Hsp90 Regulates the Phosphorylation and Activity of Serum- and Glucocorticoid-regulated Kinase-1*

SGK-1 (serum- and glucocorticoid-regulated kinase-1), a member of the AGC protein kinase family, plays an important role in regulating ion channel expression and contributes to malignant epithelial cell proliferation and survival. SGK-1 activity is regulated on three levels: transcriptional induction following a variety of environmental and intracellular stresses, proteasomal degradation, and phosphorylation. Here we report that phosphoinositide 3-kinase (PI3K)-dependent phosphorylation of SGK-1 requires formation of a complex between SGK-1 and heat-shock protein 90 (Hsp90). Inactivation of Hsp90 by geldanamycin led to decreased SGK-1 phosphorylation independently of increased proteasomal protein degradation, and inhibition of PI3K activity by LY294002 appeared to eliminate SGK-1 phosphorylation at the same residues as those affected by geldanamycin treatment. Interestingly, geldanamycin-targeted phosphorylation sites were not limited to the known conserved PI3K-dependent sites Thr-256 and Ser-422 in SGK-1 but included additional unknown PI3K-dependent residues. Inhibition of Hsp90 also resulted in a complete loss of SGK-1 kinase activity, suggesting that Hsp90 activity is essential for regulating the PI3K/SGK-1 pathway.

To maintain homeostasis, an effective response to environmental challenge is required of all living organisms. Stress-response mechanisms have evolved at both the organismal and cellular levels. One of the important effectors of the cellular stress response that has been identified in several cell types as well as in both simple and complex eukaryotic organisms is SGK-1 (serum-and glucocorticoid-regulated kinase-1). SGK-1 belongs to the "AGC family" of serine-threonine kinases that includes protein kinase B (also known as Akt1), p70 S6 kinase, and protein kinase C. Unlike other members of the AGC family, SGK is transcriptionally induced following various forms of cellular stress, including hyperosmolar or heat shock, UV irradiation, and also following steroid receptor activation (1). After induction, SGK-1 is phosphorylated via a phosphoinositide 3-kinase (PI3K) 3 -dependent pathway at residues Thr-256 and Ser-422 (2,3). These phosphorylation events activate SGK-1 kinase activity thereby initiating a potent survival signal in epithelial cells; the mechanism downstream of SGK-1 activation is mediated in part via phosphorylation and inactivation of the pro-apoptotic transcription factor FOXO3a (4,5). Recent findings have similarly demonstrated that SGK-1 functions as a critical component of insulin signaling in Caenorhabditis elegans via phosphorylation and inactivation of the FOXO3a homolog Daf 16 (6,7). In addition, SGK-1 regulates the activity and abundance of ion channels in the lung (8,9), kidney (10), and heart (11,12), thereby integrating sodium homeostasis, fluid balance, blood pressure, and heart function.
In addition to transcriptional induction following a physiological stressor, SGK-1 protein levels are tightly regulated via ubiquitin modification and rapid proteasome-mediated degradation (13). We and others have shown that the endoplasmic reticulum (ER)-associated E3 ligase C terminus of the Hsc70interacting protein (CHIP) (14) and HRD1 (15) are involved in ubiquitin modification of SGK-1 and contribute to the rapid return of SGK-1 protein expression to base-line levels during recovery from stress. Because many stress-induced proteins require molecular chaperones for proper folding and activity, we hypothesized that SGK-1 activity might be regulated by chaperone interaction. Previously, we showed that heat shock protein 70 (Hsp70) forms a complex with SGK-1 and CHIP (14); here we present evidence that phosphorylation and activity of SGK-1 are also regulated via the ubiquitous molecular chaperone heat shock protein 90 (Hsp90).
Hsp90 has recently been linked to the E3 ligase CHIP through their mutual association with several common stressresponse proteins. CHIP is a heat shock protein-dependent E3 ligase that participates in protein quality control through ubiquitination and targeting of misfolded proteins to the proteasome (42)(43)(44). The fact that many CHIP targets are also Hsp90client proteins suggests that both molecular chaperone and protein quality control mechanisms work together to maintain proper protein folding. For example, both Hsp90 and CHIP complex with the cystic fibrosis transmembrane conductance regulator (CFTR), but each protein performs a distinct function; Hsp90 binding stabilizes CFTR in a properly folded and mature state, whereas CHIP targets misfolded immature CFTR proteins for ubiquitin modification and proteasome-mediated degradation (ERAD) (45)(46)(47)(48).
In this study we show for the first time that SGK-1 forms a functional complex with Hsp90 that is disrupted following treatment with the Hsp90 inhibitor, geldanamycin (GA). Furthermore, we demonstrate that Hsp90 is required for PI3K-dependent phosphorylation of SGK-1, and that the phosphorylated residues appear to include both previously described Thr-256 and Ser-422 as well as additional unknown amino acid(s). Furthermore, Hsp90-dependent phosphorylation plays a critical role in SGK-1 function because the rapid loss of phosphorylated SGK-1 species following GA treatment completely abrogates SGK-1 kinase activity. Unlike many of the Hsp90 client proteins, including Akt1, SGK-1 is not degraded following GA treatment. In addition, Akt1 is transiently phosphorylated following GA treatment (49), whereas inactivation of Hsp90 by GA leads to a very rapid loss of SGK-1 phosphorylation and subsequent complete inactivation. Taken together, these results suggest that the stress-induced PI3K/SGK-1 pathway is positively regulated by Hsp90 consistent with the distinctive role of SGK-1 in maintaining physiological homeostasis following acute cellular stress.

