Positive and Negative Roles of p85α and p85β Regulatory Subunits of Phosphoinositide 3-Kinase in Insulin Signaling*

Class IA phosphoinositide (PI) 3-kinase is composed of a p110 catalytic subunit and a p85 regulatory subunit and plays a pivotal role in insulin signaling. To explore the physiological roles of two major regulatory isoforms, p85α and p85β, we have established brown adipose cell lines with disruption of the Pik3r1 or Pik3r2 gene. Pik3r1-/- (p85α-/-) cells show a 70% reduction of p85 protein and a parallel reduction of p110. These cells have a 50% decrease in PI 3-kinase activity and a 30% decrease in Akt activity, leading to decreased insulin-induced glucose uptake and anti-apoptosis. Pik3r2-/- (p85β-/-) cells show a 25% reduction of p85 protein but normal levels of p85-p110 and PI 3-kinase activity, supporting the fact that p85 is more abundant than p110 in wild type. p85β-/- cells, however, exhibit significantly increased insulin-induced Akt activation, leading to increased anti-apoptosis. Reconstitution experiments suggest that the discrepancy between PI 3-kinase activity and Akt activity is at least in part due to the p85-dependent negative regulation of downstream signaling of PI 3-kinase. Indeed, both p85α-/- cells and p85β-/- cells exhibit significantly increased insulin-induced glycogen synthase activation. p85α-/- cells show decreased insulin-stimulated Jun N-terminal kinase activity, which is restored by expression of p85α, p85β, or a p85 mutant that does not bind to p110, indicating the existence of p85-dependent, but PI 3-kinase-independent, signaling pathway. Furthermore, a reduction of p85β specifically increases insulin receptor substrate-2 phosphorylation. Thus, p85α and p85β modulate PI 3-kinase-dependent signaling by multiple mechanisms and transmit signals independent of PI 3-kinase activation.

A large body of evidence has shown that phosphoinositide (PI) 1 3-kinase plays a pivotal role in metabolic and mitogenic actions regulated by insulin (1,2). Insulin promotes interaction of insulin receptor substrate (IRS) proteins and Class IA PI 3-kinase, which is composed of a 110-kDa catalytic subunit (p110) and a regulatory subunit (1). This process increases in the catalytic activity of the p110 subunit and also recruits the enzyme to an area enriched by its lipid substrates. Phosphatidylinositol 3,4,5-trisphosphate (PIP 3 ), a major product of Class IA PI 3-kinase, is essential for insulin action and acts as a second messenger activating downstream serine/threonine kinases, such as Akt/protein kinase B and the atypical protein kinase Cs (3).
Three genes, Pik3r1, Pik3r2, and Pik3r3 encode different isoforms of the regulatory subunits, termed p85␣ and its spliced variants (p55␣/AS53 and p50␣ and its insert forms) (4 -7), p85␤ (4), and p55 PIK (8), respectively. The products of the Pik3r1 and Pik3r2 genes are the predominant isoforms in insulin-sensitive tissues. Using knockout mice, we have shown that p85␣ and its splice variants, p55␣ and p50␣, represent 70 -80% of the total regulatory subunits in these tissues, and p85␤ represents most of the rest (9,10). Although it was reasonable to expect that insulin sensitivity would be impaired if these regulatory subunits were deleted, surprisingly, mice lacking p85␣ alone (11) and mice heterozygous for a null mutation of the Pik3r1 gene deleting p85␣, p55␣, and p50␣ (12) show improved sensitivity to insulin. Furthermore, the latter protects mice carrying mutations of the insulin receptor and insulin receptor substrate-1 (IRS-1) from development of diabetes (12). Moreover, homozygous Pik3r1 knockout mice show hypoglycemia with hypoinsulinemia, suggesting that at least some insulin actions are enhanced in these mice, although they died within a week of birth due to the importance of p85␣ and its splice variants in growth and metabolism (13,14). p85␤ knockout mice with only a 20 -30% reduction in total regulatory subunits also exhibit modestly increased insulin sensitivity with no change in insulin-induced PI 3-kinase activity or the levels of p85⅐p110 complexes (10). These data suggest that the regulatory subunits of PI 3-kinase have negative effects on insulin actions, and a reduction of the regulatory subunit enhances insulin signaling.
Using tissues and fibroblasts derived from the Pik3r1 knockout mice, we have shown that the regulatory subunits of PI 3-kinase are more abundant than the catalytic subunits and that Ͼ30% of the regulatory subunits exist as monomers (9,12). These monomeric regulatory subunits interfere with the binding of p110-p85 heterodimers to phosphorylated IRS proteins (9). We have also shown that in the Pik3r1 heterozygous knockout fibroblasts there is enhanced insulin-like growth factor (IGF)-1 action and a significant increase in PIP 3 level despite no increase in PI 3-kinase activity (9), suggesting that the regulatory subunit can also down-regulate IGF-1/insulin signaling by promoting clearance of PIP 3 . The latter mechanism also suggests the existence of unexplored function of the regulatory subunit as a signal transducer independent of its regulatory function of PI 3-kinase activity.
In this study, to better understand the multiple roles of p85␣ and p85␤ in insulin action, including effects mediated independent of PI 3-kinase activity and differences between these homologous subunit isoforms, we have established brown preadipocyte cell lines derived from mice with a knockout of either the Pik3r1 gene or the Pik3r2 gene and characterized insulin signaling after the cells were differentiated into adipocytes in culture.

EXPERIMENTAL PROCEDURES
Cells and Cell Culture-Brown preadipocyte cell lines were established and cultured, as described previously with minor modifications (15). Briefly, brown adipocytes and their precursor cells were isolated from newborn mice (wild type, Pik3r1 ϩ/Ϫ , Pik3r1 Ϫ/Ϫ , and PiK3r2 Ϫ/Ϫ mice) by collagenase digestion and immortalized by infection with the retroviral vector pBabe, encoding SV40 T antigen (kindly provided by J. De-Caprio, Dana Farber Cancer Institute, Boston, MA) using puromycin (1 mg/ml) selection. The immortalized preadipocytes were grown to confluence in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum. Adipocyte differentiation was induced by treating confluent cells for 2 days in induction medium (DMEM supplemented with 10% fetal bovine serum, 10 nM insulin, 1 nM T3, 0.5 mM isobutylmethylxanthine, 0.5 mM dexamethasone, and 0.125 mM indomethacin). After this period cells were cultured in differentiation medium (DMEM supplemented with 10% fetal bovine serum, 10 nM insulin, and 1 nM T3) for 2 days, then the medium was switched to DMEM with 10% fetal bovine serum for 2 days. At this point cells were morphologically and functionally differentiated into brown adipocytes. Cells were subjected to assays after serum starvation for 24 h. Three independent cell lines from different animals of each genotype were used and gave similar results.
Northern Blot Analysis-Total RNA was isolated from differentiated adipocytes using an RNeasy kit (Qiagen). Thirty micrograms of the total RNA was subjected to Northern blot analysis and probed with the indicated cDNA fragments, as described previously (16,17).
Affinity Purification of Regulatory Subunits of PI 3-Kinase Using a Phosphopeptide Column-Affinity purification of the regulatory subunits were performed as previously described (9). Briefly, 3 mg of tyrosine-phosphorylated peptide surrounding Tyr-608 of IRS-1 protein (Biomol) was immobilized on 2 ml of Affi-Gel 10 beads (Bio-Rad) and packed in a plastic column. Ten mg of the lysates of each genotype of cells were applied to the column and extensively washed with buffer A with 500 mM NaCl. The proteins bound to pYMXM (p is phosphorylated Tyr) peptide were eluted with a buffer composed of 2.5 M glycine, pH 4.5, and 2 M NaCl and dialyzed with phosphate-buffered saline containing 2% glycerol. The purified proteins were subjected to SDS-PAGE and visualized by silver staining.
In Vitro Kinase Assays-For PI 3-kinase, the immunoprecipitates using the indicated antibody were washed three times with buffer A and twice with PI 3-kinase reaction buffer (20 mM Tris-HCl, pH 7.4, 100 mM NaCl, and 0.5 mM EGTA) and suspended in 50 l of PI 3-kinase reaction buffer containing 0.1 mg/ml PI (Avanti Polar Lipids). The reactions were performed, and the phosphorylated lipids were separated by thin layer chromatography (9). For Akt, cells were lysed with buffer A as described above, and the lysates were subjected to immunoprecipitation with ␣Akt followed by kinase assay using Crosstide (GSK3 peptide) as a substrate (19). For p70 S6K and p90 RSK kinase assays, cells were lysed with buffer A and immunoprecipitated with ␣p70 S6K or ␣p90 RSK , and the kinase assays were performed using S6 peptide (8-mer peptide from the C-terminal sequence of ribosomal S6 protein, Santa Cruz Biotechnology) (9). Generation and Infection of Adenoviruses-The recombinant adenoviruses, Adex1CAp85␣-HA, Adex1CAp55␣-HA, Adex1CAp50␣-HA, Adex1CAp85␤-HA, and Adex1CADNRas, were generated as described previously (7,20). The recombinant adenovirus encoding a mutant p85␣ (Adex1CA⌬p85) that lacks the p110-binding site was kindly provided by Masato Kasuga (Kobe University) (21). The control adenovirus, Adex1CALacZ, and the cosmid cassette were kindly provided by Izumi Saito (University of Tokyo). For infection, the differentiated cells were cultured in media containing the adenoviruses for 1 h at 37°C, and DMEM supplemented with serum was added and cultured for 24 h. Cells were subjected to experiments after 24 h of serum deprivation. The adenoviruses were applied at the indicated m.o.i. (multiplicity of infection) in each experiment.
2-Deoxyglucose Uptake Assays-2-Deoxyglucose uptake assays were performed using differentiated adipocytes grown in 12-well plates (20). Before initiating glucose uptake assay, cells were washed 3 times with phosphate-buffered saline and incubated in 1 ml of serum-free DMEM for 3 h at 37°C. Cells were then washed once with Krebs-Ringer phosphate/HEPES buffer (KRHB) containing 130 mM NaCl 2 , 5 mM KCl 2 , 1.3 mM CaCl 2 , 1.3 mM MgSO 4 , 10 mM Na 2 HPO 4 , and 25 mM HEPES, pH 7.4, and incubated in 1 ml of KRHB containing 0.1% bovine serum albumin without or with 10 nM insulin for 15 min at 37°C. Glucose uptake was initiated by the addition of 2-deoxy-D-[2,6-3 H]glucose to a final concentration of 0.5 Ci for 5 min at 37°C and termi-nated by 2 washes with ice-cold Krebs-Ringer phosphate/HEPES buffer. Cells were solubilized with 0.4 ml of 0.1% SDS and counted by scintillation counter. Nonspecific glucose uptake was measured in the presence of 20 M cytochalasin B and subtracted from each assay to obtain specific uptake.
Glycogen Synthase Assays-Glycogen synthase activity was measured as previously described (20). After serum starvation cells were washed twice and incubated with Krebs-Ringer phosphate/HEPES buffer without or with 100 nM insulin for 20 min. Cells were lysed with lysis buffer containing 25 mM Tris-HCl, pH 7.0, 30% glycerol, 10 mM EDTA, 100 mM KF, 1 mM phenylmethylsulfonyl fluoride. The lysates were centrifuged, and 30 l of the supernatant was added to 60 l of the assay mixture containing 50 mM Tris-HCl, pH 7.4, 25 mM NaF, 20 mM EDTA, 1 mg/ml glycogen, and 0.1 Ci of UDP-[ 14 C]glucose with 0.25 or 10 mM glucose 6-phosphate. After incubation at 30°C for 30 min, aliquots were spotted on Whatman No. 3MM paper and washed four times with ice-cold 70% ethanol, and radioactivity was counted with a scintillation counter.
Apoptosis Assay-Apoptotic rate was determined using an enzymelinked immunosorbent assay to determine the amount of nucleosomes (Roche Applied Science) (9). An equal number of cells were plated in 24-well culture plates and differentiated to adipocytes. At that time the cells were washed with phosphate-buffered saline and treated with or without10 nM insulin for 8 h. The cells (including floating FIG. 2. Changes in Class IA PI 3-kinase complexes by deletion of the p85␣ isoform or the p85␤ isoform. a, complete deletion of the each isoform in cells with the null mutation of the gene. The immunoprecipitates (IP) with the specific antibody for p85␣ (left panels) or p85␤ (right panels) from the differentiated adipocytes were subjected to immunoblot (IB) analysis probed with ␣p85pan (top panels) or PI 3-kinase assay (bottom panels). Results are representative of at least three independent experiments. PI(3)P, phosphatidylinositol-3 phosphate; Ins., insulin. b, expression levels of the regulatory subunits of PI 3-kinase in cells of each genotype. Cell lysates were subjected to immunoblot analysis probed with ␣p85pan. c, affinity purification of the regulatory subunits of PI 3-kinase from cells of each genotype using a phospho-peptide column. The cell lysates were applied to the column coupled with the phosphorylated p85 binding domain peptide of IRS-1 as described under "Experimental Procedures." The collected proteins were visualized by silver staining (top panel). In the bottom panel, each bar represents the mean Ϯ S.E. of the eluate from the results of three independent experiments. The value is expressed as a ratio to the total p85 protein level in wildtype cells (*, p Ͻ 0.05 wild type versus p85␣ Ϫ/Ϫ ; **; p Ͻ 0.05 wild type versus p85␤ Ϫ/Ϫ ). d, interaction of the regulatory subunit with p110␣ and p110␤ in each genotype of cells. The immunoprecipitates with ␣p110␣ (left panels) or ␣p110␤ (right panels) were subjected to immunoblot analysis probed with the same antibody (top panels) or ␣p85pan (middle panels) or PI 3-kinase assay (bottom panels). Results are representative of at least three independent experiments.
FIG. 3. Effects of disruption of the Pik3r1 gene or the Pik3r2 gene on insulin-induced tyrosine phosphorylation and PI 3-kinase activation. a, insulin-induced tyrosine phosphorylation of IRS proteins in each genotype of the cells. After insulin (Ins.) stimulation for 5 min, cell lysates were subjected to immunoprecipitation (IP) with ␣IRS-1 (left panels) or ␣IRS-2 (right panels) followed by immunoblot (IB) analysis probed cells) were collected and applied to the enzyme-linked immunosorbent assay (9).

Effects of Disruption of the Pik3r1 and Pik3r2 Genes on
Differentiation of Adipocytes-To assess roles of the regulatory subunits of PI 3-kinase, we established brown preadipocyte cell lines from Pik3r1 Ϫ/Ϫ (p85␣ Ϫ/Ϫ ), Pik3r1 ϩ/Ϫ (p85␣ ϩ/Ϫ ), and Pik3r2 Ϫ/Ϫ (p85␤ Ϫ/Ϫ ) neonatal mice as well as the wild-type control mice. PI 3-kinase activity has been shown to be required for adipocyte differentiation (22,23). However, all genotypes of cells were able to be differentiated into mature adipocytes with accumulation of multiocular lipid droplets, as estimated by oil-red O staining (Fig. 1a), although there was a slight delay in differentiation with the p85␣ Ϫ/Ϫ cells when cultured in induction media. Consistent with the results of oil red O staining, when differentiation was complete, there was no significant difference in the adipogenic differentiation markers, such as peroxisome proliferator-activated receptor ␥ (PPAR␥), C/EBP␣, and GLUT4, among all genotypes of cells, as estimated by Northern blot analysis and immunoblot analysis ( Fig. 1, b and c).
Effects of Disruption of the Pik3r1 and Pik3r2 Genes on the Molecular Balance in Class IA PI 3-Kinase-The molecular balance between the catalytic and regulatory subunits is important in the regulation of PI 3-kinase-dependent signaling (9,20,24); thus, we assessed the changes in the regulatory and catalytic subunits in p85␣ Ϫ/Ϫ and p85␤ Ϫ/Ϫ cells. Using antibodies specific for p85␣ or p85␤, there was complete deletion of the p85␣ protein and p85␣-associated PI 3-kinase in p85␣ Ϫ/Ϫ cells (Fig. 2a, left), whereas these cells showed modest increases in p85␤ and its associated PI 3-kinase activity (Fig. 2a, right). On the other hand, in p85␤ Ϫ/Ϫ cells, the p85␤ protein was completely abolished (Fig. 2a, right), with no prominent alteration in p85␣ protein or its associated PI 3-kinase activity (Fig. 2a, left). Using ␣p85pan antibody, which recognizes all isoforms of the Pik3r1 gene products equally and also detects p85␤ and p55 PIK to a lesser extent, p85 protein was decreased by more than 90% in p85␣ Ϫ/Ϫ cells, whereas only a very minor reduction was detected in p85␤ Ϫ/Ϫ cells (Fig. 2b). Shorter isoforms were hardly detected in all genotypes of cells (Fig. 2b).
To more precisely estimate the changes in the regulatory subunits by these knockouts, we purified the SH2-containing proteins interacting with pYMXM motif by an affinity column containing the phospho-YMPM peptide corresponding to a region surrounding Tyr-608 of IRS-1 (9). This revealed that there was a ϳ70% reduction of total p85 protein in p85␣ Ϫ/Ϫ cells; the remaining p85 protein represented the p85␤ isoform (Fig. 2c). Using the same technique a ϳ25% reduction occurred in p85␤ Ϫ/Ϫ cells (Fig. 2c). The difference between these two estimates of p85␤ (25 and 30%) is due to a modest increase in p85␤ protein in p85␣ Ϫ/Ϫ cells (Fig. 2a, right).
We have previously shown that a major reduction in the p85␣ regulatory subunit causes the secondary decrease in the p110 catalytic subunits (9,14). This is due to the fact that the p110 subunit monomer is unstable and easily degraded without the association of the regulatory subunit (25). Consistent with this, in p85␣ Ϫ/Ϫ cells both p110␣ and p110␤ were decreased by 80%, and this was accompanied by a comparable reduction in PI 3-kinase activities associated with both isoforms (Fig. 2d).
We have also previously shown that more than 30% of the p85 protein exists as a monomer (9,12). Thus, a 25% reduction in the p85 protein, which occurs in p85␤ Ϫ/Ϫ cells, would not be expected to affect the amount of p85⅐p110 heterodimer complex. Indeed, in p85␤ Ϫ/Ϫ cells, deletion of the p85␤ isoform did not appear to decrease in either p110 catalytic subunit proteins or the p85⅐p110 complexes, maintaining PI 3-kinase activity associated with the p110␣ or p110␤ isoform (Fig. 2d).
Effects of Disruption of the Pik3r1 and Pik3r2 Genes on Insulin-induced PI 3-Kinase Activation-Insulin activates Class IA PI 3-kinase-mediated signaling by promoting the interaction between phosphorylated IRS proteins, especially IRS-1 and IRS-2, and the p85⅐p110 complex (26). To estimate the role of each regulatory subunit of PI 3-kinase in insulin signaling it was important to assess the PI 3-kinase activity associated with each IRS protein as well as that associated with the total tyrosine-phosphorylated proteins.
As shown in Fig. 3a, there was no alteration in phosphorylation of IRS-1 or IRS-2 by deletion of p85␣. However, as expected, the p85 protein associated with IRS-1 and IRS-2 was dramatically decreased in p85␣ Ϫ/Ϫ cells, leading to a significant reduction in IRS-1-associated and IRS-2-associated PI 3-kinase activity (Fig. 3b). As a consequence, the p85 protein interacting with total tyrosine-phosphorylated proteins was also dramatically decreased with a significant reduction in PI 3-kinase activity associated with phosphotyrosine complexes (Fig. 3c, left and right).
On the other hand, in p85␤ Ϫ/Ϫ cells there was no reduction in p85 protein interacting with IRS-1 (Fig. 3a, left) and PI 3-kinase activity associated with IRS-1 (Fig. 3b, left), suggesting that despite a 25% decrease in the regulatory subunit, IRS-1⅐p85⅐p110 complex is preserved in these cells. Interestingly, phosphorylation of IRS-2 was prominently up-regulated in p85␤ Ϫ/Ϫ cells, although there was almost no increase in the p85 protein interacting with IRS-2 (Fig. 3a, right) and only a slight increase in IRS-2-associated PI 3-kinase activity (Fig. 3b, right). We had previously observed this in liver and muscle in vivo (10), suggesting that this is a generalized effect of p85␤ deletion. However, there was no obvious increase in interaction of IRS-2 with other SH2-containing proteins including Grb2 and SHP2 (data not shown). Nevertheless, total tyrosine phosphorylation in response to insulin was not increased in p85␤ Ϫ/Ϫ cells (Fig. 3c, left), presumably due to the relatively minor contribution of IRS-2 compared with IRS-1 in adipocytes (27). This led to a comparable amount of insulin-induced PI 3-kinase activity associated with phosphotyrosine complexes in p85␤ Ϫ/Ϫ cells (Fig. 3c, right).
Effects of Disruption of the Pik3r1 and Pik3r2 Genes on Downstream Signaling of PI 3-Kinase-One of the important components downstream of PI 3-kinase in insulin action is a Ser/Thr kinase, Akt/protein kinase B (3). Both phosphorylation and activity of Akt were significantly decreased in p85␣ Ϫ/Ϫ cells (Fig. 4a, left), although the reduction in Akt activity (30% compared with wild type) was smaller that that of insulininduced PI 3-kinase activity (50% reduction, Fig. 3c). On the other hand, Akt activity as well as its phosphorylation was significantly increased in p85␤ Ϫ/Ϫ cells (Fig. 4a) despite no increase in insulin-induced PI 3-kinase activity (Fig. 3c). By contrast, p70 S6 kinase (p70 S6K ) activity, which is also reguwith ␣PY (top panels), ␣p85pan (middle panels), or the same antibody (bottom panels). b, PI 3-knase activity associated with IRS protein. Cell lysates were subjected to immunoprecipitation with ␣IRS-1 (left panels) or ␣IRS-2 (right panels) followed by PI 3-kinase assay. Top panels show representative results, and in bottom panels, each bar represents the mean Ϯ S.E. of the relative PI 3-kinase activity calculated from the results of three independent experiments (*, p Ͻ 0.05 wild type versus p85␣ Ϫ/Ϫ ; **, p Ͻ 0.05 wild type versus p85␤ Ϫ/Ϫ ). c, total PI 3-kinase activity induced by insulin. Cell lysates were subjected to immunoprecipitation with ␣PY followed by immunoblot analysis probed with ␣PY (left top) or ␣p85pan (left bottom) or PI 3-kinase assay (right). In the right panel, each bar represents the mean Ϯ S.E. of the relative PI 3-kinase activity calculated from the results of three independent experiments (*, p Ͻ 0.05 wild type versus p85␣ Ϫ/Ϫ ). PI (3) (28), was unchanged in all genotypes of cells (Fig. 4b). These findings are consistent with those in fibroblasts with IGF-1 stimulation (9) and in tissues with in vivo insulin stimulation (10,12,14).
Based on its activation mechanism, Akt activity generally reflects the level of PIP 3 , which is determined by the balance between the activity of PI 3-kinase and lipid phosphatases, such as phosphatase and tensin homologue deleted on chromosome 10 (PTEN) (3). In previous studies we and others found a discrepancy between PI 3-kinase activity and levels of PIP 3 in cells lacking the p85␣ isoform alone knockout and in Pik3r1 heterozygous and homozygous knockouts (9,11). We have proposed two mechanisms to explain these findings. One is that under normal conditions the regulatory subunits of PI 3-kinase are more abundant than p110 catalytic subunits and the monomeric p85 inhibits the IRS proteins-mediated signal by competing with the p85-p110 heterodimer. This is also the case in this study, as evidenced by the fact that the 25% reduction in the regulatory subunit created by deletion of p85␤ did not affect p85⅐p110 complexes and insulin-induced PI 3-kinase activity. The other is that p85 may promote the clearance of PIP 3 independent of PI 3kinase activity.
To assess the latter reconstitution experiments were performed. Reconstitution of p85␣ Ϫ/Ϫ cells by adenovirus-mediated gene transfer (m.o.i. 500) to produce a level of p85␣ equal to 20 -30% that of the wild-type increased p110␣ and the p85⅐p110 complex to 35-40% that of the wild-type levels (Fig.  4c). This led to a partial restoration of PI 3-kinase activity and up-regulation of insulin-induced Akt activity, almost to the level of wild-type cells (Fig. 4c). Increasing the titer of the virus to m.o.i. 1000 and 2000 resulted in higher levels of expression of p85␣ protein, up to 70% that of the wild-type level, but produced no further increase in the amount of p110␣ or the p85⅐p110 complexes (Fig. 4c). This necessarily led to a marked increase in p85 monomer, but under these conditions, all of the p85 monomer and p85-p110 heterodimer were able to bind to phosphorylated IRS proteins. Indeed, insulin-induced PI 3-kinase activity was not decreased by increasing p85 (Fig. 4c). Despite the preserved PI 3-kinase activity, however, increased the p85␣ protein down-regulated Akt activity (Fig. 4c), indicating the existence of the mechanism of PIP 3 clearance dependent on p85 but independent of PI 3-kinase Effects of Disruption of the Pik3r1 and Pik3r2 Genes on Insulin-dependent Biological Responses-PI 3-kinase has been shown to regulate most of insulin biological responses including activation of glucose transport, glycogen synthesis, and anti-apoptosis (1, 3). Glucose uptake was modestly, but significantly, decreased (31%) in p85␣ Ϫ/Ϫ cells, paralleling the reductions in insulin-induced PI 3-kinase and Akt activities (Fig. 5a, left, compared with Figs. 3c and 4a). By contrast, despite the up-regulation of insulin-stimulated Akt activation in p85␤ Ϫ/Ϫ cells, glucose uptake was not increased (Fig. 5a, left). This was at least in part due to a 37% decrease in the GLUT4 protein in p85␤ Ϫ/Ϫ cells, although there was a modest increase (18%) in the GLUT1 protein (Fig. 5a, right). Surprisingly, despite the decrease in Akt activity in p85␣ Ϫ/Ϫ cells, both p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells showed significantly increased glycogen synthase activity (Fig. 5b, left, compared with Fig. 4a). Phosphorylation of GSK3␤ was comparable in all genotypes of cells (Fig. 5b,  right), suggesting that a relatively small amount of Akt activity is sufficient for full inhibition of GSK3 or that an alternative pathway of GSK3␤ phosphorylation exists. In adipocytes, the other pathway in regulation of glycogen synthase, protein phosphatase (PP) 1-dependent regulation, appears to be more dominant (19,29). Although the regulatory mechanism of PP1 in adipocytes still remains unclear, PP1 activity seems to be regulated in a PIP 3 -dependent fashion in 3T3L-1 adipocytes (30). Site 3a in GS (Ser-640 in muscle and Ser-641 in liver) is one of the key residues for regulation of GS activity controlled by phosphorylation and dephosphorylation (31). Indeed, phosphorylation of Site 3a was significantly decreased in both p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells (Fig. 5b, left) compared with wild type, consistent with the increased GS activity. Because GSK3 activity was indistinguishable in all genotype of cells, the decrease in phosphorylation state is likely to reflect PP1 activity, which must be up-regulated in p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells.
We also investigated insulin effects on apoptosis after serum depletion by measuring the nucleosomes using an enzymelinked immunosorbent assay. Both in the absence and the presence of insulin, apoptotic rate was significantly increased in p85␣ Ϫ/Ϫ cells, whereas the apoptotic rates under both conditions were significantly decreased in p85␤ Ϫ/Ϫ cells (Fig. 5c,  left). This is consistent with changes in Akt activity created by deletion of the regulatory subunits, since Akt plays an important role in anti-apoptosis (32). One of the mechanisms by which Akt inhibits apoptosis is via phosphorylation of proapoptotic protein Bad. Phosphorylation of Bad promotes its interaction with 14-3-3 and prevents this complex from translocating to the mitochondrial membrane, thus reducing apoptosis (32). Another Akt-dependent anti-apoptotic effect is inhibition of translocation of forkhead transcription factor Foxo1 to the nucleus by phosphorylation (33). Indeed, in parallel with Akt activity, both, 14-3-3 bound to Bad and phosphorylation of Foxo1 were prominently decreased in p85␣ Ϫ/Ϫ cells, whereas they were increased in p85␤ Ϫ/Ϫ cells (Fig. 5c, right).
Signaling Mediated by p85␣ and p85␤, Independent of Their Regulatory Role of PI 3-Kinase Activation-As shown in this and a previous study (9), clearance of PIP 3 in response to insulin/IGF-1 appears to correlate with the level of the p85 protein, but not PI 3-kinase activity, suggesting the existence of the signaling pathway mediated by p85 but independent of its regulation of PI 3-kinase. In the course of searching for such a signal regulated by p85, we have found that the kinase activities of two members of stress-activated MAP kinase family, JNK and p38 mitogen-activated protein kinase (p38 MAPK), were greatly attenuated in response to IGF-1 in embryonic fibroblasts by disruption of the Pik3r1 gene (data not shown). Thus, we assessed these kinases activities in p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells as well as p85␣ ϩ/Ϫ cells, where the p85 protein was decreased by 40%, whereas Akt activity was greatly up-regulated, and PI 3-kinase activity was unchanged (data not shown). In wild-type cells, insulin robustly activated JNK1 but not JNK2. Insulin-stimulated JNK1 activity was decreased by decreasing p85␣ isoform and to a lesser extent by  (Fig. 6a, left top). Insulin-stimulated p38 MAPK activity was also decreased, corresponding to a reduction in p85␣ isoform, but no obvious reduction was detected in p85␤ Ϫ/Ϫ cells (Fig. 6a, left middle). However, in all genotypes of cells both kinases were fully activated by anisomysin, indicating that activation machineries for JNK and p38 MAPK were intact despite the changes in the regulatory subunit of PI 3-kinase. Moreover, the p42 and p44 MAP kinases were equally activated in response to insulin in all genotypes of cells.
To confirm that JNK activity depended on the p85 protein rather than PI 3-kinase activity, we reconstituted p85␣ Ϫ/Ϫ cells with p85␣, p85␤, and a p85 mutant (⌬p85), which lacks the p110 binding domain using adenovirus-mediated gene transfer. All three p85 proteins were able to preserve JNK activity, whereas only wild-type p85␣ and p85␤ were able to up-regulate Akt phosphorylation (Fig. 6b, left). Thus, the p85 protein mediates a signal that activates JNK independent of activation of PI 3-kinase or Akt. PI 3-kinase activation mainly depends on the p85 C-terminal region, that includes the SH2 domains and the p110 binding region, whereas p85-mediated JNK activation seems to occur through its N-terminal region, presumably through one of the multiple potential protein-protein interaction motifs. To confirm the role of the N terminus of p85, we reconstituted p85␣ Ϫ/Ϫ cells with p85␣, p55␣, and p50␣, all of which shared the common C-terminal region but differ in major ways in the N terminus. Only p85␣ was able to restore insulinstimulated JNK activation (Fig. 6b, right), indicating the importance of the N-terminal region of p85 in the regulation of JNK activation.
To explore the mechanism of JNK activation regulated by p85, we assessed the Ras/Rac//MEKK1/SEK1/JNK signaling cascade, which has been shown to be activated in growth factor signaling (34 -37). Indeed, expression of a dominant negative form of Ras, but not treatment of a MEK inhibitor, almost completely blocked insulin-stimulated JNK activation in wildtype cells (Fig. 6c). We also assessed the effect of p85␣ deletion on insulin-induced Rac activation by measuring the GTPbound form of Rac interacting with the p21 binding domain of p21-activated kinase 1. As shown in Fig. 6d, p85␣ ϩ/Ϫ cells displayed the highest activity of Rac stimulated by insulin, whereas wild-type and p85␣ Ϫ/Ϫ cells exhibited the comparable levels of activation. This is consistent with the fact that the guanine nucleotide exchange factors for Rac, such as Vav, have been shown to activate Rac in a PIP 3 -dependent fashion (38), and p85␣ ϩ/Ϫ cells exhibit higher levels of PIP 3 compared with wild-type and p85␣ Ϫ/Ϫ cells (9). MEKK1 activity was estimated by phosphorylation of Ser-251 on SEK1, which is one of the MEKK1 phosphorylation sites, using phospho-specific antibody. MEKK1 activity was not decreased in p85␣ Ϫ/Ϫ cells, and it was not increased in p85␣ Ϫ/Ϫ cells reconstituted with p85␣, although the reconstitution completely normalized JNK activity (Fig. 6e). These data suggest that insulin activates a signaling cascade of Ras/Rac//MEKK1/SEK1 to activate JNK, but this cascade is not altered by changes in the p85 protein level. Furthermore, although JNK has been reported to attenuate tyrosine phosphorylation of IRS-1 by increasing its serine phos-phorylation and negatively modulating insulin sensitivity (39 -41), there was no significant difference in the time course of tyrosine phosphorylation of IRS-1 in p85␣ Ϫ/Ϫ cells versus wildtype cells (Fig. 6f).
Finally, as noted above, tyrosine phosphorylation of IRS-2 was up-regulated only in p85␤ Ϫ/Ϫ cells. This up-regulation was normalized by expression of p85␤, but not p85␣, using adenovirus-mediated gene transfer (Fig. 7). Because deletion of p85␤ does not affect insulin-induced PI 3-kinase activity, it is likely that a molecule specifically interacting with p85␤, but not p85␣, mediates a signal to enhance IRS-2 phosphorylation, independent of PI 3-kinase activity. DISCUSSION Through the use of pharmacological inhibitors and dominant negative mutants, PI 3-kinase has been shown to be required for a wide variety of the metabolic effects of insulin, including stimulation of glucose transport and glycogen synthesis (1,28,(42)(43)(44). This occurs through Class IA PI 3-kinase, which is composed of a p110 catalytic subunit and one of several isoforms of regulatory subunits (1). To better understand the role of these regulatory subunits in insulin signaling and glucose homeostasis, we have generated and analyzed mice and cell lines with disruption of the Pik3r1 gene, which encodes p85␣ and its splice variants (p85␣ Ϫ/Ϫ mice), and the Pik3r2 gene, which encodes p85␤ (p85␤ Ϫ/Ϫ mice). p85␣ Ϫ/Ϫ mice die within a week (13,14) and have a 70% reduction in total regulatory subunits, whereas p85␣ ϩ/Ϫ mice with a 40% reduction in regulatory subunits and p85␤ Ϫ/Ϫ mice with a ϳ25% reduction in regulatory subunits are viable and show increased insulin sensitivity (10,12). Thus, a moderate reduction in the regulatory subunits seems to improve insulin sensitivity, whereas a severe defect is lethal.
Two potential mechanisms have been suggested to explain these findings. One is that the moderate reduction of the regulatory subunits decreases primarily p85 monomer, allowing the remaining heterodimeric holoenzyme to better bind to phosphorylated IRS proteins (9,12). Another possibility is that the reduction in regulatory subunits increases PIP 3 levels by a mechanism independent of PI 3-kinase activity (9), as suggested by the finding that both p85␣ ϩ/Ϫ mice and p85␤ Ϫ/Ϫ mice exhibit significantly increased Akt activity with no increase in PI 3-kinase activity (10,12). In the present study we have assessed these potential mechanisms and explored functional differences between p85␣ and p85␤ at the cellular level using brown adipose cell lines from p85␣ Ϫ/Ϫ and p85␤ Ϫ/Ϫ mice (15).
Interestingly, although PI 3-kinase activity is required for differentiation of preadipocytes to adipocytes (16,22), cells with deletion of the Pik3r1 gene differentiate normally despite a Ͼ50% decrease in PI 3-kinase activity, suggesting that a relatively small amount of PI 3-kinase is sufficient for adipocytes differentiation or that some of the signaling events essential for differentiation are maintained in p85␣ Ϫ/Ϫ cells. Moreover, the levels of adiponectin (also known as Acrp30), an adipokine that has been shown to improve insulin sensitivity in vivo (17,45,46)  that changes in adiponectin concentrations do not play a major role in improvement of insulin sensitivity or hypoglycemic phenotypes in the knockout mice. In any case, because p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells can be differentiated to adipocytes, these provide an excellent system to investigate insulin signaling and biological responses in the absence of these gene products (15).
Affinity purification of proteins interacting with the pYMXM consensus binding motif for SH2 domains of the regulatory subunits of PI 3-kinase in p85␣ Ϫ/Ϫ cells revealed a 70% reduction in the regulatory subunits with only p85␤, whereas in p85␤ Ϫ/Ϫ cells there is a 25% reduction of the regulatory subunits, and p85␣ is the major remaining isoform, consistent with our findings in tissues and embryonic fibroblasts from these knockout mice (9,10). In p85␣ Ϫ/Ϫ cells this reduction of the regulatory subunits leads to the secondary decrease in p85-p110 dimers as well as the catalytic subunits, since the p110 monomer is unstable (13,25). When combined, these changes result in approximately a 50% decrease in PI 3-kinase activities associated with IRS-1, IRS-2, and phosphotyrosine complexes. By contrast, 25% of the reduction of the regulatory subunit in p85␤ Ϫ/Ϫ cells does not decrease p85⅐p110 complexes, probably because more than 30% of the regulatory subunit exists as a monomer in the wild-type state, and this portion is preferentially decreased by knockout (9,12). Thus, despite the absence of p85␤, PI 3-kinase activity associated with IRS-1 is preserved. In fact, PI 3-kinase activity associated with IRS-2 is slightly increased due to a marked up-regulation of IRS-2 phosphorylation in p85␤ Ϫ/Ϫ cells. Interestingly, there was no appreciable increase in p85␣, Grb2, or SHP2 interacting with IRS-2, suggesting that phosphorylation occurs at other sites or that these SH2-containing proteins are in a limiting amount and cannot increase above control levels. Although this up-regulation of IRS-2 phosphorylation is most prominent in adipocytes, milder degrees of up-regulation are also observed in liver and muscle of p85␤ Ϫ/Ϫ mouse but not detected in p85␣ ϩ/Ϫ or p85␣ Ϫ/Ϫ mice (9). Moreover, this up-regulation is normalized specifically by reconstitution of p85␤ Ϫ/Ϫ cells with p85␤, but not p85␣, indicating a specific effect of the p85␤ isoform. Because the N-terminal regions of p85␣ and p85␤ have a lower degree of homology than the C-terminal regions (4), it is possible that a molecule interacting with the N-terminal region of p85␤ specifically suppresses tyrosine phosphorylation of IRS-2 in the wild-type state. Although some proteins have been reported to preferentially bind to one isoform versus the other in vitro (1), this is the first functional difference between p85␣ and p85␤ demonstrated in vivo. The exact physiological significance and mechanism of the up-regulation of IRS-2 phosphorylation in the absence of p85␤ require further study but could contribute to increased insulin sensitivity in these mice.
In p85␣ Ϫ/Ϫ cells, the decrease in PI 3-kinase activity associated with phosphotyrosine complexes results in a relatively modest decrease in Akt activity. This was also observed in fibroblasts and in vivo (9,14). By contrast, Akt activity is significantly up-regulated in p85␤ Ϫ/Ϫ cells without an increase in PI 3-kinase activity, also consistent with in vivo finding (10).
These are presumably caused by (a) a reduction in the monomeric p85, which interferes the effective signal transmission by p85-p110 dimer and (b) a reduction in the p85-mediated PIP 3decreasing signal, independent of PI 3-kinase activity (9,10,12). The finding of preserved insulin-stimulated PI 3-kinase activity in p85␤ Ϫ/Ϫ cells with the 25% of reduction of the regulatory subunit supports the first mechanism. This is different from the finding in p85␣ Ϫ/Ϫ cells reconstituted with p85␣ to the level of 20 -30% that of wild-type cells, which leads to only 50% normalization of insulin-stimulated PI 3-kinase activity but a full restoration of Akt activity. Higher levels of reconstitution up to 60 -70% that of the wild-type level fail to further increase p110 or insulin-stimulated PI 3-kinase activity but do increase p85 monomer. However, under these conditions, because both the p85 monomers and p85-p110 heterodimers can bind stoichiometrically to IRS proteins, the p85 monomers do not decrease formation of p85⅐p110⅐IRS complexes or insulin-stimulated PI 3-kinase activity. By contrast, increasing expression of p85␣ results in a decrease in Akt activity despite no reduction in insulin-stimulated PI 3-kinase activity, clearly indicating the existence of the second mechanism. We have previously shown that p85␣ ϩ/Ϫ fibroblasts and under some conditions even p85␣ Ϫ/Ϫ fibroblasts have higher levels of PIP 3 than wild-type cells. Because these cells have an equal or less amount of PI 3-kinase activity than wild type cells, the increased PIP 3 levels must be due to the decreased lipid phosphatase activity (Fig. 8). Indeed, we have found that there is a reduction of lipid phosphatase PTEN activity in p85␣ ϩ/Ϫ cells and p85␣ Ϫ/Ϫ cells. 2 Taken together these data suggest that the p85 proteins have six functions, (i) serving as a bridge between p110 and tyrosine-phosphorylated IRS proteins, (ii) regulating p110 catalytic activity, (iii) stabilizing the p110 subunit by binding, (iv) when monomeric, modulating PI 3-kinase by competing with the p85-p110 holoenzyme for binding to IRS proteins, (v) downregulating PIP 3 levels independent of regulation of PI 3-kinase activity, and (vi) in the case of p85␤, regulating tyrosine phosphorylation of IRS-2. Thus, the magnitude of PI 3-kinase-dependent signaling events is determined by the balance between the positive signal derived from p85⅐p110⅐IRS complexes, the amount of p85 monomer and the negative signaling derived from p85.
Several lines of evidence including recent knockout studies suggest that Akt regulates most of the biological responses to insulin downstream of PI 3-kinase, including activation of glycogen synthesis, anti-apoptosis, and glucose transport (3,19,(47)(48)(49), although the latter is controversial (50). Indeed, in the present study, in parallel with the changes in insulin-induced Akt activity, anti-apoptotic effect of insulin is down-regulated in p85␣ Ϫ/Ϫ cells and up-regulated in p85␤ Ϫ/Ϫ cells. At the molecular level at least two pathways seem to be involved in these changes. One is proapoptotic protein Bad. Bad is phos- phorylated on two serine residues (112 and 136) upon growth factor stimuli by the actions of Akt and p90 RSK , respectively (51,52). Phosphorylation of these sites promotes association of Bad to 14-3-3 proteins, thus preventing Bad from translocating to mitochondrial membrane, where it binds to the anti-apoptotic proteins Bcl-2 and Bcl-xL and suppresses their activity (32). The interaction between Bad and 14-3-3 in response to insulin is decreased in p85␣ Ϫ/Ϫ cells and increased in p85␤ Ϫ/Ϫ cells, paralleling Akt activity and the anti-apoptotic effect of insulin in these cells. Another molecular mechanism is that insulin-induced inhibition of apoptosis involves the forkhead transcription factor Foxo1. Survival factors, such as insulin, stimulate phosphorylation of Foxo1 by Akt. This promotes the interaction between Foxo1 and 14-3-3, thereby inhibiting translocation of Foxo1 to nucleus and transcription of proapoptotic factors (32,33). Indeed, insulin-induced phosphorylation of Foxo1 is decreased in p85␣ Ϫ/Ϫ cells, whereas it is modestly increased in p85␤ Ϫ/Ϫ cells. The combined effects of these two pathways appear to define anti-apoptotic effects by insulin in p85␣ Ϫ/Ϫ cells and p85␤ Ϫ/Ϫ cells.
As noted above, whereas insulin-induced glucose transport activity in p85␣ Ϫ/Ϫ cells decreased in parallel with Akt activity, it was not increased in p85␤ Ϫ/Ϫ cells despite the up-regulation of Akt activity. This can be explained, in part, by the fact that in p85␤ Ϫ/Ϫ cells, there is a significant reduction of GLUT4 protein with a modest up-regulation of GLUT1. This appears to be due to degradation of GLUT4 protein, since mRNA levels of GLUT4 are indistinguishable in all genotypes of cells. The sentrin (SUMO-1)-conjugating enzyme mUbc9 has been reported to increase GLUT4 and decrease GLUT1 by binding to their C-terminal region (53). It is possible that Akt negatively regulates mUbc9 expression, its activity or downstream molecules, thereby modulating protein levels of GLUT4 and GLUT1. Whatever the mechanism, it was specific to adipocytes and was not detected in muscle in either type of knockout mouse (data not shown).
GS activity is also believed to be regulated by PI 3-kinase/ Akt pathway at least in muscle (3,19,47). In muscle, Akt phosphorylates and deactivates GSK3, leading to activation of GS (3,19,47), whereas PP1 dephosphorylates and activates GS in response to insulin in muscle via phosphorylation of its regulatory subunit G M (54). However, in 3T3L-1 adipocytes, overexpression of a constitutively active Akt fails to increase GS activity (19,48), possibly due to a relatively minor contribution of GSK3 in GS regulation in adipocytes (19,29). In the present study, insulin-induced GSK3 phosphorylation is similar in all genotypes of cells, whereas insulin-induced dephosphorylation of Site 3a in GS, which occurs through activity of PP1, is significantly up-regulated in both p85␣ Ϫ/Ϫ and p85␤ Ϫ/Ϫ cells, leading to an up-regulation of GS activity in these cells. On the other hand, although the exact mechanisms of regulation of PP1 in adipocytes are unknown (29,30), it is possible that the reduction of the negative effects by deletion of p85␣ or p85␤ up-regulates some signaling pathway downstream of PI 3-kinase, thereby enhancing GS activity through PP1. This up-regulation of GS activity and subsequent glycogen synthesis may contribute to hypoglycemic phenotype in p85␣ Ϫ/Ϫ mice and p85␤ Ϫ/Ϫ mice.
The most unexpected finding in this study and our previous study is that p85 down-regulates PIP 3 -dependent signaling and PIP 3 itself at a level independent of its regulation of PI 3-kinase (9,11). The existence of p85-dependent, but PI 3-kinase-independent, signaling pathway has also been suggested by studies showing that expression of a membrane-localized form of ⌬p85 increases transcription of interleukin-2 in T cells (55). We have found that insulin-induced JNK1 activity is also