Insulin Receptor Substrate-2 (IRS-2) Can Mediate the Action of Insulin to Stimulate Translocation of GLUT4 to the Cell Surface in Rat Adipose Cells*

Insulin receptor substrates-1 and -2 (IRS-1 and -2) are important substrates of the insulin receptor tyrosine kinase. Previous studies have focused upon the role of IRS-1 in mediating the actions of insulin. In the present study, we demonstrate that IRS-2 can mediate translocation of the insulin responsive glucose transporter GLUT4 in a physiologically relevant target cell for insulin action. Co-immunoprecipitation experiments performed on cell lysates derived from freshly isolated rat adipose cells incubated in the presence or absence of insulin indicated that twice as much phosphatidylinositol 3-kinase was associated with endogenous IRS-1 as with IRS-2 after insulin stimulation. When rat adipose cells in primary culture were transfected with expression vectors for IRS-1 or IRS-2, we observed 40-fold overexpression of human IRS-1 or murine IRS-2. In addition, anti-phosphotyrosine immunoblotting experiments confirmed that the recombinant substrates were phosphorylated in response to insulin stimulation. To examine the role of IRS-2 in insulin-stimulated translocation of GLUT4, we studied the effects of overexpression of IRS-1 and -2 on translocation of a co-transfected epitope-tagged GLUT4 (GLUT4-HA). Overexpression of IRS-1 or IRS-2 in adipose cells resulted in a significant increase in the basal level of cell surface GLUT4 (in the absence of insulin). Interestingly, at maximally effective concentrations of insulin (60 nm), the level of cell surface GLUT4 in cells overexpressing IRS-1 or -2 significantly exceeded the maximal recruitment observed in the control cells (160 and 135% of control, respectively; p < 0.003). Our data directly demonstrate that IRS-2, like IRS-1, is capable of participating in insulin signal transduction pathways leading to the recruitment of GLUT4. Thus, IRS-2 may provide an alternative pathway for critical metabolic actions of insulin.

IRS-2 is a protein homologous to IRS-1, originally identified as a protein phosphorylated in response to interleukin-4 (22,23). When the gene for IRS-1 was inactivated by gene targeting in transgenic mice, the IRS-1-deficient mice had reduced growth, impaired glucose tolerance, and decreased insulinstimulated glucose transport, but retained significant residual response to insulin (8,9). These observations suggested the existence of parallel pathways with the ability to mediate the metabolic actions of insulin and led to the identification and cloning of IRS-2 (23).
In vivo studies with IRS-1-deficient mice demonstrated that IRS-2 undergoes tyrosine phosphorylation, and associates with and activates PI 3-kinase in liver and muscle following insulin stimulation (24). Moreover, both the phosphorylation of IRS-2 and the association of PI 3-kinase activity with IRS-2 are enhanced in IRS-1-deficient mice. Furthermore, it appears that the IRS-2 pathway may be sufficient to compensate for IRS-1 deficiency in liver because insulin-induced activation of PI 3kinase is not significantly impaired in livers of IRS-1 deficient mice (25). In contrast, IRS-1 deficiency impairs the ability of insulin to stimulate glucose transport in isolated soleus muscle. This raises the provocative question of whether IRS-2 can substitute effectively for IRS-1 in mediating the activity of insulin to stimulate glucose transport.
Although functional similarities between IRS-1 and IRS-2 have been suggested based on their structural similarities, some data suggest that the two molecules may not be totally interchangeable. For example, the two proteins are not identical with respect to distribution and level of expression in various tissues. Furthermore, Brü ning et al. recently reported that IRS-2 cannot substitute for IRS-1 in mediating the mitogenic action of insulin-like growth factors (26). In light of these important questions, we addressed the issue of whether IRS-2 can participate in a similar fashion to IRS-1 in mediating the metabolic actions of insulin. We have used a previously described transient transfection system (27) to directly demonstrate the ability of recombinant IRS-2 to mediate the translocation of GLUT4 in rat adipose cells, a physiologically relevant target cell for insulin action in vivo. It is likely that this activity of IRS-2 is explained by its ability to bind and activate PI 3-kinase.
