Biochemical Journal

Review article

Insulin action on glucose transporters through molecular switches, tracks and tethers

Hilal Zaid, Costin N. Antonescu, Varinder K. Randhawa, Amira Klip


Glucose entry into muscle cells is precisely regulated by insulin, through recruitment of GLUT4 (glucose transporter-4) to the membrane of muscle and fat cells. Work done over more than two decades has contributed to mapping the insulin signalling and GLUT4 vesicle trafficking events underpinning this response. In spite of this intensive scientific research, there are outstanding questions that continue to challenge us today. The present review summarizes the knowledge in the field, with emphasis on the latest breakthroughs in insulin signalling at the level of AS160 (Akt substrate of 160 kDa), TBC1D1 (tre-2/USP6, BUB2, cdc16 domain family member 1) and their target Rab proteins; in vesicle trafficking at the level of vesicle mobilization, tethering, docking and fusion with the membrane; and in the participation of the cytoskeleton to achieve optimal temporal and spatial location of insulin-derived signals and GLUT4 vesicles.

  • actin dynamics
  • glucose transporter 4 (GLUT4)
  • insulin signalling
  • muscle cells
  • phosphoinositide
  • vesicle traffic


Glucose is the major fuel of most mammalian cells, and insulin is its principal regulator. A key function of insulin is to allow glucose entry into muscle and fat cells. Skeletal muscle is the largest site for disposal of dietary glucose, and GLUT4 (glucose transporter-4) is the port of entry for glucose into this tissue. One of the most fascinating biological discoveries of the past two decades is the mechanism whereby insulin regulates GLUT4 availability at the surface of muscle and fat cells. It is noteworthy that the original ‘translocation hypothesis’, whereby insulin would rapidly increase the number of glucose transporters at the fat cell membrane at the expense of an intracellular pool, was postulated long before the cloning of GLUT4 or the availability of dynamic assays of vesicle movement within cells. To date, the concept of GLUT4 translocation is highlighted in textbooks and pharmaceutical pamphlets, and in the scientific literature the transporter has been the subject of almost 4000 publications. The dynamic nature of this process has been further elucidated through elegant and creative imaging methodologies, implemented by many groups around the world. Understandably, the regulation of GLUT4 by insulin has been the subject of scholarly reviews in previous years [18]. In the present review we attempt to cover the fundamental concepts that have been experimentally tested and to highlight some ideas that are currently being discussed on the dynamic nature of GLUT4 trafficking and how insulin-derived signals regulate it. We also review the dynamic regulation of the cytoskeleton by the hormone as a potential mechanism for integration of insulin signalling and GLUT4 vesicle trafficking. Particular attention is given to the current debate of whether insulin signals impinge on one or numerous steps along the intracellular itinerary of GLUT4.

Finally, we focus the present review on studies in both muscle and fat cells. Of note, human and rodent skeletal muscle or muscle cells respond to insulin with a 2- to 3-fold gain in surface GLUT4 units; human adipocytes mount a 2- to 5-fold gain and their rodent counterparts typically respond with a 10- to 20-fold gain. Overall, the basic principles of GLUT4 translocation are similar in all of these systems, and the quantitative differences may be related to the fidelity of each step and, in some cases, to the reliance on distinct isoforms of the individual insulin signals and vesicle trafficking machinery.


Kinetic models of GLUT4 cycling

In unstimulated muscle or adipose cells, 4–10% of GLUT4 is located at the cell surface and >90% at intracellular compartments. This steady-state distribution of GLUT4 is the balance of its fast endocytosis and slow recycling, which in muscle cells have t1/2 values of 3.5 and 120 min respectively [9,10]. The cycling properties of GLUT4 in muscle cells are very similar to those in adipose cells, although its recycling is even slower in the latter (t1/2 is 230 min) [7,11,12]. Insulin shifts the steady-state distribution of GLUT4 from intracellular pools towards the PM (plasma membrane) largely by elevating its exocytic rate. In muscle cells, insulin does not alter its endocytic rate [9,10,12a], but it modestly reduces it in adipocytes [13,14]. Within 10–15 min of hormonal stimulation of muscle cells, surface GLUT4 levels double and remain stable for at least 30 min. Figure 1 illustrates the gain in surface GLUT4 in muscle and adipose cells in culture. GLUT4 is retrieved from the PM by clathrin-dependent and -independent routes of endocytosis [12a,1416]. Internalized GLUT4 is routed through a poorly understood series of endosomal compartments, which ultimately concentrate GLUT4 into SCs (‘specialized’ or ‘storage’ compartments). The SC is also known as the insulin-responsive compartment or GSVs (GLUT4-storage vesicles), described as discrete endosomes and/or regions of the TGN (trans-Golgi network) [4,17].

Figure 1 GLUT4 concentrates in perinuclear compartments and translocates to the PM upon insulin stimulation

Adhered L6 myoblasts (A), rounded-up L6 myoblasts (B) or L6 myotubes (C) stably expressing GLUT4myc with or without insulin stimulation (100 nM insulin for 20 min) were subsequently fixed and processed for immunofluorescence using an anti-myc antibody as described previously [10,96,153]. (i) Permeabilized cells showing total cellular GLUT4myc. (ii) Non-permeabilized cells showing GLUT4myc exclusively at the cell surface. (D) 3T3-L1 adipocytes with or without insulin stimulation (100 nM insulin for 20 min) were fixed, permeabilized and labelled with anti-GLUT4 antibodies to detect total cellular GLUT4. Scale bars: (A and C), 20 μm; (B and D), 5 μm.

The complex GLUT4 itinerary has made it difficult to characterize SC biochemically or spatially. Nonetheless, we and others have provided evidence of partial segregation of GLUT4 from the REs (recycling endosomes), as follows: (i) by velocity gradient centrifugation of adipose cells, GLUT4 is found in two endomembrane fractions, only one containing the transferrin receptor, TfR, a marker of the RE [18]; (ii) by ablation of the TfR-containing compartment in intact 3T3-L1 adipocytes, only 40% of GLUT4 is affected [19,20]; (iii) by immunoprecipitation of GLUT4-rich endomembranes from L6 muscle cells, recycling proteins are partly excluded [21]. These and related studies have lent support for the existence of two pools of GLUT4, differentiated by their content of RE markers. The SC is considered to be the non-RE pool of GLUT4, and it has been debated whether it is static in the basal state or whether it slowly recycles (either to the PM or back to the RE). In L6 muscle cells, the entire intracellular complement of GLUT4 recycles to the PM within 6 h [9,10,22], whereas such behaviour is controversial in 3T3-L1 adipocytes. McGraw and co-workers [11] suggested the dynamic-exchange model, whereby GLUT4 continuously cycles between the SC and RE, and can slowly escape this dynamic exchange to reach the PM. Insulin would shift the exocytic and endocytic rates, thereby elevating GLUT4 at the PM [11]. An alternative model, termed static retention, was proposed by James and co-workers [23], suggesting that GLUT4 in the SC/GSV never reaches the PM under basal conditions. Instead, GLUT4 would either be unable to cycle out of the SC/GSV in the basal state, or engage in an unproductive idle cycle between the TGN and/or RE and the SC/GSV [23]. Insulin would either shift the balance towards the RE or trigger GLUT4 exit directly from the SC/GSV to the PM, and would further stimulate the fusion of RE-derived vesicles with the PM, effectively drawing GLUT4 away from idle intracellular cycling [23]. Both models propose intracellular idle cycling of GLUT4 (whether or not including the TGN), but critically differ in the proposed ability or inability of GLUT4 to slowly recycle to the PM in the basal state. Figure 2 depicts the main differences between these models. Interestingly, a recent report suggests that the different results that support each model may have arisen from the culture conditions employed [24]. In confluent cultures, as used to propose the static model, approx. 80% of GLUT4 is retained in non-recycling compartments. However, re-plating of the cells, as used to propose the dynamic model, shifts 80% of GLUT4 to the recycling mode.

Figure 2 Static and dynamic models of GLUT4 cycling

(A) In the static-retention model, GLUT4 in the SC does not recycle to the PM, but may engage in an idle cycle with the TGN. Insulin (ins) redistributes GLUT4 through exit from the SC directly to the PM. (B) In the dynamic model(s), GLUT4 recycles to the PM through the RE, but an idle cycle between the RE and SC reduces the efficiency of recycling. Insulin (ins) redistributes GLUT4 to the PM through exit from the SC directly to the PM, but potentially also through the RE. See text for details.

Contrary to the case of 3T3-L1 cells, in L6 myoblasts, GLUT4 recycles to the PM continuously, if slowly, in either confluent or sparse cultures [9,10,22]. As only 50% of intracellular GLUT4 coincides with the TfR in REs at steady state measured by a ‘contents mixing approach’ (V.K. Randhawa and A. Klip, unpublished work), it is conceivable that, in myoblasts, GLUT4 undergoes idle cycling between the SC and RE, and this retards, but does not preclude, its availability to recycle to the PM.

GLUT4 compartments

As described above, in unstimulated cells, nearly half of the intracellular GLUT4 complement fails to share the lumen of intracellular compartments containing markers of the RE [25,26]. Important questions remaining to be answered are whether GLUT4 vesicles reaching the PM in the basal and insulin-stimulated states emanate from the same or different compartments (RE or SC), and whether insulin-derived signals regulate intracellular idle cycling, exit from these compartments and/or their interaction and fusion with the PM. Our current understanding of these events is summarized next.

One way to analyse the provenance of GLUT4 vesicles reaching the PM in the basal or insulin-stimulated states is to compare their biochemical characteristics. Although this has not been feasible through standard biochemical approaches, use of live-cell imaging and TIRF (total internal reflection fluorescence) microscopy should be more revealing. In addition, functional assays have aided in this distinction. Fusion of GLUT4 with the PM requires the v-SNARE (vesicular soluble N-ethylmaleimide-sensitive factor-attachment protein receptor) VAMP2 (vesicle-associated membrane protein 2), whereas fusion in the basal state is independent of this protein ([27,28] and references therein). Moreover, in the basal state, GLUT4 insertion into the membrane depends on the v-SNARE VAMP7/TI-VAMP (tetanus neurotoxin insensitive VAMP) [29]. Neither basal-state recycling nor insulin-dependent gain in surface GLUT4 depend on VAMP3 [2729]. Collectively, these studies suggest that GLUT4 vesicles exiting the RE and SC may be differentiated from each other and from other recycling vesicles by virtue of their VAMP isoform. Moreover, these findings raise the hypothesis that VAMP7 is present in GLUT4 vesicles emanating from the RE and VAMP2 in GLUT4 vesicles emanating from the SC.

The differential presence of VAMP2 or VAMP7 could potentially be used to isolate and characterize the SC, and efforts to this effect are underway. To date, characterization of GLUT4-containing intracellular compartments separated by differential centrifugation and relatively devoid of cargo-loaded TfR [18] revealed enrichment in IRAP (insulin-regulated aminopeptidase), IGF (insulin-like growth factor) II/mannose 6-phosphate receptor, sortilin, VAMP2 and VAMP3 [3032], yet it is not clear whether all of these are specific markers of the SC. Through a separate approach, immunopurified GLUT4-containing compartments examined by MS showed enrichment in a cohort of proteins, among which IRAP and VAMP2 stood out [33]. More discerning purifications are needed to identify bona fide markers of the SC. Towards this goal, some studies have investigated the proteins that bind directly to GLUT4. A detailed analysis of these proteins [including TUG (tether containing UBX domain for GLUT4), aldolase and Ubc9] has been presented in [6] and is not further discussed in the present review.

Strikingly, it is not known where the SC is located within the adipose or muscle cells. By immunofluorescence, the majority of GLUT4 is found in a perinuclear region in both 3T3-L1 adipocytes and L6 myoblasts and myotubes. Figure 1 illustrates this perinuclear GLUT4 localization in these cell types. Perinuclear GLUT4 exhibits partial co-localization with RE and TGN markers, and significant co-localization with VAMP2 [29,34,34a]. However, such co-localization is limited by the resolution of light microscopy, and does not prove spatial co-incidence. Selective compartment ablation and high-resolution fluorescence imaging techniques are needed for quantitative co-localization of the SC. Moreover, it has also been proposed that the functional SC is distributed throughout the cytosol and near the PM, particularly in mature adipocytes [35]. Once incorporated into the cell surface, the distribution of GLUT4 is also not homogeneous: in mature skeletal muscle, GLUT4 is preferentially directed towards the transverse tubules [36,37]. Two interesting recent studies suggest that local submembrane depots furnish the transporter to the transverse tubules and that tubular GLUT4 is a major determinant of insulin-dependent glucose uptake [38,39]. In adipose cells, some reports assign transitory deposition of surface GLUT4 in areas of caveolae [40], and the functional significance of this distribution merits further investigation.

Regulated steps in GLUT4 exocytosis

This question has been the subject of recent, elegant reviews [1,3,7], reflecting the burgeoning information and continuously revised models that should in the future be reconciled into a molecular and temporal map of the participating events. Here we offer a bird's eye view of the current knowledge of these phenomena.

Vectorial transfer

There is current debate whether: (i) insulin induces net vectorial transfer of GLUT4-containing vesicles towards the cell periphery where additional signals accelerate their fusion; or, (ii) GLUT4-containing vesicles continuously travel to and from the cell periphery, and insulin-derived signals only promote their interaction and fusion with the PM. The debate arose from the difficulty in detecting net vectorial transfer, and from the comparatively well-documented regulation of events near/at the PM. The observation that purified PM fractions from insulin-stimulated rat adipocytes retain the molecular information to promote fusion with intracellular GLUT4-containing vesicles derived from unstimulated cells suggested that a major contribution to regulation occurs at the PM [41]; however, this extrapolation from in vitro studies does not rule out that there can be important regulation in order to make the vesicles available to those membrane events in intact cells. Even when the quantitative contribution of PM-proximal events in adipocytes is large, there is support for regulation of earlier steps in the GLUT4 itinerary, as follows: by labelling GLUT4 at the surface of muscle cells and tracking their intracellular itinerary, it was observed that the transporters rapidly internalize (t1/2 3.5 min) from the cell surface, reach the RE within 20 min and exit this compartment 20 min later [10]. As recycling back to the PM is very slow (t1/2 120 min), these results suggest that GLUT4 exiting the RE is probably routed to the SC. Insulin significantly accelerated the transit of GLUT4 in and out of the RE [10], and such acceleration required input of PI3K (phosphoinositide 3-kinase) and PKB (protein kinase B) (see section ‘Signals regulating GLUT4 traffic’ below), suggesting that refilling the SC from the RE is under insulin regulation. Additional observations supporting the concept that insulin-derived signals regulate GLUT4 exit from intracellular compartments are: (i) a reorganization of GLUT4 in the perinuclear region that is not the result of initial changes at the muscle cell surface [34a]; (ii) in vitro budding of GLUT4-containing small vesicles (that do not sediment at 16000 g for 20 min) from purified intracellular membranes, and increased abundance of these vesicles in isolates from insulin-stimulated cells [42]; and (iii) increased abundance of GFP (green fluorescent protein)–GLUT4 moving along linear tracks in live cells [43].

Following release from idle cycling, GLUT4-containing vesicles translocate towards the cell membrane where they tether, dock and fuse. Although different definitions have been proposed for tethering and docking, it can be assumed that tethering is a cortical immobilization that primes or directs vesicles towards docking sites. Docking is the high-energy, long-lasting but non-covalent binding of GLUT4 vesicles to the PM, probably mediated by SNARE proteins, and fusion is the integration of the vesicular bilayer into the PM, resulting in exposure of the lumen of the GLUT4 vesicle to the extracellular medium. There is suggestive evidence that each of these three distinct steps is regulated by insulin.

