The Sugar Is sIRVed: Sorting Glut4 and Its Fellow Travelers


  • Konstantin V. Kandror,

    Corresponding author
    1. Department of Biochemistry, Boston University School of Medicine, 72 E. Concord Street, Boston, MA 02118, USA
      Konstantin V. Kandror, and Paul F. Pilch,
    Search for more papers by this author
  • Paul F. Pilch

    Corresponding author
    1. Department of Biochemistry, Boston University School of Medicine, 72 E. Concord Street, Boston, MA 02118, USA
    2. Department of Medicine, Boston University School of Medicine, 72 E. Concord Street, Boston, MA 02118, USA
      Konstantin V. Kandror, and Paul F. Pilch,
    Search for more papers by this author

Konstantin V. Kandror, and Paul F. Pilch,


Translocation of Glut4 to the plasma membrane of fat and skeletal muscle cells is mediated by specialized insulin-responsive vesicles (IRVs), whose protein composition consists primarily of glucose transporter isoform 4 (Glut4), insulin-responsive amino peptidase (IRAP), sortilin, lipoprotein receptor-related protein 1 (LRP1) and v-SNAREs. How can these proteins find each other in the cell and form functional vesicles after endocytosis from the plasma membrane? We are proposing a model according to which the IRV component proteins are internalized into sorting endosomes and are delivered to the IRV donor compartment(s), recycling endosomes and/or the trans-Golgi network (TGN), by cellugyrin-positive transport vesicles. The cytoplasmic tails of Glut4, IRAP, LRP1 and sortilin play an important targeting role in this process. Once these proteins arrive in the donor compartment, they interact with each other via their lumenal domains. This facilitates clustering of the IRV proteins into an oligomeric complex, which can then be distributed from the donor membranes to the IRV as a single entity with the help of adaptors, such as Golgi-localized, gamma-adaptin ear-containing, ARF-binding (GGA).

inline image

The regulation of blood glucose levels in mammals is achieved by insulin-dependent translocation of the glucose transporter isoform 4 (Glut4) to the plasma membrane of fat and skeletal muscle cells. Multiple lines of independent evidence, such as transgenic and knock-out mouse models (reviewed in 1) and human nuclear magnetic resonance studies in vivo (reviewed in 2), demonstrate that Glut4-mediated glucose uptake represents the rate-limiting step of insulin-stimulated glucose disposal. Consequently, the inability of normal doses of insulin to stimulate translocation of Glut4, i.e. insulin resistance, may represent the primary defect in the development of type 2 diabetes mellitus (3). Therefore, deciphering the mechanistic steps of Glut4 regulation will be crucial for our understanding of the molecular nature of insulin resistance and diabetes mellitus and for possible treatment modalities. In addition, Glut4 trafficking represents an important paradigm for cell biologists in numerous ways.

Insulin resistance in fat and skeletal muscle tissues may be caused by defective insulin signalling, aberrant Glut4 recycling or both. Recently, most research efforts directed towards understanding insulin resistance have focused on potential defects in the upstream insulin signalling pathway, which consists of the insulin receptor, its substrates (IRSs), phosphatidylinositol-3 kinase (PI-3-kinase), PI-3-kinase-dependent kinase (PDK) and the serine/threonine kinase, Akt (4,5). In particular, counter regulatory phosphorylation of IRS1 on ser/thr residues resulting in blunted insulin signaling has been observed in numerous experimental contexts with the overall effect of leading more to confusion than illumination of insulin resistance mechanisms (6). Moreover, these upstream IRS phosphorylation events do not necessarily correlate with decreased glucose uptake, which can be demonstrated to occur in vitro and in transgenic animals in vivo in an insulin-independent manner (7). These findings point to the cell biology of Glut4 recycling as a potential site of primary diabetes-related abnormalities and underscore the need for describing the many aspects of Glut4 traffic that remain unknown.

