Dynamic Modulation of the Glycosphingolipid Content in Supported Lipid Bilayers by Glycolipid Transfer Protein

Abstract

Supported lipid bilayers (SLBs) are popular models of cell membranes. Owing to the importance of glycosphingolipids (GSLs) in modulating structure and function of membranes and membrane proteins, methods to tune the GSL content in SLBs would be desirable. Glycolipid transfer protein (GLTP) can selectively transfer GSLs between membrane compartments. Using the ganglioside GM1 as a model GSL, and two mass-sensitive and label-free characterization techniques—quartz crystal microbalance with dissipation monitoring and ellipsometry—we demonstrate that GLTP is an efficient and robust biochemical tool to dynamically modulate the GSL content of SLBs up to 10 mol % GM1, and to quantitatively control the GSL content in the bulk-facing SLB leaflet. By exploiting what we believe to be a novel tool, we provide evidence that GM1 distributes highly asymmetrically in silica-supported lipid bilayers, with ∼85% of the ganglioside being present in the bulk-facing membrane leaflet. We report also that the pentameric B-subunit of cholera toxin binds with close-to-maximal stoichiometry to GM1 in SLBs over a large range of GM1 concentrations. Furthermore, we quantify the liganding affinity of GLTP for GM1 in an SLB context to be 1.5 μM.

Introduction

Supported lipid bilayers (SLBs) have become very popular as models of cell membranes with potential biotechnological applications (1–4). They have provided insight into a plethora of membrane processes from the molecular to the cellular scale, including lipid domain formation and dynamics (5–8), protein adsorption and self-assembly at lipid membranes (9,10), intermembrane interactions (11), and juxtacrine signaling (12,13).

The interest in confining lipid membranes on surfaces has been driven by the emergence of a multitude of surface-sensitive characterization techniques, and the possibility to spatially control membrane deposition (14–16), composition (17), and shape (18,19). The confinement to a macroscopic surface is also attractive from a practical point of view: the adjacent aqueous solution can be readily exchanged and membranes sequentially incubated with different sample solutions without the need of elaborate separation steps that are typically required for vesicles in solution, thus facilitating both the biochemical modification of membranes and functional studies.

Glycosphingolipids (GSLs) represent a small (typically 5–10 weight %) but functionally important fraction of membrane lipids in eukaryotes (20). They provide the plasma membrane with mechanical stability, and take part in fundamental biological processes including cell differentiation, cell-cell interaction, and transmembrane signaling. GSLs are enriched in liquid-ordered membrane microdomains which putatively function as organization sites for signaling proteins. GSL trafficking, accumulation, or cluster formation has been associated with many pathological conditions, including autoimmune disorders and neurodegenerative diseases (21).

In functional studies, it would be desirable that the composition of model membranes can be tightly controlled and dynamically modulated. The overall lipid composition of SLBs can be varied with relative ease, e.g., by adjusting the mixing ratio of lipids in the liposomes if the SLB is formed by the method of vesicle spreading (3). Attractive or repulsive interactions of some lipid types with the underlying support, though, do often lead to an asymmetric distribution of lipids between the two membrane leaflets (3,22–25). The exact lipid composition in the bulk-facing SLB leaflet is hence more difficult to control. Notably, it has hitherto hardly been possible to dynamically modulate the lipid composition of existing SLBs in a controlled way (26–30). Methyl-β-cyclodextrin has become popular as a shuttle for the insertion or extraction of cholesterol (31,32), another functionally important membrane component that profoundly affects lipid ordering and domain formation. Similar molecular tools for the enrichment or depletion of lipid bilayers with a selected lipid type would be very attractive.

GLTP is a soluble 24 kDa protein with a pI of ∼9.0 that specifically promotes intermembrane transfer of sphingoid- and glycerol-based glycosphingolipids in which the first sugar residue is β-linked to the hydrophobic lipid moiety (33). In vitro assays with lipid vesicles (34–36) and structural studies (37,38) have led to the transfer model (see Fig. S1 in the Supporting Material) in which GLTP first extracts a glycosphingolipid molecule via transient interaction with the donor membrane. GLTP recognizes the sugar head and encapsulates one or both lipid tails within its hydrophobic interior to form a complex, which then can diffuse in solution. The liganding affinity of GLTP for double-chained GSLs was found to be in the lower μM range (36). A transient encounter with an acceptor membrane eventually leads to lipid release and terminates the transfer.

The objective of this study was to evaluate the potential of GLTP as a tool to modulate the GSL content of SLBs. To this end, we employed the ganglioside GM1 and phosphatidylcholine (PC) as model GSL and inert lipid, respectively. Two mass-sensitive and label-free characterization techniques, i.e., quartz crystal microbalance with dissipation monitoring (QCM-D) and ellipsometry, were used to monitor the formation of SLBs and the transfer activity of GLTP. The specific recognition of GM1 by the B-subunit of cholera toxin was exploited to quantify the amount of the ganglioside in the bulk-facing SLB leaflet.

