Characterization of the p90 Ribosomal S6 Kinase 2 Carboxyl-terminal Domain as a Protein Kinase*

The carboxyl-terminal domain (CTD) of the p90 ribosomal S6 kinases (RSKs) is an important regulatory domain in RSK and a model for kinase regulation of FXXFXF(Y) motifs in AGC kinases. Its properties had not been studied. We reconstituted activation of the CTD inEscherichia coli by co-expression with active ERK2 mitogen-activated protein kinase (MAPK). GST-RSK2-(aa373–740) was phosphorylated in the P-loop (Thr577) by MAPK, accompanied by increased phosphorylation on the hydrophobic motif site, Ser386. Activated GST-RSK2-(aa373–740) phosphorylates synthetic peptides based on Ser386. The peptide RRQLFRGFSFVAK, which was termed CTDtide, was phosphorylated with K m and V max values of ∼140 μm and ∼1 μmol/min/mg, respectively. Residues Leu at p −5 and Arg at p −3 are important for substrate recognition, but a hydrophobic residue at p +4 is not. RSK2 CTD is a much more selective peptide kinase than MAPK-activated protein kinase 2. CTDtide was used to probe regulation of hemagglutinin-tagged RSK proteins immunopurified from epidermal growth factor-stimulated BHK-21 cells. K100A but not K451A RSK2 phosphorylates CTDtide, indicating a requirement for the CTD. RSK2-(aa1–389) phosphorylates the S6 peptide, and this activity is inactivated by S386A mutation, but RSK2-(aa1–389) does not phosphorylate CTDtide. In contrast, RSK2-(aa373–740) containing only the CTD phosphorylates CTDtide robustly. Thus, CTDtide is phosphorylated by the CTD but not the NH2-terminal domain (NTD). Epidermal growth factor activates the CTD and NTD in parallel. Activity of the CTD for peptide phosphorylation correlates with Thr577 phosphorylation. CTDtide activity is constrained in full-length RSK2. Interestingly, mutation of the conserved lysine in the ATP-binding site of the NTD completely eliminates S6 kinase activity, but a similar mutation of the CTD does not completely ablate kinase activity for intramolecular phosphorylation of Ser386, even though it greatly reduces CTDtide activity. The standard lysine mutation used routinely to study kinase functions in vivo may be unsatisfactory when the substrate is intramolecular or in a tight complex.

The RSKs 1 (reviewed in Ref. 1) and the closely related nuclear enzymes MSK1 (2) and MSK2 (also known as Rsk-b (3)) are unusual among protein serine/threonine kinases because they possess two kinase domains. This feature was predicted from analysis of the first RSK cDNA after its isolation from Xenopus laevis (4). There are four human RSK genes encoding RSK1, RSK2, RSK3, and RSK4, 2 and all have the same conserved domain structure (1). The NH 2 -terminal catalytic domain (NTD) of the RSKs is most similar to p70 ribosomal S6 kinases followed by members of the protein kinase B and protein kinase C families. All of these enzymes, including the RSK NTDs, require PDK1 phosphorylation of a conserved serine (Ser 227 in RSK2) in the activation loop for activity (5,6). The NTD once activated is functional for phosphorylation of several physiologic substrates (1), perhaps best demonstrated by the ability of a constitutively active form of RSK, containing only the NTD, to induce metaphase arrest in cleaving Xenopus embryos (7).
The RSK CTD evaluated together with the adjacent COOH terminus is most similar in a standard BLAST search to CaMactivated protein kinases I and II. However, features in the activation loop and the COOH terminus reveal its functional relationship to the single domain MAPKAP kinases 3 such as the MNKs (8) and MAPKAP kinase 2 (9). The RSK and MSK CTDs and the single domain MAPKAP kinases have a threonine residue followed by proline in the kinase activation loop, nine and eight residues from the APE motif. None of the calmodulin-activated protein kinases have threonine followed by proline at this position, a necessary feature for MAPK phosphorylation. Similarly, all of the MAPKAP kinases have a MAPK docking motif in the COOH terminus (10) that we believe to be similar in structural disposition to the calmodulinbinding domain encompassing the ␣R2 helix in CaM kinase I (11). Similar to other MAPKAP kinases, the CTD is activated by MAPK via phosphorylation of a conserved T-P site in its activation loop (12), facilitated by the MAPK docking motif (10). The CTD phosphorylates one known substrate, a conserved Ser (Ser 386 in RSK2) (12) in the linker domain that joins the NTD and the CTD. The majority of this linker corresponds to the carboxyl terminus of p70 ribosomal S6 kinase by alignment and thus belongs to the NTD by inference. Within it are several regulatory sites of phosphorylation that correspond to phosphorylation sites in p70 ribosomal S6 kinase, protein kinase B, and protein kinase C enzymes.
