G bg Stimulates Phosphoinositide 3-Kinase- g by Direct Interaction with Two Domains of the Catalytic p110 Subunit*

Class I phosphoinositide 3-kinases (PI3Ks) regulate important cellular processes such as mitogenesis, apo-ptosis, and cytoskeletal functions. They include PI3K a , - b , and - d isoforms coupled to receptor tyrosine kinases and a PI3K g isoform activated by receptor-stimulated G proteins. This study examines the direct interaction of purified recombinant PI3K g catalytic subunit (p110 g ) and G bg complexes. When phosphatidylinositol was used as a substrate, G bg stimulated p110 g lipid kinase activity more than 60-fold (EC 50 , ; 20 n M ). Stimulation was inhibited by G a o -GDP or wortmannin in a concen- tration-dependent fashion. Stoichiometric binding of a monoclonal antibody to the putative pleckstrin homology domain of p110 g did not affect G bg -mediated enzymatic stimulation, whereas incubation of G bg with a synthetic peptide resembling a predicted G bg effector domain of type 2 adenylyl cyclase selectively inhibited activation of p110 g . G bg complexes bound to N- as well as C-terminal deletion mutants of p110 g . Correspond-ingly, these enzymatically inactive N- and C-terminal mutants inhibited G bg

Class I enzymes purify as heterodimers with a molecular mass of about 200 kDa containing a catalytic subunit of 110 kDa (p110) and a regulatory subunit (12)(13)(14)(15)(16). Several mechanisms for regulating their enzymatic activity in response to extracellular stimuli have been elucidated. Among them the class I A members p110␣, ␤, and ␦ are stimulated by tyrosinephosphorylated proteins through interaction with regulatory PI3K subunits such as p85 or p55 (17,18). They in turn bind to the N terminus of the catalytic p110 subunit, thereby inducing PI3K activity. In contrast, the only known class I B member p110␥ does not bind to p85 adaptors, but instead associates with a noncatalytic p101 subunit (19). However, several lines of evidence indicate that G proteins stimulate p110␥ in the absence of p101 both in vitro and in vivo (20 -23). G␤␥ are thought to be the dominant physiological stimulus, while G␣ subunits of the G i but not G q or G 12 subfamilies only moderately activate p110␥.
Whereas a precise picture of the molecular mechanisms for receptor-induced activation of class I A PI3Ks has been drawn, little is known about how G␤␥ activates PI3K␥ and which structures of the enzyme are involved. Comparison of the deduced amino acid sequences of the catalytic subunits of class I members revealed several highly conserved regions of homology (HR) but also parts which are quite diverse (6). All enzymes have a C-terminally located catalytic domain (HR1). Interestingly, this domain requires N-terminal regions for enzymatic activity (9). Therefore a horseshoe-like folding of the p110, enabling interaction between the N-and C-terminal half of the enzyme, has been assumed (24). HR2 represents a PIK domain found in all PI3-and PI4-kinases. Other regions of homology are specific for PI3Ks (HR3) and class I PI3Ks (HR4). All class I PI3Ks also contain a Ras-binding motif (25). In contrast, only class I A enzymes have an N-terminal stretch assumed to interact with its p85 subunits, whereas only p110␥ exhibits a Ras-GAP homology region, which may fold to form a pleckstrin homology (PH) domain (6,20,26). This region of p110␥ has been speculated to be involved in G␤␥-mediated activation of PI3K␥, since PH domains of G␤␥-regulated proteins such as ␤-adrenergic receptor kinase or phosducin have previously been identified to bind G␤␥ (27,28). However, other enzymes with an inherent PH domain such as phospholipase C␥ are insensitive to modulation by G␤␥, while some effectors lacking PH domains, e.g. adenylyl cyclases and potassium or calcium channels, are regulated by G␤␥ complexes (29 -32).
Therefore, the aim of the present study was to examine whether the putative PH domain of the catalytic subunit of PI3K␥ is critical for interaction of p110␥ with G␤␥. Binding of a monoclonal antibody (mAb) to p110␥ that blocked the PH domain did not reduce G␤␥-mediated stimulation. Furthermore, results obtained with deletion mutants of p110␥ indicated that G␤␥ binds to an N-terminal region as well as to a region near or within the C-terminally located catalytic core. Correspondingly, inactive N-and C-terminal mutants inhibited G␤␥-mediated activation of wild type p110␥ by sequestration of G␤␥. We conclude that the PH domain of p110␥ is not the only region interacting with G␤␥ and hypothesize that Nand C-terminal stretches of p110␥ contribute to form a common G␤␥ effector region.

