Insulin Receptor Substrate 3 Is Not Essential for Growth or Glucose Homeostasis*

The insulin receptor substrates (IRS) 1 and 2 are required for normal growth and glucose homeostasis in mice. To determine whether IRS-3, a recently cloned member of the IRS family, is also involved in the regulation of these, we have generated mice with a targeted disruption of the IRS-3 gene and characterized them. Compared with wild-type mice, the IRS-3-null mice showed normal body weight throughout development, normal blood glucose levels in the fed and fasted state and following an oral glucose bolus, and normal fed and fasted plasma insulin levels. IRS-3 is most abundant in adipocytes and is tyrosine-phosphorylated in response to insulin in these cells. Therefore, isolated adipocytes were analyzed for changes in insulin effects. Insulin-stimulated glucose transport in the adipocytes from the IRS-3-null mice was the same as in wild-type cells. The extent of tyrosine phosphorylation of IRS-1/2 following insulin stimulation was similar in adipocytes fromIRS-3-null and wild-type mice, and the insulin-induced association of tyrosine-phosphorylated IRS-1/2 with phosphatidylinositol 3-kinase and SHP-2 was not detectably increased by IRS-3 deficiency. Thus, IRS-3 was not essential for normal growth, glucose homeostasis, and glucose transport in adipocytes, and in its absence no significant compensatory augmentation of insulin signaling through IRS-1/2 was evident.

The IRSs 1 play a key role in signal transduction from the insulin receptor (reviewed in Ref. 1). They are the major intracellular targets for phosphorylation by the activated insulin receptor tyrosine kinase. In addition, they are also substrates for the insulin-like growth factor I receptor and for tyrosine kinases associated with the receptors for growth hormone and some interleukins and interferons. Four members of the IRS family (IRS-1, -2, -3, and -4) are now known. Each IRS contains at its N terminus a pleckstrin homology domain and a phosphotyrosine binding domain, both of which are required for efficient phosphorylation by the insulin receptor tyrosine ki-nase. The large C-terminal portion of each IRS contains many tyrosine phosphorylation sites in motifs that can associate with SH2 domain-containing proteins, notably PI 3-kinase, the adaptor protein Grb-2 (complexed with Sos, the guanine nucleotide exchange factor for Ras), and the tyrosine phosphatase SHP-2. The interaction of the tyrosine-phosphorylated IRSs with PI 3-kinase and Grb-2/Sos stimulates their activities, leading to the elevation of PI 3,4,5-trisphosphate and the GTP form of Ras. These in turn activate protein kinase cascades, resulting in the well established cellular effects of insulin, including the stimulation of glucose transport, glycogen synthesis, and protein synthesis and alterations in gene transcription (reviewed in Refs. 1 and 2).
The occurrence of four members of the IRS family raises the question of the physiological roles of each. Targeted disruption of the IRS genes provides an approach to address this question. To date mice lacking IRS-1 and IRS-2 have been described. Mice deficient in IRS-1 are growth-retarded and moderately insulin-resistant (3,4). Mice lacking IRS-2 develop diabetes early in life due to severe insulin resistance combined with a failure of the pancreatic ␤ cells to proliferate (5). From these phenotypes it can be concluded that IRS-1 and 2 play important roles in the regulation of growth and glucose homeostasis.
We recently purified and cloned IRS-3 (6). To determine the physiological role of IRS-3, we have now generated mice with targeted disruption of the IRS-3 gene. We report here that IRS-3-null mice do not exhibit abnormalities in growth or glucose homeostasis. In adipocytes IRS-3 is a prominent insulinelicited Tyr(P) protein that associates with PI 3-kinase and SHP-2, and weakly with Grb-2 (7). Consequently, we examined insulin effects in adipocytes from the IRS-3-null mice. Insulin stimulation of glucose transport was not impaired. Moreover, the absence of IRS-3 did not lead to detectable increases in the tyrosine phosphorylation of IRS-1/2 or their association with PI 3-kinase or SHP-2.

