90-kDa Ribosomal S6 Kinase Is Phosphorylated and Activated by 3-Phosphoinositide-dependent Protein Kinase-1*

90-kDa ribosomal S6 kinase-2 (RSK2) belongs to a family of growth factor-activated serine/threonine kinases composed of two kinase domains connected by a regulatory linker region. The N-terminal kinase of RSK2 is involved in substrate phosphorylation. Its activation requires phosphorylation of the linker region at Ser369, catalyzed by extracellular signal-regulated kinase (ERK), and at Ser386, catalyzed by the C-terminal kinase, after its activation by ERK. In addition, the N-terminal kinase must be phosphorylated at Ser227 in the activation loop by an as yet unidentified kinase. Here, we show that the isolated N-terminal kinase of RSK2 (amino acids 1–360) is phosphorylated at Ser227 by PDK1, a constitutively active kinase, leading to 100-fold stimulation of kinase activity. In COS7 cells, ectopic PDK1 induced the phosphorylation of full-length RSK2 at Ser227and Ser386, without involvement of ERK, leading to partial activation of RSK2. Similarly, two other members of the RSK family, RSK1 and RSK3, were partially activated by PDK1 in COS7 cells. Finally, our data indicate that full activation of RSK2 by growth factor requires the cooperation of ERK and PDK1 through phosphorylation of Ser227, Ser369, and Ser386. Our study extend recent findings which implicate PDK1 in the activation of protein kinases B and C and p70S6K, suggesting that PDK1 controls several major growth factor-activated signal transduction pathways.

90-kDa ribosomal S6 kinase-2 (RSK2) belongs to a family of growth factor-activated serine/threonine kinases composed of two kinase domains connected by a regulatory linker region. The N-terminal kinase of RSK2 is involved in substrate phosphorylation. Its activation requires phosphorylation of the linker region at Ser 369 , catalyzed by extracellular signal-regulated kinase (ERK), and at Ser 386 , catalyzed by the C-terminal kinase, after its activation by ERK. In addition, the N-terminal kinase must be phosphorylated at Ser 227 in the activation loop by an as yet unidentified kinase. Here, we show that the isolated N-terminal kinase of RSK2 (amino acids 1-360) is phosphorylated at Ser 227 by PDK1, a constitutively active kinase, leading to 100-fold stimulation of kinase activity. In COS7 cells, ectopic PDK1 induced the phosphorylation of full-length RSK2 at Ser 227 and Ser 386 , without involvement of ERK, leading to partial activation of RSK2. Similarly, two other members of the RSK family, RSK1 and RSK3, were partially activated by PDK1 in COS7 cells. Finally, our data indicate that full activation of RSK2 by growth factor requires the cooperation of ERK and PDK1 through phosphorylation of Ser 227 , Ser 369 , and Ser 386 . Our study extend recent findings which implicate PDK1 in the activation of protein kinases B and C and p70 S6K , suggesting that PDK1 controls several major growth factor-activated signal transduction pathways.
The 90-kDa ribosomal S6 kinases (RSK1-3) 1 are a family of broadly expressed serine/threonine kinases that are activated by extracellular signal-regulated protein kinases (ERK1 and -2) in response to many growth factors, polypeptide hormones, and neurotransmitters (Refs. 1-3; reviewed in Refs. 4 and 5). Inactivating mutations in the RSK2 gene are responsible for the human Coffin-Lowry syndrome, which is characterized by severe mental retardation and progressive skeletal deformations (6,7). At the cellular level, RSK2 has been proposed to regulate the activity of the transcription factor cAMP response element-binding protein (CREB) (8,9) and the transcriptional co-activators p300 and CREB-binding protein (10). RSK1 can phosphorylate the estrogen receptor and enhance its transcriptional activity (11) and may also be an activator of the transcription factor NFB through phosphorylation of IB␣ (12,13). Besides a role in transcriptional control, findings in Xenopus laevis oocytes implicate RSK in stimulation of meiosis via inactivation of the p34 cdc2 -inhibitory kinase Myt1 (14). Finally, RSK can phosphorylate the Ras GTP-exchange molecule SOS and may thereby exert negative feedback of the Ras-ERK pathway (15). Recently, a family of two mitogen-and stress-activated protein kinases (MSK) has been discovered, which resembles RSK in having two kinase domains and other structural hallmarks (16,17). MSK is activated by ERK as well as by p38 mitogen-activated protein kinase in response to growth factors and various cellular stress stimuli.
The two kinase domains of RSK are connected by a ϳ100amino acid sequence, referred to here as the linker. The substrates of RSK identified so far are phosphorylated by the N-terminal kinase (NTK) (18 -20), whereas the C-terminal kinase (CTK) and the linker participate in the regulation of the NTK (18,20,21). The mechanism of activation of RSK is complex and involves phosphorylation of at least four sites ( Fig.  1), as demonstrated with RSK1 in cells treated with phorbol ester, a potent activator of ERK (21). As a probable sequence of events, ERK phosphorylates two sites, one in the linker and one in the activation loop of the CTK, leading to its activation (20 -22). The CTK then phosphorylates an additional site in the linker (21,23). Dual phosphorylation of the linker leads to increased phosphorylation of a serine residue in the activation loop of the NTK and full kinase activity (21). The critical role of this serine is indicated by the finding that its mutation to alanine abolishes the ability of all three RSK isotypes to phosphorylate exogenous substrates in vitro (6,18,21) and that this mutation in RSK2 can cause the Coffin-Lowry syndrome (6). The phosphorylation sites in the activation loop of the NTK and linker of RSK are situated in analogous positions to regulatory phosphorylation sites in p70 S6K (24), protein kinase B (PKB) (25), and protein kinase C (PKC) (26,27) (Fig. 1), suggesting a common structural basis of activation for these protein kinases.