Activation of the RSK NTD is dependent on phosphorylation of the regulatory sites in the linker domain (Ser 386 and putative MAPK sites Thr 365 and Ser 369 ). Ser 369 and Ser 386 may be the most important of these because S369A-and S386A-type mutants of full-length RSK are virtually inactive and unresponsive to agonist stimulation (6,12), whereas T365A-type mutants are nearly wild type (12). Ser 386 lies within a docking motif for PDK1, and Ser 386 phosphorylation is required for PDK1 binding and subsequent activation of the NTD (6). The role of Ser 369 phosphorylation is unknown. Ser 369 is not significantly phosphorylated in truncated RSK2 (amino acids 1-389), which has significant constitutive activity, suggesting that Ser 369 phosphorylation plays a role in the regulation of fulllength RSK. Some conformational states of inactive full-length RSK may sterically inhibit NTD activity independent of Ser 227 phosphorylation in the P-loop because a portion of RSK1 is phosphorylated basally at this site, yet is inactive (12,13). Phosphorylation of Ser 369 , and possibly Ser 386 as well, may contribute to relief of this inhibition in full-length RSK. The Ser 369 kinase(s) are U0126-inhibitable (14), pointing to ERKs1-2 or ERK5 as the upstream kinases for Ser 369 phosphorylation in vivo. Also consistent with this conclusion, phosphorylation of the equivalent serine in avian RSK1 is blocked by deletion of the MAPK docking motif (14). Thus, ERKs are the likely physiologic Ser 369 kinases.
Presently there are multiple models for RSK activation, possibly because of the existence of a multiplicity of activation mechanisms for this key signaling protein. One current model for RSK activation is vectorial (6). In this model, MAPK activates the CTD, which in turn phosphorylates Ser 386 in cis, creating a binding site for PDK1, which in turn phosphorylates Ser 227 , activating the NTD. However, some evidence suggests that NTD activation may not always proceed vectorially from CTD activation. A portion of Ser 222 in RSK1 is phosphorylated basally (12) as already mentioned, obviating the requirement for PDK1 phosphorylation. Furthermore, myristoylated avian RSK1 targeted to the plasma membrane is activated in serumstarved cells independent of evident ERK activation (15). RSK mutants rendered kinase-defective in the CTD (by mutation of the essential lysine) are still activated by growth factors but not as robustly as wild type (16,17).
In comparison with the NTD, much less is known about the CTD. In the current view, the NTD is assumed to be the only domain of the two capable of substrate phosphorylation in trans. The properties of the activated CTD as a protein kinase have never been studied. Our results demonstrate that the isolated, MAPK-activated RSK2 CTD is functional as a protein kinase toward peptide substrates. Furthermore, the CTD but not the NTD portion of the full-length protein selectively phosphorylates the best of these peptides (RRQLFRGFSFVAK), which is referred to herein as CTDtide. This peptide substrate allowed us to probe CTD regulation independently in full-length RSK2.

EXPERIMENTAL PROCEDURES
Materials-The plasmids pET-MEK1 R4F/His 6 ERK2 (18), pMT2-RSK2-(aa1-389) and its S386A mutant (6), and pGEX-5X-MK2-EE (19) were generously provided by Melanie Cobb (University of Texas South-western Medical Center, Dallas, TX), Steen Gammeltoft (Glostrup Hospital, Glostrup, Denmark), and Matthias Gaestel (Max Delbrü ck Centrum Molecular Medicine, Berlin, Germany), respectively. We obtained the murine RSK2 cDNA as pMT2 HA-RSK2 (20) from Christian Bjørbaek (Beth Israel Medical Center, Boston, MA) and have deposited its coding sequence as determined for pKH3-RSK2 (10) as GenBank TM accession number AY083469. The monoclonal antibody to RSK2 Thr(P) 577 (21) was kindly given to us by Paolo Sassone-Corsi (CNRS, Strasbourg, France). The synthetic peptides related to the NH 2 terminus of glycogen synthase were a generous gift of Sir Philip Cohen (University of Dundee, Dundee, UK), and the alcohol dehydrogenase repressor protein 1 (ADR1-g), synapsin, and glycogen synthase peptides were generous gifts from Anthony Means (Duke University, Durham, NC). Goat polyclonal anti-RSK2 (C-19) and the horseradish peroxidaselinked anti-goat antibodies were purchased from Santa Cruz Biotechnology; GammaBind Sepharose plus, glutathione-Sepharose 4B, horseradish peroxidase-linked anti-mouse, and anti-rabbit antibodies were from Amersham Biosciences; epidermal growth factor was from Collaborative Biomedical Products; microcystin LR and PD98059 were from Calbiochem; BHK-21 cells were from American Type Culture Collection; Syntide-2 was from the American Peptide Company; S6 and all RSK peptides were from the University of Virginia Biomolecular Research Facility; and the HA peptide was from the Howard Hughes Medical Institute Peptide Synthesis Facility (Duke University). All of the other reagents and products were from standard sources.