EXPERIMENTAL PROCEDURES
Generation of IRS-3-null Mice-Mouse P1 genomic clones containing the IRS-3 gene were isolated for us by Genome Systems Inc, by PCR screening of a P1 library of mouse ES cell genomic DNA (mouse strain 129) with two primers derived from the rat IRS-3 cDNA. The presence of the IRS-3 gene was confirmed by Southern blotting of restriction digests with a probe from the rat cDNA. Subsequently, we subcloned several restriction fragments of the mouse gene encoding IRS-3 into pBluescript and sequenced the coding region. In addition, we sequenced the insert of a plasmid that contains the complete coding cDNA for mouse IRS-3 (expressed sequence tag number 557145 from Genome Systems, Inc). Comparison of the two sequences revealed that the coding region of the mouse IRS-3 gene contains only a single small intron of 344 bp near the amino terminus. The identity of the two sequences in the coding region definitely established that the P1 clones contained IRS-3. As we were preparing this information for publication, another group published the same information (8), and so the details are not presented here.
In order to construct the targeting vector for disruption of the IRS-3 gene, two overlapping fragments (a 6.0-kb SpeI-SpeI and a 9.3-kb EcoRI-EcoRI fragment; see Fig. 1A), as well as a larger fragment en-* This work was supported by National Institutes of Health Grant DK42816. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  compassing these two (16.5-kb EagI-EagI fragment), were subcloned into pBluescriptSKϩ, and mapped for restriction sites. For the targeting vector, which was assembled in pBluescriptSKϩ, a 2.9-kb SpeI-EcoRI fragment containing the entire coding region of IRS-3 was replaced in opposite orientation by a 1.8-kb fragment containing the neo gene driven by the phosphoglycerate kinase promoter (PGKneo) and linked to phosphoglycerate kinase polyadenylation sequences (Fig. 1A). The neo gene was flanked by a 3.5-kb SpeI-SpeI fragment and a 2.5-kb EcoRI-EcoRI fragment derived from the IRS-3 regions 5Ј and 3Ј to the coding region, respectively. A 2.6-kb thymidine kinase gene under the regulation of the phosphoglycerate kinase promoter was placed at the 3Ј end of the targeting vector.
The targeting vector was linearized with NotI and electroporated into J1 ES cells. Colonies that were resistant to G418 (Life Technologies, Inc.) and gancyclovir (Syntex) were picked and expanded as described (9) and screened for homologous recombination by Southern blotting of EcoRI-digested genomic DNA. Seventeen ES clones out of 177 showed the 8.2-kb fragment characteristic for disruption of IRS-3 as well as the 9.3-kb fragment from the wild-type allele (see Fig. 1B). Four of these ES clones were injected into blastocysts derived from C57BL/6 and BALB/c mice in the facility directed by Dr. Arlene Sharpe (Department of Pathology, Brigham and Women's Hospital, Boston, MA). One clone yielded highly chimeric male mice, and these were bred with female wild-type C57BL/6 and BALB/c mice to obtain IRS-3 ϩ/Ϫ mice. The mice heterozygous for IRS-3 were mated to obtain IRS-3-null mice. These and their wild-type littermates were used in the experiments described in Table I  Mice were genotyped by Southern blotting and also by PCR analysis of genomic DNA obtained from tail snips. Southern blotting was performed on EcoRI-digested DNA using a digoxigenin-labeled probe that was prepared according to the manufacturer's (Roche Molecular Biochemicals) instructions by PCR amplification of a 0.9-kb SacI-XbaI fragment located upstream of the 5Ј homologous sequence of the targeting vector (Fig. 1A). Wild type and IRS-3-null mice showed the expected 9.3-and 8.2-kb fragments, respectively (Fig. 1B). For PCR analysis, the sense and antisense primers for the IRS-3 gene corresponded to nucleotides 1559 -1581 and 2017-2039, respectively, in the coding region of IRS-3 ( Fig. 1 of Ref. 8), and were expected to yield a product of 481 bp. The sense and antisense primers for the neo gene were 5Ј-AGG ATC TCG TCG TGA CCC ATG-3Ј and 5Ј-AAG GCG ATA GAA GGC GAT GC-3Ј, respectively, and were expected to yield a 233-bp product. As shown in Fig. 1C, PCR analysis yielded the fragments of the predicted sizes. Results from both methods of genotyping agreed.
