Identification of Regulatory Phosphorylation Sites in Mitogen-activated Protein Kinase (MAPK)-activated Protein Kinase-1a/p90 rsk That Are Inducible by MAPK*

Mitogen-activated protein kinase-activated protein kinase-1 (MAPKAP-K1; also known as p90 rsk ) contains two protein kinase domains in a single polypeptide. The N-terminal kinase domain is necessary for the phosphorylation of peptide substrates, whereas the C-terminal kinase domain is required for full activation of the N-terminal domain. Here we identify six sites in MAPKAP-K1a that become phosphorylated in transfected COS-1 cells. The inactive form of MAPKAP-K1a in unstimulated cells is partially phosphorylated at Ser222 and Ser733. Stimulation with phorbol 12-myristate 13-acetate induces the phosphorylation of Thr360, Ser364, Thr574, and Ser381 and increases the phosphorylation of Ser222 and Ser733. Our data indicate that mitogen-activated protein kinase activates the C-terminal kinase domain by phosphorylating Thr574 and contributes to the activation of the N-terminal kinase domain by phosphorylating Ser364. The activated C-terminal domain phosphorylates Ser381, which, together with phosphorylation of Ser364, activates the N-terminal kinase domain. The phosphorylation of Ser222and Ser733, which can be catalyzed by the N-terminal domain, does not activate MAPKAP-K1a per se, but Ser222 phosphorylation appears to be required for activation. Ser222, Ser364, and Ser381 are situated in analogous positions to phosphorylation sites in protein kinase B, protein kinase C, and p70S6K, suggesting a common mechanism of activation for these growth factor-stimulated protein kinases.

As implied by their name, the isoforms of mitogen-activated protein kinase-activated protein kinase-1 (MAPKAP-K1 1 ; also known as p90 rsk ) are only active in vitro if they have been phosphorylated by the p42 or p44 mitogen-activated protein kinases (MAPKs) (1). Several lines of evidence indicate that the MAPKAP-K1 isoforms are also activated by MAPKs in vivo via the Ras-dependent protein kinase cascade that is triggered by growth factors or tumor-promoting phorbol esters, such as phorbol 12-myristate 13-acetate (PMA). Thus, signals that activate MAPKs also activate MAPKAP-K1, and the phosphorylation (2) and activation (3) of MAPKAP-K1 induced by insulin or insulin-like growth factor-1 are prevented by PD 98059, a compound that specifically inhibits the activation of MAPK kinase-1 (3), the major upstream activator of MAPKs. Cyclic AMP-elevating agents that prevent the activation of MAPKs also prevent the activation of MAPKAP-K1 (4).
MAPKAP-K1 has a broad substrate specificity in vitro, phosphorylating proteins and peptides at serine residues that lie in Arg-Xaa-Arg-Xaa-Xaa-Ser or Arg-Arg-Xaa-Ser motifs (5,6). However, few physiological substrates have been identified so far. Upon activation, MAPKAP-K1 is known to translocate from the cytosol to the nucleus (7), where one of its substrates in neuronal cells is believed to be the transcription factor CREB (reviewed in Ref. 8). Activation of CREB by MAPKAP-K1 is intriguing because MAPKs are highly expressed in post-mitotic neurons, where they are activated by agonists known to play a role in memory formation (9), and an important role for CREB phosphorylation in memory formation is well documented (10). Moreover, inactivating mutations in the gene encoding human MAPKAP-K1b (RSK2) are associated with Coffin-Lowry syndrome, a disease that results in severe mental retardation as well as progressive skeletal deformation (11). A role of the MAPK cascade in long-term facilitation in Aplysia has also been reported recently (12).
MAPKAP-K1 is unusual in that the polypeptide contains two distinct protein kinase domains (13). The N-terminal kinase domain (NTD) is most similar to p70 S6K (almost 60% amino acid sequence identity) (14), whereas the C-terminal kinase domain (CTD) is most similar to phosphorylase kinase (40% amino acid sequence identity). The ability of MAPKAP-K1 to phosphorylate a variety of exogenous substrates is abolished by single amino acid replacements that inactivate the NTD (5,15,16). Inactivation of the CTD greatly reduces, but does not abolish, the phosphorylation of exogenous substrates (5,15). These observations and the finding that the isolated CTD is capable of phosphorylating itself (16) suggest that the CTD may play a role in the activation of the NTD. Indeed, evidence that MAPKAP-K1 can be autophosphorylated by either the NTD or the CTD has been presented (15).
Several years ago, we identified a threonine in the CTD of MAPKAP-K1b (Thr 574 in MAPKAP-K1a) as one of the residues phosphorylated by MAPKs in vitro (17). However, others (15) found that the mutation of this residue to Ala in MAPKAP-K1c (RSK3) generated an enzyme that could still be activated in response to the epidermal growth factor. This finding prompted us to identify all the sites in MAPKAP-K1a (RSK1) that become phosphorylated in vivo. Here we identify six phosphorylation sites in PMA-stimulated COS-1 cells, three of which are phosphorylated by MAPKs in vitro. Our results indicate that the MAPK-catalyzed phosphorylation of Ser 364 and Thr 574 is critical for activation of the NTD and CTD, respectively, and that the phosphorylation of Ser 381 , catalyzed by the CTD, is also important for activation of the NTD.
