p85/p110-type Phosphatidylinositol Kinase Phosphorylates Not Only the D-3, but Also the D-4 Position of the Inositol Ring*

Activation of p85/p110-type phosphatidylinositol (PI) kinase has been implicated in various cellular activities. This PI kinase phosphorylates the D-4 position with a similar or higher efficiency than the D-3 position when trichloroacetic acid-treated cell membrane is used as a substrate, although it phosphorylates almost exclusively the D-3 position of the inositol ring in phosphoinositides when purified PI is used as a substrate. Furthermore, the lipid kinase activities of p110 for both the D-3 and D-4 positions were completely abolished by introducing kinase-dead point mutations in their lipid kinase domains (ΔKinα and ΔKinβ, respectively). In addition, both PI 3- and PI 4-kinase activities of p110α and p110β immunoprecipitates were similarly inhibited by either wortmannin or LY294002, specific inhibitors of p110. Insulin induced phosphorylation of not only the D-3 position, but also the D-4 position. Indeed, overexpression of p110 in Sf9 or 3T3-L1 cells induced marked phosphorylation of the D-4 position to a level comparable to or much greater than that of D-3, whereas inhibition of endogenous p85/p110-type PI kinase via overexpression of dominant-negative p85α (Δp85α) in 3T3-L1 adipocytes abolished insulin-induced synthesis of both. Thus, p85/p110-type PI kinase phosphorylates the D-4 position of phosphoinositides more efficiently than the D-3 positionin vivo, and each of the D-3- or D-4-phosphorylated phosphoinositides may transmit signals downstream.

A variety of growth factors and hormones exert their cellular effects via interactions with specific receptors that possess protein kinase activities. The interaction of most of these ligands with their receptors induces tyrosine kinase activation and phosphorylation of the receptor and/or intracellular substrates. The tyrosine-phosphorylated protein serves as a docking protein for several cytoplasmic substrates with SH2 domains (1)(2)(3). p85/p110-type phosphatidylinositol (PI) 1 kinase has been identified through its ability to associate with these tyrosine-phosphorylated substrates (4,5) and has been thought to be an enzyme that phosphorylates the D-3 position of the inositol ring in phosphoinositides, resulting in formation of PI-3-P, PI-3,4-P 2 , and PI-3,4,5-P 3 (6), based on experiments using purified phosphatidylinositol as a substrate to determine the lipid substrate specificity.
However, our findings that the lipid products of p85/p110type PI kinase in vivo differ markedly from those produced by a conventional in vitro method used to determine the substrate specificity of PI kinase are of great interest. The results of this study show that p110␣ and p110␤ phosphorylate the D-4 position of phosphoinositides more efficiently than the D-3 position in vivo. Thus, p85/p110-type PI kinase may activate downstream targets by increasing not only D-3-phosphorylated (PI-3-P, PI-3,4-P 2 , and PI-3,4,5-P 3 ), but also D-4-phosphorylated (PI-4-P and PI-4,5-P 2 ) phosphoinositides.

EXPERIMENTAL PROCEDURES
Materials-PI and dexamethasone were purchased from Sigma. 3-Isobutyl-1-methylxanthine and 2-deoxy-D-glucose were from Wako Bioproducts, and the enhanced chemiluminescence (ECL) detection system was from Amersham Pharmacia Biotech. [␥-32 P]ATP and [ 32 P]orthophosphate were obtained from ICN. All other reagents from commercial sources were of analytical grade.
Cell Culture-Sf9 cells were maintained in TC-100 medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Life Technologies, Inc.) at 27°C. Cells were harvested 2 days post-baculovirus infection. 3T3-L1 fibroblasts were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% donor calf serum (Life Technologies, Inc.) in an atmosphere of 10% CO 2 at 37°C. Two days after the fibroblasts had reached confluence, differentiation was induced by treating the cells with DMEM containing 0.5 mM 3-isobutyl-1-methylxanthine, 4 mg/ml dexamethasone, and 10% fetal bovine serum for 48 h. Cells were re-fed with DMEM supplemented with 10% fetal bovine serum every other day. Infection with the indicated adenoviruses was carried out on day 3 post-differentiation induction, and the experiments were conducted on day 5, at which point Ͼ90% of the cells expressed the adipocyte phenotype.
