How Are GSVs Formed?

In general, the formation of vesicles is catalyzed by a budding stage where coat proteins are recruited to the donor membrane via interactions with Arf GTPases and adaptor proteins, which in turn interact with cargo on the donor membrane (38, 39, 40). GSVs may form either from endosomes or from parts of the TGN or both. GSV biogenesis has been studied in two ways: by examining either the entry of newly synthesized GLUT4 into this compartment or the re-entry of GLUT4 from the cell surface back into GSVs. One assumes that the biogenesis machinery should be the same in both cases because GSVs are renewable (41). In the case of the former, investigations have examined the time necessary for GLUT4 to acquire insulin responsiveness after electroporation of a green fluorescent protein (GFP)-tagged GLUT4 construct into adipocytes. In 3T3-L1 cells, this process was found to be slow [6–9 h (Ref. 42)], whereas in rat adipocytes it was faster [<3.5 h (Ref. 25)]. This discrepancy warrants further investigation (42). Despite these limitations, a role for Golgi localizing γ adaptin ear homology domain, ARF binding (GGAs) in GLUT4 sorting from the TGN to GSVs has been described (42). This is intriguing because GGAs bind to ubiquitinated cargo, so it will be exciting to determine whether GLUT4 and IRAP are ubiquitinated (43). Curiously, however, GGAs also play a role in sorting membrane proteins such as the mannose 6-phosphate receptor (MPR) to the lysosome (44). Because the MPR is not found in GSVs (36), this requires an additional sorting step in addition to the GGA-dependent step to distinguish between sorting to GSVs vs. lysosomes. As a possible explanation for this additional sorting step, it has been shown that sortillin, an MPR-like protein, may play an important role in GLUT4 trafficking by binding to lumenal domains in GSV cargo proteins such as GLUT4 and IRAP (45). This is surprising, however, because substitution of the IRAP lumenal domain with that of the transferrin receptor, a protein not transported to GSVs, does not perturb the insulin-responsive sorting of IRAP (46). Moreover, a GLUT4/GLUT1 chimera containing the cytosolic domains of GLUT4 and the lumenal domains of GLUT1 traffics like GLUT4 but not GLUT1 (47).

McGraw and colleagues (48) have studied the trafficking of GLUT4 from the PM after insulin stimulation. They have carefully mapped the time-course by which GLUT4 escapes ablation with Tf-HRP, thus measuring the sorting of GLUT4 out of endosomes. Using this approach, they find that 40–50% of GLUT4 is localized to endosomes in 3T3-L1 adipocytes under basal conditions. Based on these findings, it has been argued that GSVs are formed from endosomes and that GLUT4 transits from endosomes to GSVs with a t_1/2 of 20 min (48). One potential limitation of these studies is that the fate of GLUT4 after its exit from endosomes was not analyzed. This may be important because it has been reported that GLUT4 rapidly transits through endosomes to a subdomain of the TGN also containing the t-SNAREs Syntaxins 6 and 16 (49, 50). Because the Tf receptor does not follow this route (49, 50), this would also register as a non-ablatable compartment in the studies of McGraw and colleagues. It is curious that the TGN is absent from the McGraw model of GLUT4 trafficking because many studies (50, 51, 52) have shown an important role for the TGN in GSV biogenesis. A central feature of the McGraw model as described below is that a large proportion of GLUT4 (~50%) is found in endosomes at steady state. However, this observation is based on the use of the Tf-HRP ablation technique. In view of the enzymatic nature of this technique, if one molecule of Tf-HRP were to be mis-sorted into a postendosomal compartment, a likely outcome based on the nature of endosomal protein sorting, all molecules found in this compartment would be ablated and register as part of the endosomal recycling system. Because GLUT4 is likely concentrated at this step, this may vastly overestimate the amount of GLUT4 in the recycling endosomal system under basal conditions. A further caveat concerns the integrity of the cells used for experimentation. A considerable proportion of GLUT4 is targeted to endosomes in preadipocytes (53). However, as these cells differentiate into mature adipocytes, GLUT4 is withdrawn from endosomes into a more specialized compartment, presumably GSVs (15, 53). It is noteworthy that Karylowski et al. (48) use a model system involving trypsinization of mature adipocytes followed by electroporation, after which the cells are reseeded and studied 2–3 d later. Morphologically, these cells do not represent mature adipocytes due to the conspicuous absence of lipid droplets (48). Hence, whereas this model system may be useful for studying certain aspects of GLUT4 trafficking, we urge caution in using such a system for detailed modeling of GLUT4 sorting between endosomes and postendosomal compartments.