EXPERIMENTAL PROCEDURES
Construction and Purification of PI3K-GST Fusion Proteins-Construction of recombinant baculoviruses for expression of human GST-p110␥ fusion proteins and mutants thereof and of porcine p101 and GST-p101 were described previously (9,19,20,22). Recombinant baculoviruses for G␤ 1 and G␥ 2 subunits and for GST-p110␣ were generous gifts from Drs. M. Lohse (Wü rzburg) and M. D. Waterfield (London). For protein expression cells were infected at a multiplicity of infection of 1 virus per cell. After 48 -60 h of infection cells were pelleted by centrifugation (1,000 ϫ g), washed with phosphate-buffered saline twice, and resuspended in ice-cold buffer A containing 150 mM NaCl, 1 mM NaF, 1 mM EDTA, 50 mM Tris/HCl, pH 8.0, 10 mM dithiothreitol, 10 g/ml each of aprotinin, benzamidin, leupeptin, and 0.1 mM phenylmethylsulfonyl fluoride. Cells were disrupted by N 2 cavitation (30 min at 4°C, 25 bar) or by forcing the cell suspension through a 22-gauge needle (20 times) and subsequently through a 26-gauge needle (10 times). Nuclei and debris were discarded. The cytosolic and membranous fraction were recovered by centrifugation at 100,000 ϫ g for 50 min. Membrane extract (0.5% Lubrol PX in buffer A) and cytosol were incubated overnight with glutathione-Sepharose 4B beads (Pharmacia Biotech Inc.) prewashed with buffer A. The Sepharose-bound GST fusion proteins could be stored at Ϫ20°C in buffer B containing 50% glycerol, 1 mM EDTA, 40 mM Tris/HCl, pH 8.0, 1 mM dithiothreitol, and 1.57 mg/ml benzamidin. For enzymatic assays GST fusion proteins were freshly eluted with buffer C consisting of buffer A with 10 mM glutathione for 1 h at 4°C. Purified proteins were quantified by Coomassie Blue staining following SDS-PAGE with bovine serum albumin as the standard.
For coexpression experiments with recombinant G␤ 1 ␥ 2 equal multiplicity of infection numbers for all recombinant baculoviruses were used. After 58 -64 h of infection, cells were harvested by centrifugation (1,000 ϫ g, 10 min) and resuspended in 3 ml of ice-cold lysis buffer containing 0.5% Lubrol PX in buffer A. Lysates were incubated with glutathione-Sepharose 4B, and eluted GST fusion proteins were analyzed for binding of G␤␥.
Preparation of G Proteins-For isolation of bovine retinal transducin ␤␥ as well as G␣ o subunits and G␤␥ complexes from bovine brain, we employed standard techniques with modifications (33,34). Bovine brain G protein subunits were purified to apparent homogeneity in the presence of aluminum fluoride. Isolation and final purification of G␣ o and G␤␥ was achieved using a Mono Q (Pharmacia) fast protein liquid chromatography column (35). G protein subunits were identified by their immunoreactivity. Contamination by other pertussis toxin-sensitive G␣ subunits was excluded by analysis of autoradiographic signals after pertussis toxin-mediated [ 32 P]ADP-ribosylation with a BAS 1500 Fuji-Imager (Raytest, Straubenhardt, Germany) (36). Concentrations of G protein heterotrimers and their ␣ subunits were determined by binding of [ 35 S]GTP␥S (35), amounts of G protein ␤ subunits were determined by the method of Lowry et al. (37) and by Coomassie Blue staining following SDS-PAGE with bovine serum albumin as the standard (38). Purified proteins were stored at Ϫ70°C until use.
