p87 and p101 Subunits Are Distinct Regulators Determining Class IB Phosphoinositide 3-Kinase (PI3K) Specificity*

Background: p87 and p101 represent non-catalytic subunits of class IB PI3Kγ. Results: Expression and activity of PI3Kγ is modified differently by p87 and p101 in vitro and in living cells. Conclusion: Non-catalytic subunits of PI3Kγ represent two different regulators in the absence of Gβγ or Ras. Significance: p87 and p101 determine diversity within class IB PI3Kγ and allow integration in distinct PI3Kγ signaling pathways. Class IB phosphoinositide 3-kinase γ (PI3Kγ) comprises a single catalytic p110γ subunit, which binds to two non-catalytic subunits, p87 or p101, and controls a plethora of fundamental cellular responses. The non-catalytic subunits are assumed to be redundant adaptors for Gβγ enabling G-protein-coupled receptor-mediated regulation of PI3Kγ. Growing experimental data provide contradictory evidence. To elucidate the roles of the non-catalytic subunits in determining the specificity of PI3Kγ, we tested the impact of p87 and p101 in heterodimeric p87-p110γ and p101-p110γ complexes on the modulation of PI3Kγ activity in vitro and in living cells. RT-PCR, biochemical, and imaging data provide four lines of evidence: (i) specific expression patterns of p87 and p101, (ii) up-regulation of p101, providing the basis to consider p87 as a protein forming a constitutively and p101 as a protein forming an inducibly expressed PI3Kγ, (iii) differences in basal and stimulated enzymatic activities, and (iv) differences in complex stability, all indicating apparent diversity within class IB PI3Kγ. In conclusion, expression and activities of PI3Kγ are modified differently by p87 and p101 in vitro and in living cells, arguing for specific regulatory roles of the non-catalytic subunits in the differentiation of PI3Kγ signaling pathways.

In the case of class I A PI3Ks, p85 is responsible for the stabilization and inhibition of catalytic p110␣, p110␤, or p110␦ subunits (5)(6)(7)(8). Upon activation, p85 serves as an adaptor protein that interacts with phosphorylated tyrosine residues of membrane-localized receptor-tyrosine kinases, mediating translocation of PI3Ks from the cytosol to the plasma membrane (9,10). Association with the receptor induces conformational alterations within the PI3K, relieving the p85-mediated inhibition of p110 (11,12). Although p85 possesses adaptor and regulatory functions essential for appropriate spatial distribution and internal modulation in heterodimers, catalytic subunits of class I A determine the specificity and selectivity of PI3K signaling (13)(14)(15)(16)(17). This is reflected by the fact that an additional regulator, such as Ras or G␤␥, interacts directly with the catalytic subunits of PI3K␣ and PI3K␦ or PI3K␤, respectively (18 -25).
In contrast, there is only one catalytic subunit, p110␥, representing class I B . It forms two heterodimeric PI3K␥ variants, p87-p110␥ and p101-p110␥. Both variants can be regulated by G-protein-coupled receptors via interaction with G␤␥ (21, 26 -34). Based on initial data, it was proposed that p87 and p101 are functionally similar, acting as G␤␥ adapters anchoring PI3K␥ to the plasma membrane (31)(32)(33). The adapter function of the class I B non-catalytic subunits fairly resembles the recruitment process involved in the activation of class I A PI3K. However, in contrast to the well characterized inhibitory function of class I A p85, the role of p87 and p101 in the modulation of p110␥ activity remains obscure. More stringent examinations of PI3K␥ revealed that p87 and p101 do not function equivalently as G␤␥ adapters (34,35). Furthermore, we and others have demonstrated differential regulation of p87-p110␥ and p101-p110␥ and their integration in separate signaling cascades in vitro and in vivo (34, 36 -39). These findings in combination with the fact that only a single catalytic subunit of PI3K␥ is known led us to hypothesize additional non-redundant functions of p87 and p101 apart from the G␤␥ adapter function that should contribute to the specificity and selectivity of PI3K␥ signaling. Consistent with these assumptions, we have recently demonstrated that p101 but not p87 was able to rescue the stimulatory activity of G␤ 1 mutants incapable of activating p110␥ (35).
