Differential Regulation of Phosphoinositide 3-Kinase Adapter Subunit Variants by Insulin in Human Skeletal Muscle*

The role of phosphoinositide 3-kinase (PI 3-kinase) in insulin signaling was evaluated in human skeletal muscle. Insulin stimulated both antiphosphotyrosine-precipitable PI 3-kinase activity and 3-O-methylglucose transport in isolated skeletal muscle (both ≈2–3-fold). Insulin stimulation of 3-O-methylglucose transport was inhibited by the PI 3-kinase inhibitor LY294002 (IC50 = 2.5 μm). The PI 3-kinase adapter subunits were purified from muscle lysates using phosphopeptide beads based on the Tyr-751 region of the platelet-derived growth factor receptor. Immunoblotting of the material adsorbed onto the phosphopeptide beads revealed the presence of p85α, p85β, p55PIK/p55γ, and p50 adapter subunit isoforms. In addition, p85α-NSH2 antibodies recognized four adapter subunit variants of 54, 53, 48, and 46 kDa, the latter corresponding to the p50 splice variant. Serial immunoprecipitations demonstrated that these four proteins were associated with a large proportion of the total PI 3-kinase activity immunoprecipitated by p85α-NSH2 domain antibodies. Antibodies to p85β, p55PIK/p55γ, and the p50 adapter subunit also immunoprecipitated PI 3-kinase activity from human muscle lysates. A large proportion of the total cellular pool of the 53-kDa variant, p50, and p55PIK was present in antiphosphotyrosine immunoprecipitates from unstimulated muscle, whereas these immunoprecipitates contained only a very small proportion of the cellular pool of p85α, p85β, and the 48-kDa variant. Insulin greatly increased the levels of the 48-kDa variant in antiphosphotyrosine immunoprecipitates and caused smaller -fold increases in the levels of p85α, p85β, and the 53-kDa variant. The levels of p50 and p55PIK were not significantly changed. These properties indicate mechanisms by which specificity is achieved in the PI 3-kinase signaling system.

Insulin binding to its cell-surface receptor activates a range of intracellular signaling cascades that ultimately result in the regulation of a number of important metabolic events within the cell (1,2). A large body of evidence indicates that PI 3-ki-nase 1 activity is necessary for a wide range of cellular effects elicited by insulin (3)(4)(5)(6)(7)(8)(9)(10)(11). Specifically, several recent studies indicate that constitutive activation of PI 3-kinase is sufficient to induce stimulation of glucose uptake in adipocytes (12)(13)(14). However, many growth factors that stimulate PI 3-kinase do not stimulate glucose metabolism, suggesting that different signaling inputs have specific mechanisms for utilizing the PI 3-kinase signaling system to generate downstream responses (reviewed in Ref. 15).
Insulin and receptor tyrosine kinase-linked growth factors activate class 1 PI 3-kinases, which consist of a 110-kDa catalytic subunit linked to an adapter subunit (15). Three isoforms of the catalytic subunit have been identified: p110␣, p110␤, and p110␥ (16 -18). Of these, only p110␣ and p110␤ are present in insulin-responsive tissues. Two widely expressed Ϸ85-kDa forms of the adapter subunit were originally characterized: p85␣ and p85␤. These share a high degree of sequence and structural homology, with each containing two SH2 domains, two proline-rich domains (P1 and P2), an SH3 domain, and a Bcr homology domain (19). Recently, several truncated isoforms of the PI 3-kinase adapter subunit have been identified including a 55-kDa variant of p85, termed p55␥ or p55 PIK (20,21), which is encoded by a gene separate from p85␣ or p85␤. Two splice variants of the p85␣ gene have also been reported. These are a 53-55-kDa form termed AS53 (22) or p55␣ (20) and a form encoding a protein with a predicted molecular mass of 50 kDa (23,24). PI 3-kinase purified from liver has also been reported to associate with a 46-kDa protein that cross-reacts with p85 antibodies (25). All of these truncated adapter subunit variants contain two SH2 domains and the P2 proline-rich domains, but lack the SH3 domain, the Bcr homology domain, and the P1 proline-rich domain of full-length p85.
The main mechanism by which insulin regulates PI 3-kinase activity is by tyrosine phosphorylation of specific YMXM motifs on the intracellular signaling intermediates IRS-1 and IRS-2, which enables the recruitment of PI 3-kinase via the SH2 domain on the PI 3-kinase adapter subunit (26). This increases cellular PI 3-kinase activity by a combination of increasing the catalytic activity of p110 (27,28) and locating the p110 catalytic subunit in proximity to its substrate at the membrane.