This approach allowed us to quantify the interleaflet distribution of GM1 in SLBs and the stoichiometry of the multivalent interaction between GM1 and cholera toxin B subunit. Our results also provide a quantification of the liganding affinity of GLTP for GM1 and GM1 insertion/extraction rates in the SLB context.

Materials and Methods

Buffers

A HEPES buffer solution of 150 mM NaCl, 10 mM HEPES, and 3 mM NaN₃ at pH 7.4 was prepared in ultrapure water (resistivity 18.2 MΩ/cm). A quantity of 2 mM CaCl₂ was added for the formation of SLBs. A quantity of 2 mM EDTA was added for all other incubation steps.

Vesicle preparation

Lyophilized 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) and GM1 ganglioside (GM1) were purchased from Avanti Polar Lipids (Alabaster, AL). Pure DOPC and mixtures of DOPC and GM1 at different molar ratios were dissolved in chloroform and an equivolume mixture of chloroform and methanol, respectively, then dried, resuspended in HEPES buffer, and homogenized as described earlier (39). Small unilamellar vesicles (SUVs) were obtained by sonication (40). Extruded unilamellar vesicles (EUVs) were obtained by extrusion through polycarbonate membranes with 100 nm pore size (LiposoFast; Avestin Europe, Weinheim, Germany). The vesicle suspensions of ∼2 mg/mL concentration were stored at 4°C under nitrogen, and diluted to desired concentrations before use. Lipid concentrations and mixing ratios were deduced from the mass of lipids dissolved within an error of less than±10%.

Proteins GLTP and CTB₅

Glycolipid transfer protein (GLTP; molecular mass 24 kDa) was expressed and purified as described earlier (37,38). The stock solution of ∼10 mg/mL was diluted in 20 mM TRIS (pH 7.4) and 150 mM NaCl to 2 mg/mL and aliquots were stored at −80°C. The protein was verified to be functional by fluorescence binding (36) and crystallization (37,38) assays. The protein concentration was determined by optical density measurements (Nanodrop 2000; Thermo Scientific, Wilmington, DE) at a wavelength of 280 nm using an extinction coefficient of 0.79 g⁻¹·L·cm⁻¹. Lyophilized cholera toxin B-subunit pentamer (CTB₅, 58 kDa) (Sigma, St. Louis, MO) was reconstituted in ultrapure water at a stock concentration of ∼1 mg/mL, and stored at 4°C. Before use, protein stock solutions were diluted in EDTA-containing HEPES buffer to the desired concentration.

Substrates

Silica-coated QCM-D sensors (QSX303; Q-Sense, Västra Frölunda, Sweden) were immersed in an aqueous solution of 2% sodium dodecyl sulfate for 30 min, rinsed abundantly with ultrapure water, blow-dried with N₂, treated with UV/ozone (Bioforce Nanoscience, Ames, IA) for 30 min, and placed in the measurement chambers.

Quartz crystal microbalance with dissipation monitoring

Quartz crystal microbalance with dissipation monitoring (QCM-D) measures changes in resonance frequency (Δf) and dissipation (ΔD) of a sensor crystal upon interaction of (soft) matter with its surface. The QCM-D response is sensitive to the mass (including coupled solvent) and the mechanical properties of the surface-bound layer. Measurements were performed with a Q-Sense E4 system (Q-Sense, Västra Frölunda, Sweden), at a working temperature of 23°C. Unless otherwise stated, the system was operated in flow mode, with a rate of typically 20 μL/min. For quantitative interaction assays between GLTP and GM1, a purpose-designed fluid handling method was developed (see the Supporting Material). Δf and ΔD were measured at six overtones simultaneously; changes in dissipation and normalized frequency, Δf = Δf_i/i, of the ninth overtone (i = 9) are presented.

Quantification of adsorbed masses

The “acoustic” mass, m_QCM,, was calculated according to $m_{\mathrm{QCM}}=-C\times \Delta f,\mathrm{with}C=18.06\pm 0.15\mathrm{ng}·\mathrm{c}{\mathrm{m}}^{-2}·\mathrm{H}{\mathrm{z}}^{-1}$for the sensors used here. This equation is valid for films that exhibit low dissipation, as is the case for CTB₅ or SLBs, but includes hydrodynamically coupled solvent. To account for the contribution of solvent to the QCM-D response, calibration measurements with an in situ combination of QCM-D and ellipsometry on the same support were performed (see the Supporting Material).