Vesicle tethering

The exocyst, a tethering machinery conserved from yeast to mammals, is a stable complex of proteins Sec3p, Sec5p, Sec6p, Sec8p, Sec10p, Sec15p, Exo70p and Exo84p [44]. Among them, Exo70 is thought to play a central role as it interacts with almost all of the other subunits [45,46]. Interestingly, in 3T3-L1 adipocytes, Exo70 translocates to the PM upon insulin stimulation, through activation of the Rho GTPase TC10 [47]. The C-terminal of Snapin binds to the coiled-coil domain in Exo70, and Snapin expression knockdown somewhat inhibited glucose uptake in 3T3-L1 adipocytes [48]. It is still unknown, however, whether GLUT4 vesicles physically dock through the exocyst, and the participation of Exo70 in GLUT4 translocation was recently challenged [48a]. Additional or alternative components of the functional peripheral tether of GLUT4 vesicles could include elements of the cortical actin that remodels in response to insulin (see the section ‘Cytoskeleton input on GLUT4 translocation’ below).

Vesicle docking and fusion

Owing to insufficient sensitivity of light microscopy techniques, it had been difficult to distinguish between docking and fusion events in a quantitative fashion, and to define their regulation by insulin. Use of TIRF and fluorescent probes sensitive to the extracellular pH have resulted in breakthroughs in this regard. As TIRF only analyses events within 200 nm of the surface, these studies do not rule out possible regulation of earlier steps such as vesicle budding or vectorial traffic. By quantitative analysis of TIRF images in live 3T3-L1 adipocytes expressing GFP–GLUT4 chimaeras, three groups reported that GLUT4 vesicles are mobile in the vicinity of the membrane, and insulin increases the number of vesicles immobilized in that zone [35,49]. This has been interpreted as reflecting increased docking, although it could also result from tethering. Moreover, a study challenged this tenet, proposing instead that immobilization occurs post-fusion and reflects GLUT4 molecules in clathrin lattices [43]. Nonetheless, because interfering with PI3K or AS160 (Akt/PKB substrate of 160 kDa) signalling reduced the insulin-induced peripheral immobilization of GLUT4 and these signals contribute to net GLUT4 exocytosis, it is likely that cortical immobilization of GLUT4 is related to tethering/docking rather than its endocytosis.

Analysis of the dwell time of each vesicle at the membrane also suggested that docking is an obligatory step prior to vesicle fusion with the membrane. Using a GLUT4 surrogate (IRAP) linked to a GFP sensitive to the pH of the extracellular medium (pHluorin), Jiang et al. [50] recently showed that IRAP-containing vesicles ‘dock’ and fuse in response to insulin, and that AS160 regulates the docking step.

SNARE proteins are thought to mediate both docking and fusion events, and hence it will be important to use this IRAP–pHlouorin reporter to differentiate the step(s) that they mediate in the GLUT4 interaction with the PM. The evidence that SNARE proteins ultimately mediate exposure of GLUT4 at the cell surface is vast. As mentioned earlier, the SC contains the v-SNARE VAMP2, which binds to its cognate t-SNARE (target SNARE) receptors syntaxin-4 and SNAP (soluble NSF-attachment protein) 23 at the PM [28,51]. Whether insulin promotes this interaction or simply increases the proportion of GLUT4 vesicles available for SNARE complex formation is not known. It is well documented that VAMP2, syntaxin4 and SNAP23 are the elements of the SNARE complex involved in docking/fusion of insulin-sensitive GLUT4 vesicles with the PM. This has been determined by interfering with SNARE complex formation by either selective toxins (cleaving VAMP2), gene knockdown (targeting VAMP2), neutralizing antibodies (binding to SNAP23) or peptides mimicking binding regions (of all three proteins) [28]. The SNARE complex consists of four parallel α-helices formed from the coiled-coil segments of SNARE proteins in two opposite membranes [52]. Prior to docking/fusion, cis-SNARE complexes exist on vesicles and must be disassembled to allow trans-SNARE complex formation. cis-SNARE complex disassembly is driven by the ATPase activity of NSF (N-ethylmaleimide-sensitive factor). NSF recruitment to the cis-SNARE complex requires SNAP [53]. NSF is believed to associate with GLUT4 vesicles and cell membranes, but this interaction itself does not impact on the formation of fusion complexes [54]. ATPase-deficient NSF (that binds SNAREs but cannot disassemble them) predominantly affected intracellular membrane fusion events involved in GLUT4 cycling from the endosomal system to the SC, but not to the PM [55] in rat adipocytes. The same mutant precluded basal state and insulin-dependent gain in surface GLUT4 in muscle cells [22].

The regulation of SNARE complex formation is believed to be exerted via SM (Sec1p/Munc18) [56]. Among the three mammalian SM genes, Munc18c is involved in GLUT4 trafficking, stabilizing syntaxin in the closed inactive conformation in the basal state. Although there is good evidence from fixed cells or in vitro analysis that Munc18c dissociates from syntaxin4 upon insulin stimulation [57,58], by fluorescence correlation spectroscopy of live cells it was determined that Munc18c instead switches to a different binding site on the t-SNARE, thereby allowing VAMP2 and GLUT4 vesicle docking at the PM [59].

The basic parameters of vesicle vectorial transfer, tethering, docking and fusion are illustrated in Figure 3(A), along with the insulin-derived signals that may control them (see below).

Figure 3 Consolidated insulin signalling pathways regulating GLUT4 traffic

(A) Stages of GLUT4 exocytosis regulated by insulin. Mobilization: GLUT4 vesicles are transported to the cell periphery, possibly along microtubules. Tethering: GLUT4 vesicles are retained near the cell periphery by the remodelled actin cytoskeleton through ACTN4, and/or by the exocyst complex. Docking: GLUT4 vesicles bind to the PM via SNARE complexes. Fusion: irreversible incorporation of GLUT4 vesicles on to the PM is enhanced by insulin through action of Munc18c on SNARE proteins. Also shown are the possible steps of input by PtdIns3P and PtdIns(3,4,5)P3. An animated version of this Figure is available at (B) Consolidated figure summarizing all signalling pathways proposed to lead to GLUT4 translocation to the PM. The insulin receptor phosphorylates IRS1/2, which activates PI3K. The IRS-2 pathway is not involved in the GLUT4 trafficking. Downstream of IRS-1/PI3K, signalling bifurcates into two arms. One arm is characterized by PKB/Akt, which in turn phosphorylates AS160 thus shutting of its GAP activity towards target Rabs. The other arm may involve aPKC and is characterized by Rac and actin remodelling. ACTN4 possibly links GLUT4 vesicles to actin filaments. In addition, insulin can activate the CAP/Cbl pathway leading to TC10 activation, which has an impact on actin remodelling. Several phosphoinositide phosphatases and kinases shown are predicted to regulate these pathways.


There are a number of excellent reviews of insulin signalling [8,60], and we have previously reviewed signalling events leading to glucose uptake [61]. However, as new elements are identified and old controversies are reconciled, we have revisited and updated these events in a concise way, focusing on GLUT4 translocation as the specific outcome. Figure 3(B) presents a working model of insulin signals that impact on GLUT4 trafficking, which summarizes most evidence to date.

IRSs (insulin receptor substrates)

It is well known that insulin binding elicits rapid autophosphorylation of its receptor, followed by binding and tyrosine phosphorylation of IRS-1/2. Using siRNA (small interfering RNA)-mediated knockdown of IRS proteins, it was found that IRS-1, rather than IRS-2, is required for GLUT4 translocation to the muscle cell surface and glucose uptake [62,63]. This finding is consistent with the decrease of GLUT4 translocation observed upon expression of a fragment encoding the PH (pleckstrin homology) domain of IRS-1 that blocked signalling through IRS-1 [64]. A number of studies have also reported differential contributions of IRS-1 and IRS-2 to glucose metabolism. This is outside of the scope of this review, but is summarized in [65].


IRS-1 binds the regulatory p85 subunit of class I PI3K activating its catalytic p110 subunit. The PI3K family enzymes phosphorylate the third hydroxyl position of the inositol ring of phosphoinositides. The PI3K family is subdivided into three classes, the heterodimeric class I and III composed each of a catalytic and regulatory subunit, and class II composed of only a catalytic subunit [66].

Early in 1990, two groups showed that insulin stimulates PI3K activity [67,68]. Those pioneering results were followed by numerous studies geared to identify the precise mechanism of PI3K impact on insulin action. Essential to this quest were wortmannin and LY294002, unrelated inhibitors that block the catalytic activity of the enzyme [69,70], as well as a dominant-negative, class I PI3K mutant (delta-p85) [71,72]. There is a wide consensus that the two chemical inhibitors (at concentrations that inhibit classes I and III but not class II PI3K) or delta-p85 overexpression vastly reduce the insulin-dependent gains in glucose uptake and surface GLUT4 levels in adipose and muscle cells (reviewed in [61]). It also goes unchallenged that overexpression of constitutively active class I PI3K elevates surface GLUT4 levels in adipocytes [71]. These results suggest that one or more of the products of class I PI3K are essential for this function.

In vivo, the major product of class I PI3K is PtdIns(3,4,5)P3, although PtdIns(3,4)P2 and PtdIns3P also rise in muscle and fat cells in response to insulin [69,73,74], and may originate from either activation of class II [producing PtdIns(3,4)P2 and PtdIns(3)P] or class III (producing PtdIns3P), or from further metabolism of PtdIns(3,4,5)P3. It is therefore interesting that PtdIns(3,4,5)P3 introduced into L6 muscle cells was not sufficient to stimulate glucose uptake, although it effectively evoked a gain in surface GLUT4 levels [75,76]. In contrast, PtdIns3P administration induced GLUT4 arrival near the PM that did not fuse with the membrane, and PtdIns(3,4)P2 had no consequence on GLUT4. From those studies a model arose whereby insulin-induced activation of class I PI3K produces PtdIns(3,4,5)P3 that contributes, along with PtdIns3P, to mobilize GLUT4 vesicles towards the surface, whereas only PtdIns(3,4,5)P3 mediates vesicle fusion with the membrane [76]. Consistent with this prediction, PtdIns3P was insufficient to allow vesicle fusion in adipocytes, until removal of a fusion block at the level of Munc18 allowed fusion to proceed [57]. Also, overexpression of the PtdIns(3,5)P2 phosphatase ‘72-5ptase’ (72-kDa inositol polyphosphate 5-phosphatase) enhanced the production of PtdIns3P, and promoted GLUT4 vesicle mobilization [77]. Conversely, overexpression of a PtdIns3P phosphatase, myotubularin, reduced insulin-dependent GFP–GLUT4 movement to the cell periphery [78], further substantiating the participation of this phosphoinositide in insulin action. PtdIns3P is a substrate of the enzyme PIKfyve (FYVE-domain containing PtdIns3P 5-kinase), an enzyme also involved in insulin signalling (see below).

The model of dual input of PtdIns(3,4,5)P3 and PtdIns3P was further supported and expanded by the finding that PtdIns3P is produced by other members of the PI3K family. Indeed, PI3K-II C2α is activated in response to insulin [79,80] and generates PtdIns3P in L6 muscle cells upon insulin stimulation [74]. The small GTPase TC10 appeared to be needed for such PI3K-II activation based on overexpression of dominant negative or constitutively active TC10 mutants. PI3K-II knockdown attenuated GLUT4 translocation and glucose uptake in the insulin-stimulated state [74], supporting the participation of its products in GLUT4 trafficking.

The dual participation of PtdIns(3,4,5)P3 and PtdIns3P to increase surface GLUT4 does not, however, lead to productive glucose uptake [76]. Hence it has been hypothesized that the gain in glucose may involve activation of the PM-inserted transporters. Moreover, a PtdIns(4,5)P2-binding peptide (PBP10) introduced into adipocytes caused GLUT4 mobilization and insertion in the membrane, without any accompanying gain in glucose uptake [81]. Although the relationship of this peptide to PtdIns(4,5)P2 availability in the cells remains to be determined, the results contribute to the diverse literature supporting separate inputs for GLUT4 translocation and activation. Covering this topic is beyond the scope of the present review, but further discussion can be found in [82].

Downstream of class I PI3K lie three major signalling axes, typified respectively by their initiating signals PKB/Akt, aPKC (atypical protein kinase C) and Rac. We next discuss the involvement of PKB and its targets in GLUT4 trafficking, followed by analysis of the participation of aPKC, and Rac is discussed later in the section on ‘Cytoskeleton input on GLUT4 translocation’ below.


A PKB requirement for GLUT4 trafficking in muscle cells was first shown by Wang et al. [83], a result that was further confirmed by several lines of evidence. First, mice lacking PKBβ displayed insulin resistance of glucose uptake [84]. Secondly, siRNA knock-down in 3T3-L1 adipocytes impaired insulin-dependent 2-deoxyglucose uptake and PM-directed GLUT4 trafficking [85]. From these studies it emerged that PKBβ is more likely to control GLUT4 trafficking than either PKBα or PKBγ. Thirdly, constitutively active membrane-targeted PKB provoked GLUT4 translocation in the absence of insulin [8688]. Finally, expression of dominant-negative mutants or microinjection of blocking antibodies against PKB in L6 muscle cells and 3T3-L1 adipocytes inhibited insulin-induced GLUT4 translocation [83,87,89]. Surprisingly, however, a chemical inhibitor of Akt that fully precludes its activation reduced insulin-dependent GLUT4 translocation by only 50% [90].

AS160, TBC1D1 (tre-2/USP6, BUB2, cdc16 domain family member 1) and target Rabs

Recently, there has been an intensive search for the PKB targets that lead to GLUT4 trafficking. Although this enzyme can phosphorylate a number of proteins [91], AS160 (also known as TBC1D4 [92]), was recently identified as a regulator of GLUT4 trafficking in 3T3-L1 adipocytes [33,9395] and muscle cells [96], as well as of glucose uptake in skeletal muscle [97]. In all of these systems, insulin increases AS160 phosphorylation, and this response was found to be impaired in skeletal muscle obtained from insulin-resistant patients [98,99]. The latter finding correlates with diminished insulin-mediated glucose uptake and GLUT4 translocation to the PM observed in muscle and adipose cells of diabetic animals and humans [99,100]. AS160 harbours a GAP (GTPase-activating protein) domain that is thought to maintain its target Rab(s) in an inactive, GDP-bound form. The current understanding is that upon insulin stimulation, phosphorylation of AS160 shuts off its GAP activity, shifting the equilibrium of its target Rab(s) to an active GTP-bound form, enabling it to mediate GLUT4 trafficking by releasing GLUT4 from intracellular retention mechanisms [33,9395]. This mode of action is proposed based on results using AS160 mutants. An AS160 mutant lacking four of the PKB phosphorylation sites (termed ‘4P’), but not a mutant with a critical arginine/lysine residue substitution in the GAP domain, prevented the insulin-dependent GLUT4 translocation in 3T3-L1 adipocytes [94] and muscle cells [96], and reduced insulin-dependent glucose uptake in skeletal muscle [97]. Most telling, a mutant encoding both the ‘4P’ mutation and the arginine/lysine mutation in the GAP domain was no longer inhibitory, suggesting that the phosphorylation site is functionally linked to the GAP activity of AS160. Hence AS160 can be viewed as a brake that is removed upon phosphorylation by PKB.

TBC1D1 is also a PKB substrate, highly homologous with AS160. However, whereas AS160 harbours six phosphorylation sites for PKB, TBC1D1 displays only two (Thr590 and Ser501 in mouse TBC1D1 [101,102]). Differences were also found when comparing the effects of mutants of each protein on GLUT4 translocation. Thus, whereas overexpression of wild-type AS160 had no effect, wild-type TBC1D1 diminished insulin-dependent GLUT4 translocation in 3T3-L1 adipocytes [94,102]. However, on closer scrutiny, basal state levels of surface GLUT4 were also reduced by approx. 40%, complicating the analysis of this result with overexpressed protein. As in the case of AS160 mutants, overexpression of TBC1D1 mutated in the PKB target sites largely reduced insulin-dependent GLUT4 translocation, and TBC1D1 with a mutation in its GAP domain had no effect on GLUT4 translocation [102].

In addition to their GAP activity, AS160 and TBC1D1 can bind 14-3-3 proteins, which interact with phosphorylated serine or threonine residues on numerous polypeptides, often in response to growth factor stimulation. 14-3-3 binding to phosphorylated AS160 is essential for GLUT4 translocation [103], yet the exact mechanism remains to be elucidated. AS160 and possibly TBC1D1 act as a convergence point for signalling from several protein kinases in addition to PKB/Akt, such as AMPK (AMP-activated protein kinase) and PKC [104].

The in vitro GAP activity of the TBC domains of AS160 and TBC1D1 is selective towards Rabs 2A, 8A, 8B, 10 and 14 [102,105]. In isolated GLUT4-containing intracellular membranes from muscle cells and adipocytes, Rabs 2, 4, 8, 10, 11 and 14 were detected, suggesting that they may contribute in GLUT4 trafficking [33,105,106]. Indeed, in L6 muscle cells, overexpression of Rab8A and Rab14 (but not Rab10) rescued the inhibition of GLUT4 translocation caused by the constitutively active, i.e. inhibitory ‘4P’ mutant of, AS160. This suggested that Rabs 8A and 14 are the targets of AS160 leading to GLUT4 translocation [107]. This conclusion was buttressed by the finding that Rab8A or Rab14 knockdown, but not Rab10 knockdown, reduced insulin-dependent GLUT4 translocation in muscle cells (S. Ishikura and A. Klip, unpublished work). Furthermore, knockdown of Rab8A or Rab14 rescued the rise in surface GLUT4 levels elicited by AS160 or TBC1D1 knockdown, suggesting that these Rabs mediate the release of GLUT4 intracellular retention exerted by the Rab-GAP proteins. On the other hand, Rab10 was found to be the predominant AS160 target in 3T3-L1 adipocytes, where Rab10 knockdown lowered insulin-induced GLUT4 translocation to the PM, an effect restored by its re-expression. Moreover, Rab10 silencing partially overcame the rise in surface basal GLUT4 caused by AS160 knockdown. These effects were specific for Rab10 as they were not observed upon silencing Rab8A or Rab14 [108,109]. The contrasting results reported for L6 muscle cells and 3T3-L1 adipocytes could have arisen from cell-specific reliance on distinct Rabs.

Rabs are typically considered to be molecular switches, linking a signal transduction cascade to molecular effectors. As shown above, some of these effectors appear to be molecular motors. Other members of the Rab family, although not being targets of AS160 or TBC1D1, also participate in GLUT4 trafficking, particularly Rabs 4, 5, 11 and 31. Other reviews cover this topic in detail [110,111]. It will be important to unravel the molecular effectors of such Rabs and the specific steps in GLUT4 traffic that they mediate.


Of the four aPKC isoforms, PKCλ and PKCζ have been implicated in GLUT4 translocation. Diverse approaches suggest that aPKC is both necessary and sufficient to elicit this major action of insulin, although studies to the contrary have also emerged. Overall the studies can be summarized as follows. RNAi (RNA interference)-mediated PKCλ silencing in 3T3-L1 adipocytes or dominant-negative PKCλ overexpression in muscle and fat cells partially impaired insulin-elicited GLUT4 translocation and stimulation of glucose uptake [89,112115]. On the other hand, two studies failed to see any effects on these outcomes upon expression of dominant-negative aPKC mutants [116] or PKCλ/ζ silencing [117] in the same cells. Overexpression of wild-type PKCλ/ζ in rat adipocytes mimicked insulin action on glucose uptake [118] and provoked a 2-fold increase in GLUT4 translocation [119]. Compellingly, a mouse model with a muscle-targeted PKCλ gene deletion displayed whole-body insulin resistance, impaired insulin-stimulated glucose uptake into muscle In vivo and ex-vivo, and reduced GLUT4 translocation [120]. As a correlative observation, human subjects with Type 2 diabetes and/or obesity display diminished levels or poorly active aPKC in skeletal muscle [121,122].

One of the further inconsistencies in the literature is that the studies supporting a requirement for aPKC in GLUT4 translocation have also negated a required input by PKB, which is otherwise widely supported. This controversy remains one of the most fascinating roadblocks in our understanding of insulin action and one that is worth re-addressing, perhaps through the concerted action of several groups using diverse biological systems.

How could aPKC signal to GLUT4 vesicles? The connection appears to involve the actin cytoskeleton (see the section on ‘Cytoskeleton input on GLUT4 translocation’ below), as PKCλ/ζ can impinge on Rac and actin dynamics [119,123]. These findings create the attractive hypothetical scenario that PKCλ/ζ may converge with the Rac-dependent arm of insulin signalling governing actin remodelling, that would act in parallel with PKB leading to GLUT4 redistribution (see Figure 3B).

3′- and 5′-phosphoinositide regulation by PTEN (phosphatase and tensin homologue deleted on chromosome 10), SHIP2 (Src homology 2-containing inositol phosphatase-2) and PIKfyve

As seen in the preceding sections, phosphoinositides are crucial elements in insulin action, and their levels are precisely regulated by kinases and phosphatases. PTEN dephosphorylates the third hydroxyl position of the inositol ring of phosphoinositides, mainly PtdIns(3,4,5)P3. Hence its activity would be expected to be of a major consequence on insulin signalling. Indeed, microinjection of a neutralizing anti-PTEN antibody elevated basal and insulin-stimulated PtdIns(3,4,5)P3 levels, thereby increasing GLUT4 translocation to the PM [124]. Conversely, overexpression of wild-type PTEN inhibited the metabolic actions of insulin, dependent on PI3K, although a confounding result with overexpression of dominant-negative PTEN questioned the extent to which PTEN may modulate insulin action [125]. More recently, however, it was established that selective PTEN deletion in skeletal muscle protects against the development of insulin resistance In vivo and enhances insulin-dependent Akt phosphorylation and glucose uptake [126]. Furthermore, PTEN knockdown via siRNA elevated glucose uptake into adipocytes [127]. Hence the majority of the studies In vivo and in vitro support the view that PTEN is a negative regulator of insulin action. PTEN then could potentially be a key point of modulation, toning insulin signalling down (when activated/up-regulated) or up (when inhibited/down-regulated) in diverse metabolic conditions.

The levels of PtdIns(3,4,5)P3 are regulated not only by dephosphorylation of the 3′-position in the inositol ring, but also by dephosphorylation of the 5′-position. In this regard, SHIP2 dephosphorylates PtdIns(3,4,5)P3 to produce PtdIns(3,4)P2 [128], and is hence an interesting molecule that would shift the levels of these two phosphoinositides. Although PtdIns(3,4)P2 is abundant in cell membranes, there is growing evidence that its levels can be locally regulated and impact on actin dynamics, vesicle traffic and Akt activation [129]. Contrary to the case of PTEN, there is no consensus as to how relevant SHIP2 may be in insulin action. In some studies, PTEN but not SHIP2 knockdown enhanced insulin signalling in 3T3-L1 adipocytes [117,127]. In others, SHIP2-deficient embryonic fibroblasts showed elevated PtdIns(3,4,5)P3 and Akt activation in response to serum but not to IGF stimulation [130]. Interestingly, SHIP2 expression is elevated in quadricep muscle and epididymal fat tissue, but not in the liver of diabetic db/db mice [131]; more studies are required to elucidate the relevance of this phosphatase to insulin action In vivo.

As mentioned earlier, exogenous delivery of PtdIns(4,5)P2 to muscle cells does not suffice to mobilize GLUT4 [76]; however, this does not rule out the direct participation of this phosphoinositide in insulin action, beyond serving as the major source for PtdIns(3,4,5)P3 generation. Supporting such a requirement, exogenous PtdIns(4,5)P2 delivery reversed the insulin resistance in adipocytes induced by endothelin-1 and other experimental insults [132]. A compelling body of evidence shows that PtdIns(4,5)P2 is a key regulator of actin dynamics and membrane trafficking [133], and indeed the restoration of insulin sensitivity by PtdIns(4,5)P2 delivery was linked to actin repolymerization. This cellular function is addressed further below.

Among the kinases regulating 3′- and 5′- phosphoinositide levels, PIKfyve stands out. PIKfyve produces PtdIns5P and PtdIns(3,5)P2 from phosphatidylinositol and PtdIns3P respectively [134]. This enzyme is essential for cargo sorting from late endosomes to lysosomes [135]. Although PIKfyve activity is not altered by insulin, it is reportedly inactivated by autophosphorylation [136] or activated in response to Akt [137]. Moreover, expression of a non-phosphorylatable PIKfyve mutant enhanced insulin stimulated GLUT4 translocation [137].

In principle, PIKfyve could contribute to insulin action by changing the levels of its substrates (phosphatidylinositol acid and PtdIns3P) or of its products [PtdIns5P and PtdIns(3,5)P2]. Indeed, PtdIns5P levels increase transiently upon insulin stimulation of 3T3-L1 adipocytes [138], as do PtdIns(3,5)P2 levels in intracellular membranes [139]. Accordingly, microinjection of PtdIns5P into 3T3-L1 adipocytes mobilized GLUT4 towards the cell periphery, and intracellular sequestration of intracellular PtdIns5P reduced insulin-induced GLUT4 translocation [138]. Whether this is due to the action of PtdIns5P itself or to changes in other phosphoinositides is not known, as the same group observed that PIKfyve knockdown reduced PtdIns(3,5)P2 levels, insulin-induced Akt phosphorylation and glucose uptake [139]. It will be important to dissect out in greater detail the role of the two products of PIKfyve and their respective targets, as well as to reconcile the divergent results on whether the enzyme exerts positive or negative influence on GLUT4 translocation.

Finally, for all experiments analysing phosphoinositide action, it remains to be determined whether exogenous delivery or forced changes in abundance through overexpression of kinases or phosphatases can faithfully reproduce the actions of the endogenous counterparts with regards to levels and location. This caveat must be considered when analysing all results described above, while understanding that experimental deviation from equilibrium is a scientific approach to explore physiological phenomena.



Although both microfilaments and microtubules are responsible for maintaining the integrity of perinuclear GLUT4 compartments, the relevance of microtubules to GLUT4 trafficking has been controversial. Some studies show that microtubule-depolymerizing agents such as nocodazole, taxol or colchicine attenuate the insulin-stimulated PM fusion of GLUT4 [140142]. Conversely, others report GLUT4 trafficking to be independent of microtubule integrity, since low concentrations of nocodazole that disrupted microtubules did not prevent GLUT4 trafficking [143145]. Moreover, some of the agents used to alter microtubule stability affected PKB activation [141]. In light of these variable findings, several studies revisited the role of microtubules. Therein, microtubules were shown to contribute to the regulation of GLUT4 vesicle transport to the cell periphery and docking at the PM, as nocodazole pre-treatment reduced GLUT4 build-up within the TIRF zone [42,101,146]. Given that several studies report long-distance linear movement of GLUT4 vesicles that coincide with the position of tubulin, it seems reasonable to propose that microtubules could serve as tracks for the long-range trafficking of vesicles to and from the PM. Such mobilization would also require the association of GLUT4 with microtubule-based motor proteins. In fact, expression of dominant-negative dynein or kinesin motors (KIF3 and KIF5B) disrupts the internalization or exocytosis of GLUT4 respectively [147149]. An additional possibility is that microtubules co-ordinate their function with the actin cytoskeleton (see next), potentially handing on vesicles to cortical actin filamints.

Actin filament remodelling

Actin molecules are in equilibrium between monomeric G-actin and filamentous F-actin forms. In unstimulated muscle cells, F-actin is present as stress fibres running longitudinally in myotubes or radially in myoblasts, whereas, in suspended myoblasts or in adipocytes, actin filaments are located in the cortical zone beneath the PM. Insulin rapidly promotes the formation of cortical actin projections in cultured and primary muscle and fat cells [150156]. This manifests as cortical membrane ruffling in muscle and fat cells. Actin filaments that reorganize in this fashion are composed of β-actin, to be differentiated from α or sarcomeric actin that in mature skeletal muscle constitutes the contractile apparatus. The intense actin branching and repolymerization that occurs at the cell cortex occurs in part at the expense of depolymerization of actin filaments, and indeed it is expected that a large turnover of actin filaments occurs in response to the hormone. The contribution of the actin cytoskeleton to the subcellular localization of insulin-derived signals has been reviewed earlier [157]. In the present review we focus on the impact of actin dynamics on GLUT4 trafficking and highlight recent developments in this area.

There is considerable evidence for the participation of filamentous actin in insulin-induced GLUT4 exocytosis in muscle and fat cells in culture, as well as in primary muscle tissue and fat cells. Agents that disrupt actin dynamics abrogate insulin-induced GLUT4 translocation. This has been observed upon interfering with the dynamic equilibrium between G- and F-actin by cytochalasin D [140,158160], latrunculin B [152,161,162], swinholide A or jasplakinolide [150,155,160]. However, these observations do not differentiate between the need for actin cables for the purpose of processive vesicular motion, and the participation of insulin-induced actin remodelling. The analysis of signals and molecules involved in such remodelling has aided in answering this question.

Rac and TC10

In myoblasts and myotubes, actin remodelling requires the activity of class I PI3K. Indeed, expression of the dominant inhibitory mutant delta-p85, pre-treatment with inhibitors of its enzymatic activity (wortmannin or LY294002), or high overexpression of the PH domain of GRP (general receptor for phosphoinositides) that binds and mops up PI3K lipid products, prevents actin filament reorganization in muscle [154] or fat [163,164] cells. Actin remodelling downstream of PI3K is not, however, mediated by PKB, suggesting signalling bifurcation at this juncture (see Figure 3B). As is well established, all aspects of actin remodelling are under the control of small G-proteins of the Rho family (reviewed in [165]). The type of actin-dependent ruffles formed in muscle cells are reminiscent of the action of one Rho family of small GTPase, Rac. Accordingly, insulin stimulation led to Rac activation, revealed by enhanced GTP-loading of this protein; and this response was reduced when PI3K activity was inhibited [166]. Moreover, expression of a dominant-negative Rac1 mutant or silencing Rac1 expression via siRNA in muscle cells precluded insulin-induced actin remodelling [153,167]. This occurred in the absence of the disruption of stress fibres, allowing one to test the specific consequence of the remodelling on GLUT4 translocation. Importantly, both expression of the Rac mutant and silencing Rac expression largely prevented GLUT4 translocation in muscle cells [153,167]. Since Rac1 silencing did not prevent PKB activation, the PKB signalling arm alone is insufficient to evoke GLUT4 translocation. Conversely, overexpression of dominant-negative PKB or acute PKB inhibition via the chemical AktI1/2 failed to prevent actin remodelling while reducing GLUT4 translocation ([83] and A. Koshkina, V.K. Randhawa and A. Klip, unpublished work). These results prove that each signalling arm, defined by Rac/actin and PKB, is necessary but insufficient to regulate GLUT4 trafficking.

