Identification of an Extracellular Signal-regulated Kinase (ERK) Docking Site in Ribosomal S6 Kinase, a Sequence Critical for Activation by ERK in Vivo *

Glutathione S-transferase (GST)-fusion proteins containing the carboxyl-terminal tails of three p90 ribosomal S6 kinase (RSK) isozymes (RSK1, RSK2, and RSK3) interacted with extracellular signal-regulated kinase (ERK) but not c-Jun-NH2-kinase (JNK) or p38 mitogen-activated protein kinase (MAPK). Within the carboxyl-terminal residues of the RSK isozymes is a region of high conservation corresponding to residues722LAQRRVRKLPSTTL735 in RSK1. Truncation of the carboxyl-terminal 9 residues,727VRKLPSTTL735, completely eliminated the interaction of the GST-RSK1 fusion protein with purified recombinant ERK2, whereas the truncation of residues731PSTTL735 had no effect on the interaction with purified ERK2. ERK1 and ERK2 co-immunoprecipitated with hemagglutinin-tagged wild type RSK2 (HA-RSK2) in BHK cell cytosol. However, ERK did not co-immunoprecipitate with HA-RSK2(1–729), a mutant missing the carboxyl-terminal 11 amino acids, similar to the minimal truncation that eliminated in vitro interaction of ERK with the GST-RSK1 fusion protein. Kinase activity of HA-RSK2 increased 6-fold in response to insulin. HA-RSK2(1–729) had a similar basal kinase activity to that of HA-RSK2 but was not affected by insulin treatment. Immunoprecipitated HA-RSK2 and HA-RSK2(1–729) could be activated to the same extent in vitro by active ERK2, demonstrating that HA-RSK2(1–729) was properly folded. These data suggest that the conserved region of the RSK isozymes (722LAQRRVRKL730 of RSK1) provides for a specific ERK docking site approximately 150 amino acids carboxyl-terminal to the nearest identified ERK phosphorylation site (Thr573). Complex formation between RSK and ERK is essential for the activation of RSK by ERK in vivo. Comparison of the docking site of RSK with the carboxyl-terminal tails of other MAPK-activated kinases reveals putative docking sites within each of these MAPK-targeted kinases. The number and placement of lysine and arginine residues within the conserved region correlate with specificity for activation by ERK and p38 MAPKs in vivo.

Mitogen-activated protein kinases (MAPKs) 1 transduce sig-nals from the cell surface to the nucleus, altering the activity and subcellular localization of transcription factors. ERK, JNK, and p38 MAPKs lie in distinct signaling pathways that are activated by distinct stimuli. Whereas the minimal consensus phosphorylation sequence of these proline-directed kinases would suggest promiscuous phosphorylation of many proteins, the kinases play an integral role in the cellular growth machinery; therefore, substrate specificity must be tightly regulated. It is becoming clear that the substrate specificity of MAPKs with regard to transcription factors involves high affinity binding of MAPK to sequences within the substrate that are distinct from the consensus phosphorylation sequence (1,2). Such an interaction has been described for JNK and the transcription factors c-Jun and activating transcription factor (ATF-2) (3)(4)(5). Recently, a sequence within Elk-1 was shown to contain overlapping but distinct interaction sites for ERK and JNK (6). Specific targeting interactions between MAPKs and substrates may not be limited to transcription factors. One possible ERK substrate for which targeting interactions might occur is p90 ribosomal S6 protein kinase (RSK). RSK is phosphorylated and activated by ERK in vitro (7), and inhibition of the ERK pathway with the mitogen-activated protein/ERK kinase-specific inhibitor PD98059 prevents in vivo activation of RSK (8,9).
RSK is unusual in that it contains two distinct kinase catalytic domains within a single polypeptide chain (10). In vivo ERK phosphorylation sites within RSK have been identified (11,12) (Fig. 1A). Two of these sites are essential for activation of RSK: 1) Ser 363 in the linker between the two kinase domains, and 2) Thr 573 in the activation loop of the carboxyl-terminal kinase domain (12). 2 Physiological substrates for the three mammalian isozymes of RSK (RSK1, RSK2, and RSK3), which are encoded by separate genes (13), are currently under investigation. RSK has been shown to phosphorylate the cAMP response element-binding protein (14,15), c-fos (16,17), and the estrogen receptor (18), suggesting that RSK as well as ERK plays a role in transcriptional regulation.