How Do GSVs Traffic in the Absence and Presence of Insulin?

One of the conundrums in the field concerns the fate of GSVs once formed. GLUT4 could be retained in either a static or a dynamic pool (Fig. 2). The static model proposes that GLUT4 may be retained intracellularly by an interaction between GSVs and retention machinery such as TUG (tether, containing a UBX domain, for GLUT4) (54), p115 (55), or by restricting the access of GSVs to the vesicle tethering/docking/fusion apparatus at the cell surface. In the dynamic model, GLUT4 is constantly cycling between two intracellular compartments such as the endosome and the TGN. A dynamic pool could be either endosome derived or comprise a dynamic cycle between TGN and endosomes (Fig. 2, B and C). Distinguishing between these models is important because it will potentially define at which steps acute GLUT4 translocation is regulated. In the static model, regulatory factors likely act at the anchor or the vesicle docking/fusion machinery. In the dynamic model, however, vesicle formation or intracellular vesicle fusion may be important regulatory nodes. McGraw and colleagues (48) have used the model system described above to analyze the kinetics of GLUT4 recycling in adipocytes and report that GLUT4 is rapidly internalized from the cell surface into endosomes from where it exits into GSVs with a t_1/2 of 20 min. They also find that the entire cellular cohort of GLUT4 (Fig. 2A) recycles via the adipocyte plasma membrane in the basal state with a t_1/2 of 230 min. The only way to rationalize this slow recycling rate in lieu of the high (~50%) proportion of GLUT4 in endosomes at steady state is to invoke an additional sorting step whereby GLUT4 in GSVs continuously cycles back to endosomes in the absence of insulin (otherwise, the endosomal compartment would be rapidly depleted of GLUT4). However, because GLUT4 is known to recycle from endosomes back to the cell surface very rapidly (t_1/2 ~ 10 min) (48), this model seems unlikely; otherwise the 50% of GLUT4 in endosomes should freely exchange with the plasma membrane even in the absence of insulin, and this does not appear to be the case even in the studies of Karylowski et al. Hence, again this emphasizes the likelihood that the HRP ablation method overestimates endosomal GLUT4. We do not feel that one can exclude an intracellular cycle for GLUT4, and indeed we have suggested that GLUT4 cycles between GSVs and the TGN (2), the important difference being that this cycle would aid the intracellular sequestration of GLUT4 rather than expose GLUT4 inappropriately to the endosomal recycling system.

An essential difference between the static and dynamic models is that the former, although not excluding an intracellular cycle between GSVs and the TGN, predicts that GLUT4 does not readily exchange with the PM in the absence of insulin, whereas in the latter model it will. McGraw and colleagues (48) report that the entire GLUT4 pool exchanges with the PM with t_1/2 of 230 min, whereas Govers and colleagues (53) have shown that this is not the case. One possibility is that the high basal recycling reported by Karylowski et al. (48) and Martin et al. (56) reflects the state of the cells used in their system (see above), combined with the fact that in their system they don’t actually measure the amount of GLUT4 that recycles as a percentage of the total GLUT4 expressed in the cell. This could be important because Govers et al. (53) report that there is a silent nonrecycling pool of GLUT4 that, if unrecognized, will lead to a significant over estimate in GLUT4 recycling. These discrepancies emphasize the need to develop further assays to study the formation of GSVs and GLUT4 recycling in insulin-sensitive cells. Such systems would optimally measure the recycling of endogenous GLUT4 molecules expressed in their native environment. In addition, further assays enabling analysis of the dynamic movement of GLUT4 both close to the surface and deep within the cell will be required to resolve the possible existence of intracellular futile cycles.

