A Novel MAPK Phosphatase MKP-7 Acts Preferentially on JNK/SAPK and p38α and β MAPKs

Mitogen-activated protein kinases (MAPKs) are inactivated via dephosphorylation of either the threonine or tyrosine residue or both in the P-loop catalyzed by protein phosphatases which include serine/threonine phosphatases, tyrosine phosphatases, and dual specificity phosphatases. Nine members of the dual specificity phosphatases specific for MAPKs, termed MKPs, have been reported. Each member has its own substrate specificity, tissue distribution, and subcellular localization. In this study, we have cloned and characterized a novel MKP, designated MKP-7. MKP-7 is most similar to hVH5, a member of previously known MKPs, in the primary structure. MKP-7 is predominantly localized in the cytoplasm when expressed in cultured cells, whereas hVH5 is both in the nucleus and the cytoplasm. MKP-7 binds to and inactivates p38 MAPK and JNK/SAPK, but not ERK. Furthermore, we have found that MKPs have the substrate specificity toward the isoforms of the p38 family (α, β, γ, and δ). MKP-7 binds to and inactivates p38α and -β, but not γ or δ. MKP-5 and CL100/MKP-1 also bind to p38α and -β, but not γ or δ. Finally, we propose a tentative classification of MKPs based on the sequence characteristics of their MAPK-docking site.

Here we have identified and characterized a novel MKP, designated MKP-7. MKP-7 shows high homology with all known MKPs and shares all the features of dual specificity phosphatases. MKP-7 is most similar to hVH5 in the primary sequence. Both molecules have a C-terminal stretch, which is not present in other MKPs. MKP-7 possesses a nuclear export signal (NES) sequence in the C-terminal stretch region, which regulates the subcellular localization of the molecule. MKP-7 binds to and inactivates p38 MAPK and JNK/SAPK, but not ERK. Furthermore, we show for the first time that MKPs including MKP-7 have the substrate specificity toward the isoforms of the p38 MAPK family. MKP-7 binds to and inactivates p38␣ and -␤, but not ␦ or ␥. CL100/MKP-1 and MKP-5 also bind to p38␣ and -␤, but not ␦ or ␥. Finally, we propose a tentative classification of MKPs based on their sequences of the MAPK-docking site.

EXPERIMENTAL PROCEDURES
5Ј-and 3Ј-Rapid Amplification of cDNA Ends-A Super Script human fetal brain cDNA library (Life Technologies, Inc.) was used as a template. The vector sequence (5Ј-caccaaacagctatgacc-3Ј) and a gene specific primer (5Ј-gtttcgctgaatgctggatgagctc-3Ј) were used for the first PCR of 5Ј-rapid amplification of cDNA ends, and the T7 primer and a gene specific primer (5Ј-gagctcatccagcattcagcgaaac-3Ј) were used for the first PCR of 3Ј-rapid amplification of cDNA ends. The SP6 primer and gene-specific primers (5Ј-gcaaccttcgcttcataagcttggagcag-3Ј or 5Ј-ctgctc-* This work was supported in part by grants-in-aid from the Ministry of Education, Science and Culture of Japan (to E. N). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. caagcttatgaagcgaaggttgc-3Ј) were used for nested PCR. Klentaq polymerase mix (CLONTECH) was used in PCR reactions and PCR products were subcloned into TOPO TA cloning vector (Invitrogen). The fulllength MKP-7 was obtained by the PCR method using Super Script human fetal brain library as a template. The primers used were 5Јagatctatggcccatgagatgattggaactc-3Ј and 5Ј-agatcttcaggagacctcaatgatttccatgctg-3Ј.