Proteins and Peptides-Escherichia coli cells (BL21(DE3) pLys S) expressing the rat glutathione S-transferase (GST), hemagglutinintagged (HA) MAPKAP-K1a fusion protein were grown at 37°C in LB medium in a 25-liter fermenter until the cells reached an absorbance of 0.6 at 600 nm. The temperature was reduced to 28°C, and isopropyl-␤-D-thiogalactopyranoside was added to 30 M. The bacteria were grown for a further 5 h, pelleted at 5000 ϫ g, and lysed, and GST-HA-MAPKAP-K1a was purified on glutathione-agarose (18), concentrated, and stored in aliquots at Ϫ80°C. The yield of full-length fusion protein (108 kDa), which composed ϳ10% of the final material, was 20 g/liter of bacterial culture. Human MAPKAP-K2 (residues 46 -400) was expressed in E. coli as a GST fusion protein and activated with p38/ reactivating kinase MAPK (19). Human p42 MAPK was expressed in E. coli as a GST fusion protein and activated with MAPK kinase-1 (18). Restriction enzymes were from New England Biolabs (Hitchin, UK). Peptides were synthesized on an Applied Biosystems Model 430A peptide synthesizer by Dr. G. Bloomberg (University of Bristol).
Antibodies-All antibodies were raised in sheep at the Scottish Antibody Production Unit (Lanarkshire, UK). One antibody was raised against the peptide RNQSPVLEPVGRSTLAQRRGIKK (corresponding to the last 23 residues of rabbit MAPKAP-K1b) and affinity-purified on peptide CH-Sepharose columns.
Construction of Site-specific Mutant Plasmids-The pMT2 construct encoding rat MAPKAP-K1a with a hemagglutinin tag (YPYDVPDYA) at the N terminus (21) (hereafter termed HA-MAPKAP-K1a) contained an extra codon (CAG) from the published sequence, resulting in an insertion of glutamine at position 157. pBluescript phagemid containing DNA encoding rat HA-MAPKAP-K1a (p90 rsk-1 ) (21) was modified by polymerase chain reaction mutagenesis to generate the D206A and D558A mutants (5). The mutants S222A, T360A, T360E, S364A, S364E, S381A, and S381E were produced by digestion of pBluescript HA-MAPKAP-K1a with BglII and PinAI and ligation of a mutated BglII-PinAI-digested polymerase chain reaction fragment. The mutants T574A and T574D and the double mutant D558A/T574A were produced using PinAI and NdeI. The mutations were introduced by separate polymerase chain reaction procedures to generate the double mutant. The mutants T613A, T622A, and S733A and the double mutant T613A/ T622A were produced using NdeI and HindIII. The polymerase chain reaction fragments were produced essentially as described (22), and the sequences of the mutants were checked using an Applied Biosystems automatic DNA sequencer and specific oligonucleotide primers. The mutant and wild-type MAPKAP-K1a DNA constructs were expressed by subcloning from the Bluescript phagemid vector into the pMT2 and pGEX4T1 vectors using the unique EcoRI site for expression in COS-1 cells and the BL21(DE3) pLys S strain of E. coli, respectively. All constructs were prepared using the QIAGEN plasmid maxikit accord-ing to the manufacturer's protocol. 2 Measurement of MAPKAP-K1 Activity in COS-1 Cells-HA-MAP-KAP-K1a was transiently expressed in COS-1 cells (5) and incubated for 3 h in the absence of serum and in the presence or absence of 50 M PD 98059 before stimulation with 80 nM PMA. The cells were washed and lysed (5), and transfected HA-MAPKAP-K1a was immunoprecipitated as described below. The endogenous enzyme was immunoprecipitated using a MAPKAP-K1 antibody. The immunoprecipitates were assayed for MAPKAP-K1 activity using the peptide S6 218 -249 (G 245 / G 246 ) (KEAKEKRQEQIAKRRRLSSLRASTSKSGGSQK) (23). One unit of MAPKAP-K1 activity is that amount which incorporates 1 nmol of phosphate into the substrate peptide in 1 min.
Immunoprecipitation of HA-MAPKAP-K1a from 32 P-Labeled COS-1 Cells-Cells transfected with HA-MAPKAP-K1a were washed three times in phosphate-free medium, incubated for 1 h, and washed twice more in the same medium. The cells were then incubated for 4 h with carrier-free [ 32 P]orthophosphate (1 mCi/ml) in the presence or absence of PD 98059. After appropriate stimulation, the cells were washed four times with ice-cold Dulbecco's modified essential medium and then lysed in 300 l of lysis buffer (20 mM Tris-HCl, pH 7.5 (20°C), 0.1% (by volume) 2-mercaptoethanol, 2 mM EDTA, 2 mM EGTA, 1% (by mass) Triton X-100, 0.25 M NaCl, Complete™ protease inhibitor mixture (Boehringer), 1 mM sodium orthopervanadate, 10 mM sodium glycerophosphate, 50 mM NaF, 5 mM sodium pyrophosphate, 1 M microcystin-LR, and 0.27 M sucrose). The lysates were incubated with agitation for 30 min at 4°C with protein G-Sepharose incubated with preimmune sheep serum and centrifuged for 10 min at 13,000 ϫ g, and the supernatant was added to protein G-Sepharose coupled to monoclonal antibody 12CA5, which recognizes the HA epitope. After agitation for 90 min at 4°C, the complex was washed eight times with 1.0 ml of lysis buffer containing 0.5 M NaCl and then twice with 50 mM Tris-HCl, pH 6.8, and 0.1% (by volume) 2-hydroxyethanethiol.
