Main

The first PI3K to be purified and characterized was the mammalian p85 (regulatory subunit)–p110 (catalytic subunit) dimeric enzyme, which associates with protein tyrosine kinases8. Two isoforms of p85, p85α and p85β, are derived from distinct genes, and each contains two Src homology 2 (SH2) domains and one SH3 domain3,4,5. Pik3r1 (encoding p85α) generates two splicing variants, p55α (p85/AS53; Refs 9,10) and p50α (Refs 6,7), that lack the SH3 domain and bcr–homology domain of p85α and contain unique 34– or 6–aa sequences at the amino terminus, respectively. p55γ is derived from a distinct gene, lacks an SH3 domain and contains a unique 34–aa sequence at the N terminus11. Insulin activates PI3K by tyrosine phosphorylation of IRSs such as Irs1 and subsequent binding of p85α associated with p110 (refs 12, 13,14,15). PI3K inhibitors such as wortmannin, LY294002 or dominant–negative forms of p85 have been used in functional analyses of PI3K. These agents, however, have several serious drawbacks for defining the biological functions of PI3K and the multiple isoforms of its regulatory subunit16. To investigate the role of PI3K in glucose metabolism in vivo, we deleted the first exon of Pik3r1 (exon 1A, which contains the initiation codon) in mice to selectively abolish the expression of full–length Pik3r1 mRNA without disrupting the p55α and p50α splicing variants (Fig. 1a).

Figure 1: Targeted disruption of Pik3r1.
figure 1

a, Schematic representation of mouse Pik3r1 (top), the targeting vector (middle) and the targeted gene region (bottom). A neomycin resistance gene (neo) was substituted for the PstI– PstI fragment. The gene encoding diphtheria toxin A fragment (DTA) was ligated to the 3´ terminus. B, E, P and X indicate BamHI, EcoRI, PstI and XbaI sites, respectively. Not all Pst I sites are indicated. b, Southern–blot analysis of progeny from a heterozygote cross. Genomic DNA was double–digested with EcoRI and BamHI and hybridized with probe A under high stringency. The 6.0–kb band corresponds to the wild–type, the 1.6–kb band to the mutant allele. c, PI3K activities in anti–p85α–immunoprecipitates. PI3K activities in the liver were determined in immunoprecipitates with anti–p85α antibody before and after an insulin injection through the portal vein. Values are expressed as the mean±s.e.m. obtained from three independent experiments.

Full size image

Matings of Pik3r1+/– mice yielded Pik3r1−/− mice (Fig. 1b) that survived normally under pathogen–free conditions. PI3K activity immunoprecipitated with anti–p85α antibody against the SH3 region of p85α that recognized p85α, but not p85β, p55γ, p55α or p50α, before and after insulin stimulation were completely eliminated in the liver (Fig. 1c). Blood glucose levels were significantly lower in Pik3r1−/− mice than in wild–type mice (114±5 versus 140±4 mg/dl, P<0.001). Pik3r1−/− mice showed significantly smaller increase in both glucose and insulin levels when they were loaded with glucose (Fig. 2a), and hypoglycaemia when they were loaded with insulin (Fig. 2b). To evaluate insulin sensitivity more directly, we determined steady–state plasma insulin (SSPI) and steady–state plasma glucose (SSPG) levels17. Although Pik3r1−/− and wild–type mice exhibited the same SSPI levels, Pik3r1−/− animals showed a significantly decreased SSPG levels (Fig. 2c), indicating that they were indeed more sensitive to insulin. Thus, Pik3r1−/− mice showed increased insulin sensitivity and hypoglycaemia, in contrast with our original predictions.

Figure 2: Increased insulin sensitivity and hypoglycaemia in Pik3r1−/− mice.
figure 2

a, Changes in plasma glucose (top) and serum insulin (bottom) levels in a glucose tolerance test. Mice were fasted for more than 16 h and then given 1.5 mg of glucose/g of body weight via intraperitoneal injection. Values are expressed as the mean±s.e.m. obtained from the analysis of wild–type (open circles, n=10) and Pik3r1−/− mice (filled circles, n=10). b, Changes in plasma glucose levels in an insulin tolerance test. Freely fed mice were given 0.75 mU of human insulin/g of body weight. Values are expressed as the mean±s.e.m. obtained from the analysis of wild–type (open circles, n=9) and Pik3r1−/− mice (filled circles, n=9). c, SSPI and SSPG levels. A mixture of somatostatin, glucose and insulin was infused at a rate of 5 μl/min for 80 min. Plasma insulin and glucose levels were maintained at a constant level 60 min after the start of infusion, and the means of three points of plasma insulin and glucose levels were calculated as SSPI and SSPG levels, respectively. Values are expressed as the mean±s.e.m. obtained from the analysis of wild–type (open bars, n=5) and Pik3r1−/− mice (filled bars, n=4). *P<0.05; ** P<0.01 compared with wild–type mice.

Full size image

2–Deoxy–glucose uptake in the isolated soleus muscle was significantly (P<0.05) increased in Pik3r1−/− mice: 133% and 140% that in wild–type mice at basal and 10 nM insulin, respectively (Fig. 3a). Moreover, 14C glucose uptake in adipocytes was significantly (P<0.01) increased in Pik3r1−/− mice: 199% and 159% of that in wild–type mice at basal and 10 nM insulin, respectively. 3– O–methylglucose uptake in adipocytes was also increased 2–3–fold with physiological concentrations of insulin (0.1–1.0 nM) in Pik3r1−/− mice (Fig. 3b). Wortmannin suppressed insulin–stimulated glucose transport in both Pik3r1−/− and wild–type mice. We next examined the intracellular localization of Glut4 by membrane fractionation and immunoblotting with anti–Glut4 antibody. Although the total amount of Glut4 in Pik3r1−/− mice was the same as that in wild–type mice (data not shown), Glut4 levels in the PM fraction in Pik3r1−/− mice were significantly elevated to approximately 150% of those in wild–type mice at basal and maximal stimulation of insulin (Fig. 3b, top), consistent with increased glucose transport activity in both these states in Pik3r1−/− mice. We investigated the kinetics of endocytosis by measuring reversal of insulin–stimulated glucose transport activity with collagenase18. The decrease in insulin–stimulated glucose transport activity with collagenase occurred with a t1/2 of 18.3 min in wild–type and 18.8 min in Pik3r1−/− adipocytes. Thus, we concluded that increased glucose transport activity in adipocytes was due to increased insulin–induced exocytotic pathway of Glut4 translocation in Pik3r1−/− adipocytes. The mechanism underlying the elevated level of basal glucose transport in Pik3r1−/− mice, which may contribute to the decreased fasting blood glucose level, is unknown.

Figure 3: Increased glucose transport in skeletal muscle and adipocytes from Pik3r1–deficient mice.
figure 3

a, 2–Deoxy–glucose uptake in isolated soleus muscles. Results are mean±s.e.m. obtained from five independent experiments. *P<0.05 compared with wild–type mice. b, Dose–response curves of insulin–stimulated 3–O–methylglucose (3–MG) transport and the amount of Glut4 in the PM fraction of adipocytes. 3–MG uptake was measured in isolated adipocytes from wild–type (WT, open circles) and Pik3r1−/− mice (filled circles) in the presence or absence of 100 nM wortmannin. There were no significant differences in size or morphology of adipocytes from wild–type and Pik3r1−/− mice. Results are mean±s.e.m. obtained from quadruplicate samples. Top, adipocytes were incubated at 37 °C for 30 min in the presence or absence of 10−7 M insulin. Cells were fractionated, and the same amount of protein from the PM fraction was separated by SDS–PAGE and blotted with anti–Glut4 antibody.

Full size image

Despite complete abrogation of full–length p85α, PI3K activity in the immunoprecipitates with anti–phosphotyrosine antibody (anti–PY) was 63% and 50% of normal in skeletal muscle and adipocytes, respectively (Fig. 4a), and similar results were obtained with anti–Irs1 antibody (data not shown). We also examined expression levels of PI3K regulatory subunit isoforms in skeletal muscle ( Fig. 4b, lanes 1–5) and adipocytes (Fig. 4b, lanes 6–17). Lysates were immunoprecipitated with a panel of antibodies against: (i) regulatory subunits such as p50α, p55α, p55γ or p85β; (ii) the N–terminal SH2 domain of p85 (anti–p85PAN) that can recognize p85α as well as p50α, p55α, p55γ and p85β; or (iii) the p110α catalytic subunit of PI3K, followed by blotting with anti–p85PAN. In wild–type mice, p85α was the major PI3K regulatory subunit isoform (Fig. 4b, lanes 1,6) that bound to the p110α catalytic subunit (Fig. 4b, lanes 3,8). In Pik3r1−/− mice, however, p50α was the major PI3K regulatory subunit isoform ( Fig. 4b, lanes 2,5,7,11) that bound to the p110α catalytic subunit (Fig. 4b, lanes 4,9). In adipocytes, p50α was overexpressed in Pik3r1−/− mice (Fig. 4b, lanes 6 versus 7, 10 versus 11). Although p55α was also overexpressed in Pik3r1−/− adipocytes (Fig. 4b, lanes 6 versus 7, 12 versus 13), its association with p110α was much weaker than that of p50α (Fig. 4b, lane 9). p55γ was not detected in adipocytes from either mouse type using a specific antibody against p55γ (Fig. 4b, lanes 14 versus 15; Refs 7,11). Expression of p85β was not altered between wild–type and Pik3r1−/− mice (Fig. 4b, lanes 16 versus 17) and its association with p110α was weaker than that of p50α in Pik3r1−/− mice ( Fig. 4b, lane 9).

Figure 4: Switch in PI3K regulatory subunit isoform from p85α to p50α in skeletal muscle and adipocytes.
figure 4

a, PI3K activities associated with tyrosine–phosphorylated proteins. Adipocytes were incubated for 5 min at 37 °C in the presence or absence of 100 nM insulin. PI3K activities were determined in the anti–PY immunoprecipitates of the lysates. In each experiment, PI3K activity is given relative to that measured in wild–type mice under basal conditions. Results are expressed as the mean±s.e.m. of six experiments in muscle from wild–type (WT) and Pik3r1−/− mice, and of three experiments in adipocytes. b, Expression of regulatory subunits of PI3K. Lysates from skeletal muscle (lanes 1–5) or adipocytes (lanes 6–17) were directly immunoprecipitated with anti–p85PAN, anti–p110α, anti–p50α, anti–p55α, anti–p55γ, or anti–p85β and blotted with anti–p85PAN (lanes 1–15) or anti–p85β (lanes 16, 17). c, PI3K regulatory subunits bound to Irs1. Lysates from skeletal muscles (left) or adipocytes (right) with or without insulin stimulation were immunoprecipitated with anti–Irs1 and blotted with anti–PY (top) or anti–p85PAN (bottom). In some Pik3r1−/− animals, another 85–kD protein, presumably p85β, was bound to Irs1, although the relative amount of this protein was much lower than that of p50α. We also noted a 48–kD protein recognized by anti–p85PAN that bound to Irs1, although the identity of this protein is unknown. d, Irs1 bound to PI3K regulatory subunits. Lysates from skeletal muscles (left) or adipocytes (right) with or without insulin stimulation were immunoprecipitated with anti–p85PAN (top) or anti–p50α (bottom) and blotted with anti–Irs1 or anti–PY. e, Time course of the association of Irs1 with p85α or p50α. Isolated adipocytes were divided equally into five aliquots and incubated at 37 °C with 100 nM insulin for the indicated time. Cells were rapidly lysed and immunoprecipitated with anti–Irs1 followed by western blot with anti–Irs1 or anti–p85PAN. The relative intensity of the regulatory subunits normalized to the amount of Irs1 was calculated by computed densitometry, and the means of intensity obtained from two independent experiments are displayed in the bar graph.

Full size image

We next determined which regulatory subunits of PI3K were bound to Irs1 in skeletal muscle (Fig. 4c, left) and adipocytes (Fig. 4c, right). Irs1 was tyrosine phosphorylated to a similar extent in both wild–type and Pik3r1−/− mice (Fig. 4c, top). Lysates were immunoprecipitated with anti–Irs1 and blotted with anti–p85PAN. In wild–type mice, p85α was bound to Irs1 in an insulin–dependent fashion. In Pik3r1−/− mice, however, p50α was the major protein recognized by anti–p85PAN bound to Irs1 in an insulin–dependent fashion (Fig. 4c, bottom). p55α was also bound to Irs1 in an insulin–dependent fashion in adipocytes. When immunoprecipitation and immunoblotting were carried out in the reverse, Irs1 was found to be the major tyrosine–phosphorylated protein bound to p85α in wild–type mice or to p50α in Pik3r1−/− mice after insulin stimulation in skeletal muscle (Fig. 4d, left) and adipocytes (Fig. 4d, right). Irs2 (ref. 13) and Irs3 (ref. 15) were more weakly associated with p85α or p50α than Irs1 (data not shown). We thus conclude that p85α in wild–type mice and p50α in Pik3r1−/− mice have roles in insulin–stimulated PI3K activation by binding IRS family proteins, mainly Irs1, in muscle and adipocytes. In addition, p55α may have a compensatory role in insulin–stimulated activation of PI3K in Pik3r1−/− mice, whereas p55γ and p85β were not shown to be upregulated to compensate for p85α deficiency. The kinetics of the interaction between Irs1 and p85α and p50α were also different. In wild–type adipocytes the association between p85α and Irs1 reached a maximum 2 min after insulin stimulation and continued up to 20 min. In contrast, in adipocytes from Pik3r1−/− mice, although the association between p50α and Irs1 reached a maximum 2–5 min after insulin stimulation, steady–state levels of the p50α/Irs1 complex gradually decreased after 5 min (Fig. 4e).

To unravel the mechanism by which a switch in PI3K regulatory subunit isoform from p85α to p50α caused stimulation of glucose transport and Glut4 translocation in Pik3r1−/− adipocytes, we studied PI3K activities in subcellular fractions, especially in the LDM fraction, as activation of PI3K activities in the LDM fraction may have a role in Glut4 translocation2,19,20. PI3K activities associated with Irs1 were stimulated with insulin in PM (12.4–fold), LDM (9.2–fold) and cytosolic fractions (43–fold), and in homogenate (6.3–fold) in wild–type mice (Fig. 5a). Although PI3K activities associated with Irs1 were increased in response to insulin in PM (7.2–fold) and cytosolic fractions (18.1–fold), and in homogenate (2.7–fold) in Pik3r1−/− mice, there was no increase in insulin–stimulated PI3K activity associated with Irs1 in the LDM fraction. Consistent with this observation, western–blot analyses revealed that only a small amount of Irs1 was associated with p50α PI3K in the LDM fraction of Pik3r1−/− mice 5 min after insulin stimulation, whereas the amount of Irs1 associated with p85α was proportionate in the LDM fraction to that in homogenate in wild–type mice (Fig. 5b, c ). However, there was only a modest reduction in insulin–stimulated PI3K activity in the anti–p110α immunoprecipitates from LDM as well as PM (Fig. 5a, bottom).

Figure 5: Intracellular localization of PI3K and its activity and increased PtdIns(3,4,5)P3 generation in adipocytes from Pik3r1−/− mice.
figure 5

a, PI3K activities in anti–PY or anti–p110α immunoprecipitates in subcellular fractions. Adipocytes were incubated for 5 min at 37 °C in the presence or absence of 100 nM insulin before subcellular fractionation. Equal amounts of protein (20 μg) were immunoprecipitated with anti–PY or anti–p110α. In each experiment, the activity is given relative to that measured in the PM of wild–type cells under basal conditions. Results are expressed as the mean of two or three experiments. Open bars, wild–type (WT) mice; filled bars, Pik3r1−/− mice. An image of the PtdIns(3)P position in a representative experiment is also shown. b,c, Association of Irs1 with the regulatory subunit of PI3K in the homogenate and LDM fraction of adipocytes. Adipocytes were incubated for 5 min at 37 °C in the presence or absence of 100 nM insulin, followed by subcellular fractionation. b, Homogenates (100 μg) or LDM fractions (30 μg) were immunoprecipitated with anti–Irs1 and blotted with anti–p85PAN. Relative amounts of p85α or p50α associated with Irs1 are shown in the bar graph. c, Homogenates (100 μg) or LDM fractions (60 μg) were immunoprecipitated with anti–p85PAN and blotted with anti–PY. d, Intracellular distribution of the regulatory subunits of PI3K in adipocytes. Adipocytes were incubated for 5 min at 37 °C in the presence or absence of 100 nM insulin, followed by subcellular fractionation. Equal amounts of protein (60 μg) were immunoprecipitated with anti–p85PAN and blotted with anti–p85PAN. Relative amounts of p85α or p50α in each fraction are shown in the bar graph. e, In vivo generation of PtdIns(3,4,5)P3 in response to insulin. Freshly isolated adipocytes were labelled for 3 h with [32P]orthophosphate and then incubated in the presence or absence of 100 nM insulin for the indicated time. 32P–labelled phospholipids were extracted and separated by TLC. The spots were identified by the standards synthesized by PI3K in vitro. Levels of PtdIns(3,4,5)P3 normalized to the total phospholipids obtained from the analysis of wild–type (open circles) or Pik3r1−/− mice (filled circles) are shown. Experiments were repeated three times with a similar result. A representative image at 10 min after insulin stimulation is shown in the right panel. Lanes: 1, the standards synthesized by PI3K in vitro; 2, WT; 3, Pik3r1 −/−.

Full size image

We examined the levels of expression of regulatory subunits of PI3K in each subcellular fraction. Both p85α and p50α were similarly distributed in all fractions, including the LDM fraction (Fig. 5d). Expression of p50α was greatest in cytosol, followed by LDM, with the lowest expression in PM. Moreover, the distribution of p50α between control and insulin–treated adipocytes was similar. Thus, the decrease in the association of Irs1 with the p50α/p110α heterodimer in the LDM fraction in Pik3r1−/− mice was not due to a selective decrease in the amounts of p50α ( Fig. 5d) or p110α (Fig. 5a) in the LDM fraction. Together with the kinetics of binding of p50α to Irs1 (Fig. 4e), it seems likely that Irs1 became readily dissociated from the p50α/p110α heterodimer in the LDM fraction once PI3K was activated by the association of p50α/p110α heterodimer with Irs1 in response to insulin in the PM and cytosolic fractions of Pik3r1−/− mice. It has been proposed that PtdIns(3,4,5)P3 can bind directly to SH2 domains of p85α or p50α, thereby causing dissociation of PI3K from tyrosine–phosphorylated proteins such as Irs1 (ref. 16). Thus, the generation of PtdIns(3,4,5)P3 in response to insulin in adipocytes may be increased in Pik3r1−/− mice, although total PI3K activity or PI3K activity in anti–PY immunoprecipitates in vitro was decreased. To test this possibility, we measured in vivo generation of PtdIns(3,4,5)P3 in response to insulin in adipocytes. In Pik3r1−/− mice, generation of PtdIns(3,4,5)P3 in response to insulin in vivo was indeed increased 5 min after insulin stimulation (Fig. 5e), which was consistent with the kinetics of dissociation of p50α from Irs1 ( Fig. 4e). This increase in the amounts of PtdIns(3,4,5)P3 generated in response to insulin appeared to be associated with facilitating translocation of Glut4 and concomitantly promoting dissociation of p50α from Irs1. Thus, the isoform switch from p85α to p50α PI3K regulatory subunit in Pik3r1−/− mice is associated with an increase in insulin–induced generation of PtdIns(3,4,5)P3 that appears to be responsible for the increased glucose transport and hypoglycaemia. This study provides the first direct evidence that PI3K and its regulatory subunits have a crucial role in glucose homeostasis and glucose transport in vivo.

Methods

Homologous recombination.

A DNA fragment containing the first exon of Pik3r1 was cloned from the D3 genomic library using a human cDNA fragment flanking the ATG translation initiation codon as a probe. Only a single clone containing a region of genomic DNA encoding exon 1A was identified. An 880–bp neomycin resistance gene without an MC1 promoter but with a poly(A)+ addition signal was substituted for the PstI–PstI fragment which started at –8 of the ATG translation initiation codon. The gene encoding diphtheria toxin A fragment (DTA) with an MC1 promoter was ligated at the 3´ terminus of the vector backbone for negative selection21. We carried out homologous recombination experiments as described22. Chimaeric mice were generated as described23. Subsequently, three clones transmitted the mutation through the germ line and gave rise to three independent Pik3r1−/− mouse lines with the same phenotype.

Antibodies.

An anti–p85α monoclonal antibody against the N–terminal region of human p85α (anti–p85α) was purchased from MBL. Anti–p85 polyclonal antibody against a full–length p85–GST fusion protein containing the N–terminal SH2 domain of p85α (anti–p85PAN) and anti–p110α antibody against the carboxy–terminal region (aa 1,054–1,068; anti–p110α) were purchased from Upstate Biotechnology. We prepared specific antibodies against p50α (anti–p50α), p55α (anti–p55α), p55γ (anti–p55γ) or p85β (anti–p85β) as described7. Monoclonal anti–phosphotyrosine antibody (anti–PY) and anti–Irs1 antibody was purchased from Upstate Biotechnology.

Immunoprecipitations and immunoblotting.

Liver and muscle tissues were excised and homogenized in ice–cold buffer A (25 mM Tris–HCl, pH 7.4, 10 mM Na3VO4, 10 mM NaPPi, 100 mM NaF, 10 mM EDTA, 10 mM EGTA and 1 mM phenylmethylsulfonyl fluoride (PMSF)). Adipocytes were solubilized with ice–cold buffer A containing 1% NP–40. Samples were separated by SDS–polyacrylamide gel electrophoresis followed by immunoblotting. Subsequent detection was by either [125I] protein A or enhanced chemiluminescence.

PI3K assay.

PI3K activities in liver and muscle were determined in immunoprecipitates with the indicated antibodies after an insulin injection into the portal vein24. PI3K activities in total lysates or subcellular fractions of adipocytes were determined as described25 with some modification. In brief, adipocytes were incubated in the presence or absence of insulin, homogenized using a Teflon pestle and, in some cases, subjected to fractionation as described below. We lysed membrane fractions in buffer B (20 mM Tris–HCl, pH 7.4, 10% glycerol, 5 mM EGTA, 5 mM EDTA, 1 mM Na3VO4, 100 mM NaF, 10 mM NaPPi, 1 mM PMSF, 10 μg/ml leupeptin and 10 μg/ml aprotinin) containing 1% Triton X–100. PI3K was immunoprecipitated with the indicated antibody, and its activity was measured as described24.

In vivo insulin sensitivity.

SSPG and SSPI levels were determined as described17 with some modification. In brief, mice were anaesthetized with ether aspiration followed by intraperitoneal injection of uletane, and a catheter was inserted in the right jugular vein. All experiments were performed on conscious mice. A mixture of somatostatin (0.1 mg/kg/hr; Sigma), glucose (1 g/kg/hr) and insulin (0.1 U/kg/hr) was infused at a rate of 5 μl/min for 80 min. Blood samples were collected from the tail tip at 60, 70 and 80 min after the initiation of infusion. The means of three points of plasma glucose and insulin concentrations were calculated as SSPG and SSPI, respectively.

Glucose transport activity.

Measurement of 2–deoxy–D–[U–14C]glucose transport activity was performed as described24. Adipocytes were prepared by the collagenase digestion method26, and the [U–14C]glucose uptake assay was performed as described27. The 3–O–methylglucose (3–MG) transport assay was performed using a double–isotope, 3–O–[14C]methylglucose and L–[3H]glucose (Du Pon NEN) as described18. To examine the inhibition of insulin–induced glucose transport by wortmannin, adipocytes were preincubated with wortmannin (100 nM) for 30 min before measurement of 3–MG transport. To determine the kinetics of the endocytosis pathway, the insulin–treated adipocytes were further treated with crude collagenase (0.2 mg/ml), and 3–MG transport activities were measured as described18.

Preparation of subcellular fractions of adipocytes.

Subcellular fractions were prepared as described26 with some modifications. In brief, cells were suspended in 6 vol. (v/v) of homogenizing buffer (20 mM Tris–HCl, pH 7.4, 250 mM sucrose, 5 mM EGTA, 5 mM EDTA, 1 mM Na3VO4, 10 mM NaF, 10 mM NaPPi, 1 mM PMSF, 10 μg/ml leupeptin and 10 μg/ml aprotinin) and homogenized by 10 strokes in a Teflon pestle at 17 °C. We centrifuged homogenates for 15 min at 18,000g at –4 °C to remove fat cake. The plasma membrane–rich fraction was prepared as follows. The pellet was suspended in homogenizing buffer and centrifuged for 20 min at 100,000g on a sucrose cushion (1.12 M sucrose, 20 mM Tris–HCl, pH 7.4, 1 mM EDTA). The plasma membranes (PMs), collected at the interface, were resuspended in homogenizing buffer and centrifuged at 60,000g for 10 min. The 18,000g supernatant was centrifuged for 30 min at 48,000g, yielding a pellet of high–density microsome fraction. The 48,000g supernatant was centrifuged at 400,000g for 60 min, yielding a pellet of LDM fraction and a supernatant containing cytosol. All procedures were carried out at 4 °C except for the homogenizing step, and membranes were used immediately for immunoprecipitation. The distribution of marker enzymes, 5´ nucleotidase for the plasma membranes and N–acetylglucosamine galactosyltransferase for the Golgi apparatus in the various cellular fractions, was as described26.

In vivo generation of PtdIns(3,4,5)P3.

Adipocytes were washed with phosphate–free RPMI 1640 medium plus 25 mM HEPES, pH 7.4, suspended in phosphate–free RPMI 1640 medium plus 25 mM HEPES, pH 7.4, containing [32P]orthophosphate (1 mCi/ml) and incubated for 3 h. Cells were then treated with insulin for the indicated time. The reaction was stopped by methanol:1 N HCl, 1 M NaCl (1:1) and the lipids were extracted with chloroform. For thin–layer chromatography, samples were spotted onto a silica gel plate pretreated with potassium oxalate and developed as described28.