EXPERIMENTAL PROCEDURES
Cell Culture and Expression Vectors-SK-BR-3 and MDA-MB-231 human breast cancer cell lines and COS-1 cells were obtained from American Type Culture Collection (ATCC, Manassas, VA). Cells were cultured in Dulbecco's modified Eagle's medium (Lonza Walkersville Inc., Walkersville, MD) supplemented with 10% heat-inactivated fetal calf serum (Gemini Bio-Products, West Sacramento, CA) and 1% penicil-lin/streptomycin (Lonza Walkersville Inc.). Cloning of fulllength SGK-1 in-frame with a C-terminal FLAG tag into the retroviral vector pLPCX (Clontech) and creation of SGK-1 mutations S422D, T256A/S422A, and K127M (kinase-dead) SGK-1-FLAG were described previously (13,50). The mutation of serine 78 to alanine (S78A) was created using the Quick-Change site-directed mutagenesis kit (Biocrest Manufacturing, West Cedar Creek, TX) using the following primers: forward, 5Ј-CCT TCT CCT CCA CCA GCT CCT TCT CAG CAA ATC-3Ј, and reverse, 5Ј-GAT TTG CTG AGA AGG AGC TGG TGG AGG AGA AGG-3Ј (51). Myc-tagged constitutively active p110 CAAX was a kind gift of Dr. J. Downward of Cancer Research UK London Research Institute. HA-ubiquitin in the pRBG4 vector was described previously (13). Constitutively active Mek5DD was a kind gift of Dr. Mark Abe of the University of Chicago.
Endogenous Hsp90 was immunoprecipitated from either SK-BR-3 cells transiently expressing SGK-1-FLAG and treated with ALLN overnight or subconfluent MDA-MB-231 cells that were starved of all growth factors for 60 h and then stimulated for 10 h with 1 M dexamethasone, 20% serum, and ALLN. Immunoprecipitation was performed as described previously (52). Briefly, cells were lysed in 25 mM MOPS, pH 7.5, 1 mM EDTA, 0.02% NaN 3 , and 10% glycerol (MENG) containing the same protease/proteasome/phosphatase inhibitor mixture as described above. Lysates were passed 10 times through a 25-gauge needle and then centrifuged for 10 min at 10,000 ϫ g, 4°C. Supernatants were then precleared for 2 h at 4°C with 4 g of mouse IgM (Zymed Laboratories Inc.), 2 g of goat antimouse (GAM) IgM antibody (Zymed Laboratories Inc.), and 20 l of protein G-agarose (Sigma). For endogenous SGK-1 coprecipitation, supernatants were precleared for 30 min at 4°C with 4 g of TEPC-183 mouse IgM (tetramethylpentadecaneinduced plasmacytoma; Sigma), 2 g of GAM, and 25 l of protein G-agarose. Agarose was pelleted by centrifugation for 2 Hsp90 Regulates SGK-1 min at 300 ϫ g, and supernatant protein concentrations were determined by Bradford assay. Equal amounts of protein (typically, 1-2 mg) were incubated on a rotator overnight at 4°C with 4 g of anti-Hsp90 antibody MA3-011 (ABR Affinity Bioreagents, Golden, CO), then for another 1 h with fresh GAM, and finally for 1 h with protein G-agarose. Agarose beads were pelleted for 2 min at 300 ϫ g and then washed three times in MENG buffer supplemented with the inhibitor mixture and 1% Nonidet P-40. Proteins were extracted from the beads by adding equal volume of 2ϫ SDS-PAGE sample buffer and boiling for 4 min before SDS-PAGE analysis. For endogenous SGK-1 co-precipitation, immunoprecipitation was performed as above except anti-Hsp90 or TEPC-183 control antibodies were incubated for 2 h, and precipitated proteins were washed with MENG buffer containing 50 mM NaCl.
In Vivo Labeling-SK-BR-3 cells (7 ϫ 10 6 ) transiently expressing either wild type or T256A/S422A SGK-1-FLAG were starved of all growth factors overnight, additionally starved of phosphate for 2 h with either 1 M geldanamycin (GA, Sigma) or 50 M LY294002 (LC Laboratories, Woburn, MA), and then stimulated with 20% serum for 4 h in the presence of 1 mCi of [ 32 P]orthophosphoric acid (PerkinElmer Life Sciences and Analytical Science, Shelton, CT), 10 M MG 132, and either GA or LY294002. Cells were collected and lysed for a denaturing immunoprecipitation. Briefly, cell pellets were lysed in 100 l of 50 mM Tris-HCl, pH 7.5 containing 1% SDS, 5 mM DTT, sheared 10 times with a 23-gauge needle, boiled 10 min, and then centrifuged 5 min at room temperature. To each supernatant, 1.2 ml of 50 mM Tris-HCl, pH 7.5, containing 250 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40, and inhibitors listed above were added. Lysates were then subjected to the standard FLAG immunoprecipitation protocol except that precipitated proteins were eluted using SDS-PAGE sample loading dye.
In Vivo Ubiquitination Assay-SK-BR-3 cells stably expressing SGK-1-FLAG were transfected with a plasmid encoding HA-ubiquitin and treated with 1 M GA. SGK-1 was immunoprecipitated from equal amounts of protein lysate. SGK-1-FLAG species were then immunoblotted with either anti-HA-HRP or anti-SGK-1 antibodies to detect both ubiquitinated and unmodified SGK-1 species, respectively.
Inhibitor Treatments-SK-BR-3 or MDA-MB-231 cells were treated with 1 M GA or vehicle (DMSO) for 3 h or as otherwise indicated. In some experiments, 50 nM okadaic acid (EMD Chemicals, Inc.) was added for 2 h prior to GA treatment. In others, either 10 M ALLN or MG132 was added during GA treatment for 16 or 2.5 h, respectively. To inhibit potential kinases acting upstream of SGK-1, either 100 M U0126 (Cell Signaling Technology), 10 M SB203580 (Sigma), 10 M JNK inhibitor II (EMD Chemicals, Inc.), 100 nM staurosporine (Sigma), or 50 M LY294002 (LC Laboratories, Woburn, MA) was added to cells stably expressing SGK-1-FLAG for 3 h.

Identification of Hsp90 as an SGK-1-associated Protein-In
comparison with the regulation of its closely related family member, Akt1, modulation of SGK-1 activity is less well understood. To identify SGK-1-associated proteins, we immunoprecipitated ectopically expressed SGK-1-FLAG, resolved the co-precipitated proteins on a denaturing gel, and used MALDI-TOF mass spectrometry to identify silver-stained proteins that specifically co-precipitated with SGK-1. As seen in Fig. 1A, an ϳ90-kDa protein appeared to be specific to the SGK-1 lane; sequence analysis of eight peptides from this band identified it as Hsp90.
To confirm that Hsp90 forms a specific complex with SGK-1, we performed immunoprecipitation of ectopically expressed SGK-1-FLAG from membrane and cytosolic fractions of SK-BR-3 cells using an anti-FLAG antibody followed by immunoblotting with an anti-Hsp90 antibody. As shown in Fig. 1B, SGK-1 was found in both the cytosolic (S100) and membrane (P100) fractions, consistent with our previous observations (13). However, endogenous Hsp90 associated with SGK-1 primarily in the cytosolic fraction, consistent with the predominantly cytosolic localization of Hsp90 (53).
To further investigate the specificity of association of SGK-1 with Hsp90, we performed the reciprocal immunoprecipitation of endogenous Hsp90 from equal amounts of SK-BR-3 cell protein lysates. Co-precipitated proteins were resolved with SDS-PAGE and examined by immunoblotting using an anti-FLAG antibody to detect SGK-1. As shown in Fig. 1C, SGK-1 was detected in a complex with Hsp90, providing further evidence that Hsp90 associates with SGK-1.
Finally, to determine whether endogenous Hsp90 and SGK-1 associate, we immunoprecipitated Hsp90 from MDA-MB-231 cells after inducing SGK-1 expression with serum and glucocorticoid stimulation. Fig. 1D illustrates the specific immunoprecipitation of Hsp90 followed by detection of endogenous SGK-1 using an anti-SGK-1 antibody directly conjugated to horseradish peroxidase. When SGK-1 was not induced by serum and glucocorticoid stimulation, the interaction could be barely detected, presumably because of the very low levels of SGK-1. However, in stimulated cells, endogenous SGK-1 clearly co-precipitates with Hsp90. Thus, we show here that both endogenous and ectopic SGK-1 form an in vivo complex with endogenous Hsp90.
Hsp90 Inhibition Abolishes Slow Mobility Species of SGK-1 Independently of Proteasome-mediated Degradation-Hsp90specific inhibitors have been shown to be useful tools for studying Hsp90 function. An irreversible Hsp90 inhibitor, GA, binds specifically to the ATP-binding pocket of Hsp90, which has a unique conformation compared with other chaperones (e.g. Hsp70) (54). GA thereby interferes specifically with the ATPdependent association of Hsp90 and its client proteins (55). Because the affinity of GA for Hsp90 is within the micromolar range (56), we examined whether inhibition of Hsp90 by 1 M GA would result in its dissociation from SGK-1. We treated SK-BR-3 cells stably expressing SGK-1-FLAG with GA for 3 h and then performed an anti-FLAG immunoprecipitation. As seen in Fig. 2A, Hsp90-␤, as well as Hsp90-␣ (data not shown), co-immunoprecipitated with SGK-1. However, following treatment with GA, Hsp90 was no longer detected in the SGK-1 immunoprecipitation, further suggesting that SGK-1 and Hsp90 binding in vivo is ATP-dependent. Notably, GA treatment resulted only in the loss of the slowly migrating bands of SGK-1, suggesting that Hsp90 inhibition affects the post-translational modification of SGK-1.
Hsp90 has been reported to stabilize a wide variety of client proteins (for review see Ref. 57). We therefore examined the effect of GA on endogenous SGK-1 steady-state levels in MDA-FIGURE 1. Hsp90 associates with SGK-1. A, SGK-1-FLAG was immunoprecipitated from SK-BR-3 cells, resolved on an SDS-polyacrylamide gel, and stained with silver, and co-precipitated Hsp90 was identified using MALDI-TOF as the peptide sequences as indicated. B, SK-BR-3 cell lysates were fractionated into cytosolic (S100) and membranous (P100) fractions. SGK-1-FLAG was immunoprecipitated and precipitated proteins were immunoblotted using anti-Hsp90 and anti-FLAG antibodies. C, endogenous Hsp90 was immunoprecipitated from SK-BR-3 cell lysates, and ectopic SGK-1-FLAG was immunoblotted using anti-FLAG antibody. D, endogenous Hsp90 was immunoprecipitated from MDA-MB-231 cells starved for 48 h and either stimulated with 1 M dexamethasone (Dex) and 20% serum or vehicle for 10 h. Endogenous SGK-1 was detected using an anti-SGK-1-HRP antibody. Molecular markers are indicated. IP, immunoprecipitation, IB, immunoblotting.
MB-231 breast cancer cells. To induce SGK-1 expression, we starved cells for 24 h prior to overnight treatment with 1 M dexamethasone and 20% serum. In parallel, cells were treated with ALLN for 4 h prior to cell lysis. Fig. 2B (left panel) shows that endogenous SGK-1 levels significantly increased following ALLN treatment, consistent with the known proteasome-mediated degradation of SGK-1 (13,50). Between 3 and 6 h, GA caused the complete disappearance of slow mobility SGK-1 species (Fig. 2B, right panel). Surprisingly, ALLN did not rescue the disappearance of these slow mobility species, suggesting that increased protein degradation was not the cause of the GA-induced loss of the upper species. At a later time point of GA treatment (22 h), an overall reduction in endogenous SGK-1 levels was seen that could be due to the predicted loss of glucocorticoid receptor expression leading to decreased transcriptional induction of SGK. Similar results were obtained in SK-BR-3 breast cancer cells (data not shown).
We next investigated whether the more specific proteasome inhibitor MG132 could reverse the disappearance of the slowly migrating species of SGK-1. SK-BR-3 cells transiently expressing SGK-1-FLAG were treated with GA for 6 and 9 h. Parallel cell cultures were treated with or without MG132 for 2.5 h before each collection time point. SGK-1-FLAG accumulated following MG132 treatment (Fig. 2C, left), which is consistent with accumulation of endogenous SGK-1 following proteasome inhibition. Once again, the slow mobility species of SGK-1 largely disappeared following 3 h of GA treatment either in the presence or absence of MG132 (Fig. 2C, right). In addition, we observed that GA treatment increased the accumulation of the faster mobility species of SGK-1, suggesting that these species are not degraded following GA treatment. Similar results were obtained in HEK 293 cells (data not shown). Taken together, these data suggest that the slow mobility species of both endogenous and ectopic SGK-1 are affected by GA in a similar manner in different cell types, and that their rapid disappearance is not the result of an increase in proteasome-mediated protein degradation.
Despite the lack of proteasome-mediated degradation of SGK-1 following GA treatment, we hypothesized that GA might still affect SGK-1 ubiquitin modification (58 -61). We therefore examined whether the efficiency of the ubiquitinmodification of SGK-1 was altered following GA treatment. SK-BR-3 cells stably expressing SGK-1-FLAG were transiently transfected with a plasmid encoding HA-ubiquitin. Cells treated with GA or vehicle were lysed, and SGK-1 was immunoprecipitated. Levels of ubiquitinated SGK-1 were determined by immunoblotting with anti-HA antibody. As shown in Fig. 2D, slowly migrating species of immunoprecipitated SGK-1 disappeared following 1 h of GA treatment (bottom panel, lane 8). However, overall SGK-1 ubiquitination increased after 2 h of GA treatment (Fig. 2D, upper panel, lane  9). Therefore, the disappearance of the upper band of SGK-1 (1 h) precedes the increase in overall ubiquitin modification (2 h), suggesting that the loss of the slowly migrating SGK-1 species is not a result of increased ubiquitin modification. Following longer GA treatment, the accumulation of the faster migrating species of SGK-1 (Fig. 2D, bottom panel, lanes 9 -12) was accompanied by an increase in ubiquitin-modified SGK-1 (upper panel, lanes 9 -12). The ratios of ubiquitinated versus unmodified SGK-1 actually decreased over time as measured by densitometry (data not shown), suggesting that inhibiting Hsp90 activity decreases the efficiency of SGK-1 ubiquitina-

Hsp90 Regulates SGK-1
tion. Moreover, cycloheximide chase of ectopically expressed SGK-1 in SK-BR-3 cells did not reveal a significant difference in the stability of already synthesized SGK-1 following GA treatment (data not shown). Taken together, these results strongly suggest that the GA-mediated disappearance of the slowly migrating SGK-1 species is independent of both increased ubiquitin modification and proteasome-mediated degradation. Therefore, it is unlikely that Hsp90 is required for SGK-1 stability. We hypothesized that the mechanism whereby inhibition of Hsp90 function leads to increased unmodified SGK-1 could be related to a loss of phosphorylated species and the "collapse" of these bands into unmodified forms.
Hsp90 Regulates PI3K-dependent Phosphorylation of SGK-1-To examine whether the slowly moving SGK-1 species are phosphorylated, SK-BR-3 cells stably expressing SGK-1-FLAG were lysed, and equal amounts of lysates were treated for 3 h with either serine/threonine-specific protein phosphatase or tyrosine-specific protein phosphatase (YOP). Reactions were stopped with 2ϫ loading dye and then analyzed by immunoblotting of SGK-1. As shown in Fig. 3A, SGK-1 typically separated into four major bands, which remained throughout the duration of the experiment, suggesting that SGK-1 was not simply degraded over time (lanes 1 versus 2). In contrast, phosphatase treatment for 3 h completely abrogated the upper band of SGK-1 similar to the pattern observed following GA treatment in vivo (Fig. 3A, lanes 3 versus 5). This suggests that the slowly migrating SGK-1 species are phosphorylated at serine and/or threonine residues. Interestingly, the protein-tyrosine phosphatase YOP did not have a significant effect on the phosphorylated SGK-1 banding pattern, suggesting that phosphorylated tyrosines do not contribute to the upper bands.
To further investigate Hsp90-dependent phosphorylation of SGK-1, we directly examined whether Hsp90 inhibition interferes with in vivo SGK-1 phosphorylation. We labeled SK-BR-3/SGK-1-FLAG cells with [ 32 P]orthophosphate in the presence of either vehicle or GA, immunoprecipitated SGK-1, and examined the levels of either total or phosphorylated SGK-1 using immunoblotting and autoradiography, respectively. As seen in Fig. 3B, the slowly migrating SGK-1 species had electrophoretic mobility similar to the radiolabeled SGK-1 (lanes 1 and 3), strongly suggesting that these species are phosphorylated. In the presence of GA, these upper radiolabeled SGK-1 species almost completely disappeared (Fig. 3B, lane 4) consistent with the depletion of slow moving species (lane 2). These results strongly suggest that Hsp90 activity is required for SGK-1 phosphorylation in vivo.
It has been shown that Hsp90 can enhance phosphorylation of Akt1 by inhibiting Akt1-directed phosphatase activity (62)(63)(64). We therefore examined whether the phosphorylated SGK-1 species that disappear following GA treatment might consist of differentially phosphorylated species of SGK-1 that are phosphatase-sensitive. We treated SK-BR-3 cells stably expressing SGK-1-FLAG with 50 nM of the phosphatase inhibitor okadaic acid (OA) for 2 h followed by GA treatment. Whole cell lysates were collected and analyzed by immunoblotting. As shown in Fig. 4A, 30 min of GA treatment completely depleted the slowly migrating phospho-SGK-1 bands, and longer treatment caused further accumulation of the faster migrating, unmodified SGK-1 bands (Fig. 4A, top). As expected, OA, which inhibits both protein phosphatases 1 and 2A (PP1 and PP2A) at the concentrations used, caused a relative accumulation of phosphorylated SGK-1 over time as seen in Fig. 4A (bottom left). However, OA did not rescue the disappearance of the phosphorylated species of SGK-1 following GA treatment (Fig.  4A, bottom right), suggesting that GA does not increase phosphatase activity toward SGK-1.
We next asked whether Hsp90-dependent phosphorylation of SGK-1 involves a specific kinase or activated pathway upstream of SGK-1. We treated SK-BR-3 cells stably expressing SGK-1-FLAG with a panel of kinase inhibitors that target the major intracellular signaling pathways. Neither U0126 (a dual Erk1/2 and Erk5 inhibitor at the concentrations used), SB203580 (a p38 inhibitor), a JNK inhibitor II, nor staurosporine (a myosin light chain kinase inhibitor) demonstrated any significant effect on the slowly migrating phosphorylated species of SGK-1 (Fig. 4B, lanes 4 -7). In contrast, only the PI3K inhibitor LY294002 eliminated the two most slowly migrating species of SGK-1, resulting in a loss of the same SGK-1 species that disappear following Hsp90 inhibition (Fig. 4B, lanes 2 and 3).
To confirm that PI3K activity plays a role in the Hsp90-dependent phosphorylation of SGK-1, we co-expressed SGK-1-FLAG and the constitutively active catalytic subunit of PI3K p110␣ (p110 CAAX) in COS-1 cells. Following GA treatment in control cells, the slower migrating species of SGK-1 disappeared, and the faster migrating SGK-1 species rapidly accumulated as in previous experiments (Fig. 4C, lanes 1-4). Expression of p110 CAAX both significantly increased SGK-1 basal levels and impaired the depletion of the slowly migrating phos- phorylated species of SGK-1 following GA treatment (Fig. 4C,  lanes 6 -8). These results suggest that very high levels of the PI3K pathway activation by p110 CAAX can overcome a loss of Hsp90 function, perhaps by allowing SGK-1 to be phosphorylated despite suboptimal protein folding. Interestingly, in a similar experiment using constitutively active Mek5DD that should activate phosphorylation of SGK-1 at Ser-78 (65), the loss of Hsp90-dependent phosphorylated bands was not reversed (data not shown) consistent with the lack of effect of U0126 on Hsp90-dependent SGK-1 phosphorylation.
Hsp90-dependent Phosphorylation of SGK-1 Occurs at Both Canonical and Noncanonical Sites-SGK-1 has been shown to have two phosphorylation sites downstream of PI3K activity, Thr-256 and Ser-422 (2). The former has been shown to be phosphorylated by PDK1; the latter is phosphorylated by an unknown kinase, sometimes referred to as "PDK2." We next examined whether the known PI3K-dependent phosphoryla-tion sites are involved in Hsp90-dependent SGK-1 phosphorylation using anti-phospho-specific antibody. SGK-1-FLAG was immunoprecipitated from SK-BR-3 cells and immunoblotted with either anti-phospho-Ser-422 or anti-pan-SGK-1 antibody. As shown in Fig. 5A, SGK-1 lost the Ser-422 phosphorylated species following GA treatment (left). Interestingly, phospho-Ser-422 appears to contribute to the slowest migrating species, including the hyperphosphorylated bands (Fig. 5A, right). Following GA treatment, however, most of phosphorylated Ser-422 SGK-1 disappeared, suggesting that phosphorylation at this site is dependent on Hsp90 activity. We could not evaluate phosphorylation at Thr-256 because of poor specificity of the only available anti-phospho-Thr-256 antibodies. However, Thr-256 phosphorylation has been shown to require prior phosphorylation at Ser-422 (66), suggesting that both Ser-422 and Thr-256 phosphorylation are Hsp90-dependent.
