Mapping ERK2-MKP3 Binding Interfaces by Hydrogen/Deuterium Exchange Mass Spectrometry*

ERK2, a prototypic member of the MAPK family, plays a central role in regulating cell growth and differentiation. MKP3, an ERK2-specific phosphatase, terminates ERK2 signaling. To understand the molecular basis of ERK2 recognition by MKP3, we carried out hydrogen/deuterium exchange mass spectrometry experiments to map the interaction surfaces between the two proteins. The results show that the exquisite specificity of MKP3 for ERK2 is governed by two distinctive protein-protein interactions. To increase the “effective concentration” of the interacting molecules, the kinase interaction motif in MKP3 (64RRLQKGNLPVR74) and an MKP3-specific segment (101NSSDWNE107) bind the common docking site in ERK2 defined by residues in L16, L5, β7-β8, and αd-L8-αe, located opposite the kinase active site. In addition to this “tethering” effect, additional interactions between the 364FTAP367 sequence in MKP3 and the ERK2 substrate-binding site, formed by residues in the activation lip and the P+1 site (β9-αf loop), L13 (αf-αg loop), and the MAPK insert (L14-α1L14-α2L14), are essential for allosteric activation of MKP3 and formation of a productive complex whereby the MKP3 catalytic site is correctly juxtaposed to carry out the dephosphorylation of phospho-Thr183/phospho-Tyr185 in ERK2. This bipartite protein-protein interaction model may be applicable to the recognition of other MAPKs by their cognate regulators and substrates.

Mitogen-activated protein kinase (MAPK) 3 cascades are highly conserved signal transduction modules in eukaryotes that mediate the intracellular transmission and amplification of extracellular stimuli, leading to the induction of appropriate cellular responses to changes in the surrounding environment (1,2). The MAPKs are compact enzymes lacking recognizable protein-protein interaction modules normally found in signaling molecules (3,4), yet they are highly specific in their interactions with substrates, activating kinases, or inactivating phosphatases. Although the importance of MAPKs in cellular signaling is well established, there is limited understanding of the molecular basis for MAPK recognition by its activators, inactivators, and substrates. Such knowledge is essential for comprehension of the ability of MAPKs to integrate diverse biological stimuli and to transmit signals to the nucleus to generate appropriate cellular responses.
Extracellular signal-regulated protein kinase 2 (ERK2), the prototypic member of the MAPK family, has been the subject of intense study. ERK2 is activated by phosphorylation of Thr 183 and Tyr 185 in the activation lip by the dual specificity MAPK/ ERK kinase MEK1. Deactivation of ERK2 activity is carried out by MAPK phosphatase 3 (MKP3) (5,6). Genetic analyses show that MKP3 plays an important role in modulating a number of ERK-mediated embryonic developmental processes, including oogenesis and wing formation in Drosophila and neural patterning in Xenopus (7,8). Additional studies have indicated that MKP3 is responsible for the negative feedback regulation of fibroblast growth factor-induced ERK activation in developing limbs, neural plate, and somites in chicken embryos (9 -11).
MKP3 forms a physical complex with ERK2 and is highly specific for ERK2 dephosphorylation with a k cat /K m that is 10 6fold higher than that for the hydrolysis of p-nitrophenyl phosphate or the bisphosphorylated peptide derived from the ERK2 activation lip (12,13). Part of the specificity comes from docking interactions between the N-terminal domain of MKP3 and the non-catalytic regions of ERK2 (12,14,15). Interestingly, ERK2 can also stimulate the phosphatase activity of MKP3 (16). Biochemical and structural evidence suggests that ERK2 binding elicits allosteric activation of MKP3, resulting in optimum alignment of the general acid and other active-site residues in MKP3 with respect to the substrate for efficient catalysis (12,(17)(18)(19)(20). Thus, MKP3 substrate specificity is linked to the ability of the substrate to induce productive orientation in the active site. This provides a powerful mechanism to ensure high fidelity in MKP3-mediated ERK2 inactivation. In this mechanism, MKP3 exists in latent inactive states, and upon association with ERK2, MKP3 is activated, leading to selective inactivation of ERK2.
The structural basis for the exquisite specificity of ERK2 dephosphorylation by MKP3 has not been elucidated. In fact, it has been notoriously difficult to obtain co-crystals of ERK2 in complex with any of its interacting proteins. To better under-stand the molecular basis of specific ERK2 recognition by MKP3, we have employed hydrogen/deuterium exchange mass spectrometry (H/DX-MS) to map the interaction surfaces between the two proteins concomitant with complex formation. By combining the information gained from the H/DX-MS experiments and direct biochemical analyses, coupled with molecular modeling, we were able to define the structural elements in both MKP3 and ERK2 that are important for specific recognition. The results also yield a structural model for understanding how efficient and precise docking interactions between ERK2 and its cognate substrates and regulators can be achieved.

