Human phosphoinositide 3-kinase C2beta, the role of calcium and the C2 domain in enzyme activity.

The cDNA for a human Class II phosphoinositide 3-kinase (PI 3-kinase C2beta) with a C2 domain was cloned from a U937 monocyte cDNA library and the enzyme expressed in mammalian and insect cells. Like other Class II PI 3-kinases in vitro, PI 3-kinase C2beta utilizes phosphatidylinositol (PI) and PI 4-monophosphate but not PI 4, 5-biphosphate as substrates in the presence of Mg2+. Remarkably, and unlike other PI 3-kinases, the enzyme can use either Mg-ATP or Ca-ATP to generate PI 3-monophosphate. PI 3-kinase C2beta, like the Class I PI 3-kinases, but unlike PI 3-kinase C2alpha, is sensitive to low nanomolar levels of the inhibitor wortmannin. The enzyme is not regulated by the small GTP-binding protein Ras. The C2 domain of the enzyme bound anionic phospholipids such as PI and phosphatidylserine in vitro, but did not co-operatively bind Ca2+ and phospholipids. Deletion of the C2 domain increased the lipid kinase activity suggesting that it functions as a negative regulator of the catalytic domain. Although presently it is not known whether PI 3-kinase C2beta is regulated by Ca2+ in vivo, our results suggest a novel role for Ca2+ ions in phosphate transfer reactions.

The cell, through diverse surface receptors with unique binding specificities, can sense bound signal molecules and transduce responses that regulate its physiology. It seems clear that most cell surface receptors activate a phosphoinositide 3-kinase (PI 3-kinase) 1 as part of the signal transduction cascade lead-ing to the formation of phosphoinositides with 3Ј-phosphate groups (1,2). The diversity of physiological events associated with increased PI 3-kinase activity is evident from reports of enzyme activation in: processes such as cell proliferation and transformation (3)(4)(5), events linked to insulin action, including alterations in glucose transport (6), the effects of growth factors on cell shape and motility (7), T cell signaling (8,9) and apoptosis (10,11). The activation of PI 3-kinase in this array of receptor-triggered processes suggests that 3Ј-phosphoinositides have a role as second messengers. At least three 3Ј-phosphoinositides are produced in cells: PI(3)P, PI(3,4)P 2 , and PI(3,4,5)P 3 . Receptor-triggered signals have been shown to activate PI 3-kinases and generate PI(3,4)P 2 and PI(3,4,5)P 3 (12,13). PI(3)P can also be detected in cells but its production is not regulated by external signals (12,13). PI(3)P and PI(3,4)P 2 can also be generated from PI(3,4)P 2 and PI(3,4,5)P 3 through the action of phosphoinositide phosphatases which could be regulated by distinct mechanisms to those of the phosphoinositide kinases (14). Through many studies involving the purification and molecular characterization of PI 3-kinases, a family of enzymes has been defined, which can be divided into three classes, whose members have diverse substrate specificity and distinct control mechanisms (15,16).
The Class I PI 3-kinases, which can be subdivided into IA and IB, are known to be activated by receptors. Although they can phosphorylate PI, PI(4)P, and PI(4,5)P 2 in vitro, these enzymes utilize mainly PI(4,5)P 2 as a substrate in vivo (12,13). The Class IA enzymes are heterodimers that are recruited to, and activated by receptors linked to tyrosine kinases through their p85 subunits. At least three distinct p110 subunits that have kinase activity are associated with p85 subunits: these are p110␣, p110␤, and p110␦ (17)(18)(19). The p85 subunits serve as adaptors and regulators (20 -22). The catalytic subunits can be directly regulated by Ras (23). The Class IB PI 3-kinase, p110␥ (24), is associated with a p101 adaptor, which may link the kinase to serpentine receptors by the ␤␥ subunits of heterotrimeric G-proteins and thus mediate p110␥ activation (25).
The Class II PI 3-kinases have a carboxyl-terminal C2 domain. The nature of any receptor-linked activation pathway for these enzymes remains unclear. Studies of the Drosophila (26,27), murine (27,28), and human enzymes (29) show that their in vitro substrate specificity is restricted to PI and PI(4)P, and that they cannot utilize PI(4,5)P 2 . The role of the C2 domain is not understood, although studies of the Drosophila enzyme suggest that it mediates calcium-independent phospholipid binding (26). This feature is similar to that of the C2B domain of synaptotagmin, which binds to the clathrin-AP2 complex in a Ca 2ϩ -independent manner (30). The synaptotagmin C2B domain can also homodimerize in a Ca 2ϩ -dependent manner (31). The diversity in biochemical function which can be mediated by C2 domains is evident from studies which show that C2 domains of synaptotagmin, protein kinase C, and phospholipase C can bind a variety of ligands, including Ca 2ϩ , phopholipids, inositol polyphosphates, and intracellular proteins (32,33).
The third class of PI 3-kinases contains phosphatidylinositol 3-kinases that are specific for PI. Only one enzyme has yet been found in each of several species examined. In yeast, the enzyme is the product of the vesicle protein sorting mutant gene Vps34 (34). Both Vps34p and its human homologue PtdIns 3-kinase (35) associate with a serine/threonine kinase, Vps15p in yeast (36) and p150 in man (37). The biochemical function of the PI(3)P generated by these enzymes has been elusive, although recent reports of alterations in PI(3,5)P 2 levels in yeast and mammalian cells after osmotic shock suggest that PI(3)P may serve as a substrate for an as yet unknown PI(3)P 5-kinase (38) and that PI(3,5)P 2 may mediate vesicular trafficking.
Several other proteins (Tor1 and 2, RAFT, ATM, and DNAdependent protein kinase), which have regions homologous to the PI 3-kinase HRI kinase domain, have been described (39). These proteins seem to be involved in cell cycle regulation and, although in some cases they have protein kinase activity, it remains unclear if they function as PI 3-kinases.
In this article, we describe the cloning, expression, and enzymology of a human Class II PI 3-kinase with an amino acid sequence that is virtually identical to that of the PI 3-kinase HsC2, which was recently described by Brown et al. (40). Here we show that PI 3-kinase C2␤ has a substrate specificity restricted to PI and PI(4)P, similar to that of the Drosophila, human, and murine Class II enzymes previously examined. The cofactor studies described here, however, show that the enzyme is unique in being able to use Ca-ATP for its lipid kinase activity in vitro. The role of the C2 domain in calcium binding, enzyme activity, and membrane association has been investigated, and the results show that the C2 domain may function mainly as a modulator of catalytic function.