Results

Insertion of GM1 into SLBs

To study the efficiency of GLTP to transfer GM1 from vesicles to supported lipid bilayers (SLBs), the following assay was designed (Fig. S1). First, an SLB was formed by spreading of small unilamellar vesicles (SUVs) made of DOPC to a silica surface. Second, selected test solutions containing GLTP and/or GM1 in various presentations were exposed to the SLB. Third, cholera toxin B-subunit was added to determine the amount of accessible GM1 in the SLB. All steps were monitored by QCM-D (Fig. 1).

Figure 1.

Assay for the insertion of GM1 into SLBs, monitored by QCM-D. (A) Formation of an SLB upon exposure of 50 μg/mL DOPC SUVs to a silica-coated sensor surface. (B) Exposure of different donor solutions in flow mode to the SLB: mixture of 2.1 μM GLTP and 50 μg/mL SUVs containing 5 mol % (= 3.0 μM) GM1 (-○-), 50 μg/mL SUVs containing 5 mol % GM1 alone (-▿-), 2.1 μM GLTP alone (-⋄-), micellar solutions of GM1 at concentrations of 32 μM (-Δ-) and 640 μM (-□-), and pure buffer solution (-×). (C) Subsequent incubation of 0.17 μM CTB₅. Each incubation step starts at 0 min; rinses in buffer are indicated (arrowheads).

Exposure of DOPC SUVs to silica (Fig. 1 A) resulted in a two-phase behavior in both frequency and dissipation, with final changes, Δf = −25 Hz and ΔD < 0.3 × 10⁻⁶. This response is characteristic for the formation of an SLB of good quality (41).

Small but detectable and fast changes of Δf = −2 Hz and ΔD = 0.15 × 10⁻⁶ occurred upon continuous exposure of the SLB to a solution of 50 μg/mL vesicles containing 5 mol % (3.0 μM) GM1 together with 2.1 μM GLTP (Fig. 1 B). No changes were observed when adding either of the two components alone. Neither did we observe any response when presenting pure GM1, in the form of micellar solutions, to the SLB; even a >200-fold increase in the GM1 concentration (640 μM), as compared to the SUVs, did not produce any significant frequency shift. These results provide a first indication for GLTP-mediated transfer of GM1 to SLBs. From the frequency response upon transfer, we estimate an apparent initial insertion rate of −10 Hz/min. Also, the QCM-D response upon transfer reached 80% of its final value within 30 s, indicating that insertion is fast.

It is remarkable that exposure of GLTP in μM concentrations to DOPC SLBs did not result in any measurable QCM-D response. Neither did we find any unbinding when rinsing after exposure of GLTP and GM1-containing vesicles that would have been indicative for the transient presence of GLTP at the membrane. Apparently, GLTP does not bind with appreciable affinity to SLBs irrespective of their GM1 content. The interaction that leads to GM1 transfer, therefore, must be short-lived.

The direct QCM-D response upon GM1 transfer remained rather small. This limits the quantification of transfer, and minor transfer may potentially remain undetected. CTB₅ was used to enhance the signal for the presence of GM1 in the SLB (Fig. 1 C). Strongest binding (Δf = −10.5 Hz) was observed on SLBs that were incubated with GM1-containing SUVs and GLTP, corroborating successful transfer. CTB₅ remained stably bound upon rinsing in buffer, and changes in dissipation remained small (ΔD < 0.5 × 10⁻⁶) throughout the adsorption process, indicating that CTB₅ associated tightly to the SLB (42). No binding was detectable on SLBs that were not exposed to GM1, confirming that CTB₅ binding was specific. CTB₅ binding was very small (|Δf| < 1 Hz) on SLBs that were exposed to SUVs containing 3.0 μM GM1, or to 32 μM of GM1 micelles in the absence of GLTP. However, some binding (Δf = −4 Hz) was observed after incubation with large concentrations (640 μM) of micellar GM1 solutions, indicating that some spontaneous insertion does also occur.

These data provide evidence that GLTP can efficiently transfer GM1 from vesicles to SLBs. The transfer is faster, by several orders of magnitude, than the spontaneous insertion of GM1 from vesicular or micellar solutions in the absence of GLTP.

Extraction of GM1 from SLBs

In the next step, we tested whether GLTP is able to extract GM1 from SLBs. For this purpose, GM1-containing SLBs were formed from SUVs that were made of a mixture of DOPC and 5 mol % GM1 (Fig. 2 A). The SLBs were then exposed to a continuous flow of solutions of GLTP and/or SUVs containing only DOPC. The concentrations of GLTP and SUVs were kept identical to the insertion assay.

Figure 2.

