Evidence That 3-Phosphoinositide-dependent Protein Kinase-1 Mediates Phosphorylation of p70 S6 Kinase in Vivoat Thr-412 as well as Thr-252*

Protein kinase B and p70 S6 kinase are members of the cyclic AMP-dependent/cyclic GMP-dependent/protein kinase C subfamily of protein kinases and are activated by a phosphatidylinositol 3-kinase-dependent pathway when cells are stimulated with insulin or growth factors. Both of these kinases are activated in cells by phosphorylation of a conserved residue in the kinase domain (Thr-308 of protein kinase B (PKB) and Thr-252 of p70 S6 kinase) and another conserved residue located C-terminal to the kinase domain (Ser-473 of PKB and Thr-412 of p70 S6 kinase). Thr-308 of PKBα and Thr-252 of p70 S6 kinase are phosphorylated by 3-phosphoinositide-dependent protein kinase-1 (PDK1)in vitro. Recent work has shown that PDK1 interacts with a region of protein kinase C-related kinase-2, termed the PDK1 interacting fragment (PIF). Interaction with PIF converts PDK1 from a form that phosphorylates PKB at Thr-308 alone to a species capable of phosphorylating Ser-473 as well as Thr-308. This suggests that PDK1 may be the enzyme that phosphorylates both residues in vivo. Here we demonstrate that PDK1 is capable of phosphorylating p70 S6 kinase at Thr-412 in vitro. We study the effect of PIF on the ability of PDK1 to phosphorylate p70 S6 kinase. Surprisingly, we find that PDK1 bound to PIF is no longer able to interact with or phosphorylate p70 S6 kinase in vitro at either Thr-252 or Thr-412. The expression of PIF in cells prevents insulin-like growth factor 1 from inducing the activation of the p70 S6 kinase and its phosphorylation at Thr-412. Overexpression of PDK1 in cells induces the phosphorylation of p70 S6 kinase at Thr-412 in unstimulated cells, and a catalytically inactive mutant of PDK1 prevents the phosphorylation of p70 S6K at Thr-412 in insulin-like growth factor 1-stimulated cells. These observations indicate that PDK1 regulates the activation of p70 S6 kinase and provides evidence that PDK1 mediates the phosphorylation of p70 S6 kinase at Thr-412.

p70 S6 kinase (p70 S6K) 1 is activated by insulin and growth factors and mediates the phosphorylation of the 40 S ribosomal protein S6 (1). This enables efficient translation of mRNA molecules containing a polypyrimidine tract at their 5Ј-transcriptional start sites (2). p70 S6K also phosphorylates unknown proteins to permit progression through the G 1 phase of the cell cycle (3). p70 S6K is activated by insulin and growth factors, through a phosphoinositide 3-kinase-dependent pathway and becomes phosphorylated on at least 7 Ser/Thr residues in response to these agonists. The phosphorylation of two of these residues namely Thr-252 and Thr-412 on the longer splice variant of the ␣-isoform (Thr-229 and Thr-389 on the shorter splice variant) appears to make the most important contribution to the activation of p70 S6K (4 -6). Phosphorylation of Thr-252 alone or mutation of Thr-412 to glutamic acid to mimic phosphorylation of this residue results in a small activation of p70 S6K. In contrast, phosphorylation of both residues or phosphorylation of Thr-252 in the T412E mutant of p70 S6K results in a large activation of expressed p70 S6K, showing that phosphorylation of Thr-252 and Thr-412 leads to a synergistic activation p70 S6K (7,8).
p70 S6K is a member of the AGC subfamily of protein kinases, which include protein kinase B (PKB) (9), protein kinase C (PKC) isoforms (10), and serum-and glucocorticoid-regulated protein kinase (11). The residues surrounding Thr-252 and Thr-412 of p70 S6K are highly conserved in all AGC family members, and phosphorylation of the residues equivalent to Thr-252 and Thr-412 of p70 S6K is necessary for activation and/or stability of these kinases in vivo (12). Thr-252 is located between subdomains VII and VIII of the kinase domain, a region whose phosphorylation activates many kinases. The residue equivalent to Thr-252 lies in a Thr-Phe-Cys-Gly-Thr-Xaa-Glu-Tyr consensus motif (where the underlined Thr corresponds to Thr-252 and Xaa is a variable residue). Thr-412 is located C-terminal to the catalytic domain, and the residues surrounding Thr-412 lie in a Phe-Xaa-Xaa-Phe-Ser/Thr-Phe/ Tyr consensus motif. We (7) and others (8) have demonstrated that 3-phosphoinositide-dependent protein kinase-1 (PDK1) (13) can phosphorylate p70 S6K at Thr-252 in vitro and in transfection experiments. Phosphorylation of p70 S6K by PDK1 in vitro is independent of the presence of inositol 3, 4, 5 trisphosphate, and activation is increased greatly if the noncatalytic C-terminal tail of p70 S6K is deleted and if Thr-412 is mutated to an acidic residue. Other studies have shown that PDK1 phosphorylates the residue equivalent to Thr-252 of p70 S6K in PKB isoforms (14,15), PKC isoforms (16 -18), serum-and glucocorticoid-regulated protein kinase (19,20), and cAMP-dependent protein kinase (21).
Recently, we made the surprising observation that PDK1 can be converted from a form that phosphorylates Thr-308 of PKB alone (the residue equivalent to Thr-252 in p70 S6K) to a form that phosphorylates both Thr-308 and Ser-473 (the residue equivalent to Thr-412 in p70 S6K) through interaction with a region of protein kinase C-related kinase-2 (PRK2), termed the PDK1 interacting fragment (PIF) (22). In this study, we sought to identify the effect that PIF had on the ability of PDK1 to phosphorylate and activate p70 S6K. Surprisingly, we find that, in contrast to PKB, PDK1 complexed to PIF can no longer phosphorylate p70 S6K in vitro. We also report that expression of PIF or a catalytically inactive mutant of PDK1 in cells prevents the activation of p70 S6K and its phosphorylation at Thr-412 as well as Thr-252, suggesting that PDK1 may mediate the phosphorylation of both residues in vivo.

EXPERIMENTAL PROCEDURES
Materials-The peptides used to assay PKB␣ (RPRAATF) (23) and p70 S6K (GRPRTSSFAEG) (24) and the peptides used to raise and purify the T412-P antibody were synthesized by Dr. G. Blomberg (University of Bristol, United Kingdom). Protein G-Sepharose, glutathione-Sepharose, and CHX-Sepharose were purchased from Amersham Pharmacia Biotech; protease inhibitor tablets were from Roche Molecular Biochemicals, tissue culture reagents, IGF1, and microcystin-LR were from Life Technologies, Inc.; sensorChips CM5 and SA were from Bia-Core AB; biotinylated reagent and secondary antibodies coupled to horse radish peroxidase were from Pierce.
Antibodies-The phospho-specific antibody recognizing p70 S6K phosphorylated at Thr-412 was raised in sheep against the peptide SESANQVFLGFTYVAPSV (corresponding to residues 401-418 of the longer splice variant of human ␣-isoform of p70 S6K), in which the underlined residue is phosphothreonine. The antibody was affinity purified on CH-Sepharose covalently coupled to the phosphorylated peptide. The antibodies were then passed through a column coupled to the nonphosphorylated peptide, and the antibodies that did not bind to this column were selected. Monoclonal antibodies recognizing the HA or Myc epitope were purchased from Roche Molecular Biochemicals; the monoclonal antibody recognizing GST was purchased from Sigma and used to verify the level of expression of GST-PIF in cells, whereas rabbit polyclonal antibodies recognizing the 18 C-terminal residues of PRK2/ PIF were purchased from Santa Cruz Biotechnology.
Preparation of Insect Cell His-p70 S6K-p70 S6K with a His epitope tag at its N terminus lacking the C-terminal 104 residues is termed p70 S6K-T2. To prepare wild type and the mutant 412E-p70 S6K-T2 the cDNA for these constructs was amplified by polymerase chain reaction from the pMT2 vector encoding these forms of p70 S6K (6) using the following oligonucleotides: 5Ј-AGG ATC CAC CAT GCA CCA TCA CCA TCA CCA TAT GAG GCG AGC AAG GAG GCG G-3Ј and 5Ј-GCG GCC GCT CAA CTT TCA AGT ACA GAT GGA GCC-3Ј. The polymerase chain reaction products were then subcloned into the BamHI/NotI sites of the pFASTBAC 1 vector, and this vector was used to generate recombinant baculovirus using the Bac-to-Bac system (Life Technologies, Inc.). The resulting viruses encoded p70 S6K-T2 or 412E-p70 S6K-T2 with an N-terminal hexahistidine sequence and was used to infect Sf21 cells (1.5 ϫ 106/ml) at a multiplicity of infection of 5. The infected cells were harvested 72 h post-infection, and the His-p70 S6K proteins purified by Ni 2ϩ /nitrilotriacetic acid-agarose chromatography as described previously for PKB␤ (25). They were then dialyzed against 50 mM Tris/HCl, pH 7.5, 0.1 mM EGTA, 0.27 M sucrose, 0.03% (by volume) Brij-35, 0.1% (by volume) 2-mercaptoethanol, 1 mM benzamidine, and 0.2 mM phenylmethylsulphonyl fluoride, snap frozen in aliquots, and stored at Ϫ80°C. p70 S6K-T2 or 412E p70 S6K-T2 were both recovered with a yield of 60 mg/liter infected Sf21 cells and were Ͼ90% homogeneous as judged by polyacrylamide gel electrophoresis followed by Coomassie Blue staining.
Phosphorylation of GST-p70 S6K by PDK1-GST-PIF and GST-D978A-PIF were expressed in human embryonic kidney 293 cells and purified on glutathione-Sepharose, and the very small amount of endogenous PDK1 associated with GST-PIF was removed by immunoprecipitation with a PDK1 antibody (22). Phosphorylation of GST-p70 S6K-T2 by PDK1 was carried out as described previously (7) except that PDK1 was incubated with the indicated concentration of GST-PIF or PIF peptide for 10 min on ice prior to initiation of the assay with magnesium [␥-32 P]ATP. The wild type GST-p70 S6K-T2, and the mutant T252A-GST-p70 S6K-T2, T412A-GST-p70 S6K-T2 proteins were expressed in 293 cells and purified as described previously (7). Wild type and catalytically inactive GST-PDK1 were expressed either in 293 cells (26) or in Escherichia coli (27).
Transient Transfection Experiments-The DNA constructs encoding for the wild type and mutant forms of HA-p70 S6K in the pMT2 vector used in this study have been described previously (6). The constructs encoding wild type HA-PKB␣ (25), wild type Myc-PDK1 (26), and a catalytically inactive mutant of Myc-PDK1 (in which Lys-111 and Asp-223 are both mutated to Ala), termed "kinase-dead" PDK1 (26), were all in the pCMV5 vector. The constructs used to express GST-PIF and the mutant GST-F977A-PIF (22) are in the pEBG2T vector. The empty pEBG2T vector was used to express GST protein in control experiments. DNA constructs used in this study were purified from bacteria using the Qiagen plasmid Mega kit according to the manufacturer's protocol.
293 cells cultured on 10-cm diameter dishes in Dulbecco's modified Eagle's medium containing 10% (by volume) fetal bovine serum were transfected with 2 g of DNA construct encoding either wild type or mutant HA-p70 S6K or HA-PKB␣ and 10 g of DNA construct encoding GST-PIF, GST-F977A-PIF, GST, Myc-PDK1, kinase-dead Myc-PDK1, or the empty pCMV5 vector using a modified calcium phosphate method (28). 24 h posttransfection the cells were deprived of serum for 16 h and exposed to IGF1 (100 nM) for the time indicated. The cells were lysed in 1 ml of lysis buffer (50 mM Tris/HCl, pH 7.5, 1 mM EDTA, 1 mM EGTA, 1% (by volume) Triton X-100, 1 mM sodium orthovanadate, 10 mM sodium ␤-glycerophosphate, 50 mM NaF, 5 mM sodium pyrophosphate, 1 M microcystin-LR, 0.27 M sucrose, and protease inhibitor tablets) and cleared by centrifugation, and 50 g of protein was subjected to immunoprecipitation with anti-HA monoclonal antibody. The protein concentrations of the lysates were determined by the Bradford method.
Kinase Assays-The HA-p70 S6K or HA-PKB␣ immunoprecipitates were washed and assayed for kinase activity using the peptide Crosstide (GRPRTSSFAEG) as described previously for PKB␣ (28). One unit of activity was the amount that catalyzed the phosphorylation of 1 nmol of substrate in 1 min.
Immunoblotting for Dephosphorylated and Thr-412-phosphorylated p70 S6 Kinase-Cell lysates were made with 1% SDS, and the indicated amounts of protein were subjected to SDS-polyacrylamide gel electrophoresis, subsequently transferred to nitrocellulose, and immunoblotted using the indicated monoclonal antibody or the T412-P phosphospecific antibody (0. 4 g/ml) in 50 mM Tris/HCl, pH 7.5, 0.15 M NaCl, 0.5% (by volume) Tween, and 10% (by mass) skimmed milk. Detection was performed using the enhanced chemiluminescence reagent (Amersham Pharmacia Biotech).
Surface Plasmon Resonance Measurements of PDK1 Binding to p70 S6 Kinase-p70 S6K-T2 and T412E p70 S6K-T2 mutant were aminecoupled to a CM5 sensor chip (BIAcore AB) according to the manufacturer's instructions. The indicated concentrations of His-PDK1 were injected over the chip at a flow rate of 30 l/min, and the steady-state binding was determined in the presence or absence of PIF peptide. The apparent equilibrium dissociation constant (K d ) for the binding of His-PDK1 to p70 S6 kinase was determined by fitting the increase in steady-state binding upon increasing PDK1 concentration to a rectangular hyperbola using SigmaPlot 4 (SPSS Inc.). The measure of response in our experiments is termed RU; 1000 RU ϭ 1 ng/mm 2 of protein bound to the surface.

Phosphorylation of p70 S6K by PDK1 Is Inhibited by PIF-
PDK1 binds with submicromolar affinity to a region of PRK2 termed PIF (22). PIF is situated C-terminal to the kinase domain of PRK2, and the binding of this region of PRK2 to PDK1 is mediated by a consensus motif similar to that encompassing Thr-412 of p70 S6K, except that the residue at this position is Asp (Asp-978), rather than Thr or Ser. In Fig. 1, we demonstrate that PDK1, when complexed to either GST-PIF or a 24 residue synthetic peptide whose sequence encompasses the PDK1 binding site on PIF, was unable to phosphorylate GST-p70 S6K-T2 (a deletion mutant of p70 S6K that lacks the C-terminal 104 residues) in vitro. In a parallel experiment it was verified that PDK1 complexed to GST-PIF or the PIF peptide was able to phosphorylate PKB at both Thr-308 and Ser-473 to near stoichiometric levels (data not shown) as reported previously (22). GST-p70 S6K-T2 was used as a PDK1 substrate ( Fig. 1) rather than the full-length p70 S6K, which is very poorly phosphorylated by PDK1 in vitro (7,8). Truncation of the C-terminal 104 residues of p70 S6K is likely to be benign, as p70 S6K-T2 2 when expressed in cells possesses indistinguishable properties to the full-length protein as it is still activated by insulin and growth factors in a rapamycin-and wortmannin-sensitive manner (5,6).
A mutant form of GST-PIF or the 24-residue PIF peptide in which the amino acid equivalent to Asp-978 in PRK2 is mutated to Ala (GST-D978A-PIF) possesses markedly reduced affinity for PDK1 (22). Consistent with this, GST-D978A-PIF or the mutant D978A-PIF peptide poorly inhibited the phosphorylation of GST-p70 S6K-T2 by PDK1 (Fig. 1). If Asp-978 in the PIF peptide is mutated to a phosphoserine residue instead of an Ala, to restore the negative charge, the resulting peptide interacted with PDK1 with the same affinity as the wild type PIF peptide (22) and prevented PDK1 from phosphorylating GST-p70 S6K-T2 (Fig. 1).
PDK1 Phosphorylates p70 S6 Kinase at Thr-412 in Vitro-To determine whether PDK1 could phosphorylate p70 S6K at Thr-412, we raised phospho-specific antibodies that only recognize p70 S6K phosphorylated at Thr-412 (termed T412-P antibody). This antibody did not recognize GST-p70 S6K-T2 that had been incubated with magnesium ATP in the absence of PDK1. However, following the addition of PDK1, which had either been expressed in 293 cells or bacteria, the GST-p70 S6K-T2 became recognized by the T412-P antibody (Fig. 2). Incubation of the T412-P antibody with the phosphorylated Thr-412 peptide immunogen used to raise the antibody (but not with the dephosphorylated peptide) abolished its recognition of GST-p70 S6K-T2 (see Fig. 3). The rate at which PDK1 phosphorylated Thr-412 (as well as Thr-252) of GST-p70 S6K-T2 was not increased in the presence of lipid vesicles containing phosphatidylinositol 3,4,5-trisphosphate (data not shown). PDK1 phosphorylated the T252A mutant of GST-p70 S6K-T2 at Thr-412 to the same extent as the wild type GST-p70 S6K-T2. The T412A mutant of GST-p70 S6K-T2 was not recognized by the T412-P antibody after incubation with PDK1/magnesium ATP. The 24-residue PIF peptide prevented PDK1 from phosphorylating the p70 S6K at Thr-412. A kinase-dead mutant of PDK1 was unable to phosphorylate GST-p70 S6K-T2 at Thr-412 (Fig. 2).
The rate at which PDK1 phosphorylates Thr-412 is likely to be significantly lower than that at which it phosphorylates Thr-252. 32 P-labeled GST-p70 S6K-T2 phosphorylated with PDK1 was digested with either trypsin or V8 protease and then subjected to peptide map analysis on HPLC as described previously (7). This analysis revealed that the major 32 P-labeled peptide containing ϳ20 -30% of the total radioactivity applied to the HPLC column corresponded to the peptide phosphorylated at Thr-252. Although several minor peptides were present in this analysis, which each comprised Ͻ5% of the total phosphate, we were unable to attribute any of these to a peptide phosphorylated at Thr-412. This analysis does not exclude the possibility that the recovery of the 32 P-labeled peptide phosphorylated at Thr-412 may be poor but suggests that the stoichiometry at which PDK1 phosphorylated p70 S6K at Thr-412 is much lower than that which it phosphorylates Thr-252.