Isolation and Characterization of cDNA Clones Encoding PI 3-Kinase C2␤-Reverse transcriptase-PCR was performed on a U937 cDNA library, using degenerate primers corresponding to the conserved peptides GDDLRQD and FHIDFG, and screening was performed as described previously in Refs. 35 and 41. Sequencing of the cDNA clones was carried out using the Taq DyeDeoxy terminator Cycle Sequencing system (ABI) and an automated DNA sequencer (ABI 373).
Rapid Amplification of cDNA Ends PCR and Assembling of the Full-length cDNA-Rapid amplifiction of cDNA ends PCR was performed using U937 -ligated cDNA as a template and nested sense primers and PI 3-kinase C2␤ specific antisense primers. Taq DNA polymerase (Cetus) was used in two consecutive 50-l PCR reactions (first PCR: ATTAACCCTCACTAAAGGG (T3 promoter) sense primer; ACTGAATTCTCAACCCACGTCCACATTCCTCAGG (ϩ1170), antisense primer, 94°C for 2 min denaturation followed by 30 cycles of 95°C for 40 s, 56°C for 15 s, 72°C for 2 min; second PCR: TGCAG-GAATTCGGCACGA (cloning site) sense primer, ACTGAATTCTCAAG-TAGTCTTGGATGTCAGAGC (ϩ971) nested antisense primer, 96°C for 1 min followed by 25 cycles of 96°C for 40 s, 56°C for 15 s, 72°C for 2 min). Approximately 500 ng of cDNA ligation mixture was used in the first PCR, then 1 l from the first PCR reaction was amplified in the second reaction. The PCR reaction products were digested with EcoRI and ligated into pBluescript (Stratagene). Fifty clones were sequenced and oligonucleotides were designed in order to introduce an EcoRI site 5Ј to the most NH 2 -terminal ATG codon. Reverse transcriptase-PCR using Vent DNA polymerase (New England Biolabs) was performed using U937 cDNA as a template. The resulting 5Ј cDNA was digested with EcoRI and EcoRV, sequenced to check sequence fidelity and ligated in front of the remainder of the cDNA to produce the complete open reading frame.
Construction of Expression Vectors-The cDNA encoding PI 3-kinase C2␤ was subcloned into pcDNA3 (Invitrogen) using the EcoRI and XhoI sites. NH 2 -terminal Glu-(MEFMPME) or Myc-(MEQKLISEEDL) epitope tags were introduced into the cDNA in-frame by PCR using Vent DNA polymerase. An EcoRI site was added to the 5Ј of the tag sequences to facilitate subcloning. The PCR products encoding the tagged NH 2 termini were fused to the cDNA of PI 3-kinase C2␤ using a unique BstUI site at ϩ169 and recloned into pcDNA3 using EcoRI and XhoI. The sequence of these constructs was confirmed by NH 2 -terminal sequencing.
Construction of Mutants of PI 3-Kinase C2␤-In order to generate a deletion of the C2 domain of PI 3-kinase C2␤, the cDNA encoding the NH 2 -terminal Myc-tagged version of the enzyme cloned in pcDNA3 was digested at position ϩ4419 with ApaI. After dephosphorylization of the linearized DNA with calf intestinal phosphatase (Boehringer Mannheim), annealed 5Ј-phosphorylated oligonucleotides (sense TGAAG-TACTCAATTGGGCC, antisense CAATTGAGTACTTCAGGCC) were ligated to the cDNA to introduce a TGA in-frame at codon ϩ1474. The construct was analyzed by restriction mapping and sequencing of the 3Ј end.
Transient Expression in Mammalian Cells-HEK293 cells were grown to 50 -60% confluence on 150-mm dishes and transfected with cDNA constructs in pcDNA3 using calcium-phosphate, exactly as described (42). The cells were harvested 48 h after transfection and analyzed for gene expression.
Expression in Insect Cells-The cDNA construct encoding Glu-tagged PI 3-kinase C2␤ was subcloned into pBluescript SK using EcoRI and XhoI sites and this construct digested with EcoRI and KpnI. The resulting cDNA insert was then subcloned into the pAcSG2 baculovirus transfer vector (PharMingen). Sf9 insect cells were co-transfected with the recombinant transfer vector encoding PI 3-kinase C2␤ and Bacul-oGold linearized baculovirus DNA (PharMingen), using Lipofectin (Life Technologies, Inc.). Recombinant baculoviruses were harvested 7 days post-transfection, amplified, and screened for induction of protein expression. Positive viruses were plaque-purified, re-amplified, and used for production of recombinant enzyme. Sf9 cells were grown to 50 -60% confluency and infected with recombinant plaque-purified baculoviruses for 60 h. The cells were then harvested by centrifugation and washed once in ice-cold PBS (Life Technologies, Inc.). Recombinant PI 3-kinase C2␤ was purified from the Triton X-100 soluble fraction by immunoprecipitation with purified monoclonal anti-Glu-tag antibodies and Protein G-Sepharose, as described below.
Immunoprecipitations-Cells grown on 150-mm dishes were placed on ice, washed once in ice-cold PBS (Life Technologies, Inc.), and lysed for 20 min on ice in 2 ml of lysis buffer (20 mM HEPES-NaOH (pH 7.4), 150 mM NaCl, 1% (w/v) Triton X-100, 2 mM EDTA, 10 mM sodium fluoride, 10 mM Na 2 HPO 4 , 10% (w/v) glycerol, 1 mM phenylmethylsulfonyl fluoride, 5 mM benzamidine, 7 mM diisopropyl fluorophosphate, 1 mM N ␣ -tosyl-L-lysine chloromethyl ketone, 20 M leupeptin, 18 M pepstatin, 21 g/ml aprotinin, 2 mM dithiothreitol). The cells were then scraped from the dishes, transferred into 1.5-ml microcentrifuge tubes, and centrifuged for 20 min at 15,000 ϫ g at 4°C. The supernatant was collected and incubated for 2 h at 4°C with constant rotation with the relevant antibody. Protein A-or Protein G-Sepharose CL-4B (Pharmacia, 10 l of beads per sample) were then added and the incubation continued for 1 h at 4°C with constant rotation. The immunoprecipitates were washed once in lysis buffer, once in 50 mM Tris-HCl (pH 7.4), 0.5 M LiCl and once in TBS (50 mM Tris-HCl (pH 7.4), 150 mM NaCl). The enzyme preparations were stored at Ϫ30°C in TBS containing 50% glycerol and 1 mM dithiothreitol. In the case of the C2 deletion mutant, the washing and storage buffers were supplemented with 100 M diisopropyl fluorophosphate, since the enzyme lost activity much more rapidly than the wild-type, possibly because of an increased susceptibility to proteolysis.
