (2, 3, 4), modulators of signal transduction (4), and as receptors for microbial toxins (5).

The lipid-raft hypothesis proposes naturally occurring lipids such as sphingolipids, cholesterol, and glycosphingolipids specifically aggregate in the plane of the cell membrane, driven primarily by the saturated hydrocarbon chains of the sphingolipids that allow cholesterol molecules to be tightly intercalated (6). Although the small size and dynamic nature of lipid rafts makes them difficult to probe directly in intact membranes, these ordered regions are implicated in biological processes ranging from signal transduction to viral entry to lateral transport of proteins (6, 7). This establishes a need to determine how individual lipid components physically and chemically interact with and affect each other to understand both structure and dynamics within cell membranes. Unique structural properties of GSL suggest a strong tendency to form ordered regions enriched in GSL in phospholipid bilayers; these include the geometry of the monomer within the membrane, the capability of the hydrophilic oligosaccharide chain to interact with neighboring water molecules, the hydrogen bonding ability of the amide linkage of ceramide to form an ordered network at the membrane surface, and the sphingosine Δ⁴ double bond stabilizing a parallel orientation of the carbohydrate moiety with the hydrocarbon chains near the lipid-water interface (8). Conclusions about the lateral organization of various GSLs and their interactions with neighboring phospholipids within model membrane systems have been inconsistent, dependent on the type of GSL under study, with experimental methods utilized ranging from NMR to DSC, and environmental conditions such as pH and electrolyte concentrations (9, 10, 11, 12, 13, 14, 15, 16, 17), prompting the need for more basic and systematic comparisons.

The most biologically prevalent GSL in caveolae and lipid rafts is monosialoganglioside GM1 (see Fig. 1), a sphingolipid that contains a headgroup composed of four sugar groups and one sialic acid. The chemical and structural properties of GM1 and other related GSLs have been reviewed extensively (18, 19). Pure GM1 monolayers formed at the air-water interface at 30°C are fluid at all packing densities and surface pressures as determined by surface pressure versus molecular area isotherms, fluorescence microscopy, and grazing incidence x-ray diffraction (GIXD) (19, 20). When added to zwitterionic DPPC, GM1 condenses (reduces the average area per molecule (APM)) the monolayer at low concentrations (<25 mol %), whereas at higher concentration (>25 mol %) it fluidizes the layer (20). This condensing effect, where the binary mixture is more close-packed than either component alone, has been attributed to the headgroup geometries of the two molecules with wedge-shaped GM1 forming a pocket between them into which the DPPC headgroup can fit, enabling the hydrocarbon tails of the two molecules to align and condense (20). The driving force for this condensation effect is likely intermolecular hydrogen bonding between GSL sugar groups and their associated water molecules. In the case of ganglioside-phospholipid condensation, the dipole moment within the DPPC headgroup may align with the negatively charged sialic acid residues to counteract the electrostatic repulsion (19, 21). These results are echoed in several other monolayer (22, 23) and bilayer studies (24, 25). Differential scanning calorimetry experiments performed on DPPC bilayers indicated that inclusion of GM1 increased the gel-to-liquid condensed phase transition temperature, consistent with condensation, and the enthalpic value of this transition peaked at ∼30 mol % GM1 (24). Molecular dynamics simulations of DPPC bilayers containing up to 22 mol % GM1 show decreases in surface area and increases in the hydrocarbon chain order parameter, indicating condensation (25). This observed condensing differs from effects reported for DPPC:GM1 monolayers deposited on a 145 mM NaCl, pH 5.6 subphase where lateral expansion is seen (15).

Figure 1.

Chemical structures of DPPC, CM, GA1, and GM1 molecules used in the studies.

In contrast to GM1, monolayers composed of pure asialoganglioside (GA1) lipids (GM1 without the sialic acid residue; Fig. 1) form condensed domains upon compression due to a headgroup with a cross-sectional area that is more commensurate with the area occupied by the hydrophobic tail region, allowing the molecules to pack well together. This has been observed with fluorescence microscopy at the air-water interface and atomic force microscopy in deposited layers (26). Although pure GA1 monolayers form condensed domains, the addition of DPPC to the monolayer further condenses the layer compared to either individual component (26). Parallel surface pressure-area (π-A) isotherm experiments with ceramide (CM) (sphingolipid backbone without the carbohydrate moiety, Fig. 1) indicate that in the absence of the bulky carbohydrate moiety, the hydrogen bonding network in the headgroup region results in a fully condensed pure CM monolayer, even at lift-off pressures. At low surface pressures, CM condenses DPPC at all binary ratios, but at a physiologically relevant π-value of 30 mN/m, additivity plots indicate that DPPC and CM mix ideally (20). Surface pressure measurements in monolayers of DPPC:CM3 (structurally related to CM shown in Fig. 1 with water added across the double bond) have shown a condensation effect with a negative deviation in average area per molecule compared to ideality (27). In lipid bilayers, fluorescence probe techniques show that the addition of CM to DPPC resulted in the ordering of the phospholipid acyl chain region with the magnitude of the change dependent on the CM composition within the bilayer (28).