Proteolytic Digestion of HA-MAPKAP-K1a-Immunoprecipitated 32 P-labeled HA-MAPKAP-K1a was denatured in SDS and resolved by electrophoresis on a 7.5% SDS-polyacrylamide gel. After autoradiography, the 86-kDa band corresponding to phosphorylated HA-MAPKAP-K1a was excised, eluted, precipitated, and oxidized as described (24). The protein was then incubated at 30°C in 0.3 ml of 50 mM Tris-HCl, pH 8.0 (20°C), and 0.1% (by volume) reduced Triton X-100 containing 5 g of trypsin or chymotrypsin. After 3 h, another 5 g of protease was added, and the suspension was left for a further 12 h. Guanidinium chloride (8 M) was added to a final concentration of 1 M to precipitate any residual SDS, and after standing on ice for 10 min, the suspension was centrifuged for 10 min at 13,000 ϫ g. The supernatant containing Ͼ90% of the 32 P radioactivity was removed and fractionated as described below.
Identification of Phosphorylation Sites-A tryptic digest of 32 P-labeled HA-MAPKAP-K1a obtained from 8 mg of cell lysate was first purified on a column of chelating Sepharose (Pharmacia) saturated with iron(III) chloride (25). The phosphopeptides were recovered (70% yield) by elution with water/ethanol/triethylamine, pH 10.5; dried; redissolved in 0.1% (by volume) trifluoroacetic acid; and applied to a 24 cm ϫ 4-mm Vydac 218TP54 C 18 column (Separations Group, Hesperia, CA) equilibrated with 0.1% (by volume) trifluoroacetic acid and connected to an on-line monitor for continuous measurement of 32 P radioactivity. The column was developed with a linear acetonitrile gradient (0.33%/min) in 0.1% (by volume) trifluoroacetic acid. The flow rate was 0.8 ml/min, and fractions of 0.4 ml were collected. Typically, Ͼ80% of the 32 P radioactivity applied to the column was recovered. Six major tryptic peptides (T1-T6) and chymotryptic peptides (C1-C6) were obtained. Peptide T1 was further purified by chromatography on a microbore C 18 column equilibrated with 10 mM ammonium acetate, pH 6.5, and aliquots of peptides T1, T2, and T5 were then analyzed on an Applied Biosystems Model 476A sequencer to determine the amino acid sequence. Additional aliquots of peptides T1-T6 and chymotryptic peptides C1-C6 were coupled covalently to a Sequelon arylamine membrane and analyzed on an Applied Biosystems Model 470A sequencer. 32 32 P-labeled full-length GST-MAPKAP-K1a produced by phosphorylation with MAPKAP-K2 was excised, digested with trypsin, and chromatographed on the C 18 column as described above. The two major 32 P-labeled peptides eluting at 18.5% acetonitrile (peptide T1) and 27.7% acetonitrile (peptide T5) were further purified by chromatography on the microbore C 18 column at pH 6.5, and their amino acid sequences were determined.

Activation of MAPKAP-K1a in COS-1 Cells
MAPKAP-K1a could only be expressed in E. coli in very low yield, was highly degraded (see "Experimental Procedures"), and could not be activated to a significant extent by p42 MAPK. For these reasons, the mechanism of activation of MAPKAP-K1a was studied by transiently transfecting the cDNA into COS-1 cells, followed by immunoprecipitation of the HA-tagged enzyme from the cell lysates (see "Experimental Procedures"). Following stimulation of the cells with 80 nM PMA, the activity of transfected HA-MAPKAP-K1a increased 10 -15-fold, peaking after 10 -15 min, similar to the activation of the endogenous MAPKAP-K1 activity (Fig. 1). The rate of activation of the expressed and endogenous enzymes was strongly suppressed by PD 98059, an inhibitor of the activation of MAPK kinase-1 (3), indicating that the activation of MAPKAP-K1 is catalyzed by MAPKs. The failure of PD 98059 to suppress the PMAinduced activation of MAPKAP-K1 completely has been noted previously in Swiss 3T3 cells (3) and is likely to be explained by the finding that PD 98059 inhibits the activation of MAPK kinase by only 80 -90%. Due to the enormous amplification potential of the MAPK cascade, the 10 -20% residual MAPK kinase activity is sufficient to induce significant activation of MAPKAP-K1. However, the possibility that another PMA-activated PD 98059-insensitive kinase cascade may also contribute to the activation of MAPKAP-K1 cannot be excluded.
