Regulation of Kinase Activity of 3-Phosphoinositide-dependent Protein Kinase-1 by Binding to 14-3-3*

3-Phosphoinositide-dependent protein kinase-1 (PDK1) plays a central role in activating the protein kinase A, G, and C subfamily. In particular, PDK1 plays an important role in regulating the Akt survival pathway by phosphorylating Akt on Thr-308. PDK1 kinase activity was thought to be constitutively active; however, recent reports suggested that its activity is regulated by binding to other proteins, such as protein kinase C-related kinase-2 (PRK2), p90 ribosomal protein S6 kinase-2 (RSK2), and heat-shock protein 90 (Hsp90). Here we report that PDK1 binds to 14-3-3 proteins in vivo and in vitrothrough the sequence surrounding Ser-241, a residue that is phosphorylated by itself and is critical for its kinase activity. Mutation of PDK1 to increase its binding to 14-3-3 decreased its kinase activity in vivo. By contrast, mutation of PDK1 to decrease its interaction with 14-3-3 resulted in increased PDK1 kinase activity. Moreover, incubation of wild-type PDK1 with recombinant 14-3-3in vitro decreased its kinase activity. These data indicate that PDK1 kinase activity is negatively regulated by binding to 14-3-3 through the PDK1 autophosphorylation site Ser-241.

originally identified as a kinase that could phosphorylate Akt on its activation loop (residue Thr-308) (4 -6). Later studies, however, have showed that PDK1 is not just an Akt kinase but also a kinase phosphorylating p70 S6K , SGKs, PKC isoforms, and p90 ribosomal protein S6 kinases (RSKs) at the equivalent residues of Thr-308 of Akt (reviewed in Ref. 7). Therefore, PDK1 plays a central role in activating the AGC family of protein kinases. In the case of Akt, the interaction of phosphatidylinositol 3,4,5-trisphosphate with the pleckstrin homology domain recruits Akt to the plasma membrane and promotes conformational change, which results in phosphorylation of Akt at Thr-308 by PDK1 and at Ser-473 by an as yet unidentified kinase (so called PDK2) (7,8). In the case of p70 S6K , SGKs, and PKC isoforms, however, the precise mechanisms for PDK1dependent phosphorylation and activation are not well understood.
PDK1 itself is also a member of the AGC subfamily of protein kinases and is phosphorylated on the Ser-241 activation loop (equivalent to Thr-308 of Akt) (7). As PDK1 expressed in bacteria is active and is phosphorylated at Ser-241 (9), it is thought to phosphorylate itself at this same site. Further, mutation of Ser-241 to Ala was reported to abolish PDK1 kinase activity, and IGF-I stimulation did not cause further activation of PDK1 (9). According to these results, PDK1 was thought to be constitutively active. Several recent reports, however, suggested that PDK1 kinase activity is controlled by PDK1-associating proteins, such as PKC-related kinase-1 (PRK1)/PRK2 (10), PDK1-interacting fragment of PRK2 (11), RSK2 (12), and Hsp90 (13). The interaction of PDK1 with the PDK1-interacting fragment of PRK2 converts PDK1 from an enzyme that phosphorylates Akt only at Thr-308 into a kinase that phosphorylates both Thr-308 and Ser-473 (11). Further, the association of PDK1 with RSK2 stimulates PDK1 activation and autophosphorylation (12). We recently reported that Hsp90 participates in stability and signaling of PDK1 (13). Therefore, PDK1 is not a constitutively active kinase but a kinase regulated by other interacting proteins.
