SEARCH

SEARCH BY CITATION

Keywords:

  • curvature;
  • Gb3;
  • GM1;
  • Laurdan;
  • lipid sorting;
  • model membrane;
  • Shiga toxin;
  • sphingolipid

Abstract

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

To maintain cell membrane homeostasis, lipids must be dynamically redistributed during the formation of transport intermediates, but the mechanisms driving lipid sorting are not yet fully understood. Lowering sphingolipid concentration can reduce the bending energy of a membrane, and this effect could account for sphingolipid depletion along the retrograde pathway. However, sphingolipids and cholesterol are enriched along the anterograde pathway, implying that other lipid sorting mechanisms, such as protein-mediated sorting, can dominate. To characterize the influence of protein binding on the lipid composition of highly curved membranes, we studied the interactions of the B-subunit of Shiga toxin (STxB) with giant unilamellar vesicles containing its glycosphingolipid receptor [globotriaosylceramide (Gb3)]. STxB binding induced the formation of tubular membrane invaginations, and fluorescence microscopy images of these highly curved membranes were consistent with co-enrichment of Gb3 and sphingolipids. In agreement with theory, sorting was stronger for membrane compositions close to demixing. These results strongly support the hypothesis that proteins can indirectly mediate the sorting of lipids into highly curved transport intermediates via interactions between lipids and the membrane receptor of the protein.

Lipid sorting is a key process in eukaryotic cells required to maintain membrane homeostasis among organelles during intracellular transport (1). In particular, anterograde sorting of sphingolipids and cholesterol generates the very different lipid compositions of plasma membrane and endoplasmic reticulum. Several mechanisms have been proposed for the selection and incorporation of specific lipids during the formation of transport intermediates. The most popular has been the raft hypothesis [(2,3) and (4), for a recent review] in which specific lipids are selected and incorporated into microdomains that subsequently bud to form transport vesicles (5). Work over the last few years has suggested that membrane rafts may be quite dynamic. For example, active non-equilibrium processes such as the contraction of cortical actin can induce microdomain formation and receptor clustering (6). It has also been proposed that the very high curvature of transport intermediates could mediate lipid sorting (7,8).

Several groups have studied how couplings between membrane composition and morphology could contribute to lipid and protein re-organization (9–12). Theoretical modeling indicates that a reduction in membrane bending energy could drive the depletion of sphingolipids and cholesterol from transport intermediates, especially when aided by lipid–lipid interactions (13). This curvature-induced sorting hypothesis was recently tested using giant unilamellar vesicles (GUVs) connected to membrane nanotubes with diameters similar to transport intermediates (13). In accordance with the theory, sphingolipids were significantly depleted from membrane nanotubes when the GUV lipid composition was close to a region of phase co-existence (13–15). Recent experiments on cell membranes have reported that the plasma membrane composition is indeed close to a region of phase co-existence (16–18). Thus, curvature-induced sorting of lipids may account for the depletion of sphingolipids and cholesterol in coat protein I (COPI) vesicles formed during retrograde transport from the Golgi apparatus to the endoplasmic reticulum (19). However, the reduction of membrane bending energy cannot account for the enrichment of sphingolipids and cholesterol in vesicles formed along the anterograde pathway from the trans Golgi network (TGN) to the plasma membrane (20). Clearly, proteins must play an important role not only in forming transport intermediates but also in determining their lipid composition.

The general mechanisms of protein-mediated lipid sorting in curved membranes are likely to be conserved, although individual membrane trafficking pathways use different proteins and receptors. To study these general mechanisms, we used the B-subunit of Shiga toxin (STxB), a toxin produced by Shigella dysenteriae and enterohemorrhagic strains of Escherichia coli that use a clathrin-independent endocytosis pathway. STxB contains multiple binding sites for its glycosphingolipid receptor, Gb3 (globotriaosylceramide), and STxB binding induces the formation of tubular plasma membrane invaginations (21). STxB-induced invaginations have been reproduced in vitro using GUVs containing Gb3, showing that these deformations result from a physical process that does not require external energy or other cytosolic protein machinery. Both theoretical and experimental work suggest that the binding of multiple Gb3 receptors underneath the STxB pentamer induces local membrane compaction and negative spontaneous membrane curvature (21,22). Furthermore, in GUVs containing both liquid disordered (Ld) domains enriched in unsaturated phospholipids and liquid ordered (Lo) domains enriched in sphingolipids, STxB–Gb3 complexes were observed to strongly segregate into the Lo domains (22). In principle, preferential interactions between Gb3 lipids and lipids in the Lo domains could allow STxB to indirectly induce lipid sorting and influence the lipid composition of tubular membrane invaginations. Because Shiga toxin interacts with a specific lipid receptor and can induce membrane deformations, it is an ideal system to examine how proteins can influence the lipid composition of curved membrane structures. The physical mechanisms studied using STxB should apply generally to protein-mediated lipid sorting in curved transport intermediates.