Cloning of cDNA-Cloning and construct-reverse transcription-polymerase chain reaction were performed to amplify full-length p110␤ from human embryonic heart cDNA, based on its reported sequence (9). The entire coding region of p110␣ cDNA was obtained as described in our previous report (7). The cDNAs coding for the kinase-dead point mutants of p110␣ (⌬Kin␣) and p110␤ (⌬Kin␤), which lack amino acids 917-950 of p110␣ and amino acids 921-954 of p110␤, were designed as reported previously (10,11). A portion of human GLUT2 cDNA corresponding to residues 510 -524 was ligated to each cDNA to generate catalytic subunits of PI 3-kinase tagged at their C termini.
Gene Transduction-The entire coding regions of p110␣, p110␤, and their kinase-dead point mutants, all of which have an epitope GLUT2 tag at their C termini, were cloned into transfer vectors (either pBacPAK8 or pBacPAK9). The entire coding region of LacZ from Escherichia coli was also cloned into pBacPAK8. Recombinant baculoviruses were obtained by homologous recombination between the recombinant transfer vectors and their parental virus genome according to the manufacturer's instructions (CLONTECH). To obtain recombinant adenoviruses, the expression cosmid cassette pAdexCAwt was ligated with each of the cDNAs coding for either LacZ from E. coli or epitope-tagged p110␣, followed by homologous recombination between the recombinant cosmid cassette and its parental virus genome, as described previously (7). Recombinant adenovirus expressing ⌬p85␣ was prepared as described previously (12).
Immunoprecipitation and Western Blotting-Cells were lysed in PBS containing 1% Triton, 0.35 mg/ml phenylmethylsulfonyl fluoride, and 100 mM sodium vanadate. Cell lysates were centrifuged at 15,000 ϫ g for 10 min at 4°C to remove insoluble materials. The supernatants were incubated with the indicated antibodies, followed by the addition of protein A-Sepharose (Amersham Pharmacia Biotech). Alternatively, in some experiments, lysates were immunoprecipitated with anti-hemagglutinin antibodies, followed by the addition of protein G-Sepharose (Amersham Pharmacia Biotech). The immune complexes were collected by centrifugation, washed with PBS containing 1% Triton X-100, boiled in Laemmli sample buffer containing 100 mM dithiothreitol, and then subjected to SDS-polyacrylamide gel electrophoresis. Immunoblotting was performed with the ECL system according to the manufacturer's instructions. In some experiments, the band intensities were quantitated with a Model GS-525 molecular imager (Bio-Rad).
In Vitro Generation of 32 P-Labeled Phosphoinositides (PI 3-Kinase Assay)-Sf9 cells infected with baculovirus were lysed with PBS containing 1% Nonidet P-40 and 0.35 mg/ml phenylmethylsulfonyl fluoride and then immunoprecipitated with anti-C-terminal GLUT2 tag antibodies and protein A-Sepharose. 3T3-L1 adipocytes (in 24-well culture dishes) were serum-starved for 3 h in DMEM containing 0.2% bovine serum albumin. The cells were incubated with 10 Ϫ6 M insulin for 5 min; washed with ice-cold PBS; and then lysed with PBS containing 1% Nonidet P-40, 0.35 mg/ml phenylmethylsulfonyl fluoride, and 100 mM sodium vanadate. Cell lysates were precleared of insoluble materials by centrifugation (15,000 ϫ g, 4°C, 10 min) and were subjected to immunoprecipitation with anti-phosphotyrosine antibodies and protein A-Sepharose. The PI 3-kinase activity in the immunoprecipitates was measured as described previously (7). When indicated, the membrane fraction of Sf9 cells, prepared as described previously (13), was used as the substrate instead of purified PI from a commercial source. The reaction products were deacylated and analyzed by HPLC. In some experiments, wortmannin or LY294002 at the indicated concentrations was added to the reaction mixture of the in vitro PI kinase assay, and the inhibitory effects on 32 P-labeled phosphoinositide synthesis were investigated.
To exclude the possibility that D-3-phosphorylated phosphoinositides may activate some unknown PI 4-kinase in the membrane fraction, resulting in the synthesis of D-4-phosphorylated phosphoinositides, purified D-3-phosphorylated phosphoinositides were prepared by incubating purified PI, PI-4-P, or PI-4,5-P 2 with nonradioactive ATP and p110␣ immunoprecipitates in vitro. Phosphoinositides labeled with unlabeled phosphate groups on their D-3 positions (PI-3-P, PI-3,4-P 2 , and PI-3,4,5-P 3 ) were separated by TLC, followed by chloroform extraction. Each of them was dried and mixed with Sf9 cell membrane fraction in the presence of [␥-32 P]ATP.
In Vivo Generation of 32 P-Labeled Phosphoinositides-3T3-L1 adipocytes infected with an adenovirus containing p110␣, ⌬p85␣, or control LacZ DNA were phosphate-starved overnight in phosphate-free DMEM (Life Technologies, Inc.), followed by serum starvation for 3 h. [ 32 P]Orthophosphate (0.1 mCi/ml) was added, and the cells were cultured for an additional 2 h. Following the labeling period, cells were incubated with or without 10 Ϫ6 M insulin for 15 min. The reaction was then terminated with an ice-cold PBS wash, followed by the addition of methanol and 1 N HCl (1:1) and finally lipid extraction with chloroform.
A similar procedure was performed in the experiments using Sf9 cells. Sf9 cells infected with the indicated baculovirus were also phosphate-starved for 18 h and then labeled with [ 32 P]orthophosphate (0.1 mCi/ml) for 2 h.

RESULTS AND DISCUSSION
p85/p110-type PI kinase has been identified through its ability to associate with many tyrosine-phosphorylated substrates and has been considered to phosphorylate the D-3 position of the inositol ring in phosphoinositides (6, 16 -18). However, of great interest are our findings that the lipid products of p85/ p110-type PI kinase in vivo differ markedly from those produced by a conventional in vitro method used to determine the substrate specificity of each PI kinase.
First, recombinant baculoviruses that express either p110␣ or p110␤ with a tagged epitope at their COOH terminus were prepared. Sf9 cells were infected with either p110␣ or p110␤ recombinant virus or control LacZ virus, and similar expression levels were obtained among them (Fig. 1A). Immunoprecipitates with antibodies against the tag were subjected to in vitro lipid kinase assays using purified PI as substrates, followed by HPLC analysis (Fig. 1, B and C). The major product was PI-3-P with p110␣ immunoprecipitates, although small amounts of PI-4-P and PI-3,4-P 2 were also detected. Based on the amounts of PI-3-P, PI-4-P, and PI-3,4-P 2 produced, the in vitro lipid kinase activity of p110␣ immunoprecipitates at the D-4 position was estimated to be 4% of that at the D-3 position. On the contrary, immunoprecipitates of p110␤, another isoform of the catalytic subunit, produced Ͻ1% of the PI-3-P produced by p110␣ immunoprecipitates, and no generation of PI-4-P or PI-3,4-P 2 was detected. These results indicate that the catalytic subunits of p85/p110-type PI kinase, p110␣ and p110␤, are indeed PI kinases for the D-3 position.
Quite different results were obtained when trichloroacetic acid-treated cell membranes were used as substrates for p110␣ or p110␤, instead of purified PI. Treating the cell membranes with 10% trichloroacetic acid apparently denatures the protein contained in the cell membrane fraction, and trichloroacetic acid-treated cell membrane did not significantly incorporate 32 P into the lipid (data not shown). In the assay using trichloroacetic acid-treated membrane from Sf9 cells as a substrate (Fig. 2B), p110␣ immunoprecipitates produced PI-4-P in an amount comparable to that of PI-3-P. Similarly, PI-4-P and PI-3-P were detected in p110␤ immunoprecipitates, and the level of the former was ϳ5.4 times the level of the latter. The amount of PI-3-P produced in p110␤ immunoprecipitates was 19% of that produced in p110␣ immunoprecipitates, whereas nearly the same amount of PI-4-P was produced in p110␣ and p110␤ immunoprecipitates.
