p85α Gene Generates Three Isoforms of Regulatory Subunit for Phosphatidylinositol 3-Kinase (PI 3-Kinase), p50α, p55α, and p85α, with Different PI 3-Kinase Activity Elevating Responses to Insulin

Phosphatidylinositol 3-kinase (PI 3-kinase) is stimulated by association with a variety of tyrosine kinase receptors and intracellular tyrosine-phosphorylated substrates. We isolated a cDNA that encodes a 50-kDa regulatory subunit of PI 3-kinase with an expression cloning method using 32P-labeled insulin receptor substrate-1 (IRS-1). This 50-kDa protein contains two SH2 domains and an inter-SH2 domain of p85α, but the SH3 and bcr homology domains of p85α were replaced by a unique 6-amino acid sequence. Thus, this protein appears to be generated by alternative splicing of the p85α gene product. We suggest that this protein be called p50α. Northern blotting using a specific DNA probe corresponding to p50α revealed 6.0- and 2.8-kb bands in hepatic, brain, and renal tissues. The expression of p50α protein and its associated PI 3-kinase were detected in lysates prepared from the liver, brain, and muscle using a specific antibody against p50α. Taken together, these observations indicate that the p85α gene actually generates three protein products of 85, 55, and 50 kDa. The distributions of the three proteins (p85α, p55α, and p50α), in various rat tissues and also in various brain compartments, were found to be different. Interestingly, p50α forms a heterodimer with p110 that can as well as cannot be labeled with wortmannin, whereas p85α and p55α associate only with p110 that can be wortmannin-labeled. Furthermore, p50α exhibits a markedly higher capacity for activation of associated PI 3-kinase via insulin stimulation and has a higher affinity for tyrosine-phosphorylated IRS-1 than the other isoforms. Considering the high level of p50α expression in the liver and its marked responsiveness to insulin, p50α appears to play an important role in the activation of hepatic PI 3-kinase. Each of the three α isoforms has a different function and may have specific roles in various tissues.

A variety of growth factors and hormones mediate their cellular effects via interactions with cell surface receptors that possess protein kinase activity (1,2). The interaction of most of these ligands with their receptors induces tyrosine kinase activation and autophosphorylation of the receptor, resulting in physical association of these receptors with several cytoplasmic substrates having SH2 domains. Phosphatidylinositol 3-kinase (PI 3-kinase) 1 has been identified through its ability to associate with cellular protein kinases, including numerous growth factor receptors and oncogene products (3,4). This lipid kinase phosphorylates phosphatidylinositol at the D-3 position of the inositol ring in response to stimulation with a variety of growth factors and hormones (5). Although the role of this lipid product in cellular regulation remains unclear, recent reports suggest that the activation of PI 3-kinase leads to the activation of c-Akt, Rac, PKC-␥ isoform, and p70 S6 kinase (6 -9). As a result, PI 3-kinase has been suggested to play essential roles in the regulation of various cellular activities, including proliferation (10,11), differentiation (12), membrane ruffling (13), prevention of apoptosis (14), and insulin-stimulated glucose transport (10,15,16).
PI-3 kinase is composed of a catalytic 110-kDa protein (p110) associated with a regulatory subunit (17)(18)(19). The regulatory subunit contains two proline-rich motifs, two Src homology-2 (SH2) domains, and a domain responsible for the binding with p110 between the two SH2 domains (3,4). Many activated receptors with tyrosine kinase activity interact with the SH2 domain in the regulatory subunit through phosphorylated YXXM motifs in the receptors themselves (20), resulting in the activation or recruitment of PI 3-kinase (21). To date, four regulatory subunits of PI 3-kinase have been identified, two 85-kDa proteins (p85␣, p85␤) and two 55-kDa proteins (p55␣/ p85/AS53, p55␥/p55 PIK ) (22)(23)(24). The two recently cloned 55-kDa regulatory subunits, p55␣ and p55␥, are unique because the SH3 and bcr homology domains found in p85s are replaced by a unique 34-amino acid residue NH 2 terminus. In this study, we screened a rat liver cDNA library using a 32 P-labeled human IRS-1 protein and obtained a cDNA that encodes a novel 50-kDa regulatory subunit for PI 3-kinase. Sequence analysis of the cDNA revealed that this protein consists of a unique 6amino acid sequence at its NH 2 terminus, as well as two SH2 domains and an inter-SH2 domain of p85␣. Neither the SH3 and bcr homology domains, of the p85 regulatory subunit, nor the unique 34-amino acid residue, of the p55 regulatory subunit, were found in this 50-kDa protein. These sequence data indicate that this 50-kDa protein is generated by alternative splicing of the p85␣ gene product. We suggest that this protein be called p50␣.
In total, five regulatory subunits for PI 3-kinase have been identified in mammalian cells to date, including two 85-kDa proteins, two 55-kDa proteins, and one 50-kDa protein.
In this study, we demonstrated the tissue distributions and different roles in PI 3-kinase activation, via insulin stimulation, of these subunits. Our data suggest that these five regulatory subunits may have different roles in the various responses induced by the numerous growth factors, hormones, and oncogene products with which they interact.

EXPERIMENTAL PROCEDURES
Expression Screening of Rat Liver cDNA Library with Human 32 P-IRS-1 Protein-The recombinant human IRS-1 protein was prepared as described previously (24). The insulin receptor was partially purified from human placenta on wheat germ agglutinin-agarose as described previously (25). The 32 P-IRS-1 probe was prepared by incubating IRS-1 with activated insulin receptor in the presence of Mn 2ϩ and [␥-32 P]ATP (24). An oligo(dT)-primed rat liver cDNA library was prepared in UNI-ZAP XR (Stratagene) according to the manufacturer's instructions. Sixty 15-cm plates representing 3,000,000 independent plaques were plated and incubated for 7 h at 37°C. Then the plates were overlaid with nitrocellulose filters that had been impregnated with 10 mM isopropyl-␤-D-thiogalactopyranoside and incubated for 8 h at 37°C. The hybridization of the filters with 32 P-IRS-1 probe and washing were performed as described previously (24). The cDNA inserts in pBluescript were prepared by in vivo excision according to the manufacturer's instructions (Stratagene). The nucleotide sequences were determined using an ABI automatic sequencer.
