A new role for the p85-phosphatidylinositol 3-kinase regulatory subunit linking FRAP to p70 S6 kinase activation.

The serine/threonine kinase p70 S6 kinase (p70S6K) phosphorylates the 40 S ribosomal protein S6, modulating the translation of an mRNA subset that encodes ribosomal proteins and translation elongation factors. p70S6K is activated in response to mitogenic stimuli and is required for progression through the G(1) phase of the cell cycle and for cell growth. Activation of p70S6K is regulated by phosphorylation of seven different residues distributed throughout the protein, a subset of which depends on the activity of p85/p110 phosphatidylinositol 3-kinase (PI3K); in fact, the phosphorylation status of Thr(229) and Thr(389) is intimately linked to PI3K activity. In the full-length enzyme, however, these sites are also acutely sensitive to the action of FKBP 12-rapamycin-associated protein (FRAP). The mechanism by which PI3K and FRAP cooperate to induce p70S6K activation remains unclear. Here we show that the p85 regulatory subunit of PI3K also controls p70S6K activation by mediating formation of a ternary complex with p70S6K and FRAP. The p85 C-terminal SH2 domain is responsible for p85 coupling to p70S6K and FRAP, because deletion of the C-terminal SH2 domain inhibits complex formation and impairs p70S6K activation by PI3K. Formation of this complex is not required for activation of a FRAP-independent form of p70S6K, however, underscoring the role of p85 in regulating FRAP-dependent p70S6K activation. These studies thus show that, in addition to the contribution of PI3K activity, the p85 regulatory subunit plays a critical role in p70S6K activation.

The serine/threonine kinase p70 S6 kinase (p70S6K) phosphorylates the 40 S ribosomal protein S6, modulating the translation of an mRNA subset that encodes ribosomal proteins and translation elongation factors. p70S6K is activated in response to mitogenic stimuli and is required for progression through the G 1 phase of the cell cycle and for cell growth. Activation of p70S6K is regulated by phosphorylation of seven different residues distributed throughout the protein, a subset of which depends on the activity of p85/p110 phosphatidylinositol 3-kinase (PI3K); in fact, the phosphorylation status of Thr 229 and Thr 389 is intimately linked to PI3K activity. In the full-length enzyme, however, these sites are also acutely sensitive to the action of FKBP 12-rapamycin-associated protein (FRAP). The mechanism by which PI3K and FRAP cooperate to induce p70S6K activation remains unclear. Here we show that the p85 regulatory subunit of PI3K also controls p70S6K activation by mediating formation of a ternary complex with p70S6K and FRAP. The p85 C-terminal SH2 domain is responsible for p85 coupling to p70S6K and FRAP, because deletion of the C-terminal SH2 domain inhibits complex formation and impairs p70S6K activation by PI3K. Formation of this complex is not required for activation of a FRAP-independent form of p70S6K, however, underscoring the role of p85 in regulating FRAPdependent p70S6K activation. These studies thus show that, in addition to the contribution of PI3K activity, the p85 regulatory subunit plays a critical role in p70S6K activation. p70S6K 1 is a mitogen-induced kinase with an important role in ribosome biogenesis. p70S6K phosphorylates the 40 S ribosomal protein S6, regulating the translation of a set of mRNA transcripts containing a polypyrimidine tract at the 5Ј tran-scriptional start site (1). Recruitment of these mRNAs to translating polysomes may be the mechanism by which p70S6K regulates cell growth (2). Inhibition of p70S6K with specific antibodies or with the immunosuppressant rapamycin blocks cell cycle progression through G 1 , implicating this serine/threonine kinase in the control of cell cycle and cell growth (3)(4)(5). Recent studies support this idea, because p70S6K-deficient Drosophila and p70S6K knockout mice show a significant reduction in body size (6,7).
p70S6K activity is tightly regulated by the coordinated phosphorylation of at least seven different residues. Four prolinedirected sites have been identified in the autoinhibitory domain at the C terminus of the protein (Ser 411 , Ser 418 , Thr 421 , and Ser 424 ). These residues are rapidly phosphorylated in response to mitogenic stimuli, but the kinase(s) responsible for this modification remains to be identified. Phosphorylation of these sites contributes to but is not sufficient for p70S6K activation; other critical residues, such as Thr 229 in the activation loop, Ser 371 , and Thr 389 in the linker region must be phosphorylated for full p70S6K activity (8). Phosphorylation of Thr 229 and Thr 389 is sensitive to wortmannin, an inhibitor of PI3K activity, and p70S6K can be activated by constitutively active mutants of PI3K in the absence of mitogenic signals (9,10), suggesting a role for PI3K in the regulation of p70S6K activity.
Whereas PI3K does not appear to phosphorylate p70S6K directly, the generation of 3-phosphoinositide (3-PtdIns) products by PI3K is required for Thr 229 and Thr 389 phosphorylation (11,12). Recent studies show that Thr 229 is phosphorylated by the 3-PtdIns-dependent kinase PDK1 (12)(13)(14), whereas Thr 389 is phosphorylated by the NEK6/7 kinases, members of neverin-mitosis-Aspergillus-like family kinases (15). Other authors propose that FRAP catalyzes this step, because Thr 389 phosphorylation is blocked by rapamycin (16). FRAP has also been shown to phosphorylate Thr 389 in vitro, although this observation has not been reproduced in vivo (17,18). A distinct function for FRAP, consistent with its ability to regulate phosphorylation of multiple p70S6K residues (19), was recently proposed by Peterson et al. (20). They show that PP2A associates directly and dephosphorylates p70S6K and that FRAP inhibits PP2A-mediated p70S6K dephosphorylation. Finally, mitogen-regulated phosphorylation of Ser 371 by a still uncharacterized kinase also contributes to p70S6K activation (21). The current model for p70S6K activation thus suggests that the coordinated phosphorylation of several residues is required for enzyme activation. According to this model, the phosphorylation of residues lying within the pseudosubstrate domain by proline-directed kinases releases the catalytic domain from the C-terminal autoinhibitory domain; this permits subsequent Thr 389 phosphorylation by the NEK6/7 kinases (15), further facilitating Thr 229 access to PDK1 (13,14,22). In addition, FRAP action would be required to protect p70S6K from PP2Amediated dephosphorylation (20).