Analysis of Blood Glucose and Plasma Insulin Concentrations and of Glucose Tolerance-Blood samples were collected from the tails cut at the tip of either randomly fed or fasted mice. For the former, samples were taken between 7 and 8 a.m.; for the latter, mice were fasted from 5 p.m. to 9 a.m., and then samples taken. Blood glucose was measured with the Precision⅐G blood glucose testing system (Medisense). For measurement of plasma insulin, heparinized (10 units/ml) blood samples were spun at 8000 rpm in a microcentrifuge for 10 min at 4°C, and the supernatant plasma was used in an insulin radioimmune assay (Linco Research Inc., catalog numbers RI and SRI 13K). For the glucose tolerance test, mice were fasted from 5 p.m. to 9 a.m., and then an oral glucose bolus of 1 g of D-glucose/kg of body weight as a solution of 0.25 g/ml was introduced with a feeding needle. Glucose levels were determined in blood drops obtained from the tail vein immediately before and at 10, 20, 30, 60, 90, and 120 min after the administration of glucose.
Isolation of Adipocytes and Assay of Glucose Transport-Epididymal or parametrial fat pads were removed from male or female 12-16-weekold mice sacrificed with CO 2 . Adipocytes were isolated by collagenase digestion of fat pads (10), as follows. Fat pads were added to collagenase (type I from Worthington Biochemical Corp.) (1 mg/ml, 2 mg/g of tissue) that was dissolved in KRH-BSA (120 mM NaCl Measurements of adipocyte number, lipid content, and relative GLUT4 content were also made with aliquots of the adipocyte preparations used for the transport assays. For determination of the cell number, adipocytes were fixed with osmium tetroxide and counted in a Coulter counter, as described in (12). For the determination of the lipid content, total lipids were extracted and weighed, according to Ref. 13, as follows. A 200-l aliquot of the 5% cell suspension was mixed with 2.7 ml of 40:10:1 isopropanol/heptane/1 N H 2 SO 4 , followed by the addition of 1.8 ml of heptane and 1.0 ml of water. The mixture was vortexed and centrifuged briefly. A 1.0-ml aliquot of the organic layer was evaporated, and the lipid weighed. For the determination of the GLUT4 content, 200 l of the 5% cell suspension was spun through dinonyl phthalate to remove the medium. The isolated adipocytes were solubilized in 100 l of SDS sample buffer and immunoblotted with antibodies against the C-terminal peptide of GLUT4 (14) at 5 g/ml, as described below. These three analyses were performed in quadruplicate.
Preparation of Adipocyte Lysates-Adipocytes were isolated as described above, except that the buffer was KRBH-BSA (120 mM NaCl, 4 mM KH 2 PO 4 , 1 mM MgSO 4 , 1 mM CaCl 2 , 10 mM NaHCO 3 , 30 mM HEPES, 1% BSA, 3 mM D-glucose, and 200 nM adenosine, pH 7.4). The cells were washed with KRBH to remove the albumin, diluted in KRBH to give an approximately 10% cell suspension, and then treated with or without 500 nM insulin (Humulin R) for 5 min at 37°C. Adipocytes were centrifuged at 80 ϫ g for 1 min to obtain a packed cell suspension. The infranatant was removed, and the cells were lysed in the following buffers (approximately 2 ml/ml of packed cells) to yield a final protein concentration of 1-2 mg/ml. When samples were used directly for SDSgel electrophoresis and immunoblotting, the lysis buffer contained 1% SDS, 20 mM dithiothreitol, 2 mM EDTA, 40 mM HEPES, 150 mM NaCl, pH 7.5, with phosphatase inhibitors (10 mM sodium pyrophosphate, 10 mM NaF, 10 mM Na 3 VO 4 ) and protease inhibitors (10 g/ml aprotinin, 10 M leupeptin, 10 M EP475, 1 M pepstatin A). Samples were held at 100°C for 5 min, and the DNA was sheared with a 23-gauge needle. When samples were used for immunoprecipitation, the SDS in the above lysis buffer was replaced with 1% nonaethylene glycol dodecyl ether (Thesit, Roche Molecular Biochemicals), and dithiothreitol was omitted. The lysates were centrifuged at 14,000 ϫ g for 10 min, and the supernatant was frozen in liquid nitrogen and stored at Ϫ80°C until use for immunoprecipitation as described below. Protein concentration was measured by a precipitating Lowry assay (15).