Gelelectrophoresis, Immunoblotting, and Antibodies-Generation of the monoclonal antibody against p110␥ and antiserum AS 398 against G␤ subunits were detailed elsewhere (22,39). For detection of p110␥ or the G protein subunit, preparations were fractionated by SDS-PAGE transferred to nitrocellulose or polyvinylidene difluoride membranes (Millipore, Eschborn, Germany). Visualization of specific antisera was performed using the ECL chemiluminescence system (Amersham, Braunschweig, Germany) or the CDP-Star chemiluminescence reagent (Tropix, Bedford, MA) according to the manufacturers' instructions.
Lipid Kinase Assay-The assays were conducted in a final volume of 50 l containing 0.1% bovine serum albumin, 2 mM EGTA, 0.2 mM EDTA, 10 mM MgCl 2 , 120 mM NaCl, 40 mM HEPES, pH 7.4, 1 mM dithiothreitol, 1 mM ␤-glycerophosphate as described previously (20) with some modifications. Briefly, 30 l of lipid vesicles (320 M phosphatidylethanolamine, 300 M phosphatidylethanolserine, 140 M phosphatidylethanolcholine, 30 M sphingomyelin, and 320 M PI) were mixed with either G␤␥ complexes or their vehicle and incubated on ice for 8 min. Thereafter the enzyme fraction (1-10 ng) was added and the mixture was incubated for further 10 min at 4°C in a final volume of 40 l. The assay was then started by adding 40 M ATP (1 Ci of [␥-32 P]ATP) in 10 l of the above assay buffer (30°C). Water-dissolved peptides (Eurogentec, Brussels, Belgium) used in this study were incubated with G␤␥ complexes before adding lipid vesicles. Wortmannin was stored in dimethyl sulfoxide (20 mM) in the dark at Ϫ20°C and added to the kinase immediately before the experiment. After 15 min the reaction was stopped with ice-cold 150 l of 1 N HCl and placing the tubes on ice. The lipids were extracted by vortexing samples with 450 l of chloroform/methanol (1:1). After centrifugation and removing of the aqueous phase, the organic phase was washed twice with 200 l of 1 N HCl. Subsequently, 40 l of the organic phase were resolved on potassium oxalate-pretreated TLC plates (Whatman, Cliffton, NJ) with 35 ml of 2 N acetic acid and 65 ml of 1-propanol as the mobile phase. Dried TLC plates were exposed to Fuji-Imaging plates, and autoradiographic signals were quantitated with a BAS 1500 Fuji-Imager.

RESULTS
Our initial experiments with the purified recombinant p110␥ yielded a 5-6-fold stimulation with 1 M G␤␥ and an EC 50 of 200 nM for transducin ␤␥ (20). As this effect of G␤␥ is moderate compared with effects on other G protein-regulated cellular targets, we optimized the conditions for p110␥ activation. Human recombinant p110␥ was expressed as GST fusion protein in Sf9 cells. Purified cytosolic enzyme exhibited a basal specific activity of 1-3 nmol of PI-3P/min/mg of p110␥ (Fig. 1A) and was much more sensitive to G␤␥ than the enzyme isolated from Sf9 membrane extracts (not shown). To investigate the stimulatory effect of G␤␥ complexes we used a mixture of highly concentrated purified bovine brain G␤␥ (see Fig. 1A) instead of using the less potent transducin ␤␥. Under these experimental conditions with phosphatidylinositol (PI) as substrate, G␤␥ complexes stimulated purified cytosolic p110␥ up to 60-fold with an EC 50 of ϳ20 nM in a bimodal manner (see Fig. 1, B and C). As expected, p110␣ was not stimulated by G␤␥ (not shown). G␤␥ increased the V max of the recombinant p110␥ for ATP, corresponding to the data obtained with the native purified PI3K␥ (23). Some variation in the efficiency of G␤␥ on p110␥ activity was probably due to either a well known variability in the quality of PI-liposome preparations used as substrates (40) or to the degree of autophosphorylation of p110␥. The specific covalent inhibitor wortmannin decreased G␤␥-stimulated enzymatic activity half-maximally at 15 nM and completely at 100 nM ( Fig. 2A). An excess of GDP-bound G␣ o also inhibited G␤␥mediated stimulation of p110␥ activity, most likely by sequestration of free G␤␥ (Fig. 2B).