The main motivation of this work is to further understand the influence of GSL on the lateral ordering and structure of surrounding phospholipids and vice versa, and to determine if the previously presented characterization of DPPC:GSL systems can be clarified using synchrotron x-ray scattering techniques. In this work, x-ray reflectivity (XR) and GIXD were used to investigate binary mixture monolayers composed of zwitterionic DPPC and GM1 at a biologically relevant π of 30 mN/m. Here, monolayers with lipids in an ordered, condensed phase at the air-water interface were used as a convenient model system to investigate the packing behavior and interactions between phospholipids and glycosphingolipids. Studying lipid packing in the condensed phase allowed detailed structural information to be obtained and can be considered as a model for lipid packing in the more ordered regions (e.g., lipid rafts) of otherwise mostly disordered biological membranes. It should also be noted that in addition to in-plane interactions between lipids in monolayers, cross-leaflet interactions between molecules in bilayer membranes may also influence lipid packing, and that these effects cannot be captured in a monolayer model (29). To determine how a particular molecular composition of the GM1 ganglioside molecule that gives rise to its unique mixing behavior with DPPC monolayers, mixtures of DPPC with either GA1 or CM were also analyzed. Lipid monolayers, used to model one leaflet of the cell membrane, provide a versatile model system that permits application of surface-sensitive scattering techniques and external control of surface pressure. Using monolayers enabled us to examine a full range of GSL concentrations from 0 to 100 mol % and amplify any weak GSL-phospholipid interactions. Although mixtures with >20% GSL will not be found in the cell membrane, the higher GSL content used in our study allow us to test and establish a molecular model that is applicable to systems at lower, more physiologically relevant concentrations. XR was used to measure the out-of-plane structure of the monolayers and characterize the extension and conformation of the GSL carbohydrate headgroup region. Using GIXD, the in-plane packing of the acyl chains and the APM of the ordered regions of the monolayers were characterized at different ratios of the two components. Here, we show that monolayers containing either GA1 or GM1 glycolipids enhance lipid ordering, resulting in larger fractions of the monolayer in the ordered phase and larger dimensions of the ordered domains. However, based on GIXD from the ordered phases, mixtures containing GA1 obeyed ideal mixing behavior, whereas GM1 containing mixtures exhibited a condensing effect on the average lipid APM.

Materials and Methods

Materials

1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC), ganglioside GM1 (brain, ovine sodium salt; the acyl chains on the d18:1 sphingosine base are predominantly 18:0 and 20:0), and n-stearoyl-d-erythro-sphingosine (18:0 CM) were obtained in powder form from Avanti Polar Lipids (Alabaster, AL) and used without further purification. Asialoganglioside GA1 (18:0) was prepared as previously described (30). Monolayer spreading solutions were prepared by dissolving in either just chloroform (HPLC grade; Thermo Fisher Scientific, Pittsburgh, PA) (e.g., DPPC) or chloroform containing 10% methanol (e.g., binary mixtures of DPPC and GA1 or GM1) at a concentration of 0.2 mg/mL. Lipid solutions were stored at −20°C in glass vials. For all Langmuir trough experiments, the subphase was ultrapure water (resistivity of 18.2 MΩcm) processed by a Milli-Q ultrapurification system (A-10 gradient; Millipore, Bedford, MA).

X-ray scattering measurements

Synchrotron XR and GIXD measurements were performed at the BW1 beamline at HASYLAB (Hamburg, Germany) using a beryllium monochromator to obtain x-rays with wavelength λ = 1.30 Å. By tilting the monochromator crystal, the beam was deflected to change the angle of incidence on the sample. X-ray scattering theory and the liquid diffractometer used here have been described previously (31, 32, 33). In XR experiments, an NaI scintillation point detector was used. In GIXD experiments, an evanescent wave was generated by an incident beam striking the water surface at an angle corresponding to a scattering vector in the z direction, q_z = 0.85·q_c, where q_c = 0.02176 Å⁻¹ is the critical scattering vector for total external reflection from the subphase. Diffracted intensities for GIXD experiments were measured using a vertically oriented 1D position-sensitive detector with an acceptance of 0 < q_z < 0.9 Å⁻¹. A Soller collimator mounted in front of the position-sensitive detector provided a horizontal resolution of Δq_xy = 0.0084 Å⁻¹ and the detector was scanned over 2θ_xy, the angle between the incident and diffracted beam projected onto the horizontal plane, to yield a q_z versus q_xy intensity distribution. A temperature-controlled (T = 30°C) Langmuir trough (460 cm² maximal working surface area and a 240 mL subphase volume, with a surface area/volume ratio of 1.92), equipped with a Wilhelmy balance and a barrier for surface pressure control, was mounted on the diffractometer. A smaller trough at the University of Chicago (145 cm² maximal working surface area and 70 mL subphase volume, with a similar surface area/volume ratio of 2.07) was used for isotherm and comparative APM measurements. The trough was enclosed in a sealed and thermostated canister flushed with helium to achieve an oxygen level below 1%; this reduced the scattering background and minimized oxidative beam damage during x-ray scans. The monolayer was compressed to the desired π before XR and GIXD were carried out to obtain out-of-plane and in-plane structural information about the sample. Our current studies as well as our previous work (20, 26, 34) and that of others (14, 35, 36) indicate that the glycosphingolipids, GM1 and GA1, form stable monolayers at the air-water interface at various surface pressures (up to 45 mN/m). Such monolayers can be kept unchanging for several hours with minimal apparent decrease in the number of molecules at the interface. As the lipid monolayer components used in this study have different reported solubilities (critical micelle concentration: DPPC – 4.6 × 10⁻¹⁰ M (37); GM1 2 × 10⁻⁸ M (38)), the area per molecule calculated from the amount spread at the interface can potentially overestimate the actual area per molecule. We resorted to XR to obtain a quantitative handle on the stability of the film. XR was used to determine the electron density contributions originating from the GSL carbohydrates (analysis fully described in the Supporting Material) and confirmed that the stable monolayers spread from 10 to 25% glycosphingolipid mixtures were within ∼5% of the nominal composition. Throughout this work, we assume a ±5% uncertainty in all monolayer compositions studied. As a precaution against beam damage, the trough was translated by 0.025 mm horizontally across the x-ray beam, in the direction along the barrier compression at every step of the q_xy scan and for XR; the sample was completely renewed by occasional 2 mm translation. The dimensions of the incoming x-ray beam footprint on the liquid surface were ∼2 × 50 mm².

