PDK1 Protein Phosphorylation at Thr354 by Murine Protein Serine-Threonine Kinase 38 Contributes to Negative Regulation of PDK1 Protein Activity*

Background: This study was done to elucidate the biochemical mechanisms underlying phosphorylation-dependent regulation of PDK1. Results: MPK38 inactivates PDK1 activity and function by phosphorylating PDK1 at Thr354. Conclusion: MPK38 acts as a putative protein kinase to negatively regulate PDK1. Significance: This study defines a novel mechanism in which MPK38 directly interacts with and phosphorylates Thr354 of PDK1, thereby inhibiting PDK1 activity. Murine protein serine-threonine kinase 38 (MPK38) is a member of the AMP-activated protein kinase-related serine/threonine kinase family, which acts as cellular energy sensors. In this study, MPK38-induced PDK1 phosphorylation was examined to elucidate the biochemical mechanisms underlying phosphorylation-dependent regulation of 3-phosphoinositide-dependent protein kinase-1 (PDK1) activity. The results showed that MPK38 interacted with and inhibited PDK1 activity via Thr354 phosphorylation. MPK38-PDK1 complex formation was mediated by the amino-terminal catalytic kinase domain of MPK38 and the pleckstrin homology domain of PDK1. This activity was dependent on insulin, a PI3K/PDK1 stimulator, as well as various apoptotic stimuli, including TNF-α, H2O2, thapsigargin, and ionomycin. MPK38 inhibited PDK1 activity in a kinase-dependent manner and alleviated PDK1-mediated suppression of TGF-β (or ASK1) signaling, probably via the phosphorylation of PDK1 at Thr354. In addition, MPK38-mediated inhibition of PDK1 activity was accompanied by the modulation of PDK1 binding to its positive and negative regulators, serine/threonine kinase receptor-associated protein and 14-3-3, respectively. Together, these findings suggest an important role for MPK38-mediated phosphorylation of PDK1 in the negative regulation of PDK1 activity.

The 3-phosphoinositide-dependent protein kinase-1 (PDK1) is a member of the protein kinase A, G, and C subfamily of protein kinases. PDK1 acts via a PH 3 domain and binds phos-phoinositides such as phosphatidylinositol 3,4-P 2 and phosphatidylinositol 3,4,5-P 3 to phosphorylate Thr 308 of PKB/Akt. Phosphorylation of both Thr 308 and Ser 473 is required for maximal activation of PKB/Akt (1,2). These residues are phosphorylated independently and are dephosphorylated by PDK1/protein phosphatase 2A (PP2A) (for Thr 308 ) and mammalian target of rapamycin complex 2 (mTORC2)/PH domain leucine-rich repeat protein phosphatases (for Ser 473 ), respectively (3)(4)(5)(6). In addition, analysis of cellular proteins that interact with PDK1 shows that the activity and function of PDK1 is regulated by PDK1-interacting proteins within cells (7)(8)(9)(10). Although an important role for PDK1 in cell survival signaling is well characterized, the mechanism by which PDK1 activity is regulated through phosphorylation remains largely unknown. It was previously thought that PDK1 is constitutively active because it shows a high level of basal activity in unstimulated cells and cannot be further activated by growth factor stimulation (11). However, recent studies strongly suggest that PDK1 activity is regulated in a phosphorylation-dependent manner. Multiple phosphorylation sites on PDK1 (Ser 25 , Ser 241 , Ser 393 , Ser 396 , and Ser 410 ) have been identified in unstimulated HEK293 cells, but only the phosphorylation of Ser 241 (Ser 244 in mouse PDK1) within the activation loop is responsible for PDK1 activity (12). Phosphorylation of Ser 396 is necessary for nuclear shuttling of PDK1 (13). Phosphorylation of mouse PDK1 at Ser 163 is also involved in fine-tuning PDK1 activity (14). In addition, PDK1 in HEK293 cells treated with pervanadate, a tyrosine phosphatase inhibitor, undergoes tyrosine phosphorylation at Tyr 9 , Tyr 373 , and Tyr 376 , leading to its activation (15). RET/PTC, a thyroidspecific oncogenic kinase (16), stimulates PDK1 activity by phosphorylating Tyr 9 (17). Protein kinase C (PKC), a kinase implicated in hyperlipidemia-induced insulin resistance, negatively regulates PDK1 activity via phosphorylation at Ser 504 and Ser 532 (18). Reduced PDK1 phosphorylation at Ser 244 , which is stimulated by insulin, occurs in the liver of obese/obese mice (19). We recently demonstrated that PDK1 undergoes apoptosis signal-regulating kinase 1 (ASK1)-dependent phosphorylation at Ser 394 and Ser 398 , which suppresses its activity (20).