To rule out the possible contamination of PI 4-kinase in the p110 immunoprecipitates and to confirm that p110 itself possesses PI 4-kinase activity, we performed two experiments using kinase-dead point mutants of p110 (Fig. 2, A and B) and lipid kinase inhibitors (Fig. 2, C and D). First, p110␣, p110␤, and their kinase-dead point mutants (⌬Kin␣ and ⌬Kin␤, respectively) were expressed at a similar level in Sf9 cells ( Fig.  2A). It was shown that the lipid kinase activities of the p110 proteins for both the D-3 and D-4 positions were completely abolished by introducing kinase-dead point mutations in their lipid kinase domains (Fig. 2B). In addition, both PI 3-and PI 4-kinase activities of p110␣ or p110␤ immunoprecipitates were inhibited similarly by either wortmannin or LY294002 (Fig. 2,  C and D). Wortmannin and LY294002 are recognized as specific inhibitors of p110, with much higher concentrations being necessary for the inhibition of other lipid kinases (19,20). Subsequently, to exclude the possibility that D-3-phosphoryl-ated phosphoinositides may activate some unknown PI 4-kinase in the membrane fraction, resulting in the synthesis of D-4-phosphorylated phosphoinositides, purified nonradioactive D-3-phosphorylated phosphoinositides (PI-3-P, PI-3,4-P 2 , or PI-3,4,5-P 3 ) were added to a mixture of trichloroacetic acid-treated cell fraction and [␥-32 P]ATP. As a result, no D-4-phosphorylated phosphoinositide labeled with 32 P was observed (Fig. 2B, last column), which excludes the possibility mentioned above. In addition, when purified PI-3-P, PI-3,4-P 2 , or PI-3,4,5-P 3 labeled with 32 P at the D-3 position was incubated with trichloroacetic acid-treated cell membrane, no phosphoinositide labeled with 32 P at the D-4 position was synthesized (data not shown). This indicates that transfer of 32 P from the D-3 position to the D-4 position is quite unlikely. Taken together, it is very likely that the PI 4-kinase activities observed in the p110 immunoprecipitate are attributable to p110 itself.
Subsequently, to demonstrate that p85/p110-type PI kinase actually exhibits PI 4-kinase activity in vivo, either p110␣ or p110␤ was first overexpressed in Sf9 cells, followed by [ 32 P]orthophosphate labeling. Sf9 cells overexpressing either p110␣ or p110␤ at a similar level significantly accumulated each phosphoinositide, as compared with control cells, the exception being PI-3,4,5-P 3 in p110␤-overexpressing cells (Fig. 3). The increases in PI-4-P and PI-4,5-P 2 were comparable to the increase in PI-3-P induced by p110␣ overexpression and even greater than the increases in PI-3,4-P 2 and PI-3,4,5-P 3 , which suggests that p110 overexpression induces accumulation of not only D-3-phosphorylated, but also D-4-phosphorylated phosphoinositides in vivo. Essentially the same results were obtained with the overexpression of both p110 and its regulatory subunit, p85 (data not shown). p110␤ overexpression induced accumulation of D-4-phosphorylated phosphoinositides at a much higher level than that of D-3 in Sf9 cells, which suggests that p110␤ may function as a PI 4-kinase predominantly in intact cells. Further study should be addressed to determine the mechanism that causes the difference in substrate specificity between p110␣ and p110␤.
Finally, to demonstrate that endogenous p85/p110 produces not only D-3-phosphorylated, but also D-4-phosphorylated phosphoinositides, and to examine the effect of insulin on the level of phosphorylation of phosphoinositides, either p110␣ or dominant-negative p85␣ was overexpressed in 3T3-L1 adipocytes, and the cellular level of each phosphoinositide in the absence or presence of insulin was investigated. p110␣ was overexpressed in 3T3-L1 adipocytes using an adenovirus expression system (Fig. 4A, upper panel) at a level approximately five times that expressed endogenously in 3T3-L1 cells (lower panel). The cellular ATP level was essentially unaltered by p110 overexpression (data not shown). The amount of overexpressed p110␣ with the adenovirus expression system is much smaller than that with the baculovirus expression system. Thus, increases in the amounts of PI-3,4-P 2 , PI-4,5-P 2 , and PI-3,4,5-P 3 accumulated in 3T3-L1 adipocytes (Fig. 4B) were smaller than those in Sf9 cells (Fig. 3), but such increases were still obvious. Interestingly, the increase in PI-4-P in 3T3-L1 adipocytes overexpressing p110␣ was much larger than that in PI-3-P, although they were almost comparable in Sf9 cells overexpressing p110␣. The reason for these differences according to cell type remains unclear, but the conformation of p110 may differ somewhat by an unknown modification, such as serine/threonine phosphorylation, according to whether p110 is expressed in Sf9 cells or in 3T3-L1 adipocytes.
Insulin showed no additive effect on the amount of each phosphoinositide in p110␣-overexpressing cells, which also agrees well with our previous report that no additive effect of insulin and p110␣ overexpression on GLUT4 translocation to the cell surface was observed (7). In addition, very clear results were obtained in the experiment using ⌬p85␣ with a hemagglutinin tag at its C terminus (Fig. 4C, upper panel). By overexpressing ⌬p85␣, which inhibits the normal activation of endogenous p110, we were able to investigate the function of endogenous p85/p110. In this experiment, the expression level of ⌬p85␣ was ϳ10 times that of endogenous p85␣ (Fig. 4C,  middle panel), which almost completely inhibited the insulininduced activation of endogenous p110 (lower panel). To determine precisely the effect of insulin on phosphoinositide phosphorylation, we calculated the amount of freshly labeled phosphoinositide during 15 min of insulin stimulation by subtracting the radioactivity from that labeled with 32 P during the 2-h incubation before insulin treatment. As indicated in Fig. 4D (by the numbers outside the parentheses), in the control cells, insulin raised 32 P-labeled PI-3-P, PI-4-P, and PI-4,5-P 2 to 7.3, 35.0, and 3.4 times the respective levels of their freshly labeled counterparts without stimuli. By overexpressing ⌬p85␣, insulin-stimulated production of PI-3,4-P 2 , PI-4,5-P 2 , and PI-3,4,5-P 3 decreased to nearly undetectable levels and that of PI-3-P and PI-4-P decreased by ϳ70 and 98%, respectively (Fig.  4D). These results clearly indicate that insulin induces accumulation of both D-3-and D-4-phosphorylated phosphoinositides via the activation of endogenous p85/p110-type PI kinase. In addition, it should be noted that p85/p110-type PI kinase, but none of the other PI-4 kinases, appears to be critical for the insulin-induced synthesis of D-4-phosphorylated phosphoinositides in 3T3-L1 cells.
As yet, there have been no reports of p85/p110-type PI kinase with PI 4-kinase activities or that growth factors/hormones stimulate PI 4-kinases. There are two possible explanations for these going unnoticed. One is that researchers have assumed that the in vitro PI kinase assay using purified PI as a substrate reflects the physiological reaction that takes place intracellularly. However, phosphoinositides are a rather minor component of phospholipids, usually composing Ͻ10%. The remaining phospholipids are phosphatidylcholine, phosphatidylethanolamine, and phosphatidylserine. Without these components around phosphoinositides, the conformation of lipid kinases might be altered. This may explain the exhibition by p85/p110-type PI kinase of a substrate specificity different from that observed with purified PI, as was reported for some FIG . 2. In vitro generation of phosphoinositides from cell membranes by p110␣, p110␤,

and their kinase-dead point mutants expressed in Sf9 cells (A and B) and the effect of PI 3-kinase inhibitors on lipid kinase activities (C and D).