Northern Blotting-Northern blotting was performed using a commercially available sheet made by Clontech (Palo Alto, CA). Nucleotides Ϫ170 -18 of p50␣ were labeled with [␣-32 P]dCTP and used as probes. The filter was hybridized and washed according to the manufacturer's instructions (Clontech). Autoradiography was performed at Ϫ80°C for 24 -48 h.
Overexpression of Regulatory Subunits in Sf9 Insect Cells-The cDNAs encoding the full-length amino acid sequences of p85␣, p85␤, p55␣, p55␥, and p50␣, as well as the HA tag amino acid sequence (YPYDVPDYA) at each COOH terminus were subcloned into pBacPAK9, and the baculoviruses were prepared according to the manufacturer's instructions (Clontech). The Sf9 cells, infected with a baculovirus containing one of the five isoforms, were cultured for 48 h and then lysed in Laemmli buffer. The samples were subjected to SDS-PAGE, and immunoblotting was performed using the antibody against the HA tag, or specific antibodies, as described previously (26).
Immunoblotting of Regulatory Subunits (p85␣, p55␣, and p50␣) of PI 3-Kinase Expressed in Various Rat Tissues-Tissues from male Wistar rats (6 -8 weeks old) were homogenized in ice-cold lysis buffer (1/10 w/v) containing 50 mM Hepes (pH 7.5), 137 mM NaCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 2 mM EDTA, 1% Nonidet P-40, 10% glycerol, 2 mg/ml aprotinin, and 34 mg/ml phenylmethylsulfonyl fluoride. Insoluble material was removed by centrifugation at 14,000 ϫ g for 60 min and incubated for 2 h at 4°C with the antibody against a whole molecule of p85␣ (␣p85 PAN-UBI ) covalently coupled to protein A-Sepharose beads purchased from UBI. The beads were washed three times in lysis buffer A, consisting of 10 mM Tris-HCl (pH 7.8), 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, and 0.5 mM PMSF, and separated from the beads by boiling in Laemmli buffer. The beads were removed by centrifugation, and the supernatants were subjected to SDS-PAGE, and immunoblotting was performed as described previously (26). To detect the PI 3-kinase activities, the immunoprecipitations were performed using the appropriate antibodies, and the PI 3-kinase activities in each immunoprecipitate were measured.
Preparation of p50␣ cRNA Probes and in Situ Hybridization Histochemistry-An antisense RNA probe was prepared by in vitro transcription from the fragment of p50␣ cDNA that had been used for Northern blot analysis, using T7 RNA polymerase, in the presence of 350 mM digoxigenin (DIG)-linked UTP in a 20-l reaction mixture, according to the manufacturer's instructions (Boehringer Mannheim). A sense probe was similarly obtained with T3 RNA polymerase.
Male Wistar rats, 4 weeks of age (supplied by the Animal Breeding Facility, Gunma University), were anesthetized with an intraperitoneal injection of sodium pentobarbital and perfused through the aorta with an isotonic sodium chloride solution to remove blood, followed by 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4. Brains were then removed, washed with phosphate-buffered saline, pH 7.4, dehydrated through graded alcohols, and embedded in paraplast wax. Serial sections (10 M thick) were cut, mounted on poly-L-lysine-coated slides, and stored at room temperature until use.
After de-waxing and rehydration, tissue sections were digested with 5 g/ml proteinase K at room temperature for 20 min. They were then fixed in 0.4% paraformaldehyde at 4°C for 10 min and incubated at 50°C overnight with a hybridization solution containing 1 g/ml DIGlabeled cRNA, 50% formamide, 10 mM Tris-HCl, pH 7.5, 300 mM NaCl, 1 mM EDTA, 0.25% SDS, 1 ϫ Denhardt's solution, 200 g/ml yeast tRNA, and 10% dextran sulfate. Following hybridization, the sections were washed in 2 ϫ SSC, 50% formamide at 58°C for 30 min, incubated in 1 g/ml RNase A solution at 37°C for 30 min, and washed once in 2 ϫ SSC and twice in 0.2 ϫ SSC at 50°C for 20 min each time. The sections were then incubated in a 1:500 diluted solution of polyclonal sheep anti-DIG Fab antibody conjugated with alkaline phosphatase, before washing and detection of the label with nitro blue tetrazolium chloride and 5-bromo-4-chloro-3-indolylphosphate. Color development was carried out at room temperature for 14 h, and the sections were then washed in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 7.5). To inhibit further color reaction, the slides were mounted in a mixture containing 24% polyvinyl alcohol, 12% glycerol, and 59 mM Tris-HCl, pH 7.5.
Chromatography of Rat Liver and Brain PI 3-Kinase on a DEAE-Sepharose Column and Gel Filtration Chromatography-The insoluble materials prepared from rat liver and brain (30 ml, 20 and 8 mg/ml, respectively) were directly loaded onto a DEAE-Sepharose Fast Flow column (200 ml). The column was then washed with buffer B, consisting of 20 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM dithiothreitol, 1 M leupeptin, 0.5 mM PMSF, and 10% glycerol. Once the absorbance eluted at 280 nm had returned to base line, the column was eluted with a gradient of KCl up to 0.5 M. Aliquots of the individual fractions were assayed for PI 3-kinase activity. The pooled fractions from the peaks (three fractions from liver chromatography, fractions A, B, and C, and one fraction from brain chromatography, fraction D) were concentrated 10-fold with Centriprep 10 (Amicon, MA) and subjected to immunoblotting analysis. Concentrated fractions A and B (1 ml) were also applied to the gel filtration column (Sephacryl S300 column, Pharmacia Biotech Inc.), previously equilibrated with buffer B, and eluted fractions (1 ml) were collected and assayed for PI 3-kinase activity. Molecular weight standards were run under the same conditions as samples.