The membranes were then blocked for 1 h in PBS containing 3% (w/v) non-fat dry milk, 0.1% (w/v) PEG 20000. The relevant primary antibodies were diluted in PBS, 0.05% (w/v) Tween 20 (PBS/Tween) and incubated with the membranes for 2 h. After extensive washing in PBS/ Tween, the blots were incubated for 1 h with goat anti-mouse or antirabbit antibodies coupled to horseradish peroxidase (Dako) at 1:2000 dilution. The membranes were then washed in PBS/Tween and the bands detected using ECL (Amersham).
Generation of Antisera-A cDNA fragment encoding amino acids 1 to 331 obtained by PCR was subcloned into pGEX-2T (Pharmacia) using the EcoRI 5Ј of the start codon and an EcoRI site introduced by PCR at the 3Ј end. Expression of this construct in Escherichia coli strain BL21/ DE3 was induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside. The bacteria were then harvested, snap-frozen, and lysed by sonication in ice-cold extraction buffer (EB, 10 mM TrisHCl (pH 7.4), 150 mM NaCl, 1% (w/v) Triton X-100, 5 mM EDTA, 5 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin). After centrifugation for 20 min at 15,000 ϫ g at 4°C, the cleared lysate was incubated with glutathione-Sepharose CL-4B (Pharmacia) for 2 h at 4°C with constant rotation. The beads were then washed three times in EB, once in TBS, and resuspended in 50 mM Tris-HCl (pH 7.4), 2.5 mM CaCl 2 . Thrombin was added (4 g/mg of recombinant protein) for 20 min at room temperature on a wheel. The supernatant containing the recombinant NH 2 -terminal fragment was treated with p-aminobenzamidine-agarose (Sigma), concentrated to 1 mg/ml using a Centriplus 10 concentrator (Amicon), and stored at Ϫ70°C. The purified recombinant NH 2 -terminal fragment was used to immunize 2 rabbits (Eurogentec, Seraing-Belgium). The animals were injected 3 times over 35 days with 100 g of antigen in Freund's adjuvant and the serum was collected after 42 days. The rabbits were then boosted every month and serum was collected 10 days after each injection.
Expression of the C2 Domain-A cDNA fragment starting at codon 1440 and encoding the C2 domain was amplified by PCR and cloned into pGEX-2T using two EcoRI sites introduced at both the 5Ј and 3Ј ends during PCR. The orientation of the construct was checked by restriction digestions and sequencing. E. coli strain BL21/DE3 was transformed with the construct and protein expression induced by adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside. The recombinant protein was purified from the Triton X-100 soluble fraction as above, and eluted from the glutathione-Sepharose CL-4B as a GST-fusion using 100 mM Tris-HCl (pH 7.4), 150 mM NaCl, 20 mM reduced glutathione. Buffer exchange to TBS was performed on a PD-10 column (Pharmacia). The purified GST-C2 domain was concentrated to 1 mg/ml using a Centriplus 10 concentrator (Amicon) and stored at Ϫ70°C. The same protocol was used to prepare recombinant synaptotagmin C2A and C2B domains as GST fusions.
Binding of Phospholipids and 45 Ca to GST-C2 Domains-The binding of purified recombinant GST-C2 domains to phospholipid vesicles was studied using a sedimentation assay, essentially as described in Ref. 44 . Phospholipids in CHCl 3 (Sigma, 160 g/assay) were dried, sonicated in 50 l of 50 mM HEPES (pH 7.2), 100 mM NaCl and centrifuged for 10 min at 12,000 ϫ g at room temperature. The pellets were resuspended in 50 l of 50 mM HEPES (pH 7.2) and added to 50 l of TBS containing 0.1 mg/ml purified recombinant GST-C2 domains and EGTA (2 mM final concentration) or CaCl 2 (1 mM final concentration). After incubation for 15 min at room temperature, the samples were centrifuged for 10 min at 12,000 ϫ g. The supernatant was precipitated using 10% trichloroacetic acid and the protein pellets washed 3 times with acetone at Ϫ20°C and boiled in SDS-PAGE sample buffer. The pellets were extracted with acetone for 30 min at Ϫ20°C, centrifuged for 10 min at 12,000 ϫ g at room temperature, and boiled in SDS-PAGE sample buffer. Equivalent amounts of the pellet and the supernatant were analyzed by SDS-PAGE and Coomassie Blue staining. The gels were scanned using a flatbed densitometer (Epson) and the data analyzed using the program Aida 1.20 beta.
To analyze binding of 45 Ca 2ϩ to the recombinant GST-C2 domains, phospholipids and proteins were prepared as above and 10 Ci of 45 CaCl 2 (Amersham, 2.2 mCi/ml) added to each sample in the presence of 1 mM unlabeled CaCl 2 . After incubation for 10 min at room temperature, 10 l of glutathione-Sepharose were added and the samples incubated with constant rotation for 30 min. The beads were then centrifuged for 20 s at 1000 ϫ g and washed twice in 50 mM HEPES (pH 7.2), 100 mM NaCl. The radioactivity bound to the beads was quantified by scintillation counting.
