p90 RSK Blocks Bad-mediated Cell Death via a Protein Kinase C-dependent Pathway*

Although activation of protein kinase C (PKC) is known to promote cell survival and protect against cell death, the PKC targets and pathways that serve this function have remained elusive. Here we demonstrate that two potent activators of PKC, 12- O -tetradecanoyl-phorbol-13-acetate and bryostatin, both stimulate phosphorylation of Bad at Ser 112 , a site known to regulate apoptotic cell death by interleukin-3. PKC inhibitors but not PI 3-kinase/Akt inhibitors block 12- O -tetradecanoyl-phorbol-13-acetate-stimulated Bad phosphorylation. PKC isoforms tested in vitro were unable to phosphorylate Bad at Ser 112 , suggesting that PKC acts indirectly to activate a downstream Bad kinase. p90 RSK and family members RSK-2 and RSK-3 are activated by phorbol ester and phosphorylate Bad at Ser 112 both in vitro and in vivo . p90 RSK stimulates binding of Bad to 14-3-3 and blocks Bad-mediated cell death in a Ser 112 -dependent manner. These findings suggest that p90 RSK can function in a PKC-dependent pathway to promote cell survival via phosphorylation and inactivation of Bad-medi-ated cell death. Survival factors prevent cells from undergoing cell death or inhibiting cell of signaling pathways activated by IL-3 1 and other survival factors the cell death has been traced to ability to stimulate phosphatidylinosi-tol (PI) 3-kinase and inactivate the apoptotic factor

Survival factors prevent cells from undergoing cell death or apoptosis by inhibiting the execution of the cell death program. Recently, the convergence of signaling pathways activated by IL-3 1 and other survival factors with the cell death machinery has been traced to their ability to stimulate phosphatidylinositol (PI) 3-kinase (1-3) and inactivate the apoptotic factor Bad (4,5). IL-3 treatment of immune cells stimulates the phosphorylation of Bad at two sites, Ser 112 and Ser 136 (5). Phosphorylation of Bad at these sites inhibits binding of Bad to Bcl-xL and induces binding of 14-3-3 proteins, which act to sequester Bad away from Bcl-xL (5).
How survival signals are transmitted to the Bcl-2/Bcl-xL checkpoint is not well understood. Survival of lymphoid progenitor cells mediated by IL-3 requires signaling through PI 3-kinase as indicated by the sensitivity to wortmannin and LY294002 (3,6), whereas survival mediated by granulocytemacrophage colony-stimulating factor (7) or the overexpression of insulin-like growth factor-1 receptors (8) does not require PI 3-kinase, suggesting the existence of additional pathways insensitive to PI 3-kinase inhibitors. Recently, the PI 3-kinasesensitive pathway was shown to involve the activation of Akt (also known as PKB) (8 -11) and the direct phosphorylation of Bad at Ser 136 (12,13). Although Akt stimulates survival by phosphorylating Bad at Ser 136 , Akt is not able to phosphorylate Bad at Ser 112 , suggesting the existence of additional pathways regulating Bad phosphorylation at Ser 112 (12). Protein kinase A localized to the mitochondrial membrane has been recently shown to phosphorylate Bad at Ser 112 (14), and the calciumactivated phosphatase calcineurin can induce apoptosis by dephosphorylating Bad at Ser 112 and Ser 136 (15).
Drugs such as TPA or bryostatin that activate protein kinase C (PKC) are also able to promote cell survival and protect against cell death (16,17). For example, TPA rescues immune cells from glucocorticoid-induced cell death and can rescue cells from growth factor withdrawal (16,18,19). Similarly, agents that inhibit PKC such as staurosporin or UCN-01 are known to be potent inducers of apoptosis (20 -22). Furthermore, tumor cell sensitivity to anti-metabolites such as araC is greatly enhanced by disruption of the PKC pathway (23) and has been traced to post-translational changes in the Bcl-2 checkpoint molecules (24,25). These observations suggest that certain PKC isoforms may act to deliver survival signals that protect against cell death. However, the targets of PKC that protect against cell death remain largely unknown.
One downstream target of PKC is the serine/threonine kinase p90 RSK , also referred to as RSK-1, which contains two distinct kinase domains within a single polypeptide (26). This enzyme is activated by TPA and functions downstream of p42/ 44MAPK (extracellular signal-regulated kinase 1/2) in the MAP kinase signaling cascade (27). Activation of p90 RSK coincides with oncogenic transformation (28), stimulation of G 0 /G 1 transition (29 -31), and differentiation of PC12 cells (32). One of the few identified physiological substrates of this enzyme is the transcription factor CREB (33). TPA activates p90 RSK by phosphorylation at several sites located within the activation loop of the C-terminal kinase domain (34). Phosphorylation at the C-terminal activation loop leads to phosphorylation and activation of the N-terminal kinase domain, which in turn phosphorylates downstream targets such as CREB (34). To date the role of p90 RSK in regulating cell survival has not been studied directly.