Genomic DNA Encoding Murine IRS-2-The mouse IRS-2 gene was cloned from a mouse P1 genomic DNA library ES 129/ola (Genome Systems, Inc., St. Louis, MO) that was screened using the polymerase chain reaction (PCR). PCR was performed using two sequence-specific oligonucleotides (sense primer, 5Ј-CTTAGGAGTGGTGGTCCCAAT-AGC-3Ј; antisense primer, 5Ј-AAGCGTGGCTGCAGGATCT -3Ј) under the following PCR conditions: denaturation, 95°C for 5 min, for the first cycle, and 1 min in later cycles; annealing, 55°C for 1 min; extension, 72°C for 2 min (extended to 10 min in the last cycle). The PCR amplified a 98-base pair band. Ten P1 clones were identified in the first screening. DNA was prepared from three clones and analyzed by PCR and Southern blotting. In two clones, the restriction pattern was identical. In Southern blot analysis, the oligonucleotide probes (identical to the PCR primers listed above) hybridized to a 6.4-kb BamHI fragment. This genomic DNA fragment was then subcloned into pBK-CMV (Stratagene), and analyzed by restriction mapping. The nucleotide sequence was determined, and we identified a 3963-base pair open reading frame spanning the entire coding sequence of murine IRS-2 (23). In addition, the clone contained 1.2-kb of 5Ј-flanking sequence and 1.2-kb of 3Јflanking sequence. The BamHI fragment was excised, blunt-ended, and ligated into the pCIS2 expression plasmid.
Expression Plasmids-pCIS2, an expression vector that generates high expression levels in transfected rat adipose cells (28), was used as the parent vector for subsequent constructions. Expression vectors for GLUT4-HA and hIRS-1 were constructed as described previously (7).
Preparation and Electroporation of Isolated Rat Adipose Cells-Isolated adipose cells were prepared from epididymal fat pads of male rats (170 -200 g, CD strain, Charles River Breeding Laboratories, Wilmington, MA) by collagenase digestion as described (27,29). Isolated adipose cells were transfected by electroporation as described (27,28). For experiments where cell surface epitope-tagged GLUT4 was measured, cells from 15 cuvettes were pooled for each of four groups transfected with pCIS2, pCIS2/GLUT4-HA, hIRS1/GLUT4-HA, and mIRS2/ GLUT4-HA. A total DNA concentration of 9 g/cuvette was used for all groups (3 g/cuvette of GLUT4-HA and 6 g/cuvette of the co-transfected plasmid).
Assay for Cell Surface Epitope-tagged GLUT4 -20 h after electroporation, adipose cells were processed as described (27) and treated without or with insulin at a concentration of 0.07 or 60 nM at 37°C for 30 min. Cell surface epitope-tagged GLUT4 was determined by using the anti-HA mouse monoclonal antibody HA-11 in conjuction with 125 Ilabeled sheep anti-mouse IgG as described (30). Cells transfected with the parent vector (pCIS2) alone were used to determine nonspecific binding of the antibodies. The lipid weight from a 200-l aliquot of cells was determined as described (31) and used to normalize the data for each sample.
Immunoprecipitation and Immunoblotting-Adipose cells were cotransfected with GLUT4-HA and either pCIS2 or expression vectors for hIRS-1 or mIRS-2 as described above. 20 h after electroporation, cells were treated without or with insulin (60 nM) for 2 min at 37°C. The cells were then washed once and resuspended in an equal volume of TES buffer (20 mM Tris, 1 mM EDTA, 8.73% sucrose, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml pepstatin, 2 mM sodium orthovanadate, 10 g/ml leupeptin, and 10 g/ml aprotinin, pH 7.4) at 18°C. The cells were homogenized by passing through a 25-gauge needle three times. The total time elapsed from the beginning of insulin stimulation to homogenization was approximately 4 min. Total membrane fractions were isolated as described previously (30,32).
The samples were boiled in Laemmli sample buffer for 5 min and subjected to SDS-polyacrylamide gel electrophoresis. Western blot analysis was carried out with the indicated antibody (1:3000 dilution). For blotting with HA-ll antibody (final concentration 5 g/ml), samples were not boiled. The enhanced chemiluminescence detection system (Amersham Life Science, Inc.) was used to visualize the bands. For quantitative studies, we adjusted the exposure time for the x-ray films in an effort to ensure that the intensities of the bands would be linearly related to the quantity of protein contained in the bands. In fact, when we analyzed multiple films that had been obtained with different exposure times, we obtained nearly identical estimates of the relative intensities of the bands. These observations support the conclusion that the exposures were indeed in the linear range.
