Involvement of Hsp90 in Signaling and Stability of 3-Phosphoinositide-dependent Kinase-1*

Serine/threonine kinase Akt is thought to mediate many biological actions toward anti-apoptotic responses. Screening of drugs that could interfere with the Akt signaling pathway revealed that Hsp90 inhibitors (e.g. geldanamycin, radicicol, and its analogues) induced Akt dephosphorylation, which resulted in Akt inactivation and apoptosis of the cells. Hsp90 inhibitors did not directly affect Akt kinase activity in vitro. Thus, we examined the effects of Hsp90 inhibitors on upstream Akt kinases, phosphatidylinositide-3-OH kinase (PI3K) and 3-phosphoinositide-dependent protein kinase-1 (PDK1). Hsp90 inhibitors had no effect on PI3K protein expression. In contrast, treatment of the cells with Hsp90 inhibitors decreased the amount of PDK1 without directly inhibiting PDK1 kinase activity. We found that the kinase domain of PDK1 was essential for complex formation with Hsp90 and that Hsp90 inhibitors suppressed PDK1 binding to Hsp90. PDK1 degradation mechanisms revealed that inhibition of PDK1 binding to Hsp90 caused proteasome-dependent degradation of PDK1. Treatment of proteasome inhibitors increased the amount of detergent-insoluble PDK1 in Hsp90 inhibitor-treated cells. Therefore, the association of PDK1 with Hsp90 regulates its stability, solubility, and signaling. Because Akt binding to Hsp90 is also involved in the maintenance of Akt kinase activity, Hsp90 plays an important role in PDK1-Akt survival signaling pathway.

The interaction of PtdIns(3,4,5)P 3 with the PH domain of Akt recruits Akt to the plasma membrane, where it is phosphorylated at two key regulatory residue sites, Thr 308 and Ser 473 . Phosphorylation at both residues is necessary for full activation of Akt and the subsequent regulation of many PI3K-regulated biological responses, including glucose uptake, protein synthesis, and apoptosis inhibition (7,8). Akt phosphorylation at Thr 308 is catalyzed by the ubiquitously expressed 3-phosphoinositide-dependent protein kinase-1 (PDK1) (9 -11). The kinase responsible for phosphorylation of Akt at Ser 473 residue is called PDK2. Recently, Balendran et al. (12) reported that interaction of the fragment of protein kinase C-related kinase-2 with PDK1 converted PDK1 from a kinase that could phosphorylate only the Thr 308 residue of Akt to one that could phosphorylate both Thr 308 and Ser 473 residues. Akt itself was reported to be associated with the Akt phosphorylation at the Ser 473 residue (13).
The 90-kDa heat-shock protein (Hsp90) was highly conserved and played an important role in refolding certain denatured proteins under stress conditions. Unlike the more general Hsp70 and Hsp60 chaperones, Hsp90 appeared to have substrate-specific folding activity. Hsp90 has an additional role in the conformational regulation of certain signal transduction molecules, such as steroid hormone receptors and some kinases. Recent reports, including ours, indicated that Hsp90 is involved in the inhibition of apoptosis by suppressing cytochrome c-mediated oligomerization of Apaf-1 (14) or by stabilizing Akt kinase activity (15). These data suggest that Hsp90 plays an important role in cell survival. This notion was supported by the facts that Hsp90 inhibitors could induce apoptosis in various types of cell lines (reviewed in Refs. 16 and 17).
We thus investigated whether Hsp90 inhibitors (i.e. geldanamycin, radicicol, and oxime derivatives of radicicol) possess the ability to modulate the Akt signaling pathway. We found that Hsp90 inhibitors induced Akt dephosphorylation and inactivation in vivo, although they had no direct effects on Akt kinase activity in vitro. Recent reports also documented the downregulation of phospho-Akt level and Akt signaling by Hsp90 inhibitors (18,19). We then examined the effects of Hsp90 inhibitors on the upstream Akt kinases, PDK1 and PI3K, and found that they down-regulated the expression level of PDK1 but did not affect the amount of PI3K. Hsp90 inhibitors had no direct effects on PDK1 kinase activity in vitro. Because binding to Hsp90 has been shown to prevent some signaling proteins from self-association or becoming insoluble (16,17), we investigated the binding capability of PDK1 to Hsp90. We found that PDK1 bound to Hsp90 via its kinase domain and that Hsp90 inhibitors suppressed the binding. Inhibition of PDK1 binding to Hsp90 led to the proteasome-dependent degradation of PDK1. Therefore, we concluded that Hsp90 is involved in the signaling and stability of PDK1.