In the present study we show that PI3K␥ functions as an obligate heterodimeric enzyme, p87-p110␥ or p101-p110␥. Based on the expression pattern in human tissues, every tissue tested expresses both non-catalytic and catalytic subunits of PI3K␥, where p87 appears to be widely distributed and represents the dominant non-catalytic subunit of PI3K␥. Hence, p87-110␥ may be defined as a constitutively expressed enzyme, whereas p101-p110␥ acquires more selective roles in an inducibly expressed manner. We demonstrate that differences in PI3K␥ activities are determined and defined by the non-catalytic subunits, p87 and p101, in vitro and in living cells. In particular, aside from their impact on the differential regulation by G␤␥ and Ras, they exhibit distinct regulatory roles even in the absence of upstream stimulators, arguing for specific regulatory roles for the constitutively expressed p87-110␥ and the inducibly expressed p101-p110␥.

Real-time PCR Analysis for Expression of PI3K␥ Subunits in
Human Tissues-TissueScan TM human major tissue quantitative PCR arrays (OriGene: catalogue number HMRT102) were used as cDNA templates for RNA expression analysis in various normal (non-cancerous) human tissues. Real-time PCR amplifications were performed using RT2 quantitative PCR primer assays (Qiagen: #PPH15199A, #PPH15703A, #PPH02226A, and #PPH00150F for p87, p101, p110␥, and GAPDH, respectively) and RT2 SYBR Green quantitative PCR mastermix (Qiagen) according to the manufacturer's specifications. The realtime PCR cycling was carried out in a LightCycler 480 (Roche Diagnostics), and the data were analyzed using LightCycler 480 (1.5.0) software. The data were normalized by relative quantification of the target gene to a reference gene, GAPDH, based on crossing point (Cp) values.
Isolation of Human Peripheral Blood Mononuclear Cells-Human peripheral blood mononuclear cells were isolated from buffy coats obtained from the blood donation center of the university hospital of Tübingen using Ficoll-Paque (GE Healthcare) density gradient centrifugation. After lysis of any remaining red blood cells in ACK buffer (0.155 M NH 4 Cl, 0.01 M KHCO 3 , 0.1 mM EDTA), the cells were washed in PBS and adjusted to a density of 2 ϫ 10 6 cells/ml cell culture medium (RPMI 1640 supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine) with or without 10% FCS (all cell culture products from PAA). After 0, 4, and 24 h, the cells were harvested and washed in PBS.
Gel Electrophoresis, Immunoblotting, and Antibodies-Generation and characterization of the antisera against the G␤ 1 subunits and p110␥ are detailed elsewhere (28,41). A specific antibody against p101 was a generous gift from Len Stephens. In the current study two preparations of anti-p87 antibodies were used. The antibody against murine p87 was a generous gift from Michael Schaefer. For generation of an antibody against human p87, rabbits were immunized with peptide conjugates corresponding to the N-terminal sequence of p87 (amino acids ESSDVELDLQRSVQAVLREC). In the actual experiment, affinity-purified antibodies were used for visualization of p87.
Anti-GFP and anti-FLAG antibodies were purchased from Cell Signaling (#29565) and from Sigma (#F3165), respectively. Total protein lysates of human adult tissues, i.e. brain, lung, rectum, thymus, were purchased from BioChain (Newark). Proteins were fractionated by SDS/PAGE (10% acrylamide) and transferred onto nitrocellulose membranes (Hybond TM -C Extra, GE Healthcare). Visualization of specific antisera was performed using the ECL chemiluminescence system (GE Healthcare) or the SuperSignal West Pico Chemiluminescent Substrate (Pierce) according to the manufacturers' instructions. Chemiluminescence signals were estimated using the VersaDoc TM 4000 MP imaging system (Bio-Rad).
To analyze the exchange of non-catalytic subunits with heterodimeric PI3K␥ variants, Sf9 cells (100 ml) were infected with baculoviruses encoding tag-free p87 or p101. After 48 h, cells were harvested and resuspended in lysis buffer containing 25 mM HEPES/NaOH, pH 7.5, 300 mM NaCl, 10 mM ␤-mercaptoethanol, 45 mM imidazole, and EDTA-free protease inhibitor mixture tablets (1 tablet/50-ml solution, Roche Diagnostics). Cells were lysed by forcing them through a 22-gauge needle 5 times and subsequently through a 26-gauge needle 7 times in 8 ml of lysis buffer. The cytosolic fraction of Sf9 cells (2 ml) was incubated with purified (His) 6 p110␥, p87-(His) 6 p110␥, or p101-(His) 6 p110␥ (10 g of catalytic p110␥ subunit in the assay) for 1 h at 4°C. Thereafter, 50 l of Ni 2ϩ -Sepharose TM 6 Fast Flow beads (GE Healthcare) was added to the mixture and incubated for a further 30 min at 4°C. After extensive washing with lysis buffer, the proteins were eluted by adding 150 l of 1ϫ sample buffer according to Laemmli (50). Alternatively, after incubation, the mixture was ultrafiltered using Amicon Ultra-4 MWCO 100-kDa (Millipore) centrifugal filter devices according to manufacturer's specifications. The proteins in the Amicon filtrate were then analyzed by immunoblotting using specific antibodies.