Assay for the extraction of GM1 from SLBs, monitored by QCM-D. (A) Formation of an SLB upon exposure of 50 μg/mL SUVs containing 5 mol % GM1 to a silica-coated sensor surface. (B) Exposure of different acceptor solutions in flow mode to the SLB: mixture of 2.1 μM GLTP and 50 μg/mL DOPC SUVs (-○-), 2.1 μM GLTP alone (-⋄-), 50 μg/mL DOPC SUVs alone (-▿-), and pure buffer solution (-×-). (C) Subsequent incubation of 0.17 μM CTB₅. Each incubation step starts at 0 min; rinses in buffer are indicated (arrowheads).

Incubation with a mixture of GLTP and acceptor SUVs for 10 min induced an increase in Δf, indicating removal of lipid material from the SLB. The value ΔD remained unchanged (Fig. 2 B), indicating that the SLB remained intact, i.e., without large-scale defects that would likely have promoted adhesion of acceptor SUVs. Subsequent addition of CTB₅ (Fig. 2 C) resulted in Δf = −3.5 Hz, whereas −23 Hz were obtained on untreated, GM1-containing SLBs. Prolonged incubation with GLTP and DOPC SUVs for 30 min decreased the CTB₅-induced absolute frequency shifts below 1.0 Hz (data not shown). In contrast, no changes were observed upon incubation of DOPC SUVs alone (Fig. 2, B and C). Neither did we observe any changes in the apparent GM1 content when flushing the SLB overnight with buffer solution (data not shown). This provides evidence that most accessible GM1 can be removed from the SLB and that it is the GLTP that extracts GM1. The apparent initial extraction rate was 0.5 Hz/min, or 20-fold lower than the initial rate of insertion under equivalent conditions. It also took much longer, ∼8 min, to reach 80% of the maximal response for extraction, as compared to 30 s for insertion.

The presence of acceptor vesicles was not required for the extraction of GM1 from SLBs: efficient extraction occurred even in the presence of GLTP alone (Fig. 2, B and C). In fact, the extraction kinetics were hardly affected by the absence of donor vesicles, indicating that extraction of GM1 is the rate-limiting step in GLTP-mediated transfer of GM1 from SLBs to acceptor vesicles.

Moreover, insertion and extraction of GM1 could be cycled several times, by alternately exposing SLBs for 10 min to a mixture of GLTP and SUVs containing 5 mol % GM1, and to GLTP alone. After each addition of the GLTP/SUV mixture, CTB₅-induced frequency shifts of −10 ± 1 Hz were obtained, whereas only −2 ± 1 Hz were found after exposure to GLTP alone (two full cycles were tested; data not shown).

After a qualitative investigation of the propensity of GLTP to modulate the composition of SLBs, further detailed studies were performed to better understand the interaction between GM1 and GLTP. To this end, we first had to establish that CTB₅ can serve as a quantitative marker for the amount of accessible GM1.

CTB₅ as a quantitative marker for accessible GM1

The B-subunits of cholera toxin are assembled into a pentamer with fivefold axial symmetry, and upon binding to lipid membranes, each subunit in the pentamer can interact specifically with one GM1 molecule (43,44). To test whether CTB₅ binds stoichiometrically to GM1, we prepared SLBs containing various molar proportions of GM1. To this end, we mixed SUVs from two different stock solutions, one containing pure DOPC and the other ∼5 mol % GM1, at controlled ratios, and exposed the resulting solutions to silica surfaces. The approach of mixing two vesicular solutions of different composition, rather than preparing a range of SUV stock solutions with varying GM1 content, was chosen because pipetting of buffer-suspended vesicles is more accurate and simpler than the mixing of lipids that are dissolved in rapidly evaporating organic solvents. In addition, SLBs were also prepared from an SUV solution containing 10 mol % GM1.

SLBs could be successfully formed with all nominal GM1 concentrations (data not shown). The absolute final frequency shifts for SLBs containing GM1 were slightly higher than for pure DOPC SLBs. An increase in the molar proportion of GM1 in the SUVs, from 0 to 10%, correlated with an increase in |Δf|, from 25 to 32 Hz. The additional shift of 7 Hz indicates an increase in the thickness of the SLB, by ∼1 nm. The increased size of the carbohydrate-bearing GM1 headgroups as compared to PC is consistent with such a scenario. An alternative explanation for the increased frequency shift would be the presence of residual vesicles in the SLB, i.e., a decreasing SLB quality. Such vesicles, however, would also induce an increase in the dissipation shift. In our experiments, we did not observe a systematic change in ΔD. Instead, the dissipation remained small (ΔD < 0.5 × 10⁻⁶), indicating that the number of residual vesicles that are embedded in the SLB is low.