Transient Transfection, Immunoprecipitation, and Western Blot Analysis-Cells were transfected with appropriate plasmids using Su-perFect transfection reagent (Qiagen) according to the manufacturers' instructions.
Kinase Assay-HT1080 cells were transfected with a pFLAG-CMV-2 vector containing WT-akt cDNA and were cultured for 24 h in medium containing inhibitors. Akt kinase activity of the immunoprecipitated FLAG-tagged Akt was estimated with GSK3 peptide (RPRAATF) as a substrate using an Akt kinase assay kit according to the manufacturer's instructions (Upstate Biotechnology). In some experiments, recombinant human active Akt (Upstate Biotechnology) was incubated with inhibitors for 30 min at 30°C and was then subjected to Akt kinase assay.
We measured PDK1 kinase activity using an assay kit according to the manufacturer's instructions (Upstate Biotechnology). Briefly, recombinant human active PDK1 (Upstate Biotechnology) protein was first incubated with inhibitors for 30 min at 30°C, following incubation with inactive SGK for 30 min at 30°C. Then the PDK1-dependent SGK kinase activity was estimated by incubating the reaction with GSK3 peptide as a substrate for 10 min at 30°C in the presence of [␥-32 P]ATP.
To examine the direct effects of Hsp90 inhibitors on PI3K activity, we transfected pcDNA3-p110␣CAAX and pcDNA3-p110␣CAAX-KD into 293T cells. Then HA-tagged p110␣-CAAX or p110␣-CAAX-KD protein was immunoprecipitated with agarose conjugated with anti-HA-agarose following incubation with Hsp90 inhibitors or LY294002 for 30 min at 30°C. Next, the immunoprecipitated PI3K was subjected to PI3K assay in the presence of these drugs, as described previously (23).
Determination of PDK1 Half-life-293T cells were transiently transfected with pCMV3-PDK1 plasmid, as described above. After 24 h of incubation, the culture medium was replaced with methionine/cysteinefree DMEM (Invitrogen, Grand Island, NY) containing 10% dialyzed fetal bovine serum, and cells were incubated with or without 1 M KF58333 for 2 h. Then cells were labeled with [ 35 S]methionine/[ 35 S]cysteine in fresh methionine/cysteine-free DMEM with or without 1 M KF58333 for an additional 2 h. Radioactive medium was removed, and cells were further incubated in nonradioactive, complete medium for various times indicated before lysis. Exposure to KF58333 was maintained in the treated culture. The amount of radiolabeled PDK1 was determined by immunoprecipitation with anti-Myc agarose and SDS-PAGE. The relative amounts of incorporated radioactivity were visualized and quantified with a BAS1500 Bio-Imaging analyzer (Fuji Film, Tokyo, Japan).

RESULTS
Hsp90 Inhibitors Induced Apoptosis with Caspase Activation-We have recently identified that Akt is associated with Hsp90 (15). To clarify the role of Hsp90 on apoptosis and Akt survival signaling pathway, we examined the effects of several inhibitors of Hsp90. We used benzoquinone ansamycin geldanamycin (16,17), antifungal antibiotic radicicol (25), and analogues of radicicol, KF25706 and KF58333 (26,27), as Hsp90 inhibitors. When 293T cells were treated with Hsp90 inhibitors, 293T cells underwent apoptosis with nuclear fragmentation (data not shown). The caspase family of aspartatespecific cysteine proteases has been demonstrated to be critical mediators in the cell death pathway (28). Among the members of the caspase family, caspase-3/CPP32/Yama/Apopain is a common and important effector of the apoptotic process. To clarify the activation of caspase-3 in vivo, we performed Western blot analysis using antibodies specifically recognizing the p85 cleaved fragment of poly(ADP-ribose) polymerase (PARP). PARP was one of the substrates of caspase-3 and produced the 85-kDa fragment after caspase-3-mediated cleavage (29,30). Treatment of 293T cells with 10 M geldanamycin or 1 M radicicol, KF25706, or KF58333 for 24-h increased the amount of cleaved fragment of PARP (Fig. 1A), indicating that caspase-3 is activated in vivo in 293T cells after inhibition of Hsp90 function. We further examined the activity of caspase-3 (-like) proteases in the KF58333-treated human HT1080 cells using fluorogenic-labeled tetrapeptide DEVD-AMC. KF58333 promoted the activation of caspase-3 (-like) proteases in HT1080 cells in a dose-dependent manner (Fig. 1B). The PARP cleavage in HT1080 cells was also induced by KF58333 in a dose-dependent manner (Fig. 1B). Activation was suppressed by adding caspase inhibitors Z-Asp and Z-VAD (data not shown). These results indicate that inhibition of Hsp90 can initiate the cells to undergo apoptosis.