Preparation of Phospholipid Vesicles-Phospholipid vesicles were prepared as described previously with some modifications (35). A 30-l phospholipid mixture containing 320 M phosphatidylethanolamine, 140 M phosphatidylcholine, 30 M sphingomyelin, and 40 M phosphatidylinositol 4,5-diphosphate with or without supplementation with different concentrations of phosphatidylserine (see below) was dried using N 2 gas and sonicated in buffer containing 40 mM Tris/HCl, pH 7.7, 0.1% BSA, 1 mM EGTA, 7 mM MgCl 2 , 120 mM NaCl, 1 mM DTT, and 1 mM ␤-glycerophosphate. To achieve equal association of each PI3K␥ variant with phospholipid vesicles in the absence of G␤ 1 ␥ 2 , the phospholipid vesicles containing 300 M phosphatidylserine were incubated with 32 nM PI3K␥ variant for 10 min at 4°C. To achieve equal association of each PI3K␥ variant in the presence of G␤ 1 ␥ 2 , p101-p110␥ (32 nM) was incubated with phospholipid vesicles lacking phosphatidylserine, whereas p110␥ or p87-p110␥ (32 nM concentrations each variant) were incubated with vesicles containing 180 M phosphatidylserine in the presence of 100 nM G␤ 1 ␥ 2 in every experimental setup for 10 min at 4°C.

RESULTS AND DISCUSSION
Expression of PI3K␥ Subunits in Human Tissues-Recent evidence suggests that PI3K␥ variants integrate into different signaling pathways (37-39). In particular, it was reported that p87-p110␥ and p101-p110␥ generate distinct cellular pools of PtdIns(3,4,5)P 3 (36). These data suggest either a cell-or tissuespecific nature of PI3K␥ variants or differential modulation of PI3K␥ depending on the non-catalytic subunit associated. To Human adult normal tissue total protein lysates obtained from BioChain were analyzed for expression of p87 and p101. A, representative blots showing expression of p87 and p101 in various human tissues. 50 g of protein was loaded per lane for brain, lung, and rectum, and 20 g of protein was loaded for thymus. B, p87 antibody raised against a human p87 peptide in rabbit was characterized by peptide blocking against human p87 overexpressed in HEK cells.

Real-time PCR analysis of mRNA expression of PI3K␥ subunits in human tissues
Real-time PCR amplification was performed using the TissueScan TM human noncancerous qPCR array (OriGene) as described under "Experimental Procedures." Shown are human tissues with p110␥ mRNA expression. The target (p87 or p101) mRNA to GAPDH mRNA ratio is indicated as follows: not detected (ND), Ͻ0.001; ϩ, 0.001 to 0.005; ϩϩ, 0.005 to 0.01; ϩϩϩ, Ͼ0.01. Human tissues expressing p110␥ always express either p87 or p101 or both, indicating that PI3K␥ exists as a heterodimeric enzyme in vivo, p87-p110␥ or p101-p110␥.

Human tissues p87 p101
Adrenal gland Determination of PI3K␥ Activity by p87 and p101 OCTOBER 25, 2013 • VOLUME 288 • NUMBER 43 examine a specific distribution of PI3K␥ subunits, we analyzed their expression levels in human non-cancerous tissues by realtime PCR and protein translation in some selected tissues by immunoblot analyses (Fig. 1). All human tissues expressing p110␥ also expressed either p87 or p101 or both (Table 1). We did not observe any tissue that contained p110␥ without noncatalytic subunits, suggesting that PI3K␥ exists as an obligate heterodimer, either p87-p110␥ or p101-p110␥. As reported earlier (31)(32)(33)36), the highest levels of PI3K␥ expression were found in tissues involved in immune responses, such as bone marrow, lymph nodes, blood leukocytes, spleen, and thymus (Table 1). However, we found significant differences in the expression of p87 and p101 in various tissues. Whereas p87 was expressed in almost all p110␥-positive tissues tested, distribution of p101 was more restricted. These unequal levels of expression and differential distribution of non-catalytic subunits in human tissues point to different cellular functions of both enzymes. Based on the broad expression of the heterodimeric p87-p110␥, it may be considered as a constitutively expressed enzyme, whereas the p101-p110␥ heterodimer fulfills selective roles as an inducible enzyme. To further support this hypothesis, we studied the protein expression of p87 and p101 in an easily available human cell model, i.e. primary cultured peripheral blood mononuclear cells, by immunoblots (Fig. 2). For semiquantitative assessment, the intensities of the p87 and p101 signals were normalized to Hsp90 expres- sion. Under these conditions p87 protein levels remained unchanged, whereas p101 protein signals increased in a timedependent manner ( Fig. 2A). Statistical analysis of at least 12 independent experiments using blood samples from 12 different donors revealed that the increase in the p101 signal was statistically significant (Fig. 2B).