Immunoprecipitation and Immunoblotting of Adipocyte Proteins-Adipocyte lysates (250 g) at 1 mg/ml were incubated on ice for 2 h with antibodies against PI 3-kinase (2 l antisera) or SHP-2 (5 g) or as a control with irrelevant rabbit immunoglobulin (5 g). Immune complexes were collected onto 20 l of protein A-Sepharose for 2 h. Beads were washed twice with the Thesit-containing lysis buffer. The immunoprecipitated proteins were solubilized by holding the beads at 100°C for 5 min in 100 l SDS sample buffer (4% SDS, 20 mM dithiothreitol, 10% glycerol, 90 mM TrisCl, 1 mM EDTA, 0.004% bromphenol blue, 10 mM Na 3 VO 4 , pH 6.8, with the same protease inhibitors as in the lysis buffer).
Statistical Analysis-For each experiment, unpaired two-tailed t tests were used to compare the mean values for the wild-type and the IRS-3-null mice. Differences between means were taken to be not significant for p values greater than 0.05. Data displayed in the text are expressed as mean Ϯ S.E. Within the same period, the former produced 252 mice in 35 litters, while the latter yielded 197 mice in 32 litters; the average litter sizes were 7 and 6, respectively. These breeding data indicate that the lack of IRS-3 does not affect fertility or embryonic viability.

Growth and Development of IRS-3-null
To assess the effects of the absence of IRS-3 on growth, groups of wild-type and IRS-3-null mice were weighed at intervals between the ages of 3 and 30 weeks (Fig. 2). Within this time period, the growth curves for the wild-type and IRS-3-null female mice overlapped. For the males, the wild-type mice were slightly heavier than their IRS-3-null counterparts at 10 -30 weeks, but the difference is not statistically significant. The weights of the major organs of the mice (brain, lung, liver, heart, kidney, spleen, pancreas, quadriceps muscle, thymus, interscapular brown fat, epididymal fat), measured for male mice at 10 months of age, were the same for the wild-type and IRS-3-null mice (data not shown). Moreover, histological examination of these tissues did not reveal any differences in morphology between the wild-type and null mice. Thus, we conclude that IRS-3-null mice show normal growth and development.
Blood Glucose and Plasma Insulin Levels in IRS-3-null Mice-To determine whether glucose homeostasis was affected by IRS-3 deficiency, the concentrations of blood glucose and plasma insulin in the fed and fasted state were compared between wild-type and IRS-3-null mice. For both male and female mice of 9 -13 weeks of age, the genotypes showed no significant difference in the fed and fasted blood glucose levels or in the fed plasma insulin levels (Table I). In the case of the fasted plasma insulin level, the female wild-type and null mice showed no significant difference, but the male IRS-3-null mice showed a level 60% that of the wild-type (p ϭ 0.009). These measurements were repeated on the same groups of male and female mice at 6 months of age, in order to determine whether any effects of IRS-3 deficiency develop with age. For both sexes, there were no significant differences between the wild-type and IRS-3-null mice in any of the four parameters, including the fasted plasma insulin concentration in the males. In order to test whether IRS-3 deficiency had an effect in a different genetic background, the same measurements were carried out on mice with a 129 ϫ BALB/c background. The mice were 6 months of age; for each sex, 6 wild-type and 6 IRS-3-null mice were examined. For both males and females, no significant differences were observed between genotypes for each of the four parameters (data not shown). In conclusion, with the possible exception of insulin in fasted young male mice, levels of blood glucose and plasma insulin in mice were not altered by IRS-3 deficiency.