A and B, Sf9 cells were infected with recombinant baculoviruses containing p110␣, p110␤, ⌬Kin␣, or ⌬Kin␤. These cells were lysed, and lysates were immunoprecipitated with anti-C-terminal GLUT2 tag antibodies. A, the immunoprecipitates were subjected to SDS-polyacrylamide gel electrophoresis, followed by immunodetection with anti-C-terminal GLUT2 tag antibodies. B, the washed immunoprecipitates were subjected to PI kinase assay. The membrane fraction from uninfected Sf9 cells was also prepared. After 10% trichloroacetic acid treatment, membrane fractions were incubated with the immunoprecipitates together with [␥-32 P]ATP, as described under "Experimental Procedures." D-3-Phosphorylated Phosphoinositide indicates PI-3-P, PI-3,4-P 2 , PI-3,4 -5-P 3 , or a mixture of these three, all of which were prepared by incubating purified PI, PI-4-P, or PI-4,5-P 2 with nonradioactive ATP and p110␣ immunoprecipitates, followed by TLC separation and chloroform extraction. They were mixed with trichloroacetic acid-treated cell membrane and [␥-32 P]ATP. The 32 P-labeled phosphoinositides generated were deacylated and separated by HPLC. Three other separate experiments yielded similar results. C and D, immunoprecipitates of p110␣ or p110␤ expressed in Sf9 cells were incubated with trichloroacetic acid-treated Sf9 cell membrane and [␥-32 P]ATP in the presence of various concentrations of wortmannin or LY294002. The amount of glycerophosphoinositides was analyzed by HPLC. PI 3-kinase activity (squares) represents the amount of PI-3-P and PI-3,4-P 2 , and PI 4-kinase activity (triangles) represents the amount of PI-4-P and PI-3,4-P 2 . Lysates were prepared from Sf9 cells overexpressing p110␣ (open symbols) or p110␤ (closed symbols). PI kinase assay was performed in the presence of wortmannin (C) or LY294002 (D). Three other separate experiments yielded similar results.
isoforms of protein kinase C (21). In addition, recently, the substrate specificity of PI 4-P-5-kinases, which has been reported based on in vitro assay, was also corrected (15). Thus, it is understandable that results obtained from the artificial in vitro assay do not reflect the actual intracellular situation. The other possible reason is that many investigators employed stimulation with epidermal growth factor, nerve growth factor, or platelet-derived growth factor to analyze the function of p85/p110-type PI kinase (22)(23)(24). Stimulation with these growth factors does activate p85/p110-type PI kinase, but phos-pholipase C␥, which hydrolyzes PI-4,5-P 2 (25), is also activated. Thus, as reported previously, epidermal growth factor, nerve growth factor, or platelet-derived growth factor does not increase the amount of PI-4,5-P 2 , despite the possibly generation by activated p85/p110-type PI kinase of PI-4,5-P 2 probably via PI-4-P production. Because of this complicated alteration in the amount of D-4-phosphorylated phosphoinositide, it seems that earlier studies did not pay attention to the effect of p85/p110type PI kinase on the amount of D-4-phosphorylated phosphoinositide. In contrast, insulin stimulation does not lead to phos-FIG. 3. Generation of 32 P-labeled lipid products in Sf9 cells overexpressing p110␣ or p110␤. Sf9 cells overexpressing LacZ, p110␣, or p110␤ were phosphate-starved as described under "Experimental Procedures." [ 32 P] Orthophosphate (0.1 mCi/ml) was added; the cells were cultured for an additional 2 h; and lipid was extracted with chloroform. The extracted 32 P-labeled lipid was analyzed by HPLC after being deacylated. The amounts of 32 P-labeled lipids generated were determined with an on-line radiochemical detector. Two other separate experiments yielded similar results.