Inhibition of Akt Kinase Activity in Vivo but Not in Vitro by Hsp90 Inhibitors-Hsp90 inhibitors were known to suppress MAPK signaling pathway by down-regulation of Raf-1 protein expression (31). However, transfection of active MEK1 could not diminish the Hsp90 inhibitor-induced apoptosis (data not shown). Therefore, inhibition of one or more other signaling pathways might be critical for the induction of apoptosis by Hsp90 inhibitors. Akt is recently reported to suppress apoptosis by phosphorylating some apoptosis-relating proteins (i.e. pro-apoptotic Bcl-2 family member Bad at Ser 136 , caspase family member caspase-9 at Ser 196 , the Forkhead transcription factor protein family, and IB kinase) (7,8). We thus examined the Akt kinase activity after treatment with Hsp90 inhibitors. Akt was continuously phosphorylated at Ser 473 residue when 293T cells were cultured in serum-containing medium. Incubation of 293T cells with radicicol, KF25706, or KF58333 induced the dephosphorylation of endogenous Akt at Ser 473 in a dosedependent manner ( Fig. 2A). The decrease in endogenous phospho-Akt (at Ser 473 ) level was also found in human fibrosarcoma HT1080 and mouse colon adenocarcinoma NL-17 cells treated with KF58333 (data not shown), suggesting that Akt dephosphorylation induced by Hsp90 inhibitors is not restricted to one particular cell line. We also examined the change of phospho-Akt (Thr 308 ) level in 293T cells. Because the anti-phospho-Akt (Thr 308 ) antibody is not sensitive enough to detect the phosphorylated form of endogenous Akt protein in 293T cells, we transfected the FLAG-tagged akt cDNA into 293T cells and then examined the effects of KF58333 on phospho-Akt (Thr 308 ) levels. KF58333 treatment promoted the dephosphorylation of Akt at Thr 308 residue in a dose-dependent manner (Fig. 2B). These results indicate that Hsp90 inhibitors suppress the phosphorylation at both Thr 308 and Ser 473 residues, although they also slightly down-regulate the expression of Akt protein itself (Fig. 2, A and B).
To confirm that the dephosphorylation of Akt was associated with the decrease in Akt-mediated signaling, we investigated the change of phosphorylated form of GSK3. Akt is known to phosphorylate GSK3␣ at the Ser 21 residue and GSK3␤ at the Ser 9 residue (7,8). Western blot analysis using an anti-phospho-GSK3␤ (Ser 9 ) antibody revealed that treatment of 293T cells with Hsp90 inhibitors decreased the endogenous phospho-GSK3␤ level (Fig. 2C), consistent with the decrease in phospho-Akt level (Fig. 2, A and B). We also confirmed the decrease in Akt kinase activity by measuring the activity by immunoprecipitating FLAG-tagged Akt from 293T cells that had been transfected with the FLAG-tagged akt cDNA. The transfected 293T cells were treated with Hsp90 inhibitors for 24 h and then harvested. As shown in Fig. 2D, treatment of Hsp90 inhibitors suppressed the Akt kinase activity. Examination of the change of phospho-Akt level by Western blotting confirmed that both geldanamycin and KF58333 almost completely dephosphorylate Akt at the Thr 308 residue. These results indicate that Hsp90 inhibitors down-regulate not only phospho-Akt level but also Akt kinase activity in vivo.