Phosphatase Assay-A catalytic activity of Myc-MKP-7 toward pnitrophenyl phosphate (pNPP) (Sigma) was measured at 37°C. Myc- Cell Cultures and Transfection-NIH3T3 cells were cultured in Dulbecco's modified Eagle's medium containing 10% calf serum. COS7 cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. These cells were maintained in 5% CO 2 at 37°C. Cells were split on 35-or 60-mm dishs at 2 ϫ 10 5 or 5 ϫ 10 5 cell number per  Cell Staining-NIH3T3 cells and COS7 cells were transfected with Myc-tagged MKP-7 or Myc-tagged hVH5. After 24 h, cells were fixed with 3.7% formaldehyde in PBS for 10 min at room temperature. After three times wash with PBS, cells were permeabilized in 0.5% Triton X-100 in PBS and washed with PBS three times. And then, cells were incubated with anti-Myc antibody (9E10) (Santa Cruz) in 3% bovine serum albumin in PBS for 16 h at 4°C. Cells were washed three times with PBS, and incubated with anti-mouse IgG secondary antibody in 3% bovine serum albumin in PBS for 1 h at 37°C. After three times wash with PBS, and twice with Milli-Q water, coverslips were mounted with mowiol. Fluorescence was viewed with a Zeiss fluorescent microscope.

RESULTS
Isolation of MKP-7 cDNA-We previously identified a member of MKP, MKP-5, and revealed its MAPK-docking site (36,50). Using the primary sequence of the MAPK-docking site of MKP-5 (CADKISRRRLQQGKITV), we searched the EST cDNA bank for cDNAs encoding a protein containing a similar sequence. We identified one clone, that has a sequence CSKLMKRRLQQD. By performing the 5Ј-and 3Ј-rapid amplification of cDNA ends/PCR method in the human fetal brain cDNA library, we obtained a 2345-base pair cDNA fragment (Fig. 1A). The cDNA contains an open reading frame encoding The GFP-tagged mutants used were: 1, residues 1-471; 2, residues 1-406; 3 (⌬C), residues 1-317; 4 (⌬N), residue 300 -665; 5, residues 300 -471; 6, residues 300 -406; 7, residues 300 -372. a protein of 665 amino acids with a calculated molecular mass of 73 kDa (Fig. 1B). It resembles all the known MKPs. We then designated it MKP-7. In its N-terminal portion, there is a "Cdc25-like domain," in which the two regions showing amino acid similarity to the Cdc25 phosphatase and the MAPK-docking site (CSKLMKRRLQQD) exist (Fig. 1, A and B). In the middle portion, there appears to be a phosphatase catalytic domain, in which the active site sequence VLVHCLAGISRSA-TIAIAYIM (residues 240 -261) exists (Fig. 1, A and B) (16,17). The two amino acids, Cys-244 and Ser-231 in this motif, to-gether with Asp-203, are likely to participate in the catalytic mechanism of dual specificity phosphatase activity (39,40). MKP-7 is most similar to hVH5 (29) (Fig. 1B). Both molecules have a long C-terminal stretch which is not present in other MKPs. In the C-terminal stretch region (residues 300 -665), two portions show high similality to hVH5 in the primary sequence. We named them HC1 and HC2 (highly conserved regions 1 and 2) (HC1, residues 373-406; HC2, residues 596 -665) (Fig. 1, B and C). Near HC1, there is a PEST-like sequence in both MKP-7 and hVH5 (indicated as dotted box in Fig. 1B). Near HC2, a serine-rich sequence exists (residues 552-588) (Fig. 1C). The C-terminal stretches of MKP-7 and hVH5 including both HC1 and HC2 show no sequence similarity to other proteins.
Chromosomal Localization and Genomic Organization of the MKP-7 Gene-During the course of this study, a BAC clone RP11-253I19 (accession number AC007619), in which the cDNA of MKP-7 is included, was reported. Then, we determined the chromosomal location of the gene for MKP-7. The MKP-7 gene localizes to human chromosome 12 between the BAC clones, RPCI11-180M15 (accession number AC008115) and RP11-161A14 (accession number AC022276) (Fig. 2, left). The organization of the exon and intron of the gene was also determined (Fig. 2, right). It is similar to that of the hVH5 gene (data not shown).
Subcellular Localization of MKP-7-Different members of MKPs show different patterns of subcellular localization (16,17). We expressed Myc-tagged MKP-7 or Myc-tagged hVH5 in COS7 cells, C2C12 cells, and NIH3T3 cells. Indirect immunofluorescence with anti-Myc antibody showed that Myc-MKP-7 predominantly localized in the cytoplasm in these cells (Fig.  3A, upper), whereas Myc-hVH5 localized in both the cytoplasm and the nucleus (Fig. 3A, lower), the pattern of which being different depending on the cell type, as previously described (29,30). While Myc-hVH5 localized predominantly in the nucleus in COS7 cells, when expressed in C2C12 cells it showed pan-cellular distribution in some cells but nuclear localization in other cells. When expressed in NIH3T3 cells, Myc-hVH5 showed almost the same subcellular localization pattern as in C2C12 cells, although the NIH3T3 cells expressing Myc-hVH5 became small (Fig. 3A, lower). Essentially the same results were obtained using GFP-tagged MKP-7 and hVH5 (data not shown). In the C-terminal portion of MKP-7, there are two regions (HC1 and HC2) that show high homology with hVH5 in the primary sequence, as described above (Figs. 1A and 3D). In HC1, there exists a putative NES sequence. Then, the cytoplasmic localization of MKP-7 in cultured cells might be caused by this putative NES sequence (Figs. 1B and 3B). To test this idea, we constructed and expressed deletion mutants of MKP-7 and examined their subcellular localizations and the effect of leptomycin B (LMB) (a specific inhibitor of NES-mediated active nuclear export) on them in COS7 cells. While full-length MKP-7 and MKP-7⌬N (residues 300 -665), in which the Nterminal portion including the phosphatase domain is deleted (see Fig. 3D), localized predominantly in the cytoplasm, MKP-7⌬C (residues 1-317), in which the C-terminal stretch is deleted (see Fig. 3D), localized both in the cytoplasm and the nucleus (Fig. 3C). Furthermore, full-length MKP-7 and MKP-7⌬N became localized to the nucleus after treatment with LMB (Fig. 3C). These results indicate the existence of a functional NES in the C-terminal portion of MKP-7. To narrow down a region responsible for nuclear export of MKP-7, we made several additional deletion mutants of MKP-7 and examined their subcellular localization. The obtained results summarized in Fig. 3D indicate that HC1 is required for the cytoplasmic localization of MKP-7. Essentially the same results were obtained by using C2C12 or NIH3T3 cells (data not shown). Therefore, these results, taken together, suggest that the putative NES sequence in HC1 is functional and regulates the cytoplasmic localization of MKP-7 in cultured cells.

MKP-7 Binds
To and Inactivates p38␣ and JNK2 but Not ERK2-It has been shown that docking interactions play an important role in regulating the efficiency and specificity of the enzymatic reactions in the MAPK cascades (15,41,(43)(44)(45)(46)(47)(48)(49)(50)(51). We examined the docking ability of MKP-7 to the three major members of MAPKs (ERK2, p38␣, and JNK2) using a co-immunoprecipitation assay. p38␣ and JNK2 co-immunoprecipitated well with MKP-7, but ERK2 did not (Fig. 4A). As the MAPK-docking site and the catalytic domain are both located in the N-terminal portion (see Fig. 1), the N-terminal half of the molecule was supposed to be sufficient to bind to MAPKs. We then examined the docking ability of both the N-terminal half and the C-terminal half of MKP-7 (MKP-7⌬C and MKP-7⌬N; see 4B). But, MKP-7⌬N did not co-immunoprecipitate with p38␣ (Fig. 4C) or JNK2 (data not shown) even