Membrane-targeted Phosphatidylinositol 3-Kinase Mimics Insulin Actions and Induces a State of Cellular Insulin Resistance*

Phosphatidylinositol (PI) 3-kinase plays an important role in various insulin-stimulated biological responses including glucose transport, glycogen synthesis, and protein synthesis. However, the molecular link between PI 3-kinase and these biological responses is still unclear. We have investigated whether targeting of the catalytic p110 subunit of PI 3-kinase to cellular membranes is sufficient and necessary to induce PI 3-kinase dependent signaling responses, characteristic of insulin action. We overexpressed Myc-tagged, membrane-targeted p110 (p110CAAX ), and wild-type p110 (p110WT) in 3T3-L1 adipocytes by adenovirus-mediated gene transfer. Overexpressed p110CAAX exhibited ∼2-fold increase in basal kinase activity in p110 immunoprecipitates, that further increased to ∼4-fold with insulin. Even at this submaximal PI 3-kinase activity, p110CAAX fully stimulated p70 S6 kinase, Akt, 2-deoxyglucose uptake, and Ras, whereas, p110WT had little or no effect on these downstream effects. Interestingly p110CAAX did not activate MAP kinase, despite its stimulation of p21 ras . Surprisingly, p110CAAX did not increase basal glycogen synthase activity, and inhibited insulin stimulated activity, indicative of cellular resistance to this action of insulin. p110CAAX also inhibited insulin stimulated, but not platelet-derived growth factor-stimulated mitogen-activated protein kinase phosphorylation, demonstrating that the p110CAAX induced inhibition of mitogen-activated protein kinase and insulin signaling is specific, and not due to some toxic or nonspecific effect on the cells. Moreover, p110CAAX stimulated IRS-1 Ser/Thr phosphorylation, and inhibited IRS-1 associated PI 3-kinase activity, without affecting insulin receptor tyrosine phosphorylation, suggesting that it may play an important role as a negative regulator for insulin signaling. In conclusion, our studies show that membrane-targeted PI 3-kinase can mimic a number of biologic effects normally induced by insulin. In addition, the persistent activation of PI 3-kinase induced by p110CAAX expression leads to desensitization of specific signaling pathways. Interestingly, the state of cellular insulin resistance is not global, in that some of insulin’s actions are inhibited, whereas others are intact.

ample, the PI 3-kinase inhibitors, wortmannin and LY 294002 prevent GLUT4 translocation and stimulation of glucose transport in rat and 3T3-L1 adipocytes (12)(13)(14). Dominant-negative mutants of the 85-kDa subunit of PI 3-kinase can also inhibit GLUT4 translocation in response to insulin (15)(16)(17). However, several other observations suggest that, although necessary, PI 3-kinase activation is not sufficient to promote glucose transporter translocation. Indeed, growth factors such as plateletderived growth factor (PDGF) can stimulate PI 3-kinase as efficiently as insulin, but have only a minor effect on GLUT4 translocation (18,19). Similarly, interleukin 4, which induces tyrosine phosphorylation of IRS-1 and PI 3-kinase activation, does not stimulate GLUT4 translocation in L6 myoblasts (18). Furthermore, subcellular fractionation analyses indicates that insulin, unlike other growth factors, stimulates PI 3-kinase activity not only in the plasma membrane fraction but also in the low density microsomal compartment (20 -22) and possibly even in GLUT4 containing subfractions of the low density microsomal of adipocytes (23,24). Thus, it appears that insulinmediated subcompartmentalization of PI 3-kinase may be unique and might be key to the specificity of the effect of insulin on glucose transport.
The aim of this study was to determine whether targeting of PI 3-kinase catalytic subunit to membranous structures is sufficient to trigger signaling events downstream of PI 3-kinase. This allows us to directly study PI 3-kinase-regulated cellular processes in the absence of insulin and to determine whether PI 3-kinase activation is sufficient to trigger signaling events specific for insulin. Furthermore, it avoids potential problems associated with the use of PI 3-kinase inhibitors in elucidating the actions of this enzyme. We, and, others have recently demonstrated that increased PI 3-kinase activity induced by expression of a constitutively active p110 subunit (p110*) can induce GLUT4 translocation (25), but it stimulates glucose transport only partially in the absence of insulin (26). In contrast, our membrane localized form of the p110 subunit of PI 3-kinase resulted in activation of downstream mitogenesis effects in COS-7 cells (27). Since gene transfer in 3T3-L1 adipocytes by conventional methods is inefficient, in the current experiments we utilized, adenovirus-mediated, high efficiency gene transfer procedures (26,28,29), and created an adenoviral vector containing the p110-␣ subunit of PI 3-kinase incorporating a CAAX box at the COOH terminus in order to target the p110 subunit to cellular membranes. Our studies showed that expression of the membrane-targeted p110 subunit of PI 3-kinase in 3T3-L1 adipocytes were sufficient to induce PI 3-kinase dependent downstream signaling events, including glucose transport.

EXPERIMENTAL PROCEDURES
Materials-Porcine insulin was kindly provided by Lilly. Phosphospecific MAP kinase antibody and phospho-specific p70 S6 kinase antibody were from New England Biolabs, Inc. Anti-human S6 kinase and anti-IRS-1 antibody were from Upstate Biotechnology Incorporated (UBI). c-Myc antibody (9E10), Akt antibody (C-20), p110␣-CT antibody, H-Ras antibody, and horseradish peroxidase-linked anti-rabbit, mouse, and goat antibodies were from Santa Cruz Biotechnology, Inc. Erk-1 antibody and PY20H were from Transduction. Dulbecco's modified Eagle's medium (DMEM) and fetal calf serum (FCS) were obtained from Life Technologies. All radioisotopes were obtained from NEN Life Science Products Inc. (Boston, MA). XAR-5 film was obtained from Eastman-Kodak (Rochester, NY). All other reagents and chemicals were purchased from Sigma.
Cell Culture-3T3-L1 cells were grown and maintained in DMEM high glucose medium containing 50 units of penicillin/ml, 50 g of streptomycin/ml, and 10% FCS in a 10% CO 2 environment. The cells were allowed to grow 2 days postconfluency, then differentiated as described earlier (29). Prior to experimentation, the adipocytes were trypsinized and reseeded in the appropriate culture dishes. The Ad-E1A-transformed human embryonic kidney cell line 293 was cultured in DMEM high glucose medium containing 50 units of penicillin/ml, 50 units of streptomycin/ml, and 10% FCS in a 5% CO 2 environment.
Plasmid Construction-pSG5-p110 WT , and pSG5-p110 CAAX were obtained as described previously (30). The 9E10 epitope was inserted at the amino terminus of bovine p110␣ cDNA, and the CAAX motif (CK-CVLS) was inserted at the COOH terminus for pSG5-p110 CAAX . To get pAC-p110 WT and pAC-p110 CAAX , p110 WT and p110 CAAX DNA digested with BamHI were cloned into pACCMVpLpA (vector containing an ampicillin selection maker), and digested with EcoRI to determine the direction.
Preparation of Recombinant Adenovirus-The recombinant adenoviruses containing the cDNA encoding the p110 CAAX or p110 WT were isolated by homologous recombination with two plasmids, pACCMV-pLpA (31) and pJM17 (32) as described previously (29). The recombinant plasmids, pAC-p110 CAAX or pAC-p110 WT , and pJM17 were purified and co-transfected into 293 cells. Since 293 cells were originally derived from adenovirus transformation, the missing E1 gene function of pJM17 is provided in trans. The resulting recombinant viruses containing the p110 CAAX or p110 WT are denoted as Ad5-p110 CAAX or Ad5-p110 WT , respectively, and are replication defective (at least in cells lacking the E1 region of adenovirus), but fully infectious.
Detection of Recombinant Ad5-p110 CAAX and Ad5-p110 WT or Wildtype Virus in Cell Culture Medium by PCR Amplification of Viral DNA-DNA templates for PCR were extracted from the supernatant of the culture medium of the 293 cells that were infected with each plaque isolates/viruses at a multiplicity of infection (m.o.i.) of 50 plaque-forming units/ml. A multiplex PCR was performed on 1/10 dilution of virus DNA using E1A and E2B region-specific primers (33) and analyzed for the presence of recombinant or wild-type adenovirus. The presence of p110 CAAX and p110 WT cDNA inserts in the recombinant virus was confirmed by PCR analysis of viral DNA with p110␣ cDNA specific primers. One clone of each of the recombinant viruses was amplified further in 293 cells.
Cell Treatment-3T3-L1 adipocytes were transduced at a m.o.i. of 1-40 plaque-forming units/cell for 16 h with stocks of either a control recombinant adenovirus (Ad5-CT) containing the cytomegalovirus promoter, pUC 18 polylinker, and a fragment of the SV40 genome, the recombinant adenoviruses Ad5-p110 CAAX or Ad5-p110 WT . Transduced cells were incubated for 48 h at 37°C in 10% CO 2 and DMEM high glucose medium with 2% heat-inactivated serum, followed by incubation in the starvation media required for the assay. The efficiency of adenovirus-mediated gene transfer was approximately 90% as measured by immunocytochemistry. The survival of the differentiated 3T3-L1 adipocytes was unaffected by incubation of cells with the different adenovirus constructs since the total cell protein remained the same in infected or uninfected cells.
Western Blotting-3T3-L1 adipocytes uninfected or infected with Ad5-p110 CAAX , Ad5-p110 WT , or Ad5-CT were starved for 16 h in DMEM regular glucose media with 0.05% FCS. The cells were stimulated with 100 ng of insulin/ml for 5-30 min at 37°C and lysed in a solubilizing buffer containing 20 mM Tris, 1 mM EDTA, 140 mM NaCl, 1% Nonidet P-40, 50 units of aprotinin/ml, 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 50 mM sodium fluoride, pH 7.5, for 30 min at 4°C. The cell lysates were centrifuged to remove insoluble materials. For Western blot analysis, whole cell lysates (20 -80 g of protein per lane) were denatured by boiling in Laemmli sample buffer containing 100 mM dithiothreitol and resolved by SDS-PAGE. Gels were transferred to nitrocellulose by electroblotting in Towbin buffer containing 0.02% SDS and 20% methanol. For immunoblotting, membranes were blocked and probed with specified antibodies. Blots were then incubated with horseradish peroxidase-linked second antibody followed by chemiluminescence detection, according to the manufacturer's instructions (Pierce).
PI 3-Kinase Assay-3T3-L1 adipocytes were infected with Ad5-p110 CAAX , Ad5-p110 WT , or Ad5-CT at the indicated m.o.i.s for 16 h at 37°C and grown in medium containing heat-inactivated serum (2%) for 52 h. Serum-starved (16 h) cells were incubated in the absence (basal) or presence of insulin (100 ng/ml) for 10 min, washed once with ice-cold phosphate-buffered saline, lysed, and subjected to immunoprecipitation (300 -500 g of total protein) with antibodies to p110␣-CT, c-Myc (2 g), or IRS-1 (4 g) overnight at 4°C. Immune complexes were precipitated from the supernatant with protein G plus (Santa Cruz) or protein A and washed as described (34). The washed immune complexes were incubated with phosphatidylinositol (Avanti) and [␥-32 P]ATP (3000 Ci/ mmol) for 10 min at room temperature. Reactions were stopped with 20 ml of 8 N HCl and 160 ml of CHCl 3 :methanol (1:1) and centrifuged, and the lower organic phase was removed and applied to a silica gel thinlayer chromatography (TLC) plate (Merck) which had been coated with 1% potassium oxalate. TLC plates were developed in CHCl 3 :CH 3 OH: H 2 O:NH 4 OH (60:47:11.3:2), dried, and visualized, and quantitated on a Molecular Dynamics PhosphorImager.
2-Deoxyglucose Transport-The procedure for glucose transport was a modification of the methods described by Klip et al. (35). Differentiated 3T3-L1 adipocytes were infected with Ad5-p110 CAAX , Ad5-p110 WT , or Ad5-CT at the indicated m.o.i.s for 16 h at 37°C and grown in medium containing heat-inactivated serum (2%) for 72 h. Serum and glucose-deprived cells were incubated in MEM in the absence (basal) or presence of 100 ng of insulin/ml for 1 h at 37°C. Glucose uptake was determined in duplicate or triplicate at each point after the addition of 10 l of substrate (2-[ 3 H]deoxyglucose or L-[ 3 H]glucose; 0.1 Ci, final concentration 0.01 mmol/liter) to provide a concentration at which cell membrane transport is rate-limiting. The value for L-glucose was subtracted to correct each sample for the contributions of diffusion and trapping.
Glycogen Synthase Activity-Glycogen synthase activity was determined as described previously (36). Differentiated 3T3-L1 adipocytes were infected with Ad5-p110 CAAX , Ad5-p110 WT , or Ad5-CT at 40 m.o.i. for 16 h at 37°C and grown in medium containing heat-inactivated serum (2%) for 72 h. The cells were serum and glucose-starved in DMEM no glucose, 0.1% BSA, 2 mM pyruvate medium for 3 h, then stimulated with or without 200 ng of insulin/ml for 30 min in 5 mM glucose containing medium. Cells were washed with ice-cold phosphatebuffered saline three times, scraped in the buffer containing 50 mM Tris-HCl, 10 mM EDTA, 100 mM potassium fluoride, pH 7.4, and sonicated. After centrifugation, protein concentration was measured. 10 g of protein was used to determine the ability to stimulate incorporation of [ 14 C]UDP-glucose into glycogen in the presence and absence of glucose 6-phosphate.
Ras GTP/GDP Assay-Differentiated 3T3-L1 adipocytes were infected with Ad5-p110 CAAX or Ad5-CT at 30 m.o.i. for 16 h at 37°C and grown in medium containing heat-inactivated serum (2%) for 48 h. Following 24 h serum starvation, the cells were stimulated with or without insulin (100 ng/ml) for 10 min, washed with phosphate-buffered saline, scraped, and frozen at Ϫ70°C immediately. Frozen cell pellets were extracted in ice-cold RIPA buffer (50 mM Hepes, pH 7.4, 10 mM MgCl 2 , 150 mM NaCl, 1% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, and 10 g/ml aprotinin, leupeptin, and pepstatin) by shaking for 5 min at 4°C. The resulting cell extracts were centrifuged at 10,000 ϫ g for 2 min. The supernatants were divided in half and either 3 g of the anti-Ras antibody Y13-259 (Santa Cruz Biotechnology) or 3 g of rat IgG (Cappel) were added. To both samples, goat anti-rat IgG and protein G-agarose were added as well as NaCl, SDS, and deoxycholate to final concentrations of 500 mM, 0.05%, and 0.5%, respectively. The samples were shaken gently for 1 h at 4°C and then the immunoprecipitates were washed 4 times in RIPA buffer containing 500 mM NaCl, 0.05% SDS, and 0.5% deoxycholate and 2 times in 20 mM Tris-PO 4 , pH 7.4. The washed immunoprecipitates were resuspended in 30 l of 5 mM Tris-PO 4 , pH 7.4, 2 mM dithiothreitol, 2 mM EDTA (TED buffer), heated to 100°C for 3 min, cooled on ice, and centrifuged at 10,000 ϫ g for 2 min. The immunoprecipitates were washed with an additional 15 l of TED buffer which was combined with the first 30 l of TED buffer and GTP and GDP were measured as described below.
GTP was converted to ATP using NDP kinase and ADP with the ATP measured in the luciferase/luciferin system according to the following reactions, where PPi is pyrophosphate. This assay is sensitive to 1 fmol of GTP and was performed essentially as described previously (37). GDP was converted to GTP using pyruvate kinase and phosphoenolpyruvate with the GTP measured as described above, Because the final product is again emitted light, this assay is also sensitive to 1 fmol. The reaction mixture was incubated for 30 min at 30°C and contained in a final volume of 15 l of 50 mM glycine, pH 7.8, 10 mM dithiothreitol, 8 mM MgSO 4 , 50 M phosphoenolpyruvate, 3 milliunits of pyruvate kinase and 5 l of sample or GDP standard. It should be noted that this assay measures the sum of GTP ϩ GDP; thus, the amount of GTP in the sample must be subtracted from the amount of GTP ϩ GDP to yield the amount of GDP. DNA was measured by a standard fluorescence method using the fluorescent dye bisbenzimidazole and protein was measured by the Bradford method. The amounts of GDP and GTP in the samples are determined from standard curves prepared with each set of samples and the data are expressed as femtomoles of GTP or GDP per microgram of DNA or milligram of protein in the cell lysate.

Expression of Ad5-p110 CAAX and Ad5-p110 WT in 3T3-L1 Adipocytes
Western Blot of Myc-tagged Proteins-The differentiated 3T3-L1 adipocytes were infected with recombinant adenoviruses expressing the membrane-localized p110 CAAX and the wild-type p110 WT at 40 m.o.i. for 16 h and protein expression was examined 72 h later by Western blotting. Immunoblotting was performed against the Myc-tagged epitope present at the amino terminus of both the recombinant constructs. Specific bands appeared at ϳ110 kDa corresponding to the p110 CAAX and the p110 WT proteins expressed in infected 3T3-L1 adipocytes ( Fig. 1, lanes 2 and 3). The level of expression of both the proteins was similar.
PI 3-Kinase Activity-The membrane-localized p110 CAAX and the wild-type p110 WT overexpressing 3T3-L1 adipocytes were incubated with or without insulin for 10 min and the PI 3-kinase activity was measured in anti-Myc and anti-p110␣ immunoprecipitates. A representative experiment utilizing anti-p110␣ immunoprecipitates is shown in Fig. 2A and quantitation of data from seven separate experiments is shown in Fig.  2B, where the PI 3-kinase activity is expressed as percent of the basal activity (observed in unstimulated, Ad5-p110 CAAX infected cells). Overexpression of the membrane-localized p110 CAAX protein resulted in a ϳ2-fold increase in p110 ␣-associated PI 3-kinase activity when infected at 40 m.o.i. in the absence of insulin (Fig. 2B). Insulin treatment of these cells further increased PI 3-kinase activity up to 4-fold (Fig. 2B), whereas, the cells infected at the same m.o.i. with control adenovirus elicited only 2.5-fold increase in PI 3-kinase activity upon insulin stimulation. In contrast, the wild-type p110 WT overexpression induced only a modest, but a dose-dependent elevation of the p110 ␣-associated PI 3-kinase activity in the absence of insulin. Upon insulin treatment, the level of p110 ␣-associated PI 3-kinase activity increased further, up to 2.5fold in the p110 WT expressing cells, similar to that observed in uninfected or control infected cells. Preincubation with 1 M wortmannin blocked the p110-associated PI 3-kinase activity in the membrane-localized p110 CAAX overexpressing 3T3-L1 adipocytes.

Biological Effects of p110 CAAX
Ser/Thr Phosphorylation of Akt in 3T3-L1 Adipocytes-Akt is a serine/threonine kinase downstream of PI 3-kinase which is activated by serine/threonine phosphorylation. (39). Akt has been implicated as a mediator of several metabolic effects of insulin, including GLUT4 translocation, glucose uptake, and glycogen synthase activation. Therefore, we determined whether PI 3-kinase can activate Akt, using the gel mobility shift assay. Cell lysates from 3T3-L1 adipocytes infected with increasing m.o.i. of the membrane-localized p110 CAAX and the wild-type p110 WT expressing adenoviruses, were analyzed by SDS-PAGE followed by Western blotting with anti-Akt antibody (Fig. 3). The retarded gel mobility indicates serine/threonine phosphorylation and activation of Akt. Overexpression of the p110 CAAX led to a significant increase in Akt activation, in a dose-dependent manner (Fig. 3, lanes 7 and 8). The extent of Akt activation by p110 CAAX at 40 m.o.i. was comparable to that observed with insulin alone (Fig. 3, lanes 2 and 8). Further addition of insulin had a modest additive effect on Akt activation (lane 9). In contrast, p110 WT overexpression did not acti-vate Akt (lanes [3][4][5]. Expression of an empty adenoviral vector, Ad5-CT, did not affect Akt activity either in the basal or insulin stimulated state (data not shown). The activation of Akt with insulin or with p110 CAAX was completely inhibitable by treatment with wortmannin (Fig. 3, lanes 6 and 10), demonstrating that PI 3-kinase is necessary for Akt activation.
p70 S6 Kinase Activation-It has been shown that p70 S6 kinase, another serine/threonine kinase is downstream of PI 3-kinase and Akt and that activation of PI 3-kinase and/or Akt is necessary for p70 S6 kinase activation. We, therefore, exam-  5, and 9), lysed, and subjected to SDS-PAGE, and immunoblotted with anti-Akt antibody. Ser/Thr-phosphorylated Akt was detected by a retarded migration of the enzyme (pAkt). The Western blot is a representative of 10 independent experiments.
FIG. 2. Effects of overexpression of p110 CAAX and p110 WT proteins on PI 3-kinase activity in 3T3-L1 adipocytes. Cells were uninfected (Ϫ) or infected with Ad5-CT (Ctrl), Ad5-p110 CAAX (p110caax), or Ad5-p110 WT (p110WT) at the indicated m.o.i. for 16 h at 37°C and grown in medium containing heat inactivated serum (2%) for 60 h. Following infection, the cells were serum starved (16 h), incubated with or without 1 M wortmannin (Wort.), stimulated with or without 100 ng of insulin/ml for 10 min, lysed, and subjected to immunoprecipitation with antibodies to p110␣. The washed immunoprecipitates were assayed for PI 3-kinase activity with PI as substrate, and the labeled PI-3 phosphate product (PI-3P) was resolved by thin-layer chromatography and visualized by autoradiography. In A, data from a representative experiment is shown. B shows mean Ϯ S.E. of seven experiments and the data is expressed as percentage of the maximal activity (ϭ100%) observed in unstimulated Ad5-p110 CAAX -infected cells.
ined the ability of the membrane-localized p110 CAAX to activate p70 S6 kinase by using both a mobility shift assay and the phospho-specific antibody that detects p70 S6 kinase only when it is phosphorylated at Thr 421 /Ser 424 . The effect of p110 CAAX overexpression on p70 S6 kinase activation is parallel to that observed for Akt activation (Fig. 4). p110 CAAX overexpression led to an insulin-independent activation of p70 S6 kinase in a dose-dependent manner (Fig. 4, upper panel, lanes 9 -11). Insulin treatment had a small additive effect on p70 S6 kinase stimulation (lane 12). p110 WT overexpression showed a modest effect to stimulate p70 S6 kinase mobility, but the extent of activation was much less than that exhibited by the p110 CAAX protein or by insulin (lanes 4 -6). The inhibitors rapamycin or wortmannin prevented insulin or p110 CAAX stimulation of p70 S6 kinase. The phospho-specific p70 S6 kinase blot showed a similar pattern of p70 S6 kinase activation as observed with the mobility shift assay (Fig. 4, lower panel).
Glucose Transport Stimulation-3T3-L1 adipocytes were infected with Ad5-p110 CAAX , Ad5-p110 WT , or Ad5-CT and 2-[ 3 H]deoxyglucose uptake was measured 86 h later, after treating cells with or without 100 ng/ml insulin for 1 h. In uninfected cells, insulin stimulated glucose uptake by ϳ8-fold, and this was not affected by Ad5-CT infection (Fig. 5). In contrast, the basal 2-deoxyglucose uptake in cells overexpressing p110 CAAX was elevated by 4 -8-fold, compared with the basal uptake measured in cells infected with the empty adenoviral vector (Ad5-CT). Thus, glucose transport in cells expressing p110 CAAX was almost the same as that observed after insulin treatment alone (Fig. 5). In contrast, 2-deoxyglucose uptake in 3T3-L1 adipocytes infected with p110 WT was comparable with the uninfected cells, or with cells infected with the empty adenoviral vector, and was further stimulated up to 8-fold by insulin treatment. Thus, membrane-targeted p110 CAAX mimics insulin-induced glucose transport activity in 3T3-L1 adipocytes, and overexpression of p110 WT was unable to stimulate glucose uptake in the absence of insulin. Pretreatment with wortmannin inhibited p110 CAAX and insulin-induced glucose transport. In addition, p110 CAAX and p110 WT overexpression did not have any effect on the expression levels of Glut 4, compared with the CT infected cells (data not shown).
p21 ras Activation-To investigate whether activated PI 3-kinase could stimulate the Ras pathway, we examined the effects of membrane-targeted p110 CAAX on basal and insulin-induced p21 ras GTP activity in 3T3-L1 adipocytes. Insulin led to ϳ1.7fold increase p21 ras GTP activity in Ad5-CT infected cells (Fig.   6). Expression of the membrane-targeted p110 CAAX increased the level of p21 ras GTP to the same extent as insulin, and insulin treatment of the p110 CAAX expressing cells had no further effect.
These results were reproducible with another method that utilizes electroporation technique (38) to incorporate [␣-32 P]GTP in 3T3-L1 adipocytes infected with control or membrane-targeted p110 CAAX adenoviruses to measure p21 ras GTP activity. Using this system, we found that insulin stimulated p21 ras GTP activity about 1.3-fold in Ad5-CT-infected cells, in contrast to the 1.7-fold stimulation described above. However, both the methods showed similar results that the membrane-targeted PI 3-kinase stimulates Ras (data not shown).

p110 CAAX Induces Cellular Insulin Resistance
p110 CAAX Stimulates Ser/Thr Phosphorylation of IRS-1 and Inhibits Its Function-It has been reported that PI 3-kinase phosphorylates serine/threonine residues of IRS-1 (9); we, therefore, examined the effect of p110 CAAX on IRS-1 by the gel mobility shift assay. p110 CAAX expression caused a mobility shift of IRS-1 in the absence of insulin without tyrosine phosphorylation, indicating phosphorylation of Ser/Thr residues (Fig. 7A). In addition, p110 CAAX expression inhibited insulinstimulated IRS-1 tyrosine phosphorylation, without affecting insulin receptor tyrosine phosphorylation (Fig. 7B).
To investigate whether this serine/threonine phosphorylation altered IRS-1 function, we measured the PI 3-kinase activity associated with IRS-1. As shown in Fig. 7C, expression of p110 CAAX resulted in ϳ40% inhibition of insulin-stimulated PI 3-kinase activity in IRS-1 antibody immunoprecipitates. This is in contrast to the ϳ45-55% increase in p110 CAAX induced PI 3-kinase activity observed in p110␣ antibody immunoprecipitates (Fig. 7C). The p110 CAAX induced inhibition of insulin stimulated IRS-1 associated PI 3-kinase activity was also observed in the cell lysates immunodepleted of the c-Myc immune complexes (Fig. 7C).
Glycogen Synthase Activity-Insulin led to an about 2.5-fold increase in glycogen synthase activity measured in control 3T3-L1 adipocytes infected with the empty adenoviral vector, Ad5-CT (Fig. 8). In cells infected with p110 CAAX , the basal level of glycogen synthesis was decreased by ϳ30% compared with control cells, while insulin-stimulated glycogen synthase activity was completely inhibited. In cells expressing p110 WT , basal glycogen synthase activity was unchanged and insulin led to an about 1.5-fold stimulation; this degree of stimulation was less  1, 3, 7, and 12). Total cell lysates (30 g) were subjected to SDS-PAGE and immunoblotted with p70 S6 kinase antibody (upper panel) or phospho-specific p70 S6 kinase antibody (lower panel). The Western blot is representative of six independent experiments. than that observed in control cells, but the differences did not reach statistical significance.
MAP Kinase Phosphorylation-We investigated whether the Ras activation induced by p110 CAAX expression was accompanied by activation of the downstream effectors Erk1 and 2. This was accomplished in uninfected cells and Ad5-p110 CAAX -infected cells by assessing MAP kinase activation using a phospho-specific MAP kinase antibody (Fig. 9, upper panel). Insulin increased MAP kinase activation in uninfected and Ad5-CTinfected cells by ϳ15-fold. Expression of the membrane-targeted p110 CAAX had no effect on the basal MAP kinase phosphorylation state, and almost completely inhibited insulininduced phosphorylation and activation of both p44 and p42 MAP kinase (Fig. 9, lane 5). In contrast, PDGF stimulated MAP kinase phosphorylation was unchanged by p110 CAAX expression (Fig. 9, lane 6). Expression of MAP kinase, as assessed by Western blotting with a polyclonal anti-Erk-1 antibody that recognizes both nonphosphorylated and phosphorylated forms, demonstrated that the protein levels were not altered by infection with either Ad5-CT or Ad5-p110 CAAX (Fig. 9, lower panel). DISCUSSION In this study, we have used adenoviral-mediated gene transfer to assess whether targeting of the catalytic p110 subunit of PI 3-kinase to cellular membranes by incorporating a CAAX box at the COOH terminus (p110 CAAX ) is sufficient to induce PI 3-kinase dependent signaling responses, characteristic of insulin action, in 3T3-L1 adipocytes. We show that when appropriately targeted, even modest levels of PI 3-kinase are sufficient to trigger full activation of the downstream serine/threonine kinases, Akt and p70 S6 kinase, and also causes stimulation of glucose transport equal to the effect of insulin. Surprisingly, insulin-mediated glycogen synthase activity was completely blocked in cells expressing p110 CAAX . Furthermore, p110 CAAX stimulated serine/threonine phosphorylation of IRS-1, and inhibited IRS-1 associated PI 3-kinase activity. Another major finding is that the membrane-localized PI 3-kinase activity was sufficient to mimic insulin-induced formation of GTP-bound p21 ras . Last, we found that expression of p110 CAAX led to inhibition of insulin-mediated MAP kinase activation, whereas PDGF-mediated MAP kinase activation was unaffected. These results lead to several predictions and conclusions.
p110 CAAX Mimics Insulin Actions-We demonstrate that membrane-targeted p110 (p110 CAAX ) promotes insulin-independent PI 3-kinase activity and is sufficient to maximally stimulate glucose uptake, in a wortmannin-sensitive manner. The level of 2-deoxyglucose uptake achieved in response to p110 CAAX expression was comparable to that seen in insulinstimulated, control adipocytes, whereas, non-targeted wildtype p110␣ (p110 WT ) had only a slight effect on 2-deoxyglucose uptake. Since, the p110 subunit of PI 3-kinase contains a COOH-terminal membrane targeting farnesylation sequence, it seems likely that the overexpressed p110 CAAX protein results in increased PI 3-kinase activity predominantly in membrane fractions, similar to insulin stimulation of endogenous PI 3-kinase. This implies that glucose uptake is not merely a function of the amount of PI 3-kinase present, but that its appropriate membrane localization is critical as well. This conclusion is quite consistent with other kinds of studies in the literature. For example, PDGF, as well as other growth factors, can stimulate PI 3-kinase in 3T3-L1 adipocytes, equally well as insulin, but only insulin leads to glucose transport stimulation (18,19). These findings suggested that insulin induced subcompartmentalization of PI 3-kinase is necessary for metabolic signal- ing. Consistent with this, Frevert and Khan (26) showed that co-expression of both the catalytic p110␣ subunit and the inter-SH2 region of the p85 regulatory subunit of PI 3-kinase in 3T3-L1 adipocytes led to a much higher level of PI 3-kinase activity than seen with insulin stimulation alone, but it had only a partial effect to stimulate glucose transport without insulin. In addition, when PI 3-kinase was activated by thiophosphorylated peptides, corresponding to the phosphotyrosine binding motif of the p85 subunit of PI 3-kinase, only a minor effect on Glut 4 translocation was observed (40). Tanti et al. (41) co-transfected rat adipose cells by electroporation with epitope-tagged Glut 4 and with either a constitutively active (p110*) or a kinase inactive form of p110 kd (41). Co-transfection with the active version of p110* resulted in stimulation of epitope-tagged Glut 4 translocation, similar to insulin, and these workers also found that the p110* was localized to the same intracellular compartment as the endogenous PI 3-kinase. Taken together with our current results, these studies support the conclusion that active PI 3-kinase is sufficient to stimulate glucose transport activity, only if it is targeted to the proper subcellular membranous compartment.
We further examined other targets of insulin action which are thought to be downstream of PI 3-kinase. Akt is a serine/ threonine kinase that is activated by insulin. It is activated by a dual mechanism involving the binding of PI-(3,4)-P2 to its PH domain, as well as by serine/threonine phosphorylation by one or more Akt kinases, which may, themselves, be stimulated by the lipid products of PI 3-kinase (42). Several lines of evidence suggest that Akt functions downstream of PI 3-kinase, e.g. insulin stimulated Akt kinase activity is inhibitable by wortmannin, a PI 3-kinase specific inhibitor, and PDGF receptor FIG. 7. Membrane-targeted p110 CAAX stimulates Ser/Thr Phosphorylation of IRS-1 and inhibits IRS-1 associated PI 3-kinase activity. Differentiated 3T3-L1 adipocytes were infected with Ad5-CT (ctrl) or Ad5-p110 CAAX (p110caax) at 40 m.o.i. in medium containing heat-inactivated serum (2%) for 16 h. Following infection, cells were serum starved (16 h), incubated in the absence or presence of insulin (100 ng/ml) for 5 (A and B) or 10 (C) min. Total cell lysates (20 g) were subjected to SDS-PAGE and immunoblotted with IRS-1 antibody (A) or PY20H (B). The Western blot is representative of three independent experiments. The cell lysates were divided into three, and subjected to immunoprecipitation with antibodies to p110␣, IRS-1, or Myc (C). After Myc antibody immunoprecipitation, the supernatants were collected, and immunoprecipitated with IRS-1 antibody. The washed immunoprecipitates were assayed for PI 3-kinase activity with PI as substrate, and the labeled PI-3 phosphate product (PI-3P) was resolved by thin-layer chromatography and visualized by autoradiography. C shows mean Ϯ S.E. of three experiments and the data is expressed as percentage of the maximal activity (ϭ100%) observed in insulin-stimulated, Ad5-CT-infected cells. Ⅺ, Ad5-Ctrl; f, Ad5-CAAX. mutants that fail to activate PI 3-kinase, also fail to activate Akt. Overexpression of a constitutively active Akt in 3T3-L1 adipocytes results in increased glucose uptake and Glut 4 translocation in the absence of insulin (43). Consistent with these findings, our data show that Akt activation is dependent on PI 3-kinase activity, and that insulin and p110 CAAX induced Akt activation is inhibitable by wortmannin, indicating that Akt activation is dependent on PI 3-kinase enzymatic function.
We also determined whether PI 3-kinase-mediated Akt activation would lead to p70 S6 kinase stimulation, since it has been shown that p70 S6 kinase is stimulated by constitutively active Akt (8) and blocked by inhibitors of PI 3-kinase (12,44). Indeed, we found that overexpression of the membrane-targeted p110 CAAX led to activation of p70 S6 kinase, which was completely inhibitable by wortmannin (Fig. 4). Our results are supported by the study of Weng et al. (44), which showed that transfection of a constitutively active form of PI 3-kinase (p110*) into 293 cells resulted in a 20 -30-fold increase in cellular PI 3-kinase activity, that resulted in activation of p70 S6 kinase by phosphorylation at Thr-252. Wortmannin resulted in selective dephosphorylation at Thr-252 concomitant with inhibition of p70 S6 kinase activity. Furthermore, Klippel et al. (27) reported that in COS-7 cells, expression of membrane-localized p110 is sufficient to trigger downstream responses, characteristic of insulin action, including stimulation of Akt and p70 S6 kinase. Their study further adds that these responses can also be triggered by expression of p110*, that is cytosolic, but exhibits a high specific activity. However, they observed maximum activation of downstream responses in cells expressing the membrane-localized p110. Thus, insulin treatment activates and targets PI 3-kinase to specific membrane compartments, and this action is mimicked by p110 CAAX , which is sufficient to trigger downstream responses characteristic of insulin action, including stimulation of Akt, p70 S6 kinase, and glucose transport.
The role of PI 3-kinase in Ras-mediated signaling is unclear. An association between p21 ras and PI 3-kinase was first demonstrated by co-immunoprecipitation in insulin and insulinlike growth factor-1 stimulated, Ras-transformed epithelial cells by Sjolander et al. (45). Subsequently, Ras was shown to bind in vitro to the p110 subunit of PI 3-kinase by Rodriguez-Viciana et al. (46). However, the relative position of PI 3-kinase with respect to Ras is confusing. Conflicting data exists suggesting that PI 3-kinase could be upstream, downstream, or independent of Ras. These alternate results are perhaps related to cell-type differences. Rodriguez-Viciana et al. (30) reported that a point mutation of the p110 subunit of PI 3-kinase at the Ras-GTP binding site elevated PI 3-kinase activity in COS cells, and the interaction of Ras-GTP, but not Ras-GDP, with PI 3-kinase led to an increase in its enzymatic activity (30). These data suggest that Ras is upstream of PI 3-kinase. However, our data are consistent with the idea that PI 3-kinase is upstream and can activate Ras (Fig. 6). We find that membrane-targeted activated PI 3-kinase activates p21 ras , resulting in increased formation of p21 ras GTP, equal to the effect of insulin. This interpretation is in agreement with earlier data from our own laboratory, in which we reported that microinjection of dominant-negative PI 3-kinase, or PI 3-kinase inhibitory antibodies, into rat fibroblasts inhibited insulin-induced Fos induction, which was rescued by activated (T-24) Ras (47). Similarly, studies by Hu et al. (48) suggest that Ras is downstream of PI 3-kinase because transfection of constitutively active PI 3-kinase resulted in Fos induction, which was blocked by both dominant-negative Ras and Raf. They also found elevated levels of GTP-bound Ras in cells transfected with constitutively active PI 3-kinase. Cellular Insulin Resistance Induced by p110 CAAX -Interestingly, we found that p110 CAAX did not mimic all of insulin's actions, and, in some cases led to a decrease in insulin signaling indicating a partial, and selective insulin resistant state. For example, we found that p110 CAAX did not mimic the effect of insulin to stimulate glycogen synthesis. Not only did p110 CAAX expression fail to enhance basal glycogen synthase activity, but it completely inhibited the ability of insulin to stimulate glycogen synthesis. Activation of glycogen synthase by insulin involves a coordinated response, including phosphorylation induced inactivation of glycogen synthase kinase 3 (GSK3) and activation of protein phosphatase 1, by phosphorylation of its G subunit (pp1G) (49). It has been suggested that GSK3 is a downstream target of Akt, which, in turn, is dependent on PI 3-kinase activity. Constitutively active Akt inhibits insulin's ability to stimulate glycogen synthesis in 3T3-L1 adipocytes (43,50,51), and our data also show that activation of Akt by the membrane-localized p110 CAAX is not sufficient to cause glycogen synthase activation in 3T3-L1 adipocytes. In theory, activated PI 3-kinase and Akt should inactivate GSK3 by phosphorylation leading to stimulation of glycogen synthase activity, whereas we, and others, show that p110 CAAX or constitutively active Akt inhibits insulin effects on this enzyme (43,50,51). However, it has been shown recently that GSK3 expression is either very low, or absent in 3T3-L1 adipocytes (52)(53)(54). Therefore, a role for GSK3 in our results is problematic. Perhaps the low (or absent) expression of GSK3 explains why p110 CAAX does not stimulate glycogen synthase by itself. An alternate pathway for glycogen synthase activation involves pp1, which has been suggested to be downstream of the IRS-1/Shc-MAP kinase pathway by some investigators (55), but a number of reports have indicated that this is not the case (56). In addition, earlier results show that the MEK inhibitor PD098059 does not lead to a decrease in insulin stimulation of glycogen synthesis (57). Thus, a role for MAPK in the regulation of glycogen synthesis seems unlikely. Another possibility is that IRS-1 directly or through its interacting proteins, but independent of PI 3-kinase, might be involved in the inhibition of glycogen synthase activity. Indeed, we find that membrane-targeted p110 CAAX serine/threonine phosphorylates IRS-1, which is inhibitable by wortmannin. This in turn prevents IRS-1 tyrosine phosphorylation and downstream signaling (Fig. 7).
Although the precise mechanisms underlying the p110 CAAX induced resistance are unknown, the current results provide some interesting insights. First, despite the fact that p110 CAAX stimulated Ras activation, it had no effect to stimulate MAP kinase phosphorylation, indicating a blockade of MAP kinase activation at a site downstream of Ras. Furthermore, in p110 CAAX expressing cells, insulin had no effect to stimulate MAP kinase phosphorylation, compared with a robust stimulation in control cells. Since insulin is thought to stimulate MAP kinase activation by activation of Ras (11), these findings also point to a post-Ras blockade of the MAP kinase pathway. On the other hand, p110 CAAX expression did not inhibit PDGFstimulated MAP kinase phosphorylation, and this is consistent with the interpretation that PDGF can lead to MAP kinase activation through a Ras-dependent as well as a non-Ras dependent pathway (58), and we would propose that expression of p110 CAAX inhibits only the Ras-dependent input into MAP kinase. These findings also demonstrate that the p110 CAAX induced inhibition of MAP kinase and insulin signaling is specific, and not due to some toxic or nonspecific effect on the cells.
Taken together, our results are consistent with the view that p110 CAAX expression inhibits the actions of insulin at a step distal to Ras activation, leading to inhibition of MAP kinase, and, possibly, glycogen synthase activation. Importantly, the cellular insulin resistance induced by p110 CAAX in these cells is not global. Thus, p110 CAAX expression stimulated AKT as well as p70 S6 kinase phosphorylation, and insulin had a further effect when added to p110 CAAX expressing cells. This would argue that this model of cellular insulin resistance is rather unique, in that some of the insulin signaling pathways are inhibited, whereas, others are intact. The fact that persistent activation of PI 3-kinase leads to desensitization of subsequent downstream events is reminiscent of the fact that hyperinsulinemia (either in vitro or in vivo) will also lead to a state of cellular insulin resistance. However, hyperinsulinemia-induced insulin resistance affects all of insulin's actions, whereas, persistent PI 3-kinase activation selectively inhibits specific insulin signaling. Since insulin's biologic effects are pleiotrophic with engagement of multiple divergent signaling pathways, further study of these cells may enhance our understanding of which signaling pathways connect to which biologic effects.
In summary, our studies show that PI 3-kinase activity can mimic a number of biologic effects normally induced by insulin, but that membrane targeting of this enzyme is necessary for activation of these events. In addition, the persistent activation induced by p110 CAAX expression leads to desensitization of specific signaling pathways. Interestingly, the state of cellular insulin resistance is not global, in that some of insulin's actions are inhibited, whereas others are intact.