Mitogen-activated Protein Kinase (MAPK) Phosphatase 3-mediated Cross-talk between MAPKs ERK2 and p38α*

MAPK phosphatase 3 (MKP3) is highly specific for ERK1/2 inactivation via dephosphorylation of both phosphotyrosine and phosphothreonine critical for enzymatic activation. Here, we show that MKP3 is able to effectively dephosphorylate the phosphotyrosine, but not phosphothreonine, in the activation loop of p38α in vitro and in intact cells. The catalytic constant of the MKP3 reaction for p38α is comparable with that for ERK2. Remarkably, MKP3, ERK2, and phosphorylated p38α can form a stable ternary complex in solution, and the phosphatase activity of MKP3 toward p38α substrate is allosterically regulated by ERK2-MKP3 interaction. This suggests that MKP3 not only controls the activities of ERK2 and p38α but also mediates cross-talk between these two MAPK pathways. The crystal structure of bisphosphorylated p38α has been determined at 2.1 Å resolution. Comparisons between the phosphorylated MAPK structures reveal the molecular basis of MKP3 substrate specificity.

MAPK pathways convert different extracellular stimuli into specific cellular responses and mediate various physiological processes, including cellular proliferation, apoptosis, differentiation, and stress responses. There are three major mammalian MAPK subfamilies: ERK, JNK and p38 (1)(2)(3)(4)(5)(6). The ERKs are typically activated by growth factors and phorbol ester, whereas JNK and p38 MAPKs are primarily activated by cytokines and environmental stress. MAPK activity is tightly controlled by phosphorylation and dephosphorylation. Full activation of the MAPKs requires phosphorylation on both threonine and tyrosine residues in the TXY motif by their specific upstream dual specificity kinases (MAPK kinases). After activation, each MAPK phosphorylates a distinct spectrum of substrates, which include key regulatory enzymes, cytoskeletal proteins, nuclear receptors, regulators of apoptosis, and many transcription factors.
The regulated dephosphorylation of MAPKs plays a key role in determining the magnitude and duration of kinase activation and hence the physiological outcome of signaling. MAPK phos-phatases (MKPs) 2 are important negative regulators of MAPK signaling. MKPs belong to a family of dual specificity phosphatases and specifically dephosphorylate both threonine and tyrosine residues in the activation loop of MAPKs. In mammalian cells, there are 10 distinct catalytically active MKPs. Several of these MKPs display distinct in vivo substrate preferences for the various MAPKs. For example, it has been shown that MKP3 selectively targets ERK, whereas MKP5 shows a preference for JNK and p38 (7)(8)(9). Therefore, there must be a fine regulatory mechanism for MAPK-MKP recognition.
MKP3 is localized predominantly in the cytoplasm and is highly specific for ERK1/2 inactivation (10 -13). MKP3 contains a conserved catalytic domain in its C terminus and a less conserved MAPK binding domain at its N terminus (supplemental Fig. S1). The inactive ERK1/2, but not p38 and JNK MAPKs, can form a tight complex with MKP3 via the interaction between the N-terminal domain of MKP3 and the C-lobe of ERK2. Unlike other PTPases and constitutively active MKPs, MKP3 exhibits very low activity toward small phosphoesters, such as para-nitrophenyl phosphate (pNPP), and binding of ERK to the N-terminal domain of MKP3 causes its catalytic activation in vitro. Neither JNK nor p38 MAPKs stimulated significant MKP3 catalytic activity over basal levels (14 -17). Thus, it has been suggested that the noncatalytic N-terminal domain of MKP3 plays an important role in controlling its substrate specificity and that the ERK-induced MKP3 activation may be responsible for the selective inactivation of ERK MAPKs by MKP3 (8). Although studies with small molecule substrates have yielded useful information about MKP3 catalysis, our knowledge of its physiological functions is somewhat less complete. Two important questions concerning the dephosphorylation mechanism of its authentic substrate remain unresolved. How does MKP3 differentially recognize and bind to MAPK isoforms? How does complex formation lead to conformational change within the catalytic domain of MKP3?
dephosphorylate the phosphotyrosine, but not phosphothreonine, in the activation loop of p38␣ MAPK in vitro and in intact cells. The catalytic efficiency of MKP3 reaction for p38␣ substrate is comparable with that for ERK2 as well as other tyrosine-specific phosphatases, including HePTP, PTP-SL, and STEP. Remarkably, binding of ERK2 to the noncatalytic domain of MKP3 increases the catalytic activity of MKP3 toward phosphorylated p38␣. These results suggest that phosphorylated p38␣ is a physiological substrate for MKP3. To gain further insight into the phosphorylation-dependent binding of p38␣ to MKP3, we determined the crystal structure of bisphosphorylated, active p38␣ at 2.1 Å resolution. Comparisons between the phosphorylated kinase structures reveal that although p38␣ and p38␥ have greater sequence similarity, the structure of phosphorylated p38␥ is more similar to ERK2 than to p38␣. The conformation of the activation lip of active p38␣ is distinct from the phosphorylated ERK2 and p38␥. This structural information reveals new insights into the mechanism of p38␣ activation by dual phosphorylation and the molecular basis of MKP3 substrate specificity.
Crystallization of Phosphorylated p38␣ and Data Collection-Crystals of active p38␣ were obtained at 20°C in 5 days using the hanging drop technique containing equal volumes of protein solution and reservoir buffer (20 -25% (w/v) polyethylene glycol 3350, 0.2 M sodium citrate, 100 mM HEPES, pH 7.5). Crystals were transferred into the cryoprotectant containing reservoir buffer with 20% glycerol before being "flash-cooled" at 100 K in a stream of nitrogen gas. X-ray data were collected on a home laboratory x-ray system (Rigaku FR-E rotating anode x-ray generator with an R-AXIS IVϩϩ image plate area detector). Data were integrated and scaled using HKL2000 (18). The crystals belonged to the orthorhombic space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 47.3 Å, b ϭ 52.3 Å, c ϭ 168.1 Å, and ␣ ϭ ␤ ϭ ␥ ϭ 90°. There was one molecule in the asymmetric unit.