Here we demonstrate that TPA, an activator of classical and novel protein kinase C isoforms, protects cells against Badmediated cell death and also stimulates phosphorylation of Bad at Ser 112 . Inhibitors of PKC block Bad phosphorylation, suggesting that PKC might directly phosphorylate Bad; however, in our hands, we have been unable to detect significant in vitro phosphorylation of Bad by several different conventional and novel PKC isoforms, suggesting that PKC activates a downstream Bad kinase. We demonstrate that TPA activates p90 RSK in a PKC-dependent fashion and that in vitro p90 RSK can directly phosphorylate Bad at Ser 112 . Overexpression of p90 RSK in HEK293 cells stimulates Bad phosphorylation at Ser112. Furthermore, treatment of HEK293 cells with TPA or overexpression of p90 RSK suppresses Bad-mediated cell death. These results suggest that the effects of phorbol ester and PKC on cell survival act via a pathway involving a p90 RSK -dependent phosphorylation of Bad at Ser 112 . Our results suggest that the PKC/p90 RSK pathway may play an important role in regulating cell death decisions at the Bcl-xL checkpoint via phosphorylation and inactivation of Bad.

EXPERIMENTAL PROCEDURES
Materials-Forskolin, 3-isobutyl-1-methyl-xanthine, TPA, fetal calf serum, and mammalian cell culture media were from Sigma. Phosphorylation state-specific Bad antibodies (Ser 112 ) and (Ser 136 ) and control Bad antibodies were from New England Biolabs. p90 RSK -specific monoclonal antibody was from Transduction Laboratories. RSK-2-and RSK-3-specific antibodies were from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). MEK protein kinase inhibitor PD98059 and protein kinase A were from New England Biolabs, and SB203580 was kindly provided by John Lee and Peter Young. PKC inhibitors GO6983, GO6976, and bisindolymaleimide, wortmannin, LY294002, and rapamycin were from Calbiochem; bryostatin was from Alexis; ␤-galactosidase stain kit was from Invitrogen; and in situ cell death detection kit for the TUNEL assay was from Roche Molecular Biochemicals. The Phototope-HRP Chemiluminescent Western Detection Kit was from New England Biolabs.
Cell Culture-HEK293 cells were maintained in modified Eagle's medium supplemented with 10% (v/v) horse serum. COS1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum.
Plasmid and DNA Constructions-The GST-Bad mammalian expression vector was constructed by fusion of the complete coding sequence of Bad amplified from a mouse brain cDNA library (CLONTECH) and cloned into the BamHI/NotI sites of eukaryotic expression vector pEBG. pEBG expresses an amino-terminal GST fusion of the cloned gene under the control of the strong, constitutively active human EF-1 ␣ promoter (63). GST-Bad S112A, S136A, and S112A/S136A mammalian expression vectors were constructed by transferring BbgI/AfeI fragments from pCMVBadS112A, pCMVBadS136A, and pCMVBadS112A/ S136A (obtained from Stanley Korsmeyer) into a GST-Bad mammalian expression vector. Maltose-binding protein (MBP)/14-3-3 ␦ and ⑀ were kindly provided by Michael Yaffe and Lewis Cantley. pMT2-p90 RSK was provided by Joseph Avruch, and pMT2RSK2 and pMT2RSK3 were provided by Christian Bjorbaek.
Phosphoantibody Production-The anti-phospho-Bad (Ser 112 ) and anti-phospho-Bad (Ser 136 ) antibodies were generated by immunizing rabbits with synthetic phosphopeptides covalently coupled to keyhole limpet hemocyanin. The presence of phosphospecific immunoreactivity was detected by enzyme-linked immunosorbent assays using both the phosphorylated and the nonphosphorylated peptides. After purification of IgG using protein A-agarose, the phosphopeptide-specific antibodies were purified by first passing IgG over immobilized, nonphosphorylated peptide, to remove antibodies reactive with the nonphosphorylated epitopes. The nonabsorbed fraction was then passed over a column of immobilized phosphopeptide. After extensive washing, the retained immunoglobulins were eluted at low pH, rapidly neutralized, dialyzed, and concentrated.
Transient Transfections-Transient transfection of HEK293 cells was performed as described previously (35) with various amounts of expression plasmid as described in the figure legends. The total amount of DNA transfected was maintained at 20 g with pGEM3. Following calcium phosphate transfections, cells were glycerol-shocked and incubated for 24 h in media containing 10% fetal calf serum and for 18 h without serum. Cells were treated with 200 nM TPA for the indicated times.
14-3-3 Fusion Protein Pull Down-MBP-14-3-3 proteins bound to maltose beads were mixed with extracts prepared from HEK293 cells transfected with Bad expression vectors and treated with or without TPA and lysed in cell lysis buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM Na 3 VO 4 , 1 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride). Cell extracts were incubated overnight at 4°C; beads were washed twice with cell lysis buffer and twice with phosphate-buffered saline; and proteins were eluted with SDS sample buffer for Western analysis.
Protein Kinase Assay-p90 RSK activity was detected by immunoprecipitation of p90 RSK from HEK293 cell extracts. HEK293 cells were pretreated with different inhibitors for 1 h and then treated with 200 nM TPA for 30 min. Cells were lysed in cell lysis buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton, 2.5 mM sodium pyrophosphate, 1 mM Na 3 VO 4, 1 g/ml leupeptin). Extracts from 5 ϫ 10 5 cells were incubated with p90 RSK antibody overnight and with protein A beads for 3 h at 4°C by gentle rocking. Immunocomplexed beads were washed twice with cell lysis buffer and twice with kinase buffer (25 mM Tris, pH 7.4, 5 mM ␤-Glycerolphosphate, 2 mM dithiothreitol, 0.1 mM Na 3 VO 4, 10 mM MgCl 2 ). Immunocomplexes were resuspended in 50 l of kinase buffer supplemented with 200 M ATP and 2 g of GST-Bad protein or MBP-Bad and Bad mutation and incubated for 30 min at 30°C. Kinase reactions were terminated with SDS sample buffer, and Bad phosphorylation was detected by immunoblotting.
Intein-mediated Protein-Peptide Ligation-Wild-type and mutant peptides spanning Bad amino acid sequences from 106 to 141 were chemically synthesized with a cysteine added to the N terminus of each peptide. Phosphopeptides were synthesized by incorporating phosphorylated serine at position 112 or 136. Mutant peptides were synthesized with the following changes: Ser 112 changed to Ala, Ser 136 changed to Ala, or the double mutant where both Ser 112 and Ser 136 were changed to Ala. Fusion proteins were prepared by incubating each peptide with were used for immunoblotting using phosphorylation state-specific antibodies to Ser 112 and Ser 136 . B, HEK293 cells were transfected with plasmids encoding wild-type GST-Bad or GST-Bad containing S112A, S136A, or both (S112A/S136A) mutants as indicated. After 24 h, the cells were deprived of serum, and 18 h later they were treated with 200 nM TPA for 30 min as indicated prior to harvest. Total cell extracts were prepared, and immunoblotting was carried out using phosphospecific antibodies directed against Bad Ser 112 (top), or phosphorylation-independent control Bad antibody (bottom).
bacterially expressed MBP-paramyosin isolated and purified using the IMPACT protein expression system (New England Biolabs) using the Mxe GyrA intein as described by Evans et al. (36). Following the peptide ligation reaction, MBP fusion proteins were purified away from free peptides by binding to maltose beads.
TUNEL Assay-HEK293 cells were transfected with pCMVBad, plus or minus p90 RSK . 24 h after transfection, cells were serum-starved for 18 h and then fixed with 4% paraformaldehyde solution for 30 min, blocked with 0.3% H 2 O 2 in methanol for 30 min, and permeabilized with 0.1% Triton X-100 in 0.1% sodium citrate. After extensive washing, cells were incubated with terminal deoxynucleotidyl transferase and fluorescein-labeled nucleotide mixture for 1 h at 37°C, washed three times with phosphate-buffered saline, incubated with anti-fluorescein antibody conjugated with horseradish peroxidase for 30 min, washed, and then developed with 3Ј3-diaminobenzidine tetrahydrochloride solution.
X-Gal Staining-After transfection and treatment, cells were fixed in 4% paraformaldehyde solution for 10 min at room temperature. After washing three times with phosphate-buffered saline, cells were incubated with 1 mg/ml X-Gal in N,N-dimethylformamide, 4 mM potassium ferricyanide, 4 mM potassium ferrocyanide, 2 mM MgCl 2 in phosphatebuffered saline buffer for 24 h. Blue color-containing cells were examined under a microscope.

Protein Kinase C Activators Stimulate Bad Phosphorylation
at Ser 112 -Phosphorylation state-specific antibodies (37) directed against Ser 112 and Ser 136 of Bad were generated as described under "Experimental Procedures." The antibodies were shown to be highly specific for the phosphorylated epitope by enzyme-linked immunosorbent assays and by Western blotting using phosphorylated and nonphosphorylated GST-Bad protein (data not shown). To further characterize antibody site specificity, a series of MBP-Bad fusion proteins was constructed by ligating to the C terminus of MBP chemically synthesized Bad peptides or Bad peptides phosphorylated at Ser 112 , Ser 136 , or both sites (see "Experimental Procedures" and Ref. 36). As shown in Fig. 1A, phospho-Bad Ser 112 antibody only recognizes MBP-Bad fusion protein when phosphorylated at Ser 112 or Ser 112 and Ser 136 , whereas phospho-Bad Ser 136 antibody only recognizes MBP-Bad fusion protein when phosphorylated at Ser 136 or Ser 112 and Ser 136 . These results demonstrate a high degree of specificity for both antibodies to their specific phosphoserine residues. We next examined antibody specificity in vivo. HEK293 cells were transfected with plasmids encoding wild-type GST-Bad or GST-Bad S112A, S136A, or S112A/S136A mutants. Cell extracts were prepared and immunoblotted using phosphospecific antibodies directed against Bad Ser 112 (Fig. 1B, top) or Bad antibody (Fig. 1B,  bottom). Treatment of HEK293 cells with TPA increased the phosphorylation of wild-type GST-Bad at Ser 112 . Mutation of Ser 112 to Ala blocked the TPA-induced immunoreactivity. Mutation of Ser 136 to Ala did not block TPA-induced Ser 112 phosphorylation, demonstrating the site specificity of the phospho-Bad Ser 112 antibodies. Wild-type and mutant GST-Bad proteins were expressed at similar levels as determined using Bad antibody (Fig. 1B, bottom).
We next explored other agents that might also stimulate phosphorylation of Bad at Ser 112 using the phosphospecific Bad Ser 112 antibody. Treatment of GST-Bad transfected HEK293 cells with TPA or epidermal growth factor produced the largest induction of GST-Bad phosphorylation at Ser 112 . Forskolin, UV irradiation, and platelet-derived growth factor also reproducibly stimulated phosphorylation at Ser 112 ( Fig. 2A). TPA rapidly induced a sustained phosphorylation of Bad, evident within 5 min of treatment, that remained elevated 8 h after treatment (Fig. 2B). Bryostatin, another activator of PKC, as well as TPA induced endogenous Bad Ser 112 phosphorylation in COS1 cells (Fig. 2C).
Protein Kinase C Inhibitors Block Bad Phosphorylation-Since phorbol esters are known to activate classical and novel PKC isoforms as well as several other growth-associated pathways including the MAP kinase cascade, we next used selective inhibitors to determine the role of several different intracellular signal transduction pathways on Bad phosphorylation at Ser 112 . Treatment of HEK293 cells with rapamycin (a selective mTOR/FRAP inhibitor) or two different PI 3-kinase inhibitors, wortmannin and LY294002, had no effect on TPA-stimulated Bad phosphorylation. Inhibitors of MEK and the MAPK cascade (PD98059), or p38 MAPK (SB203580) had very small effects on TPA-stimulated Bad phosphorylation at Ser 112 (Fig.  3A). We next examined the ability of several different PKC inhibitors to block TPA-stimulated Bad phosphorylation (Fig.  3, A and B). Bisindolymaleimide I (also known as GF 109203X or GO6850) is a potent inhibitor of PKC␣, -␤1, -␤2, -␥, -␦, and -⑀ isoforms. 10 nM bisindolymaleimide I partially blocked, while 100 nM completely blocked, TPA-induced Bad phosphorylation at Ser 112 . Furthermore, 1 M bisindolymaleimide I reduced phosphorylation at Ser 112 to below basal levels (Fig. 3B), suggesting an effect on basal Bad phosphorylation. Treatment of HEK293 cells with 100 nM GO6983, a potent inhibitor of PKC␣, -␤, -␥, and -␦ isoforms, also blocked TPA-stimulated Bad phosphorylation. Significantly, treatment with GO6976 had little or no effect on TPA-stimulated Ser 112 phosphorylation until concentrations of 1 M were reached. This inhibitor has been reported to selectively inhibit Ca 2ϩ -dependent PKC isoforms ␣, and ␤1, but to have no effect on the Ca 2ϩ -independent isoforms ⑀ and ␦ (38). The potent inhibition by bisindolymaleimide I and the inability of GO6976 to block TPA-mediated signaling to Bad suggest a role for the Ca 2ϩ -independent, novel PKC isoforms.
p90 RSK Phosphorylates Bad at Ser 112 in Vitro-We next sought to identify the TPA-activated protein kinase phosphorylating Bad at Ser 112 . Since both TPA and bryostatin induce Bad phosphorylation and PKC inhibitors block Bad phosphorylation, we first tested the ability of various PKC isoforms to directly phosphorylate Bad at Ser 112 . Several different conventional (␣, ␤1, ␤2, ␥), novel (⑀, ␦), and atypical () PKC isoforms were tested for their ability to phosphorylate Bad in vitro. Surprisingly, although the PKC isoforms were able to phosphorylate various control proteins and peptides, all of the isoforms tested were unable to phosphorylate Bad at Ser 112 (data not shown). This result suggested that PKC may activate a downstream protein kinase that in turn phosphorylates Bad at Ser 112 .
The role of p90 RSK was next investigated, since this protein kinase is activated by TPA downstream of PKC (39). The substrate specificity of p90 RSK in vitro is reported to be consistent with the RXXS motif in which Bad Ser 112 is found (28). To test whether p90 RSK could phosphorylate Bad at Ser 112 , p90 RSK was immunoprecipitated from mock-or TPA-treated HEK293 cells using a p90 RSK -specific monoclonal antibody. Immunocomplex kinase assays were performed using wild-type and mutant MBP-Bad fusion proteins as substrates, and Bad phosphorylation was detected by immunoblotting using a phospho-Bad Ser 112 antibody (Fig. 4A). TPA stimulated p90 RSK kinase activity that phosphorylated wild-type Bad and the Bad S136A mutant but not the S112A and S112A/S136A mutants. Immunocomplexes were also incubated in a kinase assay using recombinant GST-Bad fusion protein as substrate, and Bad phosphorylation was detected by immunoblotting using phospho-Bad Ser 112 (Fig. 4B, top) and phospho-Bad Ser 136 (Fig. 4B,  bottom) antibodies. As shown in Fig. 4B, TPA stimulates the ability of p90 RSK to phosphorylate recombinant GST-Bad at Ser 112 but not at Ser 136 . These results suggest that in vitro p90 RSK phosphorylates Bad at Ser 112 but not significantly at Ser 136 .
We next reasoned that if p90 RSK was downstream of PKC on a pathway leading to Bad phosphorylation at Ser 112 , then p90 RSK activity should show a profile of inhibitor sensitivity similar to that observed for Bad Ser 112 phosphorylation. To test this, we examined the sensitivity of p90 RSK kinase activity to 20 M PD98059, 100 nM GO6983, 100 nM GO6976, or 100 nM bisindolymaleimide I by pretreating cells for 1 h with each inhibitor. Following a 30-min TPA treatment, cell extracts were prepared, p90 RSK was immunoprecipitated, and kinase activity was determined as described above. As observed for Bad Ser 112 phosphorylation (Fig. 3A), pretreatment with either GO6983 or bisindolymaleimide I blocked p90 RSK activity (Fig.  4B). Pretreatment with PD98059 and GO6976 resulted in little or no effect on p90 RSK activity, in good agreement with their effects on Bad Ser 112 phosphorylation (compare Fig. 3A with Fig. 4B). To exclude the possibility that these inhibitors block p90 RSK activity directly, p90 RSK was immunoprecipitated from TPA-treated HEK293 cell extracts, and immunocomplexes were preincubated in the presence of 20 M PD98059 and 1 M of each of the three different PKC inhibitors (GO6983, GO6976, or bisindolymaleimide I) in vitro for 1 h at 37°C. Kinase assays were then carried out using GST-Bad as substrate. As shown in Fig. 4C, even a 10-fold increase over the concentrations of GO6983 and bisindolymaleimide I required to block p90 RSK activity in vivo had no effect on p90 RSK activity in vitro (Fig. 4, Overexpression of p90 RSK Stimulates Bad Phosphorylation at Ser 112 -We next examined the ability of p90 RSK to stimulate Bad phosphorylation at Ser 112 when overexpressed by transfection. HEK293 cells were cotransfected with plasmids encoding pMT2-p90 RSK and GST-Bad or encoding GST-Bad S112A, S136A, and S112A/S136A mutants; 24 h after transfection, cells were grown in the medium without serum for a further 18 h and then treated with TPA for 30 min before harvesting. Cell extracts were analyzed by immunoblotting using phospho-Bad Ser 112 antibody. Overexpression of p90 RSK induced Bad phosphorylation at Ser 112 in the absence of TPA treatment, suggesting that overexpressed p90 RSK is at least partially active in the absence of TPA (Fig. 4D, lane 4). Bad Ser 112 phosphorylation was further stimulated by TPA in cells transfected with p90 RSK (Fig. 4D, lane 5). Mutation of Bad Ser 112 to Ala, but not Ser 136 to Ala, blocks p90 RSK -induced phosphorylation.

RSK-2 and RSK-3 Also Induce Bad Ser 112 Phosphorylation in Vitro and in Vivo-
We further investigated the role of other RSK family members on Bad Ser 112 phosphorylation. RSK-2 and RSK-3 were immunoprecipitated from mock-, TPA-, or TPA plus PD98059-treated HEK293 cells using RSK-2-and RSK-3-specific antibodies. Immunocomplex kinase assays were performed using GST-Bad fusion proteins as substrates, and Bad phosphorylation was detected by immunoblotting using phospho-Bad Ser 112 antibody. As shown in Fig. 5A, TPA stimulates the ability of both RSK-2 and RSK-3 to phosphorylate Bad at Ser 112 in a fashion independent of MAP kinase. To exclude the possibility of antibody cross-reaction, we analyzed specificity by Western blotting using HEK293 cell extracts where p90 RSK , RSK-2, and RSK-3 were individually overexpressed. Antibodies showed no cross-reactivity among p90 RSK , RSK-2, and RSK-3 (data not shown). To test the ability of RSK-2 and RSK-3 to stimulate Bad phosphorylation in vivo, we transfected HEK293 cells with pMT2-RSK2, pMT2-RSK3, and GST-Bad. As for p90 RSK , overexpression of either RSK-2 or RSK-3 stimulated Bad Ser 112 phosphorylation that was further enhanced after TPA treatment (Fig. 5B).