Whereas studies describing the existence of an interaction between ERK and RSK have been reported (19 -21), analyses of the site(s) of interaction and the in vivo significance of the interaction are essential. Examination of the RSK isozymes for the ability to interact with ERK revealed that ERK co-immunoprecipitated with RSK2 and RSK3, but not RSK1 (21). The site of interaction between RSK and ERK was localized to the carboxyl-terminal 44 amino acids of RSK3 (21). MAPK-interacting kinase (Mnk) was recently identified as a serine/threonine kinase with sequence similarity to RSK that is phosphorylated and activated by MAPK (22,23). Mnk1 co-purified with (22) and was phosphorylated by both ERK and p38 MAPK (22,23), whereas Mnk2 interacted specifically with ERK (22). Truncation of the carboxyl-terminal 90 amino acids (residues 334 -424) of Mnk1 eliminated phosphorylation and activation by ERK (23). Thus, the site of interaction between MAPK and the substrate kinases RSK and Mnk appears to lie within the carboxyl-terminal regions of the substrate.
In the present studies, interaction between MAPK and three isozymes of RSK was examined. An ERK docking site critical to this interaction was identified, and the role of the ERK/RSK complex was assessed in vivo.
Vector Construction-Expression vectors were generously provided by the following: pMT2.HA-RSK1 (rat), pMT2.HA-RSK2 (mouse), and pMT2.HA-RSK3 (human), Dr. Christian Bjørbaek (Harvard Medical School, Boston, MA); pEThis.MEK1/ERK2 (active ERK2) and pGexKG.-  (24). pGEX2T bacterial expression constructs encoding the carboxyl-terminal tail residues of RSK1, RSK2, and RSK3 were created by polymerase chain reaction amplification of the 3Ј end of each RSK. pK3H.RSK2 was generated by polymerase chain reaction amplification of pMT2.HA-RSK2 (mouse) cDNA and subcloning into a unique BamHI site in pK3H. PK3H.RSK2  , which encodes a RSK2 (mouse) mutant in which the carboxyl-terminal 11 amino acids are deleted, was created by ligating annealed oligonucleotides (encoding a Bpu1102I site and a stop codon) to pK3H.RSK2 linearized with Bpu1102I. The sequences were verified by automated sequencing. Oligonucleotides used are available on request. Purification of GST-fusion proteins was performed according to a protocol supplied for the GST Gene Fusion System (Pharmacia Biotech).
Cell Culture and Transfection-BHK-21 (C-13) cells were grown in a 37°C humidified atmosphere containing 10% CO 2 in Dulbecco's modified Eagle's medium supplemented with 5% fetal calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin (DMEM/5% FCS/PS). The cells were plated (20 -40% confluence) on 150-mm dishes 16 -20 h before transfection. The cells were transfected with 10 g of CsClbanded DNA (pK3H.RSK, pK3H.RSK2  , or pK3H) per 150-mm dish as described for transfection of 100-mm dishes in the Calcium Phosphate ProFection® System manual (Promega). At 48 -72 h after transfection, the cells were serum-starved for 1-1.5 h in DMEM/PS, followed by treatment with 94 nM insulin (or vehicle) for 12 min at 37°C. Three 150-mm plates/condition were washed six times with cold phosphate-buffered saline (PBS), and the cells were scraped into lysis Buffer A (10 mM Tris base, pH 7.5, 150 mM NaCl 2 , 1% Triton X-100, 0.5% Nonidet P-40, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 10 g/ml each of leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, 200 M sodium orthovanadate, and 1 M microcystin-LR) for immune complex kinase assays or lysis Buffer B (50 mM Tris base, pH 8, 150 mM NaCl 2 , 1% Nonidet P-40, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 10 g/ml each of leupeptin and aprotinin, 1 mM phenylmethylsulfonyl fluoride, 200 M sodium orthovanadate, and 1 M microcystin-LR) for co-immunoprecipitation experiments. The cells were lysed by incubation on ice for 20 min, and the supernatant was clarified by centrifugation at 12,000 ϫ g for 10 min for immune complex kinase assays or at FIG. 1. Schematic of RSK1 and sequence alignment of the GST-RSK fusion proteins with mouse Mnk1 and Mnk2. A, schematic of RSK1 depicting the two kinase domains, the phosphorylation sites 2 identified by Dalby et al. (12), and the carboxyl-terminal 20 amino acids. B, alignment of the GST-RSK fusion proteins with mouse Mnk. Asterisks ‫)ء(‬ mark identical residues between the RSK and Mnk isozymes. A conserved region between RSK isozymes is highlighted.