PIF Inhibits IGF1-induced Activation of p70 S6K-To determine whether expression of PIF in cells could prevent the endogenous PDK1 from phosphorylating and activating p70 S6K, HA-tagged full-length p70 S6K (HA-p70 S6K) was transfected into 293 cells together with constructs encoding either GST-PIF, a mutant form of GST-PIF that interacts with PDK1 weakly (GST-F977A-PIF), or GST itself. The wild type or mutant GST-PIF and GST itself were all expressed at a similar level and were present at a much higher concentration than the endogenous PDK1 or PRK2 (data not shown). The cells were subsequently stimulated with IGF1 for 40 min (the time at which HA-p70 S6K is maximally activated) (data not shown); the cells were lysed, and the HA-p70 S6K was immunoprecipitated and assayed. Cells expressing HA-p70 S6K and GST exhibited a readily measurable basal p70 S6K activity in unstimulated cells, which was increased 10-fold in response to IGF1 (Fig. 3A). In contrast, cells expressing HA-p70 S6K and GST-PIF possessed a basal HA-p70 S6K activity that was virtually undetectable, and IGF1-stimulation caused only a very slight increase in the HA p70 S6K activity (Fig. 3A). In cells expressing HA-p70 S6K and GST-F977A-PIF, HA-p70 S6K was substantially activated by IGF1, although not to the same extent as in cells expressing HA-p70 S6K and GST (Fig. 3A). This is probably explained by a weak interaction of GST-F977A-PIF with PDK1.
PIF Inhibits IGF1-induced Phosphorylation of p70 S6K at Thr-412-As PIF inhibited p70 S6K activation in cells, we sought to determine the effect of PIF expression on the phosphorylation of p70 S6K at Thr-412 and Thr-252. We used the T412-P antibody to measure the phosphorylation of p70 S6K at Thr-412. These experiments showed that IGF1 triggered the phosphorylation of Thr-412 (Fig. 3A). This was abolished by incubation of the T412-P antibody with the phosphorylated The only other 32 P-labeled protein on the gel, which is not shown, corresponded to autophosphorylation of PDK1 and contained ϳ10-fold lower levels of 32 P radioactivity than that of the GST-p70 S6T2 phosphorylated by PDK1 in the absence of PIF. A high amount of PDK1 is used in this experiment to achieve a near maximal phosphorylation of GST-p70 S6T2. If the reactions were carried out at a 10-fold lower concentration of PDK1 under conditions where the phosphorylation of GST-p70 S6T2 by PDK1 is linear with time and the amount of substrate used, then PIF still prevented the phosphorylation of GST-p70 S6KT2 (data not shown). wt indicates wild type, and PSer indicates phosphoserine. The results of a duplicate experiment for each condition are shown, and similar findings were obtained in five separate experiments.
Thr-412 peptide immunogen used to raise the antibody but not with the dephosphorylated peptide (Fig. 3A) or a phosphopeptide corresponding to the sequence surrounding Thr-252 (data not shown). Furthermore, a mutant form of HA-p70 S6K in which Thr-412 was changed to an Ala was not recognized by the T412-P antibody following IGF1 stimulation (Fig. 5C).
When HA-p70 S6K was coexpressed in cells with GST-PIF, IGF1 failed to induce the phosphorylation of HA-p70 S6K at Thr-412 (Fig. 3A). In contrast, in cells expressing HA-p70 S6K and the mutant GST-F977A-PIF, the phosphorylation of HA-p70 S6K still occurred but at a lower level than that observed in cells expressing HA-p70 S6K and GST. The decrease in Thr-412 phosphorylation is consistent with the reduced activation of HA-p70 S6K in these cells compared with those expressing GST alone (Fig. 3A). It should be noted, however, that cotransfection of HA p70 S6K with the GST-F977A-PIF mutant induced a 50% maximal activation of P70 S6K, despite inducing a significantly greater reduction in the level of phosphorylation of Thr-412 (Fig. 3A). This finding demonstrates that the relationship between p70 phosphorylation at Thr-412 and level of p70 S6K activity does not appear to be linear. One explanation for this observation is that the F977A-PIF mutant may inhibit more potently p70 S6K phosphorylation at Thr-412 than Thr-252; however, thus far we have not been able to raise phospho-specific antibodies recognizing p70 S6K phosphorylated at Thr-252 to explore this possibility. The overexpression of GST-PIF in cells also abolished the IGF1-induced activation and phosphorylation at Thr-412 of the p70 S6K-T2 mutant, which lacks the C-terminal 104 residues (data not shown).