X-ray reflectivity

XR yields detailed information on the electron density distribution normal to the interface, ρ_e⁻(z), laterally averaged over both the ordered and disordered parts of the film. The reflectivity is defined as the ratio of reflected to incident beam intensities, in a specular geometry, as a function of the vertical momentum transfer vector q_z = 4π sin(θ)/λ, where θ is the incident angle of the x-ray on the surface. Intensities were collected over the range 0.01 < q_z < 0.75 Å⁻¹, background subtracted, and normalized to incident beam flux. Data presented are divided by the Fresnel reflectivity and with error bars representing 1 SD error for each data point. The reflectivity curve can be analyzed to obtain the in-plane averaged electron density distribution normal to the interface. To accomplish this, a model-free approach based on cubic B-splines was used to obtain the electron density profile normal to the interface (39, 40, 41). We performed several hundred refinements within the parameter space and present a family of models for each reflectivity data set, all of which satisfy the χ² ≤ χ²_min + 1 criterion (42). The superposition of the profiles, after excluding nonphysical oscillating ones, yielded a broad electron density ribbon, which is a measure of the uncertainty in the real space structure (42).

Grazing incidence x-ray diffraction

Whereas XR measurements are sensitive to the average electron density distribution normal to the interface, GIXD yields precise in-plane packing properties and, in the case of lipid acyl chains, information on their molecular tilts, in-plane arrangements, average area per molecule, and the size of the ordered domains. The monolayers at the air-water interface can be considered as azimuthally disoriented domains analogous to 2D powders. Using an evanescent wave to amplify surface sensitivity by confining the beam to an interfacial layer of several nanometers, GIXD occurs when the 2D ordering of the molecules within the film satisfies the Bragg condition: nλ = 2dsinθ_xy. Diffracted intensities were recorded as a function of both q_z = 2π sin(α_f)/λ where α_f is the out-of-plane angle of the diffracted beam and q_xy ∼ 4π sin(θ_xy)/λ (33). After background subtraction, the data were integrated over q_z to yield intensity Bragg peak profiles. Pseudo-Voigt functions were used to fit the Bragg peaks. Peak positions, identified by the q_xy at the maximum of the Bragg peak, correspond to the d-spacings, d = 2π/q_xy, in the 2D lattice, and were used to obtain the unit cell parameters associated with the lipid chain packing (43).

Results

X-ray reflectivity

Out-of-plane structures of lipid monolayers composed of DPPC, CM, GA1, and GM1 (Fig. 1) were characterized at a π-value of 30 mN/m using XR. Single component monolayers composed of these molecules as well as binary mixtures (DPPC:CM, DPPC:GA1, and DPPC:GM1) with various compositions (mol %) were studied. Fitting of XR data (see representative data sets in Fig. 2, A and B) yields the in-plane averaged electron density distribution perpendicular to the air-water interface, ρ_e⁻ (Fig. 2, C and D). For a pure DPPC monolayer, the center of the PC headgroups corresponds to a ρ_e⁻ maximum, which was set to z = 0 (Fig. 2 C). The shoulder to the left of this maximum corresponds to the hydrocarbon tails (z ∼ −15 Å) that extends into the air where ρ_e⁻ reaches zero and the region to the right where the distribution reaches a plateau corresponds to the ρ_e⁻ of water (z > 15 Å). Monolayers of CM and DPPC:CM mixtures exhibited ρ_e⁻ distributions very similar to that of pure DPPC (Fig. S1). The largest structural differences from pure DPPC were observed for monolayers that contained the glycosphingolipids GA1 or GM1 where broader high ρ_e⁻ regions corresponding to both the DPPC and GSL headgroups and the GSL carbohydrate moieties were observed (Fig. 2, C and D). Due to the difference in reported solubility of the lipid monolayer components and questions about the equilibrated lipid composition, an analysis was performed on the electron density profiles obtained by XR for each of the pure components and several binary mixtures at different mol % to determine the electron density contributions from the GSL carbohydrates (detailed description included in the Supporting Material). These electron density values can be converted to GSL concentrations in the monolayer and indicate that the GSL concentration for the mixtures was within approximately ±5% of the spreading solution and therefore all reported values of the GSL ratio within the monolayer are ±5%.

Figure 2.

Out-of-plane structures of representative DPPC:GM1 and DPPC:GA1 monolayers at 30 mN/m. (A) XR data and fits for 75:25 DPPC:GM1 and 75:25 DPPC:GA1 mixtures and (B) 25:75 DPPC:GM1 and 25:75 DPPC:GA1 mixtures are compared to pure DPPC and pure GM1. Data (symbols) and fits (shaded regions) are divided by the Fresnel reflectivity and offset vertically for clarity. Electron density distributions matching the fits are shown in (C) and (D) as pairs of lines defining the boundary of accepted fits. The center of the headgroups is located at z = 0, below which is the tail region (z < 0) and above which is the headgroup region (z > 0). To see this figure in color, go online.

For 25:75 DPPC:GSL mixtures (Fig. 2 D), significantly more ρ_e⁻ associated with the carbohydrate headgroups was observed and the GA1 and GM1 distributions were similar to that of a pure GM1 monolayer. On the other hand, ρ_e⁻ of the hydrocarbon tails differed depending on whether the monolayer contained GA1 or GM1 molecules. The pure GM1 monolayer exhibited the smallest extension of the tails. The tail extension in the 25:75 mixed monolayer containing GM1 was intermediate between the pure DPPC and pure GM1 monolayers. This suggests that for GM1 at this mole ratio, the lipid packing is strongly influenced by the steric interactions between the larger carbohydrate residues. However, for GA1, due to its smaller carbohydrate headgroup, the tail extension matches that of DPPC. Additionally, the carbohydrate regions of GA1 and GM1 have similar overall ρ_e⁻ values despite GA1 containing one less carbohydrate residue. This result suggests that, at this mole ratio, there is tight packing between both the tails and carbohydrates of GA1.