Because Akt itself was reported to be associated with phosphorylation of Akt at the Ser 473 residue (13), we examined the direct effect of KF58333 on Akt kinase activity. The recombinant active Akt protein, which had previously been activated by PDK1, was incubated with KF58333 for 30 min at 30°C in vitro before estimating its kinase activity. We could not observe the inhibitory effects of KF58333 on Akt kinase activity in vitro (Fig. 2E). This result indicates that Akt dephosphorylation and inactivation induced by Hsp90 inhibitors is not the cause of direct inhibition of Akt kinase activity.
Hsp90 Inhibitors Down-regulate PDK1 Protein Expression-Because treatment of Hsp90 inhibitors down-regulate the phospho-Akt level, Hsp90 inhibitors might interfere with the kinase activity of the upstream Akt kinases PI3K and PDK1. We first examined the change of endogenous PI3K and PDK1 protein expression after drug treatment. As shown in Fig. 3A, the treatment of 293T cells with KF58333 decreased the amount of endogenous PDK1 but not p85 subunit of PI3K protein. Down-regulation of PDK1 protein expression was found when 293T cells were treated with KF58333 from the concentration of 30 nM. Raf-1 is well known to interact with Hsp90, and the inhibition of Hsp90-Raf-1 binding results in the decrease in Raf-1 protein expression (31,32). Consistent with the previous report (27), KF58333 suppressed the endogenous Raf-1 protein expression in 293T cells (Fig. 3A). The minimum concentration of KF58333 that could promote Raf-1 destabilization was also 30 nM. Because the level of p85 subunit of PI3K was not affected by KF58333 treatment, the depletion of PDK1 and Raf-1 proteins was not the result of nonspecific protein degradation. Destabilization of endogenous PDK1 and Raf-1 proteins was also observed in human fibrosarcoma HT1080, human prostate PC-3, and mouse colon adenocarcinoma NL-17 cells (Fig. 3B), suggesting that these effects of Hsp90 inhibitors are not restricted to one particular cell line. Moreover, destabilization of PDK1 was also observed in the transfected Myctagged PDK1 protein in vivo (data not shown).
To check directly if the half-life of PDK1 in cells is shortened by treatment with KF58333, we performed pulse-chase experiments. 293T cells that had been transfected with pCMV3-PDK1 were pulse-labeled with [ 35 S]methionine/[ 35 S]cysteine for 2 h and then chased in a nonradioactive, complete medium. KF58333 (1 M) had been added to cells before the start of pulse labeling and included during both pulse-label and chase periods. As shown in Fig. 3C, PDK1 seemed to be relatively stable in control (treated with vehicle Me 2 SO alone) cells. In contrast, After transfection for 24 h, cells were treated with the indicated concentrations of KF58333. After incubation for 24 h, cells were harvested and lysed in lysis buffer for Western blot analysis. The cell lysates were electrophoresed and immunoblotted with an anti-phospho-Akt (Thr 308 ) antibody or an anti-Akt antibody. C, 293T cells were treated as described in Fig. 1A. The cell lysates were electrophoresed and immunoblotted with an anti-phospho-GSK3␤ (Ser 9 ) antibody, an anti-GSK3 antibody, or an anti-Akt antibody. D, 293T cells were transfected with pFLAG-CMV-2 vector encoding WT-Akt. After incubation for 24 h, the transfected cells were treated with medium containing vehicle Me 2 SO (Control), 10 M geldanamycin, or 1 M KF58333. After incubation for an additional 24 h, cells were harvested and lysed in lysis buffer for immunoprecipitation. The FLAG-tagged Akt was immunoprecipitated and was subjected to Akt kinase assay by incubating with GSK3 peptide (RPRAATF) as a substrate for 10 min at 30°C in the presence of [␥-32 P]ATP, as described under "Experimental Procedures." Each point represents a mean Ϯ S.D. of triplicate determinations. The immunoprecipitated proteins were also subjected to immunoblot analysis with an anti-phospho-Akt (Thr 308 ) antibody. E, recombinant active Akt protein (500 ng) was preincubated with the indicated concentrations of KF58333 for 30 min at 30°C in vitro. Then Akt kinase activity was evaluated by incubating with GSK3 peptide (RPRAATF) as a substrate for 10 min at 30°C in the presence of [␥-32 P]ATP, as described under "Experimental Procedures." In the control experiment, the reaction was performed without Akt protein (solid column). Each point represents a mean Ϯ S.D. of triplicate determinations. cpm Ϯ S.D. of the control Akt activity (open column) was 210,789 Ϯ 10,717, and this value was taken as 100%. PDK1 was much more unstable in the KF58333-treated cells. Measurement of radioactivity with a BAS1500 Bio-Imaging analyzer demonstrated that the half-life of PDK1 was shortened down to under 2 h (Fig. 3D). These results indicate that PDK1 requires the molecular chaperone function of Hsp90 for its intracellular stability.
Hsp90 Inhibitors Had No Direct Effects on PDK1 and PI3K Kinase Activity-We then examined the direct effects of Hsp90 inhibitors on PDK1 kinase activity. The recombinant active PDK1 protein was incubated with control (vehicle Me 2 SO alone), 10 M geldanamycin, or 1 M KF58333 for 30 min at 30°C in vitro before estimating its kinase activity. Hsp90 inhibitors exhibited no inhibitory effects on PDK1 kinase activity (Fig. 4A). In contrast, specific PDK1 inhibitor UCN-01 almost completely inhibited the PDK1 kinase activity, as we have recently found. 2 PDK1 phosphorylates itself at the activation loop of PDK1 (Ser 241 ), resulting in its own activation (33). Thus, KF58333 did not affect PDK1 autoactivation. We also examined the direct effects of Hsp90 inhibitors on PI3K activity. Constitutive active PI3K and kinase-dead PI3K were immunoprecipitated from 293T cells that had been transfected with pcDNA3-p110␣CAAX and pcDNA3-p110␣CAAX-KD plasmids. The immunoprecipitated PI3K proteins were preincubated with control (vehicle Me 2 SO alone), geldanamycin (10 M), radicicol (1 M), KF25706 (1 M), KF58333 (1 M), or LY294002 (50 M) for 30 min at 30°C in vitro before estimating its kinase activity. Although PI3K inhibitor LY294002 almost completely suppressed the PI3K activity to the level that seen in KD-PI3K control, all the Hsp90 inhibitors exhibited no inhibitory effects on PI3K (Fig. 4B). These results indicate that Hsp90 inhibitors had no direct effects on both upstream Akt kinases.
We then examined the role of PDK1 down-regulation in Akt phosphorylation in vivo. Mutation of Akt protein at Glu 40 with Lys (E40K) was reported to be activated in a PI3K-independent manner due to its increase in the affinity of the PH domain for phospholipids (34). We estimated the effects of Hsp90 inhibitors on 293T cells that had been transfected with E40K-akt cDNA. As shown in Fig. 4C, E40K-Akt was phosphorylated at the Thr 308 residue even in the presence of PI3K inhibitor LY294002 whereas LY294002 treatment down-regulated the phospho-Akt (Thr 308 ) level of WT-Akt. KF58333 decreased the phospho-Akt (Thr 308 ) level in both WT-Akt and E40K-Akt, indicating that destabilization of PDK1 plays an important role in KF58333-mediated Akt dephosphorylation. These results indicate that PDK1 destabilization is the main reason for Akt dephosphorylation and inactivation induced by Hsp90 inhibitors.
Association of PDK1 with Hsp90 in Vivo-Because most Hsp90 client proteins were destabilized after inhibiting the binding to Hsp90 (16,17), we hypothesized that PDK1 might be one of the Hsp90 client proteins. We checked the in vivo PDK1 binding to Hsp90 by immunoprecipitating endogenous PDK1 protein following Western blot analysis with an anti-Hsp90 antibody. As shown in Fig. 5A, endogenous Hsp90 protein was co-immunoprecipitated by an anti-PDK1 antibody but not by control sheep antibody. By transfection of WT-PDK1 cDNA in a pFLAG-CMV-2 vector and WT-hsp90 cDNA in a pcDNA3.1/GS vector into 293T cells, we confirmed the binding of FLAGtagged PDK1 to V5-tagged Hsp90 in cells (Fig. 5C). To identify the binding site in PDK1, we prepared NH 2 -terminal and COOH-terminal deletion mutants, as indicated in Fig. 5B. The V5-tagged Hsp90 and WT-PDK1 or deleted PDK1 were expressed in 293T cells. The V5-tagged Hsp90 could interact with WT-, ⌬N51-, ⌬PH-, ⌬C342-, ⌬C273-, ⌬C242-, and ⌬N155-PDK1 (Fig. 5C). Hsp90 could hardly interact with ⌬C196-PDK1 and could not bind to ⌬N223-PDK1. The relationship between the Hsp90-binding ability and the structure of PDK1 deletion mutants is summarized in Fig. 5B. These results suggest that amino acid residues around 156 -223 of PDK1 (active site of PDK1, dotted area in Fig. 5B) is involved in its binding to Hsp90.
Then we investigated the effects of Hsp90 inhibitors on PDK1 binding to Hsp90. We transfected WT-PDK1 cDNA in a pFLAG-CMV-2 vector, WT-hsp90 cDNA in a pcDNA3.1/GS vector, or both into 293T cells. After transfection for 24 h, cells were treated with 1 M KF58333 for 2 h following immunoprecipitation with an anti-Myc (PDK1) antibody. As shown in Fig.  5D, PDK1 binding to Hsp90 was inhibited by treatment with KF58333. This result indicates that Hsp90 inhibitors can suppress PDK1 binding to Hsp90. Interestingly, ⌬C196-PDK1 and ⌬N223-PDK1 were hardly expressed in 293T cells (Fig. 5C). We speculate that these mutants were unstable and degraded because of their inability to bind to Hsp90, as in the case of WT-PDK1 in Hsp90 inhibitor-treated cells (Fig. 3). These results indicate that PDK1 requires Hsp90 association for its stability.
Role cysteine protease inhibitor (leupeptin), chymotrypsin-like serine protease inhibitor (TPCK), or trypsin-like serine protease inhibitor (TLCK). After 18 h of incubation, cells were harvested and lysed with lysis buffer containing 0.5% Triton X-100. Western blot analysis with an anti-PDK1 antibody revealed that KF58333-mediated PDK1 degradation was partially blocked by proteasome inhibitors and not by other protease inhibitors (Fig.  6A). Quantification of the soluble PDK1 amount revealed that proteasome inhibitors increased the level of soluble PDK1 by 1.5-to 2-fold (Fig. 6B). Because proteasome inhibitors could not perfectly protect PDK1 from degradation, we examined the amounts of PDK1 in detergent-insoluble fractions recovered from centrifuge precipitates. Treatment of the cells with proteasome inhibitors drastically increased PDK1 amount in the insoluble fractions in the presence of KF58333 (Fig. 6A). For control experiments, we examined the effects of proteasome inhibitors on insoluble PDK1 accumulation in the absence of Hsp90 inhibitors. We found that the treatment of the cells with proteasome inhibitors alone slightly increased the accumulation of insoluble PDK1 (Fig. 6A). Blockade of endogenous PDK1 turnover might be the reason for the slight increase in insoluble PDK1 level. Because proteasome inhibitors drastically increased the insoluble PDK1 amount in the presence of Hsp90 inhibitors, PDK1 degradation in the absence of Hsp90 function was carried out by the proteasome system.
Proteasome inhibitors could not completely restore the phospho-Akt level and could not fully suppressed KF58333-induced apoptosis (Fig. 6A), whereas proteasome inhibitors drastically increased the insoluble PDK1 amount. Moreover, we found that insoluble PDK1 was dephosphorylated at the Ser 241 residue (data not shown), which is essential for PDK1 activation (33). These results suggest that insoluble PDK1 is inactive. Therefore, Hsp90 is not just a shield against degradation but is a chaperone to keep PDK1 in a soluble and in an active conformational state prior to phosphorylating its downstream substrates.