The identity of the kinase that phosphorylates the serine in the activation loop of the NTK is unclear. In RSK1, phosphorylation of this serine was greatly reduced when either the NTK or the CTK was inactivated by mutagenesis (21). These observations led to the suggestion that the NTK catalyzes the phosphorylation of the serine in its activation loop. The consensus substrate phosphorylation sequence of RSK is: Arg/Lys-X-Arg-X-X-Ser (19). However, the serine in the activation loop of the NTK has lysine at the Ϫ3 position in RSK1 and RSK2 and an acidic residue at the Ϫ5 position in all RSKs, residues that in synthetic peptides result in poor phosphorylation by RSK (19). Furthermore, the serine in the activation loop of the NTK of RSK1 showed considerable basal phosphorylation under conditions where RSK1 was inactive (21). Consequently, the site may be targeted by a protein kinase other than the NTK of RSK. Recently, several studies have indicated that the analogous threonine in the activation loop of PKB (28,29), p70 S6K (30,31), PKC (32), and protein kinase A (33) are phosphorylated by 3-phosphoinositide-dependent protein kinase-1 (PDK1). PDK1 is broadly expressed and has a high level of constitutive activity, which does not seem to be modulated by extracellular stimuli (28,29,31). PDK1 contains a pleckstrin homology (PH) domain which may bind the phospholipid products of the phosphoinositide 3-OH-kinase and localize PDK1 to the plasma membrane for regulation of, e.g., PKB (34). The regulatory serine in the activation loop of the NTK of RSK is situated in a putative PDK1 consensus phosphorylation sequence (Fig. 1).
Based on these observations, we investigated the activation mechanism of RSK and addressed the possible involvement of PDK1. We show that PDK1 phosphorylates the serine in the activation loop of the NTK of RSK2, leading to substantial activation of the kinase in vitro and in vivo. Our results indicate that constitutively active PDK1 may account for basal RSK activity in cells, whereas stimulation of RSK by growth factors requires the collaborative regulation by ERK and PDK1.

EXPERIMENTAL PROCEDURES
Materials-Human recombinant epidermal growth factor (EGF) was from PreproTech Inc. (Rocky Hill, NJ). S6 peptide (residues 231-239 of human 40 S ribosomal protein 6: RRLSSLRA) and phosphopeptidespecific antibodies to RSK (catalog nos. 06-824 and 06-826) were from Upstate Biotechnology (Lake Placid, NY). Antibody to the hemagglutinin (HA) and the myc epitope were from the 12CA5 and 9E10 mouse hybridoma cell lines, respectively. Radioisotopes were from NEN Life Science Products. Other chemicals were from Sigma.
Plamid Constructs-N-terminally HA epitope-tagged rat RSK1, murine RSK2, or human RSK3 in pMT2 (35) were kindly provided by Christian Bjørbaek (Beth Israel Hospital, Boston, MA). RSK1, originally cloned from rat hepatoma cell cDNA (36), contained a glutamine insertion at position 157. A glutamine at this site has not been reported in any RSK sequence, nor have we found it in RSK1 cDNA from rat liver 2 ; consequently, it was deleted by the QuickChange TM mutagenesis procedure (Stratagene) to generate the RSK1 used in the present study. Interestingly, the RSK1 without the glutamine insertion was more responsive to PDK1 than the RSK1 that contained the glutamine (data not shown). pcDNA3-HA-ERK2 was provided by Klaus Seedorf (Hagedorn Research Laboratory, Gentofte, Denmark). pCMV5-myc-PDK1 51-556 and pCMV5-myc-PDK1 51-556 K111Q were generated as described (31). To generate full-length PDK1 constructs (amino acids 1-556), the DNA sequence encoding amino acids 1-50 of human PDK1 was amplified from a placenta cDNA library by polymerase chain reaction (PCR) and ligated into pCMV5-myc-PDK1 51-556 and pCMV5-myc-PDK1 51-556 K111Q. Kinase-deficient PDK1 was generated using pCMV5-myc-PDK1 1-556 K111Q (which is not entirely inactive) as a template for site-directed mutagenesis of Asp 205 and Asp 223 to Asn using the QuickChange TM procedure. To delete the PH domain from PDK1, the sequence corresponding to amino acids 219 -458 of PDK1 was amplified by PCR with primers introducing a stop codon and a BamHI site at the 3Ј end. The PCR product was digested with PmlI and BamHI and inserted into pCMV5-myc-PDK1 51-556 digested with same enzymes. pMT2-HA-RSK2 1-360 will be described elsewhere. 3 All point mutants of RSK2 1-360 or full-length RSK2 were generated using the QuickChange TM procedure and pMT2-HA-RSK2 1-360 or pMT2-HA-RSK2 as template. pGEX-4T-1-HA-RSK2 was generated by introducing, by PCR, a BamHI site immediately upstream of the HA tag in pMT2-HA-RSK2, followed by subcloning of RSK2 into pGEX-4T-1 using the BamHI site and the NotI site in the murine RSK2 3Ј-untranslated region. pGEX-4T-1-HA-RSK2 1-373 K100A was generated by amplifying the corresponding DNA sequence of RSK2(K100A) by PCR, using primers introducing a BamHI site upstream and a termination signal followed by a NotI site downstream of the sequence, and insertion into pGEX. pGEX-4T-1-RSK2 1-360 S227E was constructed in the same way, except that RSK2(S227E) was used as the PCR template and a BamHI site was used as the 3Ј-end subcloning site. pGEX-4T-1-RSK2 1-360 was generated by first excising the DNA sequence corresponding to RSK2 1-373 from pGEX-4T-1-HA-RSK2 1-373 K100A, using a XhoI site between the HA tag and RSK2 and an EcoRI site in the pGEX multiple cloning site, and replacing it with the sequence of RSK2 1-360 excised from pGADGH-HA-RSK2 1-360 , described elsewhere. 3 Glutathione S-Transferase (GST)-RSK Fusion Protein Synthesis and Purification-Escherichia coli cells (BL21) transformed with various forms of RSK2 in the pGEX-4T-1 vector were grown at 30°C in yeast extract-tryptone medium to an optical density (A 600 ) of 1.2. Expression of recombinant protein was induced by the addition of 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2 h. The pelleted bacteria were lysed by four freeze/thaw cycles (37), resuspended in phosphate-buffered saline with 1% (v/v) Triton X-100 and various protease inhibitors, and incubated for 30 min. Bacterial extract was clarified by centrifugation at 12,000 ϫ g, and GST-RSK was collected on glutathione-Sepharose beads (Amersham Pharmacia Biotech), washed, and eluted with 10 mM FIG. 1. Domain structure and regulatory phosphorylation sites of RSK. RSK is composed of two kinase domains connected by a regulatory linker region. The C-terminal tail contains a docking site responsible for complex formation with ERK (42,43). The locations of the four known regulatory phosphorylation sites in RSK and the surrounding amino acid sequences (single-letter code) are shown. Amino acid numbering refers to murine RSK2. Ser 369 and Thr 577 are phosphorylated by ERK (21,22), Ser 386 by the C-terminal kinase (21,23), and Ser 227 by PDK1, as suggested by the present study. Three of the sites are similarly positioned as regulatory phosphorylation sites in p70 S6K , PKB␣, and some PKC isoforms with conserved amino acids surrounding the phosphorylated residues (indicated by boldface type). The putative PDK1 consensus phosphorylation sequence is indicated by italics.
