RSK-B, a Novel Ribosomal S6 Kinase Family Member, Is a CREB Kinase under Dominant Control of p38α Mitogen-activated Protein Kinase (p38αMAPK)*

A novel ribosomal S6 kinase (RSK) family member, RSK-B, was identified in a p38αMAPK-baited intracellular interaction screen. RSK-B presents two catalytic domains typical for the RSK family. The protein kinase C-like N-terminal and the calcium/calmodulin kinase-like C-terminal domains both contain conserved ATP-binding and activation consensus sequences. RSK-B is a p38αMAPK substrate, and activated by p38αMAPK and, more weakly, by ERK1. RSK-B phosphorylates the cAMP response element-binding protein (CREB) and c-Fos peptides. In intracellular assays, RSK-B drives cAMP response element- and AP1-dependent reporter expression. RSK-B locates to the cell nucleus and co-translocates p38αMAPK. In conclusion, RSK-B is a novel CREB kinase under dominant p38αMAPKcontrol, also phosphorylating additional substrates.

Eucaryotic cells exposed to intercellular communication or environmental signals respond with growth, adaptation, or death, all of which require restructuring of specific patterns of gene expression. The external signals are transduced to intracellular effectors by pathways that link cell surface and nuclear events and process transient input to short as well as more permanent response. Mitogen-activated protein kinases (MAPKs) 1 are ubiquitous and highly conserved elements of signal pathways under the control of growth factor, cytokine, and G protein-coupled receptors (reviewed in Refs. [1][2][3][4]. A first group of MAPKs, the extracellular signal-regulated kinase family (ERKs), initially was discovered to mediate hormone and growth factor effects on proliferation and differentiation (5). A 54-kDa kinase was reported that shared with ERKs activation through tyrosine and threonine phosphorylation and a requirement for proline C-terminal to the Ser/Thr phosphorylation site, but differed with regard to substrate specificity, the 54-kDa species being the more potent c-Jun kinase (6); molecular cloning revealed a second kinase family, the c-Jun N-terminal kinases (JNKs) or stress-activated protein kinases (7,8). A third pathway, architecturally similar to the ERK cascade, but primarily stimulated by cellular stress and cytokines was identified, with a homologous kinase called p38␣ MAPK designating the pathway (9 -13). All MAPK families, ERK, JNK, and p38 MAPK , are activated by concomitant phosphorylation of threonine and tyrosine residues in TEY (14,15), TPY (7,8), and TGY (9,10,13) sites, respectively, a few residues N-terminal to the conserved APE sequence in kinase subdomain VIII (16).
Early discoveries assigned functions in cell growth and stress to MAPKs, suggesting redundant roles in basic cell functions. However, ERK, JNK, and p38 MAPK families are encoded by multiple genes, with further diversification by alternative mRNA splicing into a growing number of isoforms. At least five isoforms of the p38 MAPK family are currently known. Some isoforms show a pronounced preference in tissue expression and selective interaction with upstream kinases and downstream substrates, pointing to highly specialized functions (17,18). The functional independence of the ERK, JNK, and p38 MAPK pathways is documented by their selective activation through distinct upstream kinases (2,19). Opposing effects of ERK as compared with JNK or p38 MAPK pathway activation have been reported, e.g. shifts in the dynamic balance between activation of JNK/p38 MAPK and concomitant inhibition of ERK may induce apoptosis in rat PC12 cells (20). Activation of p38 MAPK -dependent pathways has been linked to stress-induced apoptotic death in neutrophils and to excitotoxicity in rat cerebellar granule cells (21,22). Many studies place the p38 MAPK and JNK pathways in the context of "sounding the alarm" activity and of toxicity (for review, see Refs. [23][24][25]. However, even mild environmental stimuli further increase the high basal JNK activity in the brain (26), consistent with physiologic JNK functions.
