Identification and Characterization of the Autophosphorylation Sites of Phosphoinositide 3-Kinase Isoforms β and γ*

Class I phosphoinositide 3-kinases (PI3Ks) are bifunctional enzymes possessing lipid kinase activity and the capacity to phosphorylate their catalytic and/or regulatory subunits. In this study, in vitro autophosphorylation of the G protein-sensitive p85-coupled class IA PI3Kβ and p101-coupled class IB PI3Kγ was examined. Autophosphorylation sites of both PI3K isoforms were mapped to C-terminal serine residues of the catalytic p110 subunit (i.e. serine 1070 of p110β and serine 1101 of p110γ). Like other class IA PI3K isoforms, autophosphorylation of p110β resulted in down-regulated PI3Kβ lipid kinase activity. However, no inhibitory effect of p110γ autophosphorylation on PI3Kγ lipid kinase activity was observed. Moreover, PI3Kβ and PI3Kγ differed in the regulation of their autophosphorylation. Whereas p110β autophosphorylation was stimulated neither by Gβγ complexes nor by a phosphotyrosyl peptide derived from the platelet-derived growth factor receptor, autophosphorylation of p110γ was significantly enhanced by Gβγ in a time- and concentration-dependent manner. In summary, we show that autophosphorylation of both PI3Kβ and PI3Kγ occurs in a C-terminal region of the catalytic p110 subunit but differs in its regulation and possible functional consequences, suggesting distinct roles of autophosphorylation of PI3Kβ and PI3Kγ.

pathways in cells, whereas no activation of Akt/protein kinase B by this mutant occurred (22).
Autophosphorylation of both catalytic and regulatory subunits of PI3K␤ and PI3K␥ has been proposed, but many questions regarding the autophosphorylation sites and the functional relevance of these autophosphorylation events remain unanswered. Therefore, in the present study, we examined the in vitro autophosphorylation of the G protein-sensitive class I PI3K␤ and PI3K␥. Recombinant heterodimeric PI3Ks were purified and analyzed for their autophosphorylation. Phosphorylated amino acids were identified, and the effect of autophosphorylation on PI3K lipid kinase activity was studied. Interestingly, we observed significant differences in the regulation and functional consequences of the autophosphorylation of PI3K␤ and PI3K␥.
Recombinant viruses expressing hexahistidine-tagged p110␤, p110␥, and mutants thereof were generated using the Bac-to-Bac Expression System (Invitrogen) following the manufacturer's instructions. Expression and purification of PI3K isoforms were carried out according to published protocols (11) with the exception that the partially purified proteins were subjected to an additional chromatographic step on a 1-ml Resource Q fast protein liquid chromatography column (Amersham Biosciences). For that purpose, proteins were diluted in buffer A (20 mM Tris/HCl, pH 8.0, 10 mM ␤-mercaptoethanol) and loaded onto the column. The column was subsequently washed with buffer A, and proteins were eluted with a linear gradient of 0 -500 mM NaCl in buffer A.
G␤␥ Complexes and Peptides-Expression and purification of recombinant G␤ 1 ␥ 2-His complexes was carried out as published (11). Purified proteins were quantified by Coomassie Blue staining following SDS-PAGE with bovine serum albumin as the standard (23). The tyrosinephosphorylated peptide used in this study, CGGpYMDMSKDESVD-pYVPMLDM (where pY represents phosphotyrosine), was derived from the human platelet-derived growth factor receptor (24) and kindly donated by Dr. Andreas Steinmeyer (Schering AG, Berlin, Germany). A nonphosphorylated peptide served as a control and had no effect on PI3K enzymatic activity. The peptides derived from the C termini of p110␤ and p110␥ (WMAHTVRKDYRS and WFLHLVLGIKQGEKHSA, respectively) were kindly provided by Dr. Michael Beyermann (Forschungsinstitut fü r Molekulare Pharmakologie, Berlin, Germany).
Cell Culture, Transfection, and Preparation of Cell Lysates-HEK293 cells (American Type Culture Collection, Manassas, VA) were grown in minimal essential medium with Earle's salts supplemented with 10% fetal calf serum and antibiotics. Subconfluent cells were transfected in 3-cm dishes with pcDNA3-fMLP receptor (0.2 g), pcDNA3-p101 (0.4 g), and pcDNA3-p110␥ (0.4 g) variants, using the FuGene 6 transfection reagent (Roche Molecular Biochemicals) following the manufacturer's instructions. For preparation of whole cell lysates, cells were directly lysed in sample buffer according to Laemmli (39).
Gel Electrophoresis, Immunoblotting, and Antibodies-Characterization of the monoclonal antibody against p110␥ has been described elsewhere (7). The polyclonal anti-extracellular signal-regulated kinase and anti-phospho-Akt antibodies were purchased from New England Biolabs. Whole cell lysates were fractionated by SDS-PAGE and blotted onto nitrocellulose membranes (Amersham Biosciences). Visualization of specific antisera was performed using the ECL chemiluminescence system (Amersham Biosciences) according to the manufacturer's instructions.