Protein Quantitation-Samples were solubilized in SDS-PAGE sample buffer (2% sodium dodecyl sulfate, 62 mM Tris (pH 6.8), 10% glycerol, 5% 2-␤-mercaptoethanol, 0.1% (w/v) bromphenol blue) and fractionated by 10% PAGE. Proteins were stained by incubation of the gel for 20 min in 5000-fold dilution of Sypro Orange TM (Molecular Probes, Eugene, OR) in 7.5% (v/v) acetic acid. After a brief wash in 7.5% (v/v) acetic acid, the gel was scanned on a STORM TM FluorImager (Molecular Dynamics) and quantified by the ImageQuant TM software using a dilution series of the broad range molecular mass marker from Molecular Probes as a standard.
Transfection and Immunoprecipitation-COS7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C in atmospheric air containing 5% CO 2 . Monolayers of ϳ3.2 ϫ 10 5 cells in 9.6-cm 2 dishes were incubated for 4 -5 h in serum-free medium with a total of 1.5 g of DNA complexed with 12 l of LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instructions. In double transfections, 0.75 g of each DNA construct were used. After transfection, cells were cultured for 48 h and then washed twice with serum-free medium. After incubation for 3 h in the absence of serum, the cells were exposed (or not) to epidermal growth factor, washed with phosphate-buffered saline, and solubilized for 15 min in 500 l of lysis buffer (1% Nonidet P-40, 0.5% Triton X-100, 150 mM NaCl, 50 mM Tris-HCl (pH 7.4), 10% glycerol, 1 mM Na 3 VO 4 , 5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 M leupeptin, 10 M pepstatin, and 200 kallikrein inhibitor units/ml aprotinin) on ice. Subsequent manipulations were performed at 0 -4°C. Cell extracts were clarified by centrifugation for 15 min at 14,000 ϫ g, and the supernatant was incubated for 3 h with antibody with the addition of 20 l of 50% protein A-or protein G-agarose beads (Amersham Pharmacia Biotech) during the final 45 min. Agarose beads/antibody complexes were precipitated by centrifugation, washed five times with lysis buffer, drained, and dissolved in SDS-PAGE sample buffer. For immunocomplex kinase assays, the final two washes were with kinase assay buffer A (30 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 , 1 mM dithiothreitol) for RSK/ERK assays or with kinase assay buffer B (50 mM Tris-HCl (pH 7.5), 10 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol) for PDK1 kinase reactions.
RSK and ERK Kinase Assays-Agarose beads with immunoprecipitated kinase were drained with a syringe and resuspended in 20 l of 1.5ϫ kinase assay buffer A. The kinase reaction was initiated by the addition of 10 l of substrate mixture containing ATP (300 M, 5 Ci of [␥-32 P]ATP) and S6 peptide (800 M) or myelin basic protein (24 M) for RSK and ERK assays, respectively. After 10 min at 30°C (the reaction was linear with time under these conditions), 20 l of the supernatant was removed with a syringe (leaving behind the beads with precipitated kinase) and spotted onto phosphocellulose paper (Whatman P-81), which was washed five or six times with 150 mM orthophosphoric acid, whereafter [ 32 P]phosphate incorporated into protein substrate was quantified on a STORM TM PhosphorImager using the ImageQuant software (Molecular Dynamics) or by liquid scintillation counting. A reaction blank (a kinase assay performed on non-transfected cells) was subtracted from all values. For quantitation of RSK in the immunoprecipitates, the reaction mixture remaining after the kinase assay was solubilized in 2ϫ SDS-PAGE sample buffer and subjected to protein quantitation as described above.
In Vitro Activation of GST-RSK-myc-tagged PDK1, kinase-deficient PDK1, or RSK2 was expressed in COS7 cells and immunopurified with 9E10 antibody. Agarose beads with precipitated kinase (50 -100 ng/ assay point, derived from 2-5 cm 2 of transfected cell monolayer) were drained with a syringe and resuspended in 40 l of kinase assay buffer B with addition of 35 M ATP and active GST-ERK2 (100 ng, Upstate Biotechnology Inc, Lake Placid, NY), GST-RSK2 (ϳ200 ng), GST-RSK2 1-360 (1 g), or GST-RSK2 1-360 S227E (1 g) in the combinations indicated in the figure legends. PDK1, RSK2, or ERK were then allowed to activate the GST-RSK proteins for 1.5 h at 25°C with vigorous shaking and addition of an extra 0.5 g of 9E10 antibody after 1 h. Thereafter, the agarose beads were pelleted by centrifugation, 20 l of the supernatant, containing GST-RSK, were removed and used for RSK assay. For each assay condition, a reaction was performed without addition of S6 peptide, which was considered the blank value and which was subtracted. However, in Fig. 2B, blank values in measurements of RSK2 1-360 activation by PDK1 or RSK2 were reactions without the addition of GST-RSK2  .