PI-3 Kinase Assay-Eluted fractions from the chromatography column or the lysates from the cells were incubated with the appropriate antisera for 2 h at 4°C. Protein A-Sepharose beads were used to precipitate the immune complexes. The presence of PI-3 kinase activity in immunocomplexes was determined as described previously (24).
Detection of p110 by Wortmannin Labeling-All concentrated fractions from the peaks were immunoprecipitated with the appropriate antibodies and pelleted using protein A-Sepharose beads. The beads were washed three times with lysis buffer A. The beads were then incubated at 25°C for 30 min with 20 mM Tris-HCl, pH 7.4, 0.5 mM EDTA, 0.5 mM EGTA, and 0.1 M wortmannin and washed again three times with lysis buffer A. Wortmannin-labeled immunoprecipitates were separated from the beads by boiling in Laemmli buffer. The beads were removed by centrifugation, and the supernatants were subjected to SDS-PAGE. The wortmannin was detected by enhanced chemiluminescence using anti-wortmannin antibody, generously provided by Dr. Yuzuru Matsuda (Kyowa Hakko Co., Japan), and horseradish peroxidase-labeled anti-rat IgG (Amersham Corp.).
Cell Culture-PC12 and HepG2 cells were grown in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. CHO cells were transfected with the expression vector pCAGGS containing the fulllength human insulin receptor or human IRS-1 cDNA. CHO cells stably overexpressing human insulin receptor (CHO/IR) or human IRS-1 (CHO/IRS-1) were obtained after selection with G-418 (400 g/ml) and grown in Ham's F12 medium with 10% fetal bovine serum. Prior to insulin treatment, the medium was replaced with serum-free medium containing 0.1% bovine serum albumin, and incubation was continued for 18 h. Various concentrations of insulin were added to the medium, and the cells were incubated for 5 min at 37°C. After insulin treatment, the medium was removed, and cells were collected with lysis buffer C, consisting of 50 mM Hepes, pH 7.5, 137 mM NaCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 2 mM Na 3 VO 4 , 10 mM sodium pyrophosphate, 10 mM NaF, 2 mM EDTA, 1% Nonidet P-40, 10% glycerol, 2 g/ml aprotinin, 5 g/ml leupeptin, 0.5 M pepstatin A, and 34 g/ml PMSF. Lysates were centrifuged at 10,000 ϫ g for 10 min. The resulting supernatants were assayed for PI 3-kinase activity.

Transient Expression of Regulatory Subunits of PI 3-Kinase with an Adenovirus Expression
System-The cassette cosmid for constructing recombinant adenovirus, pAdex1wt, was the generous gift of Dr. Izumi Saito (Institute of Medical Science, University of Tokyo). The cDNAs encoding the full-length amino acid sequences of p85␣, p85␤, p55␣, p55␥, and p50␣, as well as the HA tag amino acid sequence (YPYDVP-DYA) at each COOH terminus, were ligated into the SwaI sites of pAdex1wt. Recombinant adenoviruses were obtained as described previously (27). HepG2, CHO/IR, and CHO/IRS-1 cells were infected with these viruses for 1 h and then grown for 48 h. The expression of each protein was confirmed by immnoblotting wih anti-HA antibody.
[ 35 S]Methionine Labeling of HepG2 Cells Expressing p85␣, p85␤, p55␣, p55␥, or p50␣-HepG2 cells were infected with an adenovirus expressing the full-length cDNA of p85␣, p85␤, p55␣, p55␥, or p50␣ and grown for 48 h. Prior to insulin treatment, the medium was replaced with methionine-free RPMI medium, containing 0.5% bovine serum albumin and 0.1 mCi/ml Tran 35 S-label (ICN) and incubated for 18 h. Then, 10 Ϫ6 M insulin was added to the medium, and cells were incubated for 5 min at 37°C. After insulin treatment, the medium was removed, and the cells were collected with lysis buffer C. Lysates were centrifuged at 10,000 ϫ g for 10 min. The resulting supernatants were immunoprecipitated with the anti-HA antibody and then pelleted with protein G-Sepharose beads. Immunoprecipitates were separated from the beads by boiling in Laemmli buffer. The beads were removed by centrifugation, and the supernatants were electrophoresed on 7.5% SDS-polyacrylamide gels. The gels were dried and subjected to autoradiography.

RESULTS
A novel form of the regulatory subunit of PI 3-kinase was isolated from a rat liver cDNA expression library. A rat liver cDNA expression library was screened with 32 P-labeled recombinant IRS-1, and 47 positive independent clones were isolated after three or four rounds of screening. These clones included cDNAs containing complete coding regions of p85␣ and p85␤, the nucleotide sequences of which had previously been determined (22). In addition, we obtained three independent cDNAs containing the nucleotide sequence coding the NH 2 -terminal SH2 domain of p85␣ and a previously undocumented 190nucleotide sequence at its 5Ј-upstream side. These cDNAs contained an open reading frame of 1275 nucleotides, and the deduced amino acid sequence is shown in Fig. 1A. The presence of this mRNA in rat liver was confirmed by reverse transcription-polymerase chain reaction using the 5Ј-primer in the newly identified nucleotide sequence and the 3Ј-primer in the COOH-terminal sequence of p85␣ (data not shown). The deduced amino acid sequence contains 6 unique amino acids at the NH 2 -terminal head of the proline-rich domain and two SH2 domains, which are identical to those of p85␣. Thus, this mRNA, as well as p55␣ mRNA, appears to be transcribed via alternative splicing of the p85␣ gene product. We designated this protein p50␣. The splicing site of these two variants is assumed to be the same, although the part of the p55␣ cDNA that is identical to that of p85␣ is longer than that of p50␣ by two amino acids.
Expression of the Five Regulatory Subunit Isoforms in Sf9 Cells and Preparation of Specific Antibodies-The five cDNAs coding for p85␣, p85␤, p55␣, p55␥, and p50␣, with the HA tag at their COOH termini, were subcloned into the expression vector, and baculoviruses recombined with these cDNA were prepared. Sf9 cells were infected with these baculoviruses, and cell lysates were immunoblotted with anti-HA antibody (Fig.  1C, panel a) or specific antibodies against each isoform (Fig.  1C, panels b-f). Bands corresponding to p50␣ were observed using either the anti-HA antibody or the anti-p50␣ specific antibody (␣p50␣), with an electric mobility of approximately 50 kDa (Fig. 1C, panel f, lane 6). The results shown in Fig. 1C, panels b-f, indicate that none of the specific antibodies recognize other regulatory subunit isoforms. We were thus able to measure the PI 3-kinase activity associated with each of the regulatory subunit isoforms expressed endogenously in tissues or cell lines.
p50␣ mRNA Is Most Abundant in Liver, but Is Also Abundant in Brain and Kidney-The levels of p50␣ mRNA expression in various rat tissues are shown in Fig. 2A. Northern blotting with a 5Ј-unique 188-nucleotide sequence located in the 5Ј-untranslated region and a coding region for the NH 2terminal 6 amino acid sequence in the p50␣ cDNA, neither of which is included in the p85␣ cDNA nucleotide sequence, revealed two mRNA species of 6.0 and 2.8 kb. The p50␣ mRNA was most abundant in liver but was also abundant in the brain and kidney. Northern blotting using the cDNA probe coding for the N-terminal SH2 (N-SH2) domain of p85␣ revealed four bands (7.7, 6.0, 4.2, and 2.8 kb), as previously reported (22) (Fig. 2B). Among them, the 6.0-and 2.8-kb bands matched those of p50␣, whereas the 7.7-and 4.2-kb bands matched those of the SH3 p85␣ domain (22). As we reported previously, a minor portion of the 4.2-kb band and a considerable portion of the 2.8-kb band in the brain correspond to p55␣ mRNAs (22). The quantities of p85␣ mRNA and p50␣ mRNA in liver appeared to be similar, judging from the Northern blotting data.
Although the role of PI 3-kinase in brain and neural cells remains unclear, all five known isoforms are abundantly expressed in brain tissue (Fig. 2B) (22). We further investigated their distributions in various portions of the rat brain using an RNase protection assay. As shown in Fig. 3E, p50␣ mRNA was abundant in the cerebral cortex (temporal and occipital), putamen, and cerebellum. Although minor differences were observed, the distributions of p85␤ and p55␣ were similar to that of p50␣ (Fig. 3, B and D). p85␣ mRNA was also abundant in the superior colliculus and brainstem (pons and medulla oblongata) but was detected in every part of the brain (Fig. 3A). On the other hand, the expression of p55␥ mRNA was particularly prominent in the cerebellum, while being barely detectable in other parts of the brain (Fig. 3C). As to the cerebellum, in situ hybridization histochemistry revealed p50␣ transcripts to be most abundant in the cytoplasm of Purkinje cells (Fig. 4A). As reported previously (28) the PI 3-kinase and IGF-1 receptor immunoreactivities were detected in almost all Purkinje neurons in the cerebellar cortex, so we assume that p50␣ plays a specific role in these highly specialized cells.
Immunoblotting of Three Types of Regulatory Subunits (p85␣, p55␣, and p50␣) and Their Associated PI 3-Kinase Ac-tivities-To determine the expression of p85␣ and its two splice variants in different rat tissues at the protein level, lysates from various rat tissues were immunoprecipitated with protein A-agarose beads covalently coupled to ␣p85 PAN-UBI . This antiserum, which is raised against the entire region and the nSH2 region of p85␣, recognizes not only p85␣ but also p55␣ and p50␣. The immunoblot obtained with ␣p85 PAN-UBI revealed the 85-kDa band and a broad 50 -55-kDa band (Fig. 5A). The 85-kDa protein was abundantly expressed in every tissue examined, and the 50 -55-kDa band was prominent in the brain, liver, and kidney but faint in fat and muscle. By taking into consideration that the antibody used recognizes the entire p85␣ molecule, the p55␣ and p50␣ molecules, which have neither the SH3 nor the bcr homology domain, would be less effectively detected than p85␣ in the immunoblot using ␣p85 PAN-UBI . Thus, the relative amounts of p55␣ and p50␣ proteins, as compared with the amount of p85␣, are assumed to be larger than those suggested by the data in Fig. 5A.
As shown by the immunoblot using the specific antibody against p85␣ (␣p85 ␣SH3 ) (Fig. 5B), the expression of p85␣ protein is ubiquitous. The second isoform, p55␣, is expressed abundantly in the brain but only faintly in muscle. On the other hand, the third isoform, p50␣, is expressed most abundantly in liver and in relative abundance in the brain and kidney as shown in the immunoblot using the specific antibody against p50␣ (␣p50␣) (Fig. 5B). These results regarding protein expression levels appear to be consistent with Northern blotting results. Fig. 5C shows the PI 3-kinase activities associated with p85␣, p55␣, and p50␣ proteins. p85␣-associated PI 3-kinase is ubiquitously detected, whereas p55␣-associated PI 3-kinase is detected mainly in brain and muscle, and p50␣-associated PI 3-kinase is detected in liver and brain, as well as in muscle FIG. 1. A, an alignment of the amino acid sequences of p85␣, p50␣, p55␣, p55␥, and p85␤. The amino acid residues for each protein, with the addition of gaps (Ϫ) to optimize the alignment, are numbered to the right of each sequence. Two SH2 domains and the bcr and SH3 homology domains are boxed. The nucleotide sequence of p50␣ has been submitted to the GenBank TM /EMBL Data Bank with accession number D78486. B, schematic comparison of the three groups of PI 3-kinase regulatory subunits. p85␣ and its two variants, the ␤ and ␥ isoforms, were structurally divided into three groups. C, immunoblotting of p85␣, p85␤, p55␣, p55␥, and p50␣ expressed in Sf9 cells. The Sf9 cells infected with baculoviruses containing one of each of the five isoforms were cultured for 48 h and lysed in Laemmli buffer. The cell lysates were subjected to SDS-PAGE, and immunoblotting was performed with anti-HA antibody (panel a) and the protein specific antibody indicated above each panel (panels b-f). Lane 1, control Sf9 cells; lanes 2-6, Sf9 cells expressing p85␣, p85␤, p55␣, p55␥, and p50␣, respectively. tissues.
p50␣ Binds Two Types of Catalytic Subunits of PI 3-Kinase-Kurosu et al. (29) reported the presence of a 46-/100-kDa het-erodimer form of a PI 3-kinase, in rat liver, which was isolated in the flow-through fraction of a DEAE-Sepharose column. In addition, this 46-kDa protein was reported to readily be recognized by the antibody against the whole p85␣ molecule. To determine whether this 46-kDa protein is identical to p50␣, we performed the same chromatographic procedure using DEAE-Sepharose. The soluble fractions prepared from rat liver were applied to a DEAE-Sepharose column, and fractionation was performed according to the reported procedures (29). The obtained fractions were immunoprecipitated with the specific antibodies against p85␣ or p50␣, and the PI 3-kinase activities in these immunoprecipitates were measured.
As shown in Fig. 6A, the p50␣-associated PI 3-kinase was detected in two fractions. One was the flow-through fraction, eluted at 0 mM KCl (fraction A), and the other fraction eluted at approximately 0.1 M KCl (fraction B). On the other hand, the PI 3-kinase activities associated with p85␣ proteins were observed not in the flow-through fractions but in the fractions eluted at approximately 0.2 M KCl (fraction C), which is in agreement with the results of Kurosu et al. (29). The apparent molecular masses of the p50␣-associated PI 3-kinases in both fractions (fractions A and B) were determined to be 160 -180 kDa based on the gel filtration results (Fig. 6B). These fractions (fractions A, B, and C) were then collected and immunoprecipitated with ␣p85 PAN-UBI . Western blotting with ␣p85 PAN-UBI allowed isolation of the p85␣ and p50␣ proteins with DEAE chromatography (Fig. 6C). With respect to the difference between the two p50␣associated PI 3-kinase fractions, A and B, we speculate that p50␣ may bind to the different catalytic subunits. To ascertain the different characteristics of the catalytic subunits associated with p50␣ in the two fractions, we attempted wortmannin labeling, followed by treatment with anti-wortmannin antibody for detection of the catalytic subunit. The catalytic subunit associated with p50␣ in fraction B was detected by this procedure as a band of 110 kDa (Fig. 6D, lane 4). In contrast, the catalytic subunit associated with the p50␣ in fraction A was not detectable with the same procedure (Fig. 6D, lane 2), despite fraction A containing an amount of p50␣ protein similar to that of fraction B. In contrast to the case of p50␣, the PI 3-kinase activities associated with p85␣ and p55␣ were detected only in the eluted fractions containing approximately 0.2 M (fraction C) (Fig. 6A) and 0.15 M of KCl (fraction D) (Fig. 7A), respectively, but never in the flow-through fractions. Their associated catalytic subunits were easily detected by the wortmannin labeling and the following immunoblot using wortmannin antibody, as a band of 110 kDa (Fig. 6D, lane 6, and Fig. 7C, lane 2,  respectively).
The Function of p50␣ in the Insulin-induced Activation of Associated PI 3-Kinase-The role of insulin in the induction of the PI 3-kinase activities associated with the various regulatory subunits was evaluated using PC12 and HepG2 cells. PC12 cells express all five regulatory subunit isoforms, but HepG2 cells express only p85␣, p85␤, and p50␣ (data not shown). In PC12 cells, insulin caused increases in the PI 3-kinase activities of the ␣p85␣, ␣p55␣, and ␣p50␣ immunoprecipitates up to 1.9-, 1.9-, and 3.3-fold, respectively, whereas the ␣p85␤ and ␣p55␥ immunoprecipitates showed no significant increases in PI 3-kinase activity (Fig. 8A). In HepG2 cells, p85␣ and p50␣ responded to insulin stimulation with increases of 2.7-and 4.5-fold, respectively, whereas no significant change was observed for p85␤ (Fig. 8B). In both cell lines, the degree of PI 3-kinase activation was revealed to be highest for the p50␣associated PI 3-kinase. However, the possibility that the different responses are due to specific antibodies, bound to the different portions of these regulatory subunits, cannot be excluded.
Overexpression of Regulatory Subunits in HepG2 Cells and in CHO Cells Expressing Insulin Receptors or IRS-1-The cDNA FIG. 5. Immunoblotting of the protein immunoprecipitated with ␣p85 PAN-UBI and PI-3 kinase activities in various rat tissues. Rat tissues were homogenized and solubilized in lysis buffer. Supernatants were collected after the centrifugation and incubated with beads coupled to the antibody against the entire p85␣ molecule. The beads were washed three times and resuspended in Laemmli buffer. The eluate from the beads was electrophoresed and immunoblotted with ␣p85 PAN-UBI (A) or with ␣p85 ␣SH3 , ␣p55␣, or ␣p50␣ (B). The 50 -55-kDa band, present in all lanes in B, corresponds to the heavy chain of IgG. Various rat tissues were solubilized, and the supernatants obtained by centrifugation were incubated with control antibody, ␣p85 ␣SH3 , ␣p55␣, or ␣p50␣. The PI-3 kinase activities in these immunoprecipitates were measured as described under "Experimental Procedures" (C).

FIG. 6. Fractionation of rat liver cytosolic PI 3-kinase activities on a DEAE-Sepharose and a gel filtration column and detection of p110 by wortmannin labeling.
A, insoluble materials from the rat liver (30 ml, 20 mg/ml) were directly loaded onto a DEAE-Sepharose column. The column was then washed with buffer B, consisting of 20 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM dithiothreitol, 1 M leupeptin, 0.5 mM PMSF, and 10% glycerol. Once the 280 nm absorbance of the eluate had returned to base line, protein was eluted (absorbance was monitored at 280 nm (⅐ ⅐ ⅐)) with a gradient of KCl. Aliquots of the individual fractions were immunoprecipitated with ␣p50␣ or ␣p85 ␣SH3 and then assayed for PI 3-kinase activity. Ⅺ, ␣p50␣; ࡗ, ␣p85 ␣SH3 . B, the pooled fractions from the peaks (fractions A and B) were concentrated and applied to a gel filtration column. Eluted fractions (1 ml) were collected and immunoprecipitated with ␣p50␣ and assayed for PI 3-kinase activity. ࡗ, fraction A; Ⅺ, fraction B. C, each fraction was also immunoprecipitated with ␣p85 PAN-UBI and then subjected to SDS-PAGE. Immunoblotting analysis was performed with ␣p85 PAN-UBI. Lanes 1-3, fractions A-C, respectively. D, all fractions were immunoprecipitated with control antibody (lanes 1, 3, and 5) and construct for each isoform having the HA tag at its COOH terminus was prepared, and the adenoviruses for the transient expression of these isoforms were produced. HepG2 cells and CHO cells, expressing insulin receptors (CHO/IR) or IRS-1(CHO/IRS-1), were infected with these adenoviruses to achieve similar protein expression levels, as assessed by the immunoblot using anti-HA antibody (data not shown). In HepG2 cells, the insulin stimulation induced a marked increase (25.2-fold) in PI 3-kinase activity associated with p50␣ proteins, whereas relatively small increases (2-5-fold) were observed with p85␣, p55␣, and p55␥, and the increase in p85␤associated PI 3-kinase activity was very small (Fig. 9A).
To elucidate the mechanisms accounting for the variability in the extent of PI 3-kinase activation associated with the various regulatory proteins, we investigated the amount of IRS-1 bound to the expressed regulatory subunits in response to insulin. Before and after insulin stimulation, the cells were lysed and immunoprecipitated with the anti-HA antibody. The expressed regulatory subunit proteins of the indicated molec-ular mass were observed (Fig. 9B), and [ 35 S]methionine-labeled IRS-1 proteins (approximately 180 kDa) associated with each of the five regulatory subunits were measured, and the ratios of bound IRS-1/the amount of regulatory subunit expressed were calculated for each of the isoforms (Fig. 9C). p50␣ proteins apparently associated with larger amounts of phosphorylated IRS-1 protein, as compared with other isoforms. In contrast, p55␥ associated with the smallest amount of IRS-1, in response to insulin stimulation, among the five isoforms. The IRS-1 protein was also measured by immunoblotting using the antibody against IRS-1, and no significant difference was observed between the metabolic labeling method and immunoblotting data (data not shown). These data indicate that p50␣ shows the most efficient IRS-1 binding in response to insulin. This observation may explain the high capacity of this protein to induce PI 3-kinase activity, as compared with other regulatory proteins, in response to insulin stimulation.

DISCUSSION
In a study utilizing the expression cloning method with 32 Plabeled IRS-1, we previously isolated the four regulatory subunit isoforms for PI 3-kinase from a cDNA library prepared from rat brain (22). In this study, we screened a cDNA expression library prepared from rat liver by the same method and isolated cDNA encoding a novel protein, p50␣, which appears to be an alternative splicing form of the p85␣ gene product. Thus, there are five known regulatory subunit isoforms in mammals, including two 85-kDa proteins, two 55-kDa proteins, and one 50-kDa protein, as demonstrated by the results of overexpression experiments using Sf9 cells (Fig. 1C). In fact, the immunoblot using the antibody that recognizes the entire p85␣ molecule revealed marked expression of 50 -55-kDa proteins in various tissues (Fig. 5A). It appears that these 50 -55-kDa proteins have, to date, been regarded as degradation products of p85␣ or p85␤ and have thus attracted little attention. In   FIG. 7. Fractionation of rat brain cytosolic PI 3-kinase activities on a DEAE-Sepharose column and detection of p110 by wortmannin labeling. A, insoluble materials from the rat liver (30 ml, 8 mg/ml) were directly loaded onto a DEAE-Sepharose column. The column was then washed with buffer B. Protein was eluted (absorbance was monitored at 280 nm (⅐ ⅐ ⅐)) with a gradient of KCl. Aliquots of the individual fractions were immunoprecipitated with ␣p50␣, ␣p55␣, or ␣p85 ␣SH3 and then assayed for PI 3-kinase activity. Ⅺ, ␣p50␣; f, ␣p55␣; ࡗ, ␣p85 ␣SH3 . B, fraction D was also immunoprecipitated with ␣p85 PAN-UBI and then subjected to SDS-PAGE. Immunoblotting analysis was performed with ␣p85 PAN-UBI . C, fraction D was immunoprecipitated with control antibody (lane 1) and ␣p55␣ (lane 2) and labeled with 0.1 M wortmannin. Wortmannin-labeled immunoprecipitates were subjected to SDS-PAGE. The wortmannin labeling was detected by enhanced chemiluminescence.
One of the two 55-kDa proteins and the 50-kDa protein were revealed to share the two SH2 domains and an inter-SH2 domain with p85␣ and are thus considered to be alternative splicing products from the p85␣ gene. Thus, the p85␣ gene produces three different isoforms with molecular sizes of 85 (p85␣), 55 (p55␣), and 50 kDa (p50␣). These isoforms share the same two SH2 domains and an inter-SH2 domain but contain different NH 2 -terminal sequences. The most well-known isoform, p85␣, contains SH3 and bcr homology domains in its NH 2 terminus. The second ␣ type isoform, p55␣, contains a unique 34-amino acid sequence, which shows considerable similarity to the corresponding region of p55␥. The third ␣ type isoform, p50␣, contains only the unique 6-amino acid sequence in its NH 2 -terminal portion. These sequence data suggest that p50␣ may be the most primitive regulatory subunit form and that p55␣ and p85␣ may have more functions than p50␣, mediated via the 34-amino acid portion, as well as the respective SH3 and bcr homology domains. In fact, the SH3 domain of p85␣ was shown to interact with dynamin, a GTP-binding microtubule-associated protein (30), and also with microtubules (31) or ␣/␤-tubulin (32). The role of the 34-amino acid sequence in the NH 2 termini of the two 55-kDa regulatory subunits remains unknown, although the highly conserved sequence shared by p55␣ and p55␥ may suggests a specific function of the 34-amino acid portion. p50␣ contains a unique sequence of only 6 amino acids, apparently too short to associate with other molecules. This would presumably limit the functional capacity of this protein.
Northern blotting using the cDNA probe coding for the N-SH2 domain of p85␣ revealed four bands (7.7, 6.0, 4.2, and 2.8 kb), as previously reported (22) (Fig. 2B). Among these bands, those of 6.0-and 2.8-kb bands matched those of p50␣, whereas the 7.7-and 4.2-kb bands matched those of the SH3 p85␣ domain. As we reported previously, a minor portion of the 4.2-kb band and a considerable portion of the 2.8-kb band in brain correspond to p55␣ mRNA. It should be noted that these three regulatory isoforms are not specific to the rat and are also present in the mouse, hamster, and human, based on the results of immunoblotting using specific antibodies (data not shown).
As shown in our previous report (22) and in this study, all regulatory subunit isoforms for PI 3-kinase are abundantly expressed in brain tissue. Although there is one report describing the important role of PI 3-kinase in neuronal differentiation (14), no study has demonstrated a PI 3-kinase function in neuronal cells after differentiation. The yeast homolog of PI 3-kinase, VPS34, is required for trafficking of proteins from the Golgi apparatus to vacuoles (33,34). In addition, the activation of PI 3-kinase has been implicated in histamine release (35). Thus, PI 3-kinase may play a major role in the secretion of various neurotransmitters. The data on the distribution of each regulatory isoform in rat brain differed among the isoforms, possibly offering clues as to the mechanisms regulating neurotransmitter secretion.
There has been only one report describing a regulatory subunit with a molecular size of approximately 50 kDa, which was recognized by the antibody against the entire p85␣ molecule (29). In that study, the 46-kDa regulatory subunit formed a heterodimer with a 100-kDa catalytic subunit in rat liver, and this heterodimer could be separated in the flow-through fraction of a DEAE-Sepharose column. Moreover, the 46-/100-kDa PI 3-kinase was activated by G␤␥. We suspected this 46-kDa regulatory subunit to be identical to p50␣ and performed the same chromatographic procedure on a rat liver lysate using a DEAE-Sepharose column. We found p50␣ to be the only major protein detected in the flow-through fraction that associates with PI 3-kinase and is recognized by the antibody against the entire p85␣ molecule. To determine whether the molecular size of the catalytic subunit in the flow-through fraction is 100 kDa, wortmannin labeling followed by immunoblotting using the antibody against wortmannin was performed. The gel chromatographic results suggest that the actual size of the p50␣containing PI 3-kinase in the flow-through fraction, based on our calculation, is approximately 160 kDa. Thus, we speculate that p50␣ in the flow-through fraction associates with the catalytic subunit that has a molecular size of approximately 110 kDa. However, immunoblotting using the antibody against wortmannin failed to detect the catalytic subunit associated with p50␣ in the flow-through fraction. The other fraction containing p50␣-associated PI 3-kinase was eluted at a KCl concentration of 0.1 M. The catalytic subunit in this fraction was detected by wortmannin labeling, followed by immunoblotting with the antibody against wortmannin, and the molecular FIG. 9. A, insulin responsiveness of overexpressed regulatory subunits for PI 3-kinase in HepG2 cells. HepG2 cells were infected with viruses expressing the full-length amino acid sequences of p85␣ (Ⅺ), p85␤ (f), p55␣ (E), p55␥ (Ç), or p50␣ (q), as well as the HA tag amino acid sequence at each COOH terminus, for 1 h, and then grown for 48 h. The indicated concentrations of insulin were added to the medium, and the cells were incubated for 5 min at 37°C. After insulin treatment, the cells were collected with lysis buffer C. The resulting supernatants were immunoprecipitated with anti-HA antibody and assayed for PI 3-kinase activity. B, [ 35 S]methionine labeling of HepG2 cells expressing p85␣, p85␤, p55␣, p55␥, or p50␣. HepG2 cells were infected with adenoviruses expressing the full-length cDNAs of p85␣, p85␤, p55␣, p55␥, and p50␣ and then grown for 48 h. After metabolic labeling with [ 35 S]methionine, 10 Ϫ6 M insulin was added to the medium and the cells were further incubated for 5 min at 37°C. Then, all lysates were centrifuged, and the resulting supernatants were immunoprecipitated with control antibody or the anti-HA antibody and pelleted with protein G-Sepharose beads. The immunoprecipitates were subjected to SDS-PAGE analysis. The gels were dried and subjected to autoradiography. These experiments were conducted three times each and yielded similar results. C, ratios of %IRS-1 proteins/expressed regulatory proteins were calculated. size was determined to be approximately 110 kDa. Thus, there is an apparent difference between the two catalytic subunits associated with p50␣ in the flow-through fraction and that associated with p50␣ in the 0.1 M KCl fraction. In contrast to p50␣, p55␣ or p85␣ having PI 3-kinase was detected only in the 0.15 M KCl or 0.2 M KCl fraction from the DEAE-Sepharose column, respectively, and the molecular sizes of their associated catalytic subunits were determined to be 110 kDa by wortmannin labeling and immunoblotting with the antibody against wortmannin. The binding motif of the p85␣ regulatory subunit with the p110 catalytic subunit was reported to reside in the inter-SH2 domain (18), and this portion of the protein was completely conserved among p85␣, p55␣, and p50␣. Thus, it appears quite unlikely that p50␣ associates with a different catalytic subunit to which p85␣ and p55␣ cannot bind. We speculate that the difference between the two catalytic subunits bound to p50␣ might be due to a modification of the catalytic subunit, such as serine phosphorylation, although further study is needed to clarify this issue. Taking these observations together, we cannot rule out the possibility that p50␣ is identical to the reported 46-kDa protein.
It is now clear that at least four different types of growth factor-regulated PI 3-kinase exist, including mammalian homologs of Saccharomyces cerevisiae VPS34 (33), a G-proteinactivated form termed p110␥ (36), and the recently cloned p170 (37). In the case of insulin signaling, the activation of PI 3-kinase is thought to be particularly important. Insulin stimulation induces glucose transporter translocation to the plasma membrane in muscle and fat cells, resulting in an increase in glucose uptake (15,16). In addition, insulin leads to an increase in hepatic glycogen synthesis, recently reported to occur via the activation of c-Akt (38). These important functions of insulin have been shown to be blocked by specific inhibitors of PI 3-kinase, such as wortmannin (39) and LY294002 (40), as well as by the overexpression or microinjection of the dominant negative mutant p85␣ (41,42). Furthermore, overexpression of wild-type or constitutively active p110 induces glucose transporter translocation to the plasma membrane, irrespective of insulin stimulation, in 3T3-L1 cells (43,44). These findings strongly suggest the importance, in various insulin actions, of activation of the SH2 domain-containing PI 3-kinase.
We also investigated the levels of the PI 3-kinase activities associated with each of the five isoforms. First, the activation of endogenous PI 3-kinase by insulin was measured using the appropriate specific antibodies in PC12 cells and HepG2 cells. In addition, these five regulatory subunits were expressed in HepG2 cells, CHO/IR cells, and CHO/IRS-1 cells, and the extent of activation of their associated PI 3-kinases was compared. The results of a series of PI 3-kinase assays can be summarized as follows; p50␣ shows a much higher level of activation of its associated PI 3-kinase, in response to insulin stimulation, than the other regulatory subunits. The p85␣, p55␣, and p55␥ subunits exhibited only moderate responsiveness to insulin. The activation of PI 3-kinase associated with p85␤ was confirmed to be low, in agreement with the results of a previous report (45). In an effort to elucidate the molecular mechanism underlying the marked insulin-induced PI 3-kinase activation, we demonstrated that p50␣ exhibits the highest affinity for phosphorylated IRS-1 in response to insulin in vivo. Although the reason for this high affinity of p50␣ for IRS-1, despite p50␣, p55␣, and p85␣ sharing the same two SH2 domains, is unknown, we speculate that the NH 2 -terminal domains of p85␣ and p55␣ form complexes with other molecules resulting in an inability to bind IRS-1 as efficiently as p50␣. Considering both the high level of p50␣ expression in the liver and its marked responsiveness to insulin, p50␣ may play a more critical role than the other isoforms in hepatic insulininduced activation of PI 3-kinase. p55␥-associated PI 3-kinase was shown to be activated by insulin with a low affinity for IRS-1, whereas p85␤ did not respond significantly to insulin despite its association with IRS-1. Further study is needed to clarify these issues.
In summary, there are five regulatory subunit isoforms of PI 3-kinase which can be classified into three groups, an 85-kDa protein, a 55-kDa protein, and a 50-kDa protein. Each isoform has a different tissue distribution and was shown to exhibit a different level of activation, of the associated PI 3-kinase, in response to insulin stimulation. Given the idea that PI 3-kinase is involved in a series of systems, it is conceivable that PI 3-kinase plays a variety of roles in response to various stimuli. Further study is required to ascertain which isoform corresponds to which biological phenomenon.

FIG. 10. Insulin responsiveness of overexpressed regulatory subunit for PI 3-kinase in CHO/IR (A) and CHO/ IRS-1 cells (B)
. CHO/IR and CHO/IRS-1 cells were infected with various viruses for 1 h and grown for 48 h. Then 10 Ϫ6 M insulin was added to the medium, and the cells were incubated for 5 min at 37°C. After insulin treatment, the cell lysates were centrifuged at 10,000 ϫ g for 10 min. The resulting supernatants were immunoprecipitated with anti-HA antibody and assayed for PI 3-kinase activity.