The activity of transfected HA-MAPKAP-K1a was 10-fold higher than the endogenous MAPKAP-K1 activity after stimulation with PMA ( Fig. 1), even though the efficiency of transfection was only 30%. Thus, the intracellular concentration of HA-MAPKAP-K1a is far higher than that of the endogenous MAPKAP-K1. All the experiments presented below were carried out on HA-MAPKAP-K1 immunoprecipitated from cells stimulated for 15 min with PMA.

Recombinant HA-tagged MAPKAP-K1a Is Constitutively Phosphorylated at Ser 222 and Ser 733
We first wished to identify the residues in MAPKAP-K1a that were phosphorylated under basal conditions where MAP-KAP-K1a was inactive. 32 P-Labeled COS-1 cells were therefore treated for 3 h with PD 98059 in the absence of PMA (Fig. 1), and phosphoamino acid analysis revealed that, under these conditions, inactive MAPKAP-K1a was phosphorylated only at serine residues. Following tryptic digestion and chromatography on a C 18 column, two major 32 P-labeled peptides were recovered (termed T1 and T2) that eluted at 18.5 and 24% acetonitrile, respectively ( Fig. 2A). Both peptides were phosphorylated at serine, and a single burst of 32 P radioactivity was observed after the fourth (peptide T1) and third (peptide T2) cycles of Edman degradation (data not shown). Further experiments described below established that peptides T1 and T2 corresponded to residues 730 -736 and 220 -237, respectively, and that in unstimulated cells treated with PD 98059, MAP-KAP-K1a was phosphorylated at Ser 222 in the "activation loop" of the NTD and at Ser 733 four residues from the extreme C terminus.

PMA Induces the Phosphorylation of Thr 360 , Ser 364 , Ser 381 , and Thr 574 in HA-MAPKAP-K1a
Evidence from Tryptic Digestion-To identify the residues that became phosphorylated when MAPKAP-K1a was activated, the experiments described above were repeated after stimulation of the COS-1 cells with PMA. Stimulation with PMA was accompanied by a 5-6-fold increase (three experiments) in the 32 P labeling of MAPKAP-K1a with respect to cells incubated with PD 98059 in the absence of PMA ( Fig. 2A). Phosphoamino acid analysis revealed that threonine residues became phosphorylated upon cell stimulation with PMA, although the majority of the phosphate was still present as phosphoserine (data not shown). Tryptic digestion and C 18 chromatography revealed that stimulation with PMA not only caused a 2-3-fold increase in the phosphorylation of peptides T1 and T2 (three experiments), but also induced the 32 P labeling of two new major peptides termed T4 (21.8% acetonitrile) and T5 (27.7% acetonitrile) along with several minor peptides termed T3 (13% acetonitrile), T6 (28.8% acetonitrile), and T7 (30% acetonitrile) (Fig. 2C).
Peptide T4 was also a phosphoserine-containing peptide, and a single burst of 32 P radioactivity occurred after the second cycle of Edman degradation (Fig. 3D). There are three serine residues in human MAPKAP-K1a that occur two residues after trypsin-sensitive peptide bonds, namely Ser 154 , Ser 364 , and Ser 631 . The predicted tryptic phosphopeptides containing Ser 154 (LSK) and Ser 631 (IGSSK) would elute from C 18 columns at much lower acetonitrile concentrations than peptide T4. In addition, peptide T4 comigrated during C 18 chromatography and isoelectric focusing with a synthetic phosphopeptide comprising residues 363-377 and phosphorylated at Ser 364 . These results, and further experiments described below, established that Ser 364 was phosphorylated in vivo in response to PMA.
Peptide T7 was a phosphothreonine-containing peptide, and a single burst of radioactivity occurred after eight cycles of Edman degradation (Fig. 3E). The only threonine residues in MAPKAP-K1 located eight residues after an Arg or Lys residue are Thr 574 and Thr 519 . Thr 574 is located at the position equivalent to a Thr residue in rabbit MAPKAP-K1b phosphorylated by p42 MAPK in vitro (17). Further evidence presented below indicates that this residue is indeed the site of phosphorylation.
The minor phosphoserine-containing peptide T3 and the minor phosphothreonine-containing peptide T6 could not be purified in sufficient amounts for the sites of phosphorylation to be identified. Peptide T6 may contain Thr 360 since evidence presented below demonstrates that this residue is phosphorylated in vivo. However, peptide T6 was not present in all tryptic digests (cf. Fig. 8). The smallest tryptic peptide containing Thr 360 would be expected to comprise residues 334 -362. Because of its size and hydrophobicity, this peptide may be insoluble and hence not recovered from the C 18 column.