PIF Inhibits IGF1-induced Phosphorylation of p70 S6K at Thr-252-A mutant form of HA-p70 S6K in which Thr-412 was altered to glutamic acid to mimic the presence of a phosphorylated residue at this position possessed an elevated basal activity that was further activated by IGF1 when co-expressed with GST or the mutant GST-F977A-PIF (Fig. 3B). Previous work has established that the basal and IGF1-stimulated activity of this mutant are mediated through phosphorylation of Thr-252 (6). In Fig. 3B, we demonstrate that co-expression of HA-412E p70 S6K with PIF greatly reduced the basal activity of this mutant and largely prevented its activation by IGF1. This suggests that PIF also inhibits the phosphorylation of p70 S6K at Thr-252. The overexpression of PIF in cells also greatly reduced the basal and IGF1-stimulated activity of T412E p70 S6K-T2 mutant in cells (data not shown).
PIF Does Not Inhibit the Activation of PKB␣ or Its Phosphorylation at Ser-473-Previous work has shown that PIF does not prevent PDK1 from phosphorylating PKB in the presence of 3-phosphoinositide lipids but instead enables PDK1 to phosphorylate PKB at both Thr-308 and Ser-473 (see the introduction). Here we show that in marked contrast to the effect of GST-PIF on p70 S6K activation, expression of GST-PIF in 293 cells does not prevent IGF1 from inducing a ϳ20-fold activation of HA-PKB␣. This activation is similar to that observed when HA-PKB␣ is coexpressed with GST (Fig. 4). Expression of GST-PIF did not inhibit or potentiate the IGF1-induced phosphorylation of HA-PKB␣ at Ser-473 (the residue equivalent to Thr-412 in p70 S6K) (Fig. 4). GST-PIF is expressed at a similar level when cotransfected with PKB and HA-p70 S6K (data not shown), indicating that the inability of PIF to affect the activation of PKB in cells is not because of it being expressed at a low level.
A Catalytically Inactive Mutant of PDK1 Prevents the Activation and Phosphorylation of p70 S6K-Consistent with earlier findings (7,8), co-expression of HA-p70 S6K with wild type PDK1 induced a large activation of HA p70 S6K, which was not FIG. 2. PDK1 phosphorylates p70 S6K at Thr-412 in vitro and this is inhibited by PIF. 0.5 g of wild type GST-p70 S6K-T2 (wt), T252A-GST-p70 S6K-T2 (252A), or T412A-GST-p70 S6K-T2 (412A) were incubated for 90 min at 30°C with magnesium ATP in the presence or absence of wild type (wt) or kinase-dead (kd) GST-PDK1 expressed in either 293 cells or bacteria in the presence (ϩ) or absence (Ϫ) of the wild type PIF peptides (4 M) in a final volume of 20 l. The reactions were terminated by making the solutions 1% in SDS; the samples were subjected to SDS-polyacrylamide gel electrophoresis, and the phosphorylation of p70 S6K at Thr-412 was assessed by immunoblotting with the T-412P antibody. Similar results were obtained in three separate experiments.

FIG. 3. PIF inhibits p70 S6K activation and phosphorylation at
Thr-252 and Thr-412. 293 cells were co-transfected with constructs expressing the wild type (wt) full-length HA-p70 S6K (A) or the fulllength HA-T412E p70 S6K (B) with either GST-PIF, GST-F977A-PIF, or GST. 24 h posttransfection the cells were serum-starved for 16 h and then stimulated for 40 min with 100 nM IGF1. The cells were lysed and HA-p70 S6K was immunoprecipitated and assayed as described under "Experimental Procedures." Protein from each lysate (10 g for the HA blots or 20 g for the T412-P blot) was electrophoresed on a 10% SDS/polyacrylamide gel and immunoblotted using HA-antibody or the T412-P antibody. The T412-P antibody was incubated with either the synthetic peptide (10 g/ml) corresponding to residues 401 to 418 of p70 S6K phosphorylated at Thr-412 (phospho-412E peptide) or the unphosphorylated peptide (dephospho-Thr-412 peptide). The T412-P antibody consistently recognizes a protein termed "nonspecific band" in cell lysates, which migrates at (75 kDa) derived from nontransfected and transfected cells. The intensity of this band does not change with IGF1. It is not co-immunoprecipitated with HA-p70 S6K (data not shown). The HA-p70 S6K activities shown are the average Ϯ S.E. for a single experiment carried out in triplicate. Similar result were obtained in eight separate experiments (A) and two experiments (B).