Construction of pAC-pET RSK2 CTD-A PCR-based strategy was used to engineer muRSK2-(aa373-740) (wild type and a K451A kinasedefective mutant) in frame into the BamHI and XhoI sites of pET41b (Novagen). The GST-RSK2 CTD and the kinase-defective mutant were transferred to pACYC184, a plasmid that has a p15A origin of replication, using a strategy suggested to us by Dr. Peter Sheffield (Center for Cell Signaling, University of Virginia Health Sciences Center, Charlottesville, VA). PCR was used to add BclI sites to a ϳ2.3-kb fragment amplified with Pfu polymerase (Promega) from pET41a-RSK2-(aa373-740). This fragment spans the T7 promoter for RNA polymerase, the GST tag, the pET41b multicloning site, RSK2-(aa373-740) with its stop codon, and the T7 terminator. The amplified fragment was cloned into the single BamHI site of pACYC184 (GenBank TM accession number X06403). The construct was verified by sequencing. The plasmids pAC-pET RSK2-(aa373-740) and pAC-pET RSK2-(aa373-740)(K451A) have a chloramphenicol resistance marker. The expressed proteins (exclusive of GST) have a 72-amino acid leader polypeptide (derived from codons within pET41b) that contains His 6 and S tags (Novagen) in addition to thrombin and enterokinase cleavage sites.
Phosphospecific Antibody to RSK2 Ser(P) 386 -The immunogen was [Cys]-Gly-Arg-Phe-Ser(P)-Phe-Val-Ala conjugated to keyhole limpet hemocyanin via the cysteine, and the antisera were produced in rabbits by Research Genetics (Huntsville, AL). The IgG fraction was purified from the production bleed by affinity chromatography on immobilized protein G.
Protein Production and Purification-The vectors encoding GST-and His 6 -tagged proteins were transformed into Escherichia coli BL21 cells with or without the bicistronic vector that encodes constitutively activated MEK1 and wild type ERK2 (18). These cells were grown at 37°C to an A 600 of ϳ0.3 and then induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside for 6 -8 h at room temperature. The proteins were purified from the cell lysates using either glutathione Sepharose or Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) essentially as directed by the manufacturer but with the addition of 2 M lithium chloride washes for the GST-tagged RSK proteins. The purified proteins were quantified using the Bio-Rad protein assay with bovine serum albumin as the standard. Active EE-MAPKAPK2 was made according to Engel et al. (19) using pGEX-5X-MK2-EE.
Kinase Activity and K m Determination-P81 paper assays were used to monitor kinase activity of the purified GST or His 6 -tagged proteins. The kinase assays were performed in 40-l reactions containing (final concentrations) 25 mM Hepes, pH 7.4, 2 mM dithiothreitol, 0.25 mg/ml bovine serum albumin, 10 mM MgCl 2 , a peptide substrate (as indicated in the figure legends), 50 M [␥-32 P]ATP (ϳ4000 cpm/pmol) at 30°C for 13 min or the indicated times in the figure legends. Phosphate incorporation into peptide substrate was determined using P81 phosphocellulose paper as described previously (10).
Cell Culture and Transfection-BHK-21 cells were grown in a humidified incubator at 37°C with 10% CO 2 in Dulbecco's modified Eagle's medium supplemented with 10% newborn calf serum. During transfections, the cells were also supplemented with 100 units/ml of penicillin and 100 g/ml of streptomycin. The cells were transfected with 20 g of DNA using a calcium phosphate profection system (Promega) on 150-mm dishes as described previously (22). Post-transfection (ϳ45 h) the cells were serum-starved for 3 h with or without 50 M PD 98059 and then treated with epidermal growth factor (100 ng/ml) for the times indicated. The cell lysates were made as described previously (22).
Immunoprecipitations and Kinase Activity of HA-tagged Proteins-HA-tagged proteins were immunoprecipitated from cleared lysates with 25 g of 12CA5 antibody and subsequently eluted as described previously (22) except that GammaBind-Sepharose was used, and the secondary antibody was omitted. Kinase activity of 3-l portions of eluted immunoprecipitations was determined by the P81 method in 40-l reactions containing (final concentrations) 25 mM Hepes, pH 7.4, 5 mM ␤-glycerophosphate, pH 7.4, 1.5 mM dithiothreitol, 6 M cAMP-dependent protein kinase inhibitor peptide, 15 mM MgCl 2 , 100 M [␥-32 P]ATP (ϳ4000 cpm/pmol), and either S6 (RRRLSSLRA) or CTDtide (RRQL-FRGFSFVAK) at 100 M. These reactions were initiated with ATP and incubated at 30°C for 13 min. Incorporation was corrected by subtracting the peptide phosphorylation observed from control immunoprecipitations performed in parallel from cells transfected with empty vector and treated equivalently. Specific activity of immunoprecipitation kinase assays (to HA Western signal) was determined from quantitative densitometry using ImageQuant software (Molecular Dynamics) essentially as described (22).