To characterize the effects of STxB lipid distribution, we performed fluorescence microscopy imaging of GUVs containing Gb3, brain sphingomyelin (BSM), cholesterol (Chol), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) and fluorescent lipid analogs [Bodipy-FL-C5-hexadecanoyl phosphatidylcholine (HPC*), Bodipy-FL-C5-ganglioside GM1 (GM1*) and Texas-Red 1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine triethylammonium salt (HPE*)]. Despite its much simpler composition, this model system mimics several key features of biological membranes and has been widely used to study lipid sorting (23,24). Similar to previous studies of cholera toxin and simian virus 40 (25,26), we observed that the binding of STxB to Gb3 receptors could induce domain formation when the GUV membrane composition was close to a region of phase co-existence. Conversely, membrane composition modulated the ability of STxB to induce membrane deformations (22). Finally, and most importantly, the partitioning of fluorescent lipid markers in STxB-induced membrane invaginations was consistent with an enrichment of sphingomyelin (SM) and cholesterol in these highly curved, tubular structures. Furthermore, this lipid sorting was stronger for membrane compositions close to a phase boundary. Thus, we show that the binding of a protein to its membrane receptor can indirectly induce the sorting of other lipid species and produce highly curved membranes enriched in SM and cholesterol. Such a mechanism could account for the enrichment of sphingolipids and cholesterol in vesicles formed along the anterograde pathway.

Results

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

We first present the effects of STxB on GUV domain formation before examining the lipid composition of STxB-induced membrane invaginations.

Effects of Gb3 and STxB on membrane domain formation

The capacity of STxB binding to change lipid distribution was studied using GUVs consisting primarily of BSM, cholesterol and the unsaturated phospholipid, DOPC. For some compositions of this ternary system, the unfavorable interactions between BSM and DOPC cause the formation of co-existing liquid membrane domains enriched in DOPC (Ld phase) and BSM (Lo phase) (27). In GUVs with ∼33% (by mole) cholesterol, domain formation was only observed to occur when the BSM concentration was greater than ∼33% (15,24,28,29). Figure 1A shows a confocal fluorescence image of a GUV with 28.5% BSM (composition BSM28, Table 1) in which the membrane was visualized using 1 mol% HPC*, a fluorescent lipid which normally is enriched in Ld phases (30). As expected, this concentration of BSM was too low to drive phase separation and no domains are visible. However, even in the absence of STxB, the addition of 5% Gb3 can induce domain formation as is evident for the GUV shown in Figure 1B (composition BSM27). Because Gb3 promotes phase separation, less BSM was required to induce domain formation and with 5% Gb3 the phase boundary shifted to ∼27% BSM (Figure 1E).

image

Figure 1. Displacement of phase diagram boundaries because of the addition of Gb3 lipids and binding of STxB. A–D) Equatorial confocal fluorescence images of GUVs labeled with 1% HPC* (green, scale bar 5 µm). A) GUV with no Gb3 and 28.5% BSM (composition BSM28) showing homogenous fluorescence consistent with a single lipid phase. B) GUV with 5% Gb3 and 27% BSM (composition BSM27) with macroscopic domains. C) GUV with less BSM (5% Gb3, 19% BSM, composition BSM21) showing homogenous fluorescence. D) GUV with identical composition (BSM21) following incubation with 200 nm STxB* with macroscopic domains enriched in HPC* (green) and STxB* (red). E) Approximate position of phase boundary for GUVs with 0% Gb3, 5% Gb3 and 5% Gb3+STxB* (see also Table 1). Compositions forming homogenous GUVs are marked with black circles while the compositions forming macroscopic domains are marked with half-filled circles. The compositions corresponding to GUVs shown in panels A–D are indicated along with the PC and SM compositions.

Download figure to PowerPoint

Table 1.  Specific membrane compositions (given as percentage of total by mol)
Composition nameBSMCholDOPCGb3HPC*Visible domains
No STxB200 nm STxB
  1. aExperiments also performed using 1% GM1* in place of HPC*, and also with both 1%GM1* + 0.5%HPE* in place of HPC*.

BSM3333333301NoNo
BSM2828.53337.501NoNo
BSM2727323551YesYes
BSM2121324151NoYes
SMa16324651NoNo
BSM1212325051NoNo
BSM88325451NoNo
PCa0326251NoNo

To determine if STxB binding further enhanced domain formation, GUVs were incubated with a solution containing 200 nm STxB labeled with Cy3 (red; denoted as STxB*). For GUVs containing 21% BSM and 5% Gb3 (composition BSM21), the membrane was uniform in the absence of STxB (Figure 1C). As shown in Figure 1D, the binding of STxB* caused the formation of distinct domains enriched in either HPC* or STxB*. The green HPC* is expected to segregate into the Ld phase, implying that the red STxB* is strongly enriched in the Lo phase. As noted previously, tubular invaginations were not observed for conditions in which STxB binding induced co-existing phases (22). This STxB-induced demixing shows that the binding of the toxin to its membrane receptor can re-organize lipids by promoting phase separation. The effect of STxB is quite strong as domain formation was observed in membranes with as little as 21% BSM (Figure 1E). These experiments show that Gb3 itself can promote domain formation and that domain formation is further promoted by the binding of STxB to Gb3.