PI 3-Kinase Assays-PI 3-kinase activity of the immunoprecipitates was assayed by resuspending them in 25 l of 2 ϫ kinase buffer (40 mM Tris-HCl (pH 7.4), 200 mM NaCl, 2 mM dithiothreitol). Phospholipids (PI, PI(4)P, and PI(4,5)P 2 , Sigma) stored in CHCl 3 were dried, sonicated for 15 min in 50 mM Tris-HCl (pH 7.4), and added at 0.2 mg/ml final concentration to the samples. The reactions (50 l final volume) were started by the addition of 40 M ATP and 10 Ci of [␥-32 P]ATP (3000 Ci/mmol, Amersham) and 3.5 mM of the relevant divalent cations. In order to determine the kinetic parameters of the enzyme for ATP salts, the immunoprecipitates and the phospholipids were incubated in a 40-l reaction volume, and various concentrations of Mg-ATP, Ca-ATP, or Na 2 -ATP (all from Sigma) containing [␥-32 P]ATP (0.2 Ci/nmol) added in 10 l. After incubation for 15 min at 37°C, 100 l of 1 N HCl and 200 l of CHCl 3 /MeOH (1:1) (v/v) were added. The organic phase was collected, re-extracted with 40 l of MeOH, 1 N HCl (1:1) (v/v) and radioactivity measured by Cerenkov counting. The samples were then dried, resuspended in 30 l of CHCl 3 / MeOH (1:1) (v/v) and spotted onto channelled Silica Gel 60 TLC plates (Whatman), which had been pretreated in 1% (w/v) oxalic acid, 1 mM EDTA, 40% MeOH (v/v) and baked for 15 min at 110°C. The plates were developed in propanol, 2 M acetic acid (65:35) (v/v) and the radiolabeled spots quantified using a PhosphorImager (Molecular Dynamics).
Phospholipid Deacylation and HPLC Analysis-These analyses were performed as described in Ref. 45. Radioactive spots corresponding to specific phospholipids were scraped from the TLC plates and incubated with methylamine (33% (w/w) in EtOH) for 1 h at 53°C. The samples were dried in a SpeedVac and the silica gel extracted three times with 0.25 ml of H 2 O. After centrifugation, the supernatant was extracted twice with an equal volume of butanol/petroleum ether/ethyl formate (20:4:1, v/v/v). The aqueous phase containing the deacylated phospholipids was then analyzed on a Partisphere SAX HPLC column (Whatman). A gradient of 0 to 25% B over 60 min followed by 25 to 70% B over 30 min (A: H 2 O, B: 1 M (NH 4 ) 2 HPO 4 pH 3.8) was used. Radioactive peaks were monitored using a continuous flow scintillation counter (Reeve Analytical, Glasgow, United Kingdom). The retention time of the samples was compared with that of a glycero-PI(3)P standard produced in an in vitro PI kinase assay using purified recombinant bovine p110␣ and of a glycero-PI(4)P standard produced by an in vitro PI kinase assay using purified membrane preparations of A431 cells in the presence of 0.5% (w/v) Nonidet P-40.
Modeling and Alignment-The alignment of the C2 domain of PI 3-kinase C2␤ with other C2 domains was initially obtained using Multalign (46) and modified within Quanta TM to align it only to synaptotagmin C2A domain, which was used as the template for the PI 3-kinase C2␤ model. Equivalent co-ordinates were transferred from the template structure to the PI 3-kinase C2␤ C2 domain target. To build insertions or deletions a high resolution (Ͻ2.0 Å) fragment data base was searched for segments with the appropriate length, for close sequence similarity and good fit on the PI 3-kinase C2␤ C2 domain loop termini. These regions were minimized locally before the entire model was subjected to Steepest Descent minimization until the energy fluctuation leveled out. The electrostatic potential field was calculated using GRASP (47) with grids contoured at ϩ1.5 and Ϫ1.5 kilotesla.
Subcellular Fractionation-HEK 293 cells were scraped into 50 mM Tris-HCl (pH 7.4), 0.25 M sucrose (supplemented with protease inhibitors) and homogenized by passing through a G-25 needle. The homogenate was centrifuged for 5 min at 2,500 ϫ g and 4°C to remove unbroken cells and nuclei. The plasma membrane fraction was collected by centrifugation at 18,000 ϫ g for 20 min and 4°C. The low density microsomal fraction (pellet) and the cytosol (supernatant) were collected by centrifugation for 2 h at 100,000 ϫ g and 4°C.
Ras Binding Assays-These assays were performed essentially as described in Ref. 23. Recombinant Glu-tagged p110␣ and PI 3-kinase C2␤ were purified from infected Sf9 cells by immunoprecipitation and immobilized on Protein G-Sepharose. These samples were then incubated with purified recombinant H-Ras which had been preloaded with GDP or GTP; the complexes were analyzed by SDS-PAGE and anti-Ras immunoblotting. Aliquots of p110␣ and PI 3-kinase C2␤ were assayed for lipid kinase activity in parallel.

RESULTS
Cloning of PI 3-Kinase C2␤-Degenerate primers with nucleotide sequences based on conserved amino acid sequences (GDDLRQD and FHIDFG) in the kinase domain of yeast Vps34p and bovine p110␣, were used in reverse transcriptase-PCR reactions, with mRNA from the human cell line U937, to obtain a partial cDNA clone, which was then used to screen a -Zap cDNA library from the human cell line U937. Fifteen cDNA clones that contained nucleotide sequences identical to that of the PCR product were isolated. None of these cDNA clones were long enough to encode a complete open reading frame. Therefore rapid amplifiction of cDNA ends-PCR was performed on U937 cDNA and a 0.5-kilobase extension of the cDNA, which contained several potential 5Ј start (ATG) codons, was obtained. The second, but not the first, ATG from the 5Ј end, was preceded by a Kozak consensus sequence which is required for the efficient initiation of translation (48). The 5.4-kilobase composite full-length cDNA included an open reading frame which could encode a protein of 1609 amino acids with a predicted molecular mass of 182 kDa.
The amino acid sequence of the predicted protein showed that it was most closely related (42.1% identity) to that of PI 3-kinase C2␣ (29) and it was therefore referred to as PI 3-kinase C2␤. The 2 proteins were, however, only 20% identical within the amino-terminal 300 residues. Sequence alignment with other Class II PI 3-kinases revealed three regions of homology: HRI and PIK which are found in all 3-kinases (15) and the C2 domain (see Fig. 1A). The HRI catalytic domain had a conserved lysine at position 1058 that is thought to be involved in ATP binding, and included a conserved aspartate at position 1170, asparagine at position 1175, and aspartate-phenylalanine-glycine motif at positions 1188 -1190, which are all thought to be involved in ligand binding (15). The HRI domain and C2 domains show 76.3 and 56.1% identity, respectively, with the corresponding domains of human PI 3-kinase C2␣ (Fig. 1B) (29). Additional features of the sequence were two proline-rich stretches near the NH 2 terminus of the protein, between amino acids 131-137 and 144 -149, which could be SH3-binding motifs.
When this work was complete, the amino acid sequence of PI 3-kinase C2␤ was found to be almost identical to that of HsC2-PI3K which has recently been cloned from a human breast cDNA library using our primers (40). The proteins encoded by the cDNAs of PI 3-kinase C2␤ and HsC2-PI3K differ by 11 residues throughout their open reading frames. Presumably HsC2-PI3K and PI 3-kinase C2␤ are the same protein.
Expression and Biochemical Characterization of PI 3-Kinase C2␤ in Mammalian and Insect Cells-The full-length cDNA encoding PI 3-kinase C2␤ was subcloned into a mammalian expression vector and a Glu-tag epitope was added as an amino-terminal extension to the protein, using PCR methods, in order to facilitate purification of the expressed protein. Transfection of HEK293 cells with this construct induced expression of a 180-kDa protein, which was detected in anti-Glu immunoprecipitates (data not shown). A rabbit polyclonal antiserum against an NH 2 -terminal fragment of PI 3-kinase C2␤ was generated and used to immunoprecipitate endogenous PI 3-kinase C2␤ from HEK293 cells. The native enzyme, which was immunoprecipitated with rabbit polyclonal antiserum, migrated on a 7.5% SDS-PAGE with an apparent molecular weight identical to that of the recombinant protein, which had been immunoprecipitated with anti-Glu-tag antibodies (data not shown). This indicates that the cloned cDNA could encode the full-length protein. The rabbit polyclonal antiserum raised against PI 3-kinase C2␤ did not cross-react with human PI 3-kinase C2␣ immunoprecipitated from a U937 lysate with a specific rabbit antiserum (data not shown), which is consistent with the low sequence identity of the two human Class II PI 3-kinases at their NH 2 terminus (20% in the first 300 residues).
Western blot analysis of human cell lines suggest that PI 3-kinase C2␤ is widely expressed. The protein could be detected in U937, HL-60, W138VA13 fibroblasts, A431 carcinoma cells, and HEK293 cells. The PI 3-kinase C2␤ antiserum also detected a protein of 180 kDa in mouse NIH3T3 fibroblasts and PC-12 cells (data not shown).
The cDNA encoding the Glu-tagged PI 3-kinase C2␤ was subcloned into a baculovirus transfer vector and recombinant baculoviruses were generated and plaque-purified. Recombinant full-length PI 3-kinase C2␤ could be purified from infected Sf9 cells by immunoprecipitation with anti-Glu-tag antibodies (data not shown).
The purified recombinant Glu-tagged PI 3-kinase C2␤ protein expressed in both HEK293 cells and Sf9 cells was used to characterize the in vitro lipid kinase activity of the enzyme. Results obtained using enzyme from both mammalian and insect cells were essentially the same. HPLC analysis of the deacylated PIP generated by the enzyme in vitro showed that the PI substrate was phosphorylated on the 3-position of the inositol ring (Fig. 2). PI 3-kinase C2␤ was able to phosphorylate PI and PI(4)P but not PI(4,5)P 2 in the presence of Mg 2ϩ (Fig. 3A), and the enzyme appeared to have a preference for PI in vitro, as its activity toward PI(4)P was only about 10% of that directed toward PI. There was no detectable PI(4,5)P 2 3-kinase activity when the substrate was presented in vesicles. The addition of phosphatidylserine (PS), phosphatidylcholine (PC), or phosphatidylethanolamine (PE) to the in vitro reaction, either alone or in combination, did not increase the activity of PI 3-kinase C2␤ toward PI(4)P significantly, when compared with PI. A very low PI(4,5)P 2 3-kinase activity could be detected with mixed vesicles containing equimolar amounts of PI, PS, PC, and PE. This represented about 1% of the PI 3-kinase activity under these conditions (data not shown).
Surprisingly, PI 3-kinase C2␤ was able to phosphorylate PI, but not PI(4)P, in the presence of Ca 2ϩ . HPLC analysis confirmed that the PIP produced by the enzyme in the presence of Ca 2ϩ was PI(3)P (data not shown). We investigated this further and compared the divalent cation specificities of PI 3-kinase C2␤ with that of the recombinant Class IA PI 3-kinase, p110␣ in complex with p85␣. The results demonstrated that, while PI 3-kinase C2␤ had a similar PI 3-kinase activity in the presence of Mg 2ϩ , Mn 2ϩ , or Ca 2ϩ , the PI 3-kinase activity of p85␣/p110␣ was much greater in the presence of Mg 2ϩ than Mn 2ϩ , and was undetectable in the presence of Ca 2ϩ (Fig. 3B). These results are in agreement with previous reports (49). PI 3-kinase C2␤ showed similar kinase activities toward PI and PI(4)P in the presence of Mg 2ϩ or Mn 2ϩ , while no kinase activity for either substrate was detected in the presence of ZnCl 2 or CuCl 2 (not shown). The dependence of the PI-kinase activity of PI 3-kinase C2␤ on Mg 2ϩ and Ca 2ϩ concentrations were found to be similar (Fig. 4A), but the enzyme phosphorylated PI(4)P only in the presence of Mg 2ϩ and not Ca 2ϩ (Fig. 4A).
The lipid kinase activity of PI 3-kinase C2␤ was then measured in the presence of the Mg 2ϩ , Ca 2ϩ , and Na ϩ salts of ATP, in the absence of any other source of divalent cations. The enzyme preparation displayed only very low activity in the presence of Na-ATP, which demonstrates that it had not been contaminated by divalent cations during immunoprecipitation (Fig. 4B). We found that the PI kinase activity of PI 3-kinase C2␤ had a similar K m (120 M) and V max (737 nmol/mg/min) for both Mg-ATP and Ca-ATP (Fig. 5B). When the enzyme was incubated in the presence of a fixed concentration of Mg-ATP and increasing concentrations of Ca 2ϩ , the PI(4)P-kinase activity was inhibited in a dose-dependent manner (not shown).
The sensitivity of the lipid kinase activity of PI 3-kinase C2␤ to the PI 3-kinase inhibitor wortmannin (50) was found to be similar to p85␣/p110␣ (IC 50 1.6 nM) (Fig. 5A), however, its sensitivity to the inhibitor LY294002 (51) was about 6-fold higher than that of the Class I enzyme (6.9 M versus 1.2 M) (Fig. 5B).
Characterization of the C2 Domain-In order to study the function of the C2 domain of PI 3-kinase C2␤, we first investigated the phospholipid-binding properties of the isolated recombinant domain. The C2 domain of PI 3-kinase C2␤ was expressed in E. coli as a GST fusion protein and purified to near homogeneity by glutathione-Sepharose affinity chromatography. A sedimentation assay was used to assess the ability of this domain to bind phospholipid vesicles and to compare FIG. 3. Substrate and cation specificity of PI 3-kinase C2␤. A, purified recombinant PI 3-kinase C2␤ was assayed for in vitro kinase activity using PI, PI(4)P, or PI(4,5)P 2 as substrates in the presence of divalent cations Mg 2ϩ or Ca 2ϩ . Purified recombinant p85␣/p110␣ was assayed in the presence of Mg 2ϩ and PI, PI(4)P, or PI(4,5)P 2 . Radioactive phospholipid products were extracted and analyzed by TLC. B, purified recombinant PI 3-kinase C2␤ or p85␣/p110␣ were assayed for in vitro kinase activity using PI as a substrate in the presence of no cation, EDTA, EGTA, EDTA and EGTA, Mg 2ϩ , Mn 2ϩ , or Ca 2ϩ . Radioactive phospholipid products were extracted and analyzed by TLC.

FIG. 4. Dose dependence of the PI 3-kinase C2␤ lipid kinase activity for divalent cations and ATP salts.
A, purified recombinant PI 3-kinase C2␤ was assayed for in vitro kinase activity in the presence of PI (squares) or PI(4)P (circles) and increasing concentrations of Mg 2ϩ (closed symbols) or Ca 2ϩ (open symbols). Radioactive lipid products were extracted, analyzed by TLC, and quantified using a PhosphorImager. The data are mean with S.E. (smaller than symbols where not indicated) from two independent experiments. B, the enzyme was assayed in the presence of increasing concentrations of Na-ATP (triangles), Mg-ATP (closed circles), or Ca-ATP (squares). The products were analyzed and quantified as above. The data are mean with S.E. (smaller than symbols where not indicated) from two independent experiments. these properties with those of the purified recombinant GST-C2A and GST-C2B domains of synaptotagmin. Under these conditions, the recombinant proteins did not sediment significantly in the absence of phospholipids, and GST itself interacted only weakly with the phopholipid vesicles (Fig. 6). The GST-C2 domain of PI 3-kinase C2␤ bound to vesicles of PI or PS with about 4-fold higher affinity than to vesicles of PC (Fig. 6, after subtraction of the background binding to GST). The binding of the PI 3-kinase C2␤ C2 domain to vesicles of PI or PS was not affected significantly by the presence or absence of Ca 2ϩ in the assay. These results were in contrast to those obtained with the GST-C2A domain of synaptotagmin, which bound to vesicles of PI or PS in the presence of Ca 2ϩ but not in the presence of EGTA (Fig. 6), which is in agreement with previous reports (52). The C2 domain of PI 3-kinase C2␤ showed similar phospholipid binding properties to the C2B domain of synaptotagmin in this assay, as we previously found for the GST-C2 domain of the Drosophila PI 3-kinase_68D (26).
The ability of the individual GST-C2 domains to bind 45 Ca 2ϩ in the presence or absence of phospholipid vesicles was examined next. The individual C2 domains were able to bind 45  These observations indicate that the C2 domain of PI 3-kinase C2␤ binds weakly to Ca 2ϩ in the presence of anionic phospholipids, however, this binding does not appear to be co-operative, as the binding of phospholipids is not affected by the presence or absence of Ca 2ϩ ions (Fig. 6). Taken together, these studies suggest that there could be two independent, low-affinity, binding sites for Ca 2ϩ and phospholipids on the C2 domain of PI 3-kinase C2␤.
The function of the C2 domain in the context of the fulllength PI 3-kinase C2␤ was studied by comparing the properties of this construct with those of a C2 domain deletion mutant of the enzyme; this construct was generated by introducing a stop codon at the start of the cDNA sequence encoding the C2 domain. The construct was expressed as a NH 2 -terminal Myctagged protein in HEK293 cells and purified with anti-Myc antibodies. The expression levels of the mutant protein were found to be similar to those of the wild-type Myc-tagged protein, when assessed by SDS-PAGE and Coomassie Blue staining of immunoprecipitates obtained from transfected HEK293 cells (data not shown). The C2 deletion mutant also displayed an apparent molecular mass on SDS-PAGE consistent with a COOH-terminal deletion of about 15 kDa (136 amino acids, data not shown). When assayed in an in vitro lipid kinase assay, the C2 deletion mutant protein showed increased activity toward PI in the presence of Mg 2ϩ , when compared with the activity of the full-length enzyme (Fig. 7A). Similar results were obtained in the presence of Ca 2ϩ (data not shown). While the lipid kinase activity of the full-length enzyme was maximal at 100 M PI and decreased significantly at 1 mM PI (Student's t test, p ϭ 0.0266), the activity of the C2 deletion mutant increased with the concentration of PI. At 1 mM PI, the activity of the C2 deletion mutant was four times higher than that of the wild-type enzyme (Fig. 7A). This increase in lipid kinase FIG. 5. Inhibition of PI 3-kinase C2␤ lipid kinase activity by wortmannin and LY 294002. A, purified recombinant PI 3-kinase C2␤ was assayed for in vitro kinase activity in the presence of PI and increasing concentrations of wortmannin (in Me 2 SO). The control was treated with an equal volume of Me 2 SO. Radioactive phospholipid products were extracted, analyzed by TLC, and quantified using a Phospho-rImager. Data are mean with S.E. from three independent experiments and are plotted in percentage of control (100%). B, purified recombinant PI 3-kinase C2␤ (closed circles) or p85␣/p110␣ (closed squares) were assayed as above in the presence of increasing concentrations of LY294002 in Me 2 SO. The control was treated with an equivalent volume of Me 2 SO. The data are mean with S.E. from two independent experiments for PI 3-kinase C2␤ and one experiment for p85␣/p110␣ and are plotted as percentage of control (100%).