increased further by IGF1 stimulation (Fig. 5A). We consistently observed a slight decrease in HA-p70 S6K activity in cells overexpressing PDK1 following IGF1 stimulation. The co-expression of wild type PDK1 with HA-p70 S6K or T252A-p70 S6K also resulted in a large increase in Thr-412 phosphorylation in unstimulated cells (Fig. 5, A and B). In contrast, no immunoreactive band was detected after immunoblotting with the T412-P antibody when wild type PDK1 and the HA-T412A p70 S6K mutant were co-expressed (Fig. 5C). When a kinasedead mutant of PDK1 was co-expressed with HA-p70 S6K, the latter was no longer activated following IGF1-stimulation of cells nor was it phosphorylated at Thr-412 (Fig. 5A). In Fig. 5D we demonstrate that co-expression of HA-412E p70 S6K with a catalytically inactive PDK1 reduced the basal level of HA-412E p70 S6K and largely prevented its activation by IGF1. This provides evidence that overexpression of a kinase-dead PDK1 in cells also inhibits the phosphorylation of p70 S6K at Thr-252.
PIF Prevents the Interaction of PDK1 with p70 S6 Kinase-A recent study by Blenis and colleagues (29) reported that, when PDK1 and p70 S6K were cotransfected into cells, a small amount of PDK1 was coimmunoprecipitated with p70 S6K, suggesting that these proteins may interact directly. Using surface plasmon resonance measurements, we were able to detect a significant interaction (apparent K d 8 M) between PDK1 and p412E-p70 S6K-T2 (Fig. 6). This interaction was abolished in the presence of the 24-residue wild type PIF peptide but not the D978A mutant of the PIF peptide (Fig. 6), suggesting that both the PIF and 412E-p70 S6K-T2 mutant compete for the same binding site on PDK1. In parallel experiments, a significantly weaker interaction between PDK1 and wild type p70 S6K-T2 kinase was detected (Fig. 6).

DISCUSSION
Recent work has shown a high affinity-interaction between PIF and the kinase domain of PDK1, which enhances the rate at which PDK1 phosphorylates PKB␣ and allows it to phosphorylate Ser-473 as well as Thr-308. In this study we have made the surprising observation that PIF prevents PDK1 from phosphorylating p70 S6K (Figs. 1 and 2), and expression of PIF in 293 cells prevents the IGF1-induced activation of p70 S6K (Fig.  3) without affecting the activation of PKB␣ (Fig. 4). Mutant forms of PIF, which interact weakly with PDK1, were much less effective at inhibiting the phosphorylation of p70 S6K by PDK1 in vitro or at inhibiting the IGF1-induced activation of the p70 S6K. These observations could be explained if p70 S6K, but not PKB␣, needed to interact with PDK1 at a site that overlaps with the PIF binding site before p70 S6K can become phosphorylated by PDK1. This conclusion is supported by the findings in Fig. 6 that p70 S6K does interact with PDK1, and this interaction is abolished in the presence of the PIF peptide. The finding that the T412E mutant form of p70 S6K interacts with PDK1 with higher affinity than the wild type enzyme may  (6) (data not shown). Protein from each lysate (10 g for the HA blots or 20 g for the T412-P blot) was electrophoresed on a 10% SDS/polyacrylamide gel and immunoblotted using HA-antibody or the T412-P antibody. The T412-P antibody was incubated in the presence of the dephosphorylated peptide corresponding to residues 401-418 of p70 S6K. The HA-p70 S6K activities shown are the average Ϯ S.E. for a single experiment carried out in triplicate. Similar results were obtained in at least three separate experiments. Comparable results to the HA and T412-P blots shown were also obtained in at least three separate experiments. also explain why the T412E mutant of p70 S6K was observed in previous studies to be a better substrate for PDK1 than the wild type or T412A mutant of p70 S6K (7,8). Phosphorylation of PKB␣ by PDK1 is not inhibited by the presence of PIF, and we could not detect any significant interaction between PKB␣ and PDK1 in vitro by surface plasmon resonance (data not shown). As PKB␣ and PDK1 both interact with 3-phosphoinositides through their pleckstrin homology domains, it is possible that this is the primary determinant for co-localizing these molecules at the plasma membrane and hence allowing PDK1 to phosphorylate PKB␣. In contrast, substrates for PDK1 such as p70 S6K, which do not interact with 3-phosphoinositides, may actually need to interact with PDK1 with relatively high affinity before they can become phosphorylated. Previous evidence that PDK1 is an activator of p70 S6K rested largely on the demonstration that PDK1 phosphorylates and activates p70 S6K in vitro and in cotransfection experiments. The finding in this study that expression of PIF can prevent the activation of p70 S6K in vivo, presumably by binding to PDK1, provides further evidence that PDK1 is required for the activation of p70 S6K in cells.