These findings suggest that the phosphorylation of PDK1 plays an important role in regulating PDK1 activity and function.
Therefore, to explore the phosphorylation-dependent regulation of PDK1, we investigated whether MPK38 contributes to the phosphorylation of PDK1 and whether it plays a regulatory role in the PDK1 activity. We showed that MPK38 physically interacts with and phosphorylates PDK1 at Thr 354 , thereby inhibiting its activity and function. Our work also suggests that MPK38, like PKC and ASK1 (18,20), acts as a putative protein kinase to negatively regulate PDK1 in cells.
Assays for in Vivo and in Vitro Protein Interactions-Assays were carried out as described previously (20,23).
Preparation of Recombinant Proteins and the PDK1 Kinase Assay-Recombinant glutathione S-transferase (GST) or Histagged wild-type and deletion constructs of PDK1 and MPK38 were purified by affinity chromatography on glutathione-Sepharose 4B or His columns (Amersham Biosciences). The PDK1 kinase assay was performed as described previously (29) using immunoprecipitated or recombinant PDK1 proteins. Approximately 500 ng of recombinant SGK (Upstate) was used as the substrate.
Luciferase Reporter Assay-Assays were carried out in 293T cells as described previously (24). Luciferase activity was assayed using the Dual-Luciferase assay system (Promega) according to the manufacturer's instructions and normalized to ␤-galactosidase activity.
GFP-based Cell Death Assay-Green fluorescent protein (GFP)-based cell death assays were carried out in HEK293, 293T, and HaCaT cells as described previously (23,24). The nuclei of GFP-positive cells were stained with 4Ј,6Ј-diamidino-2-phenylindole dihydrochloride (DAPI) and analyzed for apoptotic morphology under a fluorescence microscope. The percentage of apoptotic cells was calculated as the number of GFP-positive cells with apoptotic nuclei divided by the total number of GFP-positive cells.
Cell Cycle Analysis-Assays were performed in HaCaT cells as described previously (23). Cells were transfected with the indicated plasmid combinations using WelFect-Ex TM Plus (WelGENE, Daegu, Korea). The cell fraction at each stage of the cell cycle was analyzed after treatment with 10% serum for 24 h in the presence or absence of 2 ng/ml porcine TGF-␤1 (R & D Systems). Flow cytometry analysis was performed using the FACSCalibur-S system (BD Biosciences).
Statistical Analysis-Values represent the means Ϯ S.E. A p value Ͻ0.05 calculated using the Student's t test was considered statistically significant.

MPK38 Interacts with PDK1 Both in Vitro and in Vivo-We
previously showed that PDK1 inhibits Smad-mediated signaling via direct interaction with Smad proteins (29). In addition, MPK38 physically interacts with and phosphorylates Smad proteins, resulting in the stimulation of TGF-␤ signaling (23). Therefore, we speculated that there may be a direct or indirect functional link between MPK38 and PDK1 signaling pathways in cells.
To test this hypothesis, we examined whether MPK38 physically interacts with PDK1 in cells using cotransfection experiments incorporating HEK293 cells expressing GST-MPK38 and FLAG-PDK1. The interaction between MPK38 and PDK1 was analyzed by immunoblotting with an anti-FLAG antibody. The presence of PDK1 was detected in the coprecipitate only when coexpressed with GST-MPK38 but not with GST alone (control) (Fig. 1A, left panel). The endogenous interaction between MPK38 and PDK1 was subsequently determined by coimmunoprecipitation experiments in HEK293, NIH 3T3, and R1.1 cells. Immunoprecipitation of endogenous MPK38 by the anti-MPK38 antibody followed by immunoblotting with an anti-PDK1 antibody identified a physical interaction between the two endogenous proteins (Fig. 1A, right panel). These observations prompted us to map the interaction domain of MPK38, which is involved in binding to PDK1. We performed in vivo binding assays using HEK293 cells expressing wild-type MPK38 and two deletion constructs of MPK38 as follows: MCAT harboring the catalytic kinase domain (amino acids 7-269), and MPKC comprising the carboxyl-terminal regulatory domain (amino acids 270 -643). Wild-type MPK38 and MCAT were able to bind PDK1, but no binding of MPKC to PDK1 was detected (Fig. 1B, left panel). This indicates that the amino-terminal kinase domain of MPK38 is responsible for PDK1 binding. Using the same approach, we determined which domain of PDK1 contributes to MPK38 binding. The carboxyl-FIGURE 1. MPK38 interacts with PDK1. A, in