Data Analysis-Representative data are shown from experiments that were repeated at least twice (see legends).

RESULTS AND DISCUSSION
Design of a Plasmid for Reconstitution of Active RSK2 CTD in Bacteria-To address whether the RSK2 CTD is functional for phosphorylation of an exogenous substrate, we felt it necessary to test an active, recombinant protein from E. coli for several reasons. Kinase activity detectable from a CTD recovered from Sf9 or mammalian cells could be ascribed to a contaminant in the preparation. E. coli do not express protein serine/threonine kinases, and protein kinase cascades can be reconstituted in bacteria. In particular, a bicistronic plasmid for inducible expression of untagged active MEK1 together with His 6 -tagged ERK2 was created to produce active ERK2 for crystallization (18). The active MEK1 phosphorylates ERK2 in bacteria, producing doubly phosphorylated, fully active ERK2. Active ERK2 activates RSK2 (reviewed in Ref. 1). We hypothesized that this strategy could be used to produce active recombinant MAPKAP kinases, including the RSK2 CTD, provided the MAPKAP kinase would fold correctly in bacteria. That the RSK2 CTD might fold correctly was suggested by previous work (17), wherein GST-RSK1-(aa386 -752) was expressed in E. coli as a soluble protein and was active as revealed by its ability to autophosphorylate via an intramolecular (concentration-independent) mechanism.
For the above reasons, we engineered muRSK2-(aa373-740) to be expressed as a GST fusion protein (see "Experimental Procedures"). The plasmid pAC-pET RSK2 CTD ( Fig. 1) derived from pACYC184 has a p15A origin of replication and is compatible with the ColE1 replicator in pET-MEK1 R4F/His6 ERK2.
GST-RSK2 CTD Is Activated by ERK2 in E. coli-We used phosphospecific antibodies in Western blots to compare the levels of phosphorylation of Ser 386 and Thr 577 in wild type GST-RSK2 CTD and K451A RSK2 CTD obtained from two conditions: expression alone or expression together with active ERK2 (Fig. 2A). RSK2-(aa373-740) encompasses the Ser 386 site, which is known to be an intramolecular substrate in RSK2 for the CTD. The blots were first probed with anti-RSK as a loading control ( Fig. 2A, top panel) and then stripped and reprobed with anti-Ser(P) 386 ( Fig. 2A, middle panel). No signal was detected for the K451A mutant expressed by itself. Ser 386 was phosphorylated in wild type protein expressed in the absence of active ERK2. Expression with active ERK2 greatly increased Ser 386 phosphorylation for both the mutant and wild type proteins. The specificity of the phosphospecific Ser 386 antibody was verified using RSK2 CTD treated with and without calf intestine alkaline phosphatase (Fig. 2B). Because Ser 386 is phosphorylated by the CTD (12, 23) and is not a MEK or ERK2 substrate, RSK2 CTD enzymatic activity is up-regulated by ERK2 in bacteria.
The blot was stripped and reprobed a third time with a monoclonal antibody (21) to Thr(P) 577 , and the regulatory site was phosphorylated by MAPK in the activation loop of the RSK2 CTD ( Fig. 2A, bottom panel). Neither GST-RSK2 CTD nor the kinase-defective K451A mutant expressed alone was immunoreactive with this antibody. Co-expression with active ERK2 induced a large and easily detectable signal from both active and kinase-defective RSK2 CTD, establishing that active ERK2 produced from the bicistronic plasmid phosphorylates RSK2 CTD on Thr 577 in bacteria.
It was somewhat surprising that the K451A protein was detectably phosphorylated on Ser 386 because this serine is not a substrate for ERK2 or MEK1 R4F. Although it is remotely possible that some portion of the Ser 386 phosphorylation occurring in situ is due to the activating kinases, it is more likely that the K451A mutant retains kinase activity. In kinases that contain the conserved lysine in subdomain II (not all do), mutation of that residue decreases but does not completely eliminate kinase activity. The residual retained activity of these mutated kinases is variable and dependent upon the kinase and the amino acid substitution. For example, ERK2 K52R retains ϳ5% of the activity of wild type ERK2 (24) and still weakly autophosphorylates (25). Our results indicate that the K451A mutant of RSK2 CTD retains reduced activity that is also up-regulated by ERK2. This reaction should be facilitated because it can occur intramolecularly (17), if, as seems almost certain, the unidentified site in that study was Ser 386 .
The residual Ser 386 kinase activity of K451A RSK helps to rationalize the ability of full-length K451A-type RSK mutants to be partially activated by ERK (Refs. 17 and 22 and this study). The Ser 386 site must be phosphorylated or else mutated with phosphomimetic residues to bind and possibly to activate PDK1 (6). This is true for either full-length or truncated RSK. Although RSK2-(aa1-389) is phosphorylated by unidentified Ser 386 kinase(s), the phosphorylation of Ser 386 in full-length RSK is most likely catalyzed by the CTD. (Data showing that EGF can induce phosphorylation of Ser 386 in K451A RSK2 are presented later in the proper context (see Fig. 8D).) Thus, NTD activity in K451A-type RSK mutants can be rationalized in part by the residual intramolecular phosphorylation of these lysine mutants.