FIG. 6. Binding of phospholipids and Ca 2؉ to GST-C2 domains.
Purified recombinant GST, synaptotagmin GST-C2A, or C2B domains or the GST-C2 domain of PI 3-kinase C2␤ were incubated in the presence of 2 mM EGTA (Ϫ) or 1 mM CaCl 2 (ϩ), without (Ϫ) or with sonicated phospholipids (PI, PS, or PC). The phospholipids were pelleted by centrifugation and the proteins present in the pellet or the supernatant analyzed by SDS-PAGE and Coomassie Blue staining. The gels were scanned using a densitometer. The data are mean with S.E. of two independent experiments. activity was not caused by a significant change in the K m for Mg-ATP of the C2 deletion mutant (120 M), as compared with the wild-type enzyme (data not shown). Furthermore, both the wild-type and the C2 deletion mutant enzymes displayed similar sensitivities to the inhibitor wortmannin (data not shown), suggesting that deletion of the C2 domain does not alter the conformation of the catalytic domain of PI 3-kinase C2␤. It can thus be hypothesized that binding of PI to the C2 domain inhibits the lipid kinase activity of the catalytic domain. To gain further insight into the mechanism of this enzymatic inhibition, the wild-type enzyme and the C2 deletion mutant were incubated with a fixed concentration of PI (100 M) and increasing concentrations of PS, which has been shown to bind to the C2 domain (Fig. 6A). The addition of PS increased the lipid kinase activity of PI 3-kinase C2␤ at 10 M, and the effect was maximal (3-fold increase) at 100 M PS (Fig. 7B). The lipid kinase activity of the C2 deletion mutant also increased with the concentration of PS, but the maximal effect was reached at 1 mM PS (Fig. 7B). Taken together, these results suggest that, although PI and PS both bind to the C2 domain with similar affinities, only the binding of PI can inhibit the lipid kinase activity of PI 3-kinase C2␤. PS appears to enhance the lipid kinase activity independently of the C2 domain, although the presence of this domain decreases the concentration of PS necessary for maximal activation of the enzyme. The simplest model to account for these observations is that the C2 domain competes with the catalytic domain for binding to PI vesicles. To test this hypothesis, the recombinant enzyme was incubated with increasing concentrations of GST, GST-C2A, or GST-C2 domain of PI 3-kinase C2␤. At 10 M PI, the IC 50 of the recombinant C2 domain of PI 3-kinase C2␤ for inhibition of the lipid kinase activity was 0.09 M, which is 7-8 times lower than for GST or the GST-C2A in the presence of EGTA (Fig.  7C). The GST-C2A domain inhibited the lipid kinase activity even more efficiently in the presence of Ca 2ϩ (78% inhibition at 0.1 M). Taken together, these results support a model in which the C2 domain of PI 3-kinase C2␤ functions as a negative regulator of the lipid kinase activity of the enzyme, by directly competing with the catalytic domain for binding to substrate vesicles.
Modeling of the C2 Domain of PI 3-Kinase C2␤-The crystal structures of the C2A of synaptotagmin (53) and of the C2 domain of phospholipase C-␦1 (54) have been determined. Both structures have an eight-stranded ␤-sandwich with a calciumbinding pocket that is formed by the loops at one end of the sandwich. Topologically the two structures differ in the arrangement of the ␤-strands within the domain. A sequence alignment of the C2 domain of PI 3-kinase C2␤ with those of synaptotagmin and phospholipase C-␦1 showed that the C2 domain of PI 3-kinase C2␤ would be expected to adopt the synaptotagmin C2A fold (data not shown).
The electrostatic potential field prediction showed the existence of a large positively polarized field and the modeled C2 domain had a potential Ca 2ϩ -binding pocket. These predictions are consistent with the experimental results obtained above, where the recombinant C2 domain of PI 3-kinase C2␤ was shown to bind Ca 2ϩ and anionic phospholipids independently, but did not display a co-operative binding of these two ligands as was seen with the synaptotagmin C2A domain.
Subcellular Localization of PI 3-Kinase C2␤-The subcellular distribution of PI 3-kinase C2␤ was studied in HEK 293 cells using a hypotonic lysis protocol followed by differential centrifugation steps to separate the plasma membrane fraction, the low density microsomal fraction, and the cytosol. Endogenous PI 3-kinase C2␤ was found to be localized mainly in the low density microsomal fraction, with a small amount being present in the plasma membrane and the cytosol fractions (Fig. 8). Overexpressed recombinant PI 3-kinase C2␤ was associated only with the plasma membrane and low density microsomal fractions and was not found in the cytosol. These results demonstrate that, at least prior to activation, PI 3-kinase C2␤ associates almost exclusively with cellular membranes in epithelial cells. In an attempt to identify the domain(s) responsible for the interaction of PI 3-kinase C2␤ with cellular membranes, the distribution of the C2 domain deletion mutant was examined and found to be identical to that of the wild-type enzyme (Fig. 8). These results demonstrate that the C2 domain of PI 3-kinase C2␤ is not responsible for the localization of the enzyme to membranes in vivo.
Lack of Interaction of PI 3-Kinase C2␤ with Ras and p85␣-To compare the regulation of PI 3-kinase C2␤ with that of the Class I PI 3-kinases, the enzyme was assayed for its ability to interact in vitro with Ras and p85␣. Under conditions where recombinant Glu-tagged p110␣ interacted with the GTPbound form of Ras, there was no binding of purified recombinant Glu-tagged PI 3-kinase C2␤ to either GDP-or GTP-bound Ras (data not shown). Similar experiments using purified recombinant p85␣ and PI 3-kinase C2␤ failed to show any interaction between the two proteins (data not shown).

DISCUSSION
Class II PI 3-kinases have now been characterized from human, mouse, Drosophila, and Caenorhabditis elegans sources. The enzymes all have COOH-terminal C2 domains and, where examined, have been shown to be able to phosphorylate PI and PI(4)P but not PI(4,5)P 2 at the 3Ј position of the inositol ring in vitro. The human Class II enzymes include three distinct isotypes: PI 3-kinase C2␣, ␤, and the recently described PI 3-kinase C2␥ (55). The cDNA sequence of PI 3-kinase C2␤ is virtually identical to that of HsC2-PI3K which was recently reported without any functional characterization (40). The overall amino acid sequences of the two human Class II PI 3-kinase ␣ and ␤ isotypes differ most significantly in their amino-terminal regions (residues 1-300): the HRI, PIK, and C2 homology domains being similar. The restricted substrate specificity of the Class II enzymes was first reported for the Drosophila enzyme (26) and consequently for the human PI 3-kinase C2␣ (29). The enzymes are predominantly phosphatidylinositol 3-kinases in vitro, and can utilize PI(4)P but not PI(4,5)P 2 as substrates.