In Vitro Phosphorylation of GST-RSK-One g of GST-RSK2 1-375K100A, GST-RSK2 1-360¸o r GST-RSK2 1-360 S227E was incubated with immunoprecipitated PDK1, RSK2, or alone as described under "In Vitro Activation of GST-RSK" except that the 1.5-h incubation was performed in the presence of 10 Ci of [␥-32 P]ATP. After the incubation period, unlabeled ATP was added to 800 M to stop the reaction. The agarose beads were pelleted by centrifugation, and 25 l of the supernatant, containing the GST-RSK, were incubated with thrombin to remove the GST moiety, whereafter an aliquot was subjected to SDS-PAGE and autoradiography.
[ 32 P]Orthophosphate Metabolic Labeling-Transfected COS7 cells were washed twice with RPMI medium containing no phosphate or serum and incubated for 3.5 h in this medium supplemented with carrier-free [ 32 P]orthophosphate (0.75 mCi/ml). After two washes with phosphate-buffered saline, the cells were solubilized with lysis buffer. Immunoprecipitation was carried out as described above, except that the clarified cell extracts were first incubated for 30 min with protein A-agarose to adsorb proteins that bind unspecifically to the beads, then centrifuged, whereafter the supernatant was incubated with antibody. Immunoprecipitates were washed six times with lysis buffer, drained with a syringe, and subjected to SDS-PAGE and autoradiography.
Immunoblotting-Samples dissolved in SDS-PAGE sample buffer were fractionated by 10% SDS-PAGE and electroblotted onto Hybond polyvinylidene difluoride membrane (Amersham Pharmacia Biotech). Membranes were blocked overnight with 5% (w/v) nonfat dry milk (Carrefour, France) and 0.1% (v/v) Tween 20 in Tris-buffered saline (pH 7.6), followed by incubation with primary antibody as indicated in the figure legend. The primary antibody was visualized by incubation with an appropriate anti-primary antibody coupled to either alkaline phosphatase or to horseradish peroxidase followed by enhanced chemifluorescence development (Amersham Pharmacia Biotech) and STORM TM scanning or by enhanced chemiluminescence development (Amersham Pharmacia Biotech) and autoluminography.

PDK1 Phosphorylates and Activates the N-terminal Kinase of RSK2 in Vitro-
We first investigated whether the isolated NTK of RSK2 can autophosphorylate or be phosphorylated by RSK2 or PDK1 in vitro. The deletion mutant RSK2 1-360 , which lacks all known phosphorylation sites in RSK, except for Ser 227 in the activation loop, was expressed as a GST fusion protein in E. coli and purified. PDK1 and RSK2 were transiently expressed in COS7 cells and immunopurified. Prior to lysis, RSK2 was activated by exposure of the cells to EGF. RSK2 1-360 was incubated for 1.5 h with Mg[␥-32 P]ATP alone or with PDK1 or active RSK2 followed by SDS-PAGE and autoradiography. As shown in Fig. 2A (lanes 1 and 2), RSK2 1-360 did not autophosphorylate and was poorly phosphorylated by RSK2. In contrast, RSK2 1-360 was heavily phosphorylated after incubation with PDK1 ( Fig. 2A, lane 4). This phosphorylation was catalyzed by PDK1, rather than by a co-immunopurified kinase, since RSK2  showed no phosphorylation by incubation with a kinase-deficient mutant of PDK1 ( Fig. 2A, lane 3). PDK1 appeared to phosphorylate RSK2 1-360 at Ser 227 , since the mutant RSK2 1-360 S227E was not phosphorylated by incubation with PDK1 ( Fig. 2A, compare lanes 5 and 7). No change in phosphorylation was observed by incubation of RSK2 1-360 S227E with active RSK2.
These findings indicate that activation loop phosphorylation is necessary and sufficient for activation of the NTK of RSK and that PDK1 can catalyze this event. In contrast, RSK2 is incapable of intra-or intermolecular autophosphorylation at this site and hence RSK2 is incapable of autoactivation.
PDK1 Phosphorylates and Activates the N-terminal Kinase of RSK2 in Vivo-We next investigated whether HA epitopetagged RSK2 1-360 is phosphorylated by PDK1 or RSK2 when coexpressed in COS7 cells. Lysates prepared from transiently transfected cells were subjected to SDS-PAGE and immunoblotting with anti-HA antibody in order to detect decreased electrophoretic mobility of RSK2 1-360 , indicative of its phosphorylation. Coexpression with RSK2, followed by exposure to EGF, did not affect the electrophoretic migration of RSK2 1-360 , whereas coexpression with PDK1 induced a profound mobility shift (Fig. 3A, lanes 1-3). Scanning of the two bands in lane 3 showed that more than 80% of RSK2 1-360 migrated with decreased mobility, indicating that PDK1 efficiently phosphorylates the NTK of RSK2 in vivo. RSK2 1-360 S227E migrated at an intermediate position relative to phosphorylated/unphosphorylated RSK2 1-360 , and the band was not shifted by coexpression with PDK1 (Fig. 3A, lanes 4 -6), indicating that PDK1 phosphorylates the NTK of RSK2 at Ser 227 .
The kinase activity of immunoprecipitated RSK2 1-360 was very low when RSK2 1-360 was expressed alone in COS7 cells (Fig. 3B). Coexpression with RSK2, followed by exposure to EGF, induced a 5-fold increase in RSK2 1-360 activity (Fig. 3B,  lane 2). The activation of RSK2 1-360 was mediated by coexpressed RSK2, since it was not observed in control cells treated only with EGF (data not shown). Coexpression with PDK1 increased 120-fold the RSK2 1-360 activity in the precipitates (Fig. 3B, lane 3). Depending on the experiment, however, coexpression with PDK1 also increased the protein level of RSK2 1-360 by 2-3-fold (see Fig. 3A), so the actual degree of stimulation of RSK2 1-360 by PDK1 was 40 -60-fold. The specific activity of RSK2 1-360 activated by PDK1 corresponded to 30 -60% of the activity of full-length RSK2 activated by EGF (data not shown). PDK1 and RSK2 were equally abundant in the cells (Fig. 3B, inset), showing that their differential ability to activate RSK2 1-360 was not due to a difference in protein levels. PDK1 and RSK2 did not increase the activity of RSK2 1-360 S227E above the basal level (Fig. 3B, lanes 4 -6), indicating that both kinases stimulate RSK2 1-360 by phosphorylation of Ser 227 .
