Signaling Mechanisms That Regulate Glucose Transport*

Insulin, the major hormonal regulator of glucose transport in humans, has served as a prototypic molecule for understanding cell signaling pathways since its discovery in 1922. It was among the first proteins for which primary amino acid sequence and three-dimensional structure were determined. The insulin receptor was likewise among the first peptide receptors to be identified by ligand binding, and its subunit structure was deduced in the earliest days of modern receptor biology (1). The structure of its tyrosine kinase domain was the first of the tyrosine kinases to be solved by x-ray crystallography (2). Further, much is now known about downstream insulin receptor signaling components (3). Despite these advances, our understanding of how insulin and the related proteins, insulin-like growth factors I and II, stimulate glucose trans- port (via the translocation of glucose transporter proteins from intracellular to plasma membranes) is fragmentary. One reason for this is that the major target for insulin signaling to glucose transport is a complex membrane trafficking pathway that is likely to contain many unknown components. intracellular

Insulin, the major hormonal regulator of glucose transport in humans, has served as a prototypic molecule for understanding cell signaling pathways since its discovery in 1922. It was among the first proteins for which primary amino acid sequence and threedimensional structure were determined. The insulin receptor was likewise among the first peptide receptors to be identified by ligand binding, and its subunit structure was deduced in the earliest days of modern receptor biology (1). The structure of its tyrosine kinase domain was the first of the tyrosine kinases to be solved by x-ray crystallography (2). Further, much is now known about downstream insulin receptor signaling components (3). Despite these advances, our understanding of how insulin and the related proteins, insulin-like growth factors I and II, stimulate glucose transport (via the translocation of glucose transporter proteins from intracellular to plasma membranes) is fragmentary. One reason for this is that the major target for insulin signaling to glucose transport is a complex membrane trafficking pathway that is likely to contain many unknown components.
As summarized in the preceding introductory comments (4) and detailed in the companion minireviews to be published subsequently (5,6), glucose uptake via the GLUT4 isoform of mammalian hexose transporters accounts for most of the stimulatory effect of insulin on this process in muscle and fat cells. GLUT4 rapidly recycles through the plasma membrane/endosomal membrane system in the presence of insulin. However, in the basal state this transporter protein is directed to and retained within specific intracellular membranes through the action of distinct elements within the GLUT4 structure (7). Insulin causes movement of GLUT4 out of this sequestered localization, leading to an increase in its steady state concentration on the cell surface membrane where it can catalyze glucose uptake. However, to accomplish this, insulin signaling is likely to regulate GLUT4 trafficking at multiple steps, consistent with data showing insulin also enhances cell surface display of proteins such as transferrin receptor and GLUT1 that recycle to a large extent independently of GLUT4. To further complicate our understanding of this system, different signaling elements may be required to modulate each of these different putative regulated steps in the GLUT4 trafficking pathway.

Insulin Receptor Signaling Circuits
Substrate phosphorylation by the insulin receptor tyrosine kinase appears to involve the binding of phosphorylated receptor tyrosine 960 to phosphotyrosine-binding (PTB) 1 domains of substrate proteins (8). Adjacent Pleckstrin homology (PH) domains on some substrate proteins also appear critical for receptor binding and phosphorylation. Tyrosine kinase signaling is often initiated by the recruitment of signaling proteins through their Src homol-ogy 2 (SH2) or PTB domains to phosphotyrosine sites. In the case of insulin receptor, tyrosine phosphorylation of four related substrate (IRS) proteins (8) and Gab-1 (9) causes many candidate signaling proteins to be recruited, including: 1) the p110-type phosphatidylinositol 3-kinase (PI 3-kinase) through the SH2 domains of p85 regulatory subunits; 2) Grb2 and the protein tyrosine phosphatase SH-PTP2, which appear to be necessary for p21 ras activation (10); 3) the tyrosine kinase Fyn, which in turn may also activate the PI 3-kinase and p21 ras pathways; and 4) Rho-associated protein serine/threonine kinase ROK␣ (11), which may modulate processes such as actin assembly and mitogenesis under control of the small GTPase Rho. Insulin receptor signaling can also engage p21 ras through tyrosine phosphorylation of Shc and its subsequent binding to complexes of Grb2 and Sos (12). Recently, it has been discovered that proteins can bind directly to the autophosphorylated insulin receptor through their SH2 domains, opening new avenues for investigation (13).