Interaction with PIF converts PDK1 into a kinase that is capable of phosphorylating both Thr-308 and Ser-473 sites of PKB. This demonstrates that PDK1 has the intrinsic ability to phosphorylate the residue in the T-loop as well as the PDK2 motif of at least one AGC kinase family member. As the residues surrounding Thr-412 of p70 S6K are highly homologous to those surrounding Ser-473 of PKB, it could be argued that PDK1, perhaps in complex with another protein(s), would also possess the intrinsic ability to phosphorylate p70 S6K at Thr-252 and Thr-412. The present study supports this hypothesis because first, overexpression of wild type PDK1 triggers the phosphorylation of p70 S6K at Thr-412 (Fig. 5A). Second, expression with PIF prevents the IGF1-induced phosphorylation of p70 S6K at Thr-412 in cells (Fig. 3A). Third, the PDK1catalyzed phosphorylation of p70 S6K at Thr-412 in vitro was prevented by PIF (Fig. 2). Finally, the overexpression of a kinase-dead mutant of PDK1 in cells not only prevented the activation of p70 S6K, as reported by others (8), but also prevented the phosphorylation of p70 S6K at Thr-412 (Fig. 5). Taken together, the data suggest that PDK1 could phosphorylate p70 S6K at Thr-412 in vivo. As PDK1 phosphorylation of Thr-412 of p70 S6K in vitro is not dependent upon 3-phosphoinositde lipids, it is possible that the sensitivity of PDK1 to these lipids in cells is conferred by the interaction of PDK1 with another protein. In this respect it should be recalled that the interaction of PDK1 with PIF enables PDK1 to be directly activated by 3-phosphoinositides (22). It is also possible that a PDK1-interacting protein(s) could increase the rate at which PDK1 phosphorylates both Thr-252 and Thr-412 of p70 S6K.
It has been recently reported that catalytically inactive mutants of PKC (30) and PKC (29) antagonize the ability of agonists to activate p70 S6K in cells. These observations were interpreted as indicating that PKC/PKC may have a role in activating p70 S6K in cells. However, PKC and PKC are both AGC kinase family members, which are likely to be activated by PDK1 in vivo, and possess an acidic residue rather than Ser/Thr in their PDK2 consensus motif. Furthermore, PKC, like PIF has been shown to interact directly with the kinase domain of PDK1 (16,18). It is therefore possible that both PKC and PKC interact with PDK1 in the same way as PIF and so prevent PDK1 from inducing the activation of p70 S6K. Recent work also implicated PKC in mediating a rapamycinsensitive phosphorylation of the novel PKC isoform (PKC␦) at the residue equivalent to Thr-412 of p70 S6K (31). This study did not, however, rule out the possibility that PDK1 complexed to PKC acquires the ability to phosphorylate PKC␦ at this residue, rather than PKC itself directly phosphorylating this residue. To complicate matters further, it has also recently been shown that conventional PKC␣ is capable of autophosphorylating itself at the residue equivalent to Thr-412 of p70 S6K (18, reviewed in Ref. 33). Sabatini and colleagues (34,35) have reported that mTOR phosphorylates p70 S6K directly at Thr-412. However, much recent evidence suggests that the ability of rapamycin, an inhibitor of the mTOR kinase, to suppress the activity of p70 S6K is mediated primarily through the rapamycin-induced activation of an mTOR-regulated protein phosphatase, which dephosphorylates p70 S6K (32,36,37). It will be important to establish whether mTOR or any other insulinstimulated kinase, which can phosphorylate p70 S6K at Thr-412, is inhibited by PIF. FIG. 6. Quantitative analysis of the binding of PDK1 to p70 S6K. Surface plasmon resonance measurements were carried out on a BiaCore instrument as described under "Experimental Procedures." His-PDK1 was injected at the indicated concentrations over (A) 2000 RUs of p70 S6K-T2 (closed squares) or 412E-p70 S6K-T2 (closed circles), which was immobilized by amine coupling to a CM5 Sensorchip. Experiments carried out in the presence of either 10 M wild type PIF peptide (octagons) or 10 M D978A mutant PIF peptide (triangles) are indicated. The responses at steady state binding were recorded. All data are single determinations from a representative experiment that was repeated at least three times with similar results. The data on the binding of wild type p70 S6K to PDK1 at concentrations above 2 M are not shown, as our analysis of the data suggested that nonspecific protein binding was contributing to part of the observed binding response under these conditions.