Preparation of COS Cell Cytosol and Purified ERK-Confluent COS-1 cells growing in 10% fetal calf serum were treated in the absence or presence of 2 M PDB for 15 min at 37°C. The plates were washed with cold PBS, and cells were scraped into D-PBS (2.7 mM KCl, 1.1 mM KH 2 PO 4 , 138 mM NaCl, and 8.1 mM Na 2 HPO 4 , pH 7.4). Cells were pelleted, and an equal volume of lysis Buffer C (50 mM HEPES, pH 7.4, 1% Nonidet P-40, 0.5% Triton X-100, 10% glycerol, 150 mM NaCl, 1 mM Na 3 VO 4 , 2 M microcystin LR, 2 mM phenylmethylsulfonyl fluoride, and 1 Complete TM protease inhibitor mixture tablet (Boehringer Mannheim)/50 ml) was added to the cell pellet. The cell pellet was resuspended, and cells were lysed by sonication. The lysate was centrifuged at 12,000 ϫ g for 10 min in a microcentrifuge. Cytosol was frozen in liquid nitrogen and stored at Ϫ70°C. Kinase-defective ERK2 (K52R ERK2) and active ERK2 were prepared as described previously (25,26).
Incubation of GST-RSK Carboxyl-Terminal Proteins with Cytosol or Purified ERK-Glutathione-Sepharose 4B beads (25 l bed volume) were copiously washed with PBS. The washed beads were rotated with 30 g of GST or GST-fusion protein in 25 mM HEPES, pH 7.4, 1 mM EDTA, and 1 mM dithiothreitol for 1 h at 4°C. The washed GST-bound beads were incubated with COS-1 cell cytosol (200 l; 17 mg/ml total protein) or 100 nM purified ERK2 in lysis Buffer C for 1.5 h at 4°C with gentle mixing. The beads were pelleted by centrifugation, and the supernatant was removed and saved. The beads were washed three times with 750 l of cold PBS. SDS-sample buffer was added to the washed beads and supernatant, and the samples were boiled for 5 min and processed for Western analysis. Unless otherwise indicated, monoclonal antibody (M12320) was used for immunoblotting ERK.
Immunoprecipitation and Immune Complex Kinase Assay-BHK cell cytosol (1 ml) from cells expressing HA-RSK2 or HA-RSK2 (1-729) treated in the absence or presence of 94 nM insulin for 12 min was pre-cleared by incubation for 1 h with immobilized protein A/G-agarose beads (20 l bed volume). Beads were pelleted by centrifugation, and the supernatant was removed. For co-immunoprecipitation, the supernatant was incubated with 30 g of monoclonal anti-HA antibody (12CA5) at 4°C for 1 h, followed by incubation of the supernatant with anti-mouse IgG:horseradish peroxidase-linked for 30 min on ice. Immobilized pro-

FIG. 2. Interaction of GST-RSK carboxyl termini with MAPK in cytosol.