It is not clear whether the different isoforms of the adapter subunit play independent or overlapping roles in intracellular signaling pathways. However, the possibility exists that specificity could be introduced into the PI 3-kinase system by differential recruitment of the individual adapter subunits in response to activation of different receptor tyrosine kinases. Such interactions would provide a mechanism by which multi-ple signaling inputs could use PI 3-kinase in different ways to obtain specific signaling outcomes.
In this study, we have investigated the PI 3-kinase signaling system in human skeletal muscle using a unique muscle strip preparation (29). The data obtained provide strong evidence that PI 3-kinase plays a crucial role in insulin signaling to glucose transport in human skeletal muscle. Furthermore, we have focused on the role of individual variants of the PI 3-kinase adapter subunit in insulin signaling. We have identified seven variants of the PI 3-kinase adapter subunit present in human skeletal muscle, and we demonstrate that these are differentially recruited into phosphotyrosine complexes in response to insulin stimulation. This study provides evidence for mechanisms by which specificity can be obtained in the PI 3-kinase signaling system.

EXPERIMENTAL PROCEDURES
Materials-125 I-Protein A, Na 125 I, [ 32 P]ATP, ECL reagents, and Rainbow protein molecular mass standards were from Amersham International (Buckinghamshire, United Kingdom). Phosphatidylinositol was from Lipid Products (Redhill, United Kingdom). All other chemicals were from Sigma. Monoclonal PY20 antiphosphotyrosine antibody was from Transduction Laboratories. Rabbit polyclonal antisera were raised against glutathione S-transferase fusion proteins corresponding to the SH3 domain (p85␣-SH3) or the N-SH2 (p85␣-NSH2) domain of human p85␣ as described previously (30). Monoclonal antibodies recognizing epitopes in the SH3 domain (U13) and in the N-SH2 domain (U2) of p85␣ and another specifically recognizing p85␤ were kindly supplied by Dr. I. Gout and Prof. M. Waterfield (Ludwig Institute, London) (31). A previously described polyclonal antiserum specific to p55 PIK was provided by Drs. S. Pons and M. White (Joslin Diabetes Center, Boston) (21). Antibodies to the p55␥ adapter subunit and the p50 splice variant of p85␣ were supplied by Dr. T. Asano (24). A peptide corresponding to amino acids 738 -755 of the platelet-derived growth factor receptor (GGYMDMSKDESVDYVPML) was phosphorylated at the tyrosine equivalent to residue 751 and cross-linked to Actigel (provided by Dr. B. Van Haesebroeck, Ludwig Institute, London).
Subjects-Muscle specimens were obtained from the vastus lateralis portion of the quadriceps femoris muscle from 19 healthy male subjects (age: 28.4 Ϯ 0.7 years; weight: 76.6 Ϯ 2.1 kg; height: 181.3 Ϯ 1.9 cm; body mass index: 23.4 Ϯ 0.4 kg/m 2 ; fasting glucose: 4.9 Ϯ 0.2 mM; and serum insulin: 6.9 Ϯ 0.4 microunits/ml). None of the subjects were taking any medication. Following an overnight fast, the subjects reported to the laboratory at 8:30 a.m. A local anesthetic (prilocain hydrochloride, 10 mg/ml) was administered subcutaneously 15 cm above the proximal border of the patella, and a 4-cm incision was made. Two muscle specimens (250 mg each) were excised for in vitro incubation, and smaller muscle strips were prepared as described (29,32).
Stimulation of Muscle and Determination of 3-O-Methylglucose Transport-After preparation, the smaller muscle strips (ϳ20 mg) were incubated at 35°C for 10 min in a recovery solution containing oxygenated Krebs-Henseleit buffer supplemented with HEPES, 38 mM mannitol, 2 mM pyruvate, and 0.1% bovine serum albumin. The muscle specimens were subsequently incubated with inhibitors and insulin as indicated in the figure legends. For measurements of 3-O-methylglucose transport, the buffer contained 5 mM 3-O-[ 3 H]methylglucose and 35 mM [ 14 C]mannitol. For glucose transport experiments, the incubations were terminated by snap-freezing the muscle in liquid nitrogen. Thereafter, the muscle samples were rapidly homogenized in ice-cold buffer containing 100 mM NaF and 10 mM EDTA. Cell debris was removed by centrifugation at 5000 ϫ g for 1 min. 3-O-Methylglucose transport was calculated as described previously (33). For immunoprecipitation experiments, the muscle samples were homogenized on ice in 50 mM Tris-Cl (pH 7.4), 1% Triton X-100, 10 mM EDTA, 1 mM NaF, 1 mM NaVO 3 , 2 mM aprotinin, 2 mM leupeptin, and 2 mM phenylmethylsulfonyl fluoride. Homogenates were precleared by centrifugation at 5000 ϫ g for 10 min prior to immunoprecipitation.