Finally, the ability of PDK1 to phosphorylate the NTK of RSK2 in vivo was analyzed in COS7 cells metabolically labeled with [ 32 P]orthophosphate. RSK2 1-360 was weakly phosphorylated in cells in the basal state or in cells treated with EGF (Fig.  4A, lanes 1 and 2). In contrast, coexpression with PDK1 resulted in strong labeling of RSK2 1-360 , but not of RSK2 1-360 S227E (Fig. 4, lanes 3 and 5).
These data confirm our findings in vitro that autophosphorylation and autoactivation of the NTK of RSK2 occurs inefficiently and show that PDK1 is an efficient activator of the NTK of RSK2 in vivo by phosphorylation of Ser 227 . After the incubation period, the agarose beads were removed by centrifugation, GST was cleaved from the NTK by thrombin digestion, whereafter the samples were subjected to SDS-PAGE and autoradiography. The experiment was repeated twice with similar results. B, after the incubation period, the agarose beads were removed by centrifugation, and the kinase activity of GST-RSK2 1-360 or GST-RSK2 1-360 S227E was determined. Data are expressed as percent of the PDK1 51-556 -stimulated value (ranging from 0.5 to 1 ϫ 10 6 cpm) and are means Ϯ S.E. of three experiments performed in duplicate. The data of the following bars were compared by non-paired t test and were different (p Ͻ 0.01): 1 versus 4. In contrast, the following data were not statistically different (p Ͼ 0.7): 1 versus 2, 1 versus 3, 5 versus 6, 5 versus 7. Inset to B, agarose beads from assays 2 and 4, containing immobilized myc-RSK2 or myc-PDK1 51-556 , respectively, were subjected to SDS-PAGE and immunoblotting with antibody to the myc epitope tag to visualize the amount of the kinases. C, GST-RSK2 1-373 K100A (kinase-deficient) was incubated alone or together with myc-PDK1 51-556 as described in A and thereafter subjected to SDS-PAGE and autoradiography. The experiment was repeated three times with similar results .   FIG. 3. PDK1 phosphorylates the isolated NTK of RSK2 at Ser 227 and stimulates its kinase activity in COS7 cells. Cells were transfected with plasmids expressing HA-RSK2 1-360 , HA-RSK2 1-360 S227E, myc-RSK2, or myc-PDK1 or with empty plasmid, indicated by Ϫ. Forty-eight hours after transfection and following a 3-h serum starvation period, the cells were lysed. Prior to lysis, cells expressing myc-RSK2 were exposed to 20 nM EGF for 15 min. A, an aliquot of the cleared lysates were subjected to SDS-PAGE and immunoblotting with antibody to the HA epitope tag on the RSK2 1-360 constructs. The experiment was performed four times with similar results. B, HA-RSK2 1-360 or HA-RSK2 1-360 S227E were precipitated from the lysates remaining from A, using antibody to the HA epitope tag, and subjected to kinase assay. Data are expressed as percent of the PDK1stimulated value (ranging from 0.6 to 1.2 ϫ 10 6 cpm) and are mean values Ϯ S.E. of three experiments performed in duplicate. The data of the following bars were compared by non-paired t test and were different (p Ͻ 0.02): 1 versus 2, 1 versus 3. In contrast, the following data were not statistically different (p Ͼ 0.5): 4 versus 5, 4 versus 6. Inset to B, to control for equal expression of myc-RSK2 and myc-PDK1, an aliquot of cleared lysate from assays 2 and 3 in B was subjected to SDS-PAGE and immunoblotting for the myc epitope tag.
Full-length RSK1, RSK2, and RSK3 Are Activated when Coexpressed with PDK1 in COS7 Cells-The ability of PDK1 to activate full-length RSK was investigated in COS7 cells transfected with HA epitope-tagged versions of RSK1, RSK2, or RSK3, alone or together with PDK1. For comparison, activation of RSK after treatment of cells with EGF was measured. The activity of RSK was measured using S6 peptide as a substrate, which is phosphorylated by the NTK, but not by the CTK of RSK (20). All three RSKs had relatively high basal activity in COS7 cells that was stimulated 3-10-fold by EGF, depending on isoform (Fig. 5A), in agreement with a previous study (35). Coexpression with PDK1, increased the activity of RSK1 to approximately 40% of the EGF-stimulated value. In contrast, PDK1 increased the activity of RSK2 and RSK3 to 70 -80% of that observed in cells exposed to EGF (Fig. 5A). Measurements of the protein amounts of RSK in the immunoprecipitates showed that PDK1 induced a ϳ1.2-fold increase of RSK2 and a 3-4-fold increase of RSK3. Accordingly, the activity data shown in Fig. 5A are normalized to RSK protein.
Immunoblotting for PDK1 showed that its protein level was ϳ50% reduced by coexpression with RSK1 compared with coexpression with RSK2 or RSK3 (Fig. 5B). However, variable expression levels of PDK1 resulted in the same activation of RSK (Fig. 6), suggesting that PDK1 is not limiting in these cotransfection experiments and that the lower level of activation of RSK1 by PDK1 is not due to less PDK1 expression.
Mechanism of Activation of RSK2 by PDK1-In the present study, PDK1 51-556 (31) and full-length PDK1 were used interchangeably. Although the two forms were expressed at somewhat different levels in COS7 cells (Fig. 6A), they stimulated RSK2 activity to the same degree (Fig. 6B).
We investigated whether the PH domain of PDK1 is required for activation of RSK2. However, a mutant of PDK1, in which the PH domain had been deleted, stimulated RSK2 activity to the same extent as wild-type PDK1 (Fig. 6B), suggesting that targeting to the plasma membrane, directed by the PH domain, is not involved in the activation of RSK2 by PDK1. Furthermore, wortmannin, an inhibitor of the phosphoinositide 3-OHkinase, did not inhibit the stimulation of RSK2 by PDK1 (data not shown).
In order to investigate whether PDK1 activates RSK indirectly, via its upstream activator ERK, we measured whether PDK1 was able to stimulate the activity of ERK2 in cells coexpressing the two kinases. PDK1, however, had no effect on ERK2 activity in the cells, in contrast to EGF, which induced a 10-fold increase (Fig. 7).