The two human isotypes C2␣ and C2␤ can be distinguished by their cation specificities: PI 3-kinase C2␣ has a preference for Mg 2ϩ when compared with Mn 2ϩ and Ca 2ϩ (29), 2 whereas the C2␤ isotype has an equivalent phosphatidylinositol 3-kinase activity in the presence of these three cations. The divalent cation specificity of the lipid kinase activity of PI 3-kinase C2␤ has been characterized for the first time, and the results show both that Ca 2ϩ ions complexed to ATP can functionally replace Mg-ATP in the active site of a lipid kinase, and that Ca 2ϩ can act as a cofactor for a phosphate transfer reaction. Whether these in vitro data are reflected in the in vivo regulation of PI 3-kinase C2␤ remains to be established. The intracellular concentrations of ATP and magnesium (1 mM) are much greater than that of calcium (50 M), thus ensuring that Mg-ATP is the predominant form of ATP in the cell. However, microdomains of high [Ca 2ϩ ] with values over 100 M have been described under the plasma membrane of living cells (56), and such sites may represent a unique environment where calcium could affect the regulation of the lipid kinase activity of PI 3-kinase C2␤. Perhaps high local concentrations of PI(4)P or PI(4,5)P 2 at the plasma membrane can exchange Mg 2ϩ , which is normally bound to ATP, for Ca 2ϩ , when its concentration under the membrane is increased by the opening of a Ca 2ϩ channel. Alternatively, the Mg 2ϩ to Ca 2ϩ exchange on ATP may take place in the active site of the lipid kinase of PI 3-kinase C2␤. If this were to happen, then an increase in Ca 2ϩ levels under the plasma membrane could activate PI 3-kinase C2␤ to phosphorylate lipids by forming Ca-ATP; however, in the presence of Mg-ATP, phosphorylation of PI(4)P by this enzyme could be inhibited. Thus an increase in [Ca 2ϩ ] would have a negative effect on the production of PI(3,4)P 2 by PI 3-kinase C2␤.
The two human isotypes can also be distinguished by their differing sensitivities to the inhibitors wortmannin and LY294002. While PI 3-kinase C2␣ is less sensitive to these compounds than is p110␣ (29), the IC 50 values for inhibition of PI 3-kinase C2␤ by both compounds are similar to those of p110␣.
The most distinguishing feature of Class II PI 3-kinases is the presence of the COOH-terminal C2 domain. Although the function of this domain has not yet been established, the C2 domain of PI3K_68D has been shown to bind anionic phospholipids in a calcium-independent manner (26). The results presented in this article demonstrate that the biochemical properties of the C2 domain of PI 3-kinase C2␤ are different from those of the C2A domain of synaptotagmin. Whereas the C2A domain of synaptotagmin co-operatively binds Ca 2ϩ and phospholipids (52), binding of negatively charged phospholipids to the C2 domain of PI 3-kinase C2␤ involves a Ca 2ϩ -independent site, this could be a positively charged strand of amino acids in the C2␤ domain. The C2 domain of PI 3-kinase C2␤ also appears to have a low-affinity calcium-binding site, which is independent of the phospholipid-binding site. The function of this calcium-binding site is unknown at present. The results of our preliminary experiments did not show any interaction between the recombinant C2 domain and other proteins, either in the presence or absence of Ca 2ϩ . The C2 domain does not appear to contribute to PI 3-kinase activity in the presence of Ca 2ϩ , as it binds Ca 2ϩ with only low affinity, when compared with the synaptotagmin C2A domain. Furthermore, deletion of the C2 domain did not affect the enzymatic activity of PI 3-kinase C2␤ in the presence of Ca 2ϩ .
The deletion of the C2 domain increased the in vitro lipid kinase activity of PI 3-kinase C2␤ toward PI. This effect was specific for PI, since binding of PS to the C2 domain did not result in an inhibition of the lipid kinase activity. The C2 domain may, therefore, compete with the kinase domain for binding to phosphatidylinositol substrate vesicles, thus behaving as a negative regulator of the lipid kinase activity of the enzyme. A putative mechanism of activation of PI 3-kinase C2␤ could thus involve proteolysis of the C2 domain, or prevention of PI binding to the C2 domain, which could be achieved by phosphorylation of this domain. It will be of interest to study the relevance of these findings in vivo.
The C2 domain does not target PI 3-kinase C2␤ to cellular membranes in vivo, as the subcellular distribution of a C2 domain deletion mutant was found to be identical to that of the full-length enzyme. The ability of PI 3-kinase C2␤ to associate with the plasma membrane and microsomal fractions of cells may be either as a result of interactions of another domain of the enzyme with phospholipids or of its association with membrane-bound proteins. The differences in biochemical properties observed between the C2A domains of synaptotagmin and PI 3-kinase C2␤ may thus explain the distinct functions of these domains. The high affinity, co-operative binding of Ca 2ϩ and phospholipids to the C2A domain is necessary for targeting synaptotagmin to the cell membrane (52). In contrast, the low affinity, constitutive PI-binding site of the C2 domain of PI 3-kinase C2␤ does not play a role in membrane targeting of the enzyme, but rather down-regulates its catalytic activity by competing for substrate binding. Our results described here demonstrate that purified recombinant PI 3-kinase C2␤ does not bind to either the p85␣ regulatory subunit of p110␣ or the GTP-bound Ras. Therefore, the regulation of this enzyme, in vivo, is distinct from that of the Class IA PI 3-kinases. The data presented in this article pose many questions about this unique PI 3-kinase that are intriguing to track down. For example, is PI 3-kinase C2␤ complexed to a regulatory subunit in vivo? Does the enzyme lie downstream of activated receptors, like the Class I family, or is it constitutively active, like the Class III phosphatidylinositol 3-kinases? The answers to these questions may help to identify yet another signaling pathway and biological process that requires PI 3-kinase for its maintenance and control.