Bound ERK2 Is Removed by Lithium Chloride Washes- 1. Plasmid map of pAC-pET RSK2 CTD. A segment from pET41B-muRSK2 CTD from the T7 promoter through to and including the T7 terminator was moved into pACYC184 at the BamHI site. This construct encodes a GST leader polypeptide-RSK2-(aa373-740) of predicted mass 72.4 kDa. The leader polypeptide (derived from pET41B (Novagen)) contains GST, His 6 , and S tags as well as thrombin and enterokinase cleavage sites.
MAPK binds to the docking motif in the carboxyl terminus of RSK, and docking facilitates RSK activation in mammalian cells (10,26). ERK2 co-purified from lysates with the GSTtagged RSK when the two proteins were expressed together in cells (Fig. 3A). The docking motif of RSK contains paired arginines that make ionic bonds to glutamic residues in the common docking domain of ERK2 (27). We tested several options for removing ERK2 from GST-RSK2 CTD while the latter was still bound to glutathione beads. Of these washes, 2 M LiCl released the majority of bound ERK2 from both wild type (Fig.  3A, top panel) and K451A RSK2 CTD (data not shown). Final preparations of both proteins contained only a small, substoichiometric amount of ERK2 that was detectable by Western blotting (Fig. 3A, middle panel) but was not evident in Coomassie-stained gels.
After purification, K451A GST-RSK CTD reproducibly retained more ERK2 than GST-RSK2 CTD (Fig. 3B, bottom  panel). The relevant difference between these proteins is the amount of Ser 386 phosphorylation (Fig. 3B, middle panel). This serine is contained within an FXF motif (in RSKs, FSF) that is an ERK docking site in proteins such as ELK1, Lin-1, KSR, and phosphodiesterase (28 -30). This motif mediates interactions with ERK2 that are independent of the D domain MAPK docking site. Binding of ERK2 to FXF occurs at physiologically relevant affinities that are decreased up to 10-fold by disruption of the motif (30). Our data suggest that ERK2 is interacting with the Ser 386 site in GST-RSK2 CTD and that phosphorylation of Ser 386 decreases the affinity of the interaction. The carboxyl-terminal docking domain is the predominant mechanism of ERK-RSK interaction. However, it is plausible that the FSF site in RSKs is contributing to interaction with ERK2. Supporting this, ERK2 binds avidly to phenyl-Superose (31). In other proteins, this motif alone can direct phosphorylation of neighboring ERK sites (28).
RSK2 CTD Phosphorylates Ser 386 Peptide-Having estab-lished that WT-RSK2 CTD was able to phosphorylate Ser 386 in bacteria ( Fig. 2A), we tested RSK2 CTD for enzymatic activity toward an exogenous Ser 386 synthetic peptide (RRQLFRGFS-FVAI) (Fig. 4). WT-RSK2 CTD activated by ERK2 phosphorylated the peptide (solid squares), but WT-RSK2 CTD did not. The residual ERK2 in the preparation should not phosphorylate the peptide because the serine is not followed by a proline. This was confirmed by finding (Fig. 4, crosses) that ERK2, with an activity toward MBP that is 20 times greater than the activity of the wild type RSK toward the Ser 386 peptide, did not phosphorylate the Ser 386 peptide. The residual ERK2 in the CTD preparations can be detected using MBP as the phosphoacceptor (data not shown), but Ser 386 is not an ERK2 substrate. To our knowledge, this is the first proof that the isolated CTD can phosphorylate any substrate in trans.
trans-Phosphorylation Correlates with Thr 577 Phosphorylation-Wild type RSK2 CTD that was not ERK2 activated, and hence not Thr 577 phosphorylated, is able to phosphorylate Ser 386 in bacteria (see above) but does not phosphorylate the Ser 386 synthetic peptide (Fig. 4, open squares). This suggests that the wild type protein is active for intramolecular phosphorylation of Ser 386 but needs Thr 577 phosphorylation and potentially a conformational change to become active toward an exogenous substrate. This conclusion is strongly supported by the finding that K451A RSK2 CTD activated by ERK2 in bacteria phosphorylates the peptide (Fig. 4, inset). Our results show that phosphorylation of RSK2 CTD at the Thr 577 site enhanced intramolecular phosphorylation at Ser 386 as expected and conferred the ability to phosphorylate Ser 386 peptide, which is novel.
A prior indication that the RSK2 CTD might function in trans is contained within a report (23) that first identified phosphorylation of the Ser 386 site in Xenopus RSK. However, that part has been ignored because their experiments had alternative explanations, chiefly coming from use of full-length RSK containing the NTD.