Here we analyzed the contribution of PI3K to p70S6K activation in greater detail. PtdIns 3-kinases are lipid kinases that modify PtdIns at position 3 of the inositol ring (23). The first characterized form of PI3K, class IA PI3K, is a heterodimer composed of a catalytic (p110␣, ␤, or ␦) and a regulatory (p85␣, ␤, or p5␥) subunit (reviewed in Refs. 24 and 25). The primary sequence of the p110 catalytic subunit comprises several regions, including the p85-binding region, a Ras-binding domain, a region homologous to phosphatidylinositol 4-kinases, and the catalytic core (24,25). The p85 regulatory subunit has an SH3 domain, a Bcr homologous region flanked by two proline-rich regions, and two SH2 domains separated by an inter-SH2 region (24,25). Stimulation of growth factor receptor Tyr kinases induces p85/p110 binding to these receptors via the p85 SH2 domains, resulting in PI3K activation (26,27). Activated mammalian p85/p110 PI3K controls several important cell functions, including cytoskeletal organization, cell growth, cell division, and survival (28 -31). The mechanism by which p85/ p110 PI3K regulates such a variety of cellular responses is not fully known but involves the binding of proteins containing SH2 and pleckstrin homology domains to 3-phosphorylated PtdIns on the cell membrane (32,33), a feature that contributes to the induction of some PI3K effectors. In addition, p85 contributes to cell stimulation by recruiting signaling molecules to the stimulated receptors via its SH3, SH2, and Bcr homologous domains (24,25). Many cellular effectors of PI3K have been identified and include the small GTP-binding protein Rac, the Ser/Thr kinase AKT/PKB, Bruton tyrosine kinase, PDK1, certain PKC isoforms, and p70S6K (9, 34 -38).
In this study, we examined the contribution of the p85 regulatory subunit to p70S6K activation. We used a mutant p85 form, p65 PI3K , which lacks one of the adapter SH2 domains but binds to p110, exhibiting higher associated lipid kinase activity than p85/p110 and enhancing receptor-stimulated PI3K activation (39). With this construct, we show that PI3K regulation of p70S6K activation requires not only PI3K enzymatic activity but also the adaptor function of the p85 regulatory subunit of PI3K. In this function, p85 mediates formation of a multimolecular complex that links FRAP to p70S6K. Formation of this complex, mediated by the C-terminal SH2 domain of p85, is required for efficient p70S6K activation. This complex, however, is not necessary for p70S6K activation when PP2A is inhibited nor for activation of a FRAP-independent form of p70S6K that does not bind PP2A. These results show that p85 contributes to p70S6K activation by bringing FRAP into complex with p70S6K, which in turn protects p70S6K from dephosphorylation. Whereas PI3K activity is required for Thr 229 and Thr 389 phosphorylation, the p85-PI3K regulatory subunit is necessary for p70S6K to retain its active/phosphorylated conformation.
Transfections-The stable transfectants were described previously (39). For transient transfections, COS7 cells were seeded the day before transfection at an appropriate concentration. The cells were transfected with LipofectAMINE (Life Technologies, Inc.) using 1 g (35-mm dishes), 2-3 g (60-mm dishes), or 4 -5 g (100-mm dishes) of total DNA, according to the manufacturer's instructions. We observed that p85 stabilizes p110 expression in cotransfections, concurring with the previously described role of p85 in protecting p110 from degradation (43). To obtain similar p110 expression levels, the amount of p85-encoding cDNA used in the transient transfected samples was one-half that of the cDNA for p65 PI3K , which yields comparable p110 expression and moderately higher p65 PI3K expression than that of p85 (see Figs. [2][3][4][5][6][7][8]. The cells were incubated for 24 h after transfection in culture medium containing 10% fetal calf serum and then incubated in serum-free medium for an additional 24-h period prior to lysis and p70S6K assay. Cell Labeling, Reprecipitation Assays, Phosphotryptic Analysis, and Western Blot-For labeling experiments, the cells were washed in methionine/cysteine-free RPMI (BioWhittaker) and incubated in this medium supplemented with 10% dialyzed fetal calf serum for 2 h prior to addition of 0.5 mCi of [ 35 S]Met/Cys (Amersham Biosciences, Inc.) for an additional 4 h. The cells were collected and lysed in Triton X-100 lysis buffer (50 mM HEPES, pH 8, 150 mM NaCl, and 1% Triton X-100 with protease inhibitors) (39). Lysates were precleared three times for 1 h with 25 l of protein-A-Sepharose (Amersham Biosciences, Inc.) prior to immunoprecipitation (44). The immune complexes were then washed extensively and disrupted using 400 l of reprecipitation solution (0.4% SDS, 50 mM triethanolamine, 100 mM NaCl, 2 mM EDTA, and 2 mM ␤-mercaptoethanol) and boiling for 5 min. Supernatant was collected and placed on ice for 5 min, when 2% Triton X-100 and 10 mM iodoacetamide (final concentration) were added. Solubilized proteins were diluted 1:2 in lysis buffer, precleared, and subjected to immunoprecipitation. Western blots and phosphotryptic analysis were as described (39,19).