These results are reminiscent of mouse data published earlier by Perino et al. (37). The authors showed increased p101 levels in mouse heart upon transverse aortic constriction, whereas p87 protein levels remained unchanged. The elevated protein levels of p101 in two different systems support not only the idea of different roles for the non-catalytic subunits of PI3K␥ but also the concept of p87-p110␥ as constitutive and p101-p110␥ as inducible heterodimers. Having found evidence for an isoform-specific regulation of PI3K␥ expression, we next looked for differences in their biochemical properties.
Biochemical Characterization of p87-p110␥ and p101-p110␥-The activation mechanism of lipid kinases may be dissected into two steps: translocation of the enzyme to its substrate at the plasma membrane and stimulation of its catalytic activity. In a previous study we showed that p87 and p101 differ with respect to membrane recruitment of PI3K␥ by upstream regulators (34). To study the impact of the non-catalytic subunits for stimulation of p110␥, we had to eliminate the impact of membrane recruitment in our biochemical approach. In doing so, we took advantage of the finding described by Kirsch et al. (44), i.e. the binding of anionic phosphatidylserine (PS) to p110␥. We first determined the optimal concentration of PS for the different experimental conditions to achieve similar amounts of p110␥ associated with phospholipid vesicles in the presence or absence of G␤ 1 ␥ 2 (Fig. 3A). In addition, we ensured that the selected PS concentrations did not interfere with G␤ 1 ␥ 2 association (Table 2). Under these optimized conditions, equal amounts of PI3K␥ variants could be detected in the vesicular fractions of each experimental setup (Fig. 3, B and C, upper panels, top rows, Table 3). In parallel setups, we determined lipid kinase activities (Fig. 3, B and C, upper panels, bottom  rows) that were subsequently quantified (Fig. 3, B and C,   graphs). In the absence of G␤ 1 ␥ 2 , the data show that the basal activity of p110␥ is slightly reduced in the presence of p87 but drastically enhanced in the presence of p101. These data indicate that the two non-catalytic subunits represent distinct regulators independent of their adaptor function and G␤␥.
Adding G␤ 1 ␥ 2 to the PI3K␥ variants resulted in increased activity of all enzymes studied (Fig. 3C, see change in y axis scale). Both non-catalytic subunits promoted stimulatory modulation of p110␥ in the G␤ 1 ␥ 2 -activated heterodimers. The effect of p101 was significantly higher as compared with p87 (Fig. 3C, graphs). This underlines that p87 and p101 have distinct profiles with respect to their role as independent regulatory subunits and as non-catalytic adaptors determining specificity toward upstream regulators. Moreover, in light of its tissue distribution, p87-p110␥ can be considered as a widely expressed enzyme that exhibits only low activity upon G␤␥ stimulation, whereas the p101-p110␥ variant may represent a selectively expressed PI3K␥ with a high rate of PtdIns(3,4,5)P 3 production. The low activity of p87-p110␥ upon stimulation by G-proteins can be interpreted as a possible way to maintain membrane homeostasis and/or serve as a coincidence detector integrating upstream signals from different pathways. The latter assumption is supported by our previous finding that the p87-p110␥ variant co-requires Ras-stimulation to gain full enzymatic activity (34). In this scenario the p87-p110␥ variant may be reminiscent of a characteristic feature of class I A PI3K␤, which was found by us and others to integrate signals from receptor-tyrosine kinases and G-protein-coupled receptors (19,21,29,45). The fact that some tissues and cells express both p87 and p101 prompted us to examine whether non-catalytic subunits of PI3K␥ can be exchanged.