Effects of membrane composition on STxB-induced membrane tubules

In addition to promoting lipid demixing, STxB binding can induce the formation of STxB-enriched clusters and tubular membrane invaginations (21). Because membrane curvature can induce lipid sorting (13), we measured the concentration of STxB and several fluorescent lipid markers in tubular invaginations and the flat portions of the GUV membrane. The effect of membrane composition on sorting was studied using two compositions in the Ld phase. The first mixture, phosphatidylcholine (PC), was devoid of BSM and far from the phase boundary. The second mixture, SM, had a composition of 17% BSM placing it relatively close to the phase boundary. Figure 2A,D shows confocal images of representative GUVs corresponding to these lipid mixtures in which the membrane was marked with 1% HPC* and GUVs were incubated in a solution of 200 nm STxB*.

image

Figure 2. Distribution of fluorescent Shiga toxin (red) and HPC* (green) between tubules and GUV, for GUVs with the PC (panels A–C, scale bar 5 µm) and SM (panels D–F, scale bar 5 µm) compositions. Fluorescence from the GUV membrane and tubules were separated using a co-ordinate system centered on the membrane contour (B and E). Tubules occupy 4% of the interior of the PC GUV and 25% of the interior of the SM GUV. The distribution of fluorescence intensity in the membrane tubules is shown in panels C and F. Signals were scaled relative to the fluorescence intensity of the membrane so differences in distribution reflect preferential sorting of STxB* or HPC* and the sorting ratio (SR) defined by eqn (1) is given for each vesicle.

Download figure to PowerPoint

For both PC and SM compositions, STxB binding can induce the formation of membrane tubules. However, GUV composition affects both the number and composition of membrane tubules, and the SM vesicle (Figure 2D) has more membrane invaginations than the PC vesicle (Figure 2A). To quantify this difference, membrane invaginations were identified via a threshold and the fraction of each GUV image occupied by membrane tubules was then calculated (Materials and Methods section, Section 2 of Supporting Information, Figure S1). For the SM vesicles, membrane tubules occupied an average area fraction of 12 ± 2% (n = 43), which was considerably larger than the average area fraction for PC vesicles (2.7 ± 0.4%, n = 26). The increased amount of membrane deformations in SM vesicles might suggest that BSM aids tubule formation. However, tubules were not observed for GUVs with high BSM concentrations (e.g. >33% BSM) and co-existing Lo and Ld phases (22). Thus, tubule formation does not have a simple dependence on membrane composition.

A second clear difference in Figure 2A,D is the relative intensity of the HPC* and STxB* signals from the GUV membrane and in the tubules. This difference is especially clear following transformation to a co-ordinate system centered on the membrane contour (Figure 2B,E). For the SM vesicle (Figure 2D), the HPC* signal is much stronger at the membrane whereas the STxB* signal is much higher in the tubules. To systematically quantify the concentration of STxB* and HPC* in the tubular and flat regions of the GUV membrane, we wrote a matlab program to distinguish these two regions and measure the STxB* and fluorescent lipid intensities in both. This procedure is described further in the Materials and Methods section and Supporting Information. Figure 2C,F shows the distribution of fluorescence intensity in the tubules normalized by the mean intensity at the membrane. The enrichment of STxB* in tubules, relative to HPC*, can then be quantified by the sorting ratio:

  • image(1)

where, inline image, inline image, inline image and inline image are the mean intensities of STxB* and HPC* fluorescence from the tubules and membrane. When STxB* is enriched in tubules relative to HPC* the sorting ratio is greater than 1, and when the sorting ratio is less than 1 STxB* is depleted in tubules relative to HPC*.

Figure 3A displays a histogram of the sorting ratio for SM (n = 42 vesicles) and PC (n = 26 vesicles) compositions, and the distribution statistics are summarized in Table 2. In both cases, STxB* is enriched in tubules (relative to HPC*) but the relative enrichment of STxB* in SM GUVs (3.0 ± 0.3) is twice that for PC GUVs (relative enrichment STxB*/HPC* = 1.5 ± 0.1). Because the sorting ratio describes the concentration of STxB* relative to HPC*, the increased sorting ratio for SM membranes could reflect greater enrichment of STxB*, partial depletion of HPC* or a combination of these two effects.

image

Figure 3. Sorting distributions of lipids and proteins for PC and SM compositions. A) Histogram of Cy3-labeled STxB (STxB*) to HPC* sorting ratio of GUVs measured using the procedure described in the text and eqn (1). Statistical parameters are also reported in Table 2. B) Histogram of STxB* to GM1* sorting ratio of GUVs. C) Histogram of GM1* to HPE* sorting ratio of GUVs incubated with non-fluorescent STxB. D) Variation of GM1*/HPE* sorting ratio as a function of BSM concentration, ×, in the lipid mixtures with 32% cholesterol and 5% Gb3 (see Table 1). The curve shows the trend of increasing GM1*/HPE* sorting close to phase separation (higher BSM concentration).