when the expression level of MKP-7⌬N was increased (Fig. 4C). These results have clearly shown that MKP-7 specifically binds to p38␣ and JNK2, not to ERK2, through the N-terminal portion of the molecule. Previously, we have identified the domain on MAPKs, termed the CD domain, which is commonly utilized for binding to MAPKKs, MAPKAPKs, and MKPs (50,51). To examine whether the CD domain is also utilized in the docking interactions with MKP-7, we used the CD domain-disrupted mutant of p38␣ (p38CDm), in which Asp-313, Asp-315, and Asp-316 in the CD domain were replaced by asparagines (50,51). As shown in Fig. 4D, while wild-type p38␣ co-precipitated well with MKP-7⌬C, p38CDm did not. Therefore, the CD domain is indispensable for the docking interaction between MKP-7 and p38 MAPK. Next, we examined the catalytic activity of MKP-7 toward MAPKs using a co-expression assay. Co-expression of full-length MKP-7 efficiently suppressed the activation of p38␣ or JNK2, but did not affect the activation of ERK2 (Fig. 4E). As the expression level of full-length MKP-7 in cultured cells was too low to examine its activity in higher expression levels (compare the lower panels of Fig. 4, E and F), we used MKP-7⌬C, in which both the MAPK-docking domain and the phosphatase catalytic domain are present (see Fig. 3D). MKP-7⌬C also efficiently inactivated p38␣ or JNK2, but not ERK2 (Fig.  4F). Furthermore, MKP-7⌬C inactivated p38␣ or JNK2 in a dose-dependent manner, but not ERK2 (Fig. 5A). These results demonstrate that MKP-7 binds to and efficiently inactivates p38␣ and JNK2, but not ERK2. Thus, we can conclude that there is good correlation between the docking ability of MKP-7 to MAPKs and its enzymatic activity toward them. It is known that when a Cys residue in the catalytic active site (VXVH-CXXGSRSXTXXXAYLM) is replaced by Ser, dual specificity phosphatases are converted to catalytically inactive forms. We tested whether MKP-7 was made catalytically inactive by such a mutation. As shown in Fig. 5B, the mutant form of MKP-7 (C244S) could not inactivate JNK2.
Catalytic Activity of MKP-7 in Vitro-To determine whether MKP-7 has a phosphatase activity, we examined the enzymatic activity of MKP-7 against pNPP, a well known artificial substrate of phosphatases. Myc-tagged MKP-7⌬C was prepared by immunoprecipitation from the cells transfected with SR␣-Myc-MKP-7⌬C. Myc-MKP-7⌬C hydrolyzed pNPP in a dose-dependent manner, and sodium vanadate, a potent inhibitor of tyrosine phosphatase, strongly inhibited the activity of MKP-7 toward pNPP (Fig. 6A). We then examined whether p38 is directly inactivated by MKP-7 in vitro. Activated p38␣ was prepared by incubating GST-p38␣ with His-MKK6 in the presence of ATP. The activated p38␣ was then incubated with immunoprecipitated Myc-MKP-7⌬C and the kinase activity of p38␣ was measured by using myeline basic protein as a substrate. As shown in Fig. 6B, p38␣ was inactivated by Myc-MKP-7⌬C. These results clearly demonstrate that MKP-7 has an intrinsic phosphatase activity and directly dephosphorylates p38␣. It has previously been reported that the binding of ERK to MKP-3/Pyst1/rVH6, Pyst2, or MKP-4/Pyst3 enhances the activity of these phosphatases (52). We then assayed the activity of Myc-MKP-7⌬C in the presence of GST-p38␣ (12 or 24 g). Incubation of Myc-MKP-7⌬C with GST-p38␣ stimulated the phosphatase activity of MKP-7⌬C toward pNPP significantly (Fig. 6C). However, incubation of Myc-MKP-7⌬C with GST-ERK2, which did not bind to MKP-7, did not stimulate the phosphatase activity markedly (Fig. 6C).