MATERIALS AND METHODS
Protein Expression and Purification-N-terminally His 6tagged ERK2 was expressed in Escherichia coli and purified following a published procedure (21). The expression and purification of ERK2/pTpY (ERK2 phosphorylated at both Thr 183 and Tyr 185 ) were carried out as described (21). Catalytically inactive MKP3 (MKP3/C293S) with a C-terminal His 6 tag and glutathione S-transferase (GST)-MKP3/C293S were expressed in E. coli and purified as described previously (14,17).
Site-directed Mutagenesis of MKP3 and ERK2-Mutant ERK2 and MKP3 were generated by PCR according to the standard procedure of the QuikChange TM site-directed mutagenesis kit (Stratagene) using pET15b-His 6 -ERK2 and pET21a-MKP3-His 6 as templates, respectively. All mutants were verified by DNA sequencing.
GST Pulldown Analysis-The binding affinity of GST-MKP3/C293S for ERK2 or ERK2/pTpY was determined by GST pulldown and Western blot analyses. Five g of GST-MKP3/ C293S in 0.5 ml of phosphate-buffered saline (140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , and 2 mM dithiothreitol, pH 7.3) was immobilized on 20 l of glutathione-Sepharose 4B beads (Amersham Biosciences) by gentle agitation at 4°C for 2 h. Different concentrations of ERK2 or ERK2/ pTpY in 200 l of phosphate-buffered saline containing 0.5% Triton X-100 were incubated with 20 l of GST-MKP3/C293Sbound beads with gentle agitation at 4°C for 2 h. After washing the beads once with phosphate-buffered saline containing 0.5% Triton X-100 and three more washes with phosphate-buffered saline, 20 l of 2ϫ SDS sample buffer was added to the beads. GST-MKP3/C293S and bound ERK2 (or ERK2/pTpY) were released by boiling the beads for 5 min at 95°C. The sample was centrifuged at 10,000 rpm for 2 min. The supernatant (15 l) was loaded onto 12.5% SDS-polyacrylamide gel. After electrophoresis, the proteins on the gel were transferred at 25 mA to nitrocellulose membrane overnight at 4°C. Rabbit anti-phospho-p44/42 MAPK antibody (catalog no. 9101, Cell Signaling Technology, Inc.) and rabbit anti-p44/42 MAPK antibody (catalog no. 9102, Cell Signaling Technology. Inc.) were used as primary antibodies to detected phospho-ERK2 and ERK2, respectively, on the membrane.
H/DX-MS-Hydrogen/deuterium exchange in MKP3/ C293S and ERK2/pTpY, either alone or in complex, was initiated by diluting 1 l of the stock solution (200 M) to 19 l of the D 2 O buffer (20 mM Tris and 100 mM NaCl, pD 7.5) on ice. The lower temperature was chosen because exchange was too rapid to be monitored at room temperature by manual quenching. The final protein concentration was 10 M. The resulting hydrogen/deuterium exchange solutions were maintained at 0°C by leaving the sample on ice and allowed for exchange for different periods of time. At appropriate intervals, the hydrogen/deuterium exchange reaction was quenched by the addition of an equal volume of cold 0.5 M phosphate buffer, pH 2.5. One l of the deuterated protein solution was loaded immediately onto a Vydac C 4 column (1.0 ϫ 50 mm) and analyzed by mass spectrometry for global exchange. Changes in deuterium incorporation within specific regions of proteins can be detected by peptidic digestion of the deuterium-labeled proteins, followed by HPLC separation and mass spectrometric analysis. To analyze hydrogen/deuterium exchange in segments of MKP3/C293S and ERK2/pTpY proteins, the hydrogen/deuterium exchange reaction was quenched by the addition of an equal volume of cold 0.5 M phosphate buffer, pH 2.5, in the presence of 20 M pepsin. Immediately, 5 l of the sample was injected into the sample loop, where pepsin digestion proceeded at 0°C for 4 min. The pepsin digests were subsequently separated on a Vydac C 18 column (1.0 ϫ 50 mm). A Shimadzu HPLC system equipped with two LC-10ADVP pumps was used to generate an acetonitrile gradient. Solvent A was 94.9% H 2 O containing 5% acetonitrile and 0.1% formic acid, and Solvent B was 95% acetonitrile containing 4.9% H 2 O and 0.1% formic acid. A low volume static mixing tee was used to minimize the delay period for HPLC. The solvent pre-cooling coil, static mixing tee, Rheodyne injector, and column were immersed in an ice bath (0°C) to minimize back-exchange with HPLC solvents. After being desalted at 5% solvent B for 5 min, the protein eluted with a 2-min 30 -50% solvent B gradient, and the peptidic peptides eluted with a 0.5-min 5-15% solvent B gradient, followed by an 8-min 15-45% solvent B gradient. The column effluent (50 l/min) was delivered directly to a Thermo Finnigan LCQ Deca XP Plus ion trap mass spectrometer for mass analyses of the deuterated protein and its peptic fragments. MagTran Version 1.0 software (written by Dr. Zhongqi Zhang) was used to determine the centroid value for each peak in the list processed from the mass spectrum. The sequences of the peptide fragments obtained after pepsin digestion were identified by tandem mass spectrometry (MS/MS). The total ion intensity of the peptides was detected in the m/z range of 400 -2000. Mass scanning was followed by collisioninduced dissociation to acquire a tandem mass spectrum within a peak during the HPLC run. Tandem mass spectra were interpreted by searching peptide data bases using the SEQUEST algorithm.