We examined the effects of PDK1 overexpression on KF58333-induced apoptosis. Unfortunately, overexpression of ⌬N51-PDK1 alone could not inhibit KF58333-induced apoptosis (Fig. 6C). When examining the phospho-Akt level in PDK1- transfected cells, we found that overexpression of PDK1 alone restored the phosphorylation of the Thr 308 residue (data not shown) but not of Ser 473 (Fig. 6C). Because phosphorylation at both Thr 308 and Ser 473 residues is necessary for full activation of Akt, PDK1 transfection could not inhibit KF58333-induced apoptosis. To examine the role of the PDK1-Akt signaling pathway during KF58333-induced apoptosis, we transfected the constitutive active form of akt cDNA (myr-Akt) and found that overexpression of active Akt suppressed KF58333-induced apoptosis (Fig. 6C). The result suggests that inhibition of the PDK1-Akt survival signaling pathway is involved in the KF58333-induced apoptosis. DISCUSSION The sensitivity of cells to apoptosis-inducing stimuli appears to be dependent on the balance between apoptosis-inducing signals and survival signals. Akt is well known to transmit the survival signals by phosphorylating its downstream substrates (7,8). The akt gene was originally identified as the cellular counterpart of the retroviral oncogene v-akt, which was present in AKT8, a retrovirus that caused T-cell lymphoma in mice. The catalytically inactive Akt exists in the cytosol. By activation of PI3K, PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 are synthesized at the plasma membrane, and Akt is recruited to the plasma membrane. Interaction of Akt to the lipids induces a conformational change of Akt. Then Akt is phosphorylated at two key regulatory sites, Thr 308 in the activation loop of the catalytic domain and Ser 473 in the COOH-terminal domain (7,8). Dual phosphorylation at both residues is necessary for full activation of Akt. Bad is one of the substrates of Akt and is phosphorylated at the Ser 136 residue. Phosphorylation of Bad prevented apoptosis by inducing the association with 14-3-3 protein isoforms. The increase in phospho-Bad/14-3-3 interaction sequestered Bad from anti-apoptotic Bcl-2 or Bcl-x L , which leads to the formation of the Bcl-2/Bax or Bcl-x L /Bax heterodimers. Akt also phosphorylates caspase family member caspase-9 at the Ser 196 residue and prevents its proteolytic activity. The antiapoptotic function of Akt was also mediated by phosphorylating and modulating the function of Forkhead transcription factors and IB kinase.
Hsp90 is an abundant and highly conserved protein involved in a diverse array of cellular processes. It is present in all species studied, from Escherichia coli to humans, with 40% amino acid identity. In contrast to other heat-shock proteins, Hsp90 is not required for maturation or maintenance of most proteins in vivo. By interacting with a wide variety of signaling proteins, Hsp90 prevents these Hsp90 client proteins from degradation, self-association, becoming active at an inappropriate time, or incorrect localization (16,17). Once Hsp90 substrates fail to be correctly folded, Hsp90 promotes the degradation of the misfolded substrates (35). Hsp90 might promote the degradation of the misfolded substrates by association with an ubiquitin ligase carboxy terminus of Hsc70-interacting protein (CHIP) (36). Therefore, Hsp90 also plays an important role in the maintenance of protein quality by regulating the balance between protein folding and degradation. Because Hsp90 is constitutively overexpressed at 2-to 10-fold higher levels in tumor cells (37), Hsp90 is thought to be a critical regulator in tumor growth and survival. Our previous work (15) demonstrated that Akt kinase activity is also regulated by binding to Hsp90. Because the binding to Hsp90 protects Akt from PP2Amediated dephosphorylation and increases Akt-mediated survival signals, overexpression of Hsp90 might protect tumor cells from undergoing apoptosis by increasing survival signals.
Hsp90 inhibitors (geldanamycin, radicicol, and their derivatives) have been reported to bind to Hsp90 and to affect Hsp90 function (16,17,(25)(26)(27). Moreover, Hsp90 inhibitors have been shown to possess anti-tumor activity in preclinical models. Hsp90 inhibitors might exhibit their anti-tumor activity by destabilizing the Hsp90 client proteins important in cancer cell proliferation (i.e. steroid receptors, Raf-1, Src family kinases, p53 mutant, cyclin-dependent kinases, and so on). As we have shown in Fig. 1, Hsp90 inhibitor had the ability to induce apoptosis with caspase activation. Therefore, it is possible that Hsp90 protects cells from undergoing apoptosis by promoting stabilization of unidentified Hsp90 client proteins that are critical for transmitting survival