In addition to Rac1, other Rho-family GTPases have also been implicated in GLUT4 translocation. In some cellular systems, Rho-family GTPases provide co-ordinated regulatory input to actin dynamics, and Cdc42 can regulate Rac. In muscle L6 cells, Cdc42 was rapidly activated (i.e. GTP-loaded) following insulin stimulation (A. Koshkina and A. Klip, unpublished work). The functional significance of this activation is currently under investigation. In 3T3-L1 adipocytes, there is controversial evidence for the participation of either Cdc42 and/or Rho activity in insulin-dependent GLUT4 trafficking [168170]. In fact, in those cells yet another different GTPase, TC10, has been highlighted to govern cortical actin polymerization and contribute to GLUT4 exocytosis [169,171,172]. Conversely, the contribution of Rac to adipocyte actin remodelling is less explored, despite the activation of both TC10 and Rac1 within minutes of insulin treatment [166]. In adipocytes, TC10 becomes activated in response to insulin and, in contrast with Rac, this activation appears to be independent of PI3K or PKB. Instead, TC10 activation is linked to the upstream GEF (guanine nucleotide exchange factor), CGD, itself the consequence of clustering of the CAP–Cbl–CrkII complex within caveolae. However, the reliance on the CAP–Cbl–CrkII–C3G relay for GLUT4 translocation has been questioned. Moreover, the inhibitory effect of the TC10 mutants occurred independently of its GTP-binding domain [169,172]. Efforts to explore this pathway further have also yielded opposing results, as silencing CAP or Cbl expression did not affect insulin-stimulated glucose uptake [173], but silencing selective TC10 isoforms precluded GLUT4 translocation [171]. More work is required to investigate the mode of action of TC10 in this function, and the recent implication of TC10 in PtdIns3P production via Rab5 [173a] and PI3K-C2α [74] may provide the connecting link. In addition, TC10 activation may also contribute to positioning aPKC near the cell surface [123]. In contrast with these extensive, if divergent, studies implicating TC10 in GLUT4 trafficking in adipocytes, there is less support for the participation of the Cbl–CAP–TC10 pathway in muscle cells. Hence, in myoblasts, TC10 activation could occur in the absence of CAP expression, and overexpression of a GDP-locked TC10 dominant-negative mutant was inconsequential on GLUT4 translocation [166]. More studies are required to investigate the relevance of this alternative pathway in skeletal muscle and its impact on whole body glucose homoeostasis.

Connection to aPKC?

As mentioned above, PI3K activates PKB, aPKC and Rac, and PKB activation does not lead to Rac activation or actin remodelling. This poses the question as to whether aPKC is related to Rac activation/actin remodelling, or whether there is a three-way split of insulin signals downstream of PI3K. The former scenario appears likely, as it has been reported that PKCζ activity lies downstream of Rac1 [174]. Moreover, overexpression of PKCζ was adequate to elicit a type of actin reorganization and, as discussed above, it also enhanced GLUT4 translocation [119,161]. There is contrasting information on whether or not aPKCs co-localize with cortical actin structures near the PM [119,154]. How aPKC is involved in GLUT4 trafficking remains to be elucidated, but in neurons, aPKC binds proteins that contribute to cellular polarity and protein compartmentation [175]. Interestingly, the kinase was also linked to serine phosphorylation of VAMP2 [176], a component of specialized insulin-responsive GLUT4 vesicles (discussed in the section on ‘Vesicle docking and fusion’). It is plausible that there are hotspots of actin remodelling and subsequent PM insertion where insulin-responsive GLUT4 vesicles accumulate as a result of aPKC input. Consistent with this possibility, GLUT4 appears to insert preferentially into ruffled areas of the PM supported by the remodelled actin mesh [155].

Rac effectors governing actin dynamics

It is reasonable to rationalize that the requirement for Rac activation in insulin-dependent GLUT4 translocation is a direct consequence of the action of Rac on actin remodelling. However, it is conceivable that Rac has an independent, parallel input on both functions. Hence it is important to map the events unleashed by Rac activation that control actin dynamics and to examine their requirement for GLUT4 traffic. These include actin filament capping and severing by gelsolin, filament severing by dephosphorylated cofilin and filament branching effected by WASP/WAVE (Wiskott–Aldrich syndrome protein/WASP verprolin homologous) acting on Arp2/3. There is no information on whether these pathways participate in insulin-induced actin remodelling. Interestingly, insulin elicits cofilin dephosphorylation in muscle cells, and cofilin knockdown via siRNA partly inhibited GLUT4 externalization [176a]. These results suggest that the dynamic remodelling of actin (cycles of severing and polymerization) helps to furnish GLUT4 to the cell surface. In adipocytes, dominant-negative N-WASP mutants reduced GLUT4 traffic to the PM [169,177].

GLUT4-associated cytoskeletal proteins: ACTN4 (α-actinin 4)

How do GLUT4 vesicles interact with the cytoskeleton? In insulin-stimulated muscle cells, GLUT4 is found within the area of cortical actin remodelling by both immunofluorescence and electron microscopy [157]. In an effort to identify molecular linkers between GLUT4 and the cytoskeleton, we embarked on a differential screen of insulin-dependent changes in protein association to immunoprecipitated GLUT4. Through stable isotope labelling of amino acids and MS analysis, the protein that most markedly increased its association with GLUT4 was ACTN4 [178]. Silencing ACTN4 protein expression by siRNA specifically diminished the increase in surface GLUT4 content by insulin [178a]. ACTN4 knockdown allowed normal PKB activation and cortical actin remodelling but precluded the localization of GLUT4 within the cortical actin mesh. ACTN4 may potentially tether the insulin-responsive GLUT4 vesicle to the actin cytoskeleton for progression towards docking and fusion. Interesting in this context is the protein TUG, which, in contrast with ACTN4, binds to GLUT4 in the absence of insulin and hormonal stimulation, causing its dissociation from the transporter in both adipose [179] and muscle (J. D. Schertzer, X. Huang and A. Klip, unpublished work) cells. It will be interesting to investigate whether TUG release might allow GLUT4 binding to ACTN4. Alternatively, TUG may be a GLUT4 tether that releases vesicles from the SC for progression towards the cell surface.

Tracks or tethering scaffold?

It is plausible that the requirement for actin dynamics inherent to the insulin-induced GLUT4 translocation is due to participation of actin filament tracks along which GLUT4 vesicles would move near the cell surface. Disrupting these tracks could potentially prevent GLUT4 exit from the SC. Consistent with this scenario would be a need for actin-based motor proteins to allow the PM-directed transport of GLUT4 vesicles. As per this prediction, several actin-based motor proteins have been implicated in GLUT4 trafficking. Partial knockdown of myosin 1c expression reduced the gain in surface GLUT4 in 3T3-L1 adipocytes [180]. Also, myosin Va is phosphorylated by Akt2, enhancing its association with actin, and myosin Va knockdown reduced GLUT4 translocation [181]. This led Yoshizaki et al. [181] to suggest that myosin Va is the motor that propels vesicles along actin tracks to the cell surface. Finally, in L6 muscle cells, a fragment of the processive myosin Vb that interferes with its interaction with Rab8 precluded insulin-dependent GLUT4 translocation (S. Ishikura and A. Klip, unpublished work). All of these results suggest that actin-based motors are required for the final gain in surface GLUT4 elicited by the hormone, but they do not directly prove that there is processive movement of the vesicles on actin tracks mediated by these motors.

On the other hand, it is also possible that insulin-induced actin remodelling into a highly branched cortical filamentous mesh serves to position vesicles and signals, effectively priming them for docking and fusion. Such actin platforms would presumably allow the insulin signalling molecules to come into close proximity with the intracellular GLUT4 vesicles. Interestingly, along with the remodelled cortical actin mesh co-localize several insulin signalling molecules (e.g. IRS-1/2, PI3K subunits, phospho-Akt, and possibly PKCζ) together with PtdIns(3,4,5)P3, VAMP2 and GLUT4 [153,154]. How GLUT4 vesicles are tethered to the actin cytoskeleton remains to be elucidated. In this regard, ACTN4 may be the missing link, as this protein is found only within remodelled actin but not along filaments in unstimulated cells, and insulin promotes cortical co-localization of actin filaments, ACTN4 and GLUT4, as well as the physical association of GLUT4 with ACTN4 [178]. The functional significance of Rabs should be revisited in this context. Rab proteins bind molecular tethers, and can regulate the actin cytoskeleton in certain cell systems [111,156,182,183]. Alternatively, the actin cytoskeleton may serve as a barrier to PM-directed vesicle release. This latter phenomenon has been well described for dense core granule release or synaptic vesicle trafficking [184]. In this final scenario, one would envision the dissolution of the actin mesh specifically about the PM by insulin to subsequently allow for vesicle fusion.