Lipid Kinase Assay-In vitro lipid kinase activity was determined basically as described previously (11). In brief, assays were conducted in a final volume of 50 l, containing 0.1% bovine serum albumin, 1 mM EGTA, 0.2 mM EDTA, 7 mM MgCl 2 , 100 mM NaCl, 40 mM HEPES, pH 7.4, 1 mM dithiothreitol, 1 mM ␤-glycerophosphate (vesicle buffer) with some modifications. Lipid vesicles (30 l containing 320 M phosphatidylethanolamine, 300 M phosphatidylserine, 140 M phosphatidylcholine, 30 M sphingomyelin, and 40 M PtdIns-4,5-P 2 in vesicle buffer) were mixed with stimuli as indicated and incubated on ice for 10 min. It should be noted that we ensured that the effects of G␤␥ on PI3K activity were not affected by their detergent-containing vehicles. Thereafter, the enzyme (20 -100 ng of PI3K␤ or 2-10 ng of PI3K␥) was added, and the mixture was incubated for further 10 min at 4°C in a final volume of 40 l. The assay was started by adding 40 M ATP (1 Ci of [␥-32 P]ATP; PerkinElmer Life Sciences) in 10 l of vesicle buffer. After an incubation period of 15 min (unless otherwise stated) at 36°C, the reaction was stopped by adding 150 l of ice-cold 1 N HCl, and lipids were extracted with 450 l of chloroform/methanol (1:1). Following centrifugation, the organic phase was washed twice with 150 l of 1 N HCl. Subsequently, 40 l of the organic phase were resolved on potassium oxalate-pretreated TLC plates (Whatman, Maidstone, UK) using a mixture of 35 ml of 2 N acetic acid and 65 ml of n-propyl alcohol as the mobile phase. Dried TLC plates were exposed to Fuji imaging plates, and autoradiographic signals were quantitated with a BAS 1500 Fuji-Imager (Raytest, Straubenhardt, Germany).
Protein Kinase Assay-In vitro protein kinase activity was determined as described for the lipid kinase activity with some modifications. The assay volume was 25 l (2 Ci of [␥-32 P]ATP/tube), the vesicle buffer contained 0 -10 mM MgCl 2 and/or 0 -10 mM MnCl 2 as indicated, and lipid vesicles lacked PtdIns-4,5-P 2 . The reaction was stopped after an incubation period of 30 min (unless otherwise stated) at 36°C by adding 10 l of 4ϫ sample buffer according to Laemmli (39). Following separation on SDS-polyacrylamide gels, proteins were transferred to nitrocellulose membranes. Dried membranes were exposed to Fuji imaging plates, and autoradiographic signals were quantitated. For the The equivalent serine residue in p110␤ is marked in boldface type. B, heterodimeric recombinant PI3K␣ (GST-p110␣/p85) and PI3K␤ (His-p110␤/p85) were purified from Sf9 cells, and the proteins were separated by SDS-PAGE and analyzed by Coomassie staining. PI3K␣ and PI3K␤ were assayed for the incorporation of 32 P into the catalytic and regulatory subunits in the presence of either Mn 2ϩ (2 mM) or Mg 2ϩ (7 mM) as indicated. Shown are representative autoradiographs and the corresponding Coomassie-stained gels as loading controls.
determination of PI3K autophosphorylation sites, 20 -50 g of PI3K␤ or PI3K␥ were phosphorylated in a final volume of 1,500 l. Samples were subjected to SDS-PAGE, and gels were stained with Coomassie Blue. Calculation of the stoichiometry of p110 autophosphorylation was based on the specific activity of [␥-32 P]ATP incorporated into p110, and counts were determined by Cerenkov counting. The amount of p110 was estimated from the Coomassie-stained gel by comparison with stained bovine serum albumin standards.
Phosphoamino Acid Analysis-Autophosphorylated PI3K␥ was sub-jected to SDS-PAGE and blotted onto a polyvinylidene difluoride membrane (Millipore Corp.), and the phosphorylated p110␥ band was excised. The protein was hydrolyzed in 6 N HCl for 1 h at 110°C. The sample was vacuum-dried, and amino acids were resuspended in 5 l of pH 1.9 buffer (0.078% (v/v) acetic acid, 0.025% (v/v) formic acid) containing 2 g each of phosphoserine, phosphothreonine, and phosphotyrosine as internal standards. The sample was applied to a cellulose thin layer plate, and electrophoresis in the first dimension was carried out in pH 1.9 buffer at 550 V for 1 h. After drying the plate, an electrophoretic separation in the second dimension was carried out in pH 3.5 buffer (0.05% (v/v) acetic acid, 0.005% (v/v) pyridine, 0.5 mM EDTA) at 500 V for 50 min. The unlabeled phosphoamino acids were visualized by spraying the plate with 0.2% (w/v) ninhydrin in acetone, and the radiolabeled phosphoamino acids were detected by autoradiography.