Glucose Tolerance Tests-To characterize glucose homeostasis in the absence of IRS-3 further, the blood glucose concentrations at various times after an oral glucose bolus were determined for male and female mice of 10 -13 weeks of age (Fig.  3). For the males, there was no difference in blood glucose levels between wild-type and IRS-3-null mice. In the case of the females, the IRS-3-null mice showed slightly greater blood glucose levels than the wild-type mice at 10 and 20 min; however, the differences were not statistically significant. In order to determine if IRS-3 deficiency affects glucose tolerance at an older age, glucose tolerance tests were also performed on male and female mice at 6 months of age. For each sex, 5 wild-type and 5 IRS-3-null mice were examined. With both sexes, there was no difference in the response to the oral glucose load (data

FIG. 2. Growth curves for wild-type and IRS-3-null mice.
Weights were determined for male and female wild-type (q) and IRS-3-null (E) mice at intervals. For all the groups, each time point presents the weights from 7-8 mice expressed as mean Ϯ S.E. not shown). Among the 5 older males of each genotype, there was one that had developed obesity (weight greater than 34 g; see Fig. 2); these two mice exhibited higher and more prolonged elevation of the blood glucose level compared with the nonobese mice. The glucose tolerance test was also carried out with 129 ϫ BALB/c mice of 6 months of age. Six wild-type and 6 IRS-3-null mice of each sex were examined. For both sexes, no difference between the genotypes was observed (data not shown). Among the 6 older 129 ϫ BALB/c males of each genotype, there was a single obese one. Unlike the obese males on the 129 ϫ C57BL/6 background, these mice did not exhibit an impaired glucose tolerance test.
Glucose Transport in Isolated Adipocytes-To determine whether the lack of IRS-3 affected glucose transport, adipocytes from wild-type and IRS-3-null mice were compared for their capacity to transport D-glucose. For adipocytes derived from male (epididymal) and female (parametrial) fat depots, the values for basal and maximal insulin-stimulated glucose transport were not significantly different between genotypes, although the average values for insulin-stimulated transport in the IRS-3-null mice were slightly lower (Fig. 4). The values for the transport rate in Fig. 4 are expressed per milligram of lipid. The lipid content per adipocyte was the same for the wild-type and IRS-3-null mice of each sex (Fig. 4, legend). In addition, the amount of the insulin-responsive glucose transporter GLUT4 per mg of lipid, as detected by immunoblotting, was the same for the wild-type and IRS-3-null mice of each sex (data not shown). Finally, the amount of epididymal or parametrial adipose tissue (grams dissected per mouse) was also the same (data not shown). Thus, the transport rates in the wild-type and IRS-3-null adipocytes are also similar when expressed per cell, per GLUT4, or per total adipose tissue.
The insulin dose that elicited a half-maximal response (ED 50 ) was determined from three or four independent experiments in which glucose uptake was measured at 0.03, 0.1, 0.3, and 10 nM insulin for epididymal adipocytes and 0.01, 0.03, 0.1, and 10 nM insulin for parametrial adipocytes. The ED 50 values were 0.064 Ϯ 0.017, 0.032 Ϯ 0.009, 0.034 Ϯ 0.011, and 0.029 Ϯ 0.011 nM for adipocytes from wild-type and IRS-3-null male, and wild-type and IRS-3-null female mice, respectively. The small differences between EC 50 values are not significant. Thus, the lack of IRS-3 in adipocytes had no significant effect on glucose transport, either in terms of the values for basal and maximal insulin-stimulated transport or the dose of insulin eliciting half-maximal stimulation.
Insulin-elicited Tyrosine Phosphorylation in Adipocytes-SDS lysates of basal and insulin-stimulated adipocytes from wild-type and IRS-3-null mice were immunoblotted with antibodies against Tyr(P) (Fig. 5). In adipocytes from wild-type mice insulin stimulated the tyrosine phosphorylation of proteins at 160 -190, 90, and 60 kDa. These correspond to IRS-1/2, the ␤ subunit of the insulin receptor, and IRS-3, respectively. Adipocytes from the IRS-3-null mice showed similar insulinelicited tyrosine phosphorylation, with the exception that the 60-kDa band was absent. The intensities of the Tyr(P) signals from IRS-1/2 and the insulin receptor ␤ subunit were the same for adipocytes from the wild-type and null mice. These results thus confirm the absence of IRS-3 protein in the IRS-3-null mice, and in addition show that the absence of IRS-3 did not result in any detectable compensatory increase in the tyrosine phosphorylation of IRS-1/2 or the insulin receptor.