Although considerable strides have been made in defining the upstream insulin signalling pathway, those biochemical events proximal to IRV mobilization and fusion with the plasma membrane remain unclear. Akt activation is critical for overall Glut4 trafficking and thus targets/substrates of this protein, such as, the Rab GAP (GTPase-activating protein), AS160/TBC1D4, have emerged as important players in IRV mobility (reviewed in 8). However, it is not clear at what stage in Glut4 trafficking that AS160/TBC1D4 participates as its knockdown does not completely mimic insulin-dependent Glut4 translocation. It is therefore feasible that AS160/TBC1D4 not only influences the translocation of the preformed IRVs, but it may also be involved in their biogenesis. The biogenesis process (9) and differentiation-dependent formation of Glut4 vesicles (10) have recently been reviewed. Hereafter, we will consider the interesting problem of how the cell sorts a number of major IRV cargo proteins with very different biochemical functions and properties.

Small Preformed Vesicles Represent the Primary form of Glut4 Storage in the Cell

On its way to and from the cell surface, Glut4 traffics through several distinct intracellular compartments (Figure 1). These include early endosomes, intermediate transport vesicles, recycling endosomes and/or the trans-Golgi network (TGN), and insulin-responsive vesicles (IRVs) (also called GSVs for glucose transporter storage vesicles), which many researchers believe represent the target of insulin regulation that results in delivery of Glut4 to the plasma membrane.

Figure 1.

Sorting IRV protein contents. Starting from the plasma membrane, the major IRV cargo proteins undergo endocytosis based on their various internalization motifs (Table 1). They are transported from early endosomes to recycling endosomes and/or the TGN via cellugyrin vesicles (depicted in blue), also based on these sorting motifs. In the latter compartments, the IRV cargo is concentrated based on luminal domain interactions and buds to IRVs using the GGA (depicted in black).

The physical nature of the major Glut4-containing compartment(s) in fully differentiated adipose and skeletal muscle cells was shown by immunoelectron microscopy to consist primarily of small (70–100 nm in diameter) vesicles and short tubules (11); biochemical fractionation shows that 60–75% of the total pool of Glut4 is present in small (ca. 80–100S) vesicles that are readily detectable by sucrose gradient centrifugation and other biochemical methods (12). The rest of the transporter is present in large rapidly sedimenting intracellular membranes that are likely to represent recycling endosomes and probably TGN structures (11).

In cultured adipose cells, the majority of Glut4-containing membranes, including small vesicles, recycling endosomes and TGN, are concentrated in the perinuclear region, which makes it difficult to dissect the nature of the ‘Glut4 pathway’ by conventional immunofluorescence. In particular, incomplete and even controversial evidence exists as to whether Glut4 traffics through recycling endosomes (13), TGN (14) or both compartments. Recent publications demonstrate that ‘classical’ Golgi proteins such as p115 (15) and Golgin-160 (16) are involved in Glut4 sorting and traffic. This suggests that the Golgi apparatus (including TGN) may be more intimately involved in the ‘Glut4 pathway’ than it has previously been appreciated. This idea is consistent with recent results showing that retraction of Glut4 from endosomes to TGN may be an essential step in the ‘Glut4 pathway’ of human myocytes (17,18).

The size of the vesicular pool of Glut4 is decreased by insulin ca. twofold concomitant with an equivalent increase in plasma membrane Glut4 content suggesting that Glut4-containing vesicles fuse with the plasma membrane. Insulin regulation of Glut4 content in endosomes and TGN has not been systematically studied and is largely unknown. Available morphological evidence suggests that insulin does not cause any major changes in the presence of Glut4 in large ‘vacuoles’ and ‘tubules’(19). This, however, may be explained by the rapid replenishing of the endosomal pool of Glut4 with newly internalized transporter (20). In any case, small membrane vesicles and short tubules with the sedimentation coefficient of ca. 80–100S seem to represent the major form of Glut4 storage in basal adipocytes and myocytes.

The process of biogenesis of small Glut4 vesicles from the large donor membranes can be reconstituted in vitro upon incubation of the donor membranes with ATP and cytosol at 37°C (21,22). Does insulin stimulate the formation of small Glut4 vesicles along with triggering translocation of the pre-existing vesicles? Some evidence shows that insulin may indeed have a positive effect on formation of Glut4 vesicles (20,21,23,24). In differentiated adipocytes, however, this effect is not likely to account for the massive increase in the plasma membrane Glut4 content in response to insulin stimulation (21) suggesting that the major effect of insulin is translocation of the preformed vesicles.