Members of the 14-3-3 protein family are highly conserved and widely expressed 28 -31 kDa proteins that naturally assemble as homodimers or heterodimers. They consists of at least seven isoforms in mammals (␤, ␥, ⑀, , , , , and ). The 14-3-3 proteins have been shown to interact with and to regulate proteins controlling a wide array of signaling pathways, including Raf-1, Bad, FKHRL1, and Cdc25c (reviewed in Ref. 14). Binding of 14-3-3 to its partners depends on phosphorylation of the Ser or Thr residue in the recognition domains. Using peptides derived from Raf-1, Muslin et al. (15) identified that the motif optimal for association with 14-3-3 proteins was RSXpSXP, where pS represents phosphorylated Ser and X represents any amino acid. Moreover, using phosphopeptide libraries, Yaffe et al. (16) and Rittinger et al. (17) revealed that there are two preferred 14-3-3 binding motifs, RSXpSXP and RXXXpSXP, and most of the 14-3-3 partners identified to date contain one of these motifs. It is important to note that several proteins that associate with 14-3-3 proteins do not contain either of these motifs, indicating that 14-3-3 can bind to some variation (e.g. RXXpS in PTPH1 and TH, RX(X)pSXX(X)S in Cbl and PKC). Furthermore, 14-3-3 bound to some proteins and peptides in a phosphorylation-independent manner (14,18).
In the course of searching protein-protein interacting motifs in PDK1, we found four 14-3-3 binding motifs (RXXpS). Thus, we started to examine the interaction between PDK1 and 14-3-3 proteins, and of the five 14-3-3 isoforms we tested, 14-3-3 and were found to form a complex with PDK1. Using sitedirected mutants, we identified the 14-3-3 recognition site in PDK1. Because the association of PDK1 with 14-3-3 reduced PDK1 kinase activity in vitro and in vivo, we concluded that 14-3-3 negatively regulates PDK1 by complex formation.

EXPERIMENTAL PROCEDURES
Reagents-The recombinant human active PDK1 and inactive SGK proteins were obtained from Upstate Biotechnology (Lake Placid, NY).
Cell Culture Conditions-Human embryonic kidney 293T, African green monkey kidney COS-7, human fibrosarcoma HT1080, and mouse fibroblast NIH/3T3 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. In some experiments, cells were precultured for 5 h (for HT1080 cells) or 21 h (for COS-7 cells) in serum-free medium (serum starvation) before serum addition.
The Myc-tagged, human, full-length wild-type PDK1 cDNA (WT-PDK1) in a pCMV3 vector was kindly provided by Drs. P. Hawkins and K. Anderson (The Babraham Institute, Cambridge, UK) (19). The NH 2terminal deleted PDK1 cDNA that encompassed residues 52-556 (⌬N51-PDK1) was generated by reverse transcription-PCR (RT-PCR) with 293T mRNA as the template and then subcloned into a pFLAG-CMV-2 vector (Sigma) (13). Substitutions of Ser at 241, 410, or 549 with Ala (S241A, S410A, or S549A, respectively), Arg at 238 with Glu (R238E), Ala at 239 with Glu or Gln (A239E or A239Q, respectively), or Val at 243 with Pro (V243P) in PDK1 cDNA was accomplished by PCR-based mutagenesis. The double-point mutant was established by further PCR mutagenesis. The wild type human akt1 cDNAs in a pFLAG-CMV-2 vector or a pHM6 vector were established in our laboratory (20). The NH 2 -terminal myristoylated (Myr) active mouse akt1 cDNA in a pUSEamp vector was purchased from Upstate Biotechnology. The NH 2 -terminal-deleted sgk cDNA that encompassed residues 61-431 (⌬N60-SGK) was generated by reverse transcription-PCR (RT-PCR) with Saos-2 mRNA as the template and then subcloned into a pFLAG-CMV-2 vector (Sigma). The wild type human v-raf-1 cDNA in a pCMV vector was purchased from Clontech (Palo Alto, CA). All the plasmid DNAs for transfection were purified using a Qiagen plasmid Maxi kit, according to the manufacturer's protocol (Qiagen, Chatsworth, CA).
Transient Transfection, Immunoprecipitation, and Western Blot Analysis-Cells were transfected with appropriate plasmids using Superfect reagent (Qiagen) or LipofectAMINE 2000 reagent (Invitrogen), according to the manufacturer's instructions.
Purification of Recombinant GST and GST⅐14-3-3 Proteins-Cultures of Escherichia coli containing a pGEX 6P-3 plasmid encoding nothing or 14-3-3 were induced for 2 h with 1 mM isopropyl-␤-Dthiogalactopyranoside at 30°C with shaking. Cells were harvested, and recombinant proteins were purified using GST Purification Modules according to the manufacturer's instructions (Amersham Biosciences).