Antibody Purification and Iodination-10 l of 100 mM Tris (pH 8.0) and 25 g of swollen protein A-agarose beads were added to 100 l of p85␣-NSH2 antiserum. The solution was mixed at 4°C for 4 h, and the beads were collected. Beads were washed with 100 mM Tris, and the antibody was eluted with 100 mM glycine (pH 3.0) for 5 min at room temperature. The glycine buffer was collected and neutralized with 10 l of 1 M Tris (pH 8.0). An aliquot of purified antibody (25 l) was iodinated by sequential addition of 5 l of NaH 2 PO 4 /Na 2 HPO 4 buffer (0.5 M, pH 7.4) and 0.5 Ci of Na 125 I followed by 10 l of chloramine T (0.5 mg/ml). The reaction was terminated 2 min later by the addition of 300 l of 25 mM Tris (pH 7.4) containing 10 mM MgCl 2 and 1% bovine serum albumin. Nonincorporated 125 I was separated from iodinated antibody on a Bio-Rad Econo-Pac 10DG disposable desalting column, and the peak iodinated antibody fraction was used for Western blotting.
Assessment of PI 3-Kinase Activity-PI 3-kinase activity was determined essentially as described previously (34). Incorporation of label from [␥-32 P]ATP into phosphatidylinositol was determined, and the lipid product was resolved by thin-layer chromatography and quantitated on a Fuji BAS2000 phosphoimager.
Statistical Analysis-Statistical differences between treatments were analyzed using Student's paired t test.

Insulin Acutely Stimulates PI 3-Kinase Activity in Human
Skeletal Muscle-The ability of insulin to stimulate PI 3-kinase activity in human skeletal muscle was initially characterized. Muscle was stimulated with insulin at 1000 microunits/ml, a concentration that maximally stimulates glucose transport (data not shown). Insulin stimulation increased PI 3-kinase activity associated with phosphotyrosine-containing proteins 3.3-fold compared with basal levels, with maximal stimulation being achieved 20 min after the addition of insulin (Fig. 1).

A Number of PI 3-Kinase Adapter Subunit Variants Are Expressed in Human
Muscle-Immunoblotting studies of whole muscle homogenate using a polyclonal antibody to the SH3 domain of p85␣ consistently revealed two immunoreactive bands at 85 and 87 kDa (Fig. 3A). Subsequent immunoblotting with isoform-specific monoclonal antibodies revealed that the faster migrating of these (lower band) represented p85␣ and the slower one (upper band) represented p85␤ (Fig. 3B). The Muscle strips were incubated with 1000 microunits/ml insulin for the indicated times and then snap-frozen in liquid nitrogen and homogenized as described under "Experimental Procedures." The homogenate was incubated with PY20 antiphosphotyrosine antibody and protein A-agarose for 1 h, at which time the beads were precipitated by centrifugation and assayed for PI 3-kinase activity as described under "Experimental Procedures." Results show a representative experiment performed in duplicate. Similar results were obtained in three separate experiments.
polyclonal antibody raised against the N-SH2 domain revealed a prominent immunoreactive protein at 85 kDa corresponding to p85␣, but did not recognize p85␤. However, this antibody also recognized immunoreactive proteins at 54, 53, 48, and 46 kDa in muscle homogenates. Whereas the relative intensity of the bands varied slightly in samples from different individuals, the 48-kDa protein was consistently the most intense (Fig. 3A). Identical immunoreactive proteins were recognized by antiserum raised against the same N-SH2 fusion protein in a separate rabbit, but these proteins were not recognized by preimmune serum from either of these rabbits (data not shown). More important, these same immunoreactive proteins were recognized by the monoclonal antibodies recognizing an epitope in the N-SH2 domain (Fig. 3B), but were not recognized by monoclonal or polyclonal antibodies recognizing epitopes in the SH3 domain of p85␣ (Fig. 3B). The lower molecular mass bands are unlikely to represent a proteolytic product of p85␣ as the intensity of the bands was not dependent on the inclusion of protease inhibitors in the homogenization mixture and did not change with time of storage. Furthermore, we have previ-ously demonstrated that the p85␣-NSH2 antiserum did not recognize any comparable lower molecular mass bands in homogenates of 3T3-L1 adipocytes or 3T3-L1 fibroblasts (Ref. 30; see also Fig. 6). The fact that the lower molecular mass bands correspond closely to those of recently reported splice variants of p85␣ (20 -25) suggests that they are in fact variants of the PI 3-kinase adapter subunit.

Identification of PI 3-Kinase Adapter Subunit Immunoreactive Bands in Fractions Purified Using Phosphopeptide
Beads-To confirm that the lower molecular mass proteins recognized by the p85␣-NSH2 antibody were variants of p85, an affinity purification step was introduced to remove any potential nonspecific reactions. In this approach, human muscle homogenates were precipitated with phosphopeptide beads based on the amino acids surrounding Tyr-751 (p85 SH2 domain-binding region) of the platelet-derived growth factor receptor. Such phosphopeptide beads have previously been shown to have a high affinity for PI 3-kinase adapter subunits (31,35). The specificity and selectivity of this approach are demonstrated by the fact that the p50 and p55 PIK antibodies used in this study recognize a number of bands in human muscle homogenates, whereas in the eluate from the phosphopeptide beads, this is reduced to a single band. The protein recognized by the p55 PIK antiserum in the phosphopeptide bead precipitates is 58 kDa, and the protein recognized by the p50 antiserum is 46 kDa (Fig. 4), which correspond closely with  the previously reported molecular masses of these variants in other species (21,24). The p85␤ antibody recognizes a single 87-kDa protein in muscle homogenates, and this protein is also present in phosphopeptide beads (Fig. 4C). Western blotting of the phosphopeptide bead eluates with the p85␣-NSH2 antibody revealed protein bands at 85, 54, 53, 48, and 46 kDa, and these were of equal intensity to the corresponding bands seen with this antibody in muscle homogenate (Fig. 4A). Several other components in the homogenate that were slightly immunoreactive with the p85␣-NSH2 antibody were not present in the precipitates and were thus confirmed as nonspecific reactions.
Association of Adapter Subunit Variants with PI 3-Kinase Activity-PI 3-kinase activity was present in PI 3-kinase adapter subunit antibody immunoprecipitates, confirming that the antibodies were recognizing functional PI 3-kinase adapter subunits. The greatest amount of PI 3-kinase activity was found associated with antibodies specific to p50, with lesser amounts associated with p85␣ and still lower levels associated with p85␤ and p55 PIK (Fig. 5A).
To investigate whether the lower molecular mass protein bands recognized by the p85␣-NSH2 antibody were associated with PI 3-kinase activity, serial immunoprecipitations followed by PI 3-kinase activity assays were performed. Total p85␣associated PI 3-kinase activity was immunoprecipitated from muscle homogenates using the p85␣-NSH2 antibody (Fig. 5B). However, immunoprecipitation of an equivalent amount of muscle homogenate with the p85␣-SH3 antibody, which recognizes both full-length p85␣ and p85␤ but not truncated variants, precipitated a much smaller amount of PI 3-kinase activity. This was not due to inefficient immunoprecipitation as both the p85␣-SH3 and p85␣-NSH2 antibodies were equally efficient, each being able to precipitate Ͼ85% of the full-length p85␣ protein (data not shown). Furthermore, subsequent immunoprecipitation with the same antibody in the supernatant from the initial p85␣-SH3 immunoprecipitation showed that no further PI 3-kinase activity could be immunoprecipitated (data not shown). The p85␣-NSH2 antibody, which recognizes truncated adapter subunit forms, was then used to immunoprecipitate PI 3-kinase activity remaining in the supernatant of the p85␣-SH3 immunoprecipitates. The amount of PI 3-kinase in these immunoprecipitates was Ϸ75% of the amount of PI 3-kinase activity directly immunoprecipitated by the p85␣-NSH2 antibody from an equivalent amount of homogenate (Fig. 5).
This indicates that truncated but functional forms of the PI 3-kinase adapter subunit are present in human skeletal muscle.
Overall, the combination of the immunoblotting data, the phosphopeptide bead experiments, and the PI 3-kinase assays in sequential immunoprecipitations provides strong evidence that the 54-, 53-, 48-, and 46-kDa proteins recognized by the p85␣-NSH2 antibody in fact represent forms of the PI 3-kinase adapter subunit containing SH2 domains, but not the SH3 domain or Bcr homology domain.
Truncated Variants of p85 Are Differentially Recruited in Response to Insulin-The regulation of the multiple, structurally distinct PI 3-kinase adapter subunit isoforms present in muscle had not been previously investigated. Therefore, the ability of insulin to induce the recruitment of the PI 3-kinase adapter subunit variants into antiphosphotyrosine immunoprecipitates was determined (Fig. 6). Direct blotting with iodinated p85␣-NSH2 antibody was performed on antiphosphotyrosine immunoprecipitates to avoid the problems associated with secondary antibodies recognizing IgG bands. Use of the p85␣-NSH2 antiserum also has the advantage of allowing direct comparison of the relative effects of insulin on several forms of the PI 3-kinase adapter subunit. The p85␣-NSH2 antiserum revealed four immunoreactive bands in the antiphosphotyrosine immunoprecipitates at 85, 53, 48, and 46 kDa. The four proteins recognized in the antiphosphotyrosine immunoprecipitates correspond to adapter subunit-specific bands seen in p85␣-NSH2 immunoblots of phosphopeptide bead eluates (Fig. 4). In these experiments, the 53-kDa protein was the most prominent immunoreactive band in antiphosphotyrosine immunoprecipitates from unstimulated muscle (Fig.  6A). The 85-and 46-kDa proteins were also readily detectable, but the 48-kDa protein was almost undetectable in antiphosphotyrosine immunoprecipitates from unstimulated muscle (Fig. 6A). The levels of the 46-kDa band, representing the p50 splice variant of p85, did not change significantly in antiphosphotyrosine immunoprecipitates following insulin stimulation (Fig. 6B). The most pronounced effect of insulin on recruitment of adapter subunit variants into antiphosphotyrosine immunoprecipitates was on levels of the 48-kDa variant (19.5-fold stimulation; p Ͻ 0.01). Insulin caused smaller -fold increases over basal levels of p85␣ (3.6-fold; p Ͻ 0.02) and the 53-kDa variant (1.7-fold; p Ͻ 0.02). This smaller -fold increase was FIG. 5. Association of PI 3-kinase activity with adapter subunit isoforms. Muscle strips were snap-frozen in liquid nitrogen and homogenized as described under "Experimental Procedures." Equal amounts of homogenate were incubated with the indicated antibody for 1 h, at which time protein A-agarose was added for a further 1 h. Beads were precipitated by centrifugation and assayed for PI 3-kinase activity as described under "Experimental Procedures." A, immunoprecipitations were performed as indicated using the p85␣ monoclonal antibody (U2), p85␤ monoclonal antibody (T4), p55 PIK polyclonal antiserum, or p50 polyclonal antiserum, and the amount of PI 3-kinase activity in the immunoprecipitates (IP's) was normalized to the amount of PI 3-kinase activity immunoprecipitated by the p85␣ monoclonal antibody (U2). Results are the mean Ϯ S.E. of at least three experiments. B, immunoprecipitations were performed with p85␣-NSH2 antisera in muscle homogenate (bar 1), p85␣-SH3 antisera in muscle homogenate (bar 2), and p85␣-NSH2 antisera in the supernatant of the p85␣-SH3 immunoprecipitates (bar 3). Results show a representative experiment performed in triplicate (mean Ϯ S.E.). Similar results were obtained in three separate experiments. largely due to the presence of larger amounts of p85␣ and particularly of the 53-kDa variant in antiphosphotyrosine immunoprecipitates from unstimulated muscle. However, when viewed in terms of absolute insulin-induced increases in the levels of adapter subunit variants in antiphosphotyrosine immunoprecipitates, it is seen that the increases in the amounts of p85␣ and the 53-and 48-kDa forms were all very similar (Fig. 6B). In these studies, the effects of insulin on the 54-kDa protein were not established as whereas it was recognized in direct blots of muscle homogenates with iodinated p85␣-NSH2 antibody, it was below the level of detection in direct blots of antiphosphotyrosine immunoprecipitates from basal and insulin-stimulated muscle (Fig. 6A). The reason for this is not clear.
As p85␤ and p55 PIK are not recognized by the p85␣-NSH2 antiserum, separate immunoblots were performed using antibodies specific to these isoforms to determine whether they associate with tyrosine-phosphorylated proteins in muscle in the unstimulated or insulin-stimulated state. The p55 PIK antiserum recognized a single protein band in antiphosphotyrosine immunoprecipitates that was identical in molecular mass to the band recognized by this antiserum in the phosphopeptide bead eluates (i.e. 58 kDa) (Fig. 6C). The p55 PIK band was present in the antiphosphotyrosine immunoprecipitates from unstimulated muscle, and the level was not significantly increased by insulin (Fig. 6D). The monoclonal antibody specific to p85␤ recognized an 87-kDa band in antiphosphotyrosine immunoprecipitates from both unstimulated and stimulated muscle (Fig. 6C). Insulin stimulation caused a 2.2-fold increase in the amount of p85␤ in antiphosphotyrosine immunoprecipitates (Fig. 6D).
Antiphosphotyrosine blotting of the phosphopeptide bead eluates was performed to determine whether tyrosine phosphorylation of the PI 3-kinase adapter subunits contributed to their association with antiphosphotyrosine immunoprecipitates. This failed to show any immunoreactive bands in phosphopeptide bead eluates despite the fact that the PY20 antibody clearly recognized a number of tyrosine-phosphorylated bands in the insulin-stimulated human muscle homogenate (Fig. 7).

DISCUSSION
Human skeletal muscle is a major insulin target tissue in the body as it is the major site of insulin-stimulated glucose disposal in vivo (36). However, the mechanisms by which insulin stimulates glucose uptake into muscle are poorly understood. The study provides evidence that PI 3-kinase activity is essential for insulin-stimulated glucose transport in human skeletal muscle. This is based on the finding that the PI 3-kinase inhibitor LY294002 inhibits insulin-stimulated 3-O-methylglucose transport in human skeletal muscle and in 3T3-L1 adipocytes with a similar potency. Furthermore, the potency of this inhibition correlates well with the potency with which this compound inhibits the activity of purified PI 3-kinase (37). Wortmannin, an alternative PI 3-kinase inhibitor, has previously been used to study the role of PI 3-kinase in insulinstimulated glucose transport in rodent muscle (11, 38 -40). However, the concentration of wortmannin required to completely inhibit insulin-stimulated glucose transport in these studies was an order of magnitude higher than required in adipocytes and cell culture models (3,(5)(6)(7)41). At these concentrations there is a possibility that wortmannin is inhibiting other signaling molecules such as PI 4-kinase (42) and phos-FIG. 6. Recruitment of p85 variants into phosphotyrosine-containing complexes by insulin. Muscle strips were incubated in the absence (Ϫ) or presence (ϩ) of 1000 microunits/ml insulin (INS) for 20 min and then snap-frozen in liquid nitrogen and homogenized as described under "Experimental Procedures." Equal amounts of the homogenates (250 g) were incubated with PY20 antiphosphotyrosine antibody and protein A-agarose for 1 h, at which time the beads were precipitated by centrifugation. Material was eluted from the beads by boiling in SDS-PAGE loading buffer containing 100 mM dithiothreitol and 2% SDS. Samples were separated by 8% SDS-PAGE and Western-blotted as indicated. A, shown is a representative experiment of direct Western blotting of antiphosphotyrosine immunoprecipitates (IP's) with iodinated p85␣-NSH2 polyclonal antiserum, including a lane containing 100 g of muscle homogenate and another containing 150 g of 3T3-L1 adipocyte lysate for comparison. B, bands corresponding to p85␣ splice variants from A were quantitated on a BAS2000 phosphoimager. Basal levels are shown by shaded bars, and insulin-stimulated levels are shown by closed bars. Bars represent the mean Ϯ range of duplicate muscle strips. Similar results were obtained in a separate experiment also performed in duplicate. C, shown is a representative Western blot of antiphosphotyrosine immunoprecipitates (Anti-ptyr IP) with the p85␤ monoclonal antibody and p55 PIK polyclonal antiserum, both visualized using ECL. D, bands corresponding to p85␤ and p55 PIK were quantitated using a Bio-Rad GS670 densitometer and are expressed as band density relative to the band density in the unstimulated state. Results in each case are the mean Ϯ S.E. of four separate muscle strips. pholipase A 2 (43). Therefore, the LY294002 data in this study strengthen the case for PI 3-kinase activity being essential for insulin stimulation of glucose transport in human skeletal muscle.
This study provides the first detailed assessment of the expression of PI 3-kinase adapter subunits and their regulation by insulin in human skeletal muscle. This is important as the effects of growth factors on PI 3-kinase are mediated through the PI 3-kinase adapter subunits, which regulate the location and specific activity of the PI 3-kinase catalytic subunit (15). Adipocytes express only the p85␣ PI 3-kinase adapter subunit (30,44), whereas this study demonstrates that human skeletal muscle expresses both p85␣ and p85␤. Additionally, we demonstrate that human skeletal muscle contains five truncated forms of the PI 3-kinase adapter subunit. These are p55 PIK and four variants recognized by the p85␣-NSH2 antibodies. The possibility that one of the proteins recognized by the p85␣-NSH2 antibodies is p55 PIK /p55␥ can be excluded as the p55 PIK antiserum only recognized a single band of 58 kDa in both phosphopeptide beads (Fig. 4) and antiphosphotyrosine immunoprecipitates (Fig. 6). This also indicates that the p85␣-NSH2 antiserum is highly selective for the p85␣ SH2 domain as it does not recognize p85␤ or p55 PIK /p55␥ despite the fact that the N-SH2 domains of p55 PIK /p55␥, p85␤, and p85␣ are Ͼ90% identical at the amino acid level. Therefore, the most likely explanation for the 54-, 53-, 48-, and 46-kDa bands is that they all contain an N-SH2 domain identical to that of p85␣ and are thus likely to be p85␣ splice variants. Both p50 (23,24) and p55␣/AS53 (20,22) are splice variants of p85␣ containing identical N-SH2 domains and would therefore be expected to crossreact with the p85␣-NSH2 antisera with a similar affinity for p85␣ itself. Indeed, the 46-kDa band recognized by the p85␣-NSH2 antibody in phosphopeptide bead eluates cross-reacts with an antibody specific to the p50 splice variant of p85␣. The 53-and 48-kDa bands are likely to represent the two described forms of the AS53 splice variant of p85␣ as antibodies raised against the unique regions of AS53 recognize bands in Western blots of rat tissue (22) that have very similar characteristics to the 53-and 48-kDa bands seen in this study. The identity of the 54-kDa band cannot be explained by any known PI 3-kinase adapter subunit isoforms, although the molecular mass slightly greater than 53 kDa suggests that it could represent the variant of AS53 that has an 8-amino acid splice insert in the inter-SH2 domain (22). However, suitable reagents to assess this were not available.