Evidence from Chymotryptic Digestion-To obtain further evidence that the assignment of phosphorylation sites was correct, 32 P-labeled HA-MAPKAP-K1a from 32 P-labeled COS-1 cells was digested with chymotrypsin and chromatographed on a C 18 column. Two prominent 32 P-labeled peptides termed C1 (17.9% acetonitrile) and C2 (18.5% acetonitrile) were isolated from cells incubated with PD 98059 in the absence of PMA (Fig.  4A). Peptide C1 contained phosphoserine, and 32 P radioactivity was released after the first cycle of Edman degradation (data not shown). This is consistent with peptide C1 containing Ser 222 and resulting from chymotryptic cleavage at Tyr 221 . Peptide C2 also contained phosphoserine, and 32 P radioactivity was released after the fourth cycle of Edman degradation (data not shown). Moreover, peptide C2 coeluted with tryptic peptide T1 (Fig. 2C). These observations indicated that peptide C2 was phosphorylated at Ser 733 and that it was generated by a pseudotryptic cleavage between Arg 729 and Lys 730 .
The 32 P labeling of peptides C1 and C2 was increased 2-3fold when cells were stimulated with PMA in the absence of PD 98059, as expected from the results presented in Fig. 2. In addition, several new chymotryptic phosphopeptides were generated. Peptides C3 (13.5% acetonitrile) and C4 (14.2% acetonitrile) both contained phosphothreonine as well as phosphoserine, and in both cases, 32 P radioactivity was released after the fourth and eighth cycles of Edman degradation. Based on additional evidence presented below, these peptides are most likely to have resulted from cleavage at Phe 356 and to be phosphorylated at Thr 360 and Ser 364 .
Peptide C5 (22.6% acetonitrile) also contained phosphoserine, and 32 P radioactivity was released after four cycles of Edman degradation (data not shown). This is consistent with peptide C5 being generated by cleavage at Phe 377 , with the site of phosphorylated being Ser 381 .
Peptide C6, a phosphothreonine-containing peptide, eluted at the same acetonitrile concentration as peptide-(566 -581) phosphorylated at Thr 574 by MAPK in vitro (17). Consistent with this assignment, 32 P radioactivity was released after the ninth cycle of Edman degradation (Fig. 3F).
Evidence from Immunoblotting-To assign the sites of phosphorylation by an independent method, antibodies were produced that recognized MAPKAP-K1a only when it was phosphorylated at particular residues. The antisera were generated in sheep against synthetic peptides from MAPKAP-K1a that were phosphorylated at Thr 360 , Ser 364 , or Ser 381 , and the phospho-specific antibodies were purified by affinity chromatography (see "Experimental Procedures"). All three antibodies recognized the 86-kDa HA-MAPKAP-K1a band from PMAstimulated COS-1 cells, but did not recognize HA-MAPKAP-K1a from unstimulated cells that had been preincubated with PD 98059 (Fig. 5A). In the absence of PD 98059, Ser 381 was recognized weakly by the anti-phospho-Ser 381 antibody. The specificity of each antibody was confirmed in two ways: first, by the finding that the antibodies did not recognize HA-MAPKAP-K1a from PMA-stimulated cells if they were first incubated with the phosphopeptide immunogen (data not shown), and second, by the finding that mutation of a particular phosphorylation site to Ala or Glu prevented the relevant antibody from recognizing HA-MAPKAP-K1a (Fig. 5B). Interestingly, MAP-KAP-K1a expressed in E. coli was phosphorylated at Ser 381 , and Ser 381 phosphorylation increased markedly following incubation with MAPK and MgATP (Fig. 5C).

Effect of Mutagenesis of Phosphorylation Sites on the Activation of MAPKAP-K1
The previous experiments identified two residues (Ser 222 and Ser 733 ) that were partially phosphorylated in unstimulated cells and four residues (Thr 360 , Ser 364 , Ser 381 , and Thr 574 ) that became phosphorylated in response to PMA. Thr 360 , Ser 364 , and Thr 574 are followed by proline residues, suggesting that they are likely to have been phosphorylated by MAPKs. Indeed, Thr 574 was identified previously as a residue phosphorylated by MAPK in vitro (17). The phosphorylation of Thr 360 and Ser 364 by MAPKs was confirmed by the finding that, after incubation of the bacterially expressed enzyme with MgATP and p42 MAPK in vitro, MAPKAP-K1a was recognized by the antibodies raised against these phosphorylated epitopes (Fig.  5C).
To investigate the role of each phosphorylation site in the activation of HA-MAPKAP-K1a, these Ser and Thr residues were initially mutated to Ala, and the mutants were transiently expressed in COS-1 cells. The PMA-induced activation of HA-MAPKAP-K1a was virtually abolished in the S364A and S381A mutants and suppressed by 85% in the S222A and T574A mutants (Fig. 6A). In contrast, the basal activation and the PMA-induced activation of HA-MAPKAP-K1a were hardly affected in the T360A and S733A mutants.
