Mechanism and regulation of GLUT-4 vesicle fusion in muscle and fat cells

Leonard J. Foster, Amira Klip


Twenty years ago it was shown that recruitment of glucose transporters from an internal membrane compartment to the plasma membrane led to increased glucose uptake into fat and muscle cells stimulated by insulin. The final step of this process is the fusion of glucose transporter 4 (GLUT-4)-containing vesicles with the plasma membrane. The identification of a neuronal solubleN-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complex as a requirement for synaptic vesicle-plasma membrane fusion led to the search for homologous complexes outside the nervous system. Indeed, isoforms of the neuronal SNAREs were identified in muscle and fat cells and were shown to be required for GLUT-4 incorporation into the cell membrane. In addition, proteins that bind to nonneuronal SNAREs were cloned and proposed to regulate vesicle fusion. We have summarized the molecular mechanisms leading to membrane fusion in nonneuronal systems, focusing on the role of SNAREs and accessory proteins (Munc18c, synip, Rab4, and VAP-33) in incorporation of GLUT-4 into the plasma membrane. Potential modes of regulation of this process are discussed, including SNARE phosphorylation and interaction with the cytoskeleton.

  • vesicle traffic
  • soluble N-ethylmaleimide-sensitive factor attachment protein receptor
  • syntaxin 4
  • 23-kDa synaptosome-associated protein-like protein
  • vesicle-associated membrane protein 2


Vesicle-membrane fusion is a fundamental cellular process that occurs at the final step of protein export to most organelles and secretion of proteins and smaller molecules. Seminal work from Rothman and colleagues (12, 20, 37, 125) in the late 1980s identified a pair of soluble proteins that could bind to the fusing membranes and were required for successful fusion of Golgi vesicles with acceptor Golgi stacks. These proteins were termed NSF (N-ethylmaleimide-sensitive factor) and SNAP (soluble NSF attachment protein) on the basis of the sensitivity of the former to N-ethylmaleimide and the ability of both proteins to bind to each other. Later, four membrane proteins from brain extracts were found to act as receptors for NSF and SNAP and were termed SNAREs (for SNAP receptors) (95). The proteins consisted of VAMP-2 (vesicle-associated membrane protein-2) (26), syntaxins A and B (10), and SNAP-25 (25-kDa synaptosome-associated proteins) (77). On the basis of their topological localization in the presynaptic bouton, these proteins were classified as vesicle (or v-) SNAREs (e.g., VAMP-2) and target (or t-) SNAREs (e.g., syntaxin and SNAP-25) (Table 1). VAMPs and syntaxins are characterized by a very short extracellularly/luminally directed COOH terminus, a single transmembrane domain, and a long cytoplasmic NH2-terminal region encompassing two coiled-coil domains (Fig.1). In contrast, SNAP-25 does not have transmembrane domains but presents two coiled-coil domains flanking a cluster of cysteine residues that are highly susceptible to palmitoylation (Fig. 1). The three proteins interact with one another through their coiled-coil domains, and it is now thought that the interaction of SNAP-25 with syntaxin is more relevant to its membrane localization than is its palmitoylation (115, 116).

View this table:
Table 1.

Nonneuronal SNAREs and interacting proteins

Fig. 1.

Common domain structure of solubleN-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins. A: syntaxins typically contain 3 coiled-coil domains. The most COOH terminal (juxtamembrane) of these domains contains the conserved glutamine (Q) residue and participates in the SNARE complex. Both vesicle-associated membrane proteins (VAMPs) and syntaxins have an extreme COOH-terminal transmembrane domain that protrudes through the membrane, leaving the COOH terminus extracellular (luminal). Similar to syntaxins, VAMPs also contain a juxtamembrane coiled-coil domain that participates in the SNARE complex, but VAMP coiled-coil domains typically contain a conserved arginine (R) residue. SNAP-25 and its homologs contain 2 coiled-coil domains, 1 at either end of the molecule, and both domains contain conserved glutamine residues. In addition, these molecules have multiple cysteine residues (cccc) in the middle of the molecule that can be palmitoylated to enhance the interaction of SNAP-25 with membranes. B: the 3 SNARE proteins condense into a complex containing 4 α-helices, 2 contributed by SNAP-25 and 1 each by VAMP and syntaxin. The helices align in a parallel fashion, presumably also parallel to the plane of the membranes. The implication is that the flexible linker between the 2 coiled-coil domains of SNAP-25 must loop back around the complex to allow both domains to align in a parallel fashion. N, NH2terminus; C, COOH terminus.

It was originally proposed that the SNARE proteins would form a link between vesicle and target membranes as a step preceding fusion (95) and that fusion would be driven by the energy released from ATP hydrolyzed by bound NSF (95). Furthermore, given that individual SNARE isoforms were found to have distinct tissular, cellular, and organellar specificity, it was proposed that the SNAREs would dictate vesicle targeting specificity (95). This hypothesis is supported by very recent work showing that only certain SNARE isoforms are able to recover disrupted norepinephrine release from cracked PC12 cells (89). It is now clear that syntaxin and SNAP-25 also populate synaptic vesicles and that VAMP is also found in target membranes (121). This has led to the suggestion that cis-complexes of v- and t-SNARE may occur within the same membrane, preventing the individual components from engaging in trans-interactions with the opposite membrane. The action of NSF and αSNAP is to dissociate thecis-complexes using the energy released by ATP hydrolysis (Fig. 2) (7, 21). The final fusion step depends on SNARE protein integrity but appears to be independent of ATP hydrolysis, suggesting that NSF is not involved at this stage (7, 21). A structure-function model of fusion has been proposed whereby SNAREs in the docked conformation “zip up” (Fig. 2) to form a tight, stable SNARE complex (40). The complex involves a four-helix coiled-coil bundle now described at atomic resolution (98). The free energy released by the formation of this exceptionally stable complex is thought to be the source of the energy used to fuse the two lipid bilayers (40). The elucidation of the crystal structure of the SNARE complex led to an alternative classification of SNAREs into Q- or R-SNAREs, based on the presence of either a glutamine (Q) or arginine (R) residue in the center of the SNARE complex (29).

Fig. 2.

Hypothetical steps in vesicle docking and fusion applied to the glucose transporter 4 (GLUT-4) system. Preformedcis-SNARE complexes on the plasmalemmal and vesicle membranes must first be dissociated by the action of the ATPaseN-ethylmaleimide-sensitive factor (NSF) and its assistant, soluble NSF attachment protein (αSNAP) (priming). The vesicle then becomes associated with the plasma membrane (tethering) through as yet unknown molecules. Rab4 may participate at this point, but tethering does not involve SNARE proteins. Once tethered, the SNAREs cantrans-associate, causing the vesicle to become more tightly associated with the plasma membrane (docking). Docking then leads to formation of the classic SNARE complex on the way to fusion of the vesicle with the plasma membrane.

Because SNARE proteins were first identified in neuronal or neuroendocrine tissues, most of our information on these protein families stems from studies in neuronal systems. However, SNARE proteins are found in all tissue and cell types. In the last decade, more than 9 VAMP isoforms, 19 syntaxin isoforms, and 3 SNAP-25 isoforms have been described across the animal and even the plant kingdoms (51). The conservation of the basic elements of these proteins has given rise to the tenet that these proteins must fulfill a universal role in the mechanism of vesicle-membrane fusion (for a comprehensive review, see Ref. 51).


An important biological phenomenon involving vesicle-membrane fusion is the incorporation of glucose transporters into the plasma membrane of muscle and fat cells. The glucose transporter of these tissues is GLUT-4, a 12-transmembrane domain protein that mediates vectoral transport of glucose in the direction of the glucose gradient (8). The hormone insulin strongly promotes GLUT-4 incorporation into the cell surface, and this translocation appears to fail in insulin resistance accompanying several forms of diabetes (60, 62, 130). Because of the physiological importance of insulin-dependent GLUT-4 translocation to the cell surface, attempts have been made to characterize the final GLUT-4 vesicle fusion step, drawing from lessons learned from neuronal synaptic transmission.

In unstimulated muscle and fat cells, the steady-state distribution of GLUT-4 favors intracellular compartments over the plasma membrane (25, 46, 52, 83). This steady state is the result of a slow mobilization of GLUT-4 to the cell surface and rapid removal from the plasma membrane (47, 53). Most studies suggest that the intracellular compartments populated by GLUT-4 include the early/sorting endosome, the recycling endosome, and a specialized vesicular compartment (41, 55, 56, 69). It is currently debated whether the latter does or does not recycle in the basal state. In rodent adipocytes, insulin promotes the externalization of the specialized vesicles and increases the recycling of GLUT-4 from the recycling endosome to the plasma membrane (41, 55, 56,69). In muscle, insulin mobilizes a specialized vesicle pool, but there is no evidence that the recycling endosome is also mobilized (1, 2). Instead, emerging studies are consistent with the possibility that muscle contraction mobilizes GLUT-4 from the recycling endosome in this tissue (24, 63, 80). Thus GLUT-4-containing vesicles incorporate into the plasma membrane in at least three circumstances: in the basal (unstimulated) state, out of the recycling endosome; in response to insulin, out of the specialized vesicle; and in response to exercise in muscle and to insulin in fat cells, out of the recycling endosome. Other possibilities are not discounted, such as additional mobilization of the specialized vesicular pool in response to exercise. This diversity of fusion events begs the question of whether similar or different molecules participate in GLUT-4 vesicle fusion with the plasma membrane in each case. In the search for answers to this question, the SNARE isoforms expressed in muscle and fat cells had first to be defined. Of the VAMP family, only VAMP-2 and VAMP-3/cellubrevin have so far been detected in muscle and fat primary tissues and corresponding cells in culture (82, 105,117). Contrary to neuronal and neuroendocrine cells, muscle and fat cells do not express syntaxin 1, but instead express syntaxin 4 (49, 102, 119). In addition, low levels of syntaxin 2 are also detected in 3T3-L1 adipocytes (119) and rat adipocytes (105), whereas small levels of syntaxin 3 are present in rat adipocytes (105). Another difference between neuronal/neuroendocrine cells and muscle and fat cells pertains to the expression of SNAP-25. This isoforms was not found in insulin-sensitive cells, which instead express SNAP-23 (3, 122,127).


Synaptic vesicle fusion is the final step in a series of events that have been described as vesicle tethering (a reversible step) and vesicle docking (an irreversible step) (Fig. 2). Information on these steps has also emerged from yeast molecular genetic studies (16) and from in vitro endosome-endosome fusion studies (19, 71). The full complement of tethering proteins has not yet been identified, but the early endosome autoantigen 1 (EEA1) protein may act as the tethering protein engaged in binding endocytic vesicles to the early endosome (19, 70). EEA1 is found in insulin-sensitive cell types (78), but its participation in GLUT-4 vesicle traffic has not been tested.

Once vesicles are brought into close proximity with their target membranes by the tethering process, SNARE proteins on opposite membranes associate and acquire the configuration required for fusion (Fig. 2, “docked”). The SNARE proteins in docked vesicles are thought to be in a high-energy state (40) that can be maintained for long periods of time. Indeed, large numbers of synaptic vesicles can be observed docked at the presynaptic plasma membrane (51). In contrast, GLUT-4 vesicles have rarely, if at all, been found perched at the plasma membrane of unstimulated muscle or fat cells.

In neurons, nerve terminal depolarization leading to calcium influx is the penultimate trigger for neurotransmitter exocytosis. Currently, there is no evidence supporting a need for calcium ions in insulin-dependent glucose uptake mediated by GLUT-4 in 3T3-L1 adipocytes, L6 myotubes, or cardiac myocytes (42, 59, 61). In fact, GLUT-4 insertion into the plasma membrane can be observed in cells equilibrated with calcium-free buffers by means of streptolysin O-induced cell permeabilization (17) (Foster LJ and Klip A, unpublished observation).



VAMP-2, the prototypical v-SNARE, is common to several systems in which vesicle traffic is regulated. These include neurotransmitter release in neural synapses (26), insulin-stimulated GLUT-4 translocation in fat and muscle cells (15, 117), and aquaporin-2 translocation in renal collecting ducts (54). VAMP-2 is expressed in muscle (82, 117) and fat cells (15, 118) and was originally detected in immunoisolated GLUT-4 compartments from rat fat cells (15). By subcellular fractionation of muscle and adipose cells, the protein is found to be distributed in similar proportions in the plasma membrane and intracellular membranes (15, 117, 118).

VAMP-2 is susceptible to cleavage by various clostridium neurotoxins (50). This susceptibility afforded a specific strategy to probe the function of VAMP-2 (and a closely related isoform called VAMP-3/cellubrevin) in GLUT-4 traffic. We and others have demonstrated a requirement for VAMP-2 in insulin-stimulated GLUT-4 translocation (17, 18, 31, 39, 65, 68, 75, 85, 99). Tetanus toxin and botulinum toxins B and D introduced into rodent adipocytes by electroporation, single-cell microinjection, chemical permeabilization (using streptolysin O toxin), or natural, toxin-mediated uptake (17, 18, 31, 39, 65, 99) reduced by more than one-half the insulin-stimulated GLUT-4 incorporation into the cell surface (Table2). In addition, introduction of antibodies raised against various regions of VAMP-2 as well as peptides representing different segments of VAMP-2 also diminished the insulin-dependent arrival of GLUT-4 at the plasma membrane of rodent adipocytes (17, 65, 68, 75) (Table 2).

View this table:
Table 2.

Effects of interfering with SNARE proteins on insulin-stimulated glucose uptake and/or GLUT-4 translocation

Recent work on the function of VAMPs in GLUT-4 traffic has focused on resolving whether VAMP-2 or VAMP-3/cellubrevin is the primary v-SNARE involved in insulin-stimulated GLUT-4 translocation. It has been suggested that VAMP-2 is the v-SNARE important for translocation of GLUT-4 from the insulin-sensitive compartment, because the cytosolic domain of VAMP-2, but not VAMP-3/cellubrevin or VAMP-1, reduced insulin-stimulated GLUT-4 translocation by one-half when microinjected into 3T3-L1 adipocytes (68). In addition, transfection of tetanus toxin light chain into L6 muscle cells in culture resulted in 70% inhibition of insulin-dependent GLUT-4 arrival at the cell surface (85). Basal levels of cell surface GLUT-4 were minimally affected. Cotransfection of tetanus toxin-insensitive mutants of VAMP-2, but not VAMP-3, rescued the inhibition (85). These results indicate that VAMP-2, but not VAMP-3, is involved in insulin-stimulated GLUT-4 translocation and that neither protein participates in GLUT-4 sorting to the plasma membrane in the basal state.

Syntaxin 4

Syntaxin 4 is expressed in muscle and fat cells, where it is largely, but not exclusively, located at the plasma membrane (97,105, 119). In fact, GLUT-4 vesicles contain syntaxin 4 that cannot be explained by contamination from plasma membranes (119). Unlike syntaxins 1 through 3, syntaxin 4 is not susceptible to cleavage by botulinum toxin C1 (90). For this reason, studies on the functional role of syntaxin 4 in GLUT-4 translocation have required the use of antibodies and peptides to perturb the function of syntaxin 4. Microinjection (65, 75,101), chemical permeabilization (17, 119), and adenoviral overexpression (75) have been used to introduce antibodies directed against syntaxin 4 or soluble domains of the protein. In all cases, the perturbation of syntaxin 4 resulted in ∼50% inhibition of insulin-stimulated glucose uptake (119) or GLUT-4 translocation (17, 65, 75,101) (Table 2).


SNAP-23 shares both sequence and structural homology with SNAP-25 (86). A protein cloned from a cDNA library of 3T3-L1 adipocytes, originally named syndet (122), was found to be the murine form of SNAP-23 (96). By subcellular fractionation of muscle and fat cells, SNAP-23 is found almost exclusively in the plasma membrane-enriched fraction (122,127). Neutralizing antibodies as well as peptides encoding the NH2 or COOH termini of SNAP-23 have been introduced into 3T3-L1 adipocytes by microinjection, chemical permeabilization, and adenoviral transfection. All these reagents reduced the insulin-dependent arrival of GLUT-4 at the plasma membrane, although they did not inhibit it completely (31, 34, 58, 87). The clostridium neurotoxins Bo/NT A and E have been useful to probe the function of SNAP-25, but they have been less effective in targeting SNAP-23. Bo/NT E has been shown to cleave SNAP-23 in some species, notably the canine isoform (64). In some reports, the toxin was able to cleave the murine SNAP-23, concomitantly reducing GLUT-4 translocation (31). In other studies, the toxin was ineffective toward SNAP-23 (18, 66). Because SNAP-23 is not a transmembrane protein and can bind to native syntaxin 4, it was also possible to introduce into cells full-length SNAP-23 to test its function. The microinjected full-length protein enhanced both insulin-stimulated GLUT-4 translocation and glucose uptake (34). These results suggest that the endogenous SNAP-23 may be available for SNARE complex formation in limiting amounts. A very recent report (43) has defined that 3T3-L1 adipocytes have approximately three times more SNAP-23 than syntaxin 4 (1.15 × 106 copies of SNAP-23 per cell compared with 3.74 × 105 copies of syntaxin 4). The extent of availability of each of these proteins for SNARE complex formation is still to be determined, given that these proteins have several cellular partners. Exogenous SNAP-23 may enhance the rate of fusion of GLUT-4 vesicles with the plasma membrane by enhancing the formation of productive complexes with syntaxin 4 and VAMP-2 (Table 2).


Unlike the expanded nature of the SNARE protein families, NSF and αSNAP have very few apparent homologs. High-resolution X-ray crystal structures suggest that NSF may engage αSNAP as a lever to pry the SNARE complex apart (129) (Table 1). This would then allow SNAREs to form complexes between opposing membranes that are competent for fusion. Indeed, NSF and αSNAP are found in rat adipocytes, and epitope-tagged versions of these proteins have been used to immunoprecipitate SNARE complexes from these cells. Such complexes contained syntaxin 4, VAMP-2, VAMP-3, and SNAP-23 (105). Transfection of a dominant negative mutant of NSF into rat adipocytes resulted in plasma membrane levels of GLUT-4 after insulin treatment that were not significantly different from basal, nontransfected levels (45). However, the level of plasma membrane GLUT-4 in basal cells expressing the dominant negative NSF was also lowered significantly (45).


The three SNAREs expressed in muscle and fat cells appear to be essential for a significant fraction of the insulin-stimulated GLUT-4 arrival at the cell surface. However, in all the experiments discussed above, interfering with VAMP-2, syntaxin 4, or SNAP-23 only partly inhibited insulin-stimulated GLUT-4 translocation, and the basal levels of plasma membrane GLUT-4 were not altered even after long time periods in the presence of the perturbing agent. This latter observation is especially surprising given that GLUT-4 is known to cycle dynamically to and from the membrane in the absence of insulin. These observations suggest that either different SNARE isoforms or other proteins mediate these two fusion events. Only one other VAMP has been detected in the relevant membranes of muscle and fat cells, VAMP-3/cellubrevin. However, complete hydrolysis of this protein by botulinum or tetanus toxin, or interference by peptides emulating NH2-terminal sequences of VAMP-3/cellubrevin, failed to affect either basal or insulin-mediated GLUT-4 arrival at the cell surface (68, 75, 85), whereas the analogous domain of VAMP-2 effectively inhibited one-half of the insulin action (68). It is conceivable that muscle and fat cells express other, toxin-insensitive VAMPs, which could potentially mediate the fusion events that are not accounted for by VAMP-2. Similarly, it is conceivable that syntaxin 2 or 3 could mediate fusion events because they each bind to VAMP-2 (28, 81).

The studies listed above support the notion that VAMP-2, syntaxin 4, and SNAP-23 are required for the incorporation of GLUT-4-containing vesicles into the plasma membrane. These proteins may participate in the actual membrane fusion step, by analogy to the fusogen role assigned to their neuronal counterparts (51). Indeed, purified, bacterially expressed SNAP-25, syntaxin 1, and VAMP-2 reconstituted into synthetic proteoliposomes can mediate fusion of these liposomes (73, 124). Fusion of a single vesicle with its target membrane likely requires the formation of more than one SNARE complex, and there is a suggestion that a ring of SNARE complexes aligns around the fusion pore (124). Current experiments have not been able to determine the number of complexes required for fusion of one vesicle. This will likely require detailed microcalorimetry experiments measuring the free energy released during complex formation.

The importance of the formation of a neuronal SNARE complex for synaptic vesicle fusion suggests that a high-affinity complex may also form between VAMP-2, syntaxin 4, and SNAP-23 in the process of GLUT-4 vesicle fusion. This possibility has been addressed experimentally, yielding somewhat surprising results (Table3). Although there is general agreement that a complex does form comprising SNAP-23, syntaxin 4, and VAMP-2, the biochemical properties of such a complex appear to differ from those of the VAMP-2/syntaxin 1/SNAP-25 complex. These differences depend on the experimental design including, importantly, whether the proteins used are recombinant forms produced in bacteria or endogenous proteins isolated from mammalian cell systems. One of the hallmarks of the neuronal SNARE complex involving SNAP-25, syntaxin 1, and VAMP-2 is its ability to resist denaturation by ionic detergents such as sodium dodecyl sulfate (SDS). Whereas an SDS-resistant complex containing VAMP-2, SNAP-23, and syntaxin 4 in 0.5% SDS was detected using surface plasmon resonance (87), such a complex could not be detected by SDS-PAGE or circular dichroism using samples in 2% SDS (35, 128). A second distinguishing feature of the neuronal SNARE complex is that SNAP-25 enhances the binding of VAMP-2 to syntaxin 1. In contrast, in glutathioneS-transferase-pulldown experiments, we found no evidence for cooperative binding between SNAP-23, VAMP-2, and syntaxin 4 (35); however, a cooperative effect of SNAP-23 on complex formation was reported when all three proteins were overexpressed in COS cells (58).

View this table:
Table 3.

Reported SNARE complexes containing SNAP-23, VAMP-2, and syntaxin 4


An important question is whether any of the steps involved in GLUT-4 vesicle incorporation into the plasma membrane is regulated by insulin-derived signals. The occupied insulin receptor undergoes autophosphorylation and then phosphorylates insulin-receptor substrate(s) (27, 69). This phosphorylation leads to activation of phosphatidylinositol 3′-kinase to produce phosphatidylinositol 3,4-bisphosphate and phosphatidylinositol 3,4,5-trisphosphate. These two lipids lead to the activation of protein kinase B/Akt and the atypical protein kinase C's (27, 32,69), both of which appear to be required for GLUT-4 translocation (5, 6, 31, 44, 123). The steps downstream of these serine/threonine kinases leading to GLUT-4 translocation are as yet unidentified.

Recent studies have suggested that certain physiological conditions may halt GLUT-4 vesicles at the docking state and prevent fusion. Isoproterenol pretreatment reduces insulin-dependent glucose uptake in adipocytes and the number of GLUT-4 molecules detected at the surface of intact cells with the use of an impermeant photolabel (114). However, isolated plasma membranes did not reflect any diminution in GLUT-4 levels caused by isoproterenol (114). These results led to the suggestion that isoproterenol maintained GLUT-4 in an occluded state, inaccessible to the cell surface. Because the photolabel used has a preferential reactivity with active transporters, it was also possible that the transporters were inactive but present at the outer surface of the membrane. Therefore, it was important to detect transporter exposure at the cell surface by other means. With the use of cell-surface biotinylation, it was confirmed that isoproterenol pretreatment rendered the transporter inaccessible to extracellular labels yet bound to plasma membranes upon their isolation (30). These results suggest that isoproterenol pretreatment allows GLUT-4 vesicles to dock but prevents their fusion with the plasma membrane. Preliminary results from our laboratory suggest that isoproterenol may have similar effects in muscle cells. Treatment of L6 skeletal muscle cells with isoproterenol before insulin treatment reduced the insulin-stimulated glucose uptake. This was concomitant with a decrease in the insulin-stimulated GLUT-4 translocation measured by the externalization of an myc epitope inserted in the first extracellular loop of GLUT-4 and stably transfected into these cells (Hayashi M, Bilan P, and Klip A, unpublished observations). Experiments are currently underway to determine whether isoproterenol causes GLUT-4 vesicles to associate but not fuse with the plasma membrane in these cells as well.

It will be interesting to confirm, by using electron microscopy or in vitro reconstitution assays, that GLUT-4 vesicles can be arrested at the docked state. To date, the only reconstitution assay achieved detects GLUT-4 binding to the membrane but does not differentiate between docking and fusion (48). GLUT-4 vesicles from a 3T3-L1 adipocyte cell line expressingmyc-tagged GLUT-4 were able to associate in vitro with isolated plasma membrane derived from wild-type 3T3-L1 cells. The association was dependent on calcium and could be prevented by including the recombinant soluble domain of syntaxin 4 in the binding assay. Notably, plasma membranes derived from insulin-stimulated cells were more effective than control membranes in binding intracellular GLUT-4 myc (48). Although this assay does not distinguish between docking and fusion, it may prove helpful in testing the individual participation of enzymes known to be activated by insulin on the interaction of donor and acceptor membranes.


A plausible mechanism whereby insulin- or isoproterenol-dependent signals could regulate the fusion machinery is through phosphorylation of its integral components. Indeed, considerable efforts have been dispensed toward defining which SNAREs are susceptible to phosphorylation and by which kinases (Table4). We have reported that syntaxin 4 is susceptible to phosphorylation by the serine/threonine kinases protein kinase A, casein kinase II, and conventional protein kinase C in vitro (35). SNAP-23 is also phosphorylated by conventional protein kinase C's, but the phosphorylation is inefficient (35). In addition, the newly identified SNAP-23 kinase (SNAK) can phosphorylate SNAP-23 and syntaxin 4 (14).

View this table:
Table 4.

Phosphorylation of SNAREs and ancillary proteins

To further explore whether syntaxin 4, VAMP-2, or SNAP-23 is a suitable substrate of kinases in vivo, 3T3-L1 adipocytes were loaded with [o-32P]phosphate and treated with insulin, and each SNARE was selectively immunoprecipitated. Although each protein was found to be phosphorylated, the level was not altered significantly by insulin treatment. We further explored whether isoproterenol may alter SNARE phosphorylation, to provide a possible explanation for the inability of GLUT-4 vesicles to dock with the plasma membrane. However, isoproterenol did not significantly alter the phosphorylation levels of either SNAP-23 or syntaxin 4 (Foster LJ and Klip A, unpublished observations) (Table 4).

Ancillary Proteins

A second mechanism whereby SNARE function might be controlled is through binding of ancillary proteins that may either prevent or promote productive SNARE complex formation leading to membrane fusion. A cohort of proteins have been described to interact with SNAREs, as described below and listed in Tables 1 and5.

View this table:
Table 5.

Effects of interfering with SNARE-binding proteins on insulin-stimulated glucose uptake or GLUT-4 translocation


The Munc18 proteins are mammalian homologs of the Sec1 protein inSaccharomyces cerevisiae and the unc-18 protein inCaenorhabditis elegans, both of which bind syntaxins from their respective species. Munc18a was originally cloned from neuronal tissues and has been given many different names, including n-Sec1 and rbSec1. Munc18a cDNA was used as bait to clone similar genes from a 3T3-L1 cDNA library. One of the proteins cloned by this method was Munc18c, which interacts specifically with syntaxins 2 and 4 but not syntaxins 1 or 3 (100, 102). Munc18c inhibits the binding of syntaxin 4 to VAMP-2 (101, 103) and SNAP-23 (3). Insulin causes the dissociation of a Munc18c/syntaxin 4 complex (103). A prediction of this observation is that once insulin causes the dissociation, syntaxin 4 would be available to bind SNAP-23 and VAMP-2, leading to fusion of the vesicles with the target membrane. Indeed, full-length Munc18c introduced into 3T3-L1 adipocytes by adenoviral transfection inhibited insulin-stimulated glucose uptake and GLUT-4 translocation by ∼50% (100,103). However, a peptide representing the domain of Munc18c that binds to syntaxin 4, when microinjected into 3T3-L1 adipocytes, inhibited fusion of green fluorescent protein-GLUT-4-containing vesicles with the plasma membrane. The peptide appeared to allow GLUT-4 vesicles to dock with the plasma membrane without fusing with it. Given that this peptide displaces Munc18c-binding to syntaxin 4, these results may suggest that the displaced, endogenous Munc18c catalyzes fusion (104) (Fig.3).

Fig. 3.

Hypothetical regulatory events in GLUT-4 vesicle docking and fusion. In the basal state, both Munc18c and synip are bound to syntaxin 4, whereas SNAP-23 and GLUT-4 are associated with the actin cytoskeleton. Pantophysin, VAMP-2, and Rab4 are located on GLUT-4 vesicles. Before insulin causes the release of Rab4 from the vesicles, Rab4 organizes the proteins on the vesicles into conformations required for fusion. A 33-kDa VAMP-2-associating protein (VAP-33) may also be involved as a chaperone for VAMP-2. Insulin causes the formation of actin ruffling and the dissociation of Munc18c and synip from syntaxin 4. Once brought into the proximity of fusion machinery at the plasma membrane by the cytoskeleton, the GLUT-4 vesicle can dock and subsequently fuse, exposing GLUT-4 to the extracellular milieu.


The recently cloned synip is a syntaxin 4-interacting protein, identified in a 3T3-L1 cDNA library by a yeast two-hybrid screen (72). Synip binding to syntaxin 4 prevents VAMP-2/syntaxin 4 binding but not SNAP-23/syntaxin 4 binding (72). As for Munc18c, the association of synip with syntaxin 4 is reduced in insulin-stimulated cells. Insulin-sensitivity is conferred by the NH2-terminal half of synip, whereas the COOH-terminal half modulates GLUT-4 translocation (72). Despite having unrelated primary sequences, synip and Munc18c regulate the availability of syntaxin 4 for fusion of GLUT-4 vesicles with the plasma membrane in response to insulin (Fig. 3). It will be interesting to determine whether the two proteins regulate different functional pools of syntaxin 4.


SNAK is a protein kinase identified by its ability to bind syntaxin 4 in a yeast two-hybrid assay (14). However, SNAP-23 is a better substrate of SNAK than syntaxin 4 (14). SNAK phosphorylates SNAP-23 in vivo and in vitro, selectively phosphorylating only SNAP-23 that is not bound to syntaxin 4. SNAK phosphorylation of SNAP-23 enhances t-SNARE complex assembly, that is, binding of SNAP-23 and syntaxin 4 (14). It is unknown whether SNAK is present in insulin-sensitive tissues or whether SNAK is activated by insulin. Results of in vivo phosphorylation do not support any insulin-dependent phosphorylation of SNAP-23 (Table 4).


The growth factor-induced phosphoprotein Hrs-2 can bind to SNAP-25 and SNAP-23 in vitro (110). In permeabilized PC12 cells, recombinant Hrs-2 inhibits norepinephrine release (9). Hrs-2 is expressed in muscle and fat cells, but its tyrosine phosphorylation state is not altered in response to insulin (Yaworsky K, Foster LJ, and Klip A, unpublished observations). To date, there is no evidence for its participation as a regulator of GLUT-4 traffic.


A ubiquitous homolog of the synaptic vesicle protein synaptophysin, termed pantophysin, has recently been cloned from several sources (13, 38). This protein is found on GLUT-4-containing vesicles from 3T3-L1 cells and, similarly to synaptophysin, binds VAMP-2 (13). Interestingly, although pantophysin itself was not phosphorylated, a 77-kDa phosphoprotein associates with pantophysin upon treatment of cells with insulin (13). This result suggests a potential regulation of pantophysin by insulin. Preliminary results from our laboratory suggest that pantophysin availability is required for GLUT-4 vesicle fusion (Foster LJ, Cheatham B, and Klip A, unpublished observations).


A 33-kDa VAMP-2-associating protein (VAP-33) was isolated from anAplysia californica cDNA library through a yeast two-hybrid approach (94). A human homolog was identified soon thereafter (126). VAP-33 is a single-transmembrane domain protein with the bulk of the molecule in the cytosol. Two isoforms of VAP-33 (VAP-33A and VAP-33B) bind VAMP-2 in vitro (74). We have recently shown (33) that VAP-33 is present on immunopurified VAMP-2 vesicles from L6 myotubes and 3T3-L1 adipocytes. Interestingly, overexpression of VAP-33A inhibited insulin-stimulated GLUT-4 translocation, and this effect was rescued by co-overexpression of VAMP-2. In addition, anti-VAP-33 antibodies microinjected into 3T3-L1 adipocytes also inhibited GLUT-4 translocation (33). We hypothesize that, as in the case of Munc18c, the levels of VAP-33A in the cell are critical and that shifting the balance of VAP-33A/B either above or below the critical point can have adverse effects on vesicle traffic.

Rab proteins.

Rab GTPases represent a family of >35 proteins that relay information upon binding in their GTP-bound form to downstream effectors. Rabs become membrane-associated via geranylgeranylation or farnesylation, and this posttranslational modification is required for their GTPase function, which terminates their function on effectors. By genetic complementation, Rab proteins have been implicated in vesicular traffic, specifically in the recognition of vesicles by target membranes (for review, see Ref. 91). Deletion of the yeast Rab4 (Sec4p) can be rescued by overexpression of specific t-SNAREs, and the Rab Ypt1p is required for v-SNARE-t-SNARE complex formation. An emerging model suggests that Rab proteins direct vesicle traffic through the recruitment of docking factors from the cytosol. Thus Sec4 binds to the yeast exocyst that links vesicles to bud membranes, Rab5 binds to EEA1 that links endocytic vesicles to early endosomes, and vesicular VPs21 binds to Vac1 that links to the t-SNARE Ppe12 on target vesicles (4).

To date, only Rab4 has been implicated in GLUT-4 traffic by virtue of its presence on immunopurified GLUT-4 compartments (23,111). Insulin stimulation causes Rab4 geranylgeranylation, GTP loading (92), and dissociation from GLUT-4-containing endomembranes (22, 36). Introduction of wild type or mutants of Rab4 or a peptide representing the hypervariable region of Rab4 resulted in inhibition of insulin-stimulated GLUT-4 translocation (22, 93, 120). Interestingly, a link between the Rab-mediated vesicle docking and the actin-based cytoskeleton has been established. Rabphilin, a Rab3 effector, interacts with the actin-bundling protein α-actinin (57), and Rab8 promotes polarized membrane transport through reorganization of actin filaments (79).

The Cytoskeleton

The actin cytoskeleton has been repeatedly implicated in exocytic events, both as a barrier separating the docked from stored synaptic vesicles (in essence, limiting the active zone) and as a facilitator of granule exocytosis (11, 112). Recent studies reveal that secretory granules acquire a coat of actin before exocytosis (113). Actin filaments are dynamic and, in addition to separating active zones and coating granules, they constitute stress fibers and cortical networks. The latter form in response to growth factor stimulation involving the Rho-family protein Rac (88), and in insulin-sensitive muscle cells, they present as large submembranous three-dimensional structures (59,109). We have recently shown that formation of cortical actin structures is required for GLUT-4 exocytosis. Specifically, the rapidly forming subcortical actin mesh contained GLUT-4 vesicles and insulin signaling molecules (59). Preventing cortical actin structure formation through transient expression of a dominant negative Rac mutant abrogated externalization of GLUT-4 (59). In nontransfected muscle cells, GLUT-4 is inserted into the membrane at sites of membrane ruffles supported by cortical actin structures (106). Notably, SNAP-23 and syntaxin 4 appear to concentrate at sites of contact of the actin mesh with the plasma membrane (Khayat K, Foster LJ, and Klip A, unpublished observations). In adipocytes, a requirement for an organized cytoskeleton in GLUT-4 traffic has also been demonstrated (76). It will be interesting to determine whether GLUT-4 vesicles, once delivered by the cytoskeleton to the vicinity of the plasma membrane, require Rab4 as the tethering molecule leading to SNARE complex formation.


GLUT-4 vesicle fusion bears similarities to and differences from the fusion of synaptic vesicles with their respective target membranes. VAMP-2, syntaxin 4, and SNAP-23, found in muscle and fat cells, form a SNARE complex that is similar but not identical to its neuronal counterpart, constituted by VAMP-2, syntaxin 1, and SNAP-25. Two mechanisms of insulin-dependent incorporation of GLUT-4 vesicles into the plasma membrane have been identified: one requiring VAMP-2, syntaxin 4, and SNAP-23, and one independent of these proteins. Although the fusion step is likely to be regulated by the hormone, to date there is no evidence for regulation through SNARE phosphorylation. However, it is conceivable that subtle regulation may still occur by this means. In contrast, there is emerging evidence that ancillary proteins such as Munc18c, synip, VAP-33, and pantophysin may regulate the availability of VAMP-2 or syntaxin 4 for productive GLUT-4 fusion. Cytoskeletal tethering of vesicles and SNAP-23 may provide entropic energy to the process of GLUT-4 vesicle fusion. Future studies should reveal whether the fine tuning of GLUT-4 vesicle fusion is altered in insulin-resistant states leading to or accompanying diabetes. In this regard, two recent studies (67, 84) report increases in muscle SNARE protein levels in two animal models of insulin resistance. Both reports suggest that these increases might be adaptive changes attempting to overcome the defects in GLUT-4 traffic that underlie insulin resistance.


We acknowledge the Juvenile Diabetes Foundation and the Medical Research Council of Canada for funding.


  • Address for reprint requests and other correspondence: A. Klip, Cell Biology Programme, The Hospital for Sick Children, Toronto, Ontario, Canada M5G 1X8 (E-mail: amira{at}


View Abstract