As illustrated in Fig. 1, p21 ras and the p85/p110-type PI 3-kinases represent two major initial switch elements for insulin receptor signaling. There is also evidence that the p21 ras -related GTP-binding proteins Rap (14), Rho (15), and Rac (16) are engaged by insulin receptor signal transduction. A central paradigm of insulin signaling is the involvement of multiple protein serine/ threonine kinases downstream of both the p21 ras and PI 3-kinase elements (Fig. 1). More recently, a group of non-receptor tyrosine kinases denoted Btk/Itk/Tec has been found in hemopoietic cells to respond to the generation of 3Ј-polyphosphoinositide, apparently by recruitment through their PH domains to membranes where they can be tyrosine-phosphorylated and activated (17). It is not yet known whether insulin or insulin-like growth factor-1 receptor signaling can act through 3Ј-phosphoinositides on these or other tyrosine kinases, and they are therefore not included in Fig. 1. Another recently discovered class of likely downstream effectors of 3Ј-phosphoinositides includes proteins that regulate membranerelated functions ( Fig. 1) such as actin assembly (Rac GTPase), early endosome fusion (EEA1), and guanine nucleotide exchange (GRP1, cytohesin-1, and ARNO) or possibly GTPase activation (␣-centaurin) of ARF proteins. It is certain that the list of potential downstream effectors of insulin receptor signaling will rapidly expand, as additional protein targets of 3Ј-phosphoinositides are discovered.
Much effort has been directed toward identifying which of the above insulin receptor signaling elements are actually linked to GLUT4 translocation. Although some evidence suggested p21 ras activation might influence glucose transport in fat (18) and muscle (19), the weight of most of the data available indicates no direct requirement of p21 ras function for insulin action on GLUT4 trafficking (20). Rather, strong indications that p85/p110-type PI 3-kinase activity is necessary for this insulin response have accumulated. Such evidence includes complete inhibition of insulin action on glucose transport by specific inhibitors of PI 3-kinase such as wortmannin (21) and LY29004 (22) and by microinjection or expression of dominant inhibitory constructs of the p85 regulatory subunit of PI 3-kinase (23). Further, in some experiments expression of an activated form of the p110 PI 3-kinase in cultured adipocytes mimics the stimulatory effect of insulin on glucose transport (24), suggesting PI 3-kinase activity is also sufficient for glucose transporter translocation.
Although the above results strongly support the hypothesis that p85/p110-type PI 3-kinase activation through its recruitment to phosphotyrosine sites on insulin receptor substrate proteins mediates GLUT4 translocation, it is instructive to review the limitations related to the current supporting data. First, the inhibitors wortmannin and LY29004 are not fully specific for PI 3-kinase activity and in any case are known to block PI 3-kinase isoforms lacking known regulatory subunits that contain SH2 or other domains that can be recruited to phosphotyrosine sites. Such other PI 3-kinase isoforms (25) rather than, or in addition to, the type p110 PI 3-kinase may be necessary for exocytosis of GLUT4. The use of the truncated p85 subunit containing SH2 domains as a dominant inhibitory reagent to inhibit native p85/p110 PI 3-kinase function also has the potential limitation of inadequate specificity. When expressed at high concentrations, SH2 domains can be promiscuous in binding phosphotyrosines within various amino acid sequences and thus may not be restricted to binding sites specific for endogenous p85/p110-type PI 3-kinase. Thus, although a key role for this PI 3-kinase in GLUT4 regulation by insulin seems almost certain, there are potential gaps in our understanding of even this point.

Multiple Signaling Pathways Regulate Glucose Transport
Recent results call into question the hypothesis that PI 3-kinase activation by insulin is sufficient for GLUT4 translocation. For example, it is well established that recruitment of PI 3-kinase to phosphotyrosines on the platelet-derived growth factor (PDGF) receptor in response to PDGF, which occurs in 3T3-L1 adipocytes to the same extent as recruitment of PI 3-kinase to protein phosphotyrosines in response to insulin, has virtually no effect on GLUT4 translocation (26). Recruitment of PI 3-kinase to IRS proteins in response to interleukin-4 (27) or cell surface integrin cross-linking (28) also fails to enhance glucose transport in the absence or presence of submaximal concentrations of insulin. Conversely, severe inhibition of insulin-mediated IRS protein tyrosine phosphorylation and recruitment of PI 3-kinase in response to incubation of 3T3-L1 adipocytes with PDGF failed to diminish glucose transport stimulation by insulin at all concentrations along the dose-response relationship (29).