The present study aimed to dissect the many different functions of the p38␣ MAPK MAPK pathway by identifying downstream p38␣ MAPK substrates. We performed an intracellular interaction trap screen in yeast, using a Gal4 DNA binding domain-p38␣ MAPK construct as bait, and identified a novel kinase, termed RSK-B, similar to the RSK family and containing two complete catalytic domains. RSK-B associates with and is activated by p38␣ MAPK . It is also activated, albeit more weakly, by ERK1, and may represent a convergence point between the p38␣ MAPK and ERK1 pathways. RSK-B controls CREB and AP-1 activity in luciferase reporter construct stud-ies, and CREB and c-Fos are RSK-B substrates. RSK-B is nuclear and localizes p38␣ MAPK to the nucleus.

EXPERIMENTAL PROCEDURES
Reagents and Materials-Standard reagents were from various sources as reported (45,46). Antibodies to RSK1, -2, and -3 and to epitope tags FLAG (antibody M2) and His 6 (anti-X-express) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA), International Biotechnology (New Haven, CT), and Invitrogen (Leek, The Netherlands), respectively. Transfectam procedure was from Biosepra. Luciferase assay kit was from Promega. PathDetect cis expression plasmids carrying luciferase genes under the control of CRE (x4) and AP-1 (x7) binding elements (pCRE-luc, pAP1-luc), and expression plasmids carrying constitutively active mutants of MEKK (380 -672; mMEKK), and MEK1 (S218/222E, ⌬32-51; mMEK1) under cytomegalovirus promoter; Yeast Two-hybrid Screen and Isolation of RSK-B cDNA-Intracellular interaction trap screen plasmids pAS2, pGAD424, pTD1, pVA3, and pCL1 were from CLONTECH. Plasmids pTD1 and pVA3 encoding GAL4-activation domain-SV40 large T-antigen and GAL4 binding domain-p53 fusion protein, respectively, were used as a positive control in the two-hybrid assays and to eliminate false positives. A XhoI-BamHI cDNA fragment encoding p38␣ MAPK ␣ was polymerase chain reactionamplified and cloned in-frame in pGBT9X, a modified version of the original pGBT9 vector which contains an unique XhoI site in the polylinker at position 873. A HpaI/BamHI fragment containing the Gal4 binding domain and the p38␣ MAPK region was inserted into pAS2.1 vector. A GAL4-AD/human placenta cDNA fusion library in pGAD10 vector was screened as recommended by CLONTECH. Positive clones were sequenced using a Li-Cor DNA 4000 automated sequencer. One class of cDNA identified in the intracellular p38␣ MAPK interaction trap assay (five independent clones) showed similarity with C-terminal sequences of the MAPKAPK family; it was called RSK-B and selected for further investigation. Using an EcoRI-SmaI fragment of the initial 1.3-kb RSK-B cDNA fragment (RSK-B⌬N2) from the two-hybrid screen, a human placenta gt11 library (CLONTECH) screen and 5Ј-RACE techniques yielded full-length 3.2-kb RSK-B cDNA according to standard protocols (47). Sequences were analyzed using the GCG Sequence Analysis software package (Madison, WI).
Expression Constructs and Mutagenesis-A mammalian expression vector for FLAG-tagged RSK-B was engineered by introducing RSK-B cDNA into modified pALTER plasmid (Promega). An NdeI linker (5Ј-TATGGTACCACCATGGACTACAAGGACGATGACGATAAGCA-3Ј) containing KpnI site, Kozak (48), and FLAG epitope sequences was inserted in the NdeI site 5Ј to RSK-B cDNA in pGAD424 vector. The KpnI-SalI fragment from this vector was ligated into the KpnI-SalI sites of pALTER to result in FLAG-tagged RSK-B in pALTER. The same strategy was used for RSK-B⌬N1 and RSK-B⌬C expression plasmids. RSK-B(K65A), RSK-B(K440A) and RSK-B(S360E) were generated by site-directed mutagenesis using the Altered Sites in vitro mutagenesis system (Promega). Wild-type MKK6 was polymerase chain reaction-amplified and cloned into pALTER, and the active MKK6 mutant mMKK6(S207G/T211G) was constructed using the the same strategy. p38␣ MAPK , ERK1 and JNK1␤1 were subcloned into pEBV-His expression vector (Invitrogen) as His 6 -epitope tag construct. Transient expression of all epitope-tagged proteins was tested in HEK 293 cells. Recombinant RSK-B and RSK-B⌬N1 were expressed in HEK 293 cells transfected with the respective FLAG-fusion vectors. FLAG M2 affinity gel was equilibrated in lysis buffer, washed once in lysis buffer with 300 mM NaCl added, once in lysis buffer, and incubated with transfected cell lysate (800 g of total protein, 25 l of affinity gel) for 1 h at 4°C. FLAG-tagged protein was eluted by adding 20 l of 10 mM Tris, pH 7.4, 150 mM NaCl, 400 g/ml FLAG peptide, 0.01% bovine serum albumin for 1 h at 30°C with agitation, and recovered from 14,000 rpm supernatant. By SDS-PAGE, FLAG-tagged proteins and bovine serum albumin were the two major components.