To examine the involvement of the putative PH domain of p110␥ we took advantage of a mAb raised against the purified enzyme (22). The mAb binds specifically to an amino acid stretch of the enzyme corresponding to the PH domain postulated between amino acids 87 and 302 (6, 26). Fig. 3 shows that the mAb did not recognize constructs lacking amino acids 75-398 (see Fig. 3B Preparations containing active p110␥ were incubated with different volumes of antibody solution to ensure binding of saturating amounts of the mAb to the enzyme. The p110␥-antibody complex was purified on glutathione-Sepharose beads (Fig. 4, upper panels) and subsequently stimulated with G␤␥. As seen in Fig. 4 (lower panel), blocking the p110␥ PH domain by the mAb did not prevent G␤␥-mediated stimulation. Conversely, a peptide derived from the adenylyl cyclase 2 (AC2) (amino acids 956 -982; Fig. 5, upper part) not containing a PH domain did compete with p110␥ for the G␤␥ complex (41,42). Increasing concentrations of this peptide completely reversed G␤␥-mediated stimulation of p110␥ with an IC 50 of 50 M (see Fig. 5, lower part). The adenylyl cyclase 2 peptide did not change basal activity, and a control peptide derived from the corresponding region of the G␤␥-insensitive adenylyl cyclase 3 (AC3) did not compete for the G␤␥-stimulated activity (see Fig. 5).
To encircle regions of p110␥ required for G␤␥-elicited activation of PI3K, we studied the copurification of G␤ 1 ␥ 2 with various GST-p110 constructs following coexpression of recombinant proteins in Sf9 cells. First, to test the specificity of this experimental approach we coexpressed G␤ 1 ␥ 2 with GST-p110␣ and GST-p110␥ (Fig. 6, center and upper panels). Since G␤␥ does not activate p110␣ it should not copurify with p110␣, whereas it should copurify with the G␤␥-activated p110␥. p110 isoforms were isolated from cell homogenates on a glutathione affinity matrix. Subsequently proteins were subjected to SDS gels, and copurified G␤␥ complexes were detected by immunoblotting with a G␤-specific antibody (see Fig. 6, lower panel). This assay revealed that G␤ 1 ␥ 2 indeed copurified with p110␥, whereas no G␤ immunoreactivity was detected after purification of p110␣ or GST. Further experiments showed that G␤␥ copurified only with membrane-bound p110␥ but not with cytosolic p110␥, as could be explained by the membrane association of G␤␥ complexes (not shown).
Next, we coexpressed wild type p110␥ or different p110␥ mutants as GST fusion proteins together with G␤ 1 ␥ 2 and studied copurification. As shown in Fig. 7 These results indicate that G␤␥ specifically interacts with at least two binding sites of p110␥. One binding site is localized N-terminally, whereas the other one is found at the C terminus.
As has been shown before, all p110␥ deletion mutants lacking more than the first 97 amino acids were enzymatically inactive as well as a p110␥ construct containing a Lys 3 Arg mutation at position 799 (K799R), which abolishes wortmannin binding (9). To support the assumption that G␤␥ interacts with two regions of p110␥, we tested whether the inactive N-and C-terminal mutants would compete with wild type p110␥ for G␤␥. This was done by coinfecting Sf9 cells with a constant amount of recombinant baculovirus encoding enzymatically active wild type p110␥-GST fusion protein and increasing amounts of viruses encoding the N terminus (⌬741-1068) or the C terminus (⌬1-739) of p110␥ as GST fusion proteins. The  1. Stimulation of p110␥ activity by G␤␥ subunits. A, recombinant p110␥ was expressed as a GST fusion protein in Sf9 cells and purified from cytosol, whereas G␤␥ was isolated from bovine brain membranes as detailed under "Materials and Methods." Proteins were separated by SDS-PAGE and stained by Coomassie Blue. Apparent molecular masses of marker proteins are indicated. DF indicates the dye front of the gel. B, a 60-fold stimulation of p110␥ PI3K activity from basal (ϩbuffer) is seen in the presence of 200 nM G␤␥. Note that addition of buffer alone results in a slight decrease in PI3K activity. C, representative concentration response curve of purified recombinant p110␥ PI3K activity by G␤␥ purified from bovine brain. Enzyme activity was determined by measuring formation of radiolabeled PI-3P from PI and [␥-32 P]ATP using a phosphorimaging system. The inset shows the corresponding autoradiogram of 32 P incorporation into PI. Basal PI3K activity in the absence of G␤␥ corresponds to the left most PI-3P spot and was about 1-3 nmol/min/mg of protein.