TPA and p90 RSK Stimulate Bad Binding to 14-3-3 Isoforms-Phosphorylation of Bad at Ser 112 and Ser 136 has been reported to mediate 14-3-3 binding, thereby sequestering Bad to cellular compartments that block its cytotoxicity (5). We next asked whether treatment of HEK293 cells with TPA or overexpression of p90 RSK would stimulate binding of Bad to 14-3-3 proteins. Cell extracts were prepared from control cells or cells transfected with GST-Bad and pMT2-p90 RSK or treated with TPA for 30 min. The cell extracts were incubated with MBP-14-3-3 ⑀ or ␦ fusion proteins immobilized on amylose beads. The interaction of GST-Bad with 14-3-3 proteins was monitored by Western blotting using phospho-Bad Ser 112 or Bad antibodies. As shown in Fig. 6, both TPA and p90 RSK stimulate binding of GST-Bad to 14-3-3 ⑀ or ␦ isoforms.
p90 RSK and TPA Suppress Bad-mediated Cell Death via Ser 112 -If PKC and p90 RSK function in a pathway that abrogates Bad-mediated cell death, then activation of the PKC/ p90 RSK pathway should block Bad-mediated cell death via phosphorylation and induction of 14-3-3 binding. To examine this possibility, we transfected HEK293 and COS1 cells with plasmids expressing wild-type Bad and a construct encoding ␤-galactosidase as a marker to allow identification of transfected cells. Cells transfected with ␤-galactosidase and stained with X-Gal appear normal in morphology; however, cotransfection with constructs encoding Bad results in a dramatic decrease in the overall number of X-galactosidase-positive cells and an increase in the number of X-Gal-stained cells showing FIG. 4. p90 RSK phosphorylates Bad at Ser 112 in vitro and in vivo; the profile of inhibitor sensitivity is similar between p90 RSK activation and Bad Ser 112 phosphorylation. A, TPA-induced p90 RSK activity phosphorylates Bad at Ser 112 . HEK293 cells were grown in normal medium for 24 h and deprived of serum for 18 h followed by treatment with or without 200 nM TPA for 30 min prior to harvesting. Cell extracts were prepared, and p90 RSK was immunoprecipitated using p90 RSK -specific antibody. Immunocomplexes were incubated in a kinase reaction using wild-type and mutant MBP-Bad fusion proteins as substrates, and Bad phosphorylation was detected by immunoblotting using phospho-Bad Ser 112 antibody. B, inhibitors of PKC but not extracellular signal-regulated kinase 1/2 block p90 RSK activity in HEK293 cells. HEK293 cells were grown as described above and incubated with 20 M PD98059, 100 nM GO6976, 100 nM GO6983, and 100 nM bisindolymaleimide, respectively for 1 h, and then treated with TPA for 30 min. Cells were harvested, and p90 RSK was immunoprecipitated as described above. Immunocomplex kinase assays were performed using GST-Bad as a substrate. Phosphorylation of Bad was detected by immunoblotting using phospho-Bad Ser 112 -(top) and phospho-Bad Ser 136 -(bottom) specific antibodies. C, inhibitors of PKC do not inhibit p90 RSK activity in vitro. Immunoprecipitated p90 RSK from TPA-treated 293 cells was preincubated with 20 M PD98059, 1 M GO6976, 1 M GO6983, and 1 M bisindolymaleimide for 1 h in vitro at 37°C. Kinase assays were then performed using GST-Bad as a substrate. Phosphorylation of Bad was detected by Western blot using phospho-Bad (Ser 112 ) antibody. D, p90 RSK induces Bad phosphorylation in vivo. HEK293 cells were transfected without or with plasmids encoding GST-Bad; GST-Bad S112A, S136A, or S112A/S136A mutants; and pMT2 p90 RSK as indicated. After 24 h, cells were deprived of serum, and 18 h later they were treated with medium or with medium plus 200 nM TPA for 30 min prior to harvest. Immunoblotting was carried out using phospho-Bad Ser 112 antibody (top panel), or Bad antibodies (bottom panel). p90 RSK Phosphorylates Bad at Ser 112 condensed nuclei and cytoplasmic blebing (Fig. 7A). The reduction in the number of X-Gal-stained cells is consistent with apoptotic death of a large fraction of the cells overexpressing Bad (Fig. 7, A and C). To quantitate the level of death induced by Bad, we scored cell death based upon the number of X-Gal-positive cells exhibiting condensed nuclei from five different fields in each experiment. S.D. values were determined by counting comparable fields from three separate experiments. As shown in Fig. 7A, transfection of COS1 cells with wild-type Bad followed by 18-h serum starvation resulted in the death of a large majority of transfected cells (70 Ϯ 10%). Bad-mediated cell death was substantially suppressed by treatment with TPA for 6 h (34 Ϯ 5%).