Download figure to PowerPoint

Table 2.  Effect of membrane composition on lipid sorting
Fluorophore pairPCSM
  1. Mean value of tube sorting ratio for GUVs with PC and SM compositions. The standard error for each is reported along with the number, N, of GUVs analyzed for each condition.

STxB*/HPC*1.5 ± 0.1 (N = 26)3.0 ± 0.3 (N = 42)
STxB*/GM1*2.1 ± 0.1 (N = 26)2.5 ± 0.2 (N = 28)
GM1*/HPE*1.3 ± 0.1 (N = 9)2.0 ± 0.2 (N = 30)

To address this question, we used GM1*, a fluorescent analog of GM1 that has previously been shown to remain in both Lo and Ld phases of this system [(13), Figure S2). Figure 3B shows the sorting ratio for STxB* relative to GM1* measured for PC vesicles (n = 26) and SM vesicles (n = 28). In the tubules, STxB* was enriched relative to GM1* with a sorting ratio of 2.1 ± 0.1 for PC membranes and a slightly higher sorting ratio of 2.5 ± 0.2 for SM membranes. Thus, both the STxB*/HPC* and the STxB*/GM1* sorting ratios are consistent with tubules formed from SM membranes having a higher STxB concentration. In addition, the differences between the HPC* and GM1* sorting ratios (Figure 3A versus B) suggest that HPC* (and thus DOPC) is depleted from membrane invaginations.

HPC* relative depletion from tubules depends on membrane composition

The membrane composition of the tubules can be directly probed by comparing the relative fluorescence of two fluorescent lipid markers. Because HPC* and GM1* are both green, we used HPE*, a lipid with similar phase behavior to HPC* (Figure S2). As shown in Figure 3C, both PC and SM tubules are depleted in HPE* relative to GM1*, as the sorting ratio of GM1*/HPE* is larger than unity. The relative depletion for SM GUVs (sorting ratio 2.0 ± 0.2) is substantially larger than for PC GUVs (sorting ratio 1.3 ± 0.1), suggesting that sorting is stronger when the membrane composition is close to a region of phase co-existence. To test this, we systematically varied the SM/DOPC ratio to produce vesicles with compositions at different distances from the phase boundary (Figure 1E). Figure 3D shows the variation of GM1*/HPE* sorting as a function of SM concentration. Membranes containing more SM were closer to the phase boundary and showed larger relative depletion of HPE*. Because HPE* is excluded from domains enriched in BSM and cholesterol, the depletion of HPE* from tubules implies that tubules are enriched in BSM and/or cholesterol. Thus, STxB can indirectly affect the concentration of lipids through their interactions with Gb3. Lipids with unfavorable interactions with Gb3, such as DOPC, can be depleted from membrane invaginations, whereas lipids with more favorable interactions with Gb3, such as BSM and cholesterol, can be enriched. Furthermore, because this sorting depends on the interaction of all lipid species in the membrane, the enrichment and depletion of lipid species becomes increasingly strong as the membrane composition approaches a phase boundary.

Membrane ordering upon STxB binding to model membranes

The experiments described earlier indicate that STxB*-enriched domains have a different lipid composition that could change membrane organization. The fluorescent probe, Laurdan, has been widely used to probe membrane ordering both in model membranes (31–33) in plasma membrane spheres prepared from cells (34) and in cells (21,35). The wavelength shift of fluorescence emission is characterized by the generalized polarization (GP) coefficient (see Materials and Methods section) (36), which ranges between −1 and 1. Changes in GP value generally correlate with differences in lipid packing (37,38), with lower GP values for less dense packing (disordered phase) and higher GP values for more dense packing (gel phase) (39).

In a previous study, STxB binding was reported to change the GP values of cell membranes (21). To characterize the effect of STxB binding in controlled conditions, we performed similar experiments on GUVs with the PC and SM compositions. Polarization effects arising from the geometry of tubules were avoided by using tense GUVs, as these blocked the tubulation while still allowing the formation of toxin clusters (Figure 4). GP values for PC and SM membranes with and without STxB* are presented in Figure 4G. In accordance with published results, the incorporation of BSM increased the GP coefficient (40,41). Furthermore, for both PC and SM GUVs, the GP coefficient of STxB clusters was slightly higher than the GP coefficient in GUVs without STxB (Figure 4E–G and Table S1). Although the mean lipid composition of the SM GUVs should be similar to the mean lipid composition of the external plasma membrane leaflet (18), the GP values for GUVs are substantially lower than those measured previously with cell membranes (0.49 in the absence of STxB* and 0.57 in the presence of STxB*(21), in the ordered phase). Interpreting this difference in absolute GP value is difficult, although, because of the differences in membrane geometry and the presence of integral proteins in the plasma membrane. Nevertheless, for both cells and GUVs, the binding of STxB caused an increase in GP value, which is consistent with the hypothesis that STxB induces more ordered lipid packing than the surrounding membrane.

image

Figure 4. Laurdan fluorescence from GUVs. A) Pseudo-color representation of GP values of a PC GUV in the absence of STxB* (Scale bar, 5 µm). B) GP image of SM GUV in the presence of STxB*. C) STxB* channel for the vesicle shown in (B) reveals visible STxB clusters. D) GP image corresponding to (B) using a mask selecting the STxB clusters to calculate GP values in the presence of STxB. E) Histogram of GP values of image (A). F) Histogram of GP values of image (D). G) Summary of mean GP values for PC and SM GUV compositions in the presence and the absence of toxin.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

The interaction of STxB with its Gb3 receptors can have multiple effects on membrane organization changing both membrane shape and lipid distribution.