Reconstitution of Heterodimeric PI3K␥-To address the question of the interchangeability of the non-catalytic subunits, we studied the enzymatic activity of both heterodimers in coexpression and reconstitution approaches. Therefore, we purified p85␣ (a non-catalytic subunit of class I A PI3Ks) as a control, p87, p101, p110␥ monomers, and p87-p110␥ and p101-p110␥ heterodimers to apparent homogeneity (Fig. 4A). p85␣ and p87 appeared to be stable in the absence of their catalytic subunits, whereas the yield of purified p101 was about 20 -40 times lower compared with expression of the other subunits. This confirms and extends our previous observations regarding the in vivo instability of p101 after expression in HEK 293 cells (30,33).

dimer with phospholipid vesicles
Recombinant purified G␤ 1 ␥ 2 dimers were incubated with phospholipid vesicles containing different amounts of anionic PS in the presence and absence of PI3K␥ variants. Phospholipid vesicles were prepared as detailed under "Experimental Procedures." Aliquots of sedimented phospholipid vesicles and their supernatants were subjected to SDS/PAGE followed by immunoblotting using G␤ 1-4 -specific antiserum. Chemiluminescence signals were estimated with a VersaDoc TM 4000 MP imaging system (Bio-Rad). For calculation of vesicle-associated G␤ 1 ␥ 2 , signal intensities in the sedimented phospholipid vesicles and its supernatant were added and considered as 100%. Shown here are the mean values Ϯ S.E. of four separate experiments. PI3K␥ variants do not affect the association of G␤ 1 ␥ 2 with phospholipid vesicles.

Phospholipid vesicles
Coincubation of G␤ 1   Determination of PI3K␥ Activity by p87 and p101 OCTOBER 25, 2013 • VOLUME 288 • NUMBER 43 Individually expressed and purified p87 or p101 were incubated with p110␥ in the presence of substrate-containing liposomes followed by assessment of the lipid kinase activity. Both p87 and p101 together with p110␥ were able to almost fully reconstitute G␤ 1 ␥ 2 -stimulated enzymatic activity compared with the activity determined in parallel experiments using the coexpressed heterodimers (Fig. 4B). The activity of PI3K␥ was selectively reconstituted in the presence of p87 or p101, whereas p85␣, a subunit of PI3K␣, -␤, and -␦, failed to reconstitute a functional enzyme (Fig. 4B). For concentration-response studies, we chose p101-p110␥ because it possessed the best signal-to-noise ratio. Coincubation of p101 with p110␥ enhanced G␤ 1 ␥ 2 -induced translocation of p110␥ to phospholipid vesicles, which was indistinguishable from the translocation of coexpressed and preformed heterodimer (Fig. 4C). Stimulation of lipid kinase activity by G␤ 1 ␥ 2 revealed a similar concentration-response correlation of the reconstituted complexes as compared with coexpressed p101-p110␥ preparations (Fig. 4D). Although the cellular function of p110␥ autophosphorylation is not yet clear (43), we studied the autophosphorylation of p101-p110␥ as an additional enzymatic feature of PI3K␥. The high basal and autonomous activity of monomeric p110␥ was transformed into a G␤␥dependent autophosphorylation of p101-p110␥, which was similar between coexpressed and reconstituted dimers (Fig. 4E). The reconstitution of fully active heterodimeric enzymes from individually purified PI3K␥ subunits prompted us to further validate the stability of the heterodimers and their ability to exchange noncatalytic subunits.
Interaction of p87 and p101 with p110␥ in Heterodimeric Complexes-We incubated heterodimeric p87-p110␥ and p101-p110␥ complexes with increasing concentrations of p85␣, p87, or p101 in the presence of constant concentrations of G␤ 1 ␥ 2 . Subsequently, their lipid kinase activities were tested (Fig. 5A). Class I A p85␣, unable to form a complex with p110␥, did not alter the activities of p87-p110␥ and p101-p110␥ heterodimers. Class I B p87 and p101 also did not affect the activities of p87-p110␥ and p101-p110␥, respectively. These results demonstrate the homogeneous nature of the purified heterodimeric complexes, FIGURE 4. Reconstitution of heterodimeric PI3K␥. A, recombinant class I A p85␣ subunit, individual subunits of PI3K␥, and heterodimeric variants were expressed in and purified from Sf9 cells. Proteins were subjected to SDS/PAGE (10% acrylamide) and analyzed by Coomassie Brilliant Blue staining. Molecular mass is given in kDa. B, lipid kinase activity of monomeric p110␥ subunit in the presence of 200 nM G␤ 1 ␥ 2 was significantly increased after coincubation with either p87 or p101 and reached the level of intensity of coexpressed p87-p110␥ or p101-p110␥, respectively. The assays were performed as described previously (35). For the reconstitution of heterodimeric PI3K␥ variants, p110␥ was incubated with a 10-fold molar excess of p87 or p101. Coincubation of p110␥ with class I A p85␣ served as a negative control. The data shown here are mean values Ϯ S.E. (n ϭ 3). C, G␤ 1 ␥ 2 -mediated recruitment of p110␥, p101-p110␥, and p110␥ coincubated with p101 to phospholipid vesicles. Pulldown assays were performed in the presence of 32 nM p110␥ or p110␥/p101 and 32 nM p101 subunit as detailed previously (35). Sedimented phospholipid vesicles were subjected to SDS/PAGE (10% acrylamide) followed by immunoblotting using specific anti-p110␥ antiserum. D and E, stimulation of lipid kinase activity and autophosphorylation of p110␥ (E), p101-p110␥ (•), and p110␥ coincubated with p101 () in response to increasing concentrations of G␤ 1 ␥ 2 were studied. The assays were conducted as detailed previously (35) with some modifications. The lipid kinase assays (D) were performed in the presence of 1.6 nM p110␥ or p101-p110␥ and 16 nM p101, whereas autophosphorylation (E) was studied in the presence of 6.4 nM p110␥ or p101-p110␥ and 32 nM p101. G␤ 1 ␥ 2 -induced activation of PI3K␥ is illustrated as a percentage of the maximal stimulation of coexpressed heterodimeric p101-p110␥. The lipid kinase activity of monomeric p110␥ subunit is known to be sensitive to G␤ 1 ␥ 2 -induced stimulation in vitro (26,34,35). The almost complete loss of p110␥ sensitivity to G␤ 1 ␥ 2 was due to the higher concentrations of non-ionic detergent, polyoxyethylene-10-lauryl ether, in the assay. The data shown here represent the averages of two independent experiments. as increases in lipid kinase activity in these experimental setups would be indicative of concomitant purified monomeric p110␥. The situation changed drastically in configurations where the opposing non-catalytic subunits were applied in a concentrationdependent manner. Incubation of p87-p110␥ with increasing concentrations of p101 significantly enhanced G␤ 1 ␥ 2 -stimulated lipid kinase activity in a concentration-dependent manner (Fig.  5A). In contrast, the application of increasing concentrations of p87 to p101-110␥ had no impact on lipid kinase activity. This was surprising, as one would expect a significant decrease in activity due to the formation of new p87-p110␥ complexes, which are less sensitive to G␤ 1 ␥ 2 .
In light of the instability of solitary purified p101, the inability of p87 to change the activity of p101-p110␥ argues for strong binding of p101 to p110␥ and hence a high stability for heterodimeric p101-p110␥ complexes. Further support for this feature comes from an immunoprecipitation approach (Fig.  5B). A commercially obtained anti-p110␥ antibody was used to precipitate heterodimeric PI3K␥ variants. In the case of the p101-p110␥ dimers, both subunits showed comparable signals in immunoblots of the precipitated samples and also of the unprecipitated sample. However, when the p87-p110␥ dimers were immunoprecipitated, the signals of p87, but not p110␥, were reduced compared with their starting products. This argues for a reduced recovery of p87, suggesting the complex stability of p87 with p110␥ is weaker compared with complexes with p101.
The fact that p101 and p110␥ can be purified independently and reconstituted into a stable heterodimeric complex can be described as a "click-in" mechanism in which p101 finds its final stable conformation resistant to protein degradation and protein replacement by p87. On the other hand, the concentrationdependent increase in activity after application of p101 to the heterodimeric p87-p110␥ complex raised the question of whether the increase in activity resulted from a newly formed heterodimeric or just from a heterotrimeric PI3K␥ complex. To address this question, we chose two independent experimental approaches, affinity copurification (Fig. 5C) and ultrafiltration (Fig. 5D).
The affinity copurification approach was based on immobilized p87-p110␥ complexes using hexahistidine-tagged p110␥. The immobilized complexes bound to Ni 2ϩ -Sepharose TM beads were coincubated with tag-free p101 and subsequently washed, eluted, and analyzed by immunoblot (Fig. 5C). Parallel experiments in the presence and absence of monomeric p110␥ subunits incubated with non-catalytic subunits were used as controls. Incubation of p87 or p101 with p110␥ resulted in reconstitution of the heterodimers, supporting the functional data (Fig. 5C, center panel). A strong nonspecific binding of p87 to the Ni 2ϩ -Sepharose TM beads was obvious in the control experiments (Fig. 5C, right panel), making it difficult to conclude whether a heterodimeric or a heterotrimeric complex was formed in the experiments testing the association of monomeric p101 with p87-110␥. Nevertheless, we saw a tight association of p101 with p110␥ (Fig. 5C, left panel), which corresponds to the functional data shown in Fig. 5A.