Rat MAPKAP-K1a contains three other threonine residues that are followed by proline, namely Thr 613 , Thr 622 , and Thr 712 . Although the evidence presented above indicated that these residues were not phosphorylated in vivo, Thr 613 and Thr 622 are conserved in every form of MAPKAP-K1 identified so far. However, their mutagenesis to Ala either individually (data not FIG. 3. Identification of the phosphorylation sites in peptides T1, T2, T4, T5, T7, and C6. 32 P-Labeled tryptic phosphopeptides from Fig. 2B and chymotryptic phosphopeptide C6 from Fig. 4 were subjected to Edman degradation, and 32 P radioactivity released after each cycle was measured. The amino acid sequences of peptides T1, T2, and T5 (single letter code) were determined by gas-phase sequencer analysis, whereas the sequence of peptide C6 was determined previously (17). The structure of peptide T4 was inferred from other data presented under "Results." shown) or in combination (Fig. 6) had little effect on the activation of MAPKAP-K1a by PMA. The third threonine (Thr 712 ) is not conserved in the Drosophila homologue (26) and is therefore unlikely to be an important regulatory phosphorylation site.
Ser 364 , Ser 381 , and Thr 574 were also mutated to acidic residues to try to mimic the effect of phosphorylation by introducing a negative charge. The S364E mutant, like the S364A mutant, had greatly reduced basal activity in unstimulated cells. However, in contrast to the S364A mutant, the S364E mutant could still be activated by PMA, albeit to only 6% of the activity of the wild-type enzyme. These experiments suggested that the presence of a negatively charged residue at Ser 364 was important for the activation of MAPKAP-K1a and indicated that the phosphorylation of at least one residue other than Ser 364 was important for activation.
Unlike the S381A mutant, the basal activity of the S381E mutant was the same as that of the wild-type enzyme, and it could be activated to 25% of the level of the wild-type enzyme after stimulation with PMA. These experiments demonstrated that the phosphorylation of at least one residue other than Ser 381 was important for activation of the NTD.
The T574D mutant, unlike the T574A mutant, had the same basal activity as wild-type MAPKAP-K1a and could be activated to 80% of the level of the wild-type enzyme after stimulation with PMA. This indicated that a negatively charged residue at position 574 was important for the activation of MAPKAP-K1a and also that phosphorylation of at least one residue other than Thr 574 was critical for activation of the NTD.

Effect of Inactivating the Kinase Domains on the Phosphorylation of MAPKAP-K1a in COS-1 Cells
To investigate the roles of the NTD and CTD in phosphorylating the residues in MAPKAP-K1a that were not followed by proline (Ser 222 , Ser 381 , and Ser 733 ), the phosphorylation of these sites was investigated using additional mutants of MAP-KAP-K1 in which either the NTD or the CTD had been inactivated. The D558A mutant, in which the CTD should be inactivated, like the T574A mutant, decreased the basal and PMAstimulated activities of MAPKAP-K1a by 85%. Significantly, the basal and PMA-stimulated activities of the D558A/T574A double mutant were identical to those of the single mutants (Fig. 7).
The reduction in phosphorylation of Ser 381 was confirmed by immunoblot experiments (Fig. 8C). Similar results were obtained with the T574A mutant (data not shown).
Inactivation of the NTD caused some decrease in the 32 P labeling of Ser 733 (peptide T1); had little effect on the 32 P labeling of Ser 364 (peptide T4), Ser 381 (peptide T5), and Thr 574 (peptide T7); but greatly reduced the 32 P labeling of Ser 222 (peptide T2) (Fig. 8B). The lack of effect on Ser 381 phosphorylation was confirmed by immunoblotting (Fig. 8C).

DISCUSSION
In this paper, we have identified six residues in MAPKAP-K1a that are phosphorylated in COS-1 cells. One of these (Ser 222 ) is located in the NTD, one (Thr 574 ) in the CTD, three (Thr 360 , Ser 364 , and Ser 381 ) in the intervening sequence between the NTD and CTD, and one (Ser 733 ) in the C-terminal tail (Fig. 9). Thr 360 , Ser 364 , and Thr 574 lie in the minimum consensus sequence for phosphorylation by MAPKs ((Ser/Thr)-Pro), are phosphorylated by MAPKs in vitro (Fig. 5 and Ref. 17), and become phosphorylated in response to an agonist (PMA) that activates MAPKs in vivo (Figs. 2-4). We therefore conclude that these three residues are phosphorylated by MAPKs in vivo. The phosphorylation of the other three serine residues is also induced/increased in response to PMA, but these residues are not followed by proline (Fig. 9), indicating that they are targeted by distinct protein kinases. We argue below that Ser 381 is phosphorylated by the CTD of MAPKAP-K1a and that Ser 222 and probably Ser 733 are phosphorylated by the NTD (Fig. 9).
The effect of the T574A mutation on the activation of MAP- KAP-K1a by PMA was similar to that produced by inactivation of the CTD, with either mutation reducing the PMA-induced activation of the NTD by 85%, whereas combining both mutations had no further effect (Figs. 6 and 7). Similar results were obtained by Bjørbaek et al. (15) using the MAPKAP-K1c (RSK3) isoform. These investigators found that changing Thr 570 to Ala (the residue equivalent to Thr 574 in MAPKAP-K1a) and inactivating the CTD by changing Lys 464 to Ala reduced basal and epidermal growth factor-stimulated activation to a similar extent. Fischer and Blenis (16) also showed that the CTD expressed alone was capable of undergoing a slow autophosphorylation reaction that was enhanced 7-fold after incubation with MAPK and MgATP. Taken together, these results strongly suggest that the MAPK-catalyzed phosphorylation of Thr 574 triggers activation of the CTD. The T574D mutant had a very low activity in unstimulated cells, but could still be activated in response to PMA (Fig. 6), indicating that the presence of an acidic residue at position 574 can substitute for the phosphorylated threonine, at least partially.