To determine whether p90 RSK could also block Bad-mediated cell death, we cotransfected p90 RSK together with Bad into HEK293 cells. Transfection of HEK293 cells with wild-type Bad (Fig. 7B) or GST-Bad (Fig. 7C) followed by 18-h serum starvation resulted in dramatic cell death, as determined by the TUNEL assay (Fig. 7B) or by morphological analysis of ␤-galactosidase-positive cells (Fig. 7C) as described above for COS1 cells. In HEK293 cells, Bad-mediated cell death was completely suppressed by cotransfection with p90 RSK (Fig. 7B). Similarly, cotransfection with p90 RSK completely suppressed cell death mediated by GST-Bad (Fig. 7C). To examine the Bad phosphorylation sites required for p90 RSK suppression, we cotransfected GST-Bad S112A, S136A, and S112A/S136A mutants together with p90 RSK . Like wild-type GST-Bad, all GST-Bad mutants induce the death of a majority of transfected cells, and cell death induced by Bad S136A mutants can be suppressed by cotransfection with p90 RSK . However, cell death induced by Bad S112A or S112A/S136A mutants was not rescued by p90 RSK , suggesting that phosphorylation of Bad at Ser 112 is necessary for p90 RSK to function as a death suppresser.

DISCUSSION
Although a wide variety of growth and survival factors activate PKC, and PKC activation is known to promote cell survival, the PKC effectors mediating cell survival are largely unknown. Experiments in this study identify a mechanism whereby the activation of PKC promotes cell survival. PKC activation by TPA is shown to result in the rapid phosphorylation of Bad at Ser 112 , and this is correlated with inhibition of Bad-mediated cell death. While PKC does not appear to directly phosphorylate Bad at Ser 112 in vitro, we show by several different criteria that the protein kinase, p90 RSK , acts downstream of PKC and in turn can directly phosphorylate Bad at Ser 112 . First, TPA-activated p90 RSK phosphorylates Bad at Ser 112 in vitro; second, the effects of PKC inhibitors on Bad phosphorylation are tightly correlated with their effects on p90 RSK activity and phosphorylation in vivo. Finally, overexpression of p90 RSK in HEK293 cells stimulates Bad Ser 112 phosphorylation and effectively blocks Bad-induced cell death. Taken together, these findings identify a pathway involving the sequential activation of PKC and p90 RSK and phosphorylation of Bad at Ser 112 that may function generally to suppress Bad-mediated cell death and promote cell survival.
Studies by Zha et al. (5) first identified Bad Ser 112 and Ser 136 as two sites that when phosphorylated by IL-3 were able to block the cytotoxic effects of Bad. Several recent studies have identified a survival pathway leading to Bad phosphorylation that involves PI 3-kinase-dependent activation of Akt, followed by phosphorylation of Bad at Ser 136 (12,13,40,41). Datta et al. demonstrated that Akt activation results in the selective phosphorylation of Bad at Ser 136 but not Ser 112 . The pathway identified by Datta et al. is blocked by the PI 3-kinase inhibitors wortmannin and LY294002 and involves the Akt-dependent phosphorylation of Bad Ser 136 (12). The PKC/p90 RSK pathway we have identified is not blocked by PI 3-kinase inhibitors, leads primarily to Bad phosphorylation at Ser 112 , and does not involve Akt, since TPA does not appreciably stimulate Akt activity in the cells we have examined (data not shown). The  Cells were fixed, and X-gal staining was performed as described under "Experimental Procedures." Apoptotic cells were quantitated by scoring the X-gal-stained cells showing condensed nuclei and cytoplasmic blebs from five different comparable fields. The S.D. value was calculated based on three independent experiments. B, HEK293 cells were transfected with empty vectors or plasmids encoding wild-type Bad alone or together with pMT2-p90 RSK as indicated. 24 h after transfection, cells were serum-starved for 18 h, fixed, and analyzed by the TUNEL assay as described under "Experimental Procedures." The percentage of apoptotic cells was quantitated by counting the amount of TUNEL positive cells from five different fields. The S.D. was calculated based on three independent experiments. C, HEK293 cells were transfected with the indicated combination of plasmids encoding Rous sarcoma virus-␤-galactosidase; GST-Bad; GST-Bad S112A, S136A, or S112A/S136A mutants; and/or pMT2-p90 RSK . 24 h after transfection, the cells were serum-starved for 18 h and then fixed and analyzed by X-gal staining. p90 RSK Phosphorylates Bad at Ser 112 34865 PKC/p90 RSK pathway therefore represents a route distinct from the PI 3-kinase/Akt pathway, leading to Bad phosphorylation and cell survival. Accumulating evidence suggests that multiple pathways converge on Bad to determine cell survival. Survival factors such as IL-3 also increases cAMP levels, resulting in activation of protein kinase A. Membrane-based protein kinase A has recently been identified as a Bad kinase phosphorylating Bad at Ser 112 (14). Calcineurin, a calcium-dependent phosphatase, can dephosphorylate Bad at both Ser 112 and Ser 136 and promote apoptosis in a calcium-dependent fashion (15). In addition, several recent studies have identified a Bad-independent pathway mediating cell survival in response to IL-4 and insulin-like growth factor-1 (7,8), suggesting the existence of Badindependent survival mechanisms.