Residues at p Ϫ5 and p Ϫ3 Are Critical for Ser 386 Peptide Phosphorylation-Specificity of protein kinases for physiologic substrates is dictated by several factors. Of these, positioning of the kinase with the substrate and primary sequence preference for the substrate are often paramount. The ability of a kinase to select substrates on the basis of primary sequence is due to pockets on the kinase surface that accommodate specific residues in the substrate, and the selection provided can be more or less stringent.
We tested the RSK2 CTD for stringency in substrate selection by comparing the phosphorylation of several mutant peptides to the parent Ser 386 peptide, each at 0.1 mM (Fig. 5,  bottom panel). The sequence similarity of RSK CTD to CaM kinase I and MAPKAP kinase 2 dictated inclusion of the peptide sequence surrounding Ser 386 from p Ϫ5 to p ϩ4. The Ser 386 motif (Fig. 5, top panel) is an exact match to the current consensus for CaM kinase I (BXRXX(S/T)XXXB) (32) and for MAPKAP kinase 2 (XXBXRXXSXX) (33), where B is a subset of hydrophic residues in each case. For MAPKAPK2, the optimal p Ϫ5 hydrophobic residue is bulky (Phe Ͼ Leu Ͼ Val Ͼ Ͼ Ala) and is leucine in the physiologic MAPKAPK2 substrate Hsp27 (LNRQLSS). For CaM kinase I, bulky hydrophobic residues are required at both the p Ϫ5 and p ϩ4 positions (32,34). The Ser 386 site of RSK includes bulky hydrophobic residues at p Ϫ5 and p ϩ4, which could be important. In addition, hydrophobic residues are important at other positions in the consensus sequence of other members of the CaM kinase superfamily to which the CTD is related (35,36). CaM kinase II and phosphorylase kinase both select Phe or another bulky hydrophobic residue strongly at p ϩ1 (36). The "hydrophobic motif " FXXFXF(Y) (37) for PDK1 docking contains phenylalanines that could be important for Ser 386 phosphorylation.
The L381K peptide was not a substrate, suggesting that the p Ϫ5 position is important for recognition by the CTD. Arg 383 at p Ϫ3 was required because the R383K and R383G peptides were compromised or not phosphorylated at all, respectively. These data show that p Ϫ5 and p Ϫ3 are critical for peptide phosphorylation by the CTD and suggest that like MAPKAPK2 the CTD prefers a hydrophobic residue and Arg at these positions. We made leucine substitutions for the phenylalanines. All of these mutants (F382L, F385L, and F387L) were phosphorylated by the RSK2 CTD nearly as well as the parent peptide. Some members of the CaM kinase superfamily (AMPactivated protein kinase (34) and CaM kinase I (32)) strongly prefer a hydrophobic residue at p ϩ4. The RSK CTD resembles CaM kinase II (36) in tolerating Lys replacement at p ϩ4. In the case of full-length RSK, intramolecular phosphorylation of Ser 386 a priori could be due to the combined effects of structural presentation of the site in the linker and to preference for a primary sequence. This will require detailed structure-function studies of the linker. However, the fact that residues in the linker (p Ϫ5 and p Ϫ3) are both critical for peptide phosphorylation and are conserved across RSKs suggests that recognition of Ser 386 for intramolecular phosphorylation will be dependent on the primary sequence in the linker.
As a final caveat, we do not know whether the peptide sequence is an optimal sequence. ERKs autophosphorylate intramolecularly on regulatory tyrosine and threonine sites (38,39) that do not conform to the (S/T)P consensus that is absolutely required for phosphorylation of exogenous substrates. We also do not imply existence of additional physiologic substrates for RSK2 CTD from these data, although our findings make the possibility more plausible.
RSK2 CTD Is a Selective Kinase-We tested ϳ20 peptides that have been used to characterize CaM kinase-related enzymes (Table I and data not shown). For comparison, the peptides in Table I were assayed with an equal amount of the active EE mutant of MAPKAPK2 (19) also produced in E. coli. The RSK linker peptides were MAPKAPK2 substrates, as expected. Peptides 1-4 are related to the NH 2 terminus of glycogen synthase and were used to characterize MAPKAPK2 specificity (33). Peptides 1 and 2 are excellent MAPKAPK2 substrates and were phosphorylated by RSK CTD but were poor substrates in comparison with the RSK linker peptides. Peptides that are excellent substrates for other CaM kinase-related enzymes were also tested and were not appreciably phosphorylated. Syntide 2 is a standard CaM kinase II substrate (35); ADR1-g and synapsin (4 -13) are benchmark substrates for CaM kinase I (40). The AMARA peptide is a substrate for AMP-activated protein kinase (34). None of these peptides were phosphorylated by RSK CTD. These results, with Fig. 5, show that the peptide specificity of RSK2 CTD is similar to MAPKAPK2, but it is distinct and much more selective.
