The Vesicle- and Target-SNARE Proteins That Mediate Glut4 Vesicle Fusion Are Localized in Detergent-insoluble Lipid Rafts Present on Distinct Intracellular Membranes*

Insulin stimulates the fusion of intracellular vesicles containing the glucose transporter Glut4 with the plasma membrane in adipocytes and muscle cells. Glut4 vesicle fusion is thought to be catalyzed by the interaction of the vesicle solubleN-ethyl-maleimide-sensitive fusion protein attachment protein receptor VAMP2 with the target solubleN-ethyl-maleimide-sensitive fusion protein attachment protein receptors SNAP-23 and syntaxin 4. Here, we use combined membrane fractionation, detergent solubility, and sucrose gradient flotation to demonstrate that the large majority (>70%) of SNAP-23 and a significant proportion of syntaxin 4 (∼35%) are associated with plasma membrane lipid rafts in 3T3-L1 adipocytes. Furthermore, VAMP2 is shown to be concentrated in lipid rafts isolated from intracellular membranes. Insulin stimulation had no effect on the plasma membrane raft association of SNAP-23 or syntaxin 4 but promoted VAMP2 insertion into plasma membrane rafts. Immunofluorescence analysis revealed that SNAP-23 was clustered at the plasma membrane and almost completely segregated from the transferrin receptor. SNAP-23 distribution seemed to be distinct from caveolin-1, and clusters of SNAP-23 were dispersed after cholesterol extraction with methyl-β-cyclodextrin, suggesting that the majority of SNAP-23 is associated with non-caveolar, cholesterol-rich lipid rafts. The results described implicate lipid rafts as important platforms for Glut4 vesicle fusion and suggest the hypothesis that such rafts may represent a spatial integration point of insulin signaling and membrane traffic.

Efficient intracellular membrane fusion requires SNARE 1 proteins on donor and acceptor membranes. In vitro assays have shown that SNARE proteins are sufficient to cause fusion of artificial membranes (1); in vivo, however, membrane fusion is dependent upon a large number of other proteins (2,3). Current models suggest that the formation of trans-SNARE complexes between two membranes may be the catalyst for subsequent membrane fusion. SNARE complex formation may cause structural changes in the transmembrane regions of the proteins, leading to membrane distortion and perhaps facilitating lipid mixing. Indeed in vitro fusion assays and studies in yeast have demonstrated the importance of SNARE transmembrane domains for efficient membrane fusion (4,5). Thus, the local lipid environment of SNARE proteins is likely to be of fundamental importance for effective membrane fusion.
Recent work has shown that SNARE proteins (syntaxin 1A, SNAP-25, and VAMP2) are associated with specialized cholesterol-rich membrane domains in PC12 cells and that cholesterol depletion reduced the extent of regulated exocytosis from these cells (6,7). We found that syntaxin 1A and SNAP-25 were highly enriched in cholesterol-dependent Triton-insoluble rafts (6). In addition, Jahn and co-workers showed that these tSNAREs were clustered in cholesterol-rich domains of the plasma membrane, although they could not isolate them in detergent-insoluble rafts (7). The interaction of tSNAREs with cholesterol-rich membrane domains may reflect the affinity of syntaxin 1A for cholesterol (7), and targeting of syntaxin 1A to Triton X-100-insoluble rafts has been shown to require coexpression of nSec1 (8).
Insulin stimulates glucose transport in muscle and adipose cells by inducing the translocation of intracellular Glut4-containing vesicles to the plasma membrane (PM) (reviewed in Ref. 9). The fusion of these vesicles with the PM involves the tSNAREs syntaxin 4 and SNAP-23 and the vesicle SNARE VAMP2 (reviewed in Ref. 9). The possibility that these SNARE molecules may be localized to lipid rafts offers a potential link to at least some molecules in the insulin signaling cascade, such as Cbl and the GTPase TC10, which function within lipid rafts in response to engagement of the insulin receptor and play a key role in Glut4 translocation (10 -12).
A further rationale for the study of SNARE molecule localization in adipocytes is that, unlike PC12 cells, well characterized fractionation procedures are available for adipocytes, allowing rafts to be isolated from distinct cellular compartments, thus allowing an analysis of where the raft-associated SNAREs reside-a feature absent from recent studies in PC12 cells (6). In this study, we showed that the large majority of PM SNAP-23 was raft-associated and that a significant amount of syntaxin 4 was also recovered in PM raft fractions. In addition, intracellular VAMP2 was shown to be largely raft-associated. These results are the first to show that target-and vesicle-SNARE proteins are associated with lipid rafts present on opposing membranes (i.e. PM versus intracellular membranes). Such data offer the hypothesis that the interface of insulin signaling and membrane trafficking may localize to raft domains in adipocytes, serving to spatially link the two processes.