We first examined the basal activity and serum-induced activation of p70S6K in CMN-5 and T14 thymic lymphoma cell lines; these cell lines express p65 PI3K (CMN-5) or p85␣ (T14) (39). Both prior to and following serum stimulation, CMN-5 cells showed significantly lower p70S6K kinase activity levels than T14 cells, as estimated by in vitro phosphorylation of S6 (Fig. 1B). This suggested that the absence of p85 in CMN-5 cells impairs p70S6K activation. It was nonetheless possible that the difference in p70S6K activity in CMN-5 and T14 cells was unrelated to p65 PI3K or p85␣ expression and dependent on another genetic background difference. To standardize the genetic background, we examined clones of the IL-2-dependent CTLL-2 cell line stably expressing p65 PI3K or p85␣ at levels similar to that of endogenous p85␣ (39); CTLL-2 lines expressing an active p110 form (p110-CAAX) (39) were also examined. Even in the absence of stimulation, cells expressing an active p110 form showed increased p70S6K activity (Fig. 1C), as reported (9). In contrast, cells expressing p65 PI3K , which exhibit increased AKT activity (39), had lower p70S6K activity levels than control cells. Cells expressing 2-fold higher p85␣ levels showed moderately higher p70S6K activity levels than control cells (Fig. 1C). IL-2 addition increased p70S6K activity in all four stable cell lines (Fig. 1C); however, p70S6K activity was still lower in cells expressing p65 PI3K . The difference in p70S6K activity was higher in CMN-5 cells compared with T14 cells than when comparing CTLL2 clones, because CMN-5 cells fail to express p85, whereas p65 PI3K -CTLL2 cells still express p85. In conclusion, despite its ability to activate AKT (39), p65 PI3K expression impairs p70S6K activation, suggesting that p85 may contribute to p70S6K activation.
The lower p70S6K activity in p65 PI3K -expressing cells was puzzling, because previous studies showed that p65 PI3K /p110 is an active PI3K form that induces 3-PtdIns production and activation of Rac and AKT pathways (39). We therefore examined in more detail the activation of p70S6K ( Fig. 2A) and AKT (Fig. 2B) resulting from coexpression of the distinct PI3K regulatory subunits in combination with p110 in COS7 cells. The p65 PI3K mutant has a small region at the C terminus homologous to a conserved region in Eph Tyr kinase receptors ( Fig. 1A) (39). For these assays, we also tested a mutant (⌬24p65 PI3K ) encompassing the same portion of p85 as p65 PI3K but lacking the Eph homology domain.
The transfected cells overexpressed the recombinant p85, ⌬24p65 PI3K , and p65 PI3K regulatory subunits, as well as the wt or constitutively active p110 catalytic subunit (Fig. 2, A and B). Expression levels of recombinant PI3K proteins were higher than that of endogenous PI3K (Ͼ20-fold difference), minimizing the contribution of endogenous proteins for p70S6K and AKT activation. p70S6K and AKT proteins were expressed to a similar extent in the different transfected samples (Fig. 2, A and B, respectively). However, while p65 PI3K /p110 activated AKT similarly to a constitutively active p110 construct (p110-CAAX), it failed to activate p70S6K (Fig. 2, A and B). Similar results were obtained in ⌬24p65 PI3K /p110-expressing cells ( were resolved by SDS-PAGE, the gels were transferred to nitrocellulose, and expression of p70S6K was examined by Western blot. Other extracts (400 g/sample) were immunoprecipitated with an anti-p70S6K Ab, and p70S6K activity was examined in vitro using 40 S ribosomes as substrate. The radioactivity present in S6 substrate is represented (mean of three assays). C, expression of p70S6K was examined by Western blot (40 g/sample). p70S6K in vitro kinase activity in anti-p70S6K immunoprecipitates prepared from extracts (400 g/ sample) of serum-and IL-2-deprived or IL-2-stimulated (1 h) CTLL-2 cell lines (indicated). The radioactivity in S6 substrate is represented (mean of three assays).

FIG. 2. Truncation of the p85 regulatory subunit impairs PI3Kinduced p70S6K activation.
A, COS7 cells were transiently transfected with a vector encoding p70S6K alone (Control) or in combination with vectors encoding p110-CAAX (p110*), p65 PI3K plus wild type p110, p85 plus p110, or ⌬24p65 PI3K plus p110. At 24 h after transfection, the cells were incubated in serum-free medium for an additional 24 h prior to lysis. Lysates (40 g/sample) were examined by Western blot using the appropriate Ab. Cell extracts (20 g) were also immunoprecipitated with anti-p70S6K Ab and tested in kinase assay as in Fig. 1. Parallel samples labeled with [ 35 S]Met were lysed and lysates (300 g) immunoprecipitated using anti-p110 Ab. B, the cells were transfected as above, replacing the vector encoding p70S6K with a vector encoding HA-AKT. Lysates (40 g/sample) were examined by Western blot using appropriate Ab. Cell extracts (300 g/sample) were also incubated with anti-HA Ab, and AKT kinase activity was examined using H2B histone as substrate. C and D, p70S6K (C) and AKT (D) activity in the transfected samples compared with that of control cells. The values shown are the means of four assays performed as in A and B. p85-PI3K Controls p70S6K Activation 2). In contrast, whereas p85/p110 modestly increased AKT activity, concurring with the moderate increase in 3-PtdIns product levels in these cells (39); p70S6K activation in these cells was as efficient as that observed in active p110-expressing cells (Fig. 2). The mean cpm incorporation values of several assays are shown in Fig. 2 (C and D). Consistent with a role for the PI3K holoenzyme complex in both AKT and p70S6K activation (24,25), we confirmed that overexpression of p85 alone inhibits endogenous PI3K activity (39) and decreases AKT and p70S6K activation (not shown). In conclusion, the finding that p65 PI3K /p110 is able to activate AKT yet fails to activate p70S6K suggests that the p85 C-terminal SH2 domain contributes directly to p70S6K activation, independently of the generation of 3-PtdIns products.