The open squares in Fig. 3 A show the adsorbed amount of CTB₅, derived from QCM-D frequency shifts, as a function of the molar proportion of GM1 in the SUVs that were used for the formation of SLBs. Independent of the GM1 proportion, CTB₅ bound irreversibly and binding did not depend significantly on CTB₅ concentration within the tested range (0.04–1.6 μM). The adsorbed amounts increased monotonously with the molar proportion of GM1. The relationship, however, was not linear. This is not surprising if one considers that the mass measured by QCM-D includes a substantial amount of coupled solvent, and that the relative contribution of trapped water to the frequency response can change with coverage (45,46). After correcting the QCM-D data for the contribution of coupled solvent (open circles in Fig. 3 A) with the aid of a calibration curve that was derived from a combined QCM-D and ellipsometry measurement (Fig. S2), we found a close-to-linear relationship between the nominal molar proportion of GM1 and the surface density of CTB₅, for GM1 proportions up to 5%. A plateau was reached for higher GM1 contents. The surface mass density of CTB₅ in the plateau regime was ∼280 ng/cm². Assuming hexagonal packing, this would correspond to an average center-to-center distance between neighboring CTB₅ molecules of 6.3 nm. The lateral extension of the pentamer is only slightly smaller (6.0 nm) (43), which readily explains the attenuation of CTB₅ binding at nominal GM1 proportions above 5 mol % as a consequence of steric constraints, i.e., saturation of the protein monolayer.

Figure 3.

(A) Stoichiometry of CTB₅ binding to GM1 in SLBs. Masses upon CTB₅ binding to SLBs, formed from SUVs as a function of the molar proportion of GM1 in the SUVs, as obtained by QCM-D (m_QCM, □) and after accounting for the contribution of trapped solvent ($m_{{\text{CTB}}_5}$, ○). The surface density of GM1 in the SLB was derived from the SUV composition, assuming an average surface area per lipid of 0.70 nm² and a symmetric interleaflet lipid distribution. A linear fit (solid line) to the data for $m_{{\text{CTB}}_5}$ reveals a stoichiometry of 2.5:1 between GM1 and CTB₅. $m_{{\text{CTB}}_5}$ on SLBs formed from EUVs of selected GM1 proportions is also shown before (•) and after (▴) coincubation of 0.8 μM GLTP and 200 μg/mL EUVs of identical composition as used during SLB formation for 45 min. A linear fit to the postincubation data gives a stoichiometry of 4.3, indicating close-to-maximal occupancy of the GM1 binding sites in CTB₅ and strong enrichment of GM1 in the bulk facing SLB leaflet before coincubation. (B) Surface density of GM1 in the SLB after coincubation for 30 min (○) and 60 min (•), respectively, of DOPC SLBs with 0.8 μM GLTP and 200 μg/mL donor EUVs of selected GM1 proportions. The dashed line corresponds to the surface density in the donor EUVs, i.e., the maximal possible transfer. Error bars represent variations between two measurements and experimental noise. (C) Equilibrium binding of CTB₅ to laterally mobile GM1, according to predictions by Lauer et al. (47), for CTB₅ solution concentrations of 0.17 (thick dash-dotted line), 0.017 (thin dash-dotted line, Δ) and 1.7 μM (thin dash-dotted line, ▿). The model assumes multistep binding: the CTB₅ pentamer first docks via a single GM1 molecule onto the membrane and then free binding sites are occupied by additional GM1 molecules. Equilibrium constants for initial binding and subsequent surface cross-linking of 1.0 × 10⁷μM⁻¹ and 1.1 × 10⁻¹² cm², respectively, were taken from Lauer et al. (47). At a CTB₅ solution concentration of 0.17 μM, which is typical for our binding assays, a stoichiometry of <2 GM1 molecules per CTB₅ pentamer would be predicted, and a saturation of the membrane surface (∼5 pmol/cm²) with CTB₅ would already be reached at 3 mol % GM1. Furthermore, adsorbed amounts are predicted to be highly sensitive to the solution concentration of CTB₅. None of these predictions is reproduced by our data. However, the model would be consistent with our data (gray solid line) if the equilibrium surface cross-linking constant is increased by four orders of magnitude (dashed lines).

Conversion of the adsorbed masses into molar surface densities (Fig. 3 A) revealed that, for molar GM1 proportions below 5 mol %, only 2.5 GM1 molecules were on average available per CTB₅ pentamer. Two possibilities appear reasonable to explain this observation. First, an excess of CTB₅ on the surface leads to competition for GM1, thereby lowering the average occupancy of binding sites (47). Second, the amount of GM1 that is present in the bulk-facing SLB leaflet does not correspond to 50% of the total GM1 content as assumed in the above calculation. If all gangliosides were instead partitioned in this leaflet, the stoichiometry would indeed be 5:1.