EXPERIMENTAL PROCEDURES
Materials-Antibodies recognizing SNAP-23 and VAMP2 were from Synaptic Systems (Gottingen, Germany). Syntaxin 4 antibody was purchased from Chemicon (Temecula, CA). Polyclonal and monoclonal caveolin-1 antibodies were from BD Biosciences (Oxford, UK). Transferrin receptor antibody was purchased from Cambridge Bioscience (Cambridge, UK). Cellubrevin antibody was from Abcam (Cambridge, UK). Insulin-responsive aminopeptidase (IRAP) antibody was a gift from Drs. Luis Garza and Morris Birnbaum (University of Pennsylvania, Philadelphia, PA), and cysteine-string protein (CSP) antibody was from Professor Robert Burgoyne (University of Liverpool, Liverpool, UK). Triton X-100, methyl-␤-cyclodextrin, and all other reagents were from Sigma.
Cell Culture-3T3-L1 fibroblasts were grown and differentiated as outlined previously (13), and used for experiments 9 to 12 days after differentiation and between passages 5 and 12. All cells used in insulinstimulation experiments were preincubated for at least 2 h in serumfree medium.
Triton X-100 Solubilization and Sucrose Gradient Centrifugation-Adipocytes (ϳ 40 ϫ 10 6 /ml) were lysed in 2 ml of MBS (25 mM MES and 150 mM NaCl, pH 6.5) containing 1% Triton X-100 and supplemented with a protease inhibitor mix (Roche Molecular Biochemicals). The samples were then incubated at 4°C for 20 min with end-over-end rotation. The solubilized cells were homogenized with 10 strokes of a Dounce homogenizer, and 1.5 ml of the homogenate was added to an equal volume of 80% (w/v) sucrose in MBS. The solubilized cells (in 40% sucrose) were placed at the bottom of a centrifuge tube and overlaid successively with 6 ml of 30% sucrose and 4 ml of 5% sucrose (in MBS). After centrifugation at 240,000 ϫ g in a Beckman SW40 rotor for 18 h, 1-ml fractions were collected from the top of the gradient (designated fractions number 1 (top) through 13 (bottom)) and immediately supplemented with protease inhibitors. The pellet was resuspended by Dounce homogenization in 1 ml of MBS and designated fraction 14.
The procedure used for isolation of rafts from purified PM or lowdensity microsomal membranes was similar to that used for whole cells, except membranes from 40 ϫ 10 6 adipocytes were solubilized in 1 ml of 0.2% Triton X-100 in MBS/protease inhibitors. Solubilized membranes were homogenized, made up to 1.5 ml with MBS, and mixed with 1.5 ml of 80% sucrose (w/v). Gradients were run as above. Equal volumes of all recovered fractions were separated by SDS-PAGE and transferred to nitrocellulose for immunoblotting analysis.
Immunofluorescence Analysis of PM Lawns-Plasma membrane lawns were prepared as described previously (15). Paraformaldehydefixed membranes were incubated in phosphate-buffered saline, 5% gelatin, and 1% goat serum and then with antibodies against SNAP-23, transferrin receptor, VAMP2, or caveolin-1 (all 1:50) at 4°C overnight. The membranes were washed and incubated with secondary antibodies conjugated to fluorescein isothiocyanate or tetramethylrhodamine B isothiocyanate for 2 h at room temperature. The coverslips were mounted and analyzed by confocal microscopy using either a Zeiss 410 LSM on an Axiovert inverted microscope or a Zeiss Pascal on an Ax-ioSkop2 upright microscope using X63 1.4-numerical aperture plan apochromat objectives (Carl Zeiss Ltd., Welwyn Garden City, UK). Cells labeled with red fluorophores were excited at 543 nm, and a 590-nm long pass filter was used for collecting emissions. Green fluorophores were excited using an argon laser emitting at 488 nm with a 505-520-nm band pass detection filter. The fraction of crossover was determined to be less than 3% by using single labeled samples of each fluor.