The T574A mutation (data not shown) or inactivation of the CTD greatly reduced the PMA-stimulated phosphorylation of Ser 381 , whereas inactivation of the NTD had no effect on the 32 P labeling of this site (Fig. 8). These observations indicate that the phosphorylation of Ser 381 is catalyzed by the CTD in vivo after it has been activated by the phosphorylation of Thr 574 . Two further lines of evidence indicate that the phosphorylation of Ser 381 is catalyzed by the CTD. First, Vik and Ryder (27) have shown that Ser 381 is a major site that becomes phosphorylated when native MAPKAP-K1 is incubated with MgATP in vitro. These investigators expressed wild-type Xenopus MAPKAP-K1b (termed S6KII) and the isolated NTD and CTD of avian MAPKAP-K1 in Sf9 cells. They found that the native enzyme and the isolated CTD, but not the isolated NTD, underwent a slow intermolecular autophosphorylation reaction in vitro. Second, we have shown that bacterially expressed MAPKAP-K1a is partially phosphorylated at Ser 381 (presumably as a result of autophosphorylation) and that the phosphorylation of this residue increases upon incubation with p42 MAPK and MgATP (Fig. 5C). The low level of Ser 381 phosphorylation induced by PMA after transfection of the D558A mutant of MAPKAP-K1a, in which the CTD should be inactive (Fig. 8), is presumably catalyzed by the endogenous MAP-KAP-K1 isoforms, which are present in vivo at much lower concentrations.
The S381A mutation virtually abolished the PMA-induced activation of MAPKAP-K1a, suggesting that the phosphorylation of this residue is critical for activation of the NTD. The lack of phosphorylation of Ser 381 may also explain why the isolated NTD lacking the CTD displays very low activity after transient transfection in COS cells (15). In contrast, the S381E mutation was still activated by PMA, albeit to only 25% of the level of the wild-type enzyme (Fig. 6); however, the S381E mutant was not constitutively active in unstimulated cells. These results not only demonstrate the importance of a negatively charged residue at position 381 for the activation of MAPKAP-K1a, but also indicate that phosphorylation of an additional residue is required to activate the NTD. This residue appears to be Ser 364 because its mutation to Ala abolished the activation of MAP-KAP-K1a by PMA (Fig. 6). The S364E mutant was not active in unstimulated cells, but could be activated by PMA to ϳ10% of the level of the wild-type enzyme (Fig. 6). This suggests that a negative charge at residue 364 is necessary for Ser 381 phosphorylation to be able to induce any activation. The T360A mutant had no effect on the activation MAPKAP-K1a by PMA (Fig. 6), and the role of Thr 360 phosphorylation remains to be determined.
Ser 222 and Ser 733 were partially phosphorylated in unstimulated cells pretreated with PD 98059 under conditions where transfected HA-MAPKAP-K1a was inactive. This demonstrated that the phosphorylation of Ser 222 and Ser 733 per se was insufficient for activation. The PMA-induced activation of MAPKAP-K1a was also unaffected in the S733A mutant. However, the S222A mutation suppresses PMA-induced activation of MAPKAP-K1a by 90% (Fig. 6), and the equivalent mutation in MAPKAP-K1c reduces the epidermal growth factor-stimulated activation of this isoform to a similar extent (15). Moreover, the phosphorylation of Ser 222 was greatly decreased under another condition where PMA-induced activation was suppressed, i.e. after inactivation of the CTD (Fig. 8A). Thus, the presence of phosphate at Ser 222 may also be important for the activation of MAPKAP-K1. Ser 222 lies in the position equivalent in the NTD to Thr 574 in the CTD, with these residues lying just N-terminal to the conserved Ala-Pro-Glu motif in the activation loop between kinase subdomains 7 and 8.
The phosphorylation of Ser 222 was greatly decreased in PMA-stimulated COS-1 cells when either the CTD or NTD was inactivated (Fig. 8, A and B). These observations suggest that Ser 222 is phosphorylated by the NTD and that the reduction in Ser 222 phosphorylation when the CTD is inactivated is a consequence of the 85% inhibition of the NTD under these conditions (Figs. 6 and 7). Phosphorylation of Ser 222 by the NTD can also explain why the 32 P labeling of this residue is enhanced by PMA (Figs. 2 and 4). The ability of the NTD to catalyze the phosphorylation of MAPKAP-K1c in vitro has been documented previously (15).
The phosphorylation of Ser 733 in PMA-stimulated COS-1 cells was unaffected by inactivation of the CTD and was reduced by 60 -70% when the NTD was inactivated. Ser 733 lies in the minimum consensus sequence for phosphorylation by the NTD (Arg-Xaa-Xaa-Ser), and phosphorylation by the NTD can account for the increased phosphorylation of Ser 733 that occurs in response to PMA (Fig. 2).