Interaction of MKPs with JNK2 and JNK3-There are three isoforms in the JNK/SAPK family, JNK1, -2, and -3 (9). Although they are similar to one another in the primary sequence, JNK3 is most distant. While JNK1 and -2 are ubiquitously expressed, the expression of JNK3 is restricted to brain. Then, we examined the docking ability of MKPs toward JNK2 and JNK3. As shown in Fig. 8C, CL100/MKP-1, MKP-5, and MKP-7⌬C bound to JNK2 and -3. The ability of MKP-5 to bind to JNK3 is weaker than that of CL100/MKP-5 or MKP-7⌬C. MKP-3 did not bind to either JNK2 or -3. Furthermore, JNK2 and -3 were efficiently inactivated by MKP-7⌬C (Figs. 4F and 8D). DISCUSSION We have isolated a cDNA clone encoding a novel MKP, termed MKP-7. MKP-7 possesses all the features of the dual specificity MAPK phosphatase. MKP-7 is most similar to hVH5 in the primary sequence. Both molecules have a C-terminal stretch, which is not present in other MKPs. In this portion, we noticed two regions (HC1 and HC2), whose sequences are highly homologous between these two molecules. There is an NES sequence in the N-terminal portion of HC1. This NES sequence regulates the cytoplasmic localization of MKP-7. Although the sequence of the NES portion of hVH5 is also very similar to that of MKP-7, subcellular localization of hVH5 in cultured cells differs from that of MKP-7. hVH5 is not exclusively cytoplasmic. HVH5 might have a nuclear localization signal or some binding proteins in the nucleus, or the activity of its NES might be weaker. HC2 is featured by clusters of positively charged amino acids. GFP-tagged HC2 fragment of MKP-7 (residues 596 -665) did not show any distinct localization but showed pan-cellular localization. 2 We may speculate, however, that the HC2 region is important for some common function, not defined yet, of the two molecules. MKP-7 has a PEST-like sequence in its C-terminal stretch, and hVH5 also has a PEST-like sequence, the sequence being not homologous to that of MKP-6, at the corresponding site. 2  HA-p38/anti-HA antibody complex was prepared by immunoprecipitation from the lysates of COS7 cells expressing HA-p38 MAPK by using anti-HA antibody. The immunoprecipitates (HA-p38/anti-HA antibody complex) was washed extensively (once with lysis buffer, three times with TBS, and then once with Nonidet P-40 buffer) before mixing with the lysates. B, MKP-7 inactivates p38␤, but not p38␥ or -␦. The experiments were performed as in Fig. 4E, using appropriate p38 isoforms. C, MKP-5 inactivates the p38␣ and ␤, but not p38␥ or -␦. The experiments were performed as in Fig. 4E, by using MKP-5. noticed a serine-rich region in the C-terminal stretch of MKP-7, which is not present in hVH5. However, the mouse ortholog of hVH5, M3/6, also has a serine-and glycine-rich region at the corresponding site (29). Puckered (a Drosophila phosphatase specific for JNK/SAPK) also has a serine-rich region in the C-terminal portion of the molecule (54). The PEST-like sequence and the serine-rich sequence might be involved in protein-protein interactions.
Incubation with the substrate p38␣ enhances the enzymatic activity of MKP-7, as was reported for MKP-3 (52). Recently, the activity of CL100/MKP-1 was also reported to be stimulated by p38␣ (55). Furthermore, we have previously shown that incubation with p38␣ enhances the catalytic activity of MKP-5, although the enhancement is modest (only 1.6-fold) (36). It is likely, therefore, that the binding of the substrate generally enhances the catalytic activity of MKPs. This mechanism might be important for regulation of the specificity of MKPs.