14-3-3 Proteins Bind to PDK1 In Vivo and
In Vitro-In order to find novel PDK1-binding proteins that regulate PDK1 kinase activity, we searched protein-protein interaction motifs in PDK1. We found that it has four 14-3-3 binding motifs (RXXpS, pS represents phosphorylated Ser, and X represents any amino acid). Among the four ( 3 RTTS 6 , 238 RANS 241 , 407 RSGS 410 , 546 RYQS 549 ), two Ser residues (Ser-241 and Ser-410) have been reported to be phosphorylated in cells (9). First, we confirmed the presence of 14-3-3 binding motifs in PDK1 by immunoblot analysis using an anti-phospho-14-3-3 binding motif antibody that could preferentially recognize the conserved 14-3-3 recognition motif (RXX(S/T)) in which X may be any amino acid) only when Ser or Thr was phosphorylated (14,18). As shown in Fig.  1A, ⌬N51-PDK1, which lacks the NH 2 -terminal 51 amino acids, was recognized by the anti-phospho-14-3-3 binding motif antibody (lane 2). The full-length PDK1 was also recognized by the antibody (data not shown). Raf-1 is well known to form a complex with 14-3-3 (14,18), and we observed that it was also recognized by the antibody (Fig. 1A, lane 4). To date, it has not been reported whether Akt forms a complex with 14-3-3. Fig.  1A shows that Akt was barely recognized in the same analysis (lane 3) although it contains one potential 14-3-3 binding motif (120-RSGpS-124).
To confirm the interacton between PDK1 and 14-3-3 in cells, we transfected Myc-tagged PDK1 cDNA together with HAtagged 14-3-3 isoform cDNAs into 293T cells, following immunoprecipitation of PDK1 with an anti-Myc antibody. Immunoblot analysis revealed that 14-3-3 and , but not 14-3-3␤, , or ⑀, were co-immunoprecipitaetd with PDK1 (Fig. 1B). This result indicates that among the five 14-3-3 isoforms we tested, only 14-3-3 and bind to PDK1 in cells, and 14-3-3 binds to PDK1 more tightly than 14-3-3 does (Fig. 1B). We then examined the binding of 14-3-3, ␤ and to Raf-1 and found that it bound to all the three isoforms with almost the same affinity (Fig. 1C). On the other hand, Akt bound to none of the three isoforms under the same conditions (Fig. 1D), which is in agreement with the data that Akt was not recognized by the antiphospho-14-3-3 binding motif antibody (Fig. 1A, lane 3).
Regulation of PDK1 Kinase Activity by Binding to 14-3-3-Ser-241 of PDK1 was reported to be phosphorylated by itself (9). Phosphorylation of PDK1 at Ser-241 is suggested to be essential for its kinase activity because the conversion of Ser-241 to Ala dramatically reduces PDK1 kinase activity (9). Therefore, it is possible that 14-3-3 regulates PDK1 kinase activity by associating with PDK1 at Ser-241 residue. We then investigated the role of PDK1-14-3-3 complex formation on PDK1 kinase activity. PDK1 mutant cDNAs were co-transfected with WT-akt cDNA into COS-7 cells, and the kinase activity of PDK1 and its mutants were estimated by the amount of the phosphorylated form of Akt at Thr-308 after serum stimulation. Consistent with a previous report (9), S241A exhibited weak phosphorylation activity when compared with WT-PDK1 (Fig. 4A, upper panel, compare lanes 4  and 3). It is important to note that mutating Arg-238 to Glu (R238E) to decrease the interaction between PDK1 and 14-3-3 resulted in a slight increase in PDK1 kinase activity (Fig. 4A, upper panel, lane 5). By contrast, converting Val-243 to Pro (V243P) to increase the association with 14-3-3 reduced that activity (Fig. 4A, upper panel, lane 6). Using HT1080 cells, we obtained