signals. Because Hsp90 forms a complex with Akt, we hypothesized that Hsp90 inhibitors would interfere with the Akt signaling pathway. As we have previously reported (15), exposure of the cells to geldanamycin did not inhibit the Hsp90-Akt binding. However, long-term exposure of geldanamycin suppressed the Hsp90-Akt complex formation and decreased slightly the amount of Akt (data not shown, Fig. 2). During the experiments, we found that Hsp90 inhibitors strongly promoted dephosphorylation and inactivation of Akt (Fig. 2). Recently, similar results were also reported that Hsp90 inhibitors could down-regulate phospho-Akt level and Akt signaling (18,19). Because Hsp90 inhibitors suppressed the phosphorylation of Akt, we investigated the amount of upstream Akt kinases, PI3K and PDK1. Many Hsp90 client proteins are degraded after treatment with Hsp90 inhibitors (16,17). Thus, we examined the expression level of PI3K and PDK1 after Hsp90 inhibitor treatment. Screening of the Hsp90 client proteins in the Akt signaling pathway revealed that PDK1 might be the critical, missing Hsp90 client protein for transmitting survival signals, because treatment of the cells with Hsp90 inhibitors decreased PDK1 stability (Fig.  3) and overexpression of the active form of Akt suppressed Hsp90 inhibitor-induced apoptosis (Fig. 6C). We also found that a proteasome-dependent pathway was involved in the Hsp90 inhibitor-induced PDK1 degradation and that Hsp90 functioned as a chaperone to keep PDK1 in a soluble and an active conformational state (Fig. 6). In contrast, Hsp90 inhibitors did not affect PI3K protein expression (Fig. 3A). Moreover, Hsp90 inhibitors showed no direct effects on PI3K activity in vitro (Fig. 4B). Therefore, PDK1 destabilization might be the main reason for the decrease in Akt-mediated survival signaling pathway. This notion was supported by the fact that KF58333 possessed the ability to decrease the amount of phospho-E40K-Akt that preferentially localized in plasma membrane in a PI3K-independent mechanism (Fig. 4C).
It is now clear that PDK1 plays a central role in the activation of the AGC subfamily of protein kinases. PDK1 phosphorylates all members of the AGC kinases at residues equivalent to Thr 308 of Akt (known as activation loop or T-loop) (8). PDK1 has been shown to be able to phosphorylate p70 S6K, PKC isoforms, SGK isoforms, and protein kinase A. PDK1 itself is also a member of the AGC subfamily of protein kinases. Like other members of the AGC kinases, PDK1 phosphorylates itself at the activation loop of PDK1 (Ser 241 ), which results in its own activation (33). Because Hsp90 inhibitors had no inhibitory effects on PDK1 kinase activity (Fig. 4A), Hsp90 inhibitorinduced Akt inactivation was not the cause of direct PDK1 inhibition. These results imply, again, that PDK1 degradation is the main reason for the Hsp90 inhibitor-induced Akt dephosphorylation.
Our experiments with PDK1 deletion mutants indicate that the catalytic domain of PDK1 (amino acid residues around 156 -223) is responsible for binding to Hsp90 (Fig. 5). Akt is also bound to Hsp90 via its catalytic domain (amino acid residues 229 -309) (15). Moreover, recent reports have also suggested that Hsp90 binds to the catalytic domain of Raf, pp60 v-src , male germ cell-associated kinase (MAK), MAK-re-lated kinase (MRK), and MAPK/MAK/MRK overlapping kinase (MOK) (24). However, MAPK, p38, and JNK/SAPK could not bind to Hsp90 despite their high homologies to Hsp90-interacting kinases (24). Consistent with this finding, Western blot analysis of the MAPK protein revealed that treatment of Hsp90 inhibitors did not affect the expression level of MAPK while the treatment decreased the phospho-MAPK amount ( Fig. 2A). Therefore, not all protein kinases could bind to Hsp90.
Four lines of evidences suggest that the Akt-mediated survival pathway is an attractive target for cancer chemotherapy (38,39). First, the Akt pathway is relatively inactive in resting cells; second, amplification of the akt gene occurs in some tumors; third, increased Akt kinase activity contributes to tumor progression in prostate cancer; fourth, loss of the tumor suppressor phosphatase and tensin homologue deleted on chromosome 10 (PTEN) is common in tumors, and its loss constitutively activates Akt. Some recent reports, including ours, also demonstrate that several anti-tumor agents exhibit their cytotoxicity by directly or indirectly modulating Akt signaling pathway (21,40), 2 suggesting that the PDK1-Akt survival signaling pathway is very important for tumor growth control and providing a rationale for developing new Hsp90 inhibitors that could specifically interfere with PDK1-Hsp90 complex formation, thus causing tumor cells to undergo apoptosis.