Structure Determination and Refinement-The structure was solved by molecular replacement using Phaser (19) with the search model of inactive p38␣ (Protein Data Bank code 1p38). Multiple rounds of the model building and the refinement were carried out with Coot (20) and PHENIX (21). The final R free was 18.2%, and the R factor was 23.9%. The final model spans nearly the entire molecule, comprising residues 1-354 and two covalently bound phosphate ions, with residues 355-360 missing because of disordered loops that are not visible in any electron density map. PROCHECK (22) was used to analyze the model stereochemistry. All of the residues were in the most favored and additional allowed regions of the Ramachandran plot. Data collection and structure refinement statistics are summarized in supplemental Table S1.
Enzyme Assays for Protein Phosphatases-Kinetic parameters for the dephosphorylation of the phosphorylated p38␣ and ERK2 were determined using a continuous spectrophotometric assay (23)(24)(25). This assay incorporates a coupled enzyme system, which uses purine nucleoside phosphorylase and its chromogenic substrate 7-methyl-6-thioguanosine (MESG) for the quantification of inorganic phosphate produced in the phosphatase reaction (26). All experiments were carried out at 25°C in 1.8 ml of reaction mixture (pH 7.0 buffer containing 50 mM MOPS, 100 mM NaCl, 0.1 mM EDTA, 100 M MESG, 0.1 mg/ml purine nucleoside phosphorylase). The reaction was initiated by adding a catalytic amount of phosphates to a reaction mixture, and the dephosphorylation reaction of bisphosphorylated ERK2 and p38␣ was detected by monitoring the real-time phosphate release at 360 nm. The time courses of absorbance change at 360 nm resulting from the MESG/phosphorylase coupling reaction were recorded on a Lambda 14 PerkinElmer spectrophotometer equipped with a magnetic stirrer in the cuvette holder. Initial rates were determined from the linear slope of progress curves obtained, and the experimental data were analyzed using a nonlinear regression analysis program. The change in absorbance was due to the conversion of MESG to 7-methyl-6-thioguanine in the presence of inorganic phosphate released from the dephosphorylation of ERK2 and p38␣ by MKP3. Quantitation of phosphate release was determined using an extinction coefficient of 11,200 M Ϫ1 cm Ϫ1 for the phosphate-dependent reaction at 360 nm at pH 7.0 (27). The stoichiometry of MAPK phosphorylation was then calculated by dividing the relative concentration of the phosphate by that of ERK2 or p38␣. The concentration of MESG was determined at 331 nm, using a molar extinction coefficient of 32,000 M Ϫ1 cm Ϫ1 .
Gel Filtration Analysis-Size exclusion chromatography was performed to determine the apparent molecular weight of proteins. 0.1 ml of each sample was loaded onto a Superdex 200 column, preequilibrated in the buffer (50 mM MOPS buffer, pH 7.0, 100 mM NaCl, 0.1 mM EDTA), and run at 4°C using a flow rate of 0.5 ml/min. The column was calibrated using the molecular weight markers.
Plasmids, Cell Culture, and Transfections-For expression in HEK293T cells, cDNAs coding the sequence of wild-type MKP3, mutant MKP3C293S, and HePTP were subcloned into the pCMV-Myc (Clontech) vector. Plasmid pcDL-SR␣ 296-Myc containing wild type MKP5 was kindly provided by Dr. Eisuke Nishida. The coding region of wild-type p38␣ was subcloned into pCS2ϩFLAG vector. HEK293T cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum and penicillin, streptomycin, and glutamine. HEK293T cells were transiently transfected with a total of 2 g of DNA using VigoFect transfection reagent (Vigorous Biotechnology). After 48 h of transfection, cells were treated for 30 min with 0.5 M sorbitol.
Isothermal Titration Calorimetry (ITC)-ITC studies were performed on MacroCal VP-ITC (Northampton, MA). 1.8 mM KIM peptide dissolved in the reaction buffer containing 50 mM HEPES, pH 7.5, and 150 mM NaCl was titrated against 130 M ERK2, p38␣, or p38␣/pTpY proteins in the same buffer. The data were fitted using the software Origin 7.0.

RESULTS
Recombinant Active p38␣ Is Rapidly and Specifically Dephosphorylated by MKP3-To understand the molecular basis of the substrate specificity of MKP3, we prepared milligram quantities of bisphosphorylated ERK2 and p38␣ (ERK2/pTpY and p38␣/pTpY) (25,28). This p38␣/pTpY preparation corresponds to the highest reported activity against ATF2⌬109 for this enzyme (29). The values of k cat and K m were determined to be 4.14 Ϯ 0.06 s Ϫ1 and 1.81 Ϯ 0.09 M, respectively (supplemental Fig. S2). We then examined the dephosphorylation kinetics and substrate specificity of MKP3 for ERK2/pTpY and p38␣/pTpY using a spectrophotometric method for the quantitation of inorganic phosphate (25). Fig. 1A shows typical progress curves of ERK2/pTpY and p38␣/pTpY dephosphorylation by MKP3. Both MAPKs can be dephosphorylated rapidly by MKP3. When all the substrate has been converted to product, the phosphorylation stoichiometry of ERK2 was determined to be close to 2 mol of phosphate/mol of ERK2. In comparison, the dephosphorylation rate of p38␣/pTpY was about 3-fold slower than that of ERK2/pTpY, and only one equivalent of inorganic phosphates was released upon treatment of p38␣/ pTpY with MKP3. The phosphorylation state of p38␣ was further determined by the amount of inorganic phosphate released from p38␣ upon treatment by various phosphatases (Fig. 1B). The addition of MKP3 (400 nM) to a reaction mixture contain- ing 0.75 M p38␣/pTpY yielded stoichiometric amounts of phosphate released from p38␣. The subsequent addition of HePTP to the reaction mixture did not cause any absorbance change at 360 nm, whereas further addition of MKP5 produced another equivalent of inorganic phosphatase, indicating that the purified p38␣ was phosphorylated equally at Thr 180 and Tyr 182 (to 1:1 mol of phosphate/mol of protein).
We also quantified the variously phosphorylated forms of p38␣ by performing Western blot analysis using anti-bisphosphorylated p38 and anti-phosphotyrosine antibodies (Fig. 1C). When probed with an anti-Tyr(P) antibody, only p38␣/pTpY and p38␣/pY showed immunoreactivity (top), and when probed with an anti-bisphosphorylated p38 antibody, only p38␣/pTpY and p38␣/pT displayed immunoreactivity (middle). No measurable immunoreactivity was apparent with the p38␣/pT sample toward anti-Tyr(P) antibody, indicating that dephosphorylation by MKP3 was complete. Thus, our results showed that MKP3 can efficiently hydrolyze Tyr(P) in p38␣/ pTpY, and the dephosphorylation rate of p38␣/pTpY by MKP3 is much faster than that reported by Groom et al. (10). One possible explanation for this discrepancy between the two studies might be that the activated Xenopus p38␣ used in their study was not fully phosphorylated at Tyr 183 (see Fig. 5B of Ref. 10) because it is generally difficult to obtain high phosphorylation stoichiometry with the wild type MAPK by in vitro activation (30). A second intriguing possibility could be due to different origins of p38␣ proteins because the Xenopus p38␣ homologue was used by Groom et al. (10), and ours is of a mammalian origin. MKP3 might exhibit different substrate specificity for the recombinant p38 proteins. For example, we have looked at the ability of MKP3 to dephosphorylate bisphosphorylated p38␥ and found that MKP3 is unable to dephosphorylate this isoform of p38 MAPK under identical experimental conditions (Fig. 1D).
The N-terminal Region Is Not a Major Determinant for Substrate Specificity of MKP3-MKP3 is a rather inefficient enzyme toward small phosphoesters, such as pNPP. However, in the presence of ERK, the catalytic activity of MKP3 increases Ͼ35fold toward pNPP (14,31). The activation of MKP3 is thought to be achieved via the interaction between the N-terminal domain of MKP3 and the non-catalytic regions of ERK2. Thus, MKP3 substrate specificity is linked to the ability of substrate to induce productive orientation in the active site. To evaluate the role of the N-terminal domain of MKP3 in conferring tyrosine specificity to p38␣, we examined the dephosphorylation of p38␣/pTpY by MKP3/⌬N151 (an N-terminally truncated MKP3, lacking the first 1-151 amino acids) (32). Fig. 2A shows the time dependence of P i release from p38␣/pTpY for sequen- tial additions of different phosphatase into the coupled reaction system. After completion of MKP3/⌬N151-catalyzed p38␣ reaction, the addition of HePTP to the reaction mixture did not cause detectable P i release, whereas the further addition of MKP5 produced another equivalent of inorganic phosphate. This result indicates that MKP3/⌬N151 still specifically dephosphorylated the phosphotyrosine of p38␣/pTpY, and the N-terminal domain of MKP3 is not a major determinant for the tyrosine specificity toward p38␣.
Next, we tested whether the ERK2-induced MKP3 activation is unique for small molecule aryl phosphates. In the initial experiments, the amount of ERK2 required to activate MKP3 maximally was determined by measuring the initial velocities at increasing concentrations of ERK2 and a fixed concentration of phosphatase substrate pNPP. Increasing amounts of ERK2 were added to a solution in which the concentration of MKP3 was at least 10 times higher than the dissociation constant, and the results were plotted to give the stoichiometry. Consistent with previous observations (33,34), we found that the activation of MKP3 by ERK2 is saturable, and maximal activation resulted from an MKP3/ERK2 ratio of 1:1, suggesting a 1:1 binding stoichiometry between MKP3 and ERK2 (Fig. 2B). We then examined the effects of ERK2 on the MKP3-catalyzed dephosphorylation of p38␣. Fig. 2C shows typical progress curves of the MKP3-catalyzed reaction in the absence and presence of ERK2. It can be seen from this figure that binding of ERK2 to MKP3 significantly increases the rate of MKP3-catalyzed p38␣/pTpY dephosphorylation, whereas at the same time it still renders MKP3 incapable of dephosphorylating Thr(P) in p38␣/pTpY. These results suggest that ERK2 and activated p38␣ are capable of binding to MKP3 simultaneously, and MKP3 undergoes catalytic activation toward other protein substrate (phosphorylated p38␣) when bound to ERK2. The various phosphorylation states of p38␣ were also confirmed by Western blot analysis as described above (Fig. 2D).
Finally, we evaluated the effect of the unphosphorylated ERK2 on the MKP3-catalyzed dephosphorylation of ERK2/ pTpY because ERK2 is a product of the MKP3 reaction and has high affinity for MKP3. The kinetics of the MKP3-catalyzed dephosphorylation of ERK2/pTpY were examined in the presence of various fixed concentrations of ERK2. The initial rates were measured as a function of ERK2/pTpY concentration keeping the ERK2 concentration constant (0 -0.4 M) (supplemental Fig. S3). Under these conditions, the double reciprocal plots yielded a simple competitive pattern with a K i value of 0.103 Ϯ 0.018 M. Thus, ERK2 acts as a competitive inhibitor of the ERK2/pTpY dephosphorylation by MKP3.