Very few studies have been conducted to directly compare the function of different PI 3-kinase adapter subunits. While p85␣ and p85␤ are coded by separate genes, they are structurally very similar, containing two SH2 domains, two proline-rich domains, a single SH3 domain, and a Bcr/racGAP homology domain (19). Despite these similarities, previous studies have suggested that differences may exist in the mechanism by which these two isoforms control the catalytic activity of the complex. For example, in some cell types, PI 3-kinase complexes containing p85␤ appear to be less responsive to insulin than those containing p85␣ (44). Furthermore, p85␣ is phosphorylated by p110 at serine residues (45), whereas this does not occur in p85␤ (31). Little was previously known about the mechanisms by which the different truncated adapter subunit isoforms are regulated. Insulin stimulation of rat muscle has been reported to cause a small increase in IRS-1 association of the AS53 splice variant of p85␣ (22). Furthermore, coexpression of p110, p55 PIK , the insulin receptor ␤-subunit, and IRS-1 in Sf9 insect cells has been reported to cause an increase in the specific activity of PI 3-kinase associated with p55 PIK , suggesting that insulin could potentially regulate this isoform (21).
This study compared the ability of insulin to induce recruitment of multiple adapter subunit isoforms into protein complexes containing tyrosine-phosphorylated proteins. The results provide evidence that there are important differences in the mechanisms by which the PI 3-kinase adapter subunit isoforms are regulated by insulin. We identified several major differences in the pattern of association of the p85 variants with tyrosine-phosphorylated protein complexes. First, there was a great variation in the proportion of the total pool of each p85 variant that associated with phosphotyrosine-containing proteins in unstimulated muscle. The levels of the 48-kDa variant associated with antiphosphotyrosine immunoprecipitates in unstimulated muscle were almost undetectable. Fulllength p85␣ was readily detectable in the same immunoprecipitates, although it still represented Ͻ5% of the total pool of full-length p85␣. In contrast, a much higher proportion of the total of the p55 PIK isoform and the 53-kDa and p50 variants was present in antiphosphotyrosine immunoprecipitates from unstimulated muscle. The other area where major differences existed was in the ability of insulin to increase the levels of different adapter subunit variants in antiphosphotyrosine immunoprecipitates. Insulin caused no significant increase at all in the levels of the 46-kDa band or the p55 PIK variant in antiphosphotyrosine immunoprecipitates. However, insulin was able to induce the recruitment of full-length p85␣, p85␤, and the 53-and 48-kDa variants into phosphotyrosine-containing protein complexes. This indicates that these four isoforms are the ones that contribute to PI 3-kinase-dependent insulin signaling pathways in human skeletal muscle. Insulin induced a similar increase in the absolute amount of full-length p85␣ and the 53-and 48-kDa forms in phosphotyrosine-containing protein complexes, indicating that these may each make similar contributions to the magnitude of the insulin stimulation of PI 3-kinase activity. However, it was striking that the ratio of the amount of adapter subunit present in these complexes in the basal and insulin-stimulated states varied greatly between the p85 variants. The biggest difference was in the case of the 48-kDa variant, where insulin increased the levels in antiphosphotyrosine immunoprecipitates Ͼ19-fold compared with basal levels, whereas the smallest -fold increase was for the 53-kDa Human muscle was incubated in the presence (ϩ) or absence (Ϫ) of 1000 microunits/ml insulin for 20 min and then homogenized as described under "Experimental Procedures." 200 g of total human muscle homogenate was incubated with the phosphopeptide beads for 2 h at 4°C. Following extensive washing, the material associated with the beads was eluted in buffer containing 2% SDS and 100 mM dithiothreitol. Phosphopeptide bead eluates (P) and total muscle homogenate (H; 100 g) were separated on an 8% SDS-polyacrylamide gel, transferred to polyvinylidene difluoride, incubated with PY20 antiphosphotyrosine antibody, and visualized by 125 I-protein A. form, where levels in antiphosphotyrosine immunoprecipitates increased only 1.7-fold. If, for example, the different adapter subunits are involved in different signaling complexes and hence different downstream responses, it is possible to see that this offers a mechanism by which insulin could use PI 3-kinasedependent signals to modulate different PI 3-kinase-dependent downstream responses to different extents.