Mammalian Cell Culture, Transfection, and Luciferase Assays-HEK 293 (ATCC CRL 1573) and HeLa cells (ATCC CCL2) were cultured in humidified air with 5% CO 2 at 37°C. HEK 293 cells were maintained in minimal essential medium, 10% fetal calf serum, 100 units/ml penicillin, 10 g/ml streptomycin, pH 7.4. For recombinant protein expression in HEK 293 cells, 3 ϫ 10 6 cells in 25-cm 2 flasks were transfected with 3 g of DNA with the Transfectam procedure. HeLa cells for reporter construct assays were cultured in Dulbecco's modified Eagle's medium, 10% fetal calf serum, 100 units/ml penicillin, 10 g/ml streptomycin, pH 7.4. For recombinant protein expression in HeLa cells, 5 ϫ 10 5 cells/well in six-well plates were transfected with the specified amounts of plasmid DNA with the DEAE-dextran method. Luciferase assays were done as recommended by Promega. Luciferase activity was measured with a Packard top count luminometer.
Coimmunoprecipitation and Kinase Assays-Cell lysates were prepared as reported (45,46). Lysate aliquots (50 g of total protein) were precleared (three times) for 20 min at 4°C with 30 l of protein A-Sepharose beads and incubated with 1 g of either anti-FLAG or anti-p38␣ MAPK antibodies for 1 h at 4°C under constant agitation. Immune complexes were allowed to bind to 13 l of protein A-Sepharose beads for 1 h at 4°C and washed, and immunoblots were performed as reported (45,46). Precipitation kinase assays were carried out as reported without adding exogenous substrate (45,46). SDS-PAGE "in-gel" kinase assays were performed as described (45) using MBP, CREBtide ( 123 KRREILSRRPSYRK 136 ), and Fos peptide ( 357 AHRKGSSSNEP 367 ) as substrates. Extracts with 500 g of total protein were precipitated with 30 g of M2 as described above. Samples were separated in 10.5% SDS-PAGE containing 200 g/ml MBP, CREBtide, or Fos peptide copolymerized in the running gel.