Grazing incidence x-ray diffraction

In-plane packing of lipid hydrocarbon chains within phospholipid monolayers composed of DPPC, CM, GA1, and GM1 was investigated using GIXD. The scattering signal originated only from lipid tails in well-ordered, crystalline domains of the monolayer and was indexed to obtain the lattice parameters of the 2D unit cells (Table S1). Because all studied molecules have two hydrocarbon tails, the scattering from the crystalline packing of the headgroups will correspond to larger d-spacings and thereby be detected at smaller q_xy values. Scattering from the molecule’s head and carbohydrate groups were not observed. Fig. 3 shows representative Bragg peaks and their fits. Fig. 3 A shows diffraction from various DPPC:CM mixtures at π = 30 mN/m. Three peaks can be distinguished in the diffraction from the pure DPPC monolayer corresponding to packing of hydrocarbon chains in an oblique 2D unit cell. The tails are tilted away from the surface normal consistent with previous measurements (29, 44, 45). Diffraction from the pure CM monolayer exhibited a single sharp peak indicative of untilted hydrocarbon tails packed in a hexagonal 2D unit cell, also consistent with previous measurements under similar conditions (46). Features indicating phase separation of the ordered components were observed for the DPPC and CM mixtures. For example, the 85:15 DPPC:CM monolayer, a CM-rich phase, yielded a peak at q_xy ∼ 1.5 Å⁻¹ and a DPPC-rich phase yielded two distinct peaks at q_xy ∼ 1.41 and 1.47 Å⁻¹, reminiscent of the pure DPPC diffraction. The CM-rich peak persisted at approximately the same q_xy position in binary mixtures from 15 to 50% CM.

Figure 3.

GIXD Bragg peak profiles for varying compositions of binary (A) DPPC:CM, (B) DPPC:GA1, and (C) DPPC:GM1 monolayers. Solid (red online) lines show fits to the data with dotted (red online) lines corresponding to individual diffraction peak contributions to the total calculated fit. All data were collected using the same incident flux so absolute intensities between measurements may be compared. Data sets are shifted vertically, and intensity scales of each panel vary for clarity. Error bars based on Poisson counting statistics are shown and are typically comparable to the size of the data symbols. To see this figure in color, go online.

Similar phase separation features were also observed in the diffraction from the 70:30 DPPC:CM mixture. In the case of the 50:50 mixture, the asymmetric Bragg peak shape also suggests the presence of a small amount of a DPPC-rich phase although the low intensity prevented adequate fitting of this contribution. For all DPPC:CM mixtures, the small shift of the peak corresponding to the CM-rich phase to a slightly lower q_xy position compared to the pure CM peak position represents an increased APM and indicates that DPPC is also incorporated within the CM-rich phase. The single diffraction peak at 50% CM indicates the CM-rich phase can accommodate DPPC up to a 1:1 CM:DPPC ratio. Similarly, compared to pure DPPC, the upshift of the q_xy positions of the remaining peaks in the DPPC:CM diffraction pattern indicate the presence of CM in the DPPC-rich phases. On the other hand, no corresponding diffraction signatures indicating a similar ordered-ordered phase separation were observed in the mixtures of GA1 or GM1 with DPPC (Fig. 3, B and C), although the existence of such lipid segregation cannot be ruled out.

Diffraction from binary DPPC:GA1 monolayers at π = 30 mN/m is shown in Fig. 3 B. For all mole ratios, the presence of three peaks indicates the tilted lipid tails arranged in oblique 2D unit cells. Compared to the pure lipid components, GA1 in binary mixtures containing a majority of DPPC exhibits sharper Bragg peaks and higher diffracted intensities, corresponding to an enhanced ordering of the lipids in the monolayer. Similar behavior was observed for DPPC:GM1-mixed monolayers containing between 10 and 25 mol % GM1 (Fig. 3 C). In contrast to GA1, the pure GM1 monolayer did not possess sufficient in-plane ordering of the lipid tails to diffract. Further, only very weak diffraction signals consisting of a single broad peak were observed for mixtures containing 70 mol % or more GM1. The areas underneath the measured Bragg peaks were integrated to quantitatively compare the strength of the diffracted signals from the different composition monolayers (Fig. 4). These integrated intensities of the peaks correlate with the fraction of in-plane ordered domains within the monolayer and to the degree of lipid ordering (i.e., the length of the hydrocarbon chain that is sufficiently ordered to contribute to the diffraction signal) in the domains. Within the range of investigated mole ratios, monolayers of DPPC:GM1 and DPPC:GA1 mixtures are more highly ordered than either of the pure components.

Figure 4.

Integrated diffracted intensity of GIXD from monolayers of DPPC:CM (diamonds; black online), DPPC:GA1 (circles; blue online), and DPPC:GM1 (squares; red online) at 30 mN/m and 30°C. The integrated intensity is correlated with the degree of lipid chain ordering in the monolayer and shows a maximum at ∼40 mol % GA1 and 10 mol % GM1. Dashed lines are presented as a visual guide. Error bars are based on Poisson counting statistics. To see this figure in color, go online.