The extent of hydrogen/deuterium exchange was calculated using Equation 1, where m is the measured mass of the deuterated protein or peptide; m 0% is the measured mass of a "zero-deuteration" control prepared by adding 1 l of 200 M protein solution to a 19:20 (v/v) mixture of deuterated buffer and quenching buffer; m 100% is the measured mass of a "full-deuteration" control pre-pared by incubating the protein sample for 16 h at room temperature in 3 M deuterated guanidine hydrogen chloride, followed by the addition of an equal volume of cold quenching buffer; and H is the total exchangeable amide hydrogens. Molecular Modeling of ERK2/pTpY and MKP3 Interactions-The crystal and NMR structures of ERK2 and MKP3, together with H/DX-MS and mutagenesis data, were used to generate a model for the ERK2-MKP3 complex. The ERK2-MKP3 kinase interaction motif (KIM; referred to as KIM MKP3 ; Protein Data Bank code 2FYS) structure (22) was utilized as the starting point for modeling the interaction between the N-terminal domain of MKP3 and ERK2. The KIM peptide (residues 64 -74) from the NMR structure (Protein Data Bank code 1HZM) (20) was docked onto the ERK2 structure with contacts analogous to those found in the experimentally determined ERK2-KIM MKP3 structure (22). Residues 101-107 in the NMR structure were extended in the model using program O (23). The orientation of the N-terminal domain of MKP3 relative to ERK2 was adjusted based on the H/DX-MS and mutational results. The resulting model was then stereochemically refined by energy minimization using the program CNS (24).
The phosphorylated ERK2 structure (Protein Data Bank code 2ERK) (25) and the crystal structure of the C-terminal domain of MKP3 (Protein Data Bank code 1MKP) (18) were utilized as the starting point for modeling the interaction between ERK2 and the C-terminal domain of MKP3. The ERK2/pTpY structure superimposes with a root mean square deviation of 0.59 Å with ERK2-KIM MKP3 and is thus quite similar. The model was created using program O (23) based on the following constraints: phospho-Tyr 185 in ERK2 binds the MKP3 active site; the WpD loop adopts a closed conformation; there is no contact between MKP3 and ERK2 peptides 54 -65 and 330 -358 (based on the hydrogen/deuterium exchange results); and the MKP3 peptide 362 LYFTAPSN 369 occupies the previously implicated ERK2 substrate-binding (SB) site (15,26). Finally, the model was refined using the program CNS (24).

RESULTS AND DISCUSSION
H/DX-MS has emerged in recent years as a powerful tool for mapping protein-ligand and protein-protein interfaces as well as identifying conformational and dynamic perturbations in proteins (27). In this study, we sought to define the ERK2/ pTpY-MKP3 binding sites by determining the changes in solvent accessibility of the protein backbone amides as a result of complex formation between the full-length proteins in solution. However, given the transient nature of the enzymesubstrate complex, it is difficult to study the binding interaction between wild-type MKP3 and doubly phosphorylated ERK2/pTpY. Fortunately, catalysis by the protein-tyrosine phosphatase (PTP) superfamily requires an essential nucleophilic Cys (28), and the catalytically inactive Cys-to-Ser mutant (e.g. Cys 293 in MKP3) retains the ability to bind substrates, but is unable to carry out substrate dephosphorylation. In fact, the Cys-to-Ser mutant binds substrates/ligands with the same affinity as the wild-type enzyme (29 -31). Consequently, we employed catalytically inactive MKP3/ C293S to capture the enzyme-substrate complex and to prevent the hydrolysