The Mitogen-activated Protein Kinase Phosphatase-3 N-terminal Noncatalytic Region Is Responsible for Tight Substrate Binding and Enzymatic Specificity*

We have reported recently that the dual specificity mitogen-activated protein kinase phosphatase-3 (MKP-3) elicits highly selective inactivation of the extracellular signal-regulated kinase (ERK) class of mitogen-activated protein (MAP) kinases (Muda, M., Theodosiou, A., Rodrigues, N., Boschert, U., Camps, M., Gillieron, C., Davies, K., Ashworth, A., and Arkinstall, S. (1996) J. Biol. Chem. 271, 27205–27208). We now show that MKP-3 enzymatic specificity is paralleled by tight binding to both ERK1 and ERK2 while, in contrast, little or no interaction with either c-Jun N-terminal kinase/stress activated protein kinase (JNK/SAPK) or p38 MAP kinases was detected. Further study revealed that the N-terminal noncatalytic domain of MKP-3 (MKP-3ΔC) binds both ERK1 and ERK2, while the C-terminal MKP-3 catalytic core (MKP-3ΔN) fails to precipitate either of these MAP kinases. A chimera consisting of the N-terminal half of MKP-3 with the C-terminal catalytic core of M3-6 also bound tightly to ERK1 but not to JNK3/SAPKβ. Consistent with a role for N-terminal binding in determining MKP-3 specificity, at least 10-fold higher concentrations of purified MKP-3ΔN than full-length MKP-3 is required to inhibit ERK2 activity. In contrast, both MKP-3ΔN and full-length MKP-3 inactivate JNK/SAPK and p38 MAP kinases at similarly high concentrations. Also, a chimera of the M3-6 N terminus with the MKP-3 catalytic core which fails to bind ERK elicits non selective inactivation of ERK1 and JNK3/SAPKβ. Together, these observations suggest that the physiological specificity of MKP-3 for inactivation of ERK family MAP kinases reflects tight substrate binding by its N-terminal domain.

One important unanswered question is of the structural domains within MKP-3 and M3-6 which underlie their selectivity for inactivation of different MAP kinases. Dual specificity phosphatase family members are characterized by an extended PTP active site signature sequence localized to the C-terminal half of the molecule (41,42), as well as one or two short N-terminal CH2 domains displaying limited similarity to noncatalytic regions of the Cdc25 phosphatase (43,44). Despite low homology within CH2 domains, the N-terminal halves of dual specificity * 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  phosphatases are otherwise highly divergent (21,36,39) and the functional role of this region has not been identified. We report here experiments demonstrating that the N-terminal half of MKP-3 is responsible for tight binding to its substrates ERK1 and ERK2 but not JNK/SAPKs or p38. This interaction appears to be responsible for MKP-3 enzymatic selectivity for inactivation of ERK MAP kinases.

MKP-3 GST Fusion Proteins and Expression in
Escherichia coli-To produce the full-length GST-MKP3 fusion protein, the oligonucleotides GCCGGATCCATGATAGATACGCTCAGACC and GCCCTCGAGTCA-CGTAGATT GCAGTTGCAGGGAGTC were used to PCR amplify the complete MKP3 coding sequence from pMST-SM (36) and to introduce BamHI and XhoI sites (underlined) immediately flanking the start and stop codon. Upon restriction digestion with BamHI and XhoI the amplified cDNA was ligated to the corresponding sites in the pGEX4T3. To generate an enzymatically inactive GST-MKP3 fusion protein, the essential catalytic Cys 293 was changed to Ser by site-directed mutagenesis performed using the Transformer site-directed mutagenesis kit (CLON-TECH). To produce a GST fusion protein consisting of the MKP-3 C-terminal catalytic core (GST-MKP-3⌬N; amino acids 153-381) the first 151 amino acid were deleted by cutting full length pGEX4T3/ MKP-3 (see above) with XbaI and BamHI followed by religation. To generate the MKP-3 N terminus as a GST fusion protein (GST-MKP-3⌬C; amino acids 1-221), the plasmid pGEX4T3/MKP-3 was digested with StyI and XhoI and ligated to a double-stranded oligonucleotide obtained by annealing the two oligonucleotides CAAGGAACTAAGA-AGGGGTTCGCT and TCGAGCGAACCCCTTCTTAGTC. Fig. 1A illustrates schematically the MKP-3 amino acid sequences encoded within these fusion proteins. All MKP-3 and other GST-fusion proteins were expressed in E. coli by induction with 100 M isopropyl-1-thio-␤-Dgalactopyranoside and growth overnight at 20°C. Fusion proteins were purified with glutathione-Sepharose using standard techniques. ERK2 and MEK-1 EE were further purified following cleavage of the GSTfusion as described (39,45). MKP-3 phosphatase activity was measured in vitro using p-nitrophenyl phosphate as described elsewhere (39).
To test whether MKP-3 binding to ERK MAP kinases reflects interaction through N-and/or C-terminal regions, we next performed precipitations using either GST-MKP-3⌬C or GST-MKP-3⌬N (Fig. 1). Unlike the full-length MKP-3 molecule, GST-MKP-3⌬N containing the C-terminal catalytic core is unable to precipitate detectable levels of either p44-HA-ERK1 or p42-Myc-ERK2 (Fig. 2). GST-MKP-3⌬N was also unable to bind detectably to either p54-HA-JNK3/SAPK␤ or HA-p38 MAP kinases (Fig. 2). The inability of GST-MKP-3⌬N to bind ERK MAP kinases is unlikely to reflect misfolding as this fusion protein displayed similar enzymatic activity as fulllength GST-MKP-3 when using the artificial substrate p-nitrophenyl phosphate (Fig. 3). Somewhat surprisingly, and in contrast to the C-terminal catalytic core of MKP-3, the N-terminal GST-MKP-3⌬C was found to bind tightly and precipitate both p44-HA-ERK1 and p42-Myc-ERK2 at least as well as the full length GST-MKP-3 molecule (Fig. 2). GST-MKP-3⌬C was also able to bind weakly to HA-p38 MAP kinase while this fusion was completely ineffective at binding p54-HA-JNK3/SAPK␤ (Fig. 2). Overall, the rank-order for MKP-3 N-terminal binding and precipitation of MAP kinases is p44-ERK1 ϭ p42-ERK2 Ͼ Ͼ p38 Ͼ p54SAPK␤. Together, these observations indicate that the N-terminal half of MKP-3 is alone able to bind tightly to ERK family MAP kinases and probably accounts, at least in part, for the precipitation observed with the full-length molecule.
Studies in yeast and mammalian cells have demonstrated that elements within MAP kinase cascades as well as other signal transduction pathways form multicomponent complexes through tight association with various anchoring proteins. For instance, the yeast pheromone mating response MAP kinases FUS3 or KSS1, as well as the upstream kinases STE7 and STE11 and G protein ␤-subunit STE4 all bind simultaneously to the scaffold protein STE5, while the mammalian cell protein AKAP79 maintains protein kinase A and C, as well as the serine threonine protein phosphatase calcineurin in close proximity to their target substrates (46,47). To test whether MKP-3 is able to bind to ERK MAP kinases directly and independently of additional cellular proteins, we next tested binding to p42 ERK2 protein purified following its expression in E. coli and cleavage from its GST fusion. Fig. 4 shows that, as observed with cell lysates, GST-MKP-3, GST-MKP-3 C293S, and GST-MKP-3⌬C immobilized on glutathione-Sepharose beads are all able to bind and precipitate purified p42 ERK2 from lysis buffer, while GST-MKP-3 ⌬N binds only very weakly to this  (45) was incubated with glutathione-Sepharose beads prebound either to GST alone or each of the GST-MKP-3 fusion proteins as indicated above each lane. Western analysis of washed beads was performed using antibody 122 specific for ERK2 with goat anti-rabbit monoclonal antibody horseradish peroxidase conjugate and chemiluminescence. Purified ERK2 protein (ϳ5 ng) was loaded as a positive control.

MAP kinase. This result indicates that binding between the MKP-3 N-terminal and ERK MAP kinases is direct and independent of additional cellular proteins.
To investigate the relationship between MKP-3 binding and MAP kinase inactivation we next incubated purified ERK2 activated by the constitutively active MAP kinase kinase MEK-1 mutant S218E S222E (39,45) together with increasing concentrations of either full-length GST-MKP-3 or GST-MKP-3⌬N. While 0.5 g of MKP-3 completely blocked MEK-1-dependent ERK2 activation, 10-fold higher concentrations of MKP-3⌬N were required to elicit similar inhibition of this MAP kinase (Fig. 5). In contrast to potent and selective inactivation of ERK2 by MKP-3, similarly high concentrations of MKP-3 and MKP-3⌬N were required to inhibit either JNK2/SAPK␣ or JNK3/SAPK␤ activated in vitro by purified SEK1 or to reverse p38 MAP kinase activation by MKK6 (Fig. 5). This suggests that, in contrast to full-length MKP-3, MKP-3⌬N displays little selectivity for inactivating different MAP kinase isoforms. These experiments demonstrate a clear correlation between ERK binding to full-length MKP-3 and specificity for ERK MAP kinase inactivation.