GM1 distribution in SLBs is strongly asymmetric

To test these hypotheses, we devised a coincubation assay that exploits the glycosphingolipid transfer activity of GLTP: SLBs were first formed from a given vesicle solution and then exposed to a mixture of the same vesicle solution together with GLTP. For this assay, we used extruded unilamellar vesicles (EUVs) of 100-nm nominal diameter. Thanks to the large radius of curvature of EUVs, GM1 is likely to be distributed symmetrically between the two membrane leaflets (48–50). Furthermore, the vesicle concentration was chosen such that GM1 in the vesicles was in large molar excess with respect to GLTP in solution. We rationalized that, at equilibrium, the molar proportion of GM1 in the bulk-facing SLB leaflet should adjust to the known molar proportion of GM1 in the vesicles.

The amounts of CTB₅ that bound to SLBs made from EUVs containing 1.8, 4, and 5 mol % GM1, respectively, before and after incubation with the EUV/GLTP mixture for 45 min are displayed in Fig. 3 A. The preincubation data (solid circles) matched our earlier observations on SUV-based SLBs well, indicating that the vesicle source does not affect the amount of GM1 that is displayed in the bulk-facing SLB leaflet. After incubation with the EUV/GLTP mixture, however, CTB₅ binding decreased drastically. The postincubation data could again be fitted with a straight line, and the resulting stoichiometry was 4.3 GM1 molecules per CTB₅ pentamer, close to the maximal value of 5. Prolonged incubation with EUV/GLTP did not affect CTB₅ binding, confirming that equilibrium was attained.

Our findings imply: i), that about 85% of GM1 must reside in the bulk-facing SLB leaflet, and ii), that CTB5 binding scales linearly with the amount of accessible GM1 with close-to-maximal stoichiometry and within a dynamic range of up to 8 mol % GM1 in the accessible SLB leaflet (Fig. S3).

Interaction between GLTP and GM1

Our initial assays on the extraction of GM1 from SLBs (Fig. 2) demonstrated that GLTP alone can efficiently sequester GM1 from a lipid membrane and maintain it in the soluble phase. To obtain quantitative insight into the equilibrium distribution of GM1 between the SLB and GLTP, selected amounts of GLTP were exposed, in a batch of still solution, to an SLB that contained a fixed amount of GM1 (Fig. 4). The amount of GM1 that was extracted at or close to equilibrium was then quantified indirectly, from the amount of CTB₅ that bound to the SLB after rapid removal of GLTP from the solution.

Figure 4.

Quantitative assay for the interaction between GLTP and GM1. (A) Step-by-step assembly of the interaction assay, followed by QCM-D (Δf, □; ΔD, ○). Start and duration of all incubation steps is indicated (arrows). SLBs were formed by spreading of 50 μg/mL SUVs containing 5 mol % GM1 to a silica surface. GLTP of varying concentrations, here [GLTP]_tot = 6.3 μM, was rapidly injected, incubated in still solution for 15 min and then rapidly washed out (see the Supporting Material for details). The total frequency shift upon subsequent incubation of 0.17 μM CTB₅ and a stoichiometry of 4.3 GM1 molecules per CTB₅ pentamer were used to determine [GM1], i.e., the amount of accessible GM1 that remained in the SLB (Fig. S3). The total concentration of accessible GM1, [GM1]_tot, was fixed at 0.75 μM. (B) The fraction of nonliganded GM1, [GM1]/[GM1]_tot, after the reaction is plotted as a function of the normalized total concentration of GLTP, [GLTP]_tot/[GM1]_tot. Error bars represent variations between two measurements and experimental noise. The fit (solid line) to a simple interaction model (see Eq. S2c in the Supporting Material) gives K_D = 1.5 ± 0.4 μM.

Fig. 4 B provides clear evidence that the extraction of GM1 is dependent on the amount of free GLTP: upon increasing the concentration of GLTP in the reaction chamber, CTB₅ binding decreased progressively, i.e., the amount of GM1 that was extracted from the SLB and went into the solution phase increased. Each GLTP molecule can accommodate a single GSL. Our data could be fitted well by a simple model that describes the extraction of GM1 from the SLB as an equilibrium reaction of membrane-bound GM1 and soluble GLTP into a soluble GM1•GLTP complex (see the Supporting Material for details). The resulting dissociation constant was 1.5 ± 0.4 μM.

Tuning the GM1 content in SLBs

Our data in Figs. 1 and 2 demonstrated that GLTP can efficiently extract and insert GM1 from SLBs. Can the amount of GM1 in the SLB be modulated with quantitative control?

The coincubation assay that we used earlier to test membrane asymmetry illustrates how controlled amounts of GM1 can be extracted from SLBs. To test whether the method works equally well to enrich SLBs, we applied the assay on SLBs that were formed from GM1-free vesicles, using EUVs of varying GM1 proportion. The total concentration of lipids and GLTP were kept constant, GM1 in the donor vesicles was used in sufficient excess over GLTP to avoid significant depletion of the GM1 pool into GM1•GLTP complexes, and the incubation time was fixed to 30 min.