Interactions of MKP3 with ERK2 and p38␣ in Vitro and in Vivo-Given that the dephosphorylation of p38␣/pTpY by MKP3 can be regulated by ERK2, we devised an in vitro interaction assay employing size exclusion chromatography to test whether ERK2, MKP3, and p38␣/pTpY could form a stable ternary complex in solution. First, we studied the interaction between only two of the three proteins. Due to the transient nature of the enzyme-substrate complex, it is difficult to study the binding interaction between wild-type MKP3 and bisphosphorylated p38␣/pTpY (Fig. 3A). Consequently, we employed catalytically inactive MKP3C293S to capture the enzyme-sub-strate complex, as judged by comparing gel filtration profiles of the individual proteins with their equimolar mixture (Fig. 3B). When an equimolar mixture of p38␣/pTpY and ERK2 was sized on gel filtration, only ERK2 showed a slight change in elution volume, indicative of an extremely weak interaction between those two proteins (panel 4). Both ERK2 and p38␣/pTpY, but not p38␣ (data not shown), were able to associate with MKP3 to form a stable heterodimer (panels 5 and 6). To demonstrate the formation of ERK2-MKP3-p38␣/pTpY heterotrimer in solution, we next incubated the three proteins in an equimolar ratio and performed gel filtration analysis. Gel filtration profiles showed that purified ERK2, MKP3, and p38␣/pTpY co-migrated to earlier fractions than MKP3-p38␣/pTpY and MKP3-ERK2 heterodimers (panel 7). The apparent molecular mass of the ERK2-MKP3-p38␣/pTpY complex appeared to be ϳ140 kDa, as assessed from the elution volume of the protein complex peak compared with the protein standards, indicating that these three proteins formed a stable ternary complex in solution. For comparison purposes, we also examined if an ERK2-MKP3-ERK2/pTpY ternary complex could also be formed under the same experimental conditions. In the absence and presence of ERK2, MKP3-ERK2/pTpY binary complex eluted with a similar retention time, and no ERK2-MKP3-ERK2/pTpY ternary complex was observed (supplemental Fig. S4).
In order to study the effects of MKP3 in vivo, 293T cells overexpressing the phosphatases and p38␣ were stimulated with sorbitol to induce p38 phosphorylation and activation, followed by immunoprecipitation with the anti-FLAG antibody; the phosphorylation state of FLAG-p38␣ was detected by immunoblot using bisphosphorylated p38 and anti-phospho-Tyr antibodies (Fig. 3C, first and second panels). The results showed that like HePTP and MKP5, MKP3 can effectively diminish the level of phosphotyrosine of FLAG-p38␣ in intact cells. We further examined the association of MKP3 with its substrate p38␣ by immunoblot using anti-Myc antibody (Fig.  3C, third panel, lane 6 versus lane 7). Consistent with our in vitro data, wild-type MKP3 failed to bind to p38␣, but the catalytically inactive MKP3 mutant C293S physically interacted with bisphosphorylated p38␣, indicating that the phosphorylation at Tyr 182 is essential for MKP3-p38␣ binding. In line with the fact that association of p38␣ with HePTP and MKP5 is not dependent upon phosphorylation of p38␣, these two phosphatases displayed much higher immuoreactivity than that of MKP3C293S (Fig. 3C, third panel, lane 7 versus lanes 8 and 9).
ERK2 can form a tight physical complex with MKP3 via the KIM motif in the N-terminal domain of MKP3 and the docking sites on ERK2. Thus, it would be of interest to know whether p38␣ binds to MKP3 through the same mechanism. To address this issue, we analyzed the interaction of ERK2 and p38␣ MAPKs with the N-terminal domain of MKP3 (residues 1-154) using gel filtration assays. It can be seen from Fig. 4, A-C, that in contrast to ERK2, neither the unphosphorylated nor the phosphorylated p38␣ was able to associate with MKP3(1-154) to form a stable heterodimer. To study the binding of the KIM sequence to ERK2 and p38␣ further, we synthesized peptide corresponding to the KIM sequences of MKP3 and measured the binding affinities of ERK2 and p38␣ with the KIM peptide by ITC. The KIM peptide binds to ERK2 with a dissociation constant (K d ) of 65.8 Ϯ 15.4 M, but no detectable interaction between the KIM peptide and p38␣ was observed under the same conditions (Fig. 4, D-F). These results support the notion that the KIM sequence in MKP3 is not responsible for p38␣/ pTpY binding.
Kinetic Characterization of MKP3-catalyzed Reactions-To gain further insight into the mechanism of ERK2-induced MKP3 activation, we carried out a detailed kinetic study of MKP3-catalyzed reaction at pH 7.0 and 25°C. The dephosphorylation of ERK2 and p38␣ catalyzed by MKP3 obeyed classical Michaelis-Menten kinetics (supplemental Fig. S5A). We were able to determine the kinetic parameters, k cat and K m , for MKP3 with both ERK2 and p38␣ as a substrate ( Table 1). As shown in Table 1, the k cat /K m value for the MKP3-catalyzed ERK2 dephosphorylation was determined to be 7.3 ϫ 10 6 s Ϫ1 M Ϫ1 , which is similar to those obtained from earlier studies (24). In comparison, the k cat of MKP3-catalyzed p38␣/pTpY dephosphorylation is only 3-fold lower than that of ERK2/pTpY, whereas the K m value for p38␣/pTpY substrate is about 8-fold higher than that for ERK2/pTpY. Binding of ERK2 to MKP3 results in a 3-fold increase in k cat for p38␣/pTpY, with a modest effect on K m . To quantitatively assess the contribution of the N-terminal domain to the MKP3-catalyzed p38␣ dephosphorylation, we also determined the kinetic parameters of MKP3/ ⌬N151 toward p38␣ substrate (supplemental Fig. S5B). The k cat value for MKP3/⌬N151 was comparable with that of full-length MKP3. However, its K m value was more than 20-fold higher than that of full-length MKP3, suggesting that the major function of the MKP3 N-terminal domain is to increase the "effective concentration" of the pTXpY motif in the vicinity of the FIGURE 3. Direct binding between p38␣/pTpY and MKP3. A, gel filtration analysis of MKP3 and p38␣/pTpY alone or an equimolar mixture of MKP3 and p38␣/pTpY, in which p38␣/pTpY was dephosphorylated to be p38␣/pT by MKP3 during incubation. B, gel filtration analysis of the p38␣/pTpY, mutant MKP3 (C293S), and ERK2 alone or an equimolar mixture of two of the three proteins as indicated and an equimolar mixture of the three proteins. Gel filtration chromatography experiments were performed with a Superdex 200 gel filtration column mounted on an AKTA FPLC system (Amersham Biosciences). All proteins were in 50 mM MOPS buffer, pH 7.0, 100 mM NaCl, 0.1 mM EDTA. C, in vivo association and dephosphorylation effects of MKP3, HePTP, and MKP5 on overexpressed p38␣. 293T cells were transfected with expression vectors for Myc-phosphatase (MKP3, MKP3C293S, HePTP, and MKP5) and FLAG-p38␣. After 48 h, cells were treated for 30 min with 0.5 M sorbitol. Complexes were precipitated with anti-FLAG followed by Western blot (WB) with the indicated antibodies. IP, immunoprecipitation; TCL, total cell lysate. The heavy chain of the anti-FLAG antibody, which was used for the immunoprecipitation, is marked by an asterisk.
active site of MKP3 for p38␣/pTpY dephosphorylation. In contrast to the previous conclusion obtained with small molecule substrates (17,35), our results suggest that the catalytic domain of MKP3 alone is in a catalytically active conformation, providing the first experimental evidence that substrate specificity of MKP3 can be independent of the capability of substrate-induced MKP3 activation.
Comparison with Other Tyrosine-specific Phosphatases-In addition to MKPs, biochemical and genetic studies indicated that distinct tyrosine-specific phosphatases (PTPases) are involved in MAPK inactivation. The hematopoietic proteintyrosine phosphatase, HePTP, and its brain-specific homologs, STEP and PTP-SL, can associate with ERK2 and p38␣ through the N-terminal domain and inactivate the ERK2 and p38␣ MAPK pathways by dephosphorylation of Tyr(P) in the activation loop (36,37). Fig. 5A shows the time courses of the p38␣/ pTpY dephosphorylation by these phosphatases. Both HePTP and PTP-SL can effectively dephosphorylate Tyr(P) in p38␣/ pTpY under the same conditions used for MKP3, whereas  Using the continuous spectrophotometric enzyme-coupled assay, we also quantitatively compared the catalytic efficiency of the MKP3-catalyzed p38␣ dephosphorylation with HePTP, PTP-SL, and STEP. All kinetic parameters are summarized in Table 2. As shown in this table, the kinetic parameters for the dephosphorylation of p38␣ by PTP-SL and HePTP were very similar. The k cat /K m values for PTP-SL and HePTP are greater than 10 6 s Ϫ1 M Ϫ1 and represent the largest values measured for a protein phosphatase reaction. In contrast, the k cat value for the STEP-catalyzed p38␣/pTpY dephosphorylation was about 20-fold lower than those for the PTP-SL and HePTP-catalyzed reactions. This result is consistent with the previous findings that HePTP and PTP-SL specifically down-regulate the p38␣ pathway, whereas STEP was unable to inactivate p38␣ in intact cells (37). With the monotyrosine-phosphorylated p38␣/pY as a substrate, the kinetic parameters for the p38␣/pY dephosphorylation by MKP3 are only slightly different from the corresponding values of the p38␣/pTpY dephosphorylation, indicating that dephosphorylation of the Tyr(P) residue by MKP3 does not require the presence of Thr(P) in p38␣ substrate. Unlike MKP3-catalyzed reactions, however, when the monophosphorylated p38␣/pY was used as a substrate for PTP-SL and HePTP, both the k cat and K m values decreased about 3-fold compared with those of p38␣/pTpY, indicating that the presence of Thr(P) had a significant effect on the kinetic properties of the p38␣ substrate for these two phosphatases. Interestingly, the kinetic parameters for the dephosphorylation of p38␣ by the MKP3-ERK2 complex are comparable with those of the HePTP and PTP-SL-catalyzed reaction, although they may possess different intrinsic catalytic activity. The second order rate constant k cat /K m value is a physiologically relevant parameter for the reaction of free enzyme with free substrate. This parameter reveals that the specificity of an enzymatic reaction is an important factor to consider when drawing conclusions about physiological substrates. The k cat /K m values for the dephosphorylation of p38␣ by the MKP3-ERK2 complex are only 2-4-fold lower than those by the most active PTPase, PTP-SL, suggesting that phosphorylated p38␣ is a physiological substrate for MKP3 in cell signaling.
Crystal Structure of Phosphorylated p38␣-To gain a structural understanding of the substrate specificity of MKP3, we crystallized the bisphosphorylated p38␣. The structure of p38␣/pTpY was solved by molecular replacement using the structure of unphosphorylated p38␣ as a search model. The x-ray diffraction data extended to 2.1 Å resolution (supplemental Table S1), and the phosphorylation site residues and covalently bound phosphates are well resolved in the electron density (Fig. 6A). The structure of p38␣/pTpY maintains the overall MAPK topology consisting of N-and C-lobes forming the catalytic groove between them. In p38␣/pTpY, 321 residues occupy positions corresponding to those found in p38␣. The C ␣ atoms of these residues in p38␣ and p38␣/pTpY superimpose with a root mean square deviation of 1.05 Å. The two lobes are rotated toward one another in p38␣/pTpY by 5°with respect to their relative orientations in the inactive conformation, not 25°a s previously thought (38,39). Other differences between the structure of phosphorylated and unphosphorylated p38␣ occur in the conformation of the activation lip, helix ␣C, glycine-rich loop (P-loop), C-terminal extension, and crossover connection (hinge) (Fig. 6B).