Characterization of the I390K Peptide as CTDtide-The I390K peptide was the best substrate for purified, active RSK2 CTD and is referred to as CTDtide (Fig. 5 and Table I). The apparent K m for this peptide is ϳ140 M. This K m is in the upper range for most protein serine/threonine kinases. The specific activity toward the peptide was 1 mol/min/mg (Fig. 6), indicative of an efficient and properly folded enzyme. For reference, the catalytic subunit of protein kinase A, produced in E. coli, has a K m of ϳ40 M for Leu-Arg-Arg-Ala-Ser-Leu-Gly (Kemptide) and a specific activity of ϳ20 mol/min/mg for phosphate transfer (41). For CTDtide to be useful as a probe for CTD activity, the CTD should phosphorylate the peptide in the full-length protein. The NTD should not phosphorylate the peptide or at least should do so poorly in comparison with CTD.
Specificity of the CTD was tested using constructs created to inactivate the two kinase domains in full-length RSK2 (Fig. 7,  A and B). K100A RSK2 is kinase-defective for the NTD; K451A RSK2 is kinase-defective for the CTD. The HA-tagged proteins were immunopurified from EGF-treated cells for assay. The K100A mutant is completely kinase-dead for phosphorylation of S6 peptide but still phosphorylates CTDtide. The K451A mutant is kinase-dead for phosphorylation of CTDtide but not for phosphorylation of S6 peptide (25% of wild type). Note the difference in scales. S6 peptide is a much better substrate for the NTD in full-length RSK than CTDtide is for the CTD. To eliminate the possibility that the NTD could account for a significant proportion of CTDtide activity, we compared intact and a kinase-defective RSK2-(aa1-389) (Fig. 7C). A S386A mutant is as kinase-dead as a S227A mutant of the P-loop because it prevents PDK1 from phosphorylating the P-loop (6). Mutation of S386A caused nearly complete loss of S6 kinase activity but did not significantly affect the small amount of apparent CTDtide activity co-purified in this experiment. This residual incorporation is most likely due to contaminating kinases in the immunopurified RSK. Corrections to CTDtide activity were made from mock immunoprecipitates from empty vector controls, usually around 1-5% of CTDtide activity of wild type or K100A RSK2.
We suspect that the CTDtide activity is constrained in the full-length protein. This would be the expected result if the CTD has evolved to phosphorylate only the Ser 386 site in the linker. Evidence for constraint of CTDtide activity in full-length RSK is FIG. 6. Kinetic analysis of CTDtide phosphorylation by GST-RSK2 CTD. The peptide was used as substrate (25-143 mM) in P81 kinase assays as described under "Experimental Procedures." Active GST-RSK2 CTD was used at 5 mg/ml, and the reactions were run for 5 min. The apparent K m value for the peptide is 140 M, and the specific activity is ϳ1 mol/min/mg (n ϭ 3 Ϯ S.D.).
a Activity was measured using the p81 paper assay (see "Experimental Procedures"). The concentration of enzyme and substrate were 6.25 g/ml and 100 M, respectively. The reactions were run in duplicate for 13 min; 100% incorporation was 6.7 or 7 pmol of phosphate for RSK2 CTD and EE-MAPKAPK 2, respectively.
shown in Fig. 7D, wherein kinase activities of the isolated NTD and CTD kinases are compared with the full-length wild type kinases. The data presented in Fig. 7 are normalized for total RSK protein from the Western signal for HA because we observed differences in level of expression of the separated domains. The CTD alone (RSK2-(aa373-740)) is consistently underexpressed, and the NTD alone (RSK2-(aa1-389)) is overexpressed compared with the full-length proteins (data not shown). Normalization was done as carefully as possible, making preliminary runs to find dilutions that were similar and would be in the linear range of the film after nonextended exposures. The separated CTD domain consistently had higher specific activity when compared with the full-length proteins. In additional experiments, portions of the assayed RSK proteins were also analyzed for reactivity to anti-Thr(P) 577 , and specific activities relative to Thr 577 phosphorylation were calculated (Table II). The enhanced specific activity of the isolated CTD is partially explained by increased stoichiometry of Thr 577 phosphorylation. However, the specific activity of CTD alone, normalizing to Thr 577 phosphorylation, is still 6-fold higher than that of either full-length protein (wild type or K100A). This suggests differences in conformation between the isolated CTD and the CTD in the full-length protein that alter the specific activity. Structures to determine the accessibility of the active site in the isolated CTD versus the CTD in full-length RSK would be of interest.
The I390K peptide can be considered CTDtide because the CTD, but not the NTD, portion of the RSK protein phosphorylates it. CTDtide can be used to assay CTD activity in relation to NTD activity in RSK provided that RSK is purified to remove other kinases, such as MAPKAPK2, that would also phosphorylate it and that proper controls are performed. The specific elution of bound RSK with HA peptide in our experiments may also have helped to reduce the amount of contaminating kinases that may stick to beads but are not elutable with HA peptide.