One striking capacity of STxB is its ability to induce the formation of domains in previously homogeneous membranes (Figure 1). STxB binding can strongly promote domain formation because the effect is even observed for GUVs with compositions relatively far from a region of phase co-existence (e.g. up to 7% lower BSM concentration). When domains formed, STxB* segregated into the domains depleted in HPC*, suggesting that the Gb3–STxB complex interacts preferentially with BSM and/or cholesterol. A preference of Gb3 for sphingolipids is also indicated by the observation that the addition of Gb3 to the membrane reduces the amount of BSM needed to induce domain formation (Section 4.3 of Supporting Information). While the interactions of lipid molecules are complex, it has been shown that sphingolipids and glycosphingolipids, such as SM and Gb3, can interact preferentially through hydrogen bonding at glycosidic and sphingosine moieties (42). Furthermore, it is likely that both of these lipids also have preferential interactions with cholesterol, as cholesterol is essential for the existence of Lo phases. Other intramembrane interactions, such as hydrophobic mismatch, could also play a role. For example, it has been proposed that the MAL protein involved in glycosyl-phosphatidylinositol-anchored protein sorting induces clustering of lipids via a shared preference for thicker raft domains (43). Such effects may account for the apparent preferential interactions of the Gb3–STxB complex with sphingolipids and cholesterol.

STxB's capacity to induce domain formation is similar to the reported effect of cholera toxin B-subunit (CTxB) binding to its ganglioside receptor, GM1, in synthetic membranes (25) and plasma membrane spheres (18). Domain formation can only occur when the interactions between membrane components outweigh the entropic cost of sorting lipids into the domains. Both STxB and CTxB can change this balance because they have multiple binding sites for their receptors [5 GM1 binding sites per CTxB molecule (44) and up to 15 Gb3 sites per STxB (45)]. Bound receptors move collectively as part of a protein–receptor complex, which greatly reduces the entropic penalty for sorting (13,46). Thus, the ability of STxB to induce domain formation may result from its ability to bind multiple Gb3 receptors (Section 4.3 of Supporting Information). This effect is likely to be important for transport and signaling in the cell, because it has been shown that plasma membranes can have a composition close to a phase boundary (16–18,47). Consequently, proteins that bind multiple receptors should be able to readily change lipid distribution to form highly dynamic domains or rafts (4,46).

A second important result is the observation that the highly curved membrane invaginations formed upon STxB binding are enriched in lipids that do not directly interact with STxB. For all compositions, STxB and Gb3 are strongly enriched in membrane invaginations while the enrichment and depletion of other lipids are indicated by several lines of evidence. First, in membrane invaginations the fluorescent lipid HPE* was clearly depleted relative to GM1*. Importantly, this sorting strengthened when the membrane composition was closer to the phase boundary (Figure 3D), suggesting that sorting resulted from collective interactions between multiple components in the membrane. In addition, the very large STxB*/HPC* sorting ratio in SM vesicles is consistent with a depletion of HPC* from tubules. As both HPE* and HPC* are excluded from phases enriched in BSM [Figure S2 and the work of Sorre et al. (13)], the sorting of these fluorescent markers strongly suggests that the membrane invaginations are enriched in BSM and depleted in DOPC. As described below, the observed sorting of proteins and lipids in tubules can be accounted for by considering the preference of STxB for curved membranes and the interactions between Gb3 and surrounding lipids.

Several factors should drive the enrichment of STxB in membrane invaginations. For both the PC and SM membrane compositions, STxB spontaneously forms aggregates indicating an effective attractive interaction between Gb3–STxB complexes (21). This aggregation of STxB is independent of membrane curvature, as it occurs even when the membrane tension is high and tubular invaginations cannot form (Figure 4). Because these STxB aggregates spontaneously induce inward curvature, they should have a lower energy in inwardly curved membranes (21). This preference would naturally concentrate STxB aggregates in invaginations, but individual STxB–Gb3 complexes may also be synergistically concentrated by the inward curvature of the membrane invaginations. Unlike lipids, membrane proteins are large enough that their binding energy can change significantly depending on the curvature of the membrane. Consequently, proteins can be strongly enriched or depleted in regions with different curvatures (10,13,48–51). Using a simple theoretical model (Section 4 of Supporting Information), we have found that a preference of STxB–Gb3 complexes for inward curvature could contribute to STxB enrichment in tubes and that this enrichment of STxB should increase significantly when the membrane composition is close to a region of phase co-existence (Figure S5C). This suggests that in addition to preformed STxB aggregates forming tubes, individual STxB–Gb3 complexes could also be enriched in membrane invaginations because of their curvature preference. This curvature-dependent enrichment of STxB is supported by measurements performed on membrane tubes pulled from GUVs containing 5% Gb3 (Section 1 of Supporting Information, Figure S3). STxB was not detectable on outwardly curved nanotubes when the tube diameter was less than 140 nm (Figure S3), showing that STxB was strongly depleted from outwardly curved membranes. This observation supports the intriguing hypothesis that the membrane curvature induced during the formation of transport intermediates could directly enrich these vesicles in proteins that prefer the corresponding membrane curvature.