The cytoskeleton in insulin-resistant states

The importance of identifying the molecular mechanisms by which cytoskeletal elements contribute to GLUT4 translocation is underscored by recent studies that highlight defects in actin dynamics in conditions of insulin resistance. In muscle cells incubated for prolonged times with high glucose or insulin, or for short times in the presence of ceramide or oxidative-radical-generating enzymes, Rac activation and actin remodelling by insulin were compromised [167]. All of those conditions emulate aspects of the cellular environment in Type 2 diabetes, which is accompanied by insulin resistance. Moreover, the loss of cortical actin filaments observed in insulin-resistant conditions was relieved by regeneration of PtdIns(3,4)P2, concomitantly relieving the insulin resistance of GLUT4 trafficking [185]. With the rapid advances in imaging, it should become possible to tease out the specific steps at which actin and microtubules regulate GLUT4 vesicle trafficking to better understand how their dysregulation may contribute to insulin resistance.


GLUT4 trafficking and in particular its regulation by insulin have become the paramount example of regulated vesicle trafficking. The segregation of a large portion of GLUT4 away from the continuously recycling pathway is a feature that evolutionarily was developed both in the structural characteristics of this transporter isoform, as well as of proteins that define a ‘specialized storage compartment’, SC. The molecular markers of such definers are still to be unravelled, but they appear to be exclusively characteristic of muscle and fat cells. Impressive progress has been made in mapping the signals that, in series and in parallel, collude to relay information to GLUT4 vesicles and target membranes. Yet, further work is required to elucidate the final steps in the insulin-signalling cascade downstream of molecular switches (Rabs) and molecular motors (myosins) that convey molecular information to the SC. In this regard, progress has been made in mapping biochemical (e.g. VAMP2 phosphorylation), structural (dispersion and generation of small vesicles) and organizational (association or dissociation of TUG, ACTN4) changes in the vesicles themselves. Additional changes at the PM are essential for actual docking/fusion of the vesicles (SNARE complex availability through Munc18c and probably other regulators). Finally, the entropic input of the cytoskeleton, potentially to position vesicles and signals, provide tethers or allow final access of vesicles to the PM, has emerged as an essential constituent for effective GLUT4 translocation. As we approach the 30th anniversary of the first glimpse of this phenomenon, we look back and marvel at the intricate series of intracellular events governing it. The next years should add the fine spatial and temporal resolution of this fundamental biological process.


We apologize to all of the colleagues whose work could not be covered in the present review. We thank Dr Philip J. Bilan and Dr Shuhei Ishikura for insightful comments on this manuscript. The work from our laboratory that is described in the present review was supported by grants MOP-7307 and MOP-1202 to A. K. from the Canadian Institutes of Health Research and from the Canadian Diabetes Association. H. Z. was supported by a postdoctoral fellowship from the Hospital for Sick Children (Toronto, ON, Canada). V. K. R. was supported by a doctoral studentship from the Canadian Institutes of Health Research, and C. N. A. was supported by a doctoral studentship from the Canadian Diabetes Association.

Abbreviations: ACTN4, α-actinin 4; aPKC, atypical protein kinase C; AS160, Akt substrate of 160 kDa; GAP, GTPase-activating protein; GFP, green fluorescent protein; GLUT4, glucose transporter 4; GSV, GLUT4-storage vesicle; IGF, insulin-like growth factor; IRAP, insulin-regulated aminopeptidase; IRS, insulin receptor substrate; NSF, N-ethylmaleimide-sensitive factor; PH, pleckstrin homology; PI3K, phosphoinositide 3-kinase; PIKfyve, FYVE-domain containing PtdIns3P 5-kinase; PKB, protein kinase B; PKC, protein kinase C; PM, plasma membrane; PTEN, phosphatase and tensin homologue deleted on chromosome 10; RE, recycling endosome; SC, specialized or storage compartment; SHIP2, Src homology 2-containing inositol phosphatase-2; siRNA, small interfering RNA; SM, Sec1p/Munc18; SNAP, soluble NSF-attachment protein; TBC1D1, tre-2/USP6, BUB2, cdc16 domain family member 1; TfR, transferrin receptor; TGN, trans-Golgi network; TIRF, total internal reflection fluorescence; t-SNARE, target SNARE; TUG, tether containing UBX domain for GLUT4; VAMP2, vesicle-associated membrane protein 2; v-SNARE, vesicular soluble N-ethylmaleimide-sensitive factor-attachment protein receptor; WASP, Wiskott–Aldrich syndrome protein


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 12a.
  14. 13.
  15. 14.
  16. 15.
  17. 16.
  18. 17.
  19. 18.
  20. 19.
  21. 20.
  22. 21.
  23. 22.
  24. 23.
  25. 24.
  26. 25.
  27. 26.
  28. 27.
  29. 28.
  30. 29.
  31. 30.
  32. 31.
  33. 32.
  34. 33.
  35. 34.
  36. 34a.
  37. 35.
  38. 36.
  39. 37.
  40. 38.
  41. 39.
  42. 40.
  43. 41.
  44. 42.
  45. 43.
  46. 44.
  47. 45.
  48. 46.
  49. 47.
  50. 48.
  51. 48a.
  52. 49.
  53. 50.
  54. 51.
  55. 52.
  56. 53.
  57. 54.
  58. 55.
  59. 56.
  60. 57.
  61. 58.
  62. 59.
  63. 60.
  64. 61.
  65. 62.
  66. 63.
  67. 64.
  68. 65.
  69. 66.
  70. 67.
  71. 68.
  72. 69.
  73. 70.
  74. 71.
  75. 72.
  76. 73.
  77. 74.
  78. 75.
  79. 76.
  80. 77.
  81. 78.
  82. 79.
  83. 80.
  84. 81.
  85. 82.
  86. 83.
  87. 84.
  88. 85.
  89. 86.
  90. 87.
  91. 88.
  92. 89.
  93. 90.
  94. 91.
  95. 92.
  96. 93.
  97. 94.
  98. 95.
  99. 96.
  100. 97.
  101. 98.
  102. 99.
  103. 100.
  104. 101.
  105. 102.
  106. 103.
  107. 104.
  108. 105.
  109. 106.
  110. 107.
  111. 108.
  112. 109.
  113. 110.
  114. 111.
  115. 112.
  116. 113.
  117. 114.
  118. 115.
  119. 116.
  120. 117.
  121. 118.
  122. 119.
  123. 120.
  124. 121.
  125. 122.
  126. 123.
  127. 124.
  128. 125.
  129. 126.
  130. 127.
  131. 128.
  132. 129.
  133. 130.
  134. 131.
  135. 132.
  136. 133.
  137. 134.
  138. 135.
  139. 136.
  140. 137.
  141. 138.
  142. 139.
  143. 140.
  144. 141.
  145. 142.
  146. 143.
  147. 144.
  148. 145.
  149. 146.
  150. 147.
  151. 148.
  152. 149.
  153. 150.
  154. 151.
  155. 152.
  156. 153.
  157. 154.
  158. 155.
  159. 156.
  160. 157.
  161. 158.
  162. 159.
  163. 160.
  164. 161.
  165. 162.
  166. 163.
  167. 164.
  168. 165.
  169. 166.
  170. 167.
  171. 168.
  172. 169.
  173. 170.
  174. 171.
  175. 172.
  176. 173.
  177. 173a.
  178. 174.
  179. 175.
  180. 176.
  181. 176a.
  182. 177.
  183. 178.
  184. 178a.
  185. 179.
  186. 180.
  187. 181.
  188. 182.
  189. 183.
  190. 184.
  191. 185.
View Abstract