CTD Activation Is Rapid and Parallels NTD Activation-Previously it has not been possible to determine whether the NTD and CTD kinase activities are congruent (i.e. on together/ off together) or dissociated (i.e. CTD-activated and inactivated, toward S6 (hatched bars), in arbitrary units normalized for RSK protein. In C and D: left ordinate, S6; right ordinate, CTDtide. The data are the averages of duplicates (ranges indicated by error bars) but are representative of six experiments. The CTD but not the NTD is required for phosphorylation of CTDtide. C, S6 versus CTDtide, constitutively active wild type RSK2-(aa1-389), and S386A mutant. S386A mutation reduces S6 but not CTDtide kinase activity; the NTD is selective for S6. (Note difference in scale for apparent CTDtide activity here versus in D.) D, S6 versus CTDtide activity, wild type, and K451A HA-RSK2-(aa373-740). The isolated CTD has 6-fold higher specific activity than CTD in full-length RSK2. HA-tagged RSK2 proteins (indicated) were immunopurified from BHK-21 cells stimulated with epidermal growth factor for 5 min (see "Experimental Procedures"). The procedure included specific elution from the immunoprecipitates with HA peptide. Eluted proteins were assayed with S6 peptide (hatched bars) and CTDtide (RRQLFRGFSFVAK) (solid bars). Specific activity a Activity of HA-RSK proteins, eluted from immunoprecipitates of cells treated for 15 min with EGF (3 l), was measured in duplicate using the p81 paper assay with CTDtide as substrate (see "Experimental Procedures"). The eluted proteins were then analyzed for protein content using western blot analysis (anti-HA), and the results are presented as specific CTDtide activity per the HA signal in arbitrary units.
b Eluted immunoprecipitated proteins from above were then analyzed for the amount of Thr 577 phosphorylation (anti-Thr(P) 577 , and the results are presented as ratio of Thr(P) 577 to HA signal, expressed as a percentage.
c The activity results from the p81 assays above presented as specific CTDtide activity per Thr 577 phosphorylation in arbitrary units.
whereas NTD remains active). We assayed the NTD and CTD activities of HA-RSK2 purified from cells treated with EGF for different times (Fig. 8). We found that activation of the CTD closely paralleled NTD activation for this agonist (Fig. 8A). Maximal activation occurred between 5 and 10 min. Activation was transitory for both domains. EGF also caused a time-dependent activation of CTD activity of K100A RSK2, which lacks S6 kinase activity (Fig. 8B). K100A HA-RSK2 was completely inactive toward S6 peptide, consistent with the demonstrated efficacy of K100A HA-RSK2 as a dominant negative (42). Finally, EGF failed at all times to increase phosphorylation of CTDtide by the full length K451A mutant (Fig. 8C). As noted above, growth factors cause a reduced activation of S6 kinase activity in K451A-type mutants. The K451A mutant is detectably phosphorylated at Ser 386 in unstimulated cells and EGF causes a weak but detectable increase in this phosphorylation (Fig. 8D, compare 0 min to 5 min). This is consistent with data discussed above that were obtained in bacteria. The kinetics of Ser 386 phosphorylation in the wild type, included as a control, show EGF activation at the earliest time point, 2.5 min.
These kinetics are compatible with the vectorial model for RSK activation because intramolecular phosphorylation of Ser 386 would be extremely rapid. Furthermore, PDK1 binding and phosphorylation of Ser 227 is potentially too fast to produce detectably slower NTD activation relative to the CTD. The PDK1 steps may even be preempted. There is evidence (13) for a pool of inactive RSK that is already phosphorylated on the NTD activation loop (12). Activation of this pool of RSK would only require phosphorylation of Ser 386 and/or linker MAPK sites and a conformational change. Because this pool is inactive in unstimulated cells despite phosphorylation of the NTD activation loop, RSK activation must also include relief of intrasteric inhibition (43).
The Future for Other MAPKAP Kinases-Reconstitution of activation of other MAPKAP kinases (MNKs, etc.) along similar lines should be feasible. Although it is possible to generate active mutants that circumvent phosphorylation requirements by truncation or acidic replacements, structural studies of enzyme regulation are most informative with authentic phosphorylated enzyme. This will require generation of additional bicistronic vectors to produce the specific activated MAPKs. Reconstitution of protein cascades in bacteria may also prove to be surprisingly specific. MAPKAP kinase 2 was discovered biochemically as an enzyme that was activated in vitro by ERK (44) but was later shown to be activated by p38␣,␤ MAPK specifically in mammalian cells. Consistent with this, we found that co-expression of MAPKAP kinase 2 with active ERK2 in bacteria caused only modest activation (data not shown), which we found surprising. EE-MAPKAP kinase 2 produced in bacteria, in contrast, was much more active (19) (data not shown).