The activation segment is the locus of major conformational differences between unphosphorylated and phosphorylated p38␣. The activation lip in p38␣ consists of 13 residues, corresponding to Leu 171 -Val 183 . In inactive p38␣, the two lobes in  the kinase domain exist in an open conformation, and the conformation of the activation loop is incompatible with substrate binding and/or catalysis (Fig. 6C). Phosphorylation of Thr 180 and Tyr 182 leads to a rearrangement of the lip (Fig. 6D). The phosphate moiety of Thr(P) 180 forms 11 hydrogen bonds with four invariant arginines, including Arg 67 and Arg 70 in helix ␣C, Arg 149 in the catalytic loop, and Arg 173 in the activation lip. An additional water molecule-mediated interaction occurs between the phosphate moiety and Lys 66 in helix ␣C. These interactions provide a link between helix ␣C and lip promoting the reorientation of catalytic site residues and interlobe closure. The lip refolding induces a 100°rotation of the side chain of Asp 168 in the conserved DFG motif, which adopts a conformation almost identical with that of Asp 184 in the active PKA (40). In contrast to Thr(P) 180 , the phosphate group of Tyr(P) 182 makes only two hydrogen bonds to the side chain of His 228 , FIGURE 6. Overall view of p38␣/pTpY and comparison with p38␣. A, ribbon diagram of p38␣/pTpY. The p38␣/pTpY is colored green (N-lobe) and blue (C-lobe), with its hinge region and activation segment in red. The 2F o Ϫ F c electron density map for Thr(P) 180 and Tyr(P) 182 , shown on the right, is contoured at 1.0 . B, superimposition of p38␣/pTpY and p38␣ (Protein Data Bank code 1p38) with C-lobe as reference. p38␣/pTpY is colored green, with its hinge region and activation segment in red. The p38␣ is colored orange, with its hinge region and activation segment in blue. Shown is a close-up view of the activation segment and its neighboring structures in p38␣ (C) and p38␣/pTpY (D) (color-coded as in Fig. 5B). The water molecule is shown as a red sphere, and hydrogen bonds are shown as yellow dashed lines. Shown is a close-up view of the ATP-binding site in p38␣ (E) and p38␣/pTpY (F).
which is not conserved in ERK2 and JNKs. Upon phosphorylation, Tyr 182 moves by ϳ18.1 Å relative to its position in the inactive p38␣ structure and appears to be positioned to the p ϩ 1 site. The conformational changes of the activation lip disrupt the turn of helix Val 183 -Arg 186 near the p ϩ 1 site and relieve steric constraints to substrate binding, thereby helping to form the peptide substrate binding channel (Fig. 6D). The ATP binding site is formed by the crossover connection, the flexible glycine-rich loop in the N-terminal domain, and the DFG motif in the C-terminal domain. The relative orientation of the two domains of inactive p38␣ is different from that observed in ERK2 and other kinases. The twisted arrangement of two domains results in misalignment of catalytic residues, such as residues Lys 53 in the N-terminal domain and Asp 168 from the C-terminal domain. Residues in the crossover connection (His 107 -Asp 112 ) are shifted, on average ϳ2.5 Å relative to ERK2, in toward the ATP binding site. As a consequence, the backbone of Met 109 blocks the ATP binding site observed in ERK2 (Fig. 6E). In the structure of bisphosphorylated p38␣, the crossover connection shifts away from the active site. Met 109 occupies a new pocket formed in the ␤ 7 -␤ 8 reverse turns, and the active site residues are significantly displaced relative to their orientation in the unphosphorylated p38␣ structure (Fig.  6F). Thus, the crossover connection of p38␣ is optimally arranged for accommodating ATP during the kinase activation. In addition, the network of interaction between Thr(P) 180 and the basic residues pulls the domains together, thereby reorganizing catalytic residues in the active site. The glycine-rich loop in p38␣/pTpY (Gly 31 -Val 38 ) is moved from its up position found in the inactive p38␣ structure to its down position, and this movement facilitates localization of the phosphate binding ribbon. The shift in position of helix ␣C also brings Glu 71 closer to Lys 53 to form a water molecule-mediated interaction in p38␣/pTpY (Fig. 6F).
Molecular Basis of MKP3 Substrate Specificity-So far, there have been two structural reports for the activated form of ERK2 and p38␥ from 14 members of the human MAPK family (28,41). p38␣ is a member of the p38 MAPK subfamily, which shares 48 and 63% sequence identity with ERK2 and p38␥, respectively. Although p38␣ and p38␥ have greater sequence similarity, differences in phosphorylated structures do not fall strictly along p38␣ versus p38␥ and ERK2 lines. The structure of p38␥/pTpY is more similar to ERK2/pTpY (root mean square deviation ϭ 1.2 Å) than p38␣/pTpY (root mean square deviation ϭ 1.62 Å). The relative orientation of the N-and C-terminal domains is different among the structures of active p38␣, ERK2, and p38␥. p38␣/pTpY has a wider domain separation in the phosphorylated structure than does ERK2/pTpY or p38␥/pTpY (Fig. 7A). A superposition of the C-terminal domains of ERK2 and p38␥ onto the C-terminal domain of p38␣ revealed rotations of the N-terminal domains of ERK2 and p38␥ by about 10.6°.
Large differences among the three activated MAPKs are observed in the positioning of the activation lips. The activation lip of p38␥ consists of residues Leu 174 -Val 186 and is same length as the one in p38␣, whereas the activation lip of ERK2 is six residues longer than that of p38␣ and spans amino acids Lue 168 -Val 186 . Fig. 7B highlights the lip regions from p38␣, p38␥, and ERK2. Except for a longer lip size for ERK2, the structures of the lip regions of activated p38␥ and activated ERK2 are nearly identical. However, the phosphorylation residues in p38␣ have a different position and conformation when compared with the corresponding amino acids in ERK2 and p38␥. The phosphate moieties of Thr(P) 180 and Tyr(P) 182 in p38␣ are 2.5 and 3.1 Å away from the location of the corresponding residues in ERK2 and p38␥, respectively, and the X residues (Gly 181 in p38␣, Gly 184 in p38␥, and Glu 184 in ERK2) in the pTXpY motif exhibit the largest distance of 6.3 Å. The Gly 181 in p38␣ forms hydrogen bonds with Arg 149 from C-lobe, whereas Glu 184 in ERK2 and Gly 184 in p38␥ do not hydrogen-bond directly with any other residue. Thus, in contrast to what had been suggested previously, our structure shows that the conformation of the activation lip in the high activity forms of MAPKs is not conserved.
From the biochemical studies given above, it is evident that the phosphotyrosine plays the critical role in p38␣ recognition by MKP3. To provide a structural explanation of how phosphorylation modulates the interactions between p38␣ and the MKP3 catalytic domain, we first compared the electrostatic potential surfaces of unphosphorylated p38␣ with that of phosphorylated p38␣ and estimated the potential surfaces of phosphorylated p38␣ with which the MKP3 catalytic domain might interact. Surface representations of p38␣ versus p38␣/pTpY show that the conformational difference in the activation loop significantly alters the hydrophobic and electrostatic surface distribution (Figs. 7, C and D). Protein-protein association results from both hydrophobic and electrostatic/hydrogen bonding interactions between interfaces composed of complementary nonpolar and charged/polar residues. A common type of interactive surface contains a hydrophobic patch surrounded by polar groups (42). Inspection of the phosphorylated p38␣ structure reveals one such potential protein binding site in the C-lobe, a hydrophobic cleft, formed by Thr 185 , Arg 186 , His 228 , and Ile 229 on one side and Val 183 , phospho-Tyr 182 , Met 194 , and Leu 195 on the other, that are surrounded by positively charged residues. In unphosphorylated p38␣, this site is unformed and further occluded by an acidic patch formed by Asp 176 , Asp 177 , and Glu 178 . Upon phosphorylation, the movement of the activation loop leads to ϳ23 Å displacement of the acidic patch and exposes the binding site for MKP3 interactions. Additionally, notable shifts and side chain rotations are also observed in residues contributing to the hydrophobic cleft. For example, Tyr(P) 182 and Val 183 are shifted by more than 12 Å; Met 194 , Leu 195 , and Thr 185 move by distances between 1.0 and 2.5 Å. This predicts that the activation state of p38␣ should influence MKP3 binding, by favoring stronger binding interactions; the phosphodonor (phospho-Tyr 182 ) and the hydrophobic cleft of p38␣/pTpY simultaneously contact MKP3.
MKP3 exhibits very high activity toward ERK2/pTpY and p38␣/pTpY (k cat /K m ϭ 7.3 ϫ 10 6 M Ϫ1 s Ϫ1 and 2.8 ϫ 10 5 M Ϫ1 s Ϫ1 ), as compared with that for the ERK2-derived phosphopeptide harboring the same pTEpY motif (k cat /K m ϭ 5 M Ϫ1 s Ϫ1 ) (43). In contrast, MKP3 is unable to dephosphorylate and bind p38␥/pTpY ( Fig. 1D and supplemental Fig. S6), although the conformation of the phosphorylated activation loop of p38␥ is very similar to that of ERK2 (28,41). This strikingly high sub-strate specificity displayed by MKP3 suggests that there is no structural complementarity between the activation loop of ERK2 and MKP3 and that structural features outside the immediate vicinity of the pTEpY site of ERK2 play an important role for the selective specificity of MKP3 for MAPKs. In addition to the common docking site, phosphorylated ERK2 has a second binding site utilized by transcription factor substrates and phosphatases, the hydrophobic DEF motif binding site formed by Thr 188 , Arg 189 , His 230 , Tyr 231 , and Leu 232 on one side; phospho-Tyr 185 , Val 186 , Met 197 , and Leu 198 on the other; and Asn 260 and Tyr 261 at the base (Fig. 7E) (44 -47). Interactions between this site and MKP3 are essential for allosteric activation of MKP3 and formation of a productive complex. Interestingly, the DEF motif binding site in ERK2/pTpY is very similar to the hydrophobic cleft of p38␣/pTpY proposed to mediate interactions with MKP3. By comparison, the corresponding region in p38␥/pTpY is occupied by the side chains of Asp 230 , His 231 , Leu 232 , Asp 233 , Asp 256 , Glu 257 , Asn 260 , and Tyr 261 (Fig. 7F). Thus, the DEF motif interaction pocket in the active p38␥ model is unavailable and therefore unreceptive for MKP3 binding. In agreement with this notion, both ERK2/pTpY and p38␣/ pTpY, but not p38␥/pTpY, were able to associate with the catalytic domain of MKP3 (MKP3/⌬N151) to form a stable heterodimer (Fig. 7, G-I). The structural differences in the DEF motif binding sites and the region outside the immediate vicinity of the pTXpY site suggest one possible mechanism for encoding complex target specificity among MAPK isoforms. Further structural analyses show that the phospho-Tyr side chains in both ERK2/pTpY and p38␣/pTpY structures are positioned similarly and are easily accessible to the active site of MKP3 (Fig. 7, B, D, and E). In contrast, the phospho-Thr residues of ERK2/pTpY and p38␣/pTpY are in different positions and circumstances. Thr(P) 180 of p38␣/pTpY nestles more deeply into the N-lobe and is in the proximity of a negatively charged patch formed by Asp 176 , Asp 177 , and Glu 178 , whereas the corresponding region near Thr(P) 183 in the ERK2/pTpY is largely hydrophobic (Fig. 7, D and E). Thus, the distinct position and circumstance of Thr(P) 180 probably explain the lack of MKP3 activity toward the dephosphorylating phosphothreonine residue in p38␣.