vivo association of MPK38 with PDK1. GST alone or GST-MPK38 was cotransfected into HEK293 cells along with FLAG-PDK1. GST fusion proteins were purified on glutathione-Sepharose beads (GST Purification), and complex formation was analyzed by immunoblotting with an anti-FLAG antibody (left panel). Equal amounts of cell lysate from HEK293, NIH 3T3, and R1.1 cells were immunoprecipitated with either rabbit preimmune serum (Preimm.) or anti-MPK38 antibody (␣-MPK38) followed by immunoblot analysis with an anti-PDK1 antibody to determine endogenous binding (right top panels). B, mapping of the MPK38 and PDK1 binding domains. The schematic structures of wild-type and deletion constructs of Mpk38 (left panel) and PDK1 (right panel) are shown. HEK293 cells transfected with the indicated expression vectors were lysed, precipitated using glutathione-Sepharose beads, and then immunoblotted with an anti-FLAG antibody to determine the level of MPK38-PDK1 binding. C, in vitro binding of MPK38 and PDK1. For native PAGE of the MPK38-PDK1 complex, autophosphorylated His-tagged PDK1 or MPK38 (each 2-3 g), prepared in the presence of the respective kinase buffers (24,29), were incubated with unlabeled recombinant GST-tagged kinase-dead MPK38 (or PDK1) and its deletion constructs (for MPK38, MPKC, and MCAT; for PDK1, CA, and PH) (each 5 g), together with the nonspecific control GST at room temperature for 1 h. The same blot was stripped and re-probed with anti-PDK1 and anti-MPK38 antibodies to confirm the presence of PDK1 and MPK38 on the radioactive band shifts (middle and bottom panels). The purity of recombinant PDK1 and MPK38 proteins used for this experiment was analyzed by Coomassie Blue staining (see supplemental Fig. 4A To confirm this, we analyzed the in vitro association of purified recombinant PDK1 with MPK38 using nondenaturing PAGE. Autophosphorylated recombinant PDK1 was incubated with an unlabeled recombinant kinase-dead (K40R) MPK38 with one of its deletion mutants (MPKC and MCAT) or with GST as a nonspecific control. A shift in the mobility of 32 Plabeled PDK1 was clearly detected upon incubation with kinase-dead MPK38 or MCAT, but no shift was observed upon incubation with GST alone or MPKC (Fig. 1C, left panel). Similarly, a shift in the mobility of 32 P-labeled MPK38 was clearly evident upon incubation with kinase-dead (KD) PDK1 and its deletion mutant PDK1(PH), but it was undetectable in the presence of GST alone or PDK1(CA) (Fig. 1C, right panel). Cumulatively, these results demonstrate that the physical interaction between MPK38 and PDK1 is mediated by the amino-terminal kinase domain of MPK38 and the carboxyl-terminal PH domain of PDK1.
Modulation of MPK38-PDK1 Complex Formation by Insulin and Apoptotic Stimuli-Because MPK38 and PDK1 interact with each other (see Fig. 1), and MPK38 stimulates apoptosis signal-regulating kinase 1 (ASK1) signaling (24), we assessed whether a PI3K/PDK1 stimulator (insulin) or MPK38/ASK1 stimulators (H 2 O 2 , TNF-␣, thapsigargin, and ionomycin) had an effect on MPK38-PDK1 complex formation in HEK293 cells. As shown in Fig. 2, the exposure of the cells to insulin resulted in a considerable decrease in endogenous MPK38-PDK1 complex formation, but this effect was alleviated by treatment with wortmannin (a PI3K inhibitor). In addition, the endogenous interaction between MPK38 and PDK1 appeared to be decreased by treatment with agents that stimulate MPK38/ ASK1. These results suggest that MPK38 may be associated with the PI3K/PDK1 signaling pathway.
MPK38 Inhibits PDK1 Activity via Thr 354 Phosphorylation-To determine whether MPK38 affected PDK1 kinase activity, HEK293 cells were transfected with either PDK1 alone or with both PDK1 and Mpk38. PDK1 kinase activity decreased markedly when PDK1 was coexpressed with wild-type MPK38, whereas coexpression of kinase-dead (K40R) MPK38 had no effect on the PDK1 kinase activity compared with the control expressing wild-type PDK1 alone (Fig. 3A, left panel). However, this effect was not observed under the same conditions in the presence of kinase-dead PDK1, suggesting that PDK1-mediated phosphorylation of the SGK substrate occurred without relying on additional proteins coimmunoprecipitated with PDK1. This result indicates that MPK38 inhibited PDK1 kinase activity in a kinase-dependent manner. Another approach using recombinant PDK1 and MPK38 proteins showed that recombinant wild-type MPK38 directly inhibited PDK1 kinase activity in a dose-dependent manner (Fig. 3A, right panel).
To assess whether the MPK38-mediated inhibition of PDK1 kinase activity influenced PDK1-mediated signaling, we developed a stable system for the tetracycline-inducible expression of Mpk38 shRNA in HEK293 cells (inducible MPK38 shRNA). Parental HEK293 cells, HEK293 cells expressing a scrambled shRNA (inducible Sc shRNA), or inducible MPK38 shRNA cells were either untreated or treated with doxycycline to induce the knockdown of endogenous MPK38. Anti-phospho-specific antibodies for PDK1 Ser 241 , AKT Thr 308 , AKT Ser 473 , and BAD Ser 136 were then used for immunoblot analysis to assess PDK1 downstream signaling. As shown in Fig. 3B, knockdown of endogenous MPK38 markedly stimulated PDK1 downstream signaling. These findings point to the inhibition of PDK1-mediated signaling upon MPK38 binding and suggest that MPK38 is a potential negative regulator of PDK1 activity.
We next determined whether PDK1 acts as a substrate for MPK38 using the recombinant nonphosphorylated form of the kinase-dead (KD) PDK1 protein as a substrate in the MPK38 kinase assay. Wild-type and kinase-dead MPK38 proteins, purified from GST-MPK38-expressing HEK293 cell lysates using glutathione-Sepharose beads, were incubated with [␥-32 P]ATP to allow phosphorylation of recombinant PDK1(KD). Phosphorylation of recombinant PDK1(KD) was clearly detected in the presence of wild-type MPK38 but not in the presence of GST alone or kinase-dead MPK38 (Fig. 3C, left panel), indicating that PDK1 may be a substrate for MPK38. Next, to identify the MPK38 phosphorylation sites on PDK1, we performed an alignment analysis using the AMP-activated protein kinase/ MPK38 consensus sequence (23,30); three potential MPK38 phosphorylation sites (Ser 92 , Thr 354 , and Ser 393 ) on PDK1 were selected. In vitro kinase assays using recombinant MPK38 and PDK1(KD) substitution mutants (S92A, T354A, and S393A) showed that MPK38 phosphorylated two PDK1(KD) mutants, S92A and S393A, but not the T354A mutant (Fig. 3C, right  panel). This was also confirmed by immunoblot analysis using an anti-phospho-PDK1(T354) antibody. These results suggest that MPK38 physically interacts with and phosphorylates PDK1 at Thr 354 .
MPK38 Alleviates PDK1-mediated Suppression of Apoptosis in a Kinase-dependent Manner-To investigate whether MPK38 regulates PDK1-mediated suppression of apoptosis, we analyzed the effect of MPK38 on PDK1-mediated suppression of TNF-␣-induced apoptosis using the GFP system (10,29).
Expression of PDK1 in the presence of TNF-␣ resulted in a considerable decrease in apoptotic cell death compared with that in control cells expressing an empty vector. However, wildtype MPK38, but not kinase-dead (K40R) MPK38, alleviated this suppression in a dose-dependent manner (Fig. 4A, compare 6th lane with 7th to 10th lanes). This result implies that MPK38 contributes to the negative regulation of the PDK1mediated survival signaling pathway through direct interaction and phosphorylation. To confirm this, we examined the effect of Mpk38 siRNA on TNF-␣-induced apoptosis. As expected, knockdown of endogenous MPK38 decreased TNF-␣-induced apoptosis in a dose-dependent manner compared with control cells expressing PDK1 alone (Fig. 4A, compare 6th lane with  11th to 14th lanes). To determine the role of MPK38-mediated phosphorylation of PDK1 at Thr 354 in the regulation of PDK1 signaling, we also assessed the effect of MPK38 on PDK1(T354A)-mediated signaling because the T354A mutant  MPK38 shRNA), or parental HEK293 cells, were cultured in the presence or absence of 1 g/ml doxycycline (Dox) for 72 h, and PDK1 downstream signaling was determined by immunoblot analysis. Inducible silencing of endogenous MPK38 by doxycycline was assessed by immunoblotting using an anti-MPK38 antibody. ␤-Actin was used as a loading control. C, identification of MPK38 phosphorylation sites on PDK1. After 48 h of transfection with GST-tagged wild-type or kinase-dead Mpk38, HEK293 cell lysates were subjected to precipitation with glutathione-Sepharose beads (GST purification) and analyzed in an in vitro kinase assay with recombinant kinase-dead PDK1 as the substrate (left panel). The in vitro kinase assay was performed with ϳ3 g of recombinant GST-tagged kinase-dead PDK1 or one of its substitution mutants (S92A, T354A, or S393A) in the presence or absence of recombinant His-tagged wild-type MPK38 (ϳ6 g) (right panel). The purity of recombinant MPK38 and PDK1 proteins used for this experiment was confirmed by Coomassie Blue staining (see supplemental Fig. 4B). P-MPK38 and P-PDK1(KD) indicate autophosphorylated MPK38 and phosphorylated kinasedead PDK1, respectively. was found to be defective in MPK38-mediated phosphorylation (see Fig. 3C). Neither wild-type nor kinase-dead MPK38 had an effect on PDK1(T354A)-mediated suppression of apoptosis compared with that in a control expressing PDK1(T354A) alone (Fig. 4A, compare 15th lane with 16th to 19th lanes).
These results indicate that MPK38 negatively regulates PDK1 activity by phosphorylating it on Thr 354 . We then extended our analysis to examine whether MPK38 regulated serum-induced cell growth induced by PDK1 activation. Flow cytometry analysis using HaCaT cells showed that coexpression of wild-type MPK38 considerably decreased the percentage of cells in S phase compared with that in control cells expressing PDK1 alone (Fig. 4B, upper panel, 3rd lane versus 5th lane, ϳ51 versus ϳ41%; see supplemental Fig. 1A), although kinase-dead MPK38 had no effect (Fig. 4B, upper panel, 3rd lane versus 7th lane, ϳ51 versus ϳ51%; see supplemental Fig. 1A). These results suggest that the decrease in the number of S phase cells observed in the presence of MPK38 was due to MPK38-mediated phosphorylation of PDK1, because transfected Mpk38 (wild-type or kinase-dead) itself did not influence the percentage of cells in S phase (Fig. 4B, upper panel, 2nd lane versus 4th  and 6th lanes, ϳ37 versus ϳ37%;  . With regard to PDK1 phosphorylation at Thr 354 , MPK38 had no effect on the accumulation of S phase cells in the presence of the T354A mutant (Fig.  4C, 4th lane versus 7th lane, ϳ54 versus ϳ54%; see supplemental Fig. 1B). These results suggest that MPK38-mediated phosphorylation of PDK1 at Thr 354 plays an important role in the negative regulation of PDK1-mediated cell growth.
MPK38 Alleviates PDK1-mediated Suppression of TGF-␤ Signaling in a Kinase-dependent Manner-PDK1 inhibits TGF-␤ signaling (29); therefore, to investigate whether MPK38 is involved in the regulation of PDK1-mediated TGF-␤ signaling, we examined the effect of MPK38 on PDK1-mediated suppression of TGF-␤-induced apoptosis. Wild-type MPK38, but not kinase-dead MPK38, increased apoptotic cell death in a dose-dependent manner compared with that in control cells expressing PDK1 alone (Fig. 5A, upper panel, 6th lane versus 7th to 10th lanes). However, neither wild-type nor kinase-dead MPK38 had an effect on PDK1(T354A)-mediated suppression of TGF-␤-induced apoptosis (Fig. 5A, lower panel, 6th lane versus 9th to 12th lanes). This result implies that MPK38 contributes to the regulation of PDK1-mediated TGF-␤ signaling by phosphorylating PDK1 at Thr 354 . We also performed flow cytometry analysis using HaCaT cells to determine whether MPK38 had a similar effect on PDK1-mediated suppression of TGF-␤-induced cell cycle arrest. Coexpression of wild-type MPK38 decreased the accumulation of cells in S phase after 24 h of serum stimulation in the presence of TGF-␤1 compared with that in control cells expressing PDK1 alone (Fig. 5B, 3rd lane versus 6th lane, ϳ34 versus ϳ28%; see supplemental Fig. 2). However, kinase-dead MPK38 had no effect on the accumulation of S phase cells (Fig. 5B, 3rd lane versus 9th lane, ϳ34 versus ϳ34%; see supplemental Fig. 2). These results are consistent with those obtained from the analysis of MPK38 involvement in PDK1-mediated suppression of TNF-␣-induced apoptosis and cell cycle arrest (Fig. 4). Furthermore, the effect of MPK38 on PDK1-mediated TGF-␤ signaling was due to the phosphorylation of PDK1 at Thr 354 (induced by MPK38 itself) because wild-type MPK38 had no effect on the accumulation of S phase cells in the presence of the T354A mutant (Fig.  5B, 4th lane versus 7th lane, ϳ36 versus ϳ37%; see supplemental Fig. 2). Taken together, these results suggest that MPK38 alleviates PDK1-mediated suppression of TGF-␤ signaling by phosphorylating PDK1 at Thr 354 .