Heterogeneity of Small Glut4 Vesicles

Small Glut4 vesicles are not structurally or functionally homogeneous, rather total preparations of these vesicles represent an overlapping mixture of at least two vesicular populations that can be partially resolved by gradient centrifugation. One population of smaller vesicles is marked by the presence of the protein called cellugyrin, while another population of larger vesicles does not contain this protein (25–27). Cellugyrin has four transmembrane helices with an as yet unknown physiological function and it is a ubiquitously expressed homologue of synaptogyrin, a major constituent of synaptic vesicles (28). Cellugyrin is present only in the population of Glut4 vesicles that is not recruited to the plasma membrane by insulin and thus permanently maintains its intracellular localization. On the contrary, cellugyrin-negative Glut4 vesicles translocate to the plasma membrane with virtually 100% efficiency (25–27). As cellugyrin-positive and cellugyrin-negative vesicles compartmentalize approximately equal amounts of Glut4 (25,27), depletion of the total vesicular pool of the transporter by insulin is no more than 50% in the absence of other perturbations.

It has been suggested that cellugyrin-positive vesicles represent a ubiquitous vesicular compartment (29) that transports proteins between early and recycling endosomes/TGN. Such transport vesicles are quite abundant in various cell types (22,30) and can accommodate significant amounts of recycling proteins. On the other hand, cellugyrin-negative vesicles are likely to represent a specialized insulin-responsive compartment, the IRVs, that has been postulated to be the target of insulin signalling (31). Thus, we will use the term ‘IRVs’ when we talk specifically about the vesicular Glut4-containing compartment that is translocated to the plasma membrane in response to insulin stimulation. The total vesicular pool of Glut4 that includes both insulin-sensitive and not sensitive vesicles of various cell biological nature will be referred to as ‘Glut4-vesicles'.

Major Cargo Proteins of Glut4 Vesicles and the IRVs

A popular approach to understanding and uncovering the ‘secrets’ of insulin responsiveness relies on the hypothesis that these may be revealed by a comprehensive analysis of IRV protein (or lipid) composition. Immunoadsorption procedures and other methods for enriching Glut4-containing vesicles allowed for the detection of their major protein components (see below and the reviews 9,10,32). However, given the vesicle heterogeneity noted above, highly purified preparations of Glut4 vesicles will likely contain multiple contaminating proteins and immunoadsorption with antibodies against Glut4 alone, for example, cannot serve as a proof of protein localization in Glut4 vesicles. Such a conclusion should be supported by biochemical co-fractionation, microscopic co-localization, ideally immunoelectron microscopy and functional analysis. It is also important to keep in mind that peripheral membrane proteins that are associated with Glut4 vesicles tend to dissociate from this compartment in the process of stringent washing steps required in immunoadsorption protocols. This may explain our slow progress in understanding the distal steps of the insulin signalling pathway that could presumably involve peripheral membrane components of the IRVs and specific strategies for identifying such proteins remain elusive.

To identify integral or tightly bound peripheral membrane proteins, a systematic proteomic analysis of highly purified IRVs was recently performed by Jedrychowski et al. (27) following the subtraction (and proteomic analysis) of cellugyrin-containing Glut4 compartments. The results of this study confirmed that major cargo proteins, namely Glut4, insulin-responsive amino peptidase (IRAP) and sortilin, that had previously been found in the total preparation of small Glut4 vesicles (reviewed in 9,10,32,33) were abundantly present in the IRVs as well and were fully insulin responsive. In addition to these three major cargo proteins, Jedrychowski et al. found that the low density lipoprotein receptor-related protein 1 (LRP1) is also a component of the IRVs (27). LRP1 has been long known to undergo insulin-dependent translocation to the plasma membrane in adipose cells (34) but escaped detection in previous proteomic analysis of Glut4 vesicles, most likely because of its enormous size (4544 amino acid residues). Thus, high abundance of Glut4, IRAP, sortilin and LRP1 in the vesicular membrane creates the ‘biochemical individuality’of the IRVs that differs them from other vesicles in the cell. Cellugyrin-positive transport vesicles contain much of the same proteins as the IRVs (27), but their specific content is lower. Unlike the IRVs, each cellugyrin-positive vesicle may carry not more than one molecule of either Glut4 or IRAP (26). This indicates that Glut4, IRAP and, likely, other intrinsic IRV proteins are actively sorted into this compartment. Potential mechanisms of such sorting will be discussed below.