Immunofluorescence-HEK 293 cells were plated on poly-D-lysinecoated coverslips. After 24 h, cells were transfected with various expression constructs as above. After 24 h of culture, the cells were serum-starved for 24 h in minimal essential medium containing 0.3% fetal calf serum. The coverslips were fixed with 4% formaldehyde in PBS for 5 min, placed in methanol for 1 min, followed by sequential washes with PBS and 1% horse serum in PBS. The coverslips were incubated with antibody (10 g/ml) for 1 h in the dark, and washed with PBS. Secondary labeled antibodies were applied for 1 h in the dark, followed by PBS washes. To visualize nuclei, the coverslips were placed in H 2 O for 2 min, treated with bisbenzimide (1.25 g/ml H 2 O) for 1 min, and washed once with H 2 O. Coverslips were dried, mounted with fluorescent mounting medium (Dako, Glostrup, Denmark) and viewed with a Leica confocal fluorescence microscope. Kinase reactions with M2 precipitates were performed in the absence of exogenous substrate to reveal RSK-B phosphorylation by co-precipitated p38␣ MAPK in presence or absence of 10 M FHPI or 10 M PD098059. Numbers below gels represent relative activation factors determined by PhosphorImager count. B, phosphorylation of affinitypurified recombinant RSK-B and RSK-B⌬N1 by p38␣ MAPK , ERK1/2, and JNK1 precipitates. HEK 293 cells were transfected with p38␣ MAPK His 6 , ERK1His 6 , or JNKHis 6 together with their respective upstream activating kinases, precipitated with specific antibodies p38(C20)-G, ERK1/2-CT, and JNK2(FL), and kinase assays were performed using RSK-B and RSK-B⌬N1 aliquots of about 1 g (as estimated from Gelcode staining) as substrates, as indicated. Incorporated radioactivity was quantified by PhosphorImager; relative activation factors are indicated below gel. 50 g/ml bovine serum albumin present in the assays was not labeled. C, Western blotting of pp38␣ MAPK , pERK1, and pJNK1 of enzyme equivalent amounts used in kinase assays of panel B, using antibodies specific for the respective activated phosphoenzymes. ent 3.2-kb cDNA with an open reading frame encoding a 772amino acid protein, termed RSK-B. The presumed translational start codon is preceded by a 5Ј-CCGCC sequence consistent with an optimal initiation context (48). The 733nucleotide 3Ј-nontranslated sequence contains an AATAAA polyadenylation signal 30 nucleotides in front of the poly(A) tail. The predicted RSK-B amino acid sequence revealed similarity with Caenorhabditis elegans putative protein kinase C54G4 (50% overall amino acid identity) and human RSK1 (49) (Fig. 1); two complete catalytic protein-Ser/Thr kinase domains with 53% and 43% identity with RSK1 in the N-and C-terminal domains, respectively, were predicted. The C-terminal domain is similar to MAPKAPK-2, and the N-terminal domain to p70S6 kinase. Both catalytic domains of RSK-B show 11 consensus kinase subdomains and fully conserved ATP-binding and activation domain sequences (16,50). The N-terminal domain has a protein kinase C-like and the C-terminal domain a calcium/calmodulin kinase-like sequence (16,50). A basic amino acid stretch near the C terminus represents a potential nuclear localization signal. Ser 236 in the N-terminal domain, Ser 343 in the sequence linking the two domains, and Thr 568 and Thr 687 in the C-terminal domain fit the minimal MAPK consensus phosphorylation site, ⌿X(S/T)P (⌿ ϭ proline or aliphatic) (30). Ser 343 is conserved in RSK-B, RSK1 (49), and p70 S6K (51); it has been shown to be essential in MAPK activation of the RSK1 N-terminal domain (52). Ser 360 and Ser 737 are conserved between RSK-B and RSK1, where they have been identified as a regulatory (auto-) phosphorylation sites induced by MAPK activity (52,53). Thr 568 and Thr 687 are conserved between RSK-B and MAPKAPK-2 (54). Northern blot analyses revealed that RSK-B is widely expressed in tissues and various cell lines (data not shown).

A p38␣ MAPK Bait Identifies RSK-B in Intracellular Interac
RSK-B Interacts with p38␣ MAPK and ERK-1 in Mammalian Cells-The physical association between FLAG-tagged RSK-B and MAPKs in mammalian cells was investigated by co-immunoprecipitation. HEK 293 cells were transfected with fulllength and truncated RSK-B (RSK-B⌬C, amino acids 1-334; RSK-B⌬N1, amino acids 335-772). Two days after transfection, cell lysate aliquots were precipitated with anti-FLAG antibody M2 (M2) (Fig. 2A). Parallel M2 precipitates were probed by Western analysis for co-precipitated p38␣ MAPK and ERK. Fulllength RSK-B and RSK-B⌬N1, but not RSK-B⌬C, co-precipitated p38␣ MAPK , confirming the C-terminal RSK-B sequence as p38␣ MAPK interaction domain. Converse precipitation with anti-p38␣ MAPK monoclonal antibody followed by Western blot analysis with M2 similarly revealed the RSK-B and p38␣ MAPK association (data not shown). Interestingly, a weaker, but significant, co-precipitation of ERK1 with RSK-B⌬N1 was also seen; a similar association of ERK1 with full-length RSK-B was detected in most but not all experiments. This suggested that RSK-B, in addition to p38␣ MAPK , is a substrate also of the ERK1 pathway. In contrast, no association between RSK-B and  5. A, RSK-B drives AP1-and CRE-dependent luciferase reporter expression. HeLa cells were co-transfected with either pAP1-luc or pCRE-luc together with RSK-B (or empty pALTER vector), and control, mMKK6, mMEK1, or mMEKK as indicated (1 g). Luciferase assays were performed with cell lysates after 2 days of culture. Top panels, relative luciferase activity, normalized to pALTER of control to reveal baseline luciferase activity in control, mMKK6, mMEK1, or mMEKK transfectants. Bottom panels, luciferase induction factor normalized to pALTER of each control, mMKK6, mMEK1, or mMEKK transfection, to reveal the RSK-B attributable luciferase activity. A representative experiment (n ϭ 4) is shown. B, RSK-B activity is necessary for luciferase response in either pCRE-luc or pAP1-luc co-transfectants. HeLa cells were co-transfected with mMKK6, wild type, or mutated RSK-B, and either pCRE-luc or pAP1-luc, as indicated (1 g). The cells were cultured for 2 days, and luciferase JNK1 was detected (Fig. 2B).
To investigate a possible stimulus-enhanced association of RSK-B with either p38␣ MAPK , ERK1/2, or JNK1, HEK 293 cells were co-transfected with (i) RSK-B, p38␣ MAPK His 6 , and mMKK6, and (ii) with RSK-B together with either mMEK1 or mMEKK. Cell lysates of these various transfectants were precipitated with M2, and immunoblot analyses were performed with anti-p38␣ MAPK , -ERK1/2, and -JNK1 antibodies (Fig. 2B). No enhancement of RSK-B association with p38␣ MAPK and ERK1 was seen, suggesting a stimulus-independent association of RSK-B with endogenous/transfected p38␣ MAPK and ERK1. Co-transfection of mMEKK did not lead to JNK1 coprecipitation, consistent with a lack of RSK-B and JNK1 interaction in the presence and absence JNK pathway activation.
RSK-B Is a p38␣ MAPK Substrate-RSK-B could be produced in recombinant expression only in low yield and furthermore, had a strong tendency for degradation and aggregation. This is in close parallel to the excellent study of Dalby et al. (52), who reported that recombinant RSK1 can only be expressed in very low yield, was highly degraded and could not be activated by ERK. To investigate whether RSK-B is a p38␣ MAPK substrate, we made therefore first use of the pronounced association of RSK-B and p38␣ MAPK and tested RSK-B phosphorylation by co-precipitated p38␣ MAPK in an ex vivo type experiment. RSK-B, RSK-B⌬N1, and RSK-B⌬C were precipitated by M2 from HEK 293 transfectants and assayed as substrates of coprecipitated p38␣ MAPK in standardized in vitro kinase assays, and 33 P incorporated in specific bands was PhosphorImager counted (Fig. 3A). Activation of the p38␣ MAPK pathway by mMKK6/p38␣ MAPK His 6 co-transfection resulted in a more than 5-fold increase in 33 P-incorporation in RSK-B when compared with RSK-B from unstimulated cells. The C-terminal catalytic domain, RSK-B⌬N1, showed a similar p38␣ MAPK stimulusenhanced 33 P-incorporation, whereas the N-terminal domain, RSK-B⌬C, did not show any 33 P-incorporation, consistent with a lack of interactivity with p38␣ MAPK . The specific inhibitor of p38␣/␤ MAPK , FHPI (12,55), added to the in vitro assay at 10 M reduced RSK-B and RSK-B⌬N1 phosphorylation practically to background level, whereas the specific MEK1 inhibitor PD098059 (56) (10 M) did not reduce 33 P incorporation into RSK-B and RSK-B⌬N1 (Fig. 3A). Thus, although RSK-B background phosphorylation by undetermined activities occurs, its p38␣ MAPK stimulus-dependent increase, which furthermore is blocked by the specific inhibitor FHPI, strongly supports the view that RSK-B is a p38␣ MAPK substrate.