The precise pathway leading from PKC to the activation of p90 RSK is still unclear. Since PKC has been reported to activate the MAPK kinase cascade at several levels including Ras and Raf (42,43), and MAPK can directly phosphorylate p90 RSK at Thr 562 in vitro and in vivo (34,39), it seems plausible to assume that PKC regulates p90 RSK via the MAPK cascade. However, blockade of MAPK with PD98059, a selective MEK inhibitor (44), has only a very small inhibitory effect on TPA-stimulated Bad Ser 112 phosphorylation. Because the MEK inhibitor PD98059 completely blocks TPA-stimulated MAPK phosphorylation (data not shown), it is unlikely that the inability of PD98059 to block p90 RSK activity and Bad phosphorylation arises from incomplete blockade of MAPK. These studies suggest that PKC activates p90 RSK in a fashion that does not depend upon the MAPK cascade. One such mechanism might be the recruitment of a PDK1-like enzyme to p90 RSK , since these kinases bind and phosphorylate PKC within its activation loop (45). Another might be direct phosphorylation of p90 RSK by PKC at activation sites such as Ser 381 . Ser 381 is located within the spacer region between p90 RSK NTD and CTD kinase domains and is a major target of autophosphorylation by the CTD (34,46). This site conforms to the consensus Phe-X-X-Phe-Ser-Phe located within the C-terminal sections of most PKC isoforms as well as Akt and p70 S6 kinase and is a critical site regulating the activity of Akt as well as p70 S6 kinase. Since PKC autophosphorylates at this site (47)(48)(49), it is possible that it may also phosphorylate similar sites on other kinases such as p90 RSK .
Activation of PKC by IL-3 and other survival factors has been observed in many different cell types, although the physiological role of PKC activation and the PKC isoforms mediating survival has been difficult to define. The observation that both IL-3 and granulocyte-macrophage colony-stimulating factor induce diacylglycerol without mobilizing Ca 2ϩ suggested a role for the diacylglycerol-dependent but Ca 2ϩ -independent novel PKC isoforms (50). Recently, transfection experiments overexpressing PKC⑀ demonstrated that this novel PKC isoform extends cell survival in the absence of IL-3 (51). PKC⑀ is involved in the protection of cardiac myocytes from hypoxiainduced cell death (52) as well as protection against tumor necrosis factor-␣-induced apoptosis (53). PKC⑀ is also known to increase the RNA and protein expression of Bcl-2 (51). Experiments in our study also suggest a role for the Ca 2ϩ -independent novel PKC isoforms based upon the potent inhibition seen with bisindolymaleimide I and GO6983 and the lack of inhibition using GO6976. The conventional PKC isoform PKC␣ has been recently implicated in an antiapoptotic response in COS1 cells (54) and in phosphorylating Bcl-2 at Ser 70 , a site critical for Bcl-2 antiapoptotic activity (25). The atypical PKC isoforms and also play a role in regulating cellular susceptibility to drug-induced (55) and UV-induced apoptosis (56). However, since activation of these isoforms is sensitive to wortmannin and is not activated by TPA, it seems unlikely that they contribute to TPA-induced Bad phosphorylation. p90 RSK has not been previously implicated in cell survival; however, the activity of this enzyme has long been known to be induced by many different growth and survival factors (13) Although this enzyme is phosphorylated and activated by MAPK and is considered a major output of the MAPK cascade, its biological significance has remained obscure, since few physiological substrates have been identified. The finding that NGF induces transcription of c-Fos via p90 RSK -dependent phosphorylation of CREB at Ser 133 (33) suggests that p90 RSK or its closely related isoform RSK-2 may play an important role in nerve growth factor-mediated regulation of gene expression via CREB and its DNA binding site, the cAMP-response element. TPA and PKC are reported to induce Bcl-2 expression and rescue immature B cells from apoptosis via a cAMP-responsive element site located within the Bcl-2 promoter (57). Furthermore, IL-3 stimulates phosphorylation of CREB at Ser133 (58). These observations suggest that the PKC/p90 RSK pathway outlined in this study may function more generally as part of a signal transduction cascade transmitting survival signals to the Bcl-2 checkpoint, regulating the phosphorylation status and cellular expression of Bad, Bcl-2, and perhaps other as yet unidentified targets.
Induction of apoptosis appears to be the major mechanism of action of most, if not all, effective chemotherapeutic and radiation-based cancer treatments (59). Hence, the success of chemo-and radiation-based therapy may largely be dependent upon their ability to override existing anti-apoptotic survival pathways. There is considerable evidence that the PKC survival pathway may limit apoptosis induced by radiation and chemotherapy. PKC decreases c-Myc-induced apoptosis in small lung cancer cells (60) and protects endothelial cells against radiation-induced apoptosis (61). PKC inhibitors stimulate apoptosis in human malignant glioma cells (21) and lymphoma cells (24), while TPA prevents apoptosis in chronic lymphocytic leukemia cells (62). In addition, tumor cell sensitivity to araC is greatly enhanced by disruption of the PKC pathway (23). Our findings suggest that disruption of the PKC/ p90 RSK pathway may also serve to augment chemotherapyinduced apoptosis and suggest p90 RSK as a novel target for intervention.