Determination of p110␤ and p110␥ Autophosphorylation Sites-Gelexcised autophosphorylated p110 spots (100 -150 pmol) were washed with 50% (v/v) acetonitrile in 25 mM ammonium bicarbonate, shrunk by dehydration in acetonitrile, and dried in a vacuum centrifuge. Disulfide bonds were reduced by incubation with 10 mM dithiothreitol in 100 mM ammonium bicarbonate for 45 min at 55°C. Alkylation was performed by replacing the dithiothreitol solution with 55 mM iodoacetamide in 100 mM ammonium bicarbonate. Following a 20-min incubation period at 25°C in the dark, the gel pieces were washed with 50% (v/v) acetonitrile in 25 mM ammonium bicarbonate, shrunk by dehydration in acetonitrile, and dried in a vacuum centrifuge. The gel pieces were incubated overnight at 37°C in 5 mM ammonium bicarbonate, containing 1 g of chymotrypsin (sequencing grade; Roche Molecular Biochemicals), for p110␥ or at room temperature in 50% (v/v) trifluoroacetic acid, containing 10 mg/ml cyanogen bromide, for p110␤. To extract the peptides, 0.5% (v/v) trifluoroacetic acid in acetonitrile was added, and the separated liquid was dried under vacuum, redissolved in 5 l of buffer B (0.1% (v/v) formic acid), and loaded onto a Vydac C18 column (150 ϫ 1 mm, 5 m, type 218 TP 5115) for micro-liquid chromatography separation. Elution was performed using a linear gradient of 5-80% buffer C in 60 min at an eluent flow rate of 30 l/min. Buffer C was 0.1% (v/v) formic acid in acetonitrile/water (8:2, v/v), containing 0.1% (v/v) formic acid. Fractions were collected, their radioactivity was determined by Cerenkov counting, and phosphopeptides were identified by matrixassisted laser desorption/ionization mass spectrometry (MALDI-MS). MALDI-MS measurements were performed on a Voyager-DE STR Bio-Spectrometry work station MALDI-TOF mass spectrometer (Perseptive Biosystems, Inc., Framingham, MA) using ␣-cyano-4-hydroxycinnamic acid as the matrix. The program FindMod (available on the World Wide Web at expasy.ch/tools/findmod) was used to interpret the MS spectra of protein digests. Amino acid sequences of the phosphopeptides were determined by nanoelectrospray tandem mass spectrometry (nanoESI-MS/MS). The liquid chromatography fractions were lyophilized and redissolved in 5 l of 1% (v/v) formic acid in methanol/water (1:1, v/v). The MS/MS measurements were performed with a nanoelectrospray hybrid quadrupole mass spectrometer Q-TOF (Micromass, Manchester, UK). The collision gas was argon at a pressure of 6.0 ϫ 10 Ϫ5 millibar in the collision cell.  2. p110␤ autophosphorylation at a C-terminal serine residue. A, heterodimeric recombinant PI3K␤ purified from Sf9 cells was subjected to SDS-PAGE and visualized by Coomassie staining. Apparent molecular masses (kDa) of marker proteins are indicated. B, peptide mass fingerprint analysis of p110␤. Autophosphorylated p110␤ was digested in gel using cyanogen bromide, the resulting peptides were separated by reversed-phase HPLC, and the fractions were analyzed by MALDI-MS. The peak with m/z 1312.65 (calculated m/z 1312.62) corresponds to the phosphorylated sequence 1061 AHTVRKDYRpS 1070 . C, p110␤ in which serine 1070 was mutated to alanine shows a loss of autophosphorylating activity. Heterodimeric PI3K␤ either containing wild-type p110␤ (WT) or a p110␤ mutant (S1070A) was subjected to a protein kinase assay in the presence of Mn 2ϩ . Shown are one representative autoradiograph and the corresponding Coomassie-stained protein bands as loading control. D, lipid kinase activity of mutant p110␤. Equal amounts of purified heterodimeric PI3K␤ (His-p110␤/p85) either containing wild-type p110␤ (WT) or a p110␤ mutant (S1070D, S1070E, or S1070A) were tested for their enzymatic activity in a lipid kinase assay in the absence (Ϫ) and presence (ϩ) of 120 nM purified G␤ 1 ␥ 2-His , 100 nM tyrosine-phosphorylated peptide, and both stimuli. Experiments were carried out in the presence of Mg 2ϩ . Indicated are mean values Ϯ S.D. of three independent experiments. E, Mn 2ϩ -dependent protein kinase activity of PI3K␤ in the presence of increasing amounts of a synthetic peptide derived from the C terminus of p110␤. One representative autoradiograph of three independent experiments is shown.

RESULTS
In Vitro Autophosphorylation of Class I A PI3Ks-Class I A PI3K␣ and PI3K␦ isoforms phosphorylate either their p85 adaptor subunit and/or the catalytic p110 subunit itself (14,15). Autophosphorylation of p110␦ occurs on serine 1039 within the C terminus. Alignment of the C termini of class I A catalytic subunits shows that p110␣, which does not autophosphorylate, contains no C-terminal serine, whereas p110␤ does have one (serine 1070) (Fig. 1A). Recent reports have described autophosphorylation of both subunits of PI3K␤ (i.e. p85 and p110␤) (25)(26)(27). In order to analyze autophosphorylation of PI3K isoforms in vitro, we expressed recombinant heterodimeric PI3K␣ and PI3K␤ in insect cells and measured their protein kinase activities (Fig. 1B). As anticipated, the p85 adaptor of PI3K␣ was phosphorylated in the presence of Mn 2ϩ only (see Fig. 1B, left panel). In contrast to p110␣, p110␤ autophosphorylated its catalytic subunit. This autophosphorylation of p110␤ was also largely Mn 2ϩ -dependent, since in vitro phosphorylation levels in the presence of Mg 2ϩ reached a maximum of only 5-10% compared with the level observed in the presence of Mn 2ϩ (see Fig. 1B, right panel). Furthermore, in the presence of Mn 2ϩ , a small but significant phosphorylation of the p85 subunit of PI3K␤ was evident. These data indicate that both subunits are phosphorylated, with p110␤ being the main substrate of PI3K␤ autophosphorylation.