For ≤4 mol % GM1 in the donor EUVs, the amount of accessible GM1 (Fig. 3 B) was in good agreement with the values that we had found in the extraction assays (Fig. 3 A). The concentration of accessible GM1 in the SLB was hence equal to the concentration in the EUVs. Beyond 4 mol %, however, the surface density in the SLB remained significantly below that in the EUVs. The amount of transferred GM1 could be pushed further by increasing either the concentrations of GLTP or the incubation time. An incubation time of 60 min, for example, led to maximal transfer for 5 mol % GM1 in the EUVs, and insertion at or above the detection limit of 20 pmol/cm², or 8 mol %, was observed from EUVs containing 20 mol % GM1 (Fig. 3 B).

Apparently, the incorporation of large amounts of GM1 into the SLB is rather slow. It is remarkable, however, that amounts as large as 8 mol % can be incorporated into SLBs at all, if one considers that the incorporation of GM1 is likely to lead to an increase in lipid packing density due to the confinement of SLBs to the support. In this context, it is also notable that the increase in lipid packing does not seem to influence the equilibrium distribution of GM1 between the SLB and donor vesicles. One might speculate that an increased lipid packing could provoke the expulsion of lipids from the SLB, e.g., in the form of membrane blebs or tubes. Such soft structures would be readily detected by the dissipation response. In contrast, we found that the dissipation shifts remained small (<0.5 × 10⁻⁶, data not shown), even upon incorporation of 8 mol % and more GM1, indicating that the SLB retained its planar morphology.

Discussion

We have demonstrated that GLTP can serve as an efficient and robust modulator of the GM1 content in SLBs. The GM1 content in the bulk-facing membrane leaflet can be quantitatively controlled and dynamically cycled by exposure to mixtures of GLTP and vesicles of defined GM1 content. Here, GLTP plays a catalytic role as a carrier that shuffles gangliosides from one membrane compartment to another. A net flux of lipids will be maintained until an equilibrium lipid distribution is attained. Under the employed experimental conditions, we find that the extraction of GM1 from SLBs is slower than the insertion.

Owing to its large hydrophilic head, GM1 exhibits higher solubility than most other lipids and readily forms micelles in aqueous solution (51). Hence, GM1 may insert spontaneously into preformed membranes when presented as a micellar solution. This has, indeed, been shown (28,52,53), although the process was rather slow, requiring GM1 concentrations of several 100 μM and incubation times of several hours. In comparison, GLTP-assisted transfer is readily accomplished with micromolar concentrations of GM1 and (sub)micromolar concentrations of GLTP within a few minutes.

GLTP transfers different GSLs with similar efficiency (36). Hence, it is very likely that the approach outlined here can be readily extended to modify the content of model membranes in other GSL species without affecting their content in other lipid types. The GSL content in SLBs can, in this way, be dynamically tuned within the range that is typically covered by the plasma membrane of eukaryotes (20). With the advent of genetically engineered GLTP mutants that are specific for a particular type of GSL, it may also become possible to modulate, selectively, the membrane content in specific GSL species (L. Malinina, unpublished results).

Considering the important role of GSLs in the regulation of membrane structure and function, and the robustness of GLTP, the protein has the potential to become a hitherto unique and invaluable tool to control membrane composition. In particular, the GSL content in the bulk-facing membrane leaflet can be precisely controlled and dynamically changed—features that, to our knowledge, are not readily provided by established SLB preparation techniques. Such a methodological approach is equivalent to the established use of β-cyclodextrin as a carrier to modulate the cholesterol content in membranes (32,54). GLTP has the added advantage of being very well soluble in aqueous solution.

Mechanism of GLTP-mediated transfer and liganding affinity of GLTP

We find that the interaction of GLTP with SLBs is very weak, irrespective of the absence or presence of GM1, and that GLTP does not perturb the morphology of the SLB. Despite the weak and transient interaction, GM1 transfer is efficient. These results are fully consistent with previous reports by Rao et al. (34,55) on small unilamellar vesicles; they had observed binding affinities between GLTP and lipid membranes in the lower mM range.

Zhai et al. (36) reported K_D = 4.7 ± 1.7 μM for the interaction between GLTP and GM1, and similar values for other GSLs. Considerably lower values have though also been reported (35). Our result of 1.5 ± 0.4 μM is of the same order of magnitude as the value by Zhai et al. (36), albeit in the range of good agreement. Both values were obtained with very different analytical approaches, and we cannot exclude that the small difference originates from minor systematic errors in one of the techniques. In addition, we note that the value by Zhai et al. was derived from micellar GM1 solutions whereas our study was based on SLBs. The liganding affinity could well be affected by the altered presentation of GM1.

Overall, the comparison of our data with the literature (33,36) demonstrates that the efficiency and mechanism of action of GLTP on SLBs is similar to that on vesicles.