The ultrafiltration approach was chosen to study the dissociation of p87 or p101 from p87-p110␥ or p101-p110␥, respectively. The protein mixtures were separated using a 100-kDa FIGURE 5. Different stabilities of heterodimeric PI3K␥ complexes. A, heterodimeric p87-p110␥ and p101-p110␥ complexes were incubated with increasing concentrations of individually purified p87 or p101 as indicated above for 30 min at 4°C, and the lipid kinase activities were estimated in the presence of 200 nM G␤ 1 ␥ 2 as described previously (35). The assays are represented as the percentage of p101-p110␥ stimulation by 200 nM G␤ 1 ␥ 2 . Coincubation of PI3K␥ with class I A p85␣ subunit served as a negative control. The data shown here are the averages of three independent experiments. B, purified heterodimeric p87-p110␥ and p101-p110␥ complexes were immunoprecipitated (IP) using specific anti-p110␥ antibody as detailed under "Experimental Procedures." Interaction of the antibody with catalytic p110␥ subunit leads to significant release of p87 from the heterodimeric p87-p110␥ complex, whereas the heterodimeric state of p101-p110␥ is unaltered. C, interaction of p87 and p101 with recombinant purified PI3K␥ variants. Incubation of proteins and copurification using Ni 2ϩ -Sepharose TM 6 Fast Flow beads (GE Healthcare) was performed as detailed under "Experimental Procedures." Aliquots of the eluates were separated by SDS/PAGE (10% acrylamide) and analyzed by immunoblotting using antibodies raised against p87, p101, and p110␥. p101 associated with p110␥ in heterodimeric p87-p110␥ complexes. D, p87-p110␥ releasing the non-catalytic p87 subunit. Incubation of PI3K␥ variants with non-catalytic subunits and filtration through Amiconா Ultra-4 MWCO 100 kDa (Millipore) centrifugal filtration devices was done as detailed under "Experimental Procedures." Aliquots of inputs (5 l) and filtrates (15 l) were subjected to SDS/PAGE (10% acrylamide) followed by immunoblotting using specific antibodies against p87 and p101. p87 was observed in the filtrate of the centrifugal filtration device indicating dissociation of the heterodimeric p87-p110␥ complex. molecular weight cutoff centrifugal filter device and analyzed by immunoblot (Fig. 5D). In mixtures testing p87 in the presence and absence of p101-p110␥, p87 was present in the filtrate as well as in mixtures testing p101 in the presence of p87-p110␥, indicating release of p87 from the heterodimeric complex (Fig. 5D). This finding suggests that p87-p110␥ represents a less stable PI3K␥ variant than p101-p110␥ and argues against the occurrence of a heterotrimeric complex.
Our data expand the differences between the two PI3K␥ variants with respect to several new aspects; that is, an inducible, selectively expressed, highly stable, and highly active p101-p110␥ and a constitutive, ubiquitously expressed, and less stable p87-p110␥. Their physiological roles can be discussed in different ways. Up-regulation of p101 would enable the cell to switch PI3K␥ activity from a low to a high G-protein-sensitive activity state. Alternatively, the differences in activity and stability argue for the integration of PI3K␥ into independent pathways, e.g. p101-p110␥ in a pathway dominated by G␤␥ and a second p87-p110␥ preferred pathway with yet unidentified players that may use Ras proteins and/or protein kinases (34,46).