coexpressed proteins were affinity-purified from Sf9 cytosol on glutathione-Sepharose beads and stimulated by addition of bovine brain G␤␥ complexes, and the fold activation of p110␥ was calculated. The C-terminal fragment bearing only the catalytic domain of p110␥ (⌬1-739) as well as the N terminus of p110␥ (⌬741-1068) inhibited G␤␥-mediated stimulation of wild type p110␥ in a concentration-dependent manner (Fig. 8). GST alone did not affect G␤␥-mediated stimulation of p110␥. The enzymatically inactive K799R p110␥ point mutant capable of binding to G␤␥ (see Fig. 7, lower panel, lane 10) inhibited enzymatic activity to the same extent as the deletion mutants (see Fig. 8). This observation is in accordance with results from the copurification experiments (see above) and underlines the hypothesis that p110␥ exhibits two domains interacting with G␤␥. Therefore, we suppose that both the N and C terminus bearing the catalytic domain of p110␥ are important for direct interaction with G␤␥. DISCUSSION This study examines the direct interaction between p110␥ and G␤␥. We show that the catalytic subunit of PI3K␥ is greatly stimulated by G␤␥ in the absence of the recently reported p101 subunit. It extents our previous findings (20 -22) supported by Tang and Downes (23) and points against an indispensable function of p101 for the stimulation of p110␥ by G␤␥ as recently hypothesized (19). Furthermore, preliminary results employing a purified recombinant heterodimer consisting of p110␥ and p101 showed no further increase in PI-3P formation in response to G␤␥ compared with p110␥ alone. At present there is no conclusive evidence for an essential role of p101 as an adaptor protein linking G protein-coupled receptors and p110␥ in a similar way as p85 mediates interaction of p110␣ with receptor tyrosine kinases (43). In addition, the recent observation that G␤␥ can bind directly to the related p110␤ supports an interaction with p110␥ as well (10). Accord- FIG. 3. Characterization of a monoclonal antibody (mAb) to p110␥. A, schematics of GST and GST fusion proteins of wild type p110␥ and mutants thereof, which were purified from baculovirusinfected Sf9 cells. B, immunoblot analysis of p110␥ and deletion mutants. Recombinant proteins were analyzed by immunoblotting after SDS-PAGE with a mAb to p110␥ as detailed under "Materials and Methods. "   FIG. 4. Effect of the p110␥ mAb on G␤␥-stimulated p110␥ PI3K activity. Immobilized p110␥ was preincubated in the absence or presence of two different volumes of mAb-containing solution (0.5 and 5 ml), eluted and stimulated by addition of half-maximally stimulating amounts of G␤␥ (25 nM) or vehicle only. PI3K activity was determined as described, and proteins were detected by immunoblot analysis using specific antisera (upper panels). Note that anti-IgG antibody recognized the heavy chain of the mAb (upper band).
FIG. 5. Effect of different concentrations of peptides derived from adenylyl cyclase type 2 (AC2) and type 3 (AC3) on basal and G␤␥-stimulated p110␥ PI3K activity. Peptides were preincubated without or with G␤␥ (25 nM) before addition of lipid vesicles and reaction mix containing p110␥. The lipid kinase assay was performed as described under "Materials and Methods." ingly, in reconstituted neutrophils we recently observed stimulation of p110␥ lipid kinase activity by the agonist of the G protein-coupled fMet-Leu-Phe receptor in the absence of p101 (21). A possible explanation for this apparent discrepancy is the observation that purified recombinant human p110␥ degrades rapidly in the absence of p101, this subunit could stabilize the enzymatic activity of p110␥. In addition, p101 could also affect the extent of autophosphorylation of p110␥ which may influence G␤␥-mediated activation of PI3K␥. Furthermore, the possibility that a small fraction of the recombinant human p110␥ associated with a putative insect cell-derived p101-like protein mediated all the G␤␥-sensitive activity is unlikely, since the basal specific activities of recombinant p110␥ recorded in this study was roughly 1 and 2 orders of magnitude larger than for recombinant p110␥ and p110␥/p101, respectively, as estimated from published results (19).