p65 PI3K Blocks p70S6K Activation by Constitutively Active p110 -We reasoned that if the p85 C-SH2 domain contributes to p70S6K activation, then the p65 PI3K /active-p110 complex may act as a dominant negative mutant and block p70S6K activation by constitutively active p110. For these assays we used a rCD2p110 mutant, which activates p70S6K in a similar manner as p110-CAAX (10), but whose expression can be examined by Western blot using an anti-CD2 Ab. The cells were transfected with the different cDNAs, expression was determined by Western blot (Fig. 3A), and kinase activity for each treatment was measured as before. p70S6K activity was significantly reduced in p65 PI3K /active-p110-expressing cells compared with cells expressing active p110 or p85/active p110 (Fig.  3A). The comparison of p70S6K activity in p65 PI3K /active p110 and p85/active p110 cells (Fig. 3A) supports a role for the p85 C-SH2 domain in p70S6K activation. Furthermore, these results show that p65 PI3K overexpression inhibits p70S6K activation by PI3K, probably competing with endogenous p85 for active p110 association. In contrast, p65 PI3K /active p110 expression efficiently activated PKB/AKT (Fig. 3C). Thus, the defective activation of p70S6K in cells expressing p65 PI3K /active-p110 is thus not due to a defect in PI3K activity.
Phosphorylation of p70S6K in p65 PI3K -expressing Cells-p70S6K activation depends on the phosphorylation of several residues in the pseudosubstrate region and on subsequent PI3Kdependent phosphorylation of residues Thr 229 and Thr 389 . Phosphorylation of other sites, such as Ser 371 , has also been shown to regulate p70S6K activation (21).
We performed phosphotryptic analysis of p70S6K phosphoprotein isolated from cells transfected with p85/active p110 or p65 PI3K /active p110 (as in Fig. 3A) and subsequently labeled with [ 32 P]orthophosphate. The p70S6K band from p65 PI3K / p110-transfected samples showed significantly less 32 P labeling (85% reduction) than the p70S6K band obtained from p85/ p110 samples, confirming deficient p70S6K activation in p65 PI3K -expressing cells. Nevertheless, when a fraction of the 32 P-p70S6K band containing 10 3 cpm from either p65 PI3K /p110 or p85/p110 samples was trypsinized and resolved, both samples yielded rather similar phosphotryptic maps (Fig. 3D). The analysis thus failed to reveal a clear single-residue phosphorylation defect, suggesting that rather than a phosphorylation defect in one site, p65 PI3K exerts its effects on the phosphorylation of several p70S6K residues. Western blot with anti-phospho-Thr 389 and anti-phospho-S411Ab (not shown) also showed a similar decrease in the phosphorylation of both residues.
To confirm that selective defects of the phosphorylation of Thr 389 or of residues in the pseudosubstrate region were not responsible for the defective p70S6K activation, the cells were transfected with p70S6K mutants containing acidic substitutions in the pseudosubstrate region alone (D3Ep70S6K) or in Thr 389 as well (E389D3Ep70S6K) (14, 16, 46). Two other critical sites for p70S6K activation, Thr 229 and Ser 371 , could not be examined by this method, because mutation of these sites abolishes kinase activity (16,21). D3Ep70S6K or E389D3Ep70S6K were cotransfected with the distinct PI3K constructs, and recombinant protein expression was subsequently confirmed in Western blot. The samples expressed p85, ⌬24p65 PI3K , and p65 PI3K regulatory subunits, similar amounts of p70S6K mutants, and similar amounts of p110 subunits (examined by [ 35 S]Met metabolic labeling and immunoprecipitation) (Fig.  4A). Both p70S6K mutants displayed higher basal kinase activity than wild type p70S6K, as reported previously (14,16), and their activity further increased upon expression of rCD2p110 (Fig. 4). However, p70S6K did not reach the activation levels observed in cells expressing active p110 in either p65 PI3K /p110 or ⌬24p65 PI3K /p110 cells (Fig. 4A). These results show that mimicking the phosphorylation of the pseudosubstrate region residues or that of Thr 389 is not sufficient to restore p70S6K activity in p65 PI3K cells, reflecting that a specific defect in these residues is not the cause of the lack of FIG. 3. p65 PI3K acts as a dominant negative mutant for PI3Kinduced p70S6K activation. A, COS7 cells were transfected with a vector encoding p70S6K alone (Control) or in combination with plasmids for rCD2p110 (p110*), p65 PI3K plus rCD2p110, or p85 plus rCD2p110. The cells were incubated for 24 h with serum and for an additional 24 h in the absence of serum prior to lysis. p70S6K, p85 (or p65 PI3K ), and rCD2p110 content in the cell lysates was estimated by Western blot using anti-p70S6K, anti-p85, or anti-CD2 Ab, respectively. p70S6K activity in normalized cell extracts (20 g) was estimated by immunoprecipitation with specific Ab followed by in vitro kinase assay. B, extracts from transfected cells were processed and examined as in A. The mean-fold induction obtained in three assays is represented. C, cells were transfected as in A, replacing p70S6K vector with a vector encoding HA-AKT; samples were processed as in Fig. 2D. The mean of three assays is shown. D, phosphotrytic maps of p70S6K isolated from 32 P-labeled cells transfected as in A. p85-PI3K Controls p70S6K Activation p70S6K activity. Together, these observations suggest that rather than a phosphorylation defect in one of these sites, p65 PI3K exerts its effects on overall p70S6K phosphorylation or on the phosphorylation of a large number of p70S6K residues. Impaired p70S6K Activation by p65 PI3K /p110 Is Unrelated to a Defect in PKC Action-Several proteins have been implicated in p70S6K activation, including PI3K and its effectors AKT and Rac (9,35,47). p65 PI3K enhances p110 catalytic activity and induces activation of AKT and Rac effectors (39), making it unlikely that the defective activation of p70S6K by p65 PI3K is related to defects in AKT or Rac signaling pathways. The mammalian target of rapamycin, FRAP, also called mTOR or RAFT1, and the atypical PKC isoform PKC have also been implicated in p70S6K activation (17,18,48). We first analyzed whether PKC was responsible for the defect in p70S6K activation in p65 PI3K -transfected cells.