GST-RSK fusion protein or GST bound to glutathione beads was incubated with COS-1 cell cytosol as described under "Experimental Procedures." Cells used to determine the presence of active ERK were treated with 2 M PDB. Supernatant and washed beads were subjected to Western analysis. Although the anti-ERK antibody (M12320) recognizes both ERK1 and ERK2, ERK2 is the predominant isozyme in COS cells, and immunodetection of ERK1 was not always observed in the cytosol from untreated cells. tein A/G-agarose beads (20 l bed volume) were incubated with the supernatant for 30 min on a Nutator rocker. The beads were pelleted by centrifugation, and the supernatant was removed and saved. The beads were washed once with 500 l of lysis Buffer B and twice with 500 l of cold PBS. SDS-sample buffer was added to the beads and the supernatant, and the samples were boiled for 5 min and processed for Western analysis with polyclonal anti-HA antibody and polyclonal anti-ERK antibody.
For the immune complex kinase assay, 1 g of polyclonal anti-HA antibody was added to 500 l of pre-cleared BHK cell cytosol (Fig. 5A) or 10 g of polyclonal anti-HA antibody were added to 1 ml of precleared BHK cell cytosol (Fig. 5B) and incubated for 1 h. Immobilized protein A/G-agarose beads (20 l bed volume) were incubated with the supernatant for 12 h. The beads were pelleted by centrifugation, and the supernatant was removed and saved. The beads were washed once with 500 l of lysis Buffer A, once with 500 l of cold PBS, and twice with 500 l of kinase wash buffer (75 mM ␤-glycerolphosphate, pH 7.   (1-729ϩ). B, HA-RSK2 and HA-RSK2  were immunoprecipitated from the cytosol of serum-deprived BHK cells. Immune complexes were incubated for 15 min at 30°C in the absence or presence of 1.5 M active ERK2. After incubation, the kinase activity of the immunoprecipitates was measured. Activity is presented as mean Ϯ S.D. (n ϭ 3). The kinase activity of immunoprecipitates from cells transfected with the empty vector and incubated in the absence or presence of active ERK2 has been subtracted. The insert demonstrates the equivalent quantity of immunoprecipitated enzyme as well as the SDS-PAGE mobility shift induced by the incubation of HA-RSK2 (WTϩ) and HA-RSK2   (1-729ϩ) in the presence of active ERK2 compared with the incubation of HA-RSK2 and HA-RSK2  in the absence of active ERK2 (WTϪ and 1-729Ϫ, respectively).

Interaction of GST-RSK Carboxyl Termini with MAPK in
Cytosol-GST-fusion proteins of the carboxyl termini of three RSK isozymes were constructed to examine the interaction between RSK and MAPK. Bacterially expressed GST-RSK1 (672-735) , GST-RSK2 (676 -740) , and GST-RSK3 (669 -733) contain the amino acid sequence (ϳ60 residues) from the end of the carboxyl-terminal kinase domain catalytic core to the carboxylterminal residue (Fig. 1B). GST-RSK-bound glutathione beads were incubated with cytosol from nonstarved COS-1 cells treated in the presence or absence of 2 M PDB. Washed beads were examined for the presence of ERK, JNK, and p38 MAPKs (Fig. 2). Immunoblotting revealed the presence of ERK from nontreated cells and active ERK from PDB-treated cells with each of the GST-RSK fusion proteins. Neither JNK nor p38 MAPK was found to be associated with the GST-RSK fusion proteins above the levels observed with GST alone. Under similar conditions, using purified recombinant kinase-defective ERK2 (K52R ERK; 100 nM), kinase-defective JNK3 (100 nM to 1 M), and wild type p38 MAPK (100 -500 nM), only ERK2 bound to the GST-RSK fusion proteins at levels above that bound to GST alone (data not shown). Thus, the carboxyl-terminal 60 amino acids of the various RSK isozymes specifically interact with ERK. Zhao et al. (21) reported that ERK does not co-immunoprecipitate with HA-RSK1. However, activation of the three RSK isozymes is blocked by the inhibition of mitogen-activated protein/ERK kinase (21). Thus, the GST-RSK1 interaction with ERK demonstrated here is consistent with the presence of an ERK-binding motif in the carboxyl-terminal tail of all three RSKs. The suggestion that ERK might bind to HA-RSK1 with less affinity than to HA-RSK2 or HA-RSK3 (21) is not excluded.