The lack of correlation between PI 3-kinase activation and GLUT4 translocation in the studies cited above may reflect an additional insulin-specific signaling pathway or pathways required to operate in conjunction with PI 3-kinase activation to effect this biological response. According to this model, agents such as interleukin-4 and anti-integrin antibody fail to enhance glucose transport because they are unable to mimic the effect of insulin on this other signaling pathway. Recent surprising results support this hypothesis. In these experiments, a cell-permeable analog of the PI 3-kinase product PI(3,4,5)P 3 , thought to be a downstream effector for insulin action through PI 3-kinase, was unable to cause GLUT4 translocation when added alone to cells (30). However, in the presence of insulin and wortmannin, a condition where no stimulation of glucose transport is observed, the PI(3,4,5)P 3 analog was able to enhance cellular uptake of glucose (30). These results suggest that insulin may uniquely act to initiate PI 3-kinase-independent signaling events in the presence of wortmannin that collaborate with the PI 3-kinase signaling pathway to effect GLUT4 translocation.
Substantial recent information suggests that PI 3-kinase-independent mechanisms also regulate GLUT4 translocation in skeletal muscle during contraction. For example, the stimulations of glucose transport in skeletal muscle caused by insulin and contrac-tion are additive, but only the effect of the former is blocked by wortmannin (31). In concert with this result, insulin but not contraction stimulates PI 3-kinase activity and the downstream protein kinase Akt/protein kinase B (32). Conversely, the 5Ј-AMPactivated protein kinase (AMPK), thought to be responsive to cell stress, is stimulated by rat hind limb contraction but not insulin (33). Another activator of AMPK, 5-aminoimidazole-4-carboximide ribonucleoside, mimicked the action of contraction to enhance glucose uptake in a wortmannin-insensitive mode (33). Taken together, these recent findings suggest the hypothesis that AMPK may mediate, at least in part, exercise-stimulated glucose uptake in skeletal muscle.
Another cell signaling pathway that appears to markedly stimulate glucose uptake in muscle involves nitric oxide, which stimulates guanylate cyclase to produce cyclic GMP. The NO donor sodium nitroprusside and dibutyryl-cGMP both accelerated epitrochlearis (34), soleus (35,36), and extensor digitorum longus (36) muscle glucose transport. Upon raising NO levels by nitroprusside, cellular cGMP concentrations were found to rise in epitrochlearis muscle, as expected, but contraction did not elevate cGMP (34). Although there is general agreement that the effect of insulin on glucose uptake in muscle is not mediated through the actions of NO (34 -38), studies differ on whether NO is involved in glucose transport regulation by contraction. NO synthase inhibitors failed to block increased glucose uptake in response to contraction in rat epitrochlearis muscle (34), but such inhibitors did ablate exercisemediated glucose uptake in other studies (36,38).
Phorbol ester-sensitive protein kinase C isoforms may also be significant regulators of muscle glucose uptake. Thus, the protein kinase activator 12-deoxyphorbol 13-phenylacetate 20-acetate causes a 3-4-fold stimulation of glucose transport in isolated rat epitrochlearis muscle (39). The effect of phorbol ester on muscle glucose uptake is additive to the effects of insulin or hypoxia (39), indicating protein kinase C acts through a separate pathway and is an unlikely candidate to mediate the responses to these stimulants.
Trimeric G protein-linked receptor agonists have been reported to exert significant stimulatory effects on glucose transport, as exemplified by ␤-adrenergic agonists in skeletal muscle (40) and brown fat (41), ␣-adrenergic agonists in heart muscle (42), bradykinin in skeletal muscle (43), thrombin in platelets, reflecting GLUT3 translocation (44), and adenosine in white (45) and brown fat (46). These effects of catecholamines and bradykinin are not blocked by wortmannin, suggesting independence of the actions of most known isoforms of PI 3-kinase. Particularly interesting is the reported effect of adenosine to potentiate the stimulation by insulin of glucose uptake in brown fat cells from 15-to 30-fold (46). The data suggest that the adenosine receptor-coupled G q protein species is linked to glucose transport regulation.
It should be noted that effects of most of the signaling systems discussed in this section that act to enhance glucose transport have not yet been tested in GLUT4-deficient mice (6). Thus it is not clear whether some of these insulin-independent modulations of glucose transport involve other glucose transporter isoforms. It appears, for example, that cell surface GLUT1 transporter activity can be regulated, in some cases potentially independent of its recruitment to the cell surface (47). This may be particularly relevant in tissues that express significant levels of this glucose transporter isoform, such as brown fat. Table I summarizes many of the cellular regulators reported to stimulate glucose transport, and the isoform likely to be involved.

Role of Docking Proteins and PI 3-Kinase Localization
Of the regulatory mechanisms for glucose transport described above, research on the PI 3-kinase pathways has been the most extensive. Whether IRS proteins are required for PI 3-kinase signaling to act on the GLUT4 system is unclear at present. Microinjection or expression of the dominant inhibitory PTB or SAIN domains of IRS-1 in insulin-sensitive cultured adipocytes is able to block the mitogenic and membrane-ruffling effects of insulin but not the stimulation of glucose transport (48). These dominant inhibitory constructs are expected to block binding of all IRS protein isoforms to the insulin receptor through binding to the insulin receptor juxtamembrane sequence surrounding Tyr-960, the docking site for IRS and Shc protein PTB domains. Indeed, nearly Minireview: Signaling Mechanisms That Regulate Glucose Transport 1866 complete inhibition of IRS-1 tyrosine phosphorylation by insulin was caused by expression of its PTB domain in these studies (48).
In contrast to what would be predicted from these results, expression of IRS proteins in primary rat adipocytes apparently enhances GLUT4 translocation (49). Mutation of insulin receptor tyrosine 960, the site of IRS protein binding required for its phosphorylation, abolishes insulin action of glucose transport (50). Furthermore, studies in animals in which the IRS-1 or the IRS-2 gene has been ablated provide support for a role of these proteins in glucose transport regulation. Mice lacking IRS-1 exhibit some insulin resistance although no diabetes (51,52), and heterozygotes for loss of both IRS-1 and insulin receptor genes are both insulinresistant and develop diabetes (53). IRS-2 gene ablation in mice alone causes both impaired insulin signaling to glucose uptake and diabetes (54). Insulin action on glucose uptake in adipocytes from animals lacking IRS-1 protein has been suggested to involve the IRS-3 isoform (55), which is tyrosine-phosphorylated in response to insulin. Such redundancy may account in part for some of these apparently conflicting results. The information gained from IRS knockout mice provide compelling support for a role of these proteins in glucose transport regulation by insulin, with the caveat that unrelated changes in response to IRS loss during development may contribute to the phenotypes observed. Taken together, the data reported to date related to the role of IRS proteins in GLUT4 translocation are difficult to reconcile unless we again invoke the hypothesis that PI 3-kinase⅐IRS protein complexes are necessary but not sufficient for GLUT4 translocation by insulin.
An important question about the role of PI 3-kinase docking proteins in regulating GLUT4 trafficking is whether they function to localize the enzyme to specific cellular sites required for regulation of GLUT4 trafficking. IRS-1 (56), IRS-3 (57), and to a lesser extent IRS-2 (58) are present in cells as both cytosolic and membrane-bound proteins, both in basal and insulin-stimulated conditions. Membrane binding of PI 3-kinase, through its binding to docking proteins, is expected to be necessary for the enzyme to come into contact with its phospholipid substrate. The exact cellular location of membrane-bound PI 3-kinase⅐IRS complexes is not known, but endosomes and GLUT4-containing vesicles, potentially through binding AP3 adaptin complexes (59) and cytoskeletal elements (60), have been suggested. Only a very small fraction of the insulin-stimulated, IRS-1-bound PI 3-kinase activity co-purifies with GLUT4-containing vesicles, but the downstream protein kinase Akt appears to be recruited as well (61). Directed expression of PI 3-kinase activity on these vesicles in adipocytes using a GLUT4/p85 SH2 domain chimera failed to stimulate GLUT4 translocation (62). This latter result is not inconsistent with the hypothesis that a second signaling pathway is necessary for GLUT4 translocation, however.