Statistical Analysis-Paired Student t tests were used to compare individual points where appropriate. P values of less than 0.05 were considered to indicate statistical significance.

Characterization of Endogenous Insulin Receptor Substrates in Rat Adipose
Cells-To understand the physiological significance of IRS-2 in insulin signaling, we compared IRS-2 with IRS-1 in rat adipose cells with respect to their ability to associate with the downstream signaling molecule PI 3-kinase. Isolated rat adipose cells were incubated in the presence or absence of insulin, and aliquots of cell lysates containing equal amounts of protein were immunoprecipitated by anti-IRS-1 or anti-IRS-2 antibody. We did not detect coimmunoprecipitation of the p85 subunit of PI 3-kinase by either anti-IRS-1 or anti-IRS-2 antibody when lysates were derived from cells that had not been treated with insulin (data not shown). In contrast, insulin stimulation resulted in PI 3-kinase association with both IRS-1 and IRS-2 ( Fig. 1). We detected approximately 50% more p85 coimmunoprecipitated with IRS-1 as compared with IRS-2. To interpret these observations, we compared the efficiencies of both antibodies in immunoprecipitating their respective targets. Using immunoblotting, we quantitated the amount of IRS-1 or IRS-2 in the cell lysate before and after the immunoprecipitation (data not shown). We estimated that anti-IRS-1 and anti-IRS-2 antibodies immunoprecipitated approximately 75 and 90% of their respective antigens. After correcting for the relative efficiencies of immunoprecipitation, these data suggest that both IRS-1 and IRS-2 participate in mediating the ability of insulin to activate PI 3-kinase and that IRS-1 may account for approximately two thirds and IRS-2 approximately one-third of the effect of insulin to activate PI 3-kinase.
Characterization of Overexpressed Recombinant hIRS-1 and mIRS-2 in Rat Adipose Cells-Previously, we have demonstrated that overexpression of hIRS-1 can stimulate the translocation of GLUT4 to the plasma membrane in rat adipose cells (27). We have used the same approach to inquire whether In other experiments, we demonstrated that p85 was not detectable in immunoprecipitates obtained from adipose cells that had not been treated with insulin (data not shown).
IRS-2 has similar biological activity. Therefore, we transfected adipose cells with expression vectors for either hIRS-1 or mIRS-2. Because previous studies demonstrated that hIRS-1 and mIRS-2 were located almost exclusively in total membrane fractions prepared from transfected rat adipose cells, 2 we used immunoblotting of total membrane fractions to quantitate the levels of expression of these two proteins in our experiments. Transfection of cells led to Ϸ3-fold increase in the total cellular content of IRS-1 and IRS-2, respectively ( Fig. 2A). (Note that it is not possible to directly compare the quantities of the two recombinant proteins (IRS-1 and IRS-2) since two different antibodies are used in the immunodetection.) Because only Ϸ5% of the electroporated cells are actually transfected under our experimental conditions (27), recombinant IRS-1 and IRS-2 were likely overexpressed by Ϸ40-fold in the sub-population of cells that were actually transfected. We then tested the ability of insulin to stimulate tyrosine phosphorylation of overexpressed hIRS-1 and mIRS-2. In control cells (transfected with empty expression vector pCIS2), insulin increased the phosphotyrosine content of endogenous insulin receptor substrates. In groups of cells transfected with IRS-1 or IRS-2, we observed a 1.7-or 1.5-fold increase, respectively, in insulin-stimulated phosphotyrosine content in the IRS-1/2 band, consistent with the level of overexpression of IRS-1 and IRS-2 (Fig. 2B). In this experiment, we did not resolve IRS-1 and IRS-2 as separate bands with different electrophoretic mobilities.