In other experiments we have immunoblotted SDS lysates of basal and insulin-treated adipocytes from wild-type and IRS-3-null mice for Tyr(P) with a different antibody, the monoclonal 4G10, rather than the monoclonal conjugate RC20:HRPO pre-  (7) FIG. 3. Glucose tolerance test with fasted wild-type and IRS-3-null mice. The test was performed on male and female wild-type (q) and IRS-3-null mice (E) at the age of 10 -13 weeks. Data are from 7-9 mice in each group and are expressed as mean Ϯ S.E.

FIG. 4. Glucose transport in isolated adipocytes from wildtype and IRS-3-null mice. Uptake of [U-14 C]D-glucose by isolated adipocytes from male or female mice is shown for the basal (open bar)
and maximal insulin-stimulated state (10 nM insulin) (closed bar). Each bar represents data from five to six independent experiments expressed as mean Ϯ S.E. The respective -fold stimulation by insulin was 3.4 Ϯ 0.4, 2.7 Ϯ 0.4, 5.4 Ϯ 0.9, and 5.2 Ϯ 0.4 for adipocytes from wild-type and IRS-3-null male, and wild-type and IRS-3-null female mice. In three or four of these experiments, the cell number and lipid content of adipocytes from wild-type and null male and wild-type and null female mice were measured, and the values were 0.207 Ϯ 0.058, 0.216 Ϯ 0.005, 0.120 Ϯ 0.018, and 0.107 Ϯ 0.010 g/cell, respectively. sented in Fig. 5. The results were similar to those shown in Fig.  5, with the exception that the signal from the phosphorylated ␤ subunit of the insulin receptor was relatively more intense and the background in the region of IRS-1/2 was stronger (data not shown). We have also measured the relative amounts of the IRS-1 and IRS-2 proteins by immunoblotting the SDS lysates for these; there was no difference in amounts between wildtype and IRS-3-null adipocytes (data not shown). We attempted to detect the IRS-3 protein in the SDS lysates of wild-type adipocytes by immunoblotting with affinity-purified antibodies against the C-terminal portion of mouse IRS-3, which are described in Ref. 7. Unfortunately, IRS-3 was not detected by this method, probably because of its relatively low abundance and/or the relative lack of sensitivity of the antibodies.
Association of the IRSs with PI 3-Kinase and SHP-2 in Adipocytes-It has previously been found that insulin treatment of adipocytes causes the association of PI 3-kinase with IRS-1, -2, and -3 and the association of SHP-2 with IRS-1 and -3 (7,16,17). It was therefore possible that an increased association of IRS-1 and/or -2 with these enzymes might occur as a compensation for IRS-3 deficiency. To examine this possibility, PI 3-kinase and SHP-2 were immunoprecipitated from lysates of basal and insulin-treated wild-type and IRS-3-null adipocytes and immunoblotted for associated IRSs with antibodies against Tyr(P) (Fig. 6). In the case of the PI 3-kinase immunoprecipitates, the intensities of the Tyr(P) signal from IRS-1/2 in the samples from insulin-treated wild-type and null adipocytes were similar. This finding suggests that the absence of IRS-3 did not result in significant enhanced association of IRS-1/2 with PI 3-kinase. As expected, no Tyr(P) signal from IRS-3 was present in the PI 3-kinase immunoprecipitate from insulintreated IRS-3-null adipocytes. In the case of the SHP-2 immunoprecipitates, in agreement with our previous study (7), IRS-3 was the primary IRS associated with SHP-2 in insulin-treated wild-type adipocytes. The absence of IRS-3 in the null adipocytes did not lead to increased binding of other Tyr(P) proteins to this phosphatase. The SHP-2 immunoprecipitates from basal and insulin-treated cells contained a Tyr(P) protein of approximately 120 kDa, the amount of which was unaffected by insulin treatment or IRS-3 absence. This is most likely the recently described SHPS-1, a tyrosine-phosphorylated 120-kDa protein that binds to the SH2 domains of SHP-2 (18). DISCUSSION Earlier studies presented data to suggest that the absence of IRS-3 might affect growth and development or glucose transport in adipocytes and glucose homeostasis. In the case of growth and development, it has been found that IRS-3 mRNA is expressed at day 7 of embryonic life, a time at which neither IRS-1 or IRS-4 mRNA are expressed (8,19). The expression of IRS-2 