Downstream Targets of 3-Phosphoinositides
The insulin-sensitive p85/p110 PI 3-kinases can catalyze phosphorylation of PI, PI(4)P, or PI(4,5)P 2 at the D-3 position of the inositol ring to produce the PI(3)P, PI(3,4)P 2 , or PI(3,4,5)P 3 . Recent results indicate PI(3,4,5)P 3 is necessary for insulin action on glucose transport, based on the inhibitory effect of microinjection of a phosphatase specific for this species into 3T3-L1 adipocytes (63). It is not yet clear whether the other two PI 3-kinase products are also necessary, but the recent identification of a protein target, EEA1, that binds PI(3)P through its RING finger and functions in the context of Rab5 in early endosome fusion (64) deserves further investigation (Fig. 1). Also, a class of proteins (65) denoted GRP1, ARNO, and cytohesin-1 containing PH domains that bind PI(3,4,5)P 3 with high affinity and a Sec7 domain that catalyzes guanine nucleotide exchange of ARF1, -5, and -6 proteins has been shown to be recruited to plasma membranes in response to insulin (66). The data available suggest these proteins may be involved in ARF6 function in membrane ruffling and actin rearrangements in response to insulin (67), but future studies on these and similar proteins in relation to GLUT4 translocation are warranted. The potential connection between PI 3-kinase signaling in insulin action and ARF function is highlighted by the discovery of polyphosphoinositide-binding protein ␣-centaurin, which contains some sequence similarity to ARF GTPase-activating protein (68). Taken together, these discoveries of proteins that bind directly to PI 3-kinase products and regulate known membrane modulators such as ARF proteins highlight a potentially rich interface between signaling and membrane trafficking.
Intense research has focused on protein serine/threonine kinases downstream of PI(3,4)P 2 and PI(3,4,5)P 3 that may regulate GLUT4. For example, recent reports (69) have shown that protein kinase C (Fig. 1) is activated by polyphosphoinositides, is stimulated in intact cells by insulin in a wortmannin-sensitive manner, and may contribute to glucose transport regulation. Another prime candidate has been Akt/protein kinase B, which is activated in conjunction with its binding to these lipids by the 3Ј-polyphosphoinositide-dependent protein kinase PDK1 and another protein kinase (70). Expression of membrane-directed constructs of Akt/protein kinase B in 3T3-L1 adipocytes (71), primary adipocytes (72), and L6 myotubes (73) stimulates GLUT4 translocation, and it is claimed that a dominant inhibitory construct blocks this process (72). However, in well controlled experiments using a mutant Akt/ protein kinase B with alanines substituted at phosphorylation sites threonine 308 and serine 473 as a dominant inhibitory construct in both Chinese hamster ovary cells and 3T3-L1 adipocytes, protein synthesis but not insulin-stimulated glucose transport was inhibited (74). These studies have recently been extended (75) to suggest that protein kinase C is directly involved in GLUT4 translocation. Taken together, there is not yet a clear consensus of data derived from multiple laboratories that any of the known protein kinases downstream of PI 3-kinase directly mediates insulin action on glucose transport.

Conclusions and Future Directions
Over the past few years many new components of the insulin receptor signaling network have been discovered and their cDNA clones isolated. Particularly exciting has been the identification of downstream targets of the insulin-regulated PI 3-kinase lipid products, including protein kinase C isoforms and , regulatory protein kinases of the Akt/protein kinase B system, tyrosine kinases Itk/Btk/Tec, the early endosome regulator EEA1, and ARF exchange factors GRP1, ARNO, and cytohesin-1. Because PI 3-kinase appears to be required for glucose transport regulation by insulin, future work is likely to connect one or more of these proteins (or additional ones to be discovered) to specific membrane trafficking components involved in GLUT4 translocation. We have also learned that additional cellular signaling elements such as trimeric G proteins, NO, and cGMP can regulate glucose transport and that unknown signaling events may be required in conjunction with PI  Insulin  IR, PI 3-kinase  GLUT4  Muscle, fat  This review  IGF-I  IGF-IR, PI 3-kinase  GLUT4  Muscle, fat  76  IGF-II  IGF-IR, PI 3-kinase  GLUT4  Muscle, fat  77  Contraction  5Ј-AMP-activated protein  Minireview: Signaling Mechanisms That Regulate Glucose Transport 1867 3-kinase for insulin to act. In the meantime, progress is being made on understanding the components of membrane systems through which GLUT4 traverses in insulin-sensitive cells, as will be discussed in the accompanying minireviews on this topic (5,6). Merging these two fertile research fields, signaling and membrane trafficking, will be the natural result of future exciting work on the mechanisms that underlie GLUT4 regulation.