The mechanism by which these adapter subunit variants are differentially recruited into complexes with tyrosine-phosphorylated proteins is not clear. The fact that all the adapter subunit variants have a minimal core containing the p110binding region and the two SH2 domains suggests that all have a similar capability to bind to tyrosine-phosphorylated proteins containing Tyr(P)-Xaa-Xaa-Met motifs. However, our results demonstrate that this is not the case in reality; thus, it appears likely that the N-terminal additions to the minimal adapter subunit core contain elements that modulate the adapter subunit's interaction with the tyrosine-phosphorylated proteins. Consistent with this is the observation that structurally similar forms of the adapter subunit behave in a similar manner. For example, p85␣ and p85␤ both contain an SH3 domain and a Bcr domain. A small amount of the total pool of each of these isoforms is found in antiphosphotyrosine immunoprecipitates from unstimulated muscle, and insulin causes a significant increase in the levels of each of these in antiphosphotyrosine immunoprecipitates. The 53-kDa variant and p55 PIK /p55␥ are the variants found in the greatest amounts in antiphosphotyrosine immunoprecipitates from unstimulated muscle, and both p55 PIK /p55␥ and p55␣/AS53 share significant homology in the N-terminal region as both have a novel 32-amino acid sequence in place of the SH3 and Bcr domains. These results suggest that as yet unidentified structural elements in the N-terminal 32-amino acid extension dictate that p55 PIK /p55␥ and p55␣/AS53 are regulated in a different manner from fulllength p85 and the 48-kDa form.
There are a number of potential mechanisms by which the multiple adapter subunit isoforms could introduce specificity into the PI 3-kinase signaling system. The most obvious is that this allows for a range of independently regulatable links between receptor tyrosine kinases and the PI 3-kinase system, allowing different growth factors to regulate PI 3-kinase in different ways. The recruitment of individual adapter subunit isoforms will result in a parallel recruitment of the associated catalytic subunit into the complex, causing a net increase in production of phosphatidylinositol 3,4,5-trisphosphate. We provide evidence that such a mechanism is functionally important as the adapter subunit variants are differentially recruited in response to insulin. The rate of phosphatidylinositol 3,4,5trisphosphate production can also be regulated by the adapter subunit's ability to modulate the catalytic activity of PI 3-kinase via the adapter subunit following its interaction with Tyr(P)-Xaa-Xaa-Met motifs. It is currently not clear whether all the variants of the adapter subunit are able to equally transduce this signal to p110 to alter catalytic activity. A further layer of specificity could be introduced into the system if the subcellular location of different adapter subunit isoforms was differentially regulated by different signaling inputs. We have previously shown that insulin and platelet-derived growth factor recruit PI 3-kinase activity to different intracellular locations in 3T3-L1 adipocytes (30). The fact that the adapter subunit isoforms differ in the domain structure outside the minimal core consisting of the p110-binding region and the two SH2 domains also may provide a mechanism by which signalling specificity can be achieved. These differences mean that in addition to the SH2 domain-directed interaction, the different adapter subunit isoforms could interact with different signaling complexes as a result of these differences in the N-terminal sequences. For example, the Bcr domain in fulllength p85␣ and p85␤ is potentially able to interact with Rac GTPase (46), whereas the SH3 domain may interact with molecules involved in downstream responses including Src (47), ␣-actinin (48), and dynamin (49). However, the physiological relevance of such interactions has never been fully investigated.
In summary, the results of this study provide evidence that PI 3-kinase activity is necessary for insulin stimulation of facilitated glucose transport in human skeletal muscle. We find that seven PI 3-kinase adapter subunit isoforms exist in this tissue and that these are differentially regulated by insulin. This identifies a mechanism by which insulin could utilize specific PI 3-kinase adapter subunit isoforms to regulate particular downstream responses and thus introduce specificity into the PI 3-kinase signaling system.