MPK38 Alleviates PDK1-mediated Suppression of ASK1 Signaling in a Kinase-dependent Manner-Because PDK1 interacts with ASK1 and inhibits ASK1-mediated signaling (20), we also examined whether MPK38 had an effect on PDK1-mediated suppression of H 2 O 2 -induced cell death. Wild-type MPK38, but not kinase-dead MPK38, alleviated PDK1-mediated suppression of H 2 O 2 -induced apoptosis in a dose-dependent manner (Fig. 6A, upper panel, 7th lane versus 8th to 11th  lanes). To further confirm this observation, we performed knockdown experiments using Mpk38 siRNA. Transfection of Mpk38 siRNA into HEK293 cells potentiated the PDK1-mediated suppression of H 2 O 2 -induced apoptosis in a dose-dependent manner (Fig. 6A, upper panel, 7th lane versus 12th to 17th  lanes). These results suggest that MPK38 contributes to the alleviation of PDK1-mediated suppression of ASK1 signaling by inhibiting PDK1 activity via direct interaction and phosphorylation. To verify this, we compared the effect of MPK38 on PDK1-mediated suppression of H 2 O 2 -induced apoptosis in the presence of wild-type PDK1 with its effect in the presence of the T354A mutant, which is defective in MPK38-mediated phosphorylation. In contrast with the negative effect of wild-type MPK38 on PDK1-mediated suppression of H 2 O 2 -induced apoptosis (Fig. 6A, lower panel, 7 to 11 lanes), neither wild-type nor Because PDK1 suppresses ASK1-induced AP-1 transcriptional activity (20), it is possible that MPK38 also enhances AP-1 transcriptional activity. To test this hypothesis, we performed an AP-1-responsive luciferase reporter assay to determine whether MPK38 affected the transcriptional activity of AP-1. As expected, wild-type MPK38, but not kinase-dead MPK38, increased AP-1-dependent luciferase activity in the presence of ASK1 and PDK1 in a dose-dependent manner (Fig.  6B, upper panel, 8th to 12th lanes). However, this effect was not observed in the presence of ASK1 and T354A mutants (Fig. 6B,  lower panel, 13th to 17th lanes), suggesting that MPK38-mediated phosphorylation of PDK1 at Thr 354 is also required for the alleviation of PDK1-mediated suppression of AP-1 transcriptional activity.
Phosphorylation of PDK1 at Thr 354 , Ser 394 , and Ser 398 Functions Cooperatively to Inhibit PDK1 Activity-ASK1-mediated phosphorylation of PDK1 at Ser 394 and Ser 398 contributes to the inhibition of PDK1 activity (20). Therefore, to establish whether the phosphorylation of PDK1 at Thr 354 induced by MPK38 has a similar effect on the regulation of PDK1 activity, we first analyzed the kinase activity of the T354A mutant using an in vitro kinase assay. The results showed that the T354A mutant induced PDK1 kinase activity at a level comparable with that of the S394A/S398A double mutant, which is defective in ASK1-mediated phosphorylation (Fig. 7A, 1st to 3rd lanes). Next, to determine the cooperative effects of PDK1 phosphorylation at Thr 354 , Ser 394 , and Ser 398 in the negative regulation of PDK1 activity, we measured the kinase activity of the PDK1 S394A/S398A/T354A triple mutant and compared it with that of two other PDK1 mutants, T354A and S394A/S398A. The S394A/S398A/T354A triple mutant showed higher phosphorylation of SGK than the T354A single mutant or the S394A/ S398A double mutant (Fig. 7A, 1st lane versus 2nd to 4th lanes). These results provide evidence that the phosphorylation of PDK1 at Thr 354 , Ser 394 , and Ser 398 is cooperatively involved in the negative regulation of PDK1 activity. If this is the case, the increased kinase activity of the PDK1 mutants defective in MPK38-and/or ASK1-mediated phosphorylation may influence PDK1-mediated cell survival functions. To verify this, we analyzed the effect of three PDK1 mutants (T354A, S394A/ S398A, and S394A/S398A/T354A) on PDK1-mediated suppression of TNF-␣-induced apoptosis. As expected, the suppressive effect of the S394A/S398A/T354A triple mutant was stronger than that of the T354A or S394A/S398A mutants (Fig.  7B, left panel). A similar result was also obtained when we analyzed the effect of the three PDK1 mutants on PDK1-mediated suppression of TGF-␤-induced apoptosis (Fig. 7B, right panel). Together, these results indicate that the phosphorylation of PDK1 at Thr 354 , Ser 394 , and Ser 398 by MPK38 and ASK1 has a cooperative effect on the negative regulation of PDK1 activity.