almost the same result (data not shown). To exclude the possibility that these changes in kinase activity were due to the changes in the affinity of the mutants to Akt, we investigated the binding of the PDK1 mutants to Akt. We did this by immunoprecipitating PDK1 following immunoblot analysis. Fig. 4B shows that no change in binding ability of R238E and V243P, or in S241A, was observed. Moreover, we saw no change in Akt binding to mutant PDK1 proteins when Akt was immunoprecipitated from the cell lysate (data not shown). Because SGK was also known to be phosphorylated by PDK1 at the Thr-256 residue (7), we estimated the kinase activities of PDK1 point mutants by transfecting ⌬N60-sgk cDNA together with PDK1 mutant cDNAs into COS-7 cells. Immunoblot analysis with an anti-phospho-SGK (Thr-256) antibody revealed that mutation of Ser-241 to Ala in PDK1 abrogated PDK1 activity in phosphorylating ⌬N60-SGK (Fig. 4C, upper panel, lane 4). We also observed that converting Val-243 to Pro (V243P) to increase the association with 14-3-3 decreased the PDK1-dependent phosphorylation of SGK at Thr-256 (Fig. 4C, upper panel, lane 6), compared with WT-PDK1 (lane 3). These results suggest that in cells, 14-3-3 proteins negatively regulate PDK1 kinase activity by binding to PDK1 at the residues surrounding Ser-241, but phosphorylation of PDK1 at Ser-241 positively regulates its activity.
To further confirm the role of 14-3-3 in PDK1 function, we examined the amount of the phosphorylated form of Akt at Thr-308 in COS-7 cells by transfecting the myristoylated form of akt cDNA (Myr-Akt) together with 14-3-3 or 14-3-3 cDNA. Tagging Akt with c-Src-derived residues required for myristoylation could induce the constitutive targeting of Myr-Akt to the plasma membrane in a PI3K-independent manner. As shown in Fig. 4D, transfection of 14-3-3 (lane 2) or 14-3-3 (lane 3) decreased the phospho-Akt (Thr-308) level of transfected Myr-Akt protein. A decrease in the amount of the phosphorylated form of Akt was also observed when WT-Akt was co-transfected with 14-3-3 or 14-3-3 cDNA (data not shown). These results suggest that 14-3-3 proteins suppress PDK1 kinase activity in cells, and the 14-3-3 binding to PDK1 is important for cellular signal transduction.
PDK1 Activity Was Down-regulated by Binding to 14-3-3 In Vitro-To confirm the negative regulation of PDK1 kinase activity by 14-3-3, we incubated recombinant PDK1 with recombinant GST⅐14-3-3 in vitro and estimated the change of PDK1 autophosphorylation activity. After the incubation with GST or GST-fused 14-3-3 at 30°C for 2 h, a condition sufficient to reconstitute specific interaction of PDK1 with 14-3-3 in vitro (Fig. 1F), PDK1 was incubated with phosphorylated or nonphosphorylated RSK2 peptide. Then, the autophosphorylation activity was examined by autoradiography. As shown in Fig.  5A, preincubation with GST⅐14-3-3 down-regulated PDK1 autophosphorylation activity in a dose-dependent manner. As reported previously (12), PDK1 kinase activity was elevated by adding phosphorylated RSK2 peptide (pS386 peptide). Under this condition, GST⅐14-3-3 also decreased PDK1 autophosphorylation activity (Fig. 5A). Similarly, incubation of the immunoprecipitated Myc-tagged WT-PDK1 with GST⅐14-3-3 caused a reduction of PDK1 kinase activity to phosphorylate a PDK1 substrate, SGK (Fig. 5B). In contrast, the kinase activity of R238E-PDK1, which associated less with 14-3-3 (Fig. 3C), was not affected by preincubation with 14-3-3 (Fig. 5B). These results indicate that 14-3-3 negatively regulates PDK1 kinase activity by association with the residues surrounding the PDK1 Ser-241 residue. DISCUSSION It is clear that PDK1 plays a central role in activating the AGC subfamily of protein kinases (7,24). These kinases then mediate intracellular signaling such as cell survival, cell growth, protein synthesis, and gene expression. PDK1 phosphorylates AGC kinase members at the residues equivalent to FIG. 5. Incubation of PDK1 with 14-3-3 in vitro decreases PDK1 kinase activity. A, the recombinant PDK1 was incubated with the indicated amount of GST or GST⅐14-3-3 for 2 h at 30°C, following incubation with 15 M non-phosphorylated RSK2 peptide (non-phospho) or phosphorylated RSK2 peptide (phospho) for 20 min at 30°C. PDK1 autophosphorylation assay was performed, as described under "Experimental Procedures." The relative amounts of incorporated radioactivity were quantified with a BAS1500 Bio-Imaging analyzer. B, COS-7 cells were transfected with pCMV3 vector encoding nothing (Ϫ), WT-PDK1 (WT) or R238E-PDK1 (R238E). The immunoprecipitated Myc-tagged proteins were incubated with the indicated amounts of GST or GST⅐14-3-3 for 2 h at 4°C, following incubation with 15 M phosphorylated RSK2 peptide (phospho-RSK2 peptide). Then, 500 ng of recombinant inactive SGK was added to the reactions. PDK1 kinase assay was performed, as described under "Experimental Procedures." The relative amounts of 32 P-SGK were quantified with a BAS1500 Bio-Imaging analyzer. The amount of the immunoprecipitated Myctagged PDK1 proteins was confirmed by immunoblot analysis with an anti-Myc antibody (lower panel). The positions of molecular mass standards (kDa) are shown on the right.
Thr-308 of Akt (known as activation loop or T-loop) (7). PDK1 is, itself, a member of the AGC subfamily of protein kinases. Thus, PDK1 phosphorylates itself at its activation loop (Ser-241), thereby activating itself (9). Although PDK1 kinase activity has been thought to be constitutively active and not further activated by growth factor stimulation, recent findings suggest that its activity and its character are controlled by interaction with other proteins. For example, when PDK1 interacts with the PDK1-interacting fragment of PRK2, it is converted to exhibit PDK2-like activity (11). The association of PDK1 with RSK2 up-regulates PDK1 kinase activity and autophosphorylation (12). We recently reported that PDK1 binds to Hsp90, and its binding prevents PDK1 from proteasome-dependent degradation and keeps it in a soluble and active conformational state (13). Moreover, PDK1 kinase activity is promoted by phosphorylation at tyrosine residues, presumably by a member of the Src kinase family (25,26).
Here we provide evidence that PDK1 binds to 14-3-3 in vivo and in vitro through the residues surrounding the autophosphorylation site Ser-241 and that the association is achieved only when Ser-241 has been phosphorylated (Fig. 3). Although PDK1 contains other 14-3-3 binding motifs in its amino acid sequence ( 3 RTTS 6 , 407 RSGS 410 , 546 RYQS 549 ), site-directed mutagenesis of the target Ser revealed that these sites were not involved in the PDK1-14-3-3 complex formation (Fig. 3B). It should be noted that 14-3-3 also bound to the S410A-PDK1 mutant in which Ser-410 was converted to Ala although the Ser-410 residue has been reported to be phosphorylated in vivo (9). There may be some conformational state essential for the 14-3-3 binding in addition to Ser phosphorylation in 14-3-3 binding motifs.