Fitting pseudo-Voigt functions to the Bragg peak profiles enabled the in-plane coherence lengths (L_C) of the ordered domains to be determined (47). The coherence length provides a measure of the distance over which alkyl tails are in positional registry along particular (h,k) crystallographic directions. Fig. 5 shows L_C values, corresponding to the average in-plane diameter of the ordered domains, as a function of monolayer composition. Values of L_C were estimated using the resolution-corrected average of the full width at half-maximum of the diffraction peaks obtained from fitting and the Scherrer equation: L_C ∼ 0.9[2π/FWHM (q_xy)] (47). The lipid tail packing within CM and CM-rich phases possessed significantly longer-range in-plane order than DPPC or any DPPC:GSL mixtures. Although L_C of the CM-rich phase decreased with increasing DPPC content in the monolayer, they remained more than twice as large as those of the coexisting DPPC-rich phases. Increasing CM content also appears to increase the L_C of the DPPC-rich phase. However, this trend is not outside of the experimental error due to limited horizontal resolution of Δq_xy and uncertainty in extracting the individual Bragg peak parameters for the two coexisting phases. Single component monolayers of both DPPC and GA1 exhibited domains with average in-plane coherence lengths of ∼100 Å. In the case of DPPC:GA1 mixtures, L_C values larger than either of the pure components were observed for mixtures containing between 10 and 25 mol % of GA1. This increase in L_C values corresponds to longer-range lateral ordering of the lipid tails. L_C values for pure GM1 and DPPC:GM1 monolayers containing >80 mol % of GM1 could not be extracted due to the lack of diffracted signal. Between 70 and 80 mol % of GM1, broad diffraction peaks were measured, corresponding to extremely small ordered domains (L_C ∼ 25 Å) within the monolayers. However, for lower GM1 concentrations, ≤60%, the in-plane coherence length increased to values greater than the L_C of the pure DPPC monolayer, indicating that within these concentration ranges GM1 enhances the lateral extent of lipid ordering.

Figure 5.

In-plane coherence length of ordered domains within monolayers of DPPC:CM (triangles; black online), DPPC:GA1 (circles; blue online), and DPPC:GM1 (squares; red online). Domain sizes larger than the pure components are observed for DPPC:GA1 and DPPC:GM1 mixtures. Dashed lines are presented as a visual guide. Error bars are based on χ² + 1 change from Bragg peak fit parameter variation. To see this figure in color, go online.

To compare the average APM in the entire monolayer at 30 mN/m as obtained for π-A measurements to the APM in ordered regions of the monolayer (Fig. 6), the Bragg peak positions were fit and indexed to the relevant 2D unit cell based on the number of observed peaks: hexagonal (one peak), distorted hexagonal (two peaks), and oblique (three peaks). In the 85:15 and 70:30 DPPC:CM mixtures, due to the coexistence of two condensed phases with different DPPC:CM ratios, a combination of 2D unit cells originating from a hexagonal and a distorted hexagonal phase were used to index the diffraction peaks. A lipid APM (area per two hydrocarbon chains) of 40.5 Å² was obtained for the CM monolayer, which is only slightly larger than twice the 19 Å² APM for the dense packing of alkyl chains (48). This indicates that the CM polar headgroup has only a minor impact on the in-plane lipid packing. In contrast, the packing of DPPC lipids, with an APM of 46.5 Å², is significantly more affected by the steric interactions of the bulkier PC headgroups resulting in tilted hydrocarbon chains and an oblique 2D unit cell (45, 49). Due to the phase coexistence in the DPPC:CM mixtures, the packing within the CM-rich phase and the DPPC-rich phase was determined separately as both phases were sufficiently ordered to diffract. Increased overall DPPC content in the layer resulted in larger APM within both the CM-rich and DPPC-rich phases. The measured APM of the DPPC:CM mixtures were compared to ideal mixing predictions using either the APM of the pure components within ordered domains obtained from GIXD (Fig. 6 C) or the average APM of the monolayer obtained from isotherms (20, 26) (Fig. 6 D). In the case of DPPC:CM mixtures, the APM obtained by both methods were similar, indicating that the lipid packing within the majority of the monolayer was sufficiently well ordered to diffract. However, these APM values were significantly smaller than predicted values based on ideal mixing behavior of the two pure components, indicating a condensation effect. The deviation of APM from ideal mixing was most prominent in the CM-rich phase. This suggests that the dispersion of DPPC within the CM matrix (or vice versa) relieves the steric repulsions from neighboring PC headgroups permitting tighter molecular packing.

Figure 6.

(A) Area per molecule within ordered domains obtained from GIXD and (B) average area per molecule obtained from isotherms for lipid packing in monolayers of DPPC:CM (diamonds and triangles; black online), DPPC:GA1 (circles; blue online), and DPPC:GM1 (squares; red online) at 30 mN/m (20, 26). Downward pointing triangles correspond to distorted hexagonal phases and upward pointing triangles to hexagonal phases within the DPPC:CM monolayers. Deviation of the ordered domain APM from ideal mixing is shown (C) based on the APM of ordered domains of the pure components and (D) based on the average APM of the pure components. Due to lack of diffraction, the ordered APM of the pure GM1 monolayer was estimated by a linear extrapolation from the 80, 70, and 60 mol % GM1 monolayer APMs. Dashed lines are presented as a visual guide. Error bars are on based on error propagation from the Bragg peak fitting and an estimated 1 Å² error in the isotherm APMs. To see this figure in color, go online.