mRNA during embryonic growth has not been determined. In the case of glucose transport and glucose homeostasis, a previous study proposed that approximately half of the insulin stimulation of glucose transport in adipocytes was due to signaling through IRS-3 (20). This proposal was based on the findings that for mouse adipocytes lacking IRS-1, insulin stimulates glucose transport only approximately half as well as in wild-type adipocytes, and more IRS-3 than IRS-2 immunoprecipitates with PI 3-kinase, as assessed by Tyr(P) immunoblotting. Moreover, it has recently been reported that reduction of the GLUT4 content specifically in adipose tissue in transgenic mice results in impaired glucose tolerance (21). Thus, reduction in insulin-stimulated glucose transport in adipose tissue might be expected to have the same effect.
Despite these suggested roles for IRS-3, the IRS-3-null mice were normal with regard to growth and development. Moreover, glucose homeostasis, as measured by their fed and fasted blood glucose and plasma insulin levels and by the oral glucose tolerance test, was not impaired. Finally, insulin stimulation of glucose transport in adipocytes was not significantly affected.
In contrast, transgenic mice deficient in IRS-1 or IRS-2 have unique phenotypes (3)(4)(5). Mice lacking IRS-1 are 30 -50% smaller than wild-type mice, but show no other developmental abnormalities. They are mildly insulin-resistant; the blood glucose concentration is normal, but the insulin level is approximately twice the normal value. Also, their blood glucose level does not fall as markedly upon an intraperitoneal injection of insulin. Mice deficient in IRS-2 are 10% smaller than normal. They progressively develop severe hyperglycemia at an early  6. Association of PI 3-kinase and SHP-2 with IRSs in adipocytes. Basal (Ϫ) and insulin-stimulated (ϩ) adipocytes from wildtype and IRS-3-null mice were immunoprecipitated with antibodies against the 85-kDa subunit of PI 3-kinase (PI3K) or SHP-2 and immunoblotted for Tyr(P) with monoclonal 4G10. Loads 1, 1/2, and 1/4 were immunoprecipitates derived from 250, 125, and 62.5 g of lysate protein, respectively. Molecular size markers in kDa are shown on the right. Repetition of these experiments with adipocytes from mice of the opposite sex gave similar results. Control immunoprecipitations with irrelevant rabbit immunoglobulin did not show any bands upon immunoblotting. Immunoblotting of samples of the lysates taken before and after immunoprecipitation with each immunoprecipitating antibody showed that the efficiencies for immunoprecipitation of the 85-kDa subunit and SHP-2 were approximately 90%. age due to a combination of insulin resistance of the liver and muscle and a failure of the ␤ cells of the pancreas to proliferate, which leads to insufficient compensatory insulin secretion. These phenotypes show that IRS-1 and -2 are essential for growth and glucose homeostasis and that each has unique functions, which cannot be fully substituted by another IRS.
The lack of a phenotype for the IRS-3-null mice and the occurrence of ones for the IRS-1 and -2 null mice may partially be explained in terms of the tissue distributions and relative amounts of these three IRSs. As assessed by Northern blotting, IRS-1 and -2 are expressed in all the major mouse tissues (liver, skeletal muscle, heart, brain, kidney, lung, spleen, testis), whereas among these tissues, IRS-3 mRNA is undetectable in skeletal muscle, a major target for insulin action, as well as brain, spleen, and testis (8,22). Moreover, even in those tissues where IRS-3 mRNA is expressed, it is likely that the amount of IRS-3 protein is considerably less than the amounts of IRS-1 and -2. For example, in liver, a major target for insulin action, insulin-stimulated tyrosine phosphorylation of IRS-3 is not detected by immunoprecipitation with antibodies against Tyr(P) or PI 3-kinase followed by immunoblotting with antibodies against Tyr(P), whereas IRS-1 and IRS-2 are prominent insulin-elicited Tyr(P) proteins (23,24). In fact, despite the presence of its mRNA in liver, heart, kidney, and lung, the only tissue in which the IRS-3 protein has been definitively detected to date is white adipose tissue (7).