The sorting of lipid species in membrane invaginations involves quite different considerations. Because of its direct interactions with STxB, the Gb3 receptor is enriched in membrane invaginations. The interactions of Gb3 with surrounding lipids should then favor a depletion of DOPC and enrichment of BSM. This hypothesis is supported by our theoretical model (Figure S4), in which repulsive interactions between Gb3 and DOPC cause BSM to be ‘dragged along’ by the STxB concentrating in tubes. Because the favorable enthalpy for cosorting STxB and BSM increases with BSM concentration, the cosorting of BSM should increase with BSM concentration and become quite large for compositions close to the phase boundary (52) (Figure S5D). This prediction matches well to the observed increase in GM1*/HPE* sorting ratio for membranes with higher BSM concentrations (Figure 3D).

Conversely, it has been suggested that the bending energy of highly curved membranes favors PC lipids, and in a previous study of GUVs with similar compositions, it was reported that tubes formed from vesicles with a membrane composition close to a phase boundary were enriched in DOPC (13). However, simple theoretical estimates suggest that interactions between lipids can often be more important than the bending energy of the membrane (Section 4.6 of Supporting Information). For this system, the interactions between Gb3 and the other lipids should dominate and indeed STxB membrane invaginations appear to be depleted in DOPC and enriched in BSM. Thus, protein binding can lead to the formation of transport intermediates enriched in sphingolipids and cholesterol, even if the vesicle membrane is highly curved. The effects of protein binding should be strongest when the composition is relatively close to the critical point (where perturbations produce the largest composition changes).

To test the more physiological PC species POPC (palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) and egg-PC, we identified three potential systems (POPC/BSM/ Chol, POPC/Egg-SM/Chol and Egg-PC/BSM/Chol) and tested a range of compositions for each (Table S2). For the POPC systems we did not find compositions that formed macroscopic Lo/Ld phases, while for the mixture of PC lipids (Egg-PC) the interactions were too strong and co-existing domains formed even at very low BSM fractions. Thus, none of the compositions tested in these systems matched the capacity of membranes derived from cells (18) and the widely used 1:1:1 DOPC/BSM/Chol canonical ‘raft mixture’(28) to form macroscopic co-existing liquid domains in response to small perturbations.

The effects of STxB binding on membrane structure and ordering were probed using Laurdan fluorescence. The slightly larger GP values for STxB-enriched domains are consistent with STxB inducing a more ordered/denser packing of lipids, as previously inferred from AFM and fluorescence measurements (22). These results also agree with previous Laurdan measurements on HeLa cells in which the presence of STxB was reported to increase the ordering of the plasma membrane (21). Although the membrane composition of cells is more complex than GUVs, in both cases STxB binding increases the GP of Laurdan.

In conclusion, proteins that have a multivalent interaction with membrane receptors can cause large changes in membrane composition and shape (25). The membrane curvature preference of protein–receptor complexes is able to induce membrane deformations and protein enrichment. Furthermore, protein–receptor complexes in highly curved membranes can induce the cosorting of lipids that interact preferentially with the membrane receptor. This lipid enrichment becomes stronger when the membrane composition is in the vicinity of a phase transition, as has been shown to be the case for some cell membranes (16–18,47). Importantly, protein-induced lipid cosorting can be strong enough to drive the enrichment of lipids that increase membrane bending rigidity (e.g. sphingolipids and cholesterol). For example, membrane invaginations induced by the binding of STxB appear to be enriched in BSM and depleted in DOPC.

Multiple sorting mechanisms are likely to be important in intracellular trafficking. For example, the minimization of membrane bending energy may account for the formation of curved transport intermediates enriched in PC lipids, such as COPI vesicles (19). However, protein-induced lipid cosorting is a general way in which curved transport intermediates could be enriched (or depleted) in sphingolipids and sterols through the presence of specific proteins and membrane receptors. This mechanism may account for the enrichment of sphingolipids and sterols in the transport intermediates between TGN and plasma membrane (20), and the identification of clustering proteins and receptors is now required.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Materials

DOPC, BSM and Chol were purchased from Avanti Polar Lipids. HPC*, GM1*, HPE* and 6-lauroyl-2-dimethylaminonaphthalene (Laurdan) were purchased from Molecular Probes. Gb3 lipids were purchased from Matreya LLC. STxB was purified as previously described (53). GUVs were grown using the electroformation technique (54) on conductive indium tin oxide [ITO] coated glass slides (55) and lipid compositions are summarized in Table 1. Briefly, lipid mixtures were dissolved in chloroform at a final concentration of 0.5 mg/mL. Fifteen microliters of the solution was spread onto each ITO slide and the lipid films were then dried under vacuum for at least 2 h. GUVs were grown in a 300 mOsm sucrose solution by applying an alternating electric field (1.1 V RMS, 10 Hz) for 4 h. To prevent phase separation, GUVs containing BSM were grown at 60°C. For microscopy experiments, GUVs were transferred into a chamber containing 200 nm of STxB in PBS buffer at 300 mOsm glucose.