MPK38 Modulates Complex Formation between PDK1 and Its Regulators STRAP and 14-3-3 Protein-Because STRAP, originally identified as a TGF-␤ receptor-interacting protein (32), positively regulates PDK1 activity (10), we thought that MPK38 may have an effect on the binding of PDK1 to STRAP. To test this, HEK293 cells transfected with FLAG-STRAP and Myc-PDK1 in the presence or absence of Mpk38 were subjected to immunoprecipitation using an anti-FLAG antibody followed by immunoblot analysis with an anti-Myc antibody. There was a considerable decrease in complex formation between PDK1 and STRAP in cells coexpressing MPK38 (Fig. 8A, left panel). Concordantly, coexpression of MPK38 decreased endogenous complex formation between PDK1 and STRAP (Fig. 8A, middle  panel). Because PDK1 forms an inactive complex with 14-3-3 (8), we also investigated the effect of MPK38 on PDK1-14-3-3 complex formation. The results showed that MPK38 markedly increased the interaction between PDK1 and 14-3-3 (Fig. 8B,  left and middle panels). These results suggest that PDK1 inactivation by MPK38 modulates complex formation between PDK1 and its positive and negative regulators, STRAP and 14-3-3, respectively.
We next examined the effect of MPK38-mediated phosphorylation of PDK1 at Thr 354 on complex formation between PDK1 and its regulators, STRAP and 14-3-3, using in vivo binding assays. The results showed that MPK38-mediated phosphor-ylation of PDK1 at Thr 354 stabilized complex formation between PDK1 and its negative regulator 14-3-3, but destabilized complex formation between PDK1 and its positive regulator STRAP, thereby inhibiting PDK1 activity (Fig. 8, A and B,  right panels). This finding suggests that the phosphorylation of PDK1 at Thr 354 induced by MPK38 has an important impact on the interaction of PDK1 with its regulators (STRAP and 14-3-3) and plays a key role in the progress of PDK1-mediated signaling.

DISCUSSION
In this study, we show that MPK38 inhibits PDK1, a master kinase for regulating multiple signaling pathways (34,35), through direct interaction and phosphorylation. MPK38 inactivates PDK1 by phosphorylating PDK1 at Thr 354 . This suggests that the MPK38-dependent phosphorylation of PDK1 at Thr 354 , like Ser 394 /Ser 398 phosphorylation by ASK1 and Ser 504 /Ser 532 phosphorylation by PKC (18,20), plays a negative role in the regulation of PDK1 activity and function.
We investigated whether MPK38 directly phosphorylates PDK1 using an in vitro kinase assay because MPK38 physically interacted with PDK1 both in vivo and in vitro (Fig. 1). We found that MPK38-mediated phosphorylation of PDK1 occurred exclusively at Thr 354 , leading to the inhibition of PDK1 activity and function (Figs. 4 -7). These findings indicate that the Thr 354 of PDK1 is a MPK38 phosphorylation site for the negative regulation of PDK1.
Our results also showed that the T354A mutant itself had a comparatively minor effect on PDK1 function when compared with the control wild-type PDK1 (Figs. 4 -7), suggesting the possibility of other potential phosphorylation sites on PDK1 that regulate its activity. Indeed, a similar trend was observed in our previous study showing that ASK1-dependent phosphorylation of PDK1 at Ser 394 and Ser 398 plays a negative regulatory role in the PDK1 function (20). This observation, together with that of PKC (18), strongly indicates that the phosphorylation of PDK1 plays an important role in the regulation of PDK1 activity and function. To examine whether the phosphorylation sites (Thr 354 , Ser 394 , and Ser 398 ) on PDK1, which are known to inhibit the PDK1 activity (20), had a cooperative effect on the negative regulation of PDK1 activity and function, we analyzed the kinase activity and apoptotic suppression of a PDK1 triple mutant (S394A/S398A/T354A), together with two other PDK1 mutants, T354A and S394A/S398A, which are defective in MPK38-and ASK1-mediated phosphorylation, respectively. The results showed that the PDK1 triple mutant (S394A/ S398A/T354A) had a stronger effect on the kinase activity and apoptotic suppressive function of PDK1 than T354A or S394A/ S398A (Fig. 7), indicating that PDK1 phosphorylation at Thr 354 , Ser 394 , and Ser 398 has a cooperative effect on the negative regulation of PDK1 signaling. Recent studies suggest the involvement of docking interactions between protein kinases and their substrates for achieving substrate specificity and regulation of protein kinase activities (36). Based on this, one may raise the argument that the regulation of PDK1 activity by MPK38 is due to the direct docking interaction between PDK1 and MPK38. However, MPK38, unlike other protein family kinases A, G, and C interacting with PDK1, was found to interact with PDK1 via the amino-terminal kinase domain of MPK38 (see Fig. 1B). In addition, as in the case of SGK (Fig. 7A), similar results (data not shown) were