In addition to LRP1, IRVs accommodate a small (ca. 10% of their total microsomal pools) portion of the receptors for transferrin (TfR) and mannose-6-phosphate (M6PR) as well as a variety of other known membrane recycling proteins (27,35) that represent minor vesicular components whose behaviour has not been extensively characterized. The presence of the TfR and M6PR in IRVs can explain their long-known insulin-dependent translocation to the plasma membrane in adipocytes (36) and their presence in IRVs is explained in the sorting section. Finally, the IRVs may contain more than one of the functionally redundant v-SNAREs (37) that have been known for some time to participate in Glut4 trafficking (38).

IRAP is likely to be the most abundant IRV component and is largely if not completely co-localized with Glut4 (26). The physiological role of an insulin-dependent aminopeptidase activity in fat and muscle remains unclear but IRAP's presence in IRVs could contribute to their insulin sensitivity via binding of the Rab GAP, TBC1D4 (35,39), and/or it could function in vesicle protein sorting. These possible roles will be further discussed below.

Sortilin is a homologue of the yeast vacuolar sorting receptor, Vps10p, and represents a multi-ligand protein receptor of very broad specificity in mammalian cells and has various functions in protein sorting and signalling of particular note, aside from muscle and fat, in neuronal cells (40). In the former cells, sortilin is enriched in the IRVs and plays an important role in the biogenesis of this compartment likely by protein–protein interactions in its luminal domain [see below and Refs (41–43)]. In this way, it may be thought of as a receptor for its IRV companion proteins, Glut4 and IRAP.

An interesting although somewhat teleological question is why does the cell use this functionally heterogenous mixture of proteins in order to build the insulin-responsive glucose transporting machinery? One speculation is the following. The lumenal domain of sortilin is highly homologous to that of the yeast sorting receptor, Vps10p, which is responsible for the intracellular traffic of aminopeptidase Y, carboxypeptidases Y and proteinase A in Saccharomyces cerevisiae(44,45). Thus, it may not be surprising that IRAP, an aminopeptidase, co-localizes with sortilin in the same vesicles in mammalian cells as well. Furthermore, it is possible that, in the course of evolution, Glut4 developed an affinity to the pre-existing sorting receptor, sortilin, and thus hijacked an ancient type of a vesicular carrier used for the intracellular trafficking of Vps10p/peptidase complexes in yeast. Other hypotheses, however, are quite possible as well.

Recycling of the IRV Proteins in the Cell

The IRVs are highly dynamic structures that go through cycles of assembly–disassembly numerous times throughout the lifetime of any fat or skeletal muscle cell raising the question as to what is the molecular mechanism(s) of IRV formation from the individual protein components, i.e. how does sorting occur? Much work in this regard has focused on the cytoplasmic regions of the major cargo proteins that contain several intracellular sorting motifs (46). Table 1 lists those motifs for the major Glut4 cargo proteins and these are further discussed below. Most of the studies of Glut4 and its companions [partially summarized in Ref. (47); see also Refs (14,48–50)] were performed with the help of a ‘loss of function’ approach and did not address Glut4 targeting specifically to the IRVs. Instead, the functional criterion of insulin-dependent translocation to the cell surface was used as the major or only proof of Glut4 localization in the insulin-responsive compartment. Considering that on the way to and from the plasma membrane, Glut4 passes through several intracellular compartments, interfering with the targeting sequences within the cytoplasmic termini of the transporter may inhibit not only entry into the IRVs, but upstream trafficking steps as well. Indeed, in the ‘gain of function’ experiments, it was recently demonstrated that the C-terminus of Glut4 targets the chimera reporter protein to the perinuclear region of the cell, that is, the IRV precursor compartment, either recycling endosomes and/or TGN, but is not sufficient for its entry into the IRVs per se(51).