To confirm that tight binding between the MKP-3 N terminus and ERK also parallels specific inactivation of this class of MAP kinase within intact cells, immune complex assays were performed on p44-HA-ERK1 and p54-HA-JNK3/SAPK␤ co-expressed with the same MKP-3 and M3-6 constructs. As we have reported previously (40) MKP-3 co-expression abolishes EGFstimulated activation of p44-HA-ERK1 (Fig. 8). In contrast, the catalytically inactive mutants MKP-3 (C293S) and MKP-3⌬C fail to inhibit, and rather augment by 2-3-fold the activation state of p44-HA-ERK1 (Fig. 8). Also as seen before (40), coexpression of M3-6 is completely unable to suppress EGFstimulated p44-HA-ERK1, while anisomycin-dependent stimulation of p54-HA-JNK3/SAPK␤ is inhibited by Ͼ90% (Fig. 8). It is of note that a truncated version of this dual specificity phosphatase, M3-6⌬C (Fig. 6A), elicits indistinguishable selective inhibition of p54-HA-JNK3/SAPK␤ (not shown). Dramatically, and in contrast to tight binding of CB16 to p44-HA-ERK1 (Fig. 7), CB16 is totally ineffective in its ability to inhibit EGF-stimulated activation of this MAP kinase (Fig. 8). Rather, as observed with the enzymatically inactive mutants MKP-3 C293S and MKP-3⌬C, co-expression of CB16 facilitates p44-HA-ERK1 activation state by approximately 2-fold (Fig. 8). Importantly, CB16 appears to be an active enzyme as its coexpression with p54-HA-JNK3/SAPK␤ results in an ϳ80% inhibition of anisomycin-stimulated activity (Fig. 8). The chimera CB16 therefore retains the enzymatic specificity of its M3-6 catalytic core despite tight interaction with p44-HA-ERK1. In contrast to these observations, the chimera CA8 possessing the catalytic core of MKP-3 and which fails to bind to p44-HA-ERK1 (Fig. 7), mediates complete inactivation of EGF-stimulated activation of this MAP kinase (Fig. 8). Notably however, and as observed in vitro using GST-MKP-3⌬N (Fig. 5), CA8 differs from MKP-3 insofar that it displays no selectivity for inactivation of p44-HA-ERK1 when compared with p54-HA-JNK3/SAPK␤ (Fig. 9). This indicates that binding of the MKP-3 N terminus to ERK isoforms within intact cells also Plasmid concentrations were kept constant using the empty pMT-SM vector. After culture for 40 h, cells were incubated for 2 h in serum-free medium and then stimulated with either 10 nM EGF (p44-HA-ERK1) or 10 g/ml anisomycin (p54-HA-JNK3/SAPK␤). Following cell lysis, MAP kinase immunoprecipitation and immunecomplex assays were performed as described in Fig. 8. Upper panels in A-D show autoradiograms of phosphorylated MBP (p44-HA-ERK1) or c-Jun (p54-HA-JNK3/SAPK␤) upon co-expression with the indicated amounts of MKP-3 (A and B) or CA8 (C and D) plasmid. Substrate bands were excised for counting by scintillation spectrometry and calculation of relative kinase activity which is indicated numerically below each lane. Western blot of immunoprecipitated (IP) p44-HA-ERK1 and p54-HA-JNK3/SAPK␤ used for immune complex assays is shown below each autoradiogram. Detection is as described in Fig. 8. Data are representative of three identical experiments. plays a critical role in determining specificity for inactivation of this class of MAP kinase.
The crystal structure of several PTP family members (Yersinina PTPase, PTP1B, PTP␣) including the dual-specificity phosphatase VHR have now been solved and, despite limited primary amino acid sequence homology, all have been shown to possess a very similar three-dimensional structure within their catalytic domains (42, 48 -52). Based on the crystal structure of VHR, reliable secondary structure prediction of rVH6 (identical to MKP-3) is possible and demonstrates that the C-terminal domain (amino acids 134 -381) is alone able to form the catalytic core of this dual specificity phosphatase (41). We have obtained essentially identical results modeling the primary amino acids 209 -348 of MKP-3 on the crystal structure of VHR using the SWISS MODEL program (53) (not shown). These secondary structure predictions are also consistent with experimental observations that the MKP-3 C-terminal half displays indistinguishable enzymatic activity to the full-length molecule when using the artificial substrate p-nitrophenyl phosphate (Fig. 3). What then is the function of the N-terminal noncatalytic halves of this gene family? Experiments described in this report demonstrate that one role for the N terminus of MKP-3 is likely to be tight binding to its substrate MAP kinases p44-ERK1 and p42-ERK2. Tight substrate binding may ensure highly restricted enzyme action within the environment of an intact cell. One untested hypothesis is that the highly divergent N termini of dual specificity phosphatase family members (39) bind tightly to a distinct set of target substrates. Interestingly, examination of the gene structures of CL100 and PAC1 indicates that their N termini are encoded by exons 1 and 2 and may have distinct evolutionary origins to the C-terminal catalytic core (44,54,55). Diverse substrate binding capabilities for this gene family may therefore have arisen through convergence of gene structures encoding a limited set of catalytic cores with a more varied range of N-teminal targeting sequences. The emergence of gene structures for other dual specificity phosphatase family members will help clarify this issue.
In summary, these observations demonstrate that MKP-3 binds directly the MAP kinases ERK1 and ERK2 through its non-catalytic N terminus. MKP-3 binding displays the rank order p44-ERK1 ϭ p42-ERK2 Ͼ Ͼ p38 Ͼ p54SAPK␤ which parallels its enzymatic specificity for MAP kinase inactivation (38,40). Experiments described in this report using the MKP-3⌬N catalytic core as well as with chimeras of MKP-3 and M3-6 indicate that this binding may underlie MKP-3 selectivity for ERK family MAP kinases. Restricted substrate binding by the structurally diverse N termini of different dual-specificity phosphatases may provide a general mechanism ensuring enzymatic selectivity for this gene family.