Truncation of GST-RSK1  and Identification of Residues Critical to ERK Interaction-To delineate the sequences important for ERK interaction, further truncations of the identified region of RSK1 were performed. An amino-terminal deletion to residue 716 (GST-RSK1 (716 -735) ) containing the carboxyl-terminal 20 amino acids, a region that is highly conserved between RSK and Mnk (Fig. 1B), was examined for interaction with ERK. GST-RSK1 (716 -735) was determined to be sufficient for interacting with K52R ERK2 (Fig. 3A). To localize the residues within this 20-amino acid sequence that were critical to ERK interaction, truncations of the 63-amino acid GST-RSK1 fusion protein (GST-RSK1 (672-735) ) were used in-stead of the 20-amino acid GST-RSK1 fusion protein (GST-RSK1 (716 -735) ) for two reasons: 1) using a peptide smaller than 20 amino acids fused to GST might preclude ERK interaction simply due to steric interference by GST rather than elimination of critical residues, and 2) although the carboxyl-terminal 20 residues were sufficient for interaction with ERK, additional residues might also be involved. As seen in Fig. 3B, deletion of the carboxyl-terminal 5 amino acids (GST-RSK (672-730) ) had no effect on the ability of the fusion protein to bind to K52R ERK2 compared with that of GST-RSK1 (672-735) . Removal of the carboxyl-terminal 20 amino acids (GST-RSK1 (672-715) ) or deletion of the carboxyl-terminal 9 amino acids (GST-RSK1 (672-726) ), which bisects the conserved region ( 722 LAQRRVRKL 730 of RSK1) of the RSK isozymes, completely eliminated the interaction of the fusion proteins with K52R ERK2 (Fig. 3B). Because alteration of the conserved region of RSK was deleterious to the interaction of GST-RSK1 with ERK, it is likely that a critical combination of residues within this conserved region acts as the docking site for ERK.
In Vivo Examination of RSK/ERK Interaction-To examine whether the sequence critical to in vitro interaction between GST-RSK and ERK was important for interaction within the context of the full-length kinase in vivo, the interaction between ERK and HA-tagged wild type RSK (HA-RSK2) or mutant RSK2 (HA-RSK2  ) in which the carboxyl-terminal 11 amino acids are deleted was examined in BHK cells. HA-RSK2  approximates the minimal truncation of GST-RSK1 (GST-RSK1 (672-726) ) that eliminates in vitro interaction with ERK. As seen in Fig. 4, ERK1 and ERK2 co-immunoprecipitated with HA-RSK2 but did not co-immunoprecipitate with HA-RSK2  above the levels observed in immunoprecipitates from cells transfected with the empty vector. Thus, the ERK docking site required for in vitro interaction is also essential for in vivo interaction of RSK and ERK.
To determine whether deletion of the ERK docking site altered the in vivo activation of RSK, the kinase activity of HA-RSK2 or HA-RSK2  immunoprecipitated from serumdeprived BHK cells incubated in the presence and absence of insulin was measured using the ribosomal S6 peptide as substrate. The activity of HA-RSK2 was increased ϳ6-fold by insulin treatment (Fig. 5A). Basal kinase activity of HA-RSK2  was similar to that of HA-RSK2; however, insulin treatment had no effect on the kinase activity of HA- RSK2   (Fig. 5A). These results demonstrate that although all identified ERK phosphorylation sites are intact in HA-RSK2  , the enzyme remains inactive under physiological conditions that stimulate HA-RSK2. Thus, the formation of a complex between ERK and the carboxyl-terminal tail of RSK is absolutely essential for efficient ERK activation of RSK in the living cell. Although the ERK docking site has been removed from HA-RSK  , ERK should have sufficient affinity for the consensus phosphorylation sequence to phosphorylate and activate RSK in vitro when present at supraphysiological concentrations. HA-RSK2 and HA-RSK2  were immunoprecipitated from serum-deprived BHK cells and incubated in the presence or absence of purified recombinant active ERK2. Active ERK2 increased the kinase activity of HA-RSK2  ϳ6-fold, which is identical to the levels of activation observed with HA-RSK2 (Fig.  5B), indicating that HA-RSK2 (1-729) is a functional enzyme. Additionally, truncation of the ERK docking site did not affect the ability of ERK to induce a mobility shift on SDS-PAGE as has been reported for wild type RSK (Ref. 21; Fig. 5B).
Several MAPK-activated protein kinases (MAPKAPKs) have been identified for both ERK (RSK and Mnk) and p38 MAPK (MAPKAP-2, MAPKAP-3, RSKb, and PRAK; Refs. 27-32). The most recently described MAPKAPKs are MSK1 (33) and RSKb (31); both have two kinase domains, like RSK1-3, and are novel RSKs. Additional members of each group are likely to be discovered, including a group so far undetected, activated by JNK. All identified MAPKAPKs have sequence similarities in their catalytic domains to calmodulin-activated protein kinases, residing in RSKs in the carboxyl-terminal kinase domain.
Sequence alignment of the ERK docking site of RSK1 ( 722 LAQRRVRKL) with the carboxyl-terminal tails of Mnk1/2, MAPKAPK-2, MAPKAPK-3, RSKb, PRAK, and MSK1 reveals a likely docking site within each of these MAPK-targeted kinases (Fig. 6). The order of alignment is with respect to the number of contiguous basic amino acids within the proposed docking sites (i.e. two for RSK1, and five for MAPKAPK-3). This order of alignment also allows for grouping of the kinases into ERK-specific and p38 MAPK-specific kinases, as well as a group that is regulated by both ERK and p38 MAPKs. RSK1, RSK2, RSK3, and Mnk2 are activated specifically by ERK (Fig. 6). The alignment for this group suggests that the sequence LAQRRXXXXL/I (X, any amino acid) may be important for the specific binding and activation of these MAPKAPKs by ERK. MAPKAPK-2 and MAPKAPK-3 have been extensively studied and have been found to be activated specifically by p38 MAPKs in vivo (27)(28)(29)(30). PRAK (the human homolog of mouse MAPKAPK-5) is also specific for p38 MAPK in vivo (32). RSKb (identical to MSK2) was identified in a two-hybrid screen with p38␣ MAPK and is activated predominantly by p38 MAPK and more weakly by ERK both in vitro and in cultured cells (31). Thus, alignment of the conserved region of these MAPKAPKs correlates the sequence L⌽(K/R)(K/R)(K/R)(K/R)XXXX (⌽, hydrophobic amino acid; X, any amino acid) with p38 MAPK specificity. A consensus cannot be suggested for those MAPKAPKs activated by both ERK and p38 MAPK (Mnk1 and MSK1; Refs. 22 and 33), because there are too few aligned sequences.
Thus, comparisons of MAPK specificity with residues aligned in the conserved regions correlate the number of contiguous basic amino acids (K/R) with MAPK specificity. ERK-specific MAP-KAPKs have two contiguous K/Rs, ERK/p38 MAPK-specific MAPKAPKs have three and four contiguous K/Rs, and p38 MAPK-specific MAPKAPKs have four and five contiguous K/Rs. A strong inference is that the number and position of basic residues within the context of the secondary structure (predicted to be helical) may be important determinants of specificity. Mutational analyses of these motifs and flanking sequences are warranted to assess the effect on ERK and p38 MAPK specificity. The basic residues in the putative MAPK docking sites are clustered and may also define nuclear localization sequences (34).
Specific docking sites for MAPKs in the MAPKAPKs are likely to play at least two roles in signaling. As demonstrated herein, the targeting motif for ERK is required for RSK2 activation in vivo. Thus, complex formation of MAPKs with the carboxyl-terminal tails of MAPKAPKs is likely to be a general feature that is important for the specific regulation of these MAPK-targeted kinases by MAPKs in vivo. In addition, docking sites may enable co-localization of a specific MAPK with a specific MAPKAPK in complexes where they collaborate to regulate transcriptional and translational machinery.