In general, DPPC:GSL mixtures had larger APMs compared to DPPC:CM mixtures presumably due to steric repulsions and packing constraints caused by the carbohydrates. In the case of DPPC:GA1 mixtures, the average lipid APM within ordered domains steadily increased from 43.4 Å² for pure GA1 to 46.5 Å² for pure DPPC monolayers, obeying ideal mixing behavior within the ordered regions (see curves with circles (blue online) in Fig. 6, A and C). On the other hand, when comparing the average APM over the entire monolayer, the APM of high (≥40%) GA1 concentration mixtures were more condensed than expected based on ideal mixing (see curves with circles (blue online) in Fig. 6, B and D). Because the DPPC:GA1 obeyed ideal mixing within the ordered domains (Fig. 6 C), but an overall condensation effect is seen (Fig. 6 D), this result suggests that the fraction of ordered domains increases with increasing GA1 concentration. The DPPC:GM1 monolayer did not follow the same trend in the ordered domain and a minimum APM was found for the 50:50 mol % composition (see curves with squares (red online) in Fig. 6 A). When comparing the average APM of the entire monolayer, GM1 mixtures exhibited a strong condensation behavior relative to ideal mixing, where as low as 20 Å² per molecule below ideal mixing predictions for 20:80 DPPC:GM1 mixtures was observed (see curves with squares (red online) in Fig. 6, B and D). This condensation across the entire layer can be attributed to an increased fraction of ordered regions, to condensed APM within the ordered regions, or to both effects. Fluorescence micrographs of these mixtures indicate the 75:25 DPPC:GM1 mixture to have the highest fraction of micron-scale condensed domains as indicated by fluorescence dye exclusion, but submicron domains cannot be visualized with this technique (20). To distinguish between these factors, the APM within the ordered domains could be compared to ideal mixing conditions. However, pure GM1 monolayers were not sufficiently ordered at 30 mN/m to yield diffraction peaks. To estimate the area per GM1 molecule in a hypothetical ordered domain of pure GM1, a linear extrapolation was made from the APM of 80, 70, and 60 mol % GM1 monolayers. This procedure yielded an APM of 46.0 Å², close to the APM measured for pure DPPC. This extrapolated value was used to compare the APM within ordered domains of DPPC:GM1 to ideal mixing predictions. Contrary to the case of DPPC:GA1, significant negative deviation from ideal mixing in DPPC:GM1 mixtures was observed within the ordered domains corresponding to a condensation of the lipid packing and a decrease in the average APM up to a certain point when higher GM1 contributes to dilation. Similar to the case of DPPC:CM mixtures, this result suggests that an interdispersed distribution of the two components alleviates steric repulsions from either the GM1 carbohydrates or PC headgroups, allowing tighter in-plane molecular packing.

Discussion

The influence of carbohydrate structure on the ordering of glycosphingolipids (GSLs) and surrounding phospholipids was investigated in monolayers at the air-water interface using synchrotron XR and GIXD. The phospholipid component, DPPC, is a standard lipid in biophysical studies as phosphatidylcholine headgroups constitute a major component of eukaryotic cell membranes. The three GSL components were chosen to span a range of carbohydrate complexity while maintaining essentially the same hydrophobic tail structure. This allowed the role of carbohydrate structure on molecular packing in membranes to be investigated using binary mixtures of DPPC:CM, DPPC:GA1, and DPPC:GM1. With a carbohydrate composed of four sugar groups and one sialic acid, GM1 possessed the largest carbohydrate headgroup of the molecules studied. GA1, which lacks a sialic acid but is otherwise structurally analogous to GM1, has a carbohydrate cross-sectional area that is more commensurate with the area occupied by its hydrophobic tails. CM, lacking any carbohydrate residues but maintaining the same sphingosine backbone and hydrophobic tail structure as GM1 and GA1, was used as a control.

GSL carbohydrate conformation

For pure DPPC and binary DPPC:CM monolayers, the hydrophobic tails and the hydrophilic headgroup regions were clearly distinguishable (Fig. S1). However, due to similar lengths of the hydrocarbon chains of DPPC and CM, only minor differences were observed among the compositions studied. In contrast, binary DPPC:GSL monolayers clearly show the presence of the carbohydrate headgroups in the water subphase as well as differences between the DPPC:GA1 and DPPC:GM1 mixtures.

At the 75:25 DPPC:GSL mole ratio (Fig. 2 C), the hydrocarbon tail regions of both mixtures are roughly equivalent to that of pure DPPC, suggesting that the interactions between neighboring carbohydrate headgroups do not substantially perturb the alkyl tail packing. Subtraction of the ρ_e⁻ distribution of DPPC:GM1 from the ρ_e⁻ of pure DPPC yields the ρ_e⁻ of the GM1 headgroup immersed in the water subphase (Fig. 7, top two curves (red online)). In a similar way, by subtracting the ρ_e⁻ of DPPC:GA1 from the DPPC:GM1 the precise position of the sialic acid residue can be obtained (Fig. 7, bottom two curves (blue online)). This distribution suggests that the sialic acid does not extend further into the water than the four sugars of the carbohydrate headgroup and, therefore, may play an important role in altering lateral interactions between molecules compared to GA1.

Figure 7.

Electron density difference between pure DPPC and 75:25 DPPC:GM1 (upper two lines; red online) from Fig. 2C shows GM1 headgroup position in regard to the center of DPPC headgroup (z = 0). The ρ_e⁻ difference between 75:25 DPPC:GM1 and 75:25 DPPC:GA1 (lower two lines; black online) indicates sialic acid position. The width of the ribbons represents uncertainty in the distributions. To see this figure in color, go online.

For 25:75 DPPC:GSL mixtures (Fig. 2 D), the carbohydrate headgroups were more pronounced and structurally similar for the GA1 and GM1 mixtures. However, in the case of the DPPC:GM1 the higher volume fraction of the carbohydrates and their steric repulsions significantly reduced the extensions of the hydrocarbon tails. This suggests that steric interactions between GM1 carbohydrates limit the molecular packing within the monolayer, resulting in a larger average APM. These results indicate that the dominant interactions influencing molecular packing make transition from tail-tail to carbohydrate-carbohydrate interactions with increasing GM1 concentration. No such transition is evident in GA1 mixtures, suggesting that the sialic acid conformation plays a role in this transition.

Phase separation and enhancement of lipid ordering

Whereas XR provides an average structure of the entire monolayer, GIXD signal arises only from the ordered domains and not the more liquid-like disordered portion of the film. The overall diffracted intensities and the coherence lengths of the peaks can be used to determine the average size and coverage of ordered domains in the monolayer. The integrated diffracted intensity is correlated with the degree of lipid ordering, and can be attributed to a combination of the number of lipids in ordered domains and the length of their hydrocarbon chains contributing to the diffraction. To a first approximation, all components have comparable APMs and similar chain lengths so that the diffracted intensity is proportional to the total area of ordered domains within the monolayer. The coherence length is obtained from the width of the diffracted peaks and reflects the average size of the ordered domains. These quantities can be used to ascertain the role carbohydrates play in phase separation, both between distinct ordered phases, as well as between disordered and ordered phases, and the lateral extent of lipid ordering.

In the case of the CM mixtures, comparison of the APM obtained from isotherms, which averages over both ordered and liquid-like regions of the monolayer, and those obtained from GIXD, were similar for all mixtures. This indicates that the majority of the monolayer is in ordered or condensed phases. Ceramides, like other members of the sphingolipid family, contain amide and hydroxyl groups positioned in an orientation to serve as hydrogen bond donors and acceptors to allow lateral interactions with neighboring molecules (50); this likely causes strong intermolecular interactions that dictate ordering. The formation of ceramide-enriched gel phase domains in glycerophospholipid/ceramide mixtures was also observed by Huang et al. (51) using ²H NMR spectroscopy on deuterated DPPC-C₁₆-CM mixtures and was attributed to the smaller CM headgroup, which allows for closer hydrocarbon chain interactions and complex-like formation between CM and PC (52). The single diffraction peak in mixtures containing 50% CM suggests that CM can accommodate the more bulky DPPC up to a 1:1 ratio. The CM-rich phase does not exactly match the packing of the pure component due to the inclusion of DPPC in the lattice. This CM-rich phase persists in the binary mixtures down to 15% CM. Mixtures with CM concentrations <50% exhibited phase separation into two ordered phases: one enriched in CM; and the other enriched in DPPC, with a variable DPPC:CM complex stoichiometry. A modest decrease in the overall diffracted intensity suggests that the fraction of the monolayer in ordered phases is decreasing with increasing DPPC concentration. This is consistent with the lower coherence lengths of the DPPC-enriched phases. However, all mixtures of these components exhibit continuous changes in intensities and coherence lengths between the two pure component systems. These results indicate that CM, containing a sphingosine backbone without any carbohydrate residues, is more ordered than DPPC and can accommodate the phospholipid in the CM lattice up to a 1:1 ratio after which phase separation occurs and excess CM promotes ordering in the DPPC-rich phase.

For DPPC:GA1 mixtures, the APM obtained using isotherms and GIXD measurements were close, and exhibited a small negative deviation at high GA1 concentrations. Again, this indicates that the majority of the monolayer was in the ordered phase. A single ordered phase was observed for all DPPC:GA1 mixtures though an ordered-ordered domain phase separation would likely be indiscernible if it existed. Both pure components exhibited ordered phases and no fixed optimal stoichiometry for packing was observed. The integrated diffracted intensity was greater for mixtures at all ratios compared to either of the pure components, with a maximum at ∼40% GA1 (see Fig. 4). This indicates that GA1 enhances lipid ordering by increasing the relative fraction of ordered to disordered phases. This is difficult to ascertain on the micron-scale with fluorescence microscopy, as a GA1 monolayer exhibits a homogenous condensed phase mesh at the interface rather than discrete dye-excluding micron-sized domains (26). Additionally, the coherence length of the ordered domains reaches a maximum between 10 and 25% GA1 (see Fig. 5). Interestingly, the largest ordered domains were observed at lower GA1 concentrations than the concentration exhibiting the highest area fraction of ordered lipids, suggesting a complex interplay between domain size and extent of lipid ordering. Overall, these results indicate that GA1 promotes ordering of the surrounding phospholipid molecules.

In contrast to the previous cases, GM1 mixtures exhibited large differences in isotherm and APM measured by GIXD. At GM1 concentrations above 50%, the average monolayer APM was much larger than the ordered phase APM, indicating that, at these concentrations, a large fraction of the mixtures was in a liquid-like disordered phase; this is supported by fluorescence and atomic force micrographs of the binary mixtures (20). Whereas pure GM1 monolayers did not possess sufficient lipid order to diffract, mixtures containing as much as 80% GM1 exhibited a broad diffraction peak from a single ordered phase. Similar to the previous cases, the stoichiometry within ordered domains varied as a function of GM1 concentration. Based on results from a previous study that determined APM values averaged over the entire monolayer and fluorescence imaging of condensed domain formation (20), we postulated a 3:1 DPPC:GM1 complex stoichiometry where the lipid film appeared to be the most condensed as reflected by the smallest average APM value. In this study, we focused on the packing and stoichiometry only within the ordered domains. Assuming a stable 3:1 DPPC:GM1 optimal stoichiometry for packing (20), the diffraction from monolayers containing 0–25 mol % GM1 should be a linear combination of DPPC and 75:25 DPPC:GM1 diffraction signals. Similarly, monolayers containing 25–100 mol % GM1 should exhibit the same diffraction behavior as the 75:25 DPPC:GM1 monolayer with decreasing diffracted intensity. However, this is not what we observed, suggesting that the equilibrium stoichiometry within ordered domains is interdependent with the stoichiometry of surrounding disordered phases. In general, the integrated diffracted intensity increased as the monolayer composition progressed from disordered GM1-rich to ordered DPPC-rich mixtures. In particular, a maximum intensity was observed at a GM1 concentration of 10 mol %, indicating an enhancement of lipid order relative to pure DPPC. Additionally, coherence lengths larger than that for pure DPPC were obtained from monolayers over a wide range of GM1 concentrations (10–60 mol %), even from those that exhibited significantly lower diffracted intensity. Similar to GA1 mixtures, these results indicate that GM1 promotes ordering of the surrounding phospholipid molecules.

