Insulin plays a fundamental role in whole-body glucose homeostasis. Central to this is the hormone's ability to rapidly stimulate the rate of glucose transport into adipocytes and muscle cells . Upon binding its receptor, insulin stimulates an intracellular signalling cascade that culminates in redistribution of glucose transporter proteins, specifically the GLUT4 isoform, from intracellular stores to the plasma membrane, a process termed ‘translocation’ [1,2]. This is an example of regulated membrane trafficking , a process that also underpins other aspects of physiology in a number of specialized cell types, for example neurotransmission in brain/neurons and release of hormone-containing vesicles from specialized secretory cells such as those found in pancreatic islets. These processes invoke a number of intriguing biological questions as follows. How is the machinery involved in these membrane trafficking events mobilized in response to a stimulus? How do the signalling pathways that detect the external stimulus interface with the trafficking machinery? Recent studies of insulin-stimulated GLUT4 translocation offer insight into such questions. In the present paper, we have reviewed these studies and draw parallels with other regulated trafficking systems.
GLUT4 trafficking uses SNARE proteins
In the absence of insulin, glucose transporter type 4 (GLUT4) populates two interrelated trafficking cycles. The first rapidly internalizes GLUT4 from the cell surface into the prototypical endosomal pathway. Once internalized, GLUT4 is then sorted efficiently into a more slowly recycling pathway between recycling endosomes, the trans-Golgi network and a population of specialized vesicles termed GLUT4 storage vesicles (GSVs) or insulin-response vesicles (IRVs) [1,4]. These represent the pool of GLUT4 that is mobilized rapidly to the cell surface in response to insulin. The endosomal system is highly dynamic, and, like many endosomal membrane proteins, GLUT4 traffics continuously through various compartments and the plasma membrane [2,5,6]. As with all eukaryotic membrane trafficking events, delivery of GLUT4 between different membranes is mediated by formation of specific SNARE (soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor) complexes. The formation of a complex between members of the t- (target) family of SNAREs and their cognate v- (vesicle) SNAREs is sufficient to drive membrane fusion, and SNAREs are also thought to contribute to the specificity of the trafficking event . Thus regulation of SNARE complex formation provides one means by which a specific trafficking event can be controlled.
The identification of v-SNARE proteins within purified GSVs  provided the first evidence that GLUT4 translocation requires SNARE proteins (reviewed in ). Subsequent studies have indicated the VAMP2 (vesicle-associated membrane protein 2) isoform as being the key v-SNARE in GSVs (reviewed in , but, for an alternative view, see ). Studies from numerous laboratories identified Sx4 (syntaxin 4) and SNAP23 (23 kDa synaptosome-associated protein) as the key t-SNAREs on the plasma membrane of adipocytes [9,11]. Disrupting the function of these molecules (e.g. using antibody microinjection, transgenic knockout or siRNA approaches) supports a role for these proteins in the insulin-stimulated delivery of GLUT4 to the cell surface (see, for example, [9,11–13]). Consistent with this, Sx4/SNAP23-containing liposomes could fuse with liposomes containing VAMP2, providing functional evidence that this SNARE combination is sufficient to drive membrane fusion in vitro .
Regulation of SNARE complex formation provides a means to control spatial and temporal co-ordinates of membrane trafficking events. Consequently, much interest has focused on the role of known regulators of SNARE complex assembly, most notably the Sec1/Munc18 (SM) family of proteins. Members of the SM family are conserved highly through evolution, and understanding their precise role in membrane fusion represents an important goal in cell biology . The role of these proteins remains a topic of debate, and formulation of a hypothesis for their action has been hampered by the observation that they seem to bind to syntaxins by multiple modes of binding, perhaps representative of different functions at different stages of the SNARE complex assembly/disassembly cycle (see, for example, [16–18]). The SM protein that binds Sx4 is Munc18c, and there is a wealth of evidence implicating this protein in the control of GLUT4 translocation . It is only recently, however, that we are beginning to understand how Munc18c is regulated.
Does insulin stimulate SNARE complex formation?