Microscope setup

Confocal images were taken either on a Zeiss LSMMETA microscope (63× objective) or a Nikon CS1 microscope (100× objective). Green-emitting dyes were excited with a 488 nm laser and Cy3 dyes were excited at 543 nm (Zeiss) or 561 nm (Nikon). Laurdan experiments were performed with the Nikon microscope. Laurdan was excited at 408 nm, and emission was measured with two band pass filters of 455 ± 15 and 515 ± 30 nm. Cy3 STxB was excited at 543 nm and the emission measured with a band pass filter of 605 ± 35 nm. To avoid any effects because of cholesterol oxidation produced by prolonged, strong HPC* photo-activation (15,56), fluorescence illumination intensity/duration was limited. Experiments on membrane nanotubes (Section 1 of Supporting Information) were performed on a home-made setup combining confocal microscopy, optical tweezers and micropipette aspiration as described in the work of Sorre et al. (13).

Measurement of sorting coefficients

Confocal fluorescence images of GUVs were analyzed using Matlab and the procedure for measuring sorting coefficients is described in detail in Section 2 of Supporting Information. Briefly, for each GUV, the membrane contour was fitted and the image then transformed into a co-ordinate system centered on this contour, as shown in Figure 2B,D. The effective ‘width’ of the GUV membrane was then estimated using the half-width-half-maximum (HWHM) of the membrane profile, w. Pixels at the exterior of the GUV (distances, t, from the membrane contour of 2w <|t| < 6w) were used to estimate the mean background level (inline image) and noise (σX) of each fluorescence channel. To limit the mixing of fluorescence from the GUV membrane and tubules, the mean fluorescence of the GUV membrane, inline image, was determined by averaging the intensity for pixels in the range |t|< w. Membrane tubules in the interior of the GUV were identified by selecting only pixels with STxB* fluorescence greater than the threshold, inline image, and which were sufficiently far from the membrane |t| > 2w. This thresholded region was then used to calculate the area fraction occupied by membrane tubules, the mean fluorescence signal for each fluorophore (inline image) and the sorting ratio defined by eqn (1).

Measurement of GP values

Laurdan fluorescence images were acquired using the Nikon confocal microscope described above with an acquisition time of 0.8 s per image. To prevent photoselective excitation, the polarization of the excitation laser was scrambled by placing a Nomarski prism (DIC slider) just below the objective. GUVs were imaged in the equatorial region and no fluorescence modulation along the contour was detected (Figure 4A,B). Laurdan experiments are often performed with a two-photon microscope because of the high sensitivity of Laurdan to photobleaching. However, measured GP values were constant for at least for 10 consecutive images which was considerably more than the 2–5 images needed to reliably measure the GP value of a vesicle.

The excitation of Laurdan probe was made at λexc = 408 nm with a laser power between 50 and 100%. The two emission channels correspond to at λem = 455and 515 nm. The GP is defined by (36):

  • image(2)

As the Laurdan analysis depends on experimental conditions such as gain values and laser power, we used a correction G for GP defined as follows (39):

  • image(3)

where, GPth is the theoretical GP value of Laurdan in DMSO at 22°C and GPexp is the corresponding experimental value. The corrected value GPcorr is then defined by:

  • image(4)

In the text, GP always refers to the corrected value, GPcorr. Details of the image analysis are given in Section 3 of Supporting Information.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

We thank Katharina Gaus and Luis Bagatolli for helpful discussions, Luc Fettler for help with Laurdan experiments, Sandrine Morlot and Aurélien Roux for the use of their confocal microscope and Ellen Batchelder and Timo Betz for helpful comments on the manuscript. We also thank the Cell and Tissue Imaging Platform (UMR 144, Institut Curie). This work was supported by European Commission (NoE SoftComp) and the Human Frontier Science Program Organisation. The groups belong to the French research consortium ‘CellTiss’. L. B. and B. S. have been supported by a grant from the Direction Générale pour l’Armement, B. S. by the Fondation pour la Recherche Médicale, A. C.-J. by the Association pour la Recherche contre le Cancer and G. T. by a Marie Curie post-doctoral fellowship from the European Community.

References

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Table S1: GP values for PC (N = 6) and SM (N = 12) GUVs in the presence and the absence of STxB*. Uncertainties were determined from the standard deviation of the population. The larger GP value for the SM composition is expected because this composition has more BSM and consequently the lipids should have a more ordered packing.