GSL induced APM condensation in ordered domains

To determine the capacity for GSLs to further condense lipid packing within ordered domains, the APM of binary mixtures were compared to the predicted APM based on ideal mixing of the pure components. In mixtures containing CM, significant APM condensation or tighter packing was observed within both the DPPC-enriched and the CM-enriched domains. Although it was not possible to quantitatively determine DPPC:CM mole ratios within the CM-enriched phases, at a 1:1 ratio there was no significant diffraction from a DPPC-enriched phase, suggesting that the CM-enriched phase can accommodate up to 1:1 mixing with DPPC and condense the DPPC packing to be commensurate with the tighter CM packing. In these mixtures, the CM headgroups are significantly smaller than the PC headgroups and the mechanism for CM-induced APM condensation may be related to relief of steric interactions between neighboring DPPC headgroups, allowing for tighter packing of the lipid tails with less tilt into a hexagonal 2D unit cell. Recent work showed that ceramide analogs with methylation at the 2-NH position attenuated ceramide-ordered phase formation (53), indicating that ceramide 2-NH hydrogen bonding along with acyl chain length, saturation, and relatively small headgroup are necessary for ordered CM-rich domains in PC bilayers. At higher DPPC concentrations, excess DPPC forms a second DPPC-enriched phase which, based on changes to the diffracted peak positions, contains some fraction of CM. This phase also exhibits an APM condensation relative to ideal mixing, presumably due to the same steric relief mechanism. Interestingly, ideal mixing of these two components appears to be obeyed based on the average APM obtained from isotherms even though the majority of the monolayer is in condensed or ordered phases (20). Deviation from ideal mixing observed using GIXD may be due to the elimination of contributions from the small fractions of disordered lipids or to the higher precision of the technique.

In the case of GA1 mixtures, no significant deviation from ideal mixing was observed in the ordered domains. Despite the enhancement of lipid ordering across the film indicated by increased diffracted intensity and larger domain sizes, GA1 does not induce an APM condensation, or tighter packing, in mixtures with DPPC. This suggests that GA1’s carbohydrate conformation remains the same in a local environment of either DPPC or other GA1 molecules. Of note is the narrower cross-sectional area of GA1 compared to DPPC in ordered domains (43.4 Å²/mol for GA1 and 46.5 Å²/mol for DPPC), which highlights the inherent close packing of the carbohydrate structures in agreement with conformational free energy calculations that show an extended headgroup sugar backbone with a small bend at the glycosidic II–III linkage (54). The NMR-derived tertiary structures of GA1 and GM1 show similar structural features of the 4-carbohydrate portion on the backbone (55), indicating that the differences seen in DPPC:GSL ordering and condensation arise primarily from the charged sialic acid residue.

Due to the lack of diffraction from pure GM1 monolayers, there is some degree of ambiguity in determining deviation from ideal mixing in the ordered domains of DPPC:GM1 monolayers. To approximate the APM of a hypothetical ordered domain of GM1, we performed a linear extrapolation from the APMs of ordered domains in high GM1 concentration mixtures. Using this approach, all DPPC:GM1 mixtures exhibited tighter packing and APM condensation relative to ideal mixing of the pure components with the greatest deviation occurring at a 1:1 DPPC:GM1 ratio. Again, the mechanism driving the APM condensation in these mixtures may be attributed to the relief of steric interactions between neighboring molecules and a phospholipid/GSL hydrogen bonding network. In contrast to the DPPC:CM mixtures, the GM1 carbohydrate headgroups are significantly larger than the PC headgroups and the steric interactions between neighboring GM1 molecules are likely to be relieved in the mixtures. Although there are extensive mentions in the literature of the likelihood of a GM1 headgroup hydrogen bonding network, detailed atomic-level experimental characterization of the carbohydrate-carbohydrate interactions involving GSL clusters remains largely unexplored. NMR studies on micellar GM1 structures show no evidence of intermolecular carbohydrate-carbohydrate interactions (56, 57), though each GSL headgroup has a large hydration sphere that could mediate long-range interactions among monomers and not be detected by NMR experiments (58). Conversely, ESR (9) and Fourier transform infrared attenuated total reflection (59) measurements of GM1 in phospholipid bilayers indicate that the oligosaccharide headgroup is capable of forming intermolecular hydrogen bonds. More recently, in atomistic simulations of ∼4:1 POPC:GM1 mixtures, GSL cluster formation is shown to be driven by hydrogen-bond formation primarily between the sialic acid and galactose residues at the end of the headgroup (sialic acid and galactose) (60, 61). In both studies, GM1 formed on average ∼2.5 hydrogen bonds with neighbor PC lipids with the majority of the intermolecular interactions arising between the phospholipid phosphate groups and either the terminal sugar groups (60) or the glucose residue and ceramide backbone (61). These studies suggest that an extensive DPPC-GM1 hydrogen bonding network can contribute to our observed condensation effect where the packing within the binary DPPC:GM1 mixture ordered domains is tighter than the pure components.

Conclusions

The results presented here demonstrate two types of enhanced lipid order, relative to the individual components, in phospholipid and glycosphingolipid mixtures. One type is an increase in the ratio of ordered to disordered phases in the monolayer which may, but not necessarily, be coincident with the growth of the average ordered domain sizes. Increased fractions of ordered domains were observed for DPPC mixtures containing either GA1 or GM1 but not with CM. This suggests that the carbohydrate residue is required for this type of lipid order enhancement but the mechanism driving this behavior remains to be resolved. The other type is condensation of the ordered APM in mixtures relative to ideal mixing of the two components. Such ordered APM condensation was observed in GIXD measurements from DPPC mixtures containing either CM or GM1 but not with GA1. The measurements examined mixtures where the GSL headgroups were either smaller (CM), approximately equal to (GA1), or larger (GM1) than the PC headgroups. We propose that the condensation effect can be attributed to relief of steric interactions between the headgroups of neighboring molecules and, to achieve this, there needs to be a mismatch between the headgroup size of the two components. This mismatch allows for the relief of steric interactions between the larger of the two headgroup types and drives the system to a tighter packing and condensed APM in the ordered phase.

Author Contributions

K.Y.C.L. and J.M. conceived the study. S.L.F., E.Y.C., K.D.C., E.B.W., T.P., and J.M. performed research. E.B.W., S.L.F., J.M., and K.Y.C.L. analyzed the data. E.B.W. and S.L.F. wrote the manuscript. All authors contributed to editing the manuscript.

Editor: Michael Brown.

Supporting Citations

References (62, 63) appear in the Supporting Material.