COS7 cells were cotransfected with p70S6K, p110, p85 (or p65 PI3K ) in combination with constitutively active PKC, dominant negative PKC, or constitutively active PKC⑀. All recombinant proteins were correctly expressed, as established by Western blotting (not shown). p70S6K activity in the samples was subsequently examined by in vitro kinase assay (Fig. 5). Whereas dominant negative PKC inhibited p85/p110-induced p70S6K activation, confirming the involvement of PKC in p70S6K activation (48), constitutively active PKC was unable to rescue p65 PI3K /p110 activation of p70S6K (Fig. 5). This suggests that the cause of defective p70S6K activation by p65 PI3K / p110 is not related to a defect in PKC action.
Formation of a Multimolecular Complex Including p85, p70S6K, and FRAP-p65 PI3K lacks the p85 C-SH2 domain, a region involved in protein-protein interactions (39); we therefore hypothesized that p65 PI3K may be defective in the formation of a protein complex regulating p70S6K activation. Using cells transfected with appropriate cDNA combinations, we examined whether p85 associated directly with p70S6K or with FRAP. Formation of protein-protein complexes was examined in [ 35 S]Met-labeled cells by immunoprecipitation of the p65 PI3K or the p85 regulatory subunits and subsequent examination of the associated proteins by reprecipitation. Fig. 6A shows a representative experiment in which the putative association of p70S6K with p65 PI3K or p85 was examined. The cells were transfected with cDNA encoding EE-p70S6K, Myc-rCD2p110, and either HA-p65 PI3K or HA-p85. Expression of the different proteins was examined by immunoprecipitation (Fig. 6A, left and center panels). Disaggregation of the proteins present in p70S6K immunoprecipitates, followed by a second immunoprecipitation using an anti-p85 Ab, showed that p85 was efficiently immunoprecipitated from p70S6K immune complexes, but a smaller amount of p65 PI3K appeared to be associated with p70S6K (Fig. 6A, right panel). For these assays we also used wild type p110, which yielded similar results (not shown). p70S6K therefore associates efficiently with the p85 regulatory subunit, and deletion of the C-SH2 domain in p65 PI3K results in a failure of complex formation.
To extend these studies, we investigated whether p85 was able to interact with FRAP. COS7 cells were transfected with cDNA encoding Myc-FRAP, EE-p70S6K, and untagged versions of wild type p110, as well as p65 PI3K or p85. For these experiments, Myc-tagged p110 was not used, because this tag was used for Myc-FRAP; the use of HA-FRAP fusion was also avoided, because anti-HA Ab did not immunoprecipitate HA- FIG. 4. Defective phosphorylation of the sites in the pseudosubstrate region or of Thr 389 are not responsible for p65 PI3K -impaired p70S6K activation. A, cells were transfected with cDNA encoding D3Ep70S6K or D3Ep70S6K combined with cDNA encoding rCD2p110 (150 kDa, p110*), p85 plus p110, or p65 PI3K plus p110 (left panels). The cells were processed as in Fig. 3A, and extracts were resolved and examined by Western blot using anti-p85 or anti-p70S6K Ab. Parallel samples were labeled with [ 35 S]Met, and the samples were immunoprecipitated with anti-p110␣ Ab to examine p110 expression. Lysates were also immunoprecipitated with anti-p70S6K Ab, and the kinase activity of immunoprecipitated D3Ep70S6K was examined in vitro. In the right panels, the cells were transfected and processed as in the left panels, replacing D3Ep70S6K cDNA with E389D3Ep70S6K cDNA. B, the mean of five assays performed as in A is represented.