Table 1.  Sorting motifs for IRV cargo proteins Motifs
  1. The above amino acid sequences, possibly excepting the last column, all bind clathrin adaptor proteins (APs), namely AP-1-3, GGA and ACAP1 (46). These sequences have been implicated in endocytosis from the cell surface as well as other intracellular vesicle budding steps, but the DXXLL GGA binding sequence, unique to sortilin, is thought to be exclusive to TGN to endosome sorting. Thus, we propose as outlined in the text and in Figure 1 that formation of the IRVs is mediated by GGA and sortilin, the latter in a complex with the remaining IRV cargo.

Sortilin Tm-820-DEDLLE-825Palmitoylation

By the same token, the cytoplasmic tail of IRAP targets this protein primarily to the IRV donor membranes but not to the IRVs (52). Still, the cytoplasmic tail of IRAP (53) as well as the C-terminus of Glut4 (51) can confer some degree of insulin responsiveness to the reporter molecules, which, however, is less than the insulin responsiveness of wild-type Glut4 and IRAP (51,52,54). One explanation of this phenomenon is that once a protein enters the IRV donor compartment(s), it can be captured by budding vesicles by, for example, mass action (10). This would be the case for proteins such as the TfR, M6PR, syntaxins 6/16 and Vti1a. These proteins demonstrate significant co-localization with Glut4 at the level of light microscopy but are not actively sorted into the IRVs and acquire less insulin responsiveness than the major IRV cargo proteins. Most membrane proteins, however, are not retained in the IRV donor compartment due to the lack of the targeting information in their cytoplasmic tails and, naturally, have a little chance for entry into IRVs.

Studies of sortilin targeting have also shown that the reporter protein consisting of the cytoplasmic tail of sortilin and EGFP is not targeted to the IRVs and lacks insulin responsiveness (55). On the contrary, the tagged luminal Vps10p domain of sortilin demonstrates partial co-localization with Glut4 and is, to some extent, targeted to the IRVs in 3T3-L1 adipocytes. Interestingly, intracellular trafficking of sortilin requires palmitoylation (56), although it is not yet clear whether and to what extent palmitoylation is required for IRV biogenesis.

Thus, it appears that the information contained in the cytoplasmic tails of the major IRV components, Glut4, IRAP and sortilin (also LRP1, Table 1 and below), is essential for their targeting into the IRV donor compartment(s) but not sufficient for their entry into IRVs. This implies that the lumenal domains should also be important for protein localization in the latter compartment. Interestingly, Glut4, IRAP, sortilin and a recently discovered IRV component, LRP1, can interact with each other via their luminal domains (27,41,52,54,55). Furthermore, cross-linking studies have shown that sortilin can homo-oligomerize (Huang and Kandror, unpublished data). It is possible that mutual luminal interactions between Glut4, IRAP, sortilin and LRP1 help to bring these proteins together in the donor membranes and may thus explain active protein sorting into IRV. Thus, the two major factors that define the efficiency of protein entry into the IRVs are their enrichment in the IRV donor compartment (recycling endosomes and/or TGN), which depends on the targeting information in the cytoplasmic tails and their ability to physically interact with the major IRV component proteins, Glut4, IRAP, sortilin and likely LRP1, which appear to form a large oligomeric complex. This complex may then be distributed from the donor membranes to the IRV as a single entity with the help of Golgi-localized, gamma-adaptin ear-containing, ARF-binding (GGA) proteins that recognize the DXXLL sequence in sortilin and ACAP1 (ADP-ribosylation factor (ARF) GAP with coiled-coil, ANK repeat and pleckstrin homology domains) that binds to the central loop of Glut4 (9). Most of these features are shown in Figure 1, the depiction of the Glut4 trafficking pathway. In addition, ubiquitination of Glut4 has also very recently been documented as an essential step in Glut4 targeting, so that GGA may mediate sorting of ubiquitinated Glut4 directly (57). Arf6 and PIP4 are likely to play essential roles in the recruitment of these adaptors to the donor membranes (9).