Effect of Overexpression of hIRS-1 or mIRS-2 on Translocation of GLUT4 -We next examined the ability of IRS-2 to mediate insulin-stimulated translocation of GLUT4 in rat adipose cells. The effect of IRS-2 was also compared with that of IRS-1, which we used as a positive control. Rat adipose cells were co-transfected with GLUT4-HA together with either "empty" control expression vector (pCIS2) or expression vectors for hIRS-1 or mIRS-2. The use of epitope-tagged GLUT4 enabled us to monitor translocation selectively in the small fraction of transfected cells. When control cells were co-transfected with pCIS2 and GLUT4-HA, insulin treatment (60 nM) led to a 3-fold increase in the quantity of tagged GLUT4 located on the cell surface (Fig. 3). Half-maximal translocation of GLUT4-HA was elicited at an insulin concentration of Ϸ0.07 nM. Overexpression of IRS-1 or IRS-2 in adipose cells resulted in a doubling of the basal level of cell surface GLUT4 even in the absence of insulin (to 2.3-and 2.0-fold of basal levels, respectively). In fact, the effects of overexpression of IRS-1 or -2 in the absence of insulin were comparable with that seen in control cells stimulated with 0.07 nM insulin, the half-maximally effective concentration of insulin. Interestingly, at maximally effective concentrations of insulin (60 nM), the level of cell surface GLUT4 in cells overexpressing IRS-1 or IRS-2 significantly exceeded the maximal recruitment observed in the insulinstimulated control cells (160 and 135% of control, respectively; p Ͻ 0.003). At an intermediate concentration of insulin (0.07 nM), IRS-1 and IRS-2 increased the quantity of GLUT4-HA on the cell surface by 90% and 60%, respectively. Thus, like IRS-1, IRS-2 has the ability to mediate the effect of insulin to recruit GLUT4 from intracellular membranes to the cell surface. Under all of our experimental conditions, IRS-1 appeared to have a larger effect than IRS-2 on translocation of GLUT4. However, it is not clear whether this apparent difference reflects a difference in the intrinsic biological activities of the two proteins or, alternatively, other factors such as the relative levels at which the two recombinant proteins were expressed.
As a control experiment, we inquired whether overexpression of hIRS-1 or mIRS-2 altered the expression of recombinant GLUT4-HA directed by our expression vector. When total membrane fractions derived from transfected cells were immunoblotted with anti-HA monoclonal antibody, we observed comparable levels of GLUT4-HA in cells co-transfected with pCIS2, hIRS-1, or mIRS-2 (Fig. 4). Therefore, the increase in GLUT4-HA detected at the surface of cells co-transfected with either hIRS-1 or mIRS-2 could not be explained by a change in the level of expression of GLUT4-HA. Rather, the effects of overexpressed IRS-1 and IRS-2 are most likely explained by their ability to stimulate translocation of GLUT4.  1 and 2), hIRS-1 (lanes 3 and 4) or mIRS-2 (lanes 5 and 6) after treatment with or without insulin (60 nM for 2 min). The samples were separated by SDS-polyacrylamide gel electrophoresis and immunoblotted with anti-phosphotyrosine antibody 4G10.

FIG. 3. Effects of overexpression of IRS-1 or IRS-2 on translocation of GLUT4 in rat adipose cells.
Cells were co-transfected with GLUT4-HA (3 g/cuvette) and pCIS2 (control), hIRS-1, or mIRS-2 (6 g/cuvette). We measured the amount of GLUT4-HA at the cell surface in the basal state or in response to insulin (0.07 or 60 nM) as described under "Materials and Methods." Data are expressed as a percentage of cell surface GLUT4-HA in the presence of a maximally effective insulin concentration for the control group (pCIS2). Results are the means Ϯ S.E. of five independent experiments. The actual values for the specific cell-associated radioactivity for the control group at 60 nM insulin was 1865 Ϯ 348 cpm. In the absence (basal state) or presence of insulin, cells overexpressing either hIRS-1 or mIRS-2 had significantly higher levels of GLUT4-HA at the cell surface than the corresponding control. At 60 nM insulin, the level of GLUT4-HA at the surface of cells overexpressing hIRS-1 was significantly greater than for cells overexpressing mIRS-2 (p Ͻ 0.01).

DISCUSSION
The identification and cloning of IRS-1 represented a major advance in the understanding of the mechanism of insulin action (23). Insulin binding to its receptor activates intrinsic receptor tyrosine kinase activity resulting in phosphorylation of cellular substrates which, in turn, bind to downstream effector molecules containing SH2 domains. This leads to the activation of multiple downstream pathways, among which PI 3-kinase appears to be the most important for mediating metabolic actions of insulin (17)(18)(19). When the IRS-1 gene was inactivated by gene targeting in mice, it was observed that the mice were insulin resistant but retained a partial ability to respond to insulin. Thus, there must be parallel pathways available to mediate the metabolic actions of insulin. The discovery of IRS-2, a protein with close homology to IRS-1, led to the suggestion that IRS-2 might provide one such parallel pathway. Although the overall amino acid sequence identity between IRS-1 and IRS-2 is 43%, these two molecules share domains of high homology (23). The greatest sequence identity is found in the N-terminal portion of the molecules in the pleckstrin homology and phosphotyrosine-binding domains. In addition, the tyrosine phosphorylation motifs identified in IRS-1 are well conserved in the IRS-2 molecule (23).
Despite the structural similarities, several observations suggest that IRS-1 and IRS-2 might not be entirely interchangeable from a functional point of view. First, whereas IRS-2 appears to compensate for the absence of IRS-1 in liver, inactivation of the IRS-1 gene leads to insulin resistance in soleus muscle despite the presence of IRS-2 in that tissue (25). Second, there are some differences in the sequences surrounding tyrosine phosphorylation sites in IRS-1 and IRS-2 so that it is possible that they may interact with a different collection of SH2 domain-containing proteins. Indeed, it has been suggested that IRS-2 cannot substitute for IRS-1 in mediating the mitogenic actions of IGF-I (26). This may help to explain the growth retardation observed in mice deficient in IRS-1. Third, there are other structural differences between IRS-1 and IRS-2. For example, at least two laboratories (33,34) have identified a domain in IRS-2 (but not in IRS-1) that participates in the binding interaction between IRS-2 and the insulin receptor. Finally, there are other candidate substrates that might be responsible for mediating parallel pathways of insulin action: e.g. IRS-3, a 60-kDa protein that has been shown to bind to PI 3-kinase in adipose cells (35). In addition, direct interaction of PI 3-kinase with the insulin receptor C terminus may provide an alternative pathway to activate PI 3-kinase that does not require IRS-1 (36,37).
Overexpression of Signaling Molecules in Primary Adipose Cells-In this work, we have used a transient transfection system to study the role of IRS-2 in insulin-elicited translocation of GLUT4 in primary cell cultures (27). Since our adipose cell transfection system has only ϳ5% efficiency in transfecting cells, it was not possible to directly assess the translocation of GLUT4 by measuring the cell surface levels of GLUT4. Instead, we used an epitope-tagged GLUT4 as a reporter for GLUT4 translocation to monitor GLUT4 translocation exclusively in the fraction of cells that are transfected. Since the total levels of GLUT4-HA expressed in the co-transfected cells are comparable for all groups of cells, changes in cell surface GLUT4-HA are presumably related to changes in the translocation process. In the co-transfection experiments, we used twice as much DNA for the IRS constructs as we did for GLUT4-HA to increase the likelihood that cells transfected with GLUT4-HA were also transfected with the vector of interest. If some fraction of cells were transfected only with GLUT4-HA, our results would underestimate the difference between control and experimental groups. According to previous studies (using an identical protocol) where we demonstrated nearly complete inhibition of insulin-stimulated translocation of GLUT4-HA by coexpressing a dominant inhibitory mutant of PI 3-kinase, we believe that at least 95% of cells expressing GLUT4-HA also express the co-transfected second plasmid under our experimental conditions (19).
The levels of overexpression for hIRS-1 or mIRS-2 constructs in transfected adipose cells are similar to what have been observed with other recombinant genes in this system (7,30,32). We demonstrated that overexpressed hIRS-1 and mIRS-2 were tyrosine phosphorylated upon insulin stimulation. The increase in tyrosine phosphorylation is comparable with levels of hIRS-1 and mIRS-2 overexpression. Overexpressing hIRS-1 or mIRS-2 increased insulin-stimulated translocation of GLUT4 as compared with control cells. The effect of hIRS-1 on GLUT4 translocation in response to insulin stimulation is in good agreement with the results reported previously (7). Interestingly, overexpression of hIRS-1 or mIRS-2 caused an increase in the levels of cell surface GLUT4 even in the absence of insulin. It is likely that the insulin receptor exhibits a low level of intrinsic tyrosine kinase activity even in the absence of a ligand that is capable of signaling recruitment of GLUT4 in adipose cells (27). Overexpression of IRS-1 or IRS-2 presumably amplified this basal signal. The fact that overexpression of the protein tyrosine phosphatase PTP1B in fat cells decreases basal cell surface GLUT4 is also consistent with this idea (30). Thus, in addition to increasing signaling in the presence of insulin, it is possible that overexpression of hIRS-1 or mIRS-2 is amplifying the small signal generated by unoccupied insulin receptors.