A crucial question for the field is whether insulin regulates SNARE complex assembly in vivo. This has proven difficult to address definitively for a range of mostly technical reasons. For example, co-immunoprecipitation approaches have been unable to distinguish between interactions that occur in vivo from those which are formed (or lost) post-lysis. The addition of tags, such as those required for FRET or bipartite fluorescent protein approaches, has been shown to alter the function of SNAREs, making interpretation of microscopy studies problematic . We have used recently a novel approach to address this question, using a microscopy-based proximity ligation assay (PLA) . PLA uses antibodies against two proteins of interest raised in separate species; these are detected subsequently using species-specific secondary antibodies to which short oligonucleotides of different defined sequence are attached covalently. A PLA signal will be obtained only if the oligonucleotide-coupled secondary antibodies are sufficiently close to allow hybridization and subsequent enzymatic ligation of added connector oligonucleotides. This in turn is determined by the proximity of the two proteins of interest. An in situ ligation reaction then ligates the oligonucleotides with connector oligonucleotides to provide a template for rolling PCR, whose product is identified by hybridization of a fluorescent probe. Associations are visualized by microscopy; the number of fluorescent spots detected being directly proportional to the number of associations between the two proteins. Note that a positive signal by PLA does not necessarily reflect a ‘direct’ interaction between the two proteins, only that the antigens are within ~15 nm of each other (i.e. within the size of an assembled SNARE complex).
We have used this approach to examine insulin-stimulated changes in associations between Sx4, SNAP23, VAMP2 and the SM protein Munc18c in 3T3-L1 adipocytes . We have analysed all six possible pairwise combinations between Sx4, SNAP23, VAMP2 and Munc18c under basal and insulin-stimulated conditions. To our initial surprise, the data revealed insulin-dependent changes in only two of these associations, i.e. between SNAP23 and VAMP2 and between SNAP23 and Munc18c; both of which increased in response to insulin. (Note that SNAP23 does not bind directly to either VAMP2 or Munc18c, but Munc18c binds the assembled Sx4–SNAP23–VAMP2 complex containing both SNAP23 and VAMP2 in sufficient proximity to generate a PLA signal .) These data support a model in which insulin increases the number of assembled Sx4–SNAP23–VAMP2 SNARE complexes, and that these have Munc18c associated with them. To our knowledge, this is the first direct demonstration of a stimulus-induced formation of a SNARE complex in vivo. It is possible that not all changes in associations are detected using such an assay due to the possibility of certain epitopes being masked or revealed following insulin treatment. To address this, we have performed all of our analyses with antibodies against different regions of each of the proteins .
Multiple pools of Sx4 in adipocytes
So how can we explain mechanistically the increased PLA associations between SNAP23 and Munc18c and SNAP23 and VAMP2 described above? Our experiments have revealed no increase in the association of Sx4 with SNAP23, VAMP2 or Munc18c, consistent with Sx4 being associated with all of these proteins before insulin-stimulation as well as after . They are not all in complex together because we have seen an increase in associations between VAMP2 and SNAP23 in response to insulin. Rather, we have interpreted our data as revealing the presence of two distinct pools of Sx4 under basal conditions: one in complex with SNAP23 and the other in complex with VAMP2. The presence of these distinct pools of Sx4 in basal cells would explain the increase in associations observed between SNAP23 and VAMP2 in the absence of concomitant increases in associations of Sx4 with VAMP2 and SNAP23 upon insulin-stimulated ternary complex formation. Similarly, no changes in the number of associations between Munc18c and Sx4 were observed, indicating that these two proteins are in association under basal conditions. Insight into these pools of Sx4 can be derived from consideration of Munc18c interactions with members of the SNARE complex; an increase in associations between Munc18c and SNAP23 in the absence of any change in the number of associations between Munc18c and VAMP2 indicates that Munc18c is associated with the Sx4–VAMP2 pool and not with the Sx4–SNAP23 pool. We have used these data to propose that there are two pools of Sx4 in adipocytes under basal conditions, one in complex with SNAP23 and the other in complex with VAMP2, and that Munc18c is associated with the latter, but not with the former .
Two pools, different functions?
Insight into possible functional roles of these distinct Sx4 pools was provided by our biochemical studies. A productive ternary SNARE complex requires interactions between Sx4, SNAP23 and VAMP2. Using recombinant proteins, we have observed that Sx4–SNAP23 binary complexes, when incubated with VAMP2, rapidly formed stable ternary complexes at a rate higher than that achieved without pre-forming the t-SNARE complex . By contrast, pre-forming Sx4–VAMP2 binary complexes was inhibitory to SNARE complex formation . Such observations have suggested that the two Sx4 pools are distinct functionally; we have suggested that the Sx4–SNAP23 pool is required for the constitutive recycling of GLUT4 through the plasma membrane observed in the absence of insulin, and that the pool of Sx4 associated with VAMP2 (and Munc18c) provides a reservoir of Sx4 that can be mobilized rapidly upon demand (i.e. in response to insulin).