We examined whether the absence of IRS-3 in adipocytes led to compensatory changes in other signaling components. In addition to IRS-3, adipocytes express IRS-1 and IRS-2, but not IRS-4 (17,19,24). There was no change in the amounts of IRS-1 or -2 and no detectable increase in the extent of their combined tyrosine phosphorylation in response to insulin. Compensation for the absence of IRS-3 could also occur by increased association of IRS-1 or 2 with PI 3-kinase and SHP-2, the SH2 domaincontaining proteins that normally associate with the Tyr(P) form of IRS-3. However, the absence of any significant increase in the association of the Tyr(P) forms of IRS-1/2 with either SH2 domain protein suggests that this is not the case.
It is important to note that our present data do not provide information about the effect of IRS-3 deficiency on the overall insulin-stimulated activation of PI 3-kinase and production of PI 3,4,5-trisphosphate. In future it will be of interest to compare insulin-stimulated production of PI 3,4,5-trisphosphate in wild-type and IRS-3-null adipocytes. A recent study examined the distribution of IRS-1, -2, and -3 in subcellular fractions of rat adipocytes, as well as their insulin-stimulated association with, and activation of, PI 3-kinase (17). Unfortunately, the analysis did not measure the relative amounts and activity of PI 3-kinase associated with each IRS, and so the contribution of the complex between IRS-3 and PI 3-kinase to the overall activation of PI 3-kinase in adipocytes is not known. This study did show that IRS-3 and its complex with PI 3-kinase are primarily located in the plasma membrane, whereas IRS-1 and -2 and their complexes with PI 3-kinase are primarily in low density microsomes. Thus, it remains possible that IRS-3, by virtue of activation of PI 3-kinase in a unique location, has some unique, as yet unidentified function in adipocytes. There is considerable evidence that insulin stimulation of glucose transport in adipocytes requires activation of PI 3-kinase through its association with the IRSs, although there are also findings that conflict with this conclusion (reviewed in Ref. 25). Our data show that association of PI 3-kinase with IRS-3 is not essential for this insulin effect.
In adipocytes, IRS-3 appears to be the predominant IRS that associates with the tyrosine phosphatase SHP-2 upon insulin treatment (Ref. 7 and Fig. 6). The interaction of the SH2 domains of SHP-2 with Tyr(P) motifs activates this phosphatase (26). Consequently, in the IRS-3-null adipocytes, insulinstimulated activation of SHP-2 is predicted to be substantially reduced. In the future it may be possible to test this prediction by measurements of SHP-2 activity. Results in earlier studies suggested that the association of SHP-2 with IRSs might be required for full activation of the mitogen-activated protein kinase cascade by insulin (27). However, a recent study has shown conclusively that this is not the case for a cell line that contains only IRS-1 (28). Thus, the role of the SHP-2/IRS association in insulin signaling is currently unknown, and the consequences of its reduction in the IRS-3-null adipocytes cannot be predicted. However, these adipocytes may provide a system in which to examine this question.
In conclusion, at this point we have not found any abnormalities in IRS-3-null mice, although insulin signaling through IRS-3 is absent in their adipocytes. It is possible that in wildtype mice IRS-3 plays significant roles, and that in the null mice other IRSs function in its place. Similarly, the phenotypes of the mice lacking IRS-1 and IRS-2 may not be more severe because IRS-3 functions in part in their place. Such functional redundancy has been found for other gene families, such as the Src kinases and the retinoic acid receptors (29,30). It should be possible to determine whether this is the case for the IRSs by generating and characterizing mice lacking two IRSs. The fact that each IRS gene is on a different chromosome (19) will facilitate this approach.