p85-PI3K Controls p70S6K Activation
FRAP efficiently. Expression of p65 PI3K /p110 or p85/p110 in the corresponding samples is illustrated in Fig. 6B (left panel). Similar expression levels of FRAP (Fig. 6B, center panel) and p70S6K (not shown) were also observed. The samples were immunoprecipitated with an anti-Myc antibody, disaggregated, and reprecipitated using an anti-p85 Ab. As shown in Fig. 6B (right panel), p85 efficiently associated with FRAP, whereas p65 PI3K associated poorly with FRAP. p70S6K was also selectively reprecipitated from FRAP immune complexes in the p85 sample (not shown), suggesting a ternary complex of FRAP, p70S6K, and p85 regulated through the p85 C-SH2 domain. As a negative control, we confirmed that the unrelated pp56 lck SH2 domain failed to associate with FRAP or p70S6K (not shown).
To extend these findings, we confirmed the differential efficiency of p65 PI3K and p85 in FRAP⅐p70S6K complex formation by immunoprecipitation and Western blotting. The cells were transfected with HA-tagged FRAP, combined with EE-p70S6K, Myc-wild type-p110, and p65 PI3K or p85. Appropriate expression of the transfected proteins was confirmed by Western blot (Fig. 7A). When p85 and p65 PI3K were immunoprecipitated and the immune complexes were resolved and examined by Western blotting using anti-HA Ab, HA-FRAP was detectable in association with p85 but not with p65 PI3K (Fig. 7B). Similarly, p85 and p65 PI3K immune complexes were resolved and blotted using anti-p70S6K Ab, revealing the presence of p70S6K associated with p85 (Fig. 7B). The more efficient association of p85 than of p65 PI3K with p70S6K was confirmed (Fig. 7C) using cells transfected as in Fig. 6A.
The formation of endogenous p85⅐p70S6K complexes was also examined, although the lack of good quality FRAP-specific antibodies prevented a similar analysis for endogenous FRAP. To analyze endogenous p85⅐p70S6K complexes, wild type NIH-3T3 cells were lysed, and lysates were immunoprecipitated using anti-p70S6K Ab. These immune complexes were resolved in SDS-PAGE and examined by Western blotting with an anti-p85 Ab (Fig. 7D). The data revealed that endogenous p85 associates with p70S6K. This was examined by an alternative method, taking advantage of the p85/p110 lipid kinase activity. Immune complexes of endogenous p70S6K contained associated PI3K activity, confirming that endogenous p85 associates with p70S6K (Fig. 7E).
A FRAP-resistant p70S6K Mutant Is Activated by p65 PI3K / p110 -Considering the relevance of FRAP for p70S6K activa-  2 and 3). Samples were examined by Western blot using anti-p85 Ab. D, NIH-3T3 cells were lysed, and lysates (300 g) were immunoprecipitated with normal rabbit serum (lane 1) or anti-p70S6K Ab (lane 2). Alternatively, lysates (30 g) were immunoprecipitated with anti-p85 Ab (lane 3). Samples were resolved by SDS-PAGE and examined in Western blot using anti-p85 Ab. E, NIH-3T3 cells were lysed, and lysates (300 g) were immunoprecipitated with normal rabbit serum (lane 1) or anti-p70S6K Ab (lane 2). Immunoprecipitates were assayed for associated PI3K activity using PtdIns 4,5-bisphosphate as substrate. The panels show representative results of at least three experiments with similar results.
p85-PI3K Controls p70S6K Activation tion (16), as well as the differential ability of p85 and p65 PI3K to regulate p70S6K activity and associate FRAP, we tested whether a defect in FRAP action or recruitment might explain the failure of p65 PI3K /p110 to activate p70S6K. For this, we used a previously described p70S6K mutant that lacks the Nand C-terminal regions (⌬N54⌬C104-p70S6K); although this mutant responds to mitogens, its activation is FRAP-independent and resistant to the effect of rapamycin (19). The cells were transfected with plasmids encoding ⌬N54⌬C104-p70S6K, active p110, and either p65 PI3K or p85; protein expression was confirmed by Western blotting (Fig. 8A). When samples were analyzed for p70S6K activity, p65 PI3K /active p110 and p85/ active p110 activated ⌬N54⌬C104-p70S6K in a comparable manner (Fig. 8, A and B); this contrasts with the p65 PI3K /active p110 defect in activating wild type p70S6K (Fig. 2A). Similar results were obtained using wt p110 (not shown). These data show that the defective action of FRAP in p65 PI3K /p110-expressing cells is responsible for the inefficient activation of p70S6K in these samples. These results suggest that the complex of p85/p110 with FRAP is required for p70S6K activation and that the C-SH2 region of p85 is essential for formation of this complex.
Inhibition of PP2A Restores p70S6K Activation by p65 PI3K / p110 -Peterson et al. (20) showed that PP2A directly associates and dephosphorylates p70S6K and that FRAP protects p70S6K from PP2A-mediated dephosphorylation. Activation of the ⌬N54⌬C104-p70S6K deletion mutant is FRAP-independent, because this mutant does not bind PP2A (20). Since p65 PI3K /p110 does not activate p70S6K due to a defect in FRAP association (see above), p70S6K may lack FRAP-protection in p65 PI3K cells, rendering it more susceptible to PP2A-mediated dephosphorylation. This would be consistent with the observation of an overall lower p70S6K phosphorylation status in p65 PI3K -expressing cells. If this hypothesis is correct, inhibition of PP2A would be predicted to restore p70S6K activation/phosphorylation in p65 PI3K -expressing cells.