The asymmetric distribution of GM1 in SLBs

Our coincubation assays with EUVs containing the same nominal amount of GM1 as the SLBs (Fig. 3 A) revealed that GM1 distributes highly asymmetrically in silica-supported lipid bilayers, with ∼85% of all GM1 residing in the bulk-facing SLB leaflet, over the entire range of GM1 concentrations investigated (0–5 mol %). Our results extend earlier reports that attractive or repulsive interactions between specific lipids and the support can strongly modulate the interleaflet lipid distribution (3,22–24).

Shreve et al. (25) had previously reported GM1 enrichment in the bulk-facing leaflet of silica-SLBs, although the degree of asymmetry could not be quantified in absolute terms. Both GM1 and silica carry negative charges, and their mutual repulsion could be the driving force for the asymmetric lipid distribution (25). Asymmetric distributions have also been observed for mixtures of PC with other lipids carrying charged headgroups, such as Texas red-labeled phosphoethanolamine (25) or trimethylammonium-propane (22) on silica or glass. For other charged lipids, such as phosphatidylserine, symmetric lipid distributions were though found in the presence of millimolar quantities of calcium ions (3,10). In our case, we mixed 2 mM CaCl₂ into the buffer to facilitate SLB formation (39) but typically investigated GM1 transfer and CTB₅ binding in EDTA-containing buffer, and one may speculate that these subtle changes in the buffer could affect the asymmetry. Comparative measurements in the presence of calcium ions and EDTA, respectively, within a few minutes after SLB formation, and after overnight incubation of the SLB in EDTA-containing buffer, however, did not reveal any significant change in CTB₅ binding (data not shown). We note that GM1 has a rather large headgroup of ∼1 nm in extension and, hence, steric constraints in the cleft between support and membrane may also contribute to the asymmetry.

CTB₅ is a quantitative marker for accessible GM1

Fig. 3 A provides evidence that CTB₅ binds with close-to-maximal stoichiometry to GM1-containing SLBs as long as CTB₅ binding is not limited by space constraints on the membrane surface. CTB₅ can hence be employed as a quantitative probe for the presence of GM1 in the accessible membrane leaflet up to concentrations of ∼8 mol %. This finding is not trivial and indeed remarkable, if one considers that CTB₅ pentamers bind in a multivalent manner to the membrane surface.

In our assays, CTB₅ was typically present in large excess. Earlier work by Lauer et al. (47) indicated that excess CTB₅ would compete for GM1, thereby limiting the average number of GM1 molecules per pentamer. The authors put forward a model that takes into account the “cross-linking” of up to five GM1 molecules by a CTB₅ pentamer. From the model, and the interaction parameters derived by Lauer et al. (47), one would expect that each pentamer binds on average to <1.7 GM1 molecules under the experimental conditions that we have employed in our assays (Fig. 3 C). Our data contradict this prediction and several other implications of the model (see Fig. 3 C for details). The measurements by Lauer et al. were performed at GM1 concentrations ≤0.5 mol %, which is on the lower limit of the detection range of our assay. We suggest that the kinetic parameters derived by Lauer et al. cannot be applied for higher GM1 concentrations. Interestingly, the model appears to be consistent with our data, if the equilibrium cross-linking constant that describes how easily the empty binding sites of CTB₅ can be occupied by the laterally diffusing GM1 is increased by four orders of magnitude (Fig. 3 C). This may imply that cross-linking is limited by mass transport, i.e., the lateral diffusion of GM1 within the SLB, rather than being kinetically limited, as proposed earlier (47).

It has been reported previously that GM1 forms clusters in SLBs of a similar lipid background (POPC instead of DOPC) and on a similar surface (glass instead of silica) and that the fraction of GM1 that is incorporated in these clusters increases in a nonlinear fashion with increasing GM1 proportion in the SLB (56). It is not unlikely that similar clusters do also form under the conditions employed by us. It appears that these clusters do not significantly affect the binding stoichiometry.

Conclusions

We have demonstrated that GLTP is an efficient and robust biochemical tool to quantitatively, specifically, and dynamically modulate the glycosphingolipid content of supported lipid membranes without affecting its overall morphology. Owing to the importance of GSLs in modulating structure and function of membranes and membrane proteins and their selective distribution in membrane domains, this tool may find widespread use in membrane research.

By exploiting an in situ combination of two mass-sensitive and label free techniques (i.e., QCM-D and ellipsometry) and the unique capacity of GLTP to shuttle GSLs between membrane compartments, we showed that GM1 distributes highly asymmetrically in silica-supported lipid bilayers, with most of the ganglioside residing in the bulk-facing leaflet. We do also find that CTB₅ binds with close-to-maximal stoichiometry to GM1 in SLBs over a large range of GM1 concentrations, and quantified the liganding affinity of GLTP for GM1.