Phorbol ester-induced activation of RSK1 involves phosphorylation of four sites in the linker and kinase regions (21). We analyzed whether the corresponding phosphorylation sites in RSK2 are involved in its activation by PDK1 by using RSK2 in which the phosphorylation sites had been mutated and phosphopeptide-specific antibodies directed against the two regulatory phosphorylation sites in the linker of rat RSK1 (21). These antibodies were found to cross-react with phosphorylated murine RSK2.
The mutant RSK2(S227E) had low basal activity compared with wild-type RSK2 and was not activated by coexpression with PDK1 (Fig. 8A). RSK2(S227E), however, was greatly stimulated by EGF, indicating that the S227E mutation was not obstructive to RSK2 function but specifically eliminated the serine through which PDK1 exerts its stimulatory effect. Immunoblotting of the precipitated RSK2 with phosphopeptidespecific antibodies showed that EGF induced strong phosphorylation of Ser 386 and Ser 369 in both wild-type RSK2 and RSK2(S227E) (Fig. 8, B and C, lanes 2 and 5). PDK1 induced no phosphorylation of Ser 369 in RSK2 (Fig. 8C), but, surprisingly,  5. Activation of full-length RSK1, RSK2, and RSK3 by coexpression with PDK1 in COS7 cells. Cells were transfected with plasmids expressing HA epitope-tagged versions of the three RSK isotypes together with myc-PDK1 51-556 or with empty plasmid, indicated by Ϫ. Forty-eight hours after transfection and following a 3-h serum starvation period, cells were exposed, or not, to 20 nM EGF for 15 min and lysed. A, the RSK isoforms were precipitated from the cleared lysates with antibody to the HA epitope and their kinase activity was determined. After the kinase assay, the amount of RSK in the precipitates was quantified and kinase activity was normalized to RSK protein. Data are expressed as percentage of the EGF-stimulated value and are means Ϯ S.E. of three independent experiments performed in duplicate. The values of basal versus PDK1-stimulated activity of each of the three RSK isoforms were different by non-paired t test (p Ͻ 0.02). B, cells transfected with expression plasmids as indicated were lysed with SDS sample buffer and subjected to SDS-PAGE and immunoblotting with antibody to the myc epitope tag present on PDK1.
induced strong phosphorylation of Ser 386 (Fig. 8B, lane 3). Furthermore, RSK2(S227E) showed high basal phosphorylation at Ser 386 , that was not enhanced by PDK1 (Fig. 8B, compare lanes 4 and 6). Immunoblotting for the HA tag on the RSK2 constructs confirmed that roughly equal amounts of RSK were present in the precipitates (Fig. 8D). RSK2(S386A) had low basal activity compared with wildtype RSK2 and could not be activated by EGF (Fig. 8A). Furthermore, RSK2(S386A) showed 60% decreased activation by PDK1 compared with wild-type RSK2 (Fig. 8A). Thus, induction of Ser 386 phosphorylation is important for activation of RSK2 by EGF as well as by PDK1.
We next investigated the effect of mutating the serine or threonine in the two regulatory sites phosphorylated by ERK. Activation by EGF of RSK2(S369A) (Fig. 9A) and RSK2(T577A) (data not shown) was reduced by 50% compared with wild-type RSK2, whereas mutation of both sites, abolished activation of RSK2 by EGF (Fig. 9A). In contrast, these mutations had no effect on the ability of PDK1 to activate RSK2, nor did the mutations decrease the basal level of RSK2 activity (Fig. 9A). EGF induced strong phosphorylation of Ser 386 in wild-type RSK2 and in RSK2(S369A), but induced no Ser 386 phosphorylation in RSK2(S369A/T577A) (Fig. 9B, lanes 2, 5, and 8).
Since the double mutant should contain an ERK-unresponsive CTK, this finding suggests that EGF induces the phosphorylation of Ser 386 in RSK2 by activation of the CTK. PDK1 induced robust phosphorylation of Ser 386 in wild-type RSK2 and in RSK2(S369A), whereas in RSK2(S369A/T577A) the effect was sometimes less pronounced (Fig. 9B, lanes 3, 6, and  9). Finally, immunoblotting for the HA tag on the RSK2 constructs showed roughly equal amounts of wild-type RSK2 and RSK2(S369A) in the precipitates, whereas the double mutant seemed slightly less abundant (Fig. 9C). Compared with wildtype RSK2, however, the double mutant is hypophosphorylated and will migrate less as a smear and therefore appear slightly less abundant.
Our findings indicate that EGF activates RSK2 by a mechanism similar to the one described for phorbol ester-induced activation of RSK1 (21) Forty-eight hours after transfection and following a 3-h serum starvation period, cells were exposed, or not, to 20 nM EGF for 15 min and lysed. ERK was precipitated with antibody to the HA epitope, and its kinase activity was determined. Data are expressed as percentage of the EGF-stimulated value and are means Ϯ S.E. of three independent experiments performed in duplicate. The values of basal versus PDK1stimulated ERK2 activity were not different by non-paired t test (p Ͼ 0.8).
FIG. 8. Activation and phosphorylation of wild-type RSK2 and mutant RSK2(S227E) or RSK2(S386A) by EGF or PDK1 in COS7 cells. Cells were transfected with plasmids expressing HA-RSK2, HA-RSK2(S227E), HA-RSK2(S386A), or myc-PDK1 or with empty plasmid, indicated by Ϫ. Forty-eight hours after transfection and following a 3-h serum starvation period, cells were exposed, or not, to 20 nM EGF for 15 min and lysed, whereafter RSK was precipitated with antibody to the HA epitope tag. A, the kinase activity of precipitated RSK was determined. Data are expressed as percentage of EGF-stimulated wild-type RSK2 activity and are means Ϯ S.E. of three independent experiments performed in duplicate. The data of the following bars were compared by non-paired t test and were different (p Ͻ 0.01): 1 versus 4, 3 versus 9. In contrast, bar 4 was not different from bar 6 (p Ͼ 0.8). After the kinase assay, aliquots of the precipitated RSKs were subjected to SDS-PAGE and immunoblotting with phosphospecific antibodies that recognize RSK2 when phosphorylated at Ser 386 (B) or Thr 369 (C) or with antibody to the HA epitope tag (D). ent manner.