We examined whether PP2A inhibition improves p70S6K activation in p65 PI3K -expressing cells (Fig. 8, C and D) by analyzing endogenous p70S6K activation in the CMN-5 thymoma cell line, which expresses p65 PI3K but not p85␣ (39) and shows defects in endogenous p70S6K activation (Fig. 1A). We used okadaic acid in the 50 -100 nM range to inhibit PP2A without affecting PP1 activity (49,50). IB stability (Fig. 8C), which is reduced upon PP2A inhibition (50), was measured as a control. Preincubation with 100 nM okadaic acid for 1 h moderately affected basal p70S6K activity in CMN-5 and control cells in the absence of serum. The presence of okadaic acid significantly increased both Thr 389 phosphorylation and p70S6K activity in serum-stimulated CMN-5 cells and only moderately affected p70S6K activation in control cells (Fig. 8, C  and D). These results indicate that p70S6K lacks FRAP protection in p65 PI3K cells, rendering p70S6K more susceptible to PP2A-mediated dephosphorylation. These observations also highlight the role of p85 in bringing FRAP into complex with p70S6K, thereby regulating p70S6K activation. DISCUSSION Here we show that p85/p110 PI3K has a dual role in p70S6K activation. In addition to the requirement for PI3K activity in PDK1-dependent p70S6K phosphorylation (13,14), we demonstrate that the p85 regulatory subunit of PI3K contributes to p70S6K stimulation by establishing a multimolecular complex that brings p70S6K into proximity with its activators, PI3K and FRAP. Formation of this complex requires the p85 C-SH2 domain, because the p65 PI3K mutant, which lacks this domain (39), associates moderately with p70S6K and very poorly with FRAP. Complex formation is required for p70S6K activation, which is defective in p65 PI3K /p110-expressing cells. In accordance with this, p70S6K shows a lower degree of phosphorylation in multiple residues in these cells. p65 PI3K /p110 nonetheless activated a FRAP-resistant p70S6K mutant, suggesting that impaired p70S6K activation in these cells is related to the defect in FRAP complex formation. FRAP protects p70S6K from PP2A-mediated dephosphorylation (20). We show that inhibition of PP2A restored p70S6K activation in p65 PI3K cells, further indicating that the C-SH2 deletion mutant defect is related to impaired FRAP action. These data highlight the essential role of the p85 C-SH2 domain in establishing the formation of a ternary complex of p85, p70S6K, and FRAP, required for efficient p70S6K phosphorylation and activation.
For these experiments, we used a mutant form of the PI3K regulatory subunit, p65 PI3K , which lacks part of the p85 inter- Myc-rCD2p110 (p110*) and either p65 PI3K or p85. The cells were incubated and lysed as described in the legend to Fig. 2A. Lysates (40 g/sample) were resolved in SDS-PAGE, and rCD2p110, p85/p65 PI3K , or ⌬N54⌬C104-p70S6K expression levels examined in Western blot using anti-CD2, anti-p85, or polyclonal anti-p70S6K Ab, respectively. Cell extracts (40 g/sample) were also incubated with anti-p70S6K Ab, and immunoprecipitates were tested for p70S6K kinase activity. B, p70S6K activity in the transfected samples compared with that of control cells. Shown are the means of four assays performed as in A. C, CMN-5 and T14 cells were serum-starved in medium for 4 h and then incubated with vehicle, okadaic acid (100 nM), serum (10%), or a combination of okadaic acid (100 nM) and serum (10%). The cells were lysed, and lysates (40 g/sample) were examined by Western blot using anti-IB, Ab anti-p70S6K, or anti-phospho-Thr 389 Ab. Other samples (300 g/ sample) were immunoprecipitated using anti-p70S6K Ab and examined in in vitro kinase assay. D, p70S6K activity in the different samples (indicated) compared with that of control cells. Shown are the means of four assays performed as in C.
SH2 and C-SH2 domains (39). p65 PI3K /p110 exhibits higher in vivo lipid kinase activity than p85/p110 and enhances receptorstimulated PI3K activation (39). The increase in 3-phosphorylated lipids in p65 PI3K -expressing cells appears to be sufficient for activation of the downstream effector AKT (39) but insufficient to activate p70S6K. These results suggest that PI3K contributes differently to the activation of its effectors AKT and p70S6K. Other activation requirements have been shown to differ for these proteins. AKT activation requires phosphorylation of Thr 308 and Ser 473 , which are homologous to p70S6K residues Thr 229 and Thr 389 . Whereas AKT T308 and p70S6K Thr 229 are both phosphorylated by PDK1 (13,14,38,51), there are differences in the phosphorylation mechanisms of AKT Ser 473 and p70S6K Thr 389 . AKT Ser 473 is phosphorylated by PDK1 complexed to a region of protein kinase C-related kinase-2, called the PDK1-interacting fragment, whereas phosphorylation of Thr 389 of p70S6K is mediated by NEK6/7 (12,15,52). Conus et al. (53) have also shown that depletion of calcium from intracellular stores blocks p70S6K activation, whereas the AKT response is unaffected, highlighting the differential activation requirements of these two proteins. The present report suggests that additional differences in the activation mechanism of p70S6K and AKT reside in the contribution of the p85 C-terminal region.
Several lines of evidence suggest that p70S6K is downstream of AKT. Although constitutively active membrane-targeted forms of AKT activate p70S6K, AKT does not appear to phosphorylate p70S6K directly (13). The finding that p65 PI3K /p110 activates AKT and not p70S6K shows that activation of AKT is not sufficient for p70S6K activation. It has also been suggested that AKT regulates FRAP function (54), and the data presented here do not refute this possibility. Nonetheless, the observation that kinase-dead mutants of AKT do not inhibit FRAP-dependent p70S6K activation by growth factors (36,47) does question the contribution of AKT in the process.