These data were corroborated by results of standard kinase assays using aliquots of p38␣ MAPK , ERK1/2, and JNK, which were precipitated in parallel from HEK 293 cells co-transfected with p38␣ MAPK His 6/ mMKK6, ERK1His 6/ mMEK1, or JNKHis 6/ mMEKK, respectively, and affinity-purified RSK-B⌬N1 and RSK-B from small scale recombinant HEK 293 transient expression as substrates (Fig. 3, B and C). Phosphorylation of full-length RSK-B by p38␣ MAPK was enhanced 2.5-fold when compared with a parallel assay performed in the presence of 10 M FHPI (Fig. 3B). RSK-B⌬N1 phosphorylation was enhanced by activated p38␣ MAPK and ERK1 3.1-and 1.5-fold, respectively, when compared with a consistent background phosphorylation in the p38␣ MAPK His 6 ϩ FHPI and JNK assays (Fig.  3B). In independent studies, 3-7-fold p38␣ MAPK -dependent inductions of RSK-B⌬N1 phosphorylation were found, consistent with a phosphorylation level in such assays found by other authors (31). Amounts of p38␣ MAPK , ERK1, and JNK equivalent to those used in the kinase assays were probed for activation in Western blots using antibodies selectively reacting with the respective activated phosphorylated kinases (46) (Fig. 3C).
Control of RSK-B Activity by MAPK Pathways-To differentiate the roles of p38␣ MAPK , ERK and JNK in RSK-B activation, upstream kinases of the three pathways were co-transfected with RSK-B in HEK 293 cells, and MBP in-gel kinase assays of M2 precipitates of cell lysates were performed under standard conditions (Fig. 4A). First, investigating the p38 MAPK pathway, it was found that p38␣ MAPK in mMKK6/p38␣ MAPK / RSK-B co-transfectants enhanced RSK-B activity about 10-fold when compared with mMKK6/RSK-B transfectants (Fig. 4A,  lane 4 versus lane 3; see legend). In all gels, RSK-B bands of the expected apparent molecular mass of RSK-B monomers and higher molecular mass bands were seen (see Fig. 4A, Western, lane 11). The latter bands, which may result from aggregation occurring after cell lysis, from their gel location tentatively were designated RSK-B dimers. As expected, mMKK6 co-transfection also resulted in the activation of endogenous and transfected p38␣ MAPK (Fig. 4A, lanes 3 and 4, and Western, lane 12). The requirement for transfected exogenous p38␣ MAPK to achieve RSK-B activation is a peculiarity of the HEK 293 cells used (see below). Second, investigating the ERK pathway, ERK1His 6 in mMEK1/ERK1His 6 /RSK-B co-transfectants enhanced RSK-B activity about 1.5-2-fold (Fig. 4A, lane 7 versus lane 6; see legend), supporting a functional connection between ERK1 and RSK-B already suggested by co-precipitation (Fig.  2). Finally, no connectivity between RSK-B and the JNK1 pathway was detected in JNK1␤1His 6 , mMEKK, and JNK1␤1His 6 ϩmMEKK RSK-B co-transfectants, consistent with the negative co-precipitation data.
To confirm the role of the p38␣ MAPK and ERK1 pathways in RSK-B activation, HEK 293 cells co-transfected with RSK-B together with either p38␣ MAPK His 6 /mMKK6 or ERK1His 6 / mMEK1 were treated overnight with 10 M FHPI or 10 M PD098059, and RSK-B activity was investigated in MBP in-gel kinase assays of M2 precipitates of cell lysates (Fig. 4B). FHPI treatment practically abolished RSK-B activity in RSK-Bϩp38␣ MAPK His 6 ϩmMKK6 co-transfectants. Interestingly, PD098059 treatment also caused a partial RSK-B inhibition in RSK-BϩmMEK1ϩERK1His 6 co-transfectants, supporting a link between the ERK pathway and RSK-B.