MKP-7 has a MAPK-docking site in the N-terminal portion (residues 49 -60). Based on the sequence characteristics of the docking site of MKPs, we propose that MKPs can be categorized into three groups. One group (the first group) includes CL100/ MKP-1, MKP-2, and PAC-1, and the second group MKP-3, MKP-4, Pyst-2, and B23, and the third group MKP-5, MKP-7, and hVH5 (Fig. 8A). The consensus sequence of the MAPKdocking site of each group is: XXRRRA(K/R) for the first group; XXXRRY(R/K)(Q)R/K(G) for the second group; X(K/R) RRLQQYK for the third group (where X is a hydrophobic residue, and Y is any residue except for K or R). MKPs acting on p38 MAPK (the first and third groups) have three consecutive positively charged amino acids in the MAPK-docking site, while the second group MKPs acting mainly on ERK have two (Fig. 9A). It has previously been shown that MAPKAPKs acting on p38 also have more consecutive positively charged amino acids in the MAPK-docking site than those acting on ERK (45,48,50,51). Notably, there are more negatively charged amino acids in the docking region of p38␣ than in that of ERK2 (50,51). In fact, we recently reported that the number of negatively charged amino acids in the docking groove of ERK2 and p38␣ regulates their docking specificity (51). Thus, the number of charged amino acids in the MAPK-docking site of MKPs might be important for the docking specificity and therefore the efficiency of the catalytic activity of MKPs. Not only the MAPKdocking sites but also other parts show high degree of sequence similarity within each group of MKPs. Thus, each group has its own substrate specificity, as outlined in Fig. 9A. However, the published results are not necessarily in complete agreement with this tentative classification of the substrate specificity of MKPs. Chu et al. (56) reported that while PAC-1 could inactivate JNK/SAPK in COS7 cells, it could not act efficiently on JNK/SAPK in NIH3T3 or HeLa cells. Such a difference might be caused by the difference in the expression levels of MKPs in each experiment, as also discussed by Chu et al. (56). CL100/ MKP-1 had been thought to be specific for ERK, but recent studies suggest that it is specific for p38 and JNK/SAPK, rather than ERK (see Ref. 16). Our results 2 also showed that CL100/MKP-1 binds strongly to p38 and JNK/SAPK, but not ERK2, and that CL100/MKP-1 inactivates p38 and JNK/SAPK more efficiently than ERK2. However, it is also clear that CL100/MKP-1 is able to inhibit ERK activation when expressed higher. It may be that every MKP can dephosphorylate to a greater or lesser extent every family of MAPKs in vitro. Then, quantitative comparison is required in future studies to determine strictly the substrate specificity of all the members of MKPs. The docking interaction may serve as an important factor for the specificity determination. Accumulating data including our present data indicate a good correlation between the docking ability and the enzymatic activity in actions of MKPs toward MAPKs (36,43,50). For example, MKP-7 binds to and inactivates p38␣ and JNK2, but does not bind to or inactivate ERK2. This is a basis for our tentative classification of MKPs according to the sequence features of the MAPKdocking sites.
Our results here demonstrate, for the first time, the substrate specificity of MKPs toward the isoforms of the p38 MAPK family. MKP-7 and MKP-5 selectively bind to and inactivate p38␣ and -␤, but not ␥ or ␦. CL100/MKP-1 also binds to p38␣ and -␤, but not ␥ or ␦ (Fig. 9B). A selectivity among the isoforms of the p38 MAPK family was already reported for the upstream activating kinases and the downstream substrates (57,58). MKK6 can activate all the isoforms of the p38 family, whereas MKK3 can activate only p38␣. As for the substrate specificity, p38␣ and -␤ can phosphorylate and activate MAP-KAPK2/3, but p38␥ or -␦ cannot. We also observed that p38␣ can bind to MAPKAPK-3, but p38␥ cannot. 2 According to our tentative classification of MKPs, the first and third groups of MKPs act on p38 (Fig. 8A), and our results show that their representatives, CL100/MKP-1 (the first group), MKP-5 and MKP-7 (the third group), cannot act on p38␥ or -␦. Then, inactivation of p38␥ and -␦ in cells might be achieved by other phosphatases, such as PP2C-type phosphatases or PTPs. The different mode of activation and inactivation of p38␥ and -␦ suggests that they may function in signaling pathways different from those regulated by other MAPKs. p38␥ is reported to have a motif sequence in its C-terminal tail that can associate with a PDZ domain-containing protein, ␣ 1 -syntrophin (42). We also found that p38␥ can bind to a PDZ-containing protein, hDLG, and phosphorylate it. 2 Then, p38␥ might regulate PDZ domain-containing proteins. It should be noted that p38␥ and -␦ are absent from Caenorhabditis elegans or Drosophila melanogaster. They might play some vertebrate specific roles. Addendum-During the course of the revision of this manuscript, the nucleotide sequence of human MKP-7 was deposited by another group.