To study the mechanism by which the p85 C-terminal region affects p70S6K activation, we examined whether impaired p70S6K activation in p65 PI3K /p110-expressing cells correlated with an altered p70S6K phosphorylation pattern. Acidic substitutions of residues in the pseudosubstrate domain and in Thr 389 did not compensate the activation defect of p70S6K in p65 PI3K /p110-expressing cells. This suggests that defects other than phosphorylation of the pseudosubstrate region or Thr 389 contribute to impairment of p70S6K activation in these cells. Similarly, although a significant reduction in overall p70S6K phosphorylation was found in p65 PI3K /active p110-expressing cells, phosphotryptic peptide maps of p70S6K immunoprecipitated from p85/p110 or from p65 PI3K /p110-expressing cells did not reveal a specific phosphorylation site defect. The lower overall phosphorylation may result from an inefficient early step in p70S6K activation or, alternatively, may reflect that p70S6K is dephosphorylated in p65 PI3K -expressing cells.
The observations presented demonstrate that the complex of p85, p70S6K, and FRAP plays an essential role in p70S6K activation. Romanelli et al. (48) reported that p70S6K associates with PKC and PDK1 and that formation of this complex is essential for PI3K-dependent p70S6K activation but not for PMA-induced p70S6K activation (48,55). Although atypical PKCs are PI3K effectors and modulate p70S6K activity (48,(55)(56)(57), it is unlikely that the inability of p65 PI3K /p110 to activate p70S6K depends on an atypical PKC, because constitutively active mutants of this protein failed to rescue p70S6K activation in p65 PI3K transfectants (Fig. 5). In contrast, differential association was found of p85 and p65 PI3K with p70S6K and even more clearly with FRAP. Whereas p85 associated efficiently with both p70S6K and FRAP, the association of these proteins with the mutant p65 PI3K form was significantly weaker and, in the case of FRAP, could be detected only by reprecipitation techniques after prolonged exposure of the gels. We propose that p85 association with p70S6K and FRAP involves the C-SH2 domain, because this is absent in p65 PI3K . Furthermore, the finding that the defective activation of p70S6K in p65 PI3K cells can be overcome by deletion of the Nand C-terminal domains of p70S6K, conferring FRAP sensitivity, shows that the defective association with FRAP is responsible for the impaired p70S6K activation in p65 PI3K cells. This observation underscores the role of the FRAP⅐p85 complex in regulating the endogenous p70S6K enzyme and supports the hypothesis that p85 links FRAP to p70S6K.
The use of the immunosuppressant rapamycin has positioned FRAP upstream of p70S6K in the signaling cascade. FRAP has been shown to phosphorylate Thr 389 in vitro, but this observation has not been reproduced in vivo (17,18). Indeed, the finding that the N-and C-terminally truncated p70S6K mutant is phosphorylated in vivo on Thr 389 in the presence of rapamycin (19) suggests that FRAP must regulate p70S6K at a different stage. Concurring with this, whereas deletion of the regions that confer FRAP sensitivity restored p65 PI3K /p110 activation of p70S6K, acidic substitution at p70S6K residue Thr 389 was unable to rescue p70S6K activation. Belham et al. (15) recently provided evidence suggesting that NEK6 and NEK7 are the Thr 389 kinases. A distinct function for FRAP, consistent with its ability to regulate phosphorylation of multiple p70S6K residues (19), was recently proposed by Peterson et al. (20). They show that PP2A associates directly and dephosphorylates p70S6K and that FRAP inhibits PP2A-mediated p70S6K dephosphorylation. Moreover, they show that activation of the ⌬N54⌬C104-p70S6K deletion mutant is FRAP-independent, because this p70S6K mutant does not bind PP2A (20). p65 PI3K /p110 activates the ⌬N54⌬C104-p70S6K mutant that does not bind PP2A (Fig. 8A). In addition, PP2A inhibition restored p70S6K activation in p65 PI3K -expressing cells (Fig. 8C). These observations demonstrate that the deficient phosphorylation and activation of p70S6K in p65 PI3Kexpressing cells is related to inefficient PI3K⅐p70S6K⅐FRAP complex formation by p65 PI3K , which results in dephosphorylation of p70S6K by PP2A.
Our results support a specific role for the p85 C-SH2 domain in mediating complex formation with p70S6K and FRAP. It is nevertheless possible that the p65 PI3K N-SH2 domain partially compensates for the absence of C-SH2, accounting for the moderate association of p65 PI3K with p70S6K and the very low association with FRAP. The differential specificity of the two SH2 domains of p85 (25,32) offers an explanation for the greater efficiency of the C-SH2 compared with the N-SH2 in mediating complex formation. Despite the defective p70S6K activation in p65 PI3K cells, these cells grow efficiently. Activation of the other p70S6K isoform (p70S6K-II) (7,58,59), which is less dependent on PI3K and FRAP (58), or activation of p90 rsk (60) may compensate for the p70S6K-I activation defect in these cells.
Our results highlight the importance of the PI3K regulatory subunit in stabilizing a protein complex required for p70S6K activation. This complex includes p70S6K, PI3K, FRAP, and probably PDK1 and PKC, bringing p70S6K into the proximity of its activators and thereby mediating p70S6K activation. The studies presented here show that PI3K exerts two independent actions on p70S6K activation. The first, described previously, is dependent on PI3K enzymatic activity implicated in phosphorylation of Thr 229 and Thr 389 residues (9 -12); the other, described here, depends on the p85-PI3K regulatory subunit that mediates the formation of a multimolecular complex involving p85, FRAP, and p70S6K. This complex is required for efficient p70S6K activation.