On the basis of our data, and consistent with studies using TIRF (total internal reflection fluorescence) microscopy that demonstrate that the majority of GLUT4 is within 100 nm of the plasma membrane [21–23], we have suggested that VAMP2–GLUT4-containing vesicles docked with Sx4 at the cell surface fuse with the plasma membrane upon insulin stimulation.
How might Sx4–VAMP2–Munc18c complexes be regulated?
The study from the Lienhard laboratory has revealed that Munc18c is phosphorylated on tyrosine residues in response to insulin, specifically on Tyr521 . We have shown that the insulin receptor can directly phosphorylate Munc18c in vitro on this residue , and others have revealed that a phosphomimetic mutant (Y521E) of Munc18c rescued defective insulin-stimulated GLUT4 translocation in Munc18c-knockdown adipocytes . Such data support a model in which phosphorylation of Munc18c is required for an essential step in GLUT4 vesicle fusion with the plasma membrane.
Recent data from our group offer new insight into the mechanism of this important action of Munc18c. We have reported previously that recombinant Munc18c inhibited in vitro fusion catalysed by Sx4–SNAP23–VAMP2 in liposome fusion assays . Consistent with this, we have observed that Munc18c lowers the rate of Sx4–SNARE complex formation in vitro . By contrast, Munc18c containing the Y521E mutation ‘accelerates’ the rate of SNARE complex formation . This seems to be achieved through alleviation of the inhibitory effect of the VAMP2–Sx4 binary interaction on SNARE complex formation. Hence we have proposed that Munc18c acts as a molecular switch to ‘mobilize’ a silent pool of Sx4 to facilitate increased delivery of GLUT4 to the plasma membrane in response to insulin. In agreement with this, studies from our group and others have revealed that phosphorylation of Munc18c dissociates the binary Munc18c–Sx4 complex  (see below).
Such observations pave the way to investigate the molecular basis for the action of phosphorylated Munc18c. One function of SM proteins involves switching the conformation of their cognate syntaxin molecules from a ‘closed’ (autoinhibited) conformer to a more ‘open’ fusion-competent form . It is possible that phosphorylated Munc18c facilitates this switch. Our preliminary data supports this notion, as Munc18c-Y521E does not stimulate SNARE complex formation using ‘open’ Sx4 mutants. However, further work will be required to fully define the molecular basis of this switch.
Lessons from other systems
Our data indicate that regulation of SNARE machinery by phosphorylation serves as an interface between signalling pathways and membrane trafficking machinery. Similar observations have been made in other cell systems, including those briefly reviewed below.
In the pancreatic β-cell, glucose triggers the release of insulin-containing secretory granules by regulated exocytosis, a process that shares significant similarity with that of insulin-stimulated GLUT4 vesicle translocation. Interestingly, in β-cell lines, elevations in glucose concentration in the media have been shown to trigger tyrosine phosphorylation of Munc18c, in this case on Tyr219 . Interestingly, this phosphorylation of Munc18c and triggering the dissociation of the Munc18c–Sx4 complex seem to facilitate interaction of Munc18c with the double C2 domain-containing protein Doc2β . Doc2β is a key positive regulator of insulin secretion, hence Thurmond and colleagues have proposed that a network of interactions between Sx4, Munc18c and Doc2β could regulate fusion of insulin granules with the cell surface . Modelling studies of Munc18c from the same group have suggested that Tyr219 is ‘buried’ under basal conditions, but that phosphorylation of this site results in changes in Sx4 contact surfaces, thereby facilitating Doc2β interactions, resulting in a potential switch of function of the fusion machinery.
The potential to regulate SNARE function via phosphorylation is not confined to insulin-sensitive or insulin-producing cells. Numerous stimuli have been shown to promote SNARE phosphorylation, including phosphorylation of syntaxins, VAMPs and regulator SM proteins. For example, phosphorylation of Sx3B in ribbon synapses by Ca2+/calmodulin-dependent kinase on Thr14 alters affinity for SNAP25 and the formation of SNARE complexes ; interestingly, this region of Sx3B may, by analogy with other syntaxins, be involved in interaction with the regulatory molecule Munc18a . Similarly, dramatic and rapid alterations in protein kinase C-mediated phosphorylation of SNAP25 are observed in mouse brains during cold-water restraint stress . Interestingly, this was confined to defined brain regions, suggesting that both central and peripheral regulations of neuronal function may be exerted through such a mechanism .