In our study we noticed that geranyl-geranylated G␤␥ complexes from bovine brain were 10 times more potent in stimulating p110␥ than previously used farnesylated bovine transducin ␤␥ (20). This difference in potency was also seen in studies with other G␤␥-regulated effectors (31). We found that half-maximal stimulation of the PI3K␥ catalytic subunit required only low nanomolar concentrations (ϳ20 nM) of bovine brain G␤␥. These concentrations are similar to those required for regulation of other effectors such as phospholipase C-␤, potassium, or calcium channels, but are much lower than those reported previously by other groups studying native or recombinant heterodimeric PI3K (14,16,19,23,31,40,44,45).
Only cytosolic but not membrane-extracted purified p110␥ was significantly stimulated by addition of exogenous G␤␥. The marginal responsiveness of the membrane-derived enzyme corresponds to the observation that recombinant G␤ 1 ␥ 2 copurified with membrane-extracted but not cytosolic p110␥. G␤␥ complexes may therefore function as a membrane anchor for p110␥. This in turn could facilitate the access of the enzyme to its lipid substrates thereby enhancing p110␥ activity. Interestingly, for the purified native PI3K␥ Tang and Downes (23) proposed cooperative kinetics for lipid substrates in the presence of G␤␥.
Direct interaction of p110␥ with G␤␥ raises the question which region of p110␥ is critical for interaction with G␤␥. Since p110␥ significantly differs from the receptor tyrosine kinaseregulated enzymes by an inherent PH domain it was speculated that this stretch may function as a G␤␥-binding site (6,20,26). Blocking the PH domain of enzymatically active p110␥  by an specific antibody reacting with this p110␥ domain did not reduce the stimulatory activity of G␤␥ on p110␥. Although the N-terminal half of p110␥ (⌬741-1068) did bind G␤␥ and inhibited G␤␥-mediated activation of the fully processed p110␥, we present evidence that interaction of G␤␥ with the putative p110␥ PH domain is not exclusively responsible for stimulation of PI3K␥ activity. In particular, p110␥ deletion mutants lacking the PH domain and adjacent stretches still bound G␤␥. Furthermore, coexpression of p110␥ with an enzymatically inactive C-terminal deletion mutant (⌬1-739) significantly inhibited activation of fully processed p110␥ by G␤␥. Similar results were obtained after mixing of separately purified p110␥ with increasing concentrations of mutants (not shown). Our results suggest that the catalytic core is important for G␤␥-mediated activation. This is not unlikely since the amino acid identity with p110␣ and -␤ in this region is only 45 and 46.5%, respectively. Recently, the motif QXXER has been proposed as a consensus binding site for G␤␥ (41). p110␥ but not p110␣, -␤, or -␦ contains these amino acids within the catalytic core (amino acids 888 -893), although they lack the correct spacing of this consensus motif. Unfortunately, peptides derived from corresponding regions of p110 isozymes strongly inhibited basal enzymatic activity preventing further analysis. In this context it should be noticed that a recent study on G protein-regulated calcium channels identified two G␤␥-interacting domains that do not contain a QXXER motif (32). Nevertheless, a peptide derived from the adenylyl cyclase 2, which is thought to interact with a G␤ domain, that for its part does not interact with PH-like structures, blocked G␤␥-induced activation of p110␥.
In summary, this study shows that the N as well as the C terminus of p110␥ interacts with G␤␥. This could be explained by a folding of the enzyme which results in proximity of N and C terminus. Indeed, several lines of evidence point to a horseshoe-like structure of p110 (24) since N-terminal regions outside the C-terminal catalytic core are mandatory for enzymatic activity and wortmannin binding (9). Based on the proposed horseshoe-like structure of p110␥ one may suggest that N-and C-terminal portions of p110␥ form a common G␤␥-effector region for regulation of enzymatic activity. The putative PH domain may be a part of this effector region but additional structures are required for G␤␥-mediated p110␥ activation. This finding supports the emerging concept in molecular and cell biology of PH structures being not sufficient do define molecules as G␤␥-regulated effectors.