p110␤ Autophosphorylates a C-terminal Serine Residue-Speculating that p110␤ autophosphorylates its C terminus, the in vitro phosphorylated and [ 32 P]phosphate-labeled protein ( Fig. 2A) was cleaved with cyanogen bromide in order to generate a C-terminal peptide, which was then analyzed by mass spectrometry. The resulting peptides were separated by reversed-phase HPLC, and the radioactivity of each fraction was determined. The main radioactive fraction was examined by MALDI-MS, and a peptide corresponding to the phosphorylated C terminus of p110␤ (m/z 1312.65) could be identified (Fig. 2B). Sequencing of this phosphopeptide by nanoESI-MS/MS revealed the 1061 AHTVRKDYRpS 1070 (where pS represents phosphoserine) sequence of p110␤ and serine 1070 as the site of autophosphorylation (Table I). In order to verify this finding, a p110␤ mutant in which serine 1070 was changed to alanine was created, and autophosphorylation of this mutant was compared with the wild-type enzyme in the presence of Mn 2ϩ . Fig. 2C shows that no significant phosphate incorporation into the mutant p110␤S1070A of PI3K␤ took place while this mutant was still catalytically active as a lipid kinase (see below). Hence, results obtained by mass spectrometric and mutagenic analysis demonstrate that serine 1070 represents in fact the main site of autophosphorylation in p110␤.
Both phosphorylation of the p85 adaptor by p110␣ and autophosphorylation of p110␦ down-regulate the enzymes' lipid kinase activities (14 -16, 20). p110␤, like p110␦, autophosphorylates a serine residue at the extreme C terminus, which may also affect the catalytic activity of PI3K␤. In order to test this assumption, we measured lipid kinase activities of p110␤ variants in which serine 1070 was mutated. Since the lipid kinase activity of PI3K␤ can be synergistically stimulated by G␤␥ complexes and a tyrosine-phosphorylated peptide derived from the platelet-derived growth factor receptor, we measured formation of PtdIns-3,4,5-P 3 under basal conditions and after stimulation of PI3K␤ variants with either stimuli in the presence of Mg 2ϩ (11). Wild-type p110␤ and the nonphosphorylating p110␤S1070A mutant exhibited the same enzymatic activity under basal conditions and after stimulation with G␤␥, phosphotyrosyl peptide, or both stimuli (Fig. 2D). This finding indicates that serine 1070 is not essential for the catalytic activity of p110␤. However, purified mutant PI3K␤ het- erodimers are less stable than the wild-type enzyme (data not shown). In order to mimic the effects of p110␤ autophosphorylation, mutants of p110␤ containing the negatively charged aspartic and glutamic acid instead of serine 1070 were employed. These p110␤S1070D/E mutants no longer autophosphorylated (data not shown). Moreover, as shown in Fig. 2D, the lipid kinase activity of either p110␤S1070D/E mutant was reduced by 4 -7-fold under basal conditions and following stimulation with G␤␥ and phosphotyrosyl peptides. This implies a regulatory function of the p110␤ autophosphorylation.
Next we created a peptide corresponding to the C terminus of p110␤ (WMAHTVRKDYRS) as a pseudosubstrate in order to test whether the protein kinase activity of the autophosphorylated p110␤ was still intact. Applying mass spectrometry, no phosphorylation of this peptide by wild-type PI3K␤ was observed (data not shown). Moreover, as indicated in Fig. 2E, increased amounts of the C-terminal peptide did not influence the autophosphorylation of p110␤ by competition. Therefore, we assume that protein phosphorylation by PI3K␤ requires highly specific protein-protein interactions, and due to the lack of appropriate substrates the effect of p110␤ autophosphorylation on the protein kinase activity of the enzyme remains unknown so far.
Regulation of p110␤ Autophosphorylation-The finding that autophosphorylation of p110␤ on serine 1070 results in downregulation of the lipid kinase activity of PI3K␤ (see Fig. 2D) suggests a regulatory function of this autophosphorylation. Therefore, one may suppose that the protein kinase activity like the lipid kinase activity of PI3K␤ is controlled by cell surface receptors. In order to address this hypothesis, we compared both kinase activities after incubation of PI3K␤ with increasing concentrations of G␤␥ complexes and the tyrosinephosphorylated peptide. As indicated, Fig. 3, A and B, shows that the lipid kinase activity of PI3K␤ was stimulated in a concentration-dependent manner by G␤␥ (EC 50 ϭ 20 nM) or phosphotyrosyl peptide (EC 50 ϭ 5 nM). In contrast, neither G␤␥ nor phosphotyrosyl peptide stimulated p110␤ autophosphorylation (see Fig. 3, A and B). Moreover, a combination of both stimuli led to a remarkable synergistic activation of PI3K␤ lipid kinase activity (see Fig. 2B) (11). However, even under these conditions, autophosphorylation of p110␤ was not en-hanced regardless of whether Mg 2ϩ , Mn 2ϩ , or mixtures thereof were present (data not shown). These observations suggest a high level of basal p110␤ autophosphorylation. Nonetheless, we found that in the presence of Mn 2ϩ , the stoichiometry of phosphorylation was maximally 0.5 mol of phosphate/mol of p110␤ (Fig. 3C), arguing against a high basal autophosphorylation as the reason for the missing regulation of PI3K␤ protein kinase activity in vitro. Moreover, no differences in the time course of p110␤ autophosphorylation occurred, regardless of whether G␤␥ or tyrosine-phosphorylated peptide were present. It is interesting that autophosphorylation peaked after more than 30 min under in vitro conditions. Taken together, the presented data do not exclude the possibility that p110␤ autophosphorylation may be involved in receptor-independent regulation of PI3K␤ enzymatic activity.
Autophosphorylation of p110␥ Is Stimulated by G␤␥-Major characteristics of PI3K␤ autophosphorylation such as Mn 2ϩ dependence and inhibition of lipid kinase activity resemble those of class I A PI3K␣ and -␦ autophosphorylation. However, in contrast to class I A kinases, a significant autophosphorylation of p110␥ occurs in the presence of Mg 2ϩ . Furthermore, autophosphorylation does not change the lipid kinase activity of PI3K␥ (11,28). These findings suggest a role for PI3K␥ autophosphorylation different from the class I A PI3K isoforms. Our observation that G␤␥ stimulates p110␥ autophosphorylation further supports this assumption (11,12). Since others have reported an inhibitory effect of G␤␥ on PI3K␥ protein kinase activity (29), we reexamined G␤␥-induced p110␥ autophosphorylation using recombinant purified protein (Fig. 4). In the absence of lipid vesicles, G␤␥ did not increase autophosphorylation of p110␥. In contrast, the addition of lipid vesicles led to a significant G␤␥-dependent stimulation of p110␥ autophosphorylation regardless of whether the lipid vesicles contained PtdIns-4,5-P 2 (see Fig. 4A). These data may indicate that the orientation of proteins on the lipid bilayer surface facilitates the interaction of G␤␥ with PI3K␥. Interestingly, we observed phosphorylation of p101, which could not be stimulated by G␤␥ even in the presence of lipid vesicles. Since in these experiments, a GST-p101/p110␥ heterodimer was analyzed (see Fig. 4B), we also used a His-p110␥/p101 heterodimer (Fig. 5A) in order to exclude the possibility that a bulky GST tag may influence the phosphorylation of PI3K␥ subunits. As indicated in the upper panel of Fig. 5B, phosphorylation of a p101 variant without a GST tag was not visible. Moreover, both G␤␥-stimulated p110␥ autophosphorylation of the His-p110␥/ p101 heterodimer (EC 50 ϭ 30 nM) and lipid kinase activity (EC 50 ϭ 10 nM) were comparable with the data obtained with the GST-p101/p110␥ heterodimer (see Fig. 5B) (11,12). Taken together, these results clearly demonstrate that, in contrast to class I A PI3K␤, autophosphorylation of PI3K␥ is sensitive to G␤␥. Whereas under basal conditions autophosphorylation of p110␥ increased linearly for a period of more than 90 min, the presence of G␤␥ significantly accelerated this phosphorylation (Fig. 5C). In particular, autophosphorylation of p110␥ reached its maximum after 20 min with a stoichiometry of about 0.95 mol of incorporated phosphate/mol of p110␥. The latter observation suggests the presence of one autophosphorylation site in p110␥.
Autophosphorylation of p110␥ Does Not Inhibit PI3K␥ Lipid Kinase Activity-In order to examine the effect of p110␥ autophosphorylation on PI3K␥ lipid kinase activity, p110␥ was phosphorylated in the presence of G␤␥. The catalytic activity of this autophosphorylated PI3K␥ was compared with the nonphosphorylated counterpart using PtdIns-4,5-P 2 as the substrate (Fig. 6). No differences in the production of PtdIns- 3,4,5-P 3 were detected, regardless of whether PI3K␥ was autophosphorylated. Therefore, in contrast to the C-terminal autophosphorylation of p110␤ and p110␦, which both downregulate lipid kinase activity of PI3K, autophosphorylation of p110␥ has no obvious inhibitory effects on PI3K␥ enzymatic activity. Hence, we assumed that autophosphorylation of p110␥ occurs at a site different from a serine residue at its C terminus.
Identification of the p110␥ Autophosphorylation Site-Phosphoamino acid analysis revealed that p110␥ autophosphorylates serine but not threonine or tyrosine residues (Fig. 7A). In order to identify the phosphorylated serine residue, in vitro [ 32 P]phosphate-labeled p110␥ protein was cleaved using different proteases. The resulting peptides were separated by reversed-phase HPLC, and the radioactivity of each fraction was determined. After digestion with chymotrypsin, a phosphopeptide corresponding to the C terminus of p110␥ was identified by MALDI-MS (m/z 1134.56) and nanoESI-MS (doubly charged ion with m/z 567.77), as shown in Fig. 7B. The site of modification within the C-terminal sequence 1093-1102 was determined by nanoESI-MS/MS (see Fig. 7B, lower panel). In par-ticular, the C-terminal yЉ fragment ion series and the loss of neutral H 3 PO 4 (98 mass units) confirmed the sequence and identified phosphorylation of serine 1101. Thus, the MS data demonstrate that serine 1101 of p110␥ is the site of autophosphorylation. To confirm this finding, a p110␥ mutant in which serine 1101 was changed to alanine was examined for its autophosphorylation. As indicated in Fig. 7C, no significant G␤␥stimulated phosphate incorporation into the p110␥S1101A mutant took place. Hence, both class I B p110␥ as well as class I A p110␤ and p110␦ isoforms autophosphorylate on a serine residue at the extreme C terminus.
Lipid Kinase Activity of Mutant PI3K␥-Next we examined the in vitro lipid kinase activity of heterodimeric PI3K␥ variants containing either wild-type p110␥ or a mutant p110␥ in which serine 1101 was replaced by either alanine (see above) or the negatively charged aspartic and glutamic acid (Fig. 8A). No differences in the production of PtdIns-3,4,5-P 3 by these PI3K␥ variants were observed under both basal conditions and followed by stimulation with G␤␥. These data underline that autophosphorylation of p110␥ does not inhibit PI3K␥ lipid kinase activity. Moreover, HEK293 cells were transiently transfected with wild-type or mutant PI3K␥ as well as with the G protein-coupled fMLP receptor, and Akt phosphorylation was subsequently determined. In the absence of PI3K␥, no fMLPinduced phosphorylation of Akt was observed (data not shown), whereas in the presence of any PI3K␥ variants (i.e. the wildtype enzyme or the alanine, aspartate, or glutamate mutants), Akt phosphorylation was significantly stimulated by fMLP to a comparable extent (Fig. 8B). Hence, results obtained both in an in vitro assay using recombinant proteins and in a cell-based assay suggest that autophosphorylation of p110␥ does not influence PI3K␥ lipid kinase activity and thus clearly differs from the autophosphorylation of class I A PI3K isoforms.
Mechanism of p110␥ Autophosphorylation-In order to examine the mechanism of p110␥ autophosphorylation, we used a peptide corresponding to the C terminus of p110␥ (WFLHLV-LGIKQGEKHSA). However, this C-terminal peptide was not a substrate for PI3K␥ protein kinase activity, since a phosphorylation of this peptide by wild-type PI3K␥ was not detected using mass spectrometry (data not shown). Furthermore, the peptide neither influenced p110␥ autophosphorylation under basal conditions nor in the presence of G␤␥ (Fig. 9A). Moreover, a kinase-defective p110␥K833R mutant did not autophosphorylate, emphasizing that phosphorylation of purified PI3K␥ preparations was not due to the presence of a contaminant kinase activity copurifying with the lipid kinase (Fig. 9B). Last, co-incubation of p110␥K833R with enzymatically active wildtype heterodimeric PI3K␥ did not result in phosphorylation of the mutant p110␥K833R, whereas the wild-type enzyme autophosphorylated. Hence, a transphosphorylation mechanism can be excluded. DISCUSSION The present study describes the autophosphorylation sites of the G protein-sensitive class I PI3K␤ and -␥ isoforms. The experimental approaches include mass spectrometric analysis of the posttranslationally modified proteins and site-directed mutagenesis, which are independent and complementary methods. With these strategies, we identified the C-terminal residues serine 1070 and serine 1101 of the catalytic p110␤ and p110␥ subunits, respectively, as the modified amino acids. Previously, the C-terminal serine 1039 was detected as the site of p110␦ autophosphorylation (15). Hence, class I PI3K␤, -␥, and -␦ isoforms share the extreme C terminus of the catalytic subunit as a common site of autophosphorylation, whereas p110␣ is not significantly modified, probably due to the lack of a serine residue in this region.  30. The yЉ ions in the spectrum that contain the phosphoserine are produced by consecutive fragmentation reactions breaking the amino bond and losing the H 3 PO 4 , or vice versa. C, p110␥ with a serine 1101 to alanine mutation does not show any G␤␥-stimulated autophosphorylation. Heterodimeric purified PI3K␥ (His-p110␥/p101) either containing wild-type p110␥ (WT) or a p110␥ mutant (S1101A) was subjected to a protein kinase assay in the absence (Ϫ) and presence (ϩ) of 120 nM purified G␤ 1 ␥ 2-His . One typical autoradiograph and the corresponding Coomassie-stained gel as loading control are shown.
Recently, Williams and co-workers published the crystal structure of p110␥ for a fragment comprising amino acid residues 144 -1102 (31). Although serine 1101 was not resolved in this structure it is clear from the data that it should be just beyond the C-terminal helix k␣ 12 , which lines the PtdIns-4,5-P 2 binding pocket. Therefore, from a sterical point of view, a preferential phosphorylation of this serine appears reasonable. Moreover, lipid and protein kinase activities may compete with each other (32,33). This assumption is supported by recent data from Yart et al. (27) demonstrating that a "protein kinaseonly" (PKO) mutant of p110␤ exhibited an even higher protein kinase activity than the wild-type enzyme. Our own studies did not indicate any difference in the autophosphorylation activity of p110␥, regardless of whether lipid substrates such as PtdIns-4,5-P 2 were present (data not shown). Other data are more complex. Bondeva et al. (22) reported that only those PKO p110␥ variants showed an increased autophosphorylation activity that contained a class II or class III donor activation loop but not the class IV counterpart. In contrast, wild-type p110␣ and all PKO p110␣ variants phosphorylated p85 equally well, whereas PKO p110␣ variants containing a class II or class III donor activation loop exhibited autophosphorylation of the p110␣ subunit (20). The latter finding was unexpected, since p110␣ lacks a C-terminal serine. In this context, our observation of a residual phosphorylation of the p110␤S1070A mutant may be of interest (see Fig. 2C). The fact that purified kinasedefective mutants of p110␤ were not detectably phosphorylated by a contaminating kinase activity rather indicates the presence of a second, quantitatively less important phosphorylation site in p110␤, presumably at threonine 1063 (see Fig. 1A). Support for this assumption comes from MALDI-post-source decay data, which revealed a double phosphorylated C-terminal peptide; unfortunately, the signal was too weak for sequencing by nanoESI/ MS-MS. A corresponding C-terminal threonine of p110␣ (see Fig.  1A) may be a candidate target for significant autophosphorylation by class II and III PKO p110␣ variants (20).
In contrast to the predominant autophosphorylation of p110␤ observed in this and previous studies (27), others have reported p85 phosphorylation as the major target of PI3K␤ autophosphorylation (25,26). In order to explain the apparent discrepancies, one must consider the experimental conditions used in these studies. Roche et al. (25) added purified p85 to immunoprecipitated p110␤ and detected a phosphorylated p85 band, whereas the p110␤ band was not shown. More important, PI3K␤ autophosphorylation was examined with enzyme preparations immobilized to beads in those studies (25,26). Interestingly, this may affect autophosphorylation, since we noticed an increased p85 phosphorylation when we examined PI3K␤ bound to Ni 2ϩ -nitrilotriacetic acid beads. 2 Hence, experimental conditions significantly influence in vitro autophosphorylation of PI3K␤. Likewise, in addition to p110␥ autophosphorylation (11,12,28), p101 phosphorylation by PI3K␥ has been reported (29). In fact, using a GST-p101 construct, we also observed in initial experiments a phosphorylation of p101, which, in con-2 C. Czupalla and B. Nü rnberg, unpublished observations. FIG. 8. Lipid kinase activity of mutant p110␥. A, in vitro lipid kinase activity of mutant p110␥. Equal amounts of heterodimeric purified PI3K␥ (His-p110␥/p101) containing either wild-type p110␥ (WT) or a p110␥ mutant (S1101D, S1101E, or S1101A) were tested for their enzymatic activity in a lipid kinase assay in the absence (Ϫ) and presence (ϩ) of 120 nM G␤ 1 ␥ 2-His . Basal PtdIns-3,4,5-P 3 formation of all PI3K␥ variants was similar (0.25 mol/min/mol enzyme). Shown are mean values Ϯ S.D. of three independent experiments. B, stimulation of Akt phosphorylation by mutant p110␥. HEK293 cells were transiently transfected with plasmids for the fMLP receptor, p101, and wild-type (WT) or mutant (S1101D, S1101E, or S1101A) p110␥. Serum-starved cells were treated with either vehicle (Ϫ) or 1 M fMLP. Equal amounts of whole cell lysates were subjected to SDS-PAGE followed by immunoblotting with anti-phospho-Akt, anti-p110␥, and anti-Erk antibodies as loading control.
FIG. 9. Mechanism of p110␥ autophosphorylation. A, protein kinase activity of purified PI3K␥ in the presence of increasing amounts of a synthetic peptide derived from the C terminus of p110␥. Autophosphorylation of p110␥ was monitored in the absence or presence of 120 nM purified G␤ 1 ␥ 2-His . Shown is one representative autoradiograph out of three independent experiments. B, the autophosphorylation of p110␥ is not mediated by transphosphorylation. Recombinant purified His-p110␥/p101 and GST-p110␥K833R were obtained from Sf9 cells, and proteins were subjected to SDS-PAGE and analyzed by Coomassie staining (left panel). The His-p110␥/p101 complex was incubated alone or in the presence of kinase-inactive GST-p110␥K833R with [␥-32 P]ATP and 120 nM G␤ 1 ␥ 2-His . Phosphorylated proteins were separated by SDS-PAGE and visualized by autoradiography. Autophosphorylated GST-p110␥ served as a control to indicate the size of a phosphorylated GST-p110␥K833R. One representative autoradiograph is shown (right panel).
trast to p110␥ autophosphorylation, was not stimulated by G␤␥. However, p101 phosphorylation was not detectable when a purified hexahistidine-tagged PI3K␥ heterodimer was used. Therefore, we cannot exclude the possibility that the artificial bulky GST tag may have facilitated phosphorylation of p101.