RESULTS
Adipocytes (40 ϫ 10 6 /ml) were solubilized in 1% Triton X-100 and centrifuged in a discontinuous sucrose gradient. This procedure separates solubilized proteins from insoluble lipid rafts, as shown by the separation of caveolin (a marker protein for rafts) and the transferrin receptor ( Fig. 1). Solubilized proteins remain in the 40% sucrose layer, whereas insoluble rafts accu-mulate at the interface of the 5 and 30% sucrose layers. Further analysis of recovered fractions revealed that the tSNARE SNAP-23 was an abundant raft protein, with an average of 68.6 Ϯ 15.5% (n ϭ 3) of the protein present in raft fractions (Fig. 1). The level of syntaxin 4 in rafts was 21.7 Ϯ 8.8% (n ϭ 3). The vesicle SNARE VAMP2 was also enriched in lipid raft fractions, with an average of 36.4 Ϯ 17.8% (n ϭ 3) raft associated (Fig. 1). Note that as with PC12 cells (6), the recovery of VAMP2 in rafts isolated from whole adipocytes was somewhat variable. In contrast to the abundance of SNARE proteins in adipocyte lipid rafts, no CSP or IRAP was detected in raft fractions (Fig. 1). CSP and IRAP serve as controls to indicate the efficient solubilization of both PMs and intracellular membranes, respectively (16,17).
Although the data reported above demonstrate the presence of SNARE proteins in adipocyte lipid raft domains, it is not clear which cellular membranes these isolated rafts are derived from. This is also true of our recent work showing raft association of SNAREs in PC12 cells (6). 3T3-L1 adipocytes can be readily subfractionated by differential centrifugation. Using this approach, we purified a PM fraction enriched in the Na,K-ATPase (Fig. 2) and a light membrane fraction (LDM) which was enriched in Glut4 vesicles and also contains markers of the Golgi, trans-Golgi network, and endosomes ( Fig. 2) (14,18,19). As expected, the tSNAREs syntaxin 4 and SNAP-23 are enriched in isolated PM fraction, and VAMP2 is enriched in the LDM fraction, consistent with its association with Glut4 vesicles (15). Note that insulin promotes a redistribution of Glut4 from LDM to PM fractions, representing insulin-stimulated fusion of Glut4 storage vesicles with the PM. A more modest increase in PM-associated VAMP2 is also observed after insulin stimulation, consistent with other studies in this cell type (15).
For preparation of lipid rafts from isolated plasma membranes, test experiments were performed to determine the optimal concentration of Triton for isolation of SNARE-containing rafts. Because SNAP-23 is largely raft-associated, this protein was used to monitor purification of SNARE-containing rafts. Because purified plasma membranes contain ϳ 20-fold less protein than total cellular membranes, we examined the purification of rafts using 0.2, 0.5, and 1.0% Triton. In test experiments, it was found that 0.2% Triton yielded a consistent recovery of SNAP-23-containing rafts. Increasing the Triton concentration from 0.2 to 1.0% markedly decreased the recovery of SNAP-23 in raft fractions (Fig. 3A). Caveolin-1 was also sensitive to increased Triton concentration; raft recovery of this protein decreased significantly at the higher Triton/protein ratio (see Fig. 3A). Note that the Triton/protein ratio at a detergent concentration of 0.2% is ϳ7:1 (milligrams). The protein profiles of isolated PM raft and non-raft fractions is compared in Fig. 3B, showing that these two sets of proteins have distinct protein profiles, demonstrating that the procedure isolates two distinct sets of proteins from adipocyte plasma membranes.
To extend the results of Fig. 1, raft association of SNARE proteins was examined in the purified PM fractions of basal and insulin-stimulated cells. As with experiments performed on whole cells (Fig. 1), purification of rafts from PMs successfully separated caveolin-1 and the transferrin receptor in raft and non-raft fractions, respectively (Fig. 4A). SNAP-23 (71.5 Ϯ 4.6%, n ϭ 7) and syntaxin 4 (35.2 Ϯ 3.5%, n ϭ 5) were both abundantly present in PM lipid rafts. We did not detect any effect of insulin stimulation on the raft association of these PM SNAREs, suggesting that tSNAREs are stably associated with rafts. However, we were able to detect an insulin-stimulated recruitment of VAMP2 to PM rafts, presumably because of Glut4 storage vesicle fusion with the PM (Fig. 4B). Note that despite the presence of a low level of VAMP2 in purified plasma membrane fractions (see Fig. 2), the immuno-detection of VAMP2 in gradient fractions isolated from basal cells proved inconsistent. In the experiment shown in Fig. 4B, we did not detect VAMP2 staining in any fraction from these gradients performed on PM fractions from basal cells. In other experiments, prolonged exposure of the immunoblots revealed VAMP2 present in PM raft fractions from basal cells (data not shown). This inconsistency presumably reflects the low levels of VAMP2 present at the cell surface in the absence of insulin and the dilution of the PM fractions upon the gradient. However, the presence of VAMP2 in raft fractions from insulinstimulated cells was a clear and consistent observation.
Because VAMP2 associates with PM lipid rafts (Fig. 4B), it was interesting to examine whether VAMP2 is also associated with intracellular rafts. Therefore, rafts were isolated from LDMs using a protocol similar to that used for preparation of PM rafts. VAMP2 was found to be almost exclusively present in lipid raft fractions (74.6 Ϯ 4.6%, n ϭ 3) (Fig. 4C). Interestingly, the vesicle SNARE cellubrevin (VAMP3) exhibited a much lower level of raft association (15.2 Ϯ 8.8%, n ϭ 3) than VAMP2. The protocol used was sufficient to solubilize the transmembrane proteins IRAP and the transferrin receptor, proteins that colocalize with intracellular VAMP2 and cellubrevin, respectively, arguing strongly that the raft association is not the result of incomplete solubilization.
Recent work (20,21) has suggested that Glut4 interacts with lipid rafts or caveolae at the PM. We found that Glut4 was present in PM raft fractions (37.5 Ϯ 6.1% raft-associated, n ϭ 3) (Fig. 4B). Insulin stimulation caused a broadly similar increase in Glut4 immunoreactivity in both raft and non-raft fractions of the PM (Fig. 4B). In contrast to its significant presence in PM raft fractions, Glut4 was mainly excluded from intracellular lipid rafts (Fig. 4C).
To examine the segregation of SNARE proteins into membrane microdomains in more detail, we analyzed SNAP-23 distribution in PMs using immunofluorescence labeling. Plasma membrane sheets derived from 3T3-L1 adipocytes were prepared and stained with antibodies against SNAP-23 and the transferrin receptor (a non-raft protein). Fig. 5A shows that both SNAP-23 and the transferrin receptor were present in discrete clusters that rarely overlapped. Note that SNAP-23 did not display the typical circular rosette structure of caveolar proteins (10) (Fig.  5B), implying that this protein is associated with non-caveolar lipid rafts. Because of our inability to co-stain for these proteins in the same lawns, however, we cannot rule out the idea that a proportion of SNAP-23 overlaps with caveolin. Although SNAP-23 distribution seemed distinct from caveolin-1, cholesterol depletion caused the dispersal of both SNAP-23 and caveolin, but the clusters of transferrin receptor mainly stayed intact (Fig. 5, A and B). Thus, the results support the idea that SNAP-23 is associated with detergent-insoluble, cholesterol-rich lipid rafts that are largely distinct from caveolae.
Finally, we examined the distribution of VAMP2 and SNAP23 in PM lawns from basal and insulin-stimulated cells.

FIG. 4. Association of SNARE proteins with rafts isolated from purified plasma membranes and low-density microsomes.
Plasma membranes purified from 40 ϫ 10 6 adipocytes, either untreated (basal) or stimulated with 1 M insulin for 30 min (ins), or LDM fractions from unstimulated cells were solubilized in 1 ml of MBS containing 0.2% Triton X-100. The lysate was homogenized and made up to a final concentration of 40% sucrose (3-ml volume) and overlaid successively with 6 ml of 30% sucrose and 4 ml of 5% sucrose (in MBS). After centrifugation at 240,000 ϫ g for 18 h, 1-ml fractions were collected from the top of the gradients (fraction 1 is the top fraction). Equal volumes of the recovered fractions were separated by SDS-PAGE and transferred to nitrocellulose for immunoblotting analysis using the antibodies indicated. A, caveolin and transferrin receptor (TfR) distribution in gradients prepared from plasma membranes from unstimulated cells. B, SNAP23, syntaxin 4 (syn 4), VAMP2, and Glut4 distribution in gradients prepared from plasma membranes purified from unstimulated (basal) or insulin stimulated (ins) cells. C, VAMP2, cellubrevin (Ceb), Glut4, transferrin receptor, and IRAP distribution in gradients prepared from purified LDMs.
FIG. 5. Immunofluorescence analysis of protein distribution in adipocyte plasma membranes. A and B, plasma membrane lawns were generated by sonication of 3T3-L1 adipocytes, either untreated or treated with 5 mM methyl-␤-cyclodextrin (ϩM␤CD) for 45 min. The lawns were fixed in paraformaldehyde and dual labeled with antibodies against SNAP-23 (green stain) and the transferrin receptor (red stain) (A) or with anti-caveolin-1 (B) and subsequently with secondary antibodies conjugated to fluorescein isothiocyanate or tetramethylrhodamine B isothiocyanate. Scale bar in B, 10 m. C, plasma membrane lawns were prepared from adipocytes previously treated either with (Ins) or without (Bas) 1 ⌴ insulin for 30 min. The distribution of SNAP23 (red) and VAMP2 (green) is shown. Fig. 5C shows that in unstimulated cells, there was little overlap of SNAP23 and VAMP2 staining (although VAMP2 signal in basal PM lawns was low). After insulin stimulation, there was an increase in the staining intensity of VAMP2 in PM lawns, and VAMP2 puncta displayed a partial overlap with SNAP23-positive domains. DISCUSSION Insulin-stimulated Glut4 vesicle fusion is emerging as a system with a requirement for lipid raft function at many levels. The insulin receptor has been reported to reside in caveolae (a subset of raft) (22), other important insulin signaling molecules associate with rafts (10 -12), Glut4 may itself interact with rafts (20,21), and cholesterol depletion inhibits insulin signaling to increase glucose transport (22). The current study adds another level to raft involvement in insulin-stimulated glucose transport with the implication that rafts may be important for SNARE function. We have been unable to directly assess the role of lipid rafts in Glut4 vesicle fusion, because treatments that perturb SNARE raft association also affect insulin signaling and probably other aspects of Glut4 trafficking. Note that cyclodextrin treatment causes the complete dispersal of caveolin (Fig. 5). Nevertheless, the level of SNAP-23 and VAMP2 in adipocyte lipid rafts (Ͼ70%) isolated from purified membranes strongly suggests that lipid rafts play an important role in the process of insulin-stimulated Glut4 vesicle fusion with the PM.
What is the significance of raft association of SNARE proteins? Raft association will allow clustering of the SNARE proteins (7), which is probably important because vesicle fusion has been suggested to require the cooperative formation of three SNARE complexes between vesicle and membrane (23). However, raft association is not the only way to cluster proteins, because the non-raft proteins NaK pump (7) and the transferrin receptor (Fig. 5) are also clustered. Thus, lipid rafts may play a more direct role in membrane fusion, for example, by facilitating force transduction to the membrane anchors of complexed SNARE proteins. Such proposals are further strengthened by the recent observations that viral fusion involves lipid raft structures, at least in some systems (24,25).
Raft association of SNARE proteins probably allows spatial segregation of proteins and protein complexes important for exocytosis. Previous work has shown that the level of SNAP-23 in adipocytes is 3-fold higher than syntaxin 4 (26); however, overexpression of SNAP-23 enhances Glut4 vesicle fusion (27). If the major function of SNAP-23 in adipocytes were to SNARE-pair with syntaxin 4, these findings would suggest that spatial regulation of syntaxin 4/SNAP-23 existed. This could be explained by the relative enrichment of SNAP-23 in PM rafts over syntaxin 4 (70 versus 35% raft-associated, respectively). Thus, it may be that overexpressed SNAP-23 interacts with uncomplexed, non-raft syntaxin 4, and either promotes membrane fusion at non-raft domains of the PM or drives syntaxin 4 into rafts. Regardless, these data clearly illustrate that the domain complexity of the PM is far more complex than our present models suggest.
We have shown previously that VAMP2 is associated to a variable degree (4.7-23.9%) with Triton-insoluble rafts isolated from PC12 cells (6,28). In contrast, VAMP2 was essentially all present in rafts isolated using the detergent Lubrol WX (6), a result recently extended to VAMP2 associated with pancreatic zymogen granules (29). Thus, the extensive association of VAMP2 with intracellular Triton-insoluble rafts in adipocytes suggests that Glut4 vesicles (which contain the majority of VAMP2) have a composition different from that of PC12 granules and pancreatic zymogen granules. Likewise, the significant difference in raft association of VAMP2 and cellubrevin (VAMP3) may reflect the differential distribution of these vesicle SNAREs between Glut4 storage vesicles and endosomes (15). Recent work has shown that VAMP2-positive Glut4 vesicles but not cellubrevin-containing vesicles interact with the actin network in L6 skeletal muscle cells (30). Although actin is insoluble in Triton X-100, flotation in sucrose gradients separates low buoyant density rafts from the higher density actin (31), strongly suggesting that the raft pool of VAMP2 identified in the present study is not actin-associated. In addition, we are not aware of any studies describing a direct interaction of VAMP2 with actin (30).
Lipid rafts are emerging as important players in many cellular pathways. The association of SNARE proteins with rafts isolated from adipocytes and PC12 cells suggests that these lipid microdomains play an important role in regulated exocytosis in diverse cell types.