Role of IRS-2 in Mediating Metabolic Actions of Insulin-In the present study, we have obtained direct evidence in support of the hypothesis that IRS-2 can mediate insulin's ability to stimulate translocation of GLUT4 to the plasma membrane. When we overexpressed IRS-2 in rat adipose cells, we observed the recruitment of GLUT4 to the plasma membrane. This observation demonstrates that IRS-2 possesses the intrinsic capacity to mediate this metabolic action of insulin. Nevertheless, to evaluate the physiologic significance of this observation, it is necessary to determine whether IRS-2 is capable of serving this function when expressed at physiological levels. The tissue distributions of IRS-1 and IRS-2 proteins have been extensively studied by Sun et al. (38). Differential expression of IRS-1 and IRS-2 has been observed; for example, IRS-2 is predominant in hematopoietic cells, while IRS-1 is absent from several lymphoid cell lines and cells of the myeloid lineage. Nevertheless, both IRS-1 and IRS-2 proteins are expressed in a variety of tissues and cell types. Comparable amounts of IRS-1 and IRS-2 have been detected in liver, muscle, and adipocytes. This indi- FIG. 4. Co-transfected cells express comparable levels of epitope-tagged GLUT4. Total membrane fractions prepared from cells co-transfected with GLUT4-HA and either pCIS2, hIRS-1, or mIRS-2 were immunoblotted with an anti-HA monoclonal antibody. Cells transfected with pCIS2 alone represent a negative control for immunoblotting since these cells do not express GLUT4-HA (lane 4). Comparable levels of GLUT4-HA are seen for cells co-transfected with GLUT4-HA and either pCIS2, hIRS-1, or mIRS-2 (densities ϭ 6880, 5638, and 6705 arbitrary units, in lanes 1-3, respectively). cates that they may be essential for maintaining cell growth and metabolic regulation. As has been proposed for IRS-1, it is likely that the ability of IRS-2 to associate with and activate PI 3-kinase may provide a mechanism for it to mediate insulinstimulated recruitment of GLUT4 in adipose cells. To determine whether there is sufficient IRS-2 to contribute to insulin action in rat adipose cells, we have investigated the function of IRS-2 in freshly prepared rat adipose cells. Insulin stimulated the association of both IRS-1 and IRS-2 with the p85 subunit of PI 3-kinase. While more molecules of PI 3-kinase were associated with IRS-1, IRS-2 bound approximately half the quantity of PI 3-kinase that was bound to IRS-1.
There are at least two possible explanations for our observation that endogenous IRS-1 binds more PI 3-kinase than does endogenous IRS-2. First, it is possible that there is simply more IRS-1 than IRS-2 in the relevant intracellular compartment. Second, IRS-1 might be intrinsically more efficient than IRS-2 at activating PI 3-kinase. If one assumes that the recombinant molecules (hIRS-1 and mIRS-2) are expressed at equal levels in the transfected cells, our observation that hIRS-1 has a larger effect upon translocation of GLUT4-HA is consistent with the hypothesis that IRS-1 may be intrinsically more active in mediating this metabolic action of insulin. Interestingly, in 3T3-L1 fibroblasts, both IRS-1 and IRS-2 were barely detectable by immunoblotting before differentiation. However, after differentiation the amount of IRS-1 increased dramatically, while IRS-2 was barely detected (38). Thus, 3T3-L1 adipocytes appear to be an imperfect model for fat cells in that, unlike freshly prepared adipocytes, they do not express significant quantities of IRS-2.
Conclusion-In conclusion, we have used a physiologically relevant target cell for insulin action to directly demonstrate that overexpression of IRS-2 is sufficient to recruit GLUT4 from intracellular vesicles to the surface of adipose cells. This suggests that IRS-2, together with IRS-1, participates in insulin signal transduction pathways related to the metabolic actions of insulin. Furthermore, our observations are consistent with the hypothesis that IRS-2 can partially compensate for deficiency of IRS-1 in mediating the activation of PI 3-kinase (26).