Table S2: List of the systems and compositions investigated to identify regions of Lo/Ld co-existence. ESM and BSM are egg sphingomyelin and brain sphingomyelin, respectively.

Figure S1: Histogram of percentage of GUV cross-sectional area occupied by tubules for PC (N = 26) and SM (N = 43) GUVs. The mean cross-sectional area for PC vesicles was 2.7 ± 0.4% whereas for SM vesicles it was 12.2 ± 1.7%.

Figure S2: Lipid distribution in the presence of 200 nmof STxB in GUVs containing 5% Gb3, BSM, Chol, DOPC (33:33:33 molar fraction), Bodipy-FL-C5-ganglioside fluorescent analog of GM1 (1%, GM*, green), and Texas-Red DHPE* (0.5%, DHPE*, red). DHPE* strongly segregates in the Ld phase [fluorescence ratio between Ld and Lo (, N = 6 vesicles)] while GM1* has similar concentrations in the Ld and Lo phases (, N = 6 vesicles), similarly to previous observations (1). The scale bar is 5 µm.

Figure S3: Exclusion of STxB* from membranes with large positive curvature. A) Confocal equatorial section of a PC GUV with HPC* fluorescence clearly visible for a tube pulled with an optical tweezer. B) Same vesicle in the STxB* channel. No fluorescent toxin could be detected on the tube. C) STxB*channel for a membrane tube of radius of 28 nm (no HPC* in membrane) shows no detectable toxin on the tube. D) Same GUV with a membrane tube of 68 nm radius in which toxin could just be detected on the tube. In all images the scale bar is 5 µm and the position of the bead is outlined with a white circle.

Figure S4: Image analysis procedure. A) Original image of GUV (‘PC’ composition) marked with fluorescent STxB* (red) and HPC* (green). The position of the membrane contour is shown in white (scale bar 5 µm). B) Transformation to co-ordinate system centered on membrane contour. The exterior, membrane and interior regions are marked by dotted white lines. Membrane invaginations were identified using the threshold, as indicated by the blue shading. Tubes occupy approximately 3.9% of the area of the GUV cross-section, or 0.18 µm2/µm of vesicle contour. C) Lipid fluorescence signal as a function of distance from the membrane center. The HWHM on the exterior side of the peak is w = 160 nm. This HWHM was used to define the membrane (|t| < w), exterior (−6w < t < 2w) and interior (t > 2w). D) Histogram of the red fluorescence intensity outside the GUV (−6w < t < −2w). Signals have been normalized to the median intensity in the membrane region. The median is , while the distance to the 16 percentile is σl = 0.044 and distance to 84 percentile is σr = 0.088.

Figure S5: Theoretical model of protein and lipid sorting. A) Effect of Gb3 interactions on phase behavior. The shift in the phase boundary () upon Gb3 addition is shown as a function χ12, the parameter describing the interactions of Gb3 with BSM and DOPC. The addition of Gb3 promotes phase separation when χ12> χ11 which occurs when Gb3 has a strongly unfavorable interactions with DOPC (and/or strongly favorable interactions with BSM). The calculation was made using Eq. 13 for x = 1 and χ11 = 2.1, for which () (i.e. intermediate degree of phase separation). B) Effect of STxB clustering × Gb3 receptors on the shift in the phase boundary upon addition of Gb3 () for the case of intermediate phase separation (χ11 = χ12 = 2.1). The reduction in mixing entropy following receptor clustering strongly promotes phase separation. C) Enhancement of enrichment of STxB in tubes by non-ideal mixing. The y-axis shows the sorting of STxB relative to sorting in the low concentration limit (lim φ2 [RIGHTWARDS ARROW] 0) as a function of BSM concentration (relative to BSM concentration at the onset of phase separation; φ11a). Blue = repulsive interactions between BSM and DOPC (χ11 = 2.1, χ12 = 1.05, χ22 = 0, φ2 = 0.05, x = 1); Red = repulsive interactions between Gb3 and DOPC (χ11 = 0, χ12 = 1.05, χ22 = 2.10, φ2 = 0.05, x = 1); Black = BSM and Gb3 with repulsive interactions for DOPC (χ11 = χ12 = χ22 = 2.1, φ2 = 0.05, x = 1); Green = BSM and Gb3 with repulsive interactions for DOPC and clustering of Gb3 receptors (χ11 = 2.1, χ12 = 1.29, χ22 = 0.48, φ2 = 0.05, x = 15). When Gb3 and BSM have repulsive interactions with DOPC, cosorting can significantly enhance the enrichment of STxB by curvature. D) Dependence of Gb3/BSM cosorting on membrane composition. The y-axis shows the enrichment of BSM in the tubes (/) as a function of BSM concentration (relative to BSM concentration at the onset of phase separation; φ11a). Sorting approximated by Eq. 21 is shown in blue, while a numeric minimization of Eq. 2 (Supporting Information) is shown in black (, , χ11 = χ12 = χ22 = 2.1).

FilenameFormatSizeDescription
TRA_1116_sm_SuppInfo.pdf768KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.