In Vivo Substrates of RSK-B Point to a Role in Transcriptional Control-RSK-B in vivo substrates were investigated in reporter gene assays. First, HeLa cells were co-transfected with AP1-or CRE-dependent luciferase reporter constructs, and RSK-B with or without either mMKK6, mMEK1, or mMEKK. The cells were cultured for 2 days and lysed, and the lysates were assayed for luciferase activity. As expected, mMKK6, mMEK1, and mMEKK in the absence of RSK-B had various effects on luciferase activities (Fig. 5A, top panels). However, transfected RSK-B substantially enhanced AP1-and CRE-dependent luciferase responses, if the p38␣ MAPK pathway was activated by mMKK6 co-expression, whereas activation of the ERK and JNK pathways had less and no significant effects, respectively (Fig. 5A, bottom panels). None of the RSK-B mutants, RSK-B/K65A, RSK-B/K440A, or RSK-B/K65AϩK440A, was active (Fig. 5B). The RSK-B activity depended on endogenous p38␣ MAPK , since treatment of mMKK6ϩRSK-B HeLa cell co-transfectants with 10 M FHPI overnight or for 6 h practiassays were conducted with cell lysates. A representative experiment (n ϭ 3) is shown. C, luciferase response depends on p38␣ MAPK -stimulated RSK-B activity. HeLa cells co-transfected with RSK-B and mMKK6 together with either pCRE-luc or pAP1-luc were cultured for 2 days, treated with either 10 M FHPI or 10 M PD098059 for 6 h or overnight prior to cell lysis, and assayed for luciferase activity. A representative experiment (n ϭ 3) is shown. cally abolished luciferase expression, whereas similar treatment with PD098059 enhanced luciferase activity (Fig. 5C). gene controlled by GAL4 binding site repeats ([GAL4] 5 -luc) and a Gal4-CREB activation domain fusion protein (GAL4-CREB). The cells were cultured for 2 days, lysed, and assayed for luciferase activity. The luciferase response was substantially enhanced in RSK-BϩmMKK6 co-transfectants (Fig. 6A). Treatment of the cells with 10 M FHPI for 6 h or overnight completely abolished this response, whereas similar treatment with PD098059 had no significant effect. In parallel studies, HeLa cells were transfected with RSK-B and treated with TNF␣ for 6 h (Fig. 6A). TNF␣ markedly stimulated the luciferase response, which depended on p38␣ MAPK activity as demonstrated by sensitivity to treatment of cells with 10 M FHPI 30 min prior to TNF␣. RSK-B activity was essential for this luciferase response, since inactive RSK-B mutants were ineffective in TNF␣-stimulated and mMKK6 co-transfected cells (Fig. 6B). To demonstrate that CREB-dependent reporter activation can result from a direct interaction of RSK-B and CREB, RSK-B precipitates from HEK 293 cell transfectants were analyzed by CREBtide in-gel kinase assay (Fig. 6C), which revealed that RSK-B phosphorylated CREBtide and, in parallel studies, c-Fos peptide.