Characterization of p87-p110␥ and p101-p110␥ in Living Cells-To extend and validate the in vitro data, the ability of p87 and p101 to affect the activity of p110␥ was studied in living cells. HEK 293 cells were cotransfected with the plasmids encoding different PI3K␥ variants and the PtdIns(3,4,5)P 3 sensor, GFP-Grp1 PH (Fig. 6A). Stimulus-dependent translocation  , 10 m). B, protein expression in HEK 293 cells evaluated by immunoblotting using anti-p110␥, anti-p87, anti-p101, and anti-GFP (CFP) antibodies. Anti-GFP (CFP) was used to detect CFP-p85␣ and GFP-Grp1 PH . C, the scatter plot represents the statistical evaluation of the membrane translocation of the PtdIns(3,4,5)P 3 sensor, GFP-Grp1 PH , in the corresponding experiments. The data shown here are mean values Ϯ S.E. of three independent experiments comprising a total of 15-18 cells per condition. D, generation of PtdIns(3,4,5)P 3 requires catalytic activity of the constitutively membrane-associated p110␥ CAAX . HEK 293 cells were transfected with plasmids encoding kinase-defective YFP-p110␥(K833R) CAAX mutant and GFP-Grp1 PH . Shown here are representative images of cells starved for 18 h (confocal laser-scanning microscope slices of 1 m) from three independent experiments (scale bar, 10 m). Although YFP-p110␥(K833R) CAAX is localized at the plasma membrane, loss of its catalytic activity impairs PtdIns(3,4,5)P 3 synthesis and, hence, translocation of its sensor, GFP-Grp1 PH , to the plasma membrane. E and F, activity of constitutively membrane-associated p110␥ CAAX is not affected by endogenous G␤ 1 ␥ 2 or Ras. E, HEK 293 cells were transfected with plasmids encoding PI3K␥ (p110␥ CAAX , p87-p110␥ CAAX and p101-p110␥ CAAX ), GFP-Grp1 PH , ␤-adrenergic receptor kinase (␤Ark)-CFP and FLAG-neurofibromin 1 (NF1). After starvation for 18 h the cells were imaged (confocal laser-scanning microscope slices of 1 m) and then lysed. The cell lysates were analyzed by immunoblotting for the expression of the plasmids using anti-p110␥, anti-FLAG, and anti-GFP (CFP) antibodies. F, the histogram represents statistical evaluation of the membrane translocation of GFP-Grp1 PH . The data represent the mean values Ϯ S.E. of three independent experiments comprising a total of 15-18 cells per condition. Although p87 displays some stimulation of membrane-associated p110␥ CAAX , this effect was not statistically significant. The activity of p110␥ CAAX was significantly enhanced by p101. of PI3K␥ to the membrane and, therefore, to the substrate was eliminated by using p110␥ C-terminally fused to a CAAX-box motif (p110␥ CAAX ) (30). Expression of each protein was verified by immunoblotting (Fig. 6B). The amount of catalytically active p110␥ CAAX was not affected by the coexpression of additional proteins. Accumulation of p110␥ CAAX at the plasma membrane correlated with enhanced generation of PtdIns(3,4,5)P 3 (Fig. 6, A  and C). To rule out kinase-independent effects due to the presence of a superfluous protein at the plasma membrane, we analyzed PtdIns(3,4,5)P 3 -induced translocation of GFP-Grp1 PH in the presence of the catalytically inactive p110␥(K833R) CAAX mutant (30,33,43,47,48). Although the K833R mutation did not affect localization of p110␥ CAAX at the plasma membrane, the PtdIns(3,4,5)P 3 sensor GFP-Grp1 PH was not translocated from the cytosol due to the blunted catalytic activity of the enzyme (Fig. 6D). Statistical analysis revealed a significant increase in p110␥ CAAX activity in the presence of p101 (Fig. 6C). In contrast, p87 displayed a weak and insignificant stimulatory effect on p110␥ CAAX . The inability of class I A p85␣ to modulate the activity of p110␥ CAAX validated the specificity of class I B non-catalytic subunits in these experiments.
To evaluate the impact of the known activation of PI3K␥ by endogenous G␤␥ and Ras, we studied the effect by coexpressing either with ␤-adrenergic receptor kinase, a scavenger of G␤␥, or with neurofibromin 1, a RasGAP protein, or both (Fig. 6E) (34,49). Neither ␤-adrenergic receptor kinase nor neurofibromin 1 altered the activities of the PI3K␥ variants, arguing for PtdIns(3,4,5)P 3 production by heterologously expressed PI3K␥ variants independent of endogenous G␤␥ and Ras activation (Fig. 6F). In summary, the findings in living cells support our in vitro data of differential regulation of p110␥ by p87 and p101.
Conclusion-Our study provides strong evidence for several new isoform-specific features of the non-catalytic PI3K␥ subunits, i.e. (i) direct but divergent regulation of p110␥ by p87 and p101, (ii) different complex stabilities of the two subunits with p110␥, (iii) diverse spatial and temporal distribution of the PI3K␥ variants in human tissues, considering p87 as a protein forming a constitutively and p101 as a protein forming an inducibly expressed PI3K␥. Together with the previously detected distinct G␤␥ adapter functions of p87 and p101 and their different sensitivities toward upstream activators, these differences establish the basis for a specific, multifaceted, and finely tuned PI3K-dependent signaling network. We, therefore, conclude that p87-p110␥ and p101-p110␥ represent different and non-redundant variants of PI3K␥ assigned to independent signaling cascades.