We next examined whether the Thr-256 and Ser-422 residues are the only two Hsp90-dependent phosphorylation sites of SGK-1. We reasoned that if phosphorylation at these sites plays a significant role in the formation of the slow mobility species of SGK-1, a double alanine mutant T256A/S422A SGK-1 should demonstrate a complete or partial loss of these bands even in the absence of GA. Furthermore, the T256A/ FIGURE 4. Hsp90 is required for SGK-1 phosphorylation downstream of PI3K activity. A, phosphatase activity does not play a role in the loss of the SGK-1 phosphorylated species following GA treatment. SGK-1-FLAG expressing SK-BR-3 cells were treated with or without okadaic acid for 2 h and then with or without GA. Whole cell lysates were resolved on an SDS-polyacrylamide gel and immunoblotted (IB) with anti-SGK-1 antibody. B, inhibition of PI3K is sufficient for dephosphorylation of SGK-1. SK-BR-3 cells stably expressing SGK-1-FLAG were treated with various kinase inhibitors. Whole cell lysates were collected, resolved on SDS-PAGE, and immunoblotted with anti-SGK-1 antibody. C, constitutive PI3K-p110 activity reverses GA-mediated loss of SGK-1 phosphorylation. COS-1 cells were transiently transfected with SGK-1-FLAG and p110 CAAX and treated with GA. Whole cell lysates were collected, resolved on SDS-PAGE, and immunoblotted with antibodies as indicated. The SGK-1 blot is shown in two different exposures: 2.5 min for GFP-co-expressed group and 30 s for p110 CAAX group. Fifty-and 150-kDa markers are indicated.

Hsp90 Regulates SGK-1
S422A mutant should not lose the slowly migrating bands following GA treatment. We examined WT and T256A/S422A SGK-1-FLAG in SK-BR-3 cells treated with vehicle or 1 M GA for 4 h by immunoblotting using an anti-SGK-1 antibody. Unexpectedly, T256A/S422A SGK-1 had virtually the same banding pattern as WT SGK-1 (Fig. 5B, lanes 1 and 3), suggesting that the double alanine SGK-1 mutant is still highly phosphorylated. Moreover, the slowly migrating phosphorylated species of T256A/S422A SGK-1 disappeared following GA treatment similarly to the wild type (Fig. 5B, lanes 1 and 2 versus  3 and 4), further suggesting that Thr-256 and Ser-422 are not the only residues affected by GA-mediated dephosphorylation and that additional phosphorylation sites are involved in Hsp90-mediated phosphorylation of SGK-1.
To further confirm that GA causes SGK-1 dephosphorylation beyond the canonical Thr-256 and Ser-422 residues, we performed in vivo 32 P labeling of both WT and T256A/S422A SGK-1 in the presence of GA. We reasoned that if Thr-256 and Ser-422 were the only two Hsp90-dependent phosphorylation sites, T256A/S422A SGK-1 should demonstrate less efficient phosphorylation and decreased sensitivity to GA treatment. However, both WT and the double alanine mutant SGK-1 were similarly phosphorylated in vivo (Fig. 5C, lane 1 versus 3), and furthermore, T256A/S422A SGK-1 lost significant phosphorylation following GA treatment (lane 3 versus 4), strongly sug-gesting that Hsp90-dependent phosphorylation of SGK-1 takes place at additional phosphorylation sites beyond Thr-256 and Ser-422.
Geldanamycin Treatment Results in Decreased SGK-1 Kinase Activity-Having established that Hsp90 inhibition causes the disappearance of hyperphosphorylated SGK-1 species involving Thr-256 and Ser-422, we next examined how Hsp90 function affects overall SGK-1 kinase activity. SK-BR-3 cells were transfected with the kinasedead (KD) and WT SGK-1-FLAG plasmids and then treated with GA or vehicle for 3.5 h before cell lysis. SGK-1 was immunoprecipitated, and an in vitro kinase assay was performed using "SGKtide" as described previously (3). Using the immunoprecipitates from equal amounts of cell lysate, we found that WT SGK-1 activity was abolished following GA treatment (Fig. 6A,  left). A parallel Western analysis of SGK-1 revealed that although the levels of the upper species of WT SGK-1 were reduced following GA treatment, significant amounts of total SGK-1 remained that were devoid of kinase activity (Fig. 6A,  right). Based on our previous results, we hypothesized that Hsp90dependent phosphorylation of SGK-1 takes place at the conserved (Thr-256 and Ser-422) sites as well as additional site(s). To investigate this hypothesis, we examined whether the activation mutation of Ser-422 to aspartic acid (S422D) (3) is sufficient to overcome the GA-mediated inhibition of kinase activity. Surprisingly, S422D SGK-1-FLAG lost 50% of its kinase activity following GA treatment (Fig. 6B, left), further confirming that Hsp90-dependent phosphorylation includes additional unknown residues beyond Ser-422 and Thr-256 that are required for full kinase activity. The results of these kinase assays suggest that the effect of GA on SGK-1 activity takes place through both the canonical PI3K sites Thr-256 and Ser-422 and additional unknown site(s).