Transfected HA-MAPKAP-K1a in PMA-stimulated COS-1 cells still contained a small amount of phosphate at Ser 222 or Ser 733 even when the NTD or CTD was inactivated (Fig. 8). The residual phosphorylation of these residues may result from an intermolecular autophosphorylation catalyzed by endogenous MAPKAP-K1 present in COS-1 cells, as suggested above for the trace phosphorylation of Ser 381 under these conditions. However, Ser 222 and Ser 733 were partially phosphorylated in COS-1 cells pretreated with PD 98059, i.e. under conditions where MAPKAP-K1a is inactive. Ser 222 and Ser 733 may therefore also be targeted by another protein kinase(s) that remains to be identified.
The PMA-induced activation of MAPKAP-K1a was unaffected by the S733A mutation, and the role of Ser 733 phosphorylation remains to be determined. It could be involved in modulating the activity of the CTD, but it is intriguing that it is located very near the C terminus of MAPKAP-K1, which has been shown to be the region required for the specific interaction of MAPKAP-K1c with MAPKs (28). The phosphorylation of Ser 733 may therefore affect the interaction between MAPKAP-K1a and its upstream activator. Since MAPKAP-K1 and MAPKs both translocate from the cytosol to the nucleus upon activation, an intriguing possibility is that Ser 733 may trigger the dissociation of activated MAPKAP-K1a from MAPKs, thereby facilitating the translocation of both enzymes to the nucleus. However, it should be noted that Ser 733 is lacking in MAPKAP-K1 from Drosophila (26).
In summary, we have shown that two independent MAPKcatalyzed phosphorylation events are required to activate the NTD and CTD of MAPKAP-K1a. Activation of the CTD requires the MAPK-catalyzed phosphorylation of Thr 574 , allowing it to phosphorylate Ser 381 . The phosphorylation of Ser 381 and the MAPK-catalyzed phosphorylation of Ser 364 appear to be critical for the activation of the NTD, which may then enhance the phosphorylation of Ser 222 in the NTD and Ser 733 in the CTD. Although more work is necessary to understand why the activation of MAPKAP-K1 is so complex and to define the role of each phosphorylation event precisely, it is intriguing that Ser 222 , Ser 364 , and Ser 381 lie in positions analogous to residues whose phosphorylation is important for the activation and/or stability of other growth factor-stimulated protein kinases, namely p70 S6K (29,30), protein kinase B (31), and protein kinase C (32, 33) (Fig. 9B). In MAPKAP-K1a, the PMAinduced increase in Ser 222 phosphorylation appears to be mediated by the NTD, but the equivalent residue in protein kinase B is phosphorylated by 3-phosphoinositide-dependent protein kinase-1 (34,35). It would clearly be of considerable interest to investigate whether 3-phosphoinositide-dependent protein kinase-1 can phosphorylate the equivalent residues in MAPKAP-K1, p70 S6K , and protein kinase C. As in MAPKAP-K1a, the residues equivalent to Ser 364 are always followed by proline (Fig. 9B), but only in MAPKAP-K1a is there evidence that phosphorylation is catalyzed by the MAPKs of the "classical" MAPK cascade. The protein kinase(s) that phosphorylate this residue in p70 S6K , protein kinase B, and protein kinase C are unknown. p70 S6K , but not protein kinase B or protein kinase C, also contains a residue equivalent to Thr 360 in MAP-KAP-K1a (Fig. 9B), and this site is known to be phosphorylated in vivo (29,30). The residues equivalent to Ser 381 all lie in a Phe-Xaa-Xaa-Phe-(Ser/Thr)-(Phe/Tyr) motif (Ref. 36 and Fig.  9B). In MAPKAP-K1a, Ser 381 is phosphorylated by the CTD, but in p70 S6K , protein kinase B, and protein kinase C, it is phosphorylated by a distinct protein kinase(s) that has not yet been identified. As suggested by Vik and Ryder (27), it is also possible that the CTD does not just phosphorylate the NTD in vivo and that it has additional physiological substrates, such as FIG. 9. Location of the major in vivo sites of phosphorylation in HA-MAPKAP-K1a. A, Ser 222 and Thr 574 are both located nine residues N-terminal to the conserved Ala-Pro-Glu motif in subdomain VIII of each kinase domain, whereas Ser 733 is four residues from the C terminus. Thr 360 , Ser 364 , and Ser 381 are located between the two kinase domains. B, residues equivalent to Ser 222 , Ser 364 , and Ser 381 in MAPKAP-K1 and the sequences surrounding these sites are conserved in p70 S6K , protein kinase B␣ (PKB␣), and protein kinase C␣ (PKC␣). Thr 360 in MAPKAP-K1 is also present in p70 S6K . Identical residues in all four kinases are shown in boldface type, and phosphorylation sites are underlined and are in boldface type. p70 S6K . In addition, activation of the CTD might be triggered by distinct signal transduction pathways and occur independently of the activation of the NTD.