Thus the hypothesis is that ‘self-assembly’ of the IRVs proceeds via luminal interactions between the major cargo proteins as described above and they then bud from the precursor compartment(s) with the aid of the adaptors and other protein contributors to this process. A prediction of this model is, therefore, that knockdown or loss of these major cargo proteins could result in decreased amounts of the other protein constituents. In agreement with this model, it has been shown that knockout of Glut4 decreases intracellular IRAP levels and vice versa (58–62) and that this effect takes place at a post-transcriptional level (58). In addition, partial knockdown of sortilin in 3T3-L1 adipocytes decreases Glut4 levels (41). Finally, knockout of LRP1 decreases levels of Glut4 and sortilin (27). Thus, depletion of either IRV component may disrupt the network of lumenal interactions between these proteins in the donor membranes and decrease the efficiency of protein sorting into IRVs. Correspondingly, the remaining IRV proteins that cannot be faithfully compartmentalized in the vesicles are either degraded (41,58–62) or mistargeted (59,63). However, knockdown of either IRAP, Glut4, sortilin or LRP1 may not completely block IRV biogenesis as these vesicles can probably be formed (albeit with a lesser efficiency) by remaining proteins (64).

Relationship of IRVs to Similar, Stimulus-Responsive Compartments in Other Cell Types

IRVs are often compared to other exocytosis-competent vesicular structures such as small synaptic vesicles (SSVs) in neurons, aquaporin-2-containing vesicles in kidney duct cells, H+/K+ pump-containing vesicles in gastric parietal cells, etc. that exist in various highly specialized cells (10,65). Thus, the question from a general cell biological perspective is whether or not these vesicular compartments represent different ‘organelles’ or if their existence is explained by cell-specific differences in the general membrane traffic in terminally differentiated cells (65). Interestingly, the aquaporin-2 vesicles of the kidney have been subjected to proteomic analysis, albeit, not to the extent of differentiating them into precursor compartments and IRV equivalents. They contain many of the same cargo proteins as IRVs, namely sortilin, IRAP and M6PR (66). On the other hand, recent studies have demonstrated that neuronal cells, in addition to SSVs, have another type of a translocation-competent vesicular compartment which is structurally and functionally similar to the IRVs in peripheral insulin target tissues (67). In other words, neurons are capable of making at least two different types of translocatable membrane vesicles with different protein composition and regulation. This, in turn, suggests that the IRVs and SSVs are not cell-specific variations of the same compartment but more likely represent individual recycling ‘organelles’.

SSVs currently represent the best-studied type of small membrane vesicles and it can be instructive to compare these to IRVs in some detail. SSVs are smaller in size than IRVs (42–45 nm) but accommodate many more copies of component proteins/vesicle. An average SSV compartmentalizes ca. 70 molecules of VAMP2, 32 molecules of synaptophysin and 15 molecules of synaptotagmin among at least 40 other different integral membrane proteins (68). Correspondingly, SSVs have a higher buoyant density (i.e. protein/lipid ratio) than IRVs (67). Noteworthy, major component proteins of SSVs are known to physically interact with each other that may have a direct relationship to the biogenesis of this compartment in neurons (69). In this regard, the high density of proteins in SSVs in comparison to the IRVs may be explained by high specific expression levels of the SSV proteins in neurons and/or the strength of their mutual interactions. A strong affinity of the SSV proteins for each other may underlie the kiss-and-run mechanism of their functioning when vesicular proteins do not dissociate from each other upon fusion with the plasma membrane but are internalized all together en bloc(70). Just like the case of the IRVs, knockout of any particular SSV protein does not affect their biogenesis and functioning in a major way (70–73). Thus, a network of multiple mutual interactions between vesicular proteins may allow the cell to produce a resilient structure that can survive the loss of any component without major effect on the functional properties of the vesicular compartment.

There is little doubt that multiple types of vesicular carriers coexist in fat and skeletal muscle cells. However, high insulin sensitivity is the unique property of the IRVs that differentiates them from other intracellular vesicles. As is mentioned above, the explanation of the insulin responsiveness of the IRVs should eventually come down to the specifics of their protein content. Even in the likely case that insulin responsiveness is provided to the IRVs by a peripheral membrane protein(s), this putative component(s) still has to somehow recognize one or more of the core constituents of the IRVs, such as Glut4, IRAP, sortilin or LRP1. Future experiments should identify the IRV proteins that are responsible for their insulin responsiveness and link these vesicles to the insulin signalling pathway.


This work was supported by National Institutes of Health, DK30425 to PFP and DK52057 to KVK.