Hsp90 Has Distinctive Roles in the Phosphorylation of SGK-1 Versus Akt1-SGK-1 shares ϳ50% homology with the catalytic domain of Akt1, and Akt1 has also been shown to be a client protein for Hsp90 chaperone activity. Because both kinases are phosphorylated downstream of the PI3K/PDK pathway, we examined whether or not SGK-1 and Akt1 phosphorylation is differentially regulated by Hsp90. MDA-MB-231 cells expressing either an empty vector or SGK-1-FLAG were treated with 1 M GA, and the levels of SGK-1, endogenous phospho-Akt1, and total Akt1 were determined by immunoblotting. As shown in Fig. 7 (top panel, lanes 1-5), endogenous SGK-1 was barely FIGURE 6. Hsp90 function is required for SGK-1 kinase activity. A, SGK-1 loses kinase activity following GA treatment. KD and WT SGK-1 were immunoprecipitated from GA-treated cells (3.5 h), and SGK-1 kinase activity was examined using an in vitro kinase assay. A representative experiment from three independent experiments is shown. Equal amounts of cell lysates were analyzed by immunoblotting (IB) using anti-SGK-1 antibody. B, constitutively active SGK-1 (S422D) remains sensitive to GA-mediated dephosphorylation and inactivation. Kinase assay and Western blot were performed as in A. SGK-1 activity above KD activity is presented. A representative experiment from three independent experiments is shown. Equal amounts of cell lysates were analyzed by immunoblotting using anti-SGK-1 antibody. Fifty-kDa marker is indicated.
detectable, whereas ectopic SGK-1 was easily detected. As expected, ectopically expressed SGK-1 lost the upper (phosphorylated) species after 30 min of GA treatment (Fig. 7, top  panel, lanes 6 -10). In contrast, levels of phospho-Ser-473 Akt1 were acutely increased following 30 min of GA treatment (Fig.  7, middle panel, lanes 2 and 3), consistent with previous observations that suggested GA induced early Akt1 activation via Src-dependent PI3K activation (49). In the presence of ectopic SGK-1 and GA treatment, phospho-Akt1 levels increased even more dramatically (Fig. 7, middle panel, lanes 1 and 2 versus 7 and 8), suggesting that elevated levels of SGK-1 protein increase Akt1 accessibility for phosphorylation. The observation that Akt1 is acutely phosphorylated at Ser-473 following GA is in contrast to the loss of SGK-1 phosphorylation occurring at the same early time point (Fig. 7, top panel, lane 7). These different short term effects of GA suggest that Hsp90 functions as an important differential regulator of two separate branches of the PI3K pathway; acutely, Hsp90 facilitates SGK-1 activation, although it limits Akt1 activity.

DISCUSSION
The results presented in this study suggest that Hsp90 is a critical component in the regulation of the PI3K/SGK-1 pathway and provide a new example of the relevance of Hsp90 function to PI3K signaling. SGK is transcriptionally up-regulated following various stress stimuli (1), thereby allowing a transient increase in SGK-1 protein levels and initiating a homeostatic survival signal to the cell. Because stress-induced proteins are often misfolded following induction (67), SGK-1 likely requires chaperone-mediated assistance in both folding and protein quality control. We previously showed that SGK-1 interacts with Hsp70 and the chaperone-dependent E3 ligase CHIP, both of which play a central role in protein quality control. Here we show that Hsp90 interaction is required for phosphorylation and full activation of SGK-1 independently of protein degradation. Hsp90-mediated regulation of SGK-1 phosphorylation is perhaps an alternative to the function of the plekstrin homology domain of Akt1, which anchors Akt1 to the plasma membrane, allowing for Akt1 activation and phosphorylation. SGK-1 does not contain an N-terminal plekstrin homology domain and appears to be activated in the cytosol by upstream kinases (13).
Although Hsp90 is involved in activation of both Akt1 and SGK-1, our data suggest that Hsp90 employs different mechanisms to regulate Akt1 versus SGK-1 phosphorylation.
Although Hsp90 affects Akt1 indirectly via modulating PI3K activity, SGK-1 phosphorylation is directly dependent on Hsp90 binding, which we propose to be a limiting step for SGK-1 phosphorylation. For example, we observed that Akt1 is transiently phosphorylated at Ser-473 following a short 30-min exposure to GA. This observation is consistent with a previous report (49) suggesting that GA rapidly disrupts Src association with Hsp90, allowing activated Src to phosphorylate Cbl, which recruits the p85 subunit of PI3K and activates Akt1. However, although we saw rapid phosphorylation of Akt1 at Ser-473, we concurrently saw a loss of SGK-1 phosphorylation, suggesting that GA affects SGK-1 activation via a direct disruption of the association of Hsp90 with SGK-1. Unexpectedly, following overexpression of SGK-1, we observed even higher Akt1 phosphorylation in the presence of GA. We speculate that GA allows Hsp90 to dissociate from Src but that residual active Hsp90 may be sequestered by the highly overexpressed SGK-1, thereby increasing Akt1 phosphorylation. Alternatively, overexpressed SGK-1 may feed back to activate the PI3K pathway. Another possibility is that Akt1 and SGK-1 form a complex that makes Akt1 more accessible for phosphorylation at Ser-473 (6).
Another indirect Hsp90-mediated mechanism of Akt1 activation that has been proposed is via stabilization of ErbB2 (a known activation pathway of PI3K). This mechanism does not seem to apply to SGK-1/Hsp90 interaction either. For example, we found that loss of SGK-1 phosphorylation following inhibition of Hsp90 in MDA-MB-231 cells (moderate PI3K activity because of low ErbB2) versus SK-BR-3 cells (highly active PI3K because of amplified ErbB2) occurred with similar kinetics (data not shown). This is in contrast to Akt1 phosphorylation, where Hsp90 inhibition by a GA derivative, 17(allylamino)-17demethoxygeldanamycin, results in a more rapid dephosphorylation of Akt1 in high ErbB2-expressing cells compared with low ErbB2 cells (64). In these studies, Hsp90 inhibition resulted in degradation of ErbB2, lower PI3K activity, and in the release of ErbB2-mediated repression of the phosphatase PP1, so that Akt1 becomes less phosphorylated (44,68). In contrast, we find that GA-mediated dephosphorylation of SGK-1 appears to be independent of ErbB2 levels and of phosphatase activity, suggesting that Hsp90 is more directly involved in SGK-1 phosphorylation.
What are the implications of the distinct Hsp90-mediated regulation of SGK-1 versus Akt1? SGK-1 is a transcriptionally induced protein kinase whose levels are increased by various stress-related extra-and intracellular factors. Therefore, its regulation should be rapid in response to acute stress to provide a strong but transient signal. SGK-1 protein levels are negatively regulated via post-translational ubiquitin modification rendering SGK-1 rapidly degraded. Positive regulation of SGK-1 is provided by phosphorylation, which we show here requires Hsp90 binding for maximal activation. Akt1, on the other hand, is a constitutively expressed protein that needs to be precisely tuned in response to insulin. This goal is achieved by the dual role Hsp90 plays in the functioning of Akt as follows: positive regulation maintaining Akt1 stability and phosphorylation and negative regulation limiting the unnecessary phosphorylation and sequential activation of Akt1. By integrating the regulation of both SGK-1 and Akt1, Hsp90 plays an impor- Hsp90 Regulates SGK-1 tant role in adequate functioning of the PI3K pathway in normal conditions and in response to stress.
We and others have shown previously that SGK-1 is highly expressed in about 40% of human breast cancers (69,70). The data presented here strongly suggest that Hsp90 inhibitors could benefit patients with tumors that overexpress SGK-1. However, we observed that overexpression of the constitutively active catalytic subunit of PI3K, p110␣ (p110 CAAX) decreased the effect of GA treatment. We hypothesize that Hsp90 inhibition may limit the phosphorylation of SGK-1 in the presence of physiological PI3K activity, but the requirement for Hsp90 is overcome as a consequence of the constitutively active membrane-bound, p110␣ CAAX. Therefore, tumors with highly active PI3K and overexpressed SGK-1 might be treated more effectively with a combination of drugs targeting both Hsp90 and PI3K.
One possible interpretation of our data could be that GA actually affects the activity of the immediate upstream kinase, PDK1, instead of directly affecting SGK-1. However, it has been shown previously that Hsp90 inhibition only decreases protein levels of PDK1 after 12 h (59,71). Given the much more rapid (within 1 h) loss of the phosphorylation of SGK-1 following GA treatment, it is unlikely that decreased PDK1 expression at 12 h contributes to the rapid and complete loss of the phosphorylated species of SGK-1. Furthermore, PDK1-mediated phosphorylation of Thr-256 has been shown to require Ser-422 phosphorylation, suggesting that Ser-422 plays a more critical role in Hsp90-dependent activation of SGK-1. The most likely explanation of our data is that a loss of Hsp90 function prevents SGK-1 from optimal protein conformation and inhibits phosphorylation of Ser-422 and additional site(s).
It was recently published that SGK-1 has alternative translational start site products with masses ranging from 42 to 49 kDa (72). In this study, we observed multiple bands of different electrophoretic mobilities when examining both endogenous and ectopic SGK-1 expression. It is possible that these bands consist of a mixture of both full-length and truncated SGK-1 proteins because neither Hsp90 nor PI3K inhibition completely resolved the multiplicity of the banding pattern of SGK-1. It should be noted that the majority of the Hsp90-dependent phospho-SGK-1 species have apparent molecular masses greater than the full-length SGK-1 protein; they are therefore likely to be modified full-length SGK-1 (49 kDa). However, we have also found that ectopically expressed SGK-1 lacking the first 60 amino acids expresses as two distinct bands and loses the slowly migrating upper band following GA treatment as well as following in vitro phosphatase treatment (data not shown). Therefore, we predict that the recently described alternatively translated forms of SGK-1 are also likely dependent on Hsp90 for phosphorylation.
In this study we have shown for the first time that the activity of SGK-1 requires Hsp90 binding. Through this interaction, Hsp90 aids in PI3K-dependent phosphorylation and activation of SGK-1 independently of increasing SGK-1 protein stability (Fig. 8). Interestingly, the requirement of Hsp90 to phosphorylate SGK-1 is not limited to the known conserved sites Thr-256 and Ser-422 but appears to include additional phosphorylated residues. Taken together with previous data on the role of Hsp90 in regulating Akt1 activity, our findings suggest that Hsp90 has an inhibitory effect on Akt1 phosphorylation while allowing SGK-1 phosphorylation. Because SGK-1 function appears to be an important component of the response of a cell to environmental stress, we hypothesize that rapid Hsp90-dependent SGK-1 activation (and concomitant Akt1 inactivation) may serve to fine-tune the relative activities of PI3K-regulated signaling, thereby optimizing the response of a cell to acute stress (Fig. 8).