DISCUSSION
The most commonly used approaches for studying the substrate specificity of MAPK phosphatases involve overexpression experiments followed by in vitro assay of kinase activity in the immunoprecipitate of cellular extracts. These kinds of experiments are indirect and may not provide definitive information about the identity of protein phosphatases involved in certain cellular functions if the kinase activity does not reliably reflect its phosphorylation state. Based on our previous studies, the monothreonine-phosphorylated p38␣/pT is catalytically active, and dephosphorylation of phospho-Tyr 182 in the activation loop of p38␣ primarily increases its K m for ATF2 (activating transcription factor 2) substrate, with only a slight decrease in k cat (25). Consequently, the inhibition extent of p38␣ activity by tyrosine-specific phosphatase strongly depends on the concentration of ATF2 used in the assay system (supplemental Fig.  S7). Upon increasing the concentration of ATF2, the relative activity gradually increases and reaches a limiting value 0.8 at the high ATF2 concentration. Considering this fact, our result is also consistent with the in vivo experimental observations that MKP3, PTP-SL, and HePTP can only partially inhibit stress-induced p38 activation (10,13,36,37).
MAPK signaling pathways are tightly regulated, and individual MAPKs exhibit exquisite specificity for their cognate regulators and substrates. All MAPK family members share a common phosphorylation site motif, raising questions as to how specific MAPK recognition by its binding partners is achieved. Previous studies (34,47) suggested that ERK2 is capable of forming a complex with MKP3 through a bipartite mechanism. In this model, one part of MKP3 (e.g. the kinase interaction motif sequence) docks to the D-site located on the back side of the ERK2 active site for high affinity association, whereas the interaction of another structural element with the DEF motif binding site in ERK2 is primarily responsible for ERK2-induced MKP3 activation. In this study, we show that the phosphorylated p38␣ also contains a MKP3 docking pocket that is located below the active site. This pocket is occluded in unphosphorylated p38␣; thus, the MKP3 preferentially binds to the dually phosphorylated, activated isoform of p38␣. Based on the crystal structural studies, we propose that the specific interaction between the catalytic domain of MKP3 and the DEF motif docking pocket in a MAPK may ensure the precise orientation and positioning of the catalytic residues in the phosphatase with respect to the TXY motif in the activation lip for efficient dephosphorylation and therefore possibly correlates with MKP3 substrate specificity toward the members of the MAPK.
Using a variety of approaches, including enzyme kinetics, size exclusion chromatography, and cellular transfection studies, we have identified that phosphorylated p38␣ is an authentic in vitro and in vivo substrate of MKP3. MKP3 catalyzes the specific hydrolysis of phospho-Tyr 182 from activated p38␣, resulting in partial inactivation of kinase activity. Together with published structural and biochemical data (35,48), our experimental observations give rise to a working model for the MKP3catalyzed p38␣ dephosphorylation in cells (supplemental Fig.  S8). In this model, MKP3 can exist in two distinct conformational states: a low activity ERK2-independent form with a k cat of 0.09 s Ϫ1 and a high activity ERK2-dependent form with a k cat of 0.28 s Ϫ1 . Binding of ERK2 to the noncatalytic domain of FIGURE 7. Molecular basis of MKP3 substrate specificity. A, superimposition of ERK2/pTpY (magenta, Protein Data Bank code 2ERK) and p38␥/pTpY (yellow, Protein Data Bank code 1CM8) onto p38␣/pTpY (green) with the C-lobe as a reference. B, close-up view of the activation segment of ERK2/pTpY, p38␥/pTpY, and p38␣/pTpY. Shown are electrostatic surface potential representations (red, negative charge; blue, positive charge; gray, hydrophobic) of p38␣ (C; Protein Data Bank code 1p38), p38␣/pTpY (D), ERK2/pTpY (E; Protein Data Bank code 2ERK), and p38␥/pTpY (F; Protein Data Bank code 1cm8). Shown are gel filtration profiles for the interaction of MKP3C293S/⌬N151 with p38␣/pTpY (G), ERK2/pTpY (H), and p38␥/pTpY (I), respectively. Gel filtration chromatography experiments were performed with a Superdex 200 gel filtration column mounted on an AKTA FPLC system (Amersham Biosciences). All proteins were in 50 mM MOPS buffer, pH 7.0, 100 mM NaCl, and 0.1 mM EDTA.
MKP3 results in an optimum arrangement of active site residues with a concomitant 3-fold increase in the k cat value.
Mounting evidence suggests a cross-talk between the p38 and ERK MAPK pathways in different types of cells. For example, both pathways are simultaneously activated in melanoma, with a positive feedback loop from active ERK to p38 (49). We postulate that, upon activation of both ERK2 and p38 pathways, the active ERK2/pTpY may induce transcriptional activation of the MKP3 gene, and MKP3 in turn dephosphorylates/inactivates ERK2/pTpY. Unphosphorylated ERK2, the dephosphorylation product of ERK2/pTpY by MKP3, associates with MKP3, and thereby enhances the dephosphorylation of the phospho-Tyr residue in p38␣/pTpY to down-regulate the p38␣ pathway. Therefore, the cross-talk between ERK2 and p38␣ mediated through MKP3 may play a role in coordinating regulation of these two distinct MAPK pathways, and the ERK2induced allosteric activation may reflect an additional level of fine tuning of p38␣ MAPK signaling.