RSK-B Is a New CREB Kinase-To
Nuclear Location of RSK-B-The potential nuclear localization signal toward the RSK-B C terminus prompted an investigation of the subcellular location of RSK-B by M2 immunostaining (Fig. 7). RSK-B in HEK 293 transfectants was exclusively nuclear. Interestingly, double staining for RSK-B and co-transfected p38␣ MAPK His 6 showed that RSK-B and p38␣ MAPK His 6 co-localized to the nucleus, whereas cells transfected with p38␣ MAPK His 6 in the absence of RSK-B showed a broad and diffuse staining of nucleus and cytoplasm. Upon co-transfecting RSK-B⌬N1 and p38␣ MAPK His 6 , the C-terminal RSK-B domain and p38␣ MAPK His 6 also co-localized to the nucleus, whereas the N-terminal domain RSK-B⌬C remained in the cytoplasmic compartment. The co-transfection of p38␣ MAPK His 6 was necessary, because the staining of endogenous p38␣ MAPK was weak and did not allow to draw conclusions about subcellular distribution. A similar nuclear co-localization was observed between RSK-B and ERKHis 6 , but not with JNK1His 6 in parallel immunostainings of respective transfectants (data not shown). DISCUSSION Here, we present a novel RSK family CREB kinase, RSK-B, which is under dominant control of p38␣ MAPK . Other members of the RSK family, RSK1, -2, and -3, have been shown to phosphorylate CREB at serine 133, necessary for CREB-dependent transcriptional activation, in response to activation through the ERK pathway (37,44). CREB phosphorylation through the p38 MAPK pathway also was reported, and MAP-KAPK2 was proposed as effector molecule even though the activity of MAPKAPK2 to phosphorylate CREB in vitro was questioned (57,58). The activation of the p38 MAPK pathway was found causally related to apoptosis induced by trophic factor withdrawal in PC12 cells and fibroblasts (20, 59) More recently, it was found in nerve growth factor-stimulated PC12 cells using kinase inhibitors that ERK and p38 MAPK pathway signals activate CREB, and MAPKAPK2, or possibly MNK1, were discussed as potential effector molecules (37). While we cannot rule out that MAPKAPK2 or MNK1 indeed may have had such activity, the present data support the view that RSK-B, given its potent p38␣ MAPK -dependent CREB kinase activity, may have in fact been the enzyme responsible for CREB activation in the previous studies (37). The activation of RSK-B through ERK1 is weaker than through p38␣ MAPK , but this convergence of the p38␣ MAPK and ERK pathways onto RSK-B is intriguing and reminiscent of similar convergence shown for the ternary complex factor (60). The targeted cloning of RSK-B used p38␣ MAPK as probe, and all functional studies relate to p38␣ MAPK transfections; at present it is not known if and to what extent RSK-B interacts with other p38 MAPK isoforms.
RSK-B in addition to CREB is capable to activate other transcription factors as shown by its activity to drive AP1-dependent reporter genes. It cannot be ruled out that the AP-1 response resulted from indirect activation, e.g. the expression of the c-fos gene while being critically controlled by the serum response element also depends on CREB activation (61,62). Although RSK-B did not phosphorylate c-Jun (data not shown) and, in general, c-Jun is not considered a substrate of RSK family members (23), a c-Fos peptide was a RSK-B substrate, consistent with the view that RSK-B may signal through c-Fos phosphorylation. The notion that RSK-B functions in transcriptional control is also consistent with its apparently exclusive nuclear location at the sensitivity of immunostaining. The other RSK family members show a stimulus-induced translocation to the cell nucleus (43). ERK and p38 MAPK reside in the cytoplasm, and in cell body and dendrites of neuronal cells (42,63), and they co-translocate with these RSKs to the cell nucleus upon cell stimulation. This is consistent with the present finding that RSK-B, or RSK-B⌬N1, co-localized p38␣ MAPK to the nuclear compartment; further studies must show whether the exclusive nuclear location of RSK-B is physiologic or results from transfection and overexpression.
The distinct pathway connectivities of RSK1, RSK2, RSK3, and RSK-B indicate that RSKs exert selective functions. This conclusion is further supported by the fact that marked differences in tissue and cell-specific RSK1, -2, and -3 expression have been reported (49). Furthermore, spontaneous mutations in the RSK2 gene, resulting in RSK2 proteins with impaired kinase activity, have been found to be associated with severe psychomotor retardation, facial and digital dysmorphisms, and skeletal deformations in Coffin-Lowry syndrome (64); neither RSK1 and RSK3 nor, presumably, RSK-B are capable to fully complement RSK2 in these patients. The role of MAPKs in brain and in cognitive functions was the focus of many recent studies (65)(66)(67). A fascinating role of MAPKs is emerging from studies of neuronal activity-dependent modification of synaptic connections in the adult nervous system (68,69), and in long term facilitation in Aplysia (70). Although much evidence points to a role of p38 MAPK pathways in cellular growth, toxicities, and inflammation, which all may involve RSK-B, the involvement of MAPKs and CREB in cognitive functions, and the link of RSK-B and CREB, is intriguing (70 -73).