obtained when we analyzed the effect of the three PDK1 mutants on PDK1 kinase activity using other PDK1 substrates that do not possess docking sites, including Smads (Smad2, -3, -4, and -7) (29), STRAP (10), and ZPR9 (see Ref. 31 and data not shown). In this context, it seems that the most likely mechanism by which MPK38 may inhibit the PDK1 activity would be through the change of PDK1 intrinsic activity, probably via direct phosphorylation of PDK1 at Thr 354 by MPK38, rather than docking interaction-mediated regulation of PDK1 activity. HEK293 cells, transfected with the indicated plasmid vectors, were lysed, immunoprecipitated (IP), and then immunoblotted with the indicated antibodies to assess the levels of PDK1-STRAP (A) or PDK1-14-3-3 (B) complex formation (left panels). Cell lysates from HEK293 cells transfected with or without GST-MPK38 were subjected to immunoprecipitation using either rabbit preimmune serum (Preimm.) or an anti-PDK1 antibody (␣-PDK1) followed by immunoblotting using an anti-STRAP antibody or anti-14-3-3 antibody to determine endogenous complex formation between PDK1 and STRAP or PDK1 and 14-3-3 (A and B, middle panels). To determine the effect of PDK1 phosphorylation at Thr 354 on endogenous association between PDK1 and STRAP or PDK1 and 14-3-3, HEK293 cells were transfected with the indicated plasmid vectors and lysed and purified on glutathione-Sepharose beads (GST Purification). The amount of complex formation was evaluated by immunoblotting with an anti-FLAG antibody (A and B, right panels). WB, Western blot.
Modulating the association between ASK1 and its regulators, such as TRX and 14-3-3, has been proposed as a potential mechanism for the MPK38-mediated stimulation of ASK1 activity (24). Therefore, we speculated that MPK38 may inhibit PDK1 signaling, possibly by influencing the association between PDK1 and its regulators, STRAP (10) and 14-3-3 (8).
To test this hypothesis, we examined the effect of MPK38 on STRAP and 14-3-3 binding to PDK1 using in vivo binding assays. Coexpression of MPK38 markedly decreased the association between PDK1 and its positive regulator STRAP and increased the association between PDK1 and its negative regulator 14-3-3 (Fig. 8). These results indicate that the MPK38mediated inhibition of PDK1 signaling is accompanied by the modulation of PDK1 binding to its regulators, STRAP and 14-3-3, similar to the MPK38-mediated stimulation of ASK1 signaling reported previously (24). As PDK1 phosphorylation at Thr 354 plays an important role in the negative regulation of PDK1 activity and function, it is likely that Thr 354 phosphorylation of PDK1 may influence binding between PDK1 and its regulators (STRAP and 14-3-3). We found that complex formation between PDK1 and its positive regulator STRAP increased in the presence of the T354A mutant compared with the control expressing wild-type PDK1, whereas complex formation between PDK1 and its negative regulator 14-3-3 decreased (Fig.  8, A and B, right panels). These results suggest that MPK38mediated phosphorylation of PDK1 at Thr 354 modulates the interaction between PDK1 and its regulators (STRAP and 14-3-3), which is crucial for determining the manner of PDK1 signaling and eventually inhibits PDK1 activity and function.
The fact that PDK1 physically interacts with both ASK1 and MPK38 led us to hypothesize that a ternary complex consisting of PDK1, MPK38, and ASK1 occurs within cells (supplemental Fig. 3). In fact, the PDK1-MPK38-ASK1 complex was detected in unstressed cells; however, treatment with H 2 O 2 disrupted this ternary complex and led to the formation of a binary complex between MPK38 and ASK1, which allowed the stimulation of ASK1 signaling, probably by stabilizing the MPK38-ASK1 complex in the presence of H 2 O 2 . In addition to the ternary complex, it is possible that the binary PDK1-ASK1 complex exists in unstressed cells because the interaction domains of ASK1 responsible for PDK1 and MPK38 binding, as well as the interaction domains of PDK1 responsible for ASK1 and MPK38 binding, are equivalent (20,24). Treatment with H 2 O 2 also disrupted the binary complex between PDK1 and ASK1 (20), promoting complex formation between MPK38 and ASK1 (24) and subsequently leading to ASK1 activation.
Collectively, the results of this study define a novel mechanism in which MPK38 directly interacts with and phosphorylates Thr 354 of PDK1, thereby inhibiting PDK1 activity. The results also provide evidence that Thr 354 of PDK1, like Ser 394 and Ser 398 of ASK1 (20) and Ser 504 and Ser 532 of PKC (18), represents a potential phosphorylation site for the negative regulation of PDK1 activity. Furthermore, the finding that MPK38 phosphorylates PDK1 on Thr 354 , thereby negatively regulating PDK1 activity, will contribute to a better understanding of the regulatory mechanism(s) involved in PDK1 activity and function.