FIG. 4. Effect of overexpression of p110␣ (A and B) or ⌬p85␣ (C and D) on phosphoinositide accumulation in 3T3-L1 cells. A and B,
3T3-L1 adipocytes were infected with recombinant adenovirus containing LacZ or p110␣. Lysates were immunoprecipitated and immunoblotted with anti-C-terminal GLUT2 tag antibodies (A, upper panel) or with antibodies to the p110␣ C terminus (lower panel). Cells were labeled with [ 32 P]orthophosphate, incubated with or without insulin, and analyzed for 32 P-labeled phosphoinositides (B). C and D, 3T3-L1 adipocytes were infected with recombinant adenovirus containing LacZ or ⌬p85␣. Lysates were immunoprecipitated (ip) and immunoblotted (blot) with antihemagglutinin tag antibodies (C, upper panel) or with anti-p85␣ antiserum (middle panel) or immunoprecipitated with anti-phosphotyrosine antibodies (␣PY) followed by PI 3-kinase assay (lower panel). As a positive control, immunoprecipitates from insulin-stimulated 3T3-L1 adipocytes overexpressing p85␣ were also subjected to PI 3-kinase assay. Cells were labeled with [ 32 P]orthophosphate, incubated with or without insulin, and analyzed for 32 P-labeled phosphoinositides (D). Numbers in parentheses represent the amount of 32 P-labeled phosphoinositides quantitated. Numbers outside parentheses represent the amount of newly 32 P-labeled phosphoinositides during the last 15 min without or with insulin stimulation. These values were obtained in experiments in which 3T3-L1 cells infected with each recombinant adenovirus were 32 P-labeled for either 2 h or 2 h and 15 min. In control cells, the amounts of 32 P-labeled PI-3-P, PI-4-P, and PI-4,5-P 2 were 354, 2402, and 236 cpm at 2 h and 412, 3471, and 360 cpm at 2 h and 15 min, respectively. In cells overexpressing ⌬p85␣, the amounts were 210, 2107, and 179 cpm at 2 h and 282, 2611, and 213 cpm at 2 h and 15 min, respectively. Two other separate experiments yielded similar results. pholipase C␥ activation (26). Thus, the generation of D-4phosphorylated phosphoinositides (PI-4-P and PI-4,5-P 2 ) by p85/p110-type PI kinase became as apparent as that of D-3phosphorylated phosphoinositides (PI-3-P, PI-3,4-P 2 , and PI-3,4,5-P 3 ). However, even in the case of insulin stimulation, since the cell has significant basal PI 4-kinase activity, if cells are prelabeled with 32 P for too long a period (12 h or longer), the effect of insulin for a very short time would be difficult to detect. This is the reason we labeled the cells for 2 h. In the living body, cells are always stimulated by many factors in the serum. It is thus reasonable to consider the PI 4-kinase activity of p85/p110 to be important for the regulation of D-4-phosphorylated phosphoinositides since p85/p110 is likely to be the only enzyme that phosphorylates the D-4 position in response to insulin.
The importance of D-4-phosphorylated phosphoinositides has been established by several investigators. PI-4,5-P 2 regulates the function of several actin-binding proteins (27)(28)(29)(30)(31) and is likely to mediate certain processes of protein sorting or secretion (32,33), membrane ruffling (34), and the formation of focal adhesions (28), some of which reportedly occur in response to growth factor/hormone stimuli in parallel with p85/p110type PI kinase activation. Thus, some of these cellular events might be attributable to D-4-phosphorylated phosphoinositides produced by p85/p110-type PI kinase. In addition, PI-4,5-P 2 plays an important role as a precursor for the second messengers diacylglycerol, inositol 1,4,5-P 3 (25), and PI-3,4,5-P 3 (22). In this respect, the cellular content of PI-4,5-P 2 might affect the signal intensity of phospholipase C␥. Our surprising findings also raise the possibility that the substrate specificity of various lipid kinases and phosphatases, when incubated with purified phosphoinositide in an in vitro assay, might be different from the true, i.e. cellular, substrate specificity.