Among the 14-3-3-binding partners, some proteins, like Raf-1, bind to almost all 14-3-3 isoforms with nearly equivalent affinity (18). However, others, like A20 and Cdc25B, bind to different 14-3-3 isoforms with significantly different affinities (27,28). We recently reported that p27 Kip1 binds to 14-3-3 after Akt-mediated phosphorylation at the COOH-terminal Thr-198 (29). The phosphorylated p27 Kip1 could form complexes with 14-3-3, , and ⑀ but not with 14-3-3␤ and (29), suggesting that 14-3-3 isoforms have different affinities to their partners. Among the five isoforms (␤, ⑀, , , and ), only 14-3-3 and interacted with PDK1, and 14-3-3 bound more tightly than 14-3-3 (Fig. 1B). The difference among 14-3-3, and other isoforms that accounts for the different affinities for PDK1 is still unclear, since many of the residues that form the ligandbinding groove are conserved in the seven isoforms. However, comparing the residues in 14-3-3 necessary for the interaction with Raf-1 and PDK1 did reveal some differences. To form a complex with Raf-1, Arg-56, Arg-60, and Arg-127 were necessary (Fig. 2B, lanes 3 and 4), a finding consistent with a previous report (23). In the case of PDK1, Arg-56 and Arg-60 were not essential but Arg-127 was ( Fig. 2A). By contrast, both Arg-56 and Arg-60 are necessary but Arg-127 is not for p27 Kip1 binding to 14-3-3 (Ref. 29). 2 These results indicate that the residues necessary for ligand-14-3-3 binding vary with the ligand. Moreover, there may be some residues essential for the association with ligands that are not conserved in the 14-3-3 isoforms. However, we could not exclude the possibility that the differences we observed above were a result of the differences between the ligand that possessed at least two 14-3-3 binding motifs (Raf-1) and the one that had only one (PDK1). The distinct mechanism of isoform-specific interaction is a problem to be solved in future studies.
The general mechanisms by which 14-3-3 regulates partner protein functions are 4-fold: 1) regulation of subcellular localization of the target protein; 2) direct regulation of catalytic activity of the binding partner; 3) protection of the ligands from proteolysis or dephosphorylation; and 4) regulation of interaction between bound protein and other molecules. The identified 14-3-3 binding residue of PDK1 (Ser-241) is thought to be phosphorylated by itself and it be essential for its kinase activity (9). Consistent with previous reporting (9), we confirmed that S241A-PDK1 exhibited weak Akt and SGK phosphorylation activity when compared with WT-PDK1 (Fig. 4). We thus generated PDK1 point mutants in which residues other than Ser-241 are mutated (Fig. 3C) and examined their activity to phosphorylate Akt (Fig. 4A) or SGK (Fig. 4C). Mutation of PDK1 to promote 14-3-3 (V243P-PDK1) association exhibited reduced PDK1 kinase activity to phosphorylate Akt and SGK in cells, while mutation of PDK1 to inhibit the interaction (R238E-PDK1) slightly increased its activity (Fig. 4). Moreover, incubation with recombinant 14-3-3 decreased PDK1 autophosphorylation activity (Fig. 5A) and kinase activity to phosphorylate SGK (Fig. 5B) in a dose-dependent manner in vitro.
These results indicate that 14-3-3 is not only a PDK1-associating protein but also a protein that negatively regulates PDK1 kinase activity.
Several reports indicate that PDK1 binds to phosphatidylinositol 3,4,5-trisphosphate and phosphatidylinositol 3,4bisphosphate through its pleckstrin homology domain (19,30). Because mutants of PDK1 deleted in its pleckstrin homology domain have been reported to prevent translocation of Akt, the relocalization of PDK1 from cytosol to the plasma membrane is suggested to play an important role in the recruitment of Akt to the plasma membrane and the subsequent Akt activation in stimulated cells (30). PDK1 catalytic activity is not required for the PDK1 translocation to the plasma membrane (30). Although mutating PDK1 to increase its binding to 14-3-3 (V243P) reduced kinase activity to phosphorylate Akt at Thr-308, the mutation did not affect the binding affinity of PDK1 to Akt (Fig. 4). Therefore, the decrease in kinase activity of V243P-PDK1 to phosphorylate Akt in cells may be caused by down-regulation of its kinase activity but not by suppressing translocation. This notion was supported by the fact that the amount of the phosphorylated form of Myr-Akt was reduced by 14-3-3 protein expression even though Myr-Akt constitutively targeted the plasma membrane (Fig. 4D). In the case of SGK, translocation to the plasma membrane is not required for PDK1dependent phosphorylation of SGK at Thr-256 (31). SGK phosphorylation at Thr-256 was also suppressed by PDK1 mutation to increase its binding to 14-3-3 (V243P; Fig. 4C). These results strongly suggest that inhibition of PDK1 activity but not PDK1 translocation is the main mechanism of 14-3-3-mediated inactivation of PDK1-dependent signal transduction.