Docking and Fusion

In the docking stage, SNARE proteins on both the GSV and the plasma membrane interact to generate a ternary complex of coiled-coil domains, which is extremely stable (70). VAMP2 has been shown to take part in a number of exocytic fusion events including GLUT4 vesicle fusion (11, 22, 37), aquaporin vesicle fusion (71), and synaptic vesicle fusion with the plasma membrane (72). VAMP2 has been shown to bind the t-SNAREs Syntaxin 1 and Syntaxin 4 in cooperation with the corresponding SNAP proteins SNAP-25 and SNAP-23, respectively (72). To mediate ternary complex formation and fusion, the VAMP2 sequence is dominated by a SNARE motif that can interact with t-SNAREs at the plasma membrane (72). Inhibition of VAMP2 function using either botulinum toxin (B, D, F, and G) or tetanus toxin inhibits GLUT4 vesicle fusion (73, 74, 75). In muscle and fat tissue, insulin stimulation induces the movement of VAMP2 to the plasma membrane with similar kinetics to GLUT4 (36). Regulation of the fusion step at the plasma membrane is further modified by interactions between Syntaxin-4 and the Sec1-family member Munc18c (76, 77).

A picture is now emerging whereby a major effect of insulin is to modify one of the steps in GLUT4 trafficking either at or very close to the plasma membrane. To address this issue, it was essential to develop techniques that could resolve discrete steps of GLUT4 trafficking in more detail. Total internal reflection fluorescence (TIRF) microscopy is a tool that allows fluorescent events occurring within 150 nm of the plasma membrane to be examined at high resolution, such as the fusion of a vesicle containing GLUT4-GFP (78, 79, 80). TIRF microscopy relies on generation of an evanescent wave that forms at a point where the excitation laser beam hits the interface between the glass coverslip and the aquatic phase of the sample at an angle that is greater than the critical angle at which all the light is reflected (81). By applying this technology to GFP-tagged GLUT4 expressed in adipocytes and using image analysis software capable of single particle analysis, it has been possible for the first time in two recent studies to visualize the movement of individual GLUT4 vesicles into the TIRF zone, tethering/docking of the vesicles, and their subsequent fusion (78, 80). These studies have shown that the GSV dwell time for the tethering/docking step was moderately decreased by insulin (78). In concert with a major effect of insulin distal to the tethering/docking step, the rate of GSV fusion frequency was increased in insulin-treated cells (78, 80). Intriguingly, the study by Bai et al. (78) showed that the majority of tethered/docked vesicles subsequently undocked from the membrane and only a minority of vesicles subsequently fused. In addition, this study also showed that tethering/docking was regulated by PI3-kinase and AS160 (78). The importance of these data is that they open the way for detailed molecular studies in the future. For example, the consequences of disrupting exocyst vs. SNARE function on the various parameters described above will provide new insights into this process and help to sharpen the focus of future efforts aimed at dissecting the major point of action of insulin and possibly other agonists such as exercise. In addition, by combining this approach with even further advances such as FRET, it should be feasible to monitor the assembly of polypeptide complexes in living cells so that we can begin to dissect, for example, which aspect of SNARE function might be regulated by insulin.

We thank all members of the James lab for useful discussions regarding the manuscript. D.E.J. is a senior principal research fellow of the National Health and Medical Research Council. We have attempted to acknowledge the contribution of colleagues where appropriate, but due to space limitations this was not always possible.