The data presented in this study as well as in other reports provide evidence that the C terminus of p110 is a common site of autophosphorylation for three out of four class I PI3K isoforms. Despite this conformity, PI3K␤ and -␥ differ in all other biochemical characteristics of autophosphorylation. For instance, PI3K␤, like the two other class I A isoforms, PI3K␣ and -␦, autophosphorylates preferentially in the presence of Mn 2ϩ (see Fig. 1B and Refs. 14 -16). In contrast, PI3K␥ exhibits a significantly stimulated autophosphorylation activity in the presence of Mg 2ϩ as shown previously (11,28) and in this study. Interestingly, in vitro most serine/threonine kinases are Mg 2ϩ -dependent, whereas many tyrosine kinases show a greater activity in the presence of Mn 2ϩ (34). Conversely, for phosphorylase kinase, a metal ion-dependent dual kinase specificity was reported (35). The presence of Mg 2ϩ causes serine phosphorylation of phosphorylase b, and Mn 2ϩ activates tyrosine phosphorylation of angiotensin II (35). The basis for these properties are still unclear. In this context, it should be remembered that autophosphorylation, per se, is not a good indicator of protein kinase activity, since many ATP-binding proteins that are not protein kinases are known to autophosphorylate in vitro (32).
We also addressed the control of autophosphorylation by upstream regulators. Under basal conditions, PI3K␥ slowly autophosphorylated (i.e. 0.1 mol of phosphate was incorporated into 1 mol of p110␥ within 30 min). The addition of G␤␥ accelerated phosphate incorporation by 8 -10-fold, resulting in an almost stoichiometric phosphorylation (see Fig. 5, B and C). Interestingly, this effect was only seen in the presence of lipid vesicles, which is in contrast to results obtained by Bondev and co-workers (29). Since the EC 50 values for the stimulation of lipid and protein kinase activities of PI3K␥ were concordant, we hypothesized that the molecular mechanisms of stimulation of these kinase activities are similar. In contrast, neither G␤␥ nor phosphotyrosyl peptide stimulated autophosphorylation of PI3K␤ (see Fig. 3) even under low basal autophosphorylation conditions (i.e. in the presence of Mg 2ϩ ). Unfortunately, in vitro data addressing a possible stimulation of autophosphorylation of PI3K␣ and -␦ by upstream regulators are missing. However, Vanhaesebroeck and co-workers (15) reported a CD28-mediated stimulation of C-terminal p110␦ phosphorylation under in vivo conditions. Data obtained from aspartate and glutamate mutants of p110␤ suggest that a phosphorylated PI3K␤ displays a hampered lipid kinase activity (see Fig. 2D). Unfortunately, a more direct experimental approach (e.g. assaying the effect of autophosphorylation on lipid kinase activity) was inconclusive. In particular, we were unable to completely remove Mn 2ϩ , which is necessary for autophosphorylation but disturbed PI3K␤ lipid kinase activity, without the use of immobilizing agents before carrying out the lipid kinase assay. Nevertheless, our results with the p110␤1070D/E mutants are consistent with data obtained from the other class I A kinases. Autophosphorylation of PI3K␣ and -␦ and exchange of serine 1039 of p110␦ to aspartate or glutamate inhibited lipid kinase activity (14 -16, 20). Possible explanations for this effect include an induction of structural/conformational changes of the phosphorylated enzyme or an impact on the phospho-transfer reaction or on the ATP/ PtdIns-4,5-P 2 interaction (15). Furthermore, based on the crystal structure of p110␥, Williams and associates (31) have speculated that a phosphorylated C terminus may be a sterical impediment for PtdIns-4,5-P 2 substrate binding. Surprisingly, here we provide experimental evidence that autophosphorylation of the C terminus of p110␥ does not inhibit lipid kinase activity as shown under in vitro conditions with a prephosphorylated wild-type enzyme and p110␥1101D/E mutants (see Figs. 6 and 8A). These mutants showed full activity on cellular effectors in HEK293 cells in vivo (see Fig. 8B). Hence, we assume that autophosphorylation of p110␥, which is primarily regulated by G␤␥, has functions distinct from regulating its lipid kinase activity. One possibility may be the existence of PI3K␥ binding partners that specifically interact with the autophosphorylated form of PI3K␥. In fact, recent evidence suggests that PI3K␥ interacts not only with its principal regulators (i.e. G␤␥ and Ras) but also with additional components of signaling cascades such as the ␤-adrenergic receptor kinase 1 (36).
We found that PI3K␤ and -␥ did not phosphorylate peptides derived from their respective C terminus, and vice versa the peptides did not affect the autophosphorylation capacity of the enzyme (see Figs. 2E and 9A). Similar observations were reported for PI3K␣ and -␦ (15), whereas Beeton et al. (26) described that PI3K␣ and -␤ phosphorylated a p85-derived peptide containing serine 608. Mechanistically, PI3K␥ did not transphosphorylate (see Fig. 9B), which may indicate a high degree of substrate specificity of the protein kinase activity. Notably, auto-but not transphosphorylation has also been reported for the p110␥ monomer and a phosphatidylinositol 4-kinase ␤ (37,38). Surprisingly, while we were searching for in vivo substrates of PI3K protein kinase activity, we failed to detect p110␥ autophosphorylation in HL-60 and Sf9 cells so far, which emphasizes the need for further investigations into the regulation, activity, and targets of PI3Ks in vivo.