ERK and PDK1 Cooperate in Activation of Full-length RSK2 in Vitro-Taken together, our findings suggest that growth factor-induced activation of RSK2 involves the cooperative action of PDK1 and ERK. To determine whether the two kinases cooperate in activation of RSK, GST-RSK2 was incubated in vitro with PDK1 and active ERK2, either alone or together. As shown in Fig. 10A, basal RSK2 activity was low and was slightly increased by incubation with active ERK2, whereas incubation with PDK1 resulted in strong activation of RSK2. Addition of ERK2 and PDK1 together, increased RSK2 activity 2-fold compared with incubation with PDK1 alone. In control reactions without RSK2, essentially no S6 peptide phosphorylation was observed (Fig. 10A). Analysis of RSK2 phosphorylation by immunoblotting with anti-phospho-Ser 386 antibody, showed that only RSK2 incubated with ERK2 was positive (Fig.  10B). Thus ERK2 had phosphorylated and activated the CTK of RSK2 with subsequent autophosphorylation of Ser 386 .
These findings indicate that PDK1 is sufficient to induce substantial activation of RSK2, whereas ERK2 is not. However, ERK2 can cooperate with PDK1 in stimulation of RSK2 activity. DISCUSSION In the present study, we have shown that PDK1 phosphorylates Ser 227 in the activation loop of the NTK of RSK2, leading to substantial activation of the kinase in vitro and in vivo. Furthermore, our findings suggest that constitutively active PDK1 cooperates with ERK in activation of RSK following exposure of cells to growth factor. The role of PDK1 in activation of RSK is analogous to that described for p70 S6K (30,31), PKB (28,29), and PKC (32), in that activation loop phosphorylation by PDK1 cooperates with phosphorylation of a conserved region C-terminally to the kinase domain in full activation of the kinases.
The mechanism of activation of the NTK of RSK has been enigmatic. Recently, it was suggested that the NTK autophosphorylates as part of the activation mechanism based on the observation that the Asp 205 3 Ala mutant of RSK1 has an inactive NTK and shows decreased phosphorylation of Ser 221 , equivalent to Ser 227 in RSK2 (21). We have tested this hypothesis experimentally, and find that the isolated NTK of RSK2 does not autophosphorylate at Ser 227 in vivo, nor during prolonged incubation at a high concentration in vitro. Additionally, in full-length RSK2, the NTK did not autoactivate in vitro, even when ERK had activated the CTK with subsequent phosphorylation of the linker. Our findings are consistent with two previous studies that failed to detect autophosphorylation in vitro of the NTK of avian RSK synthesized in E. coli (20,23) and consistent with the fact that Ser 227 is situated in a motif that lacks important features of a RSK consensus phosphorylation site (19). All together, our results strongly suggest that the NTK requires a heterologous kinase to catalyze the phosphorylation in its activation loop and that PDK1 may be this kinase. PDK1 acted as an efficient Ser 227 kinase, capable of phosphorylating nearly all NTK molecules in cotransfected cells and increasing NTK activity 100-fold in vitro. Some previous observations are consistent with the idea that PDK1, or a kinase with similar characteristics, phosphorylates the activation loop serine in the NTK of RSK. First, RSK1 shows high basal phosphorylation of Ser 221 in COS1 cells under conditions where no RSK1 activity was detectable (21), suggesting that the serine is phosphorylated by a constitutively active kinase, like PDK1. A kinase activity with properties very similar to PDK1 has been partially purified from COS1 cells (38). Second, the putative consensus phosphorylation site of PDK1 (33) is FIG. 9. Activation and phosphorylation of wild-type RSK2 and mutant RSK2(S369A) or RSK2(S369A/T577A) by EGF or PDK1 in COS7 cells. Cells were transfected with plasmids expressing HA-RSK2, HA-RSK2(S369A), HA-RSK2(S369A/T577A), or myc-PDK1 or with empty plasmid, indicated by Ϫ. Forty-eight hours after transfection and following a 3-h serum starvation period, cells were exposed, or not, to 20 nM EGF for 15 min and lysed, whereafter RSK was precipitated with antibody to the HA epitope tag. A, the kinase activity of precipitated RSK was determined. Data are expressed as percentage of EGF-stimulated wild type RSK2 activity and are means Ϯ S.E. of three independent experiments performed in duplicate. The data of the following bars were compared by non-paired t test and were different (p Ͻ 0.01): 2 versus 5, 2 versus 8. In contrast, the data in the following bars were not statistically different (p Ͼ 0.23): 1 versus 4, 1 versus 7, 3 versus 6, 3 versus 9. After the kinase assay, aliquots of the precipitated RSKs were subjected to SDS-PAGE and immunoblotting with phosphospecific antibodies that recognize RSK2 when phosphorylated at Ser 386 (B) or with antibody to the HA epitope tag (C).
FIG. 10. ERK and PDK1 cooperate in activation of full-length RSK2 in vitro. GST-RSK2 was incubated for 90 min at 25°C with MgATP in the absence or presence of activated ERK2 or myc-PDK1 51-556 immobilized on agarose beads. After the incubation period, PDK1 was removed by centrifugation. A, the activity of GST-RSK2 was determined as described under "Experimental Procedures." Data are expressed as radioactivity (cpm) incorporated into S6 peptide and are mean values Ϯ range of duplicate determinations. The experiment was performed twice with similar results. B, after the kinase assay, an aliquot of the reactions containing GST-RSK2 was subjected to SDS-PAGE and immunoblotting with phosphospecific antibodies that recognize RSK2 when phosphorylated at Ser 386 . conserved in all RSK isotypes and across species, including Drosophila melanogaster, which also has a homologue of PDK1 (39). Furthermore, in one Coffin-Lowry patient, a threonine to isoleucine mutation in the putative PDK1 consensus motif was reported (7). In RSK1, we have mutated this threonine to a glutamic acid and observed a complete loss of kinase activity. 2 Moreover, we have recently been able to co-immunoprecipitate RSK2 and PDK1 from transiently transfected cells. 4 So far, however, we have not been able to inhibit RSK activity by overexpression of the kinase-deficient mutant PDK1 1-556 K111Q/ D205A/D223A. Perhaps the mutant with three mutations in the active site is structurally altered and unable to compete with endogenous PDK1.
Our findings with the isolated NTK of RSK demonstrate that it is a functional catalytic entity in the absence of most of the linker and the CTK and provide positive evidence that Ser 227 phosphorylation stimulates the activity of the NTK, switching it from very low to very high. The previous notion that the isolated NTK of RSK is inactive (18,20,23) is probably due to the lack of phosphorylation of the activation loop serine under the conditions used.
Coexpression of full-length RSK2 or RSK3 with PDK1 led to substantial activation of RSK, apparently by inducing the phosphorylation of Ser 227 and Ser 386 . The phosphorylation of Ser 227 is likely catalyzed by PDK1. In contrast, the identity of the kinase that phosphorylates RSK2 at Ser 386 in PDK1-transfected cells is not clear, but the CTK is a likely candidate. PDK1 itself is also a candidate, since it was recently shown that PDK1 can phosphorylate the corresponding site in PKB under certain conditions in vitro that include the presence of a peptide homologous to the sequence surrounding Ser 386 (40). In the present study, PDK1 did not phosphorylate Ser 386 in vitro (Fig.  10). Interestingly, RSK2(S227E) showed high basal phosphorylation of Ser 386 in vivo that was not enhanced by coexpression with PDK1. This strongly indicates that phosphorylation of Ser 227 in the NTK promotes the phosphorylation of Ser 386 in the linker, possibly by causing a structural change in RSK that disposes Ser 386 to phosphorylation by the CTK, PDK1, or another kinase. Finally, PDK1 appeared to activate RSK2 without involvement of ERK, since PDK1 neither stimulated ERK2 activity in COS7 cells nor induced phosphorylation of Ser 369 in RSK2 and since mutation of Ser 369 and Thr 577 did not affect activation of RSK2 by PDK1. Moreover, basal RSK2 activity was not affected by the S369A/T577A mutations, but abolished by the S227E mutation. This raises the possibility that basal RSK2 activity in resting cells may be attributed to PDK1 in an ERK-independent manner. In support of this model, basal RSK2 (and RSK3) activity in COS7 cells was not affected by PD98059, an inhibitor of mitogen-activated protein kinase/extracellular signal-regulated kinase kinase, the kinase that activates ERK (35).
EGF was found to activate RSK2 by a mechanism similar to the one described for phorbol ester-induced and ERK-mediated activation of RSK1 in COS1 cells (21). EGF stimulated the phosphorylation of Ser 369 and Ser 386 in the linker of RSK2, apparently via activation of ERK and the CTK, respectively, and both sites contributed to the activation of RSK2 by EGF, as evidenced by mutational analysis. RSK2, however, differs from RSK1, in that mutation of Ser 369 reduced EGF-stimulated RSK2 activity by only 50%, whereas the same mutation in RSK1 abolished its activation by phorbol ester (21) or by EGF. 2 The linker may therefore exert tighter inhibitory control of the NTK in RSK1 than in RSK2, in agreement with the fact that RSK1 has lower basal activity than RSK2. ERK alone was not sufficient to activate GST-RSK2 in vitro, in agreement with the previous finding that RSK1-3, immunopurified from EGFtreated cells and dephosphorylated by incubation with protein phosphatase 2A, cannot be reactivated by incubation with active ERK1 (35). Only in the presence of PDK1 was ERK able to stimulate the activity of RSK2, suggesting that the two ERKinduced phosphorylations in the linker enhance the activity of the NTK only in conjunction with its phosphorylation by PDK1. In p70 S6K , phosphorylation of the threonine corresponding to Ser 386 in RSK2 promotes phosphorylation of the activation loop during serum stimulation of cells and mutation of the threonine to a glutamic acid enables PDK1 to catalyze activation loop phosphorylation in a deletion mutant of p70 S6K in vitro (30,41). One role of Ser 386 phosphorylation in RSK may therefore be to facilitate phosphorylation of the activation loop of the NTK by PDK1. This might partially explain how ERK and PDK1 cooperate in activation of RSK2 in vitro and why Ser 221 phosphorylation is increased 2-3-fold during ERK-mediated activation of RSK1 in COS1 cells (21). However, our finding that RSK2(S227E) was stimulated 20-fold in response to EGF clearly shows that the EGF-induced phosphorylation(s) can stimulate the activity of the NTK even after it has been phosphorylated in the activation loop. It is possible that phosphorylation of the linker may stabilize the phosphorylated NTK in an active conformation. However, since the isolated NTK phosphorylated at Ser 227 displayed high activity despite missing most of the linker (Fig. 2), we speculate that in full-length RSK phosphorylation of the linker serves to release an inhibition of the NTK exerted by the unphosphorylated linker.
In conclusion, we suggest that the level of RSK2 activity in cells is determined by the balanced input from PDK1 and ERK, which act hierarchically, in that ERK cannot activate RSK2 in the absence of PDK1 activity, whereas the opposite is possible. In resting cells, constitutively active PDK1 accounts for basal RSK2 activity, the magnitude of which is a function of the level of available PDK1. Extracellular stimuli that activate ERK increase RSK2 activity above the basal level, because the ERKinduced phosphorylations in the linker cooperate with PDK1 in stimulation of the NTK.
RSK, p70 S6K , PKB, and PKC are activated by growth factors and function in partly distinct signaling pathways that regulate proliferation, protein synthesis, cell survival, and other key processes. It will be important to elucidate the role of PDK1 as a common control mechanism for these pathways. In principle, the level of PDK1 activity may dictate the responsiveness of cells to growth factor action. Furthermore, overexpression of PDK1 results in activation of RSK in a growth factor-and ERK-independent manner. Similarly, ectopically expressed PDK1 has been found to activate PKB in some cell types in the absence of exogenous growth factor (28,29). High expression of PDK1 may therefore act as an internal activator of some growth factor signaling pathways.