The longin domain of the v-SNARE TI-VAMP (tetanus-insensitive VAMP) (VAMP7) plays an autoinhibitory function, acting to inhibit SNARE complex formation and membrane fusion. Galli and colleagues observed that TI-VAMP is phosphorylated by c-Src on Tyr45, and that replacement of this residue with phenylalanine activated both t-SNARE binding and exocytosis, thus providing a further example of tyrosine phosphorylation cascades regulating SNARE complex formation . A further example is suggested from studies in which the T-cell protein tyrosine phosphatase, TC48, was found to interact with Sx17 . Pervanadate treatment enhanced tyrosine phosphorylation of Sx17, and that this was reduced by overexpression of TC48. β-COP (β-coatomer protein) dispersal induced by Sx17 over-expression was reduced by pervanadate treatment, and a phosphomimetic mutant of Sx17 (Y156E) showed reduced interaction with COPI (coatomer protein I) vesicles. Such data hint at a balance of phosphorylation/dephosphorylation events controlling traffic through the ER (endoplasmic reticulum)–Golgi compartments in secretory cells .
Potential therapeutic interventions?
Our hypothesis that Sx4 is organized into distinct pools which can be recruited in response to a stimulus offers the potential for therapeutic intervention. One possible example of this has been provided by studies of insulin release from β-cells of diabetic patients. It is clear that dysfunctional insulin release precedes both Type 1 and Type 2 diabetes. One prediction of this model would be that overexpression of Sx4 in β-cells from such patients could reduce insulin secretion defects. Consistent with this, transduction of β-cells to overexpress Sx4 was found to significantly enhance insulin release, suggesting that Sx4 titration is sufficient to significantly improve insulin secretory function . We have suggested that this may arise by altering the distribution of Sx4 among the different plasma membrane pools alluded to above. Further work to examine this scenario in other regulated secretory systems is warranted clearly. Platelet secretion provides another illustrative example. Here, secretion is a key therapeutic target as a means to modulate haemostasis. Platelet secretion is regulated by phosphorylation of SNAP23, and phosphorylation of this site is controlled by IKKβ (inhibitor of nuclear factor κB kinase β) . Strikingly, inhibition of this kinase in vivo was found to prolong tail bleeding times in mice, raising the important observation that IKK inhibitors may adversely affect haemostasis .
It would be remiss not to mention other post-translational modifications and their potential impact on the SNARE machinery. Myristoylation and ubiquitylation of SNAREs have been discussed elsewhere, but other less well-studied modifications, such as nitrosylation, should also be considered. This is particularly intriguing, given the potential for dysfunction under conditions of nitrosative stress associated with some clinical pathologies. Sx4 is modified by S-nitrosylation in β-cell lines in response to short-term treatment with inflammatory cytokines, and this correlates with dysfunctional glucose-stimulated insulin secretion . Such studies reinforce the notion that the machinery that regulates membrane trafficking may offer unique nodes of therapeutic intervention. Collectively, such examples hint at complex regulatory circuitry underpinning SNARE complex regulation and function. Further studies are required to understand the nature of these circuits in different cell contexts.
Our studies have offered the first demonstration of insulin-dependent changes in associations between members of the SNARE complex that regulate insulin-stimulated GLUT4 translocation. We have suggested that negative regulation of Sx4-containing complex formation by VAMP2 is alleviated by insulin-dependent phosphorylation of Munc18c. Not only does this offer a unique model of insulin action, but also, given the central importance of SNAREs and SM proteins in all membrane trafficking events, our studies open the gates for further insight into biological mechanism and also for potential therapeutic development.
Membrane Morphology and Function: A Biochemical Society Focused Meeting held at Hotel del Camerlengo, Fara San Martino, Abruzzo, Italy, 5–8 May 2014. Organized and Edited by Banafshé Larijani [IKERBASQUE, Basque Foundation for Science and Unidad de Biofísica (CSIC-UPV/EHU), University of the Basque Country, Spain] and Marco Falasca (Barts and The London School of Medicine and Dentistry, U.K.)
glucose transporter type 4
GLUT4 storage vesicle
inhibitor of nuclear factor κB kinase
proximity ligation assay
soluble N-ethylmaleimide-sensitive fusion protein-attachment protein receptor
23 kDa synaptosome-associated protein
vesicle-associated membrane protein
This study in the laboratories of G.W.G. and N.J.B. was funded by the Biotechnology and Biological Sciences Research Council (studentships to D.K. and H.L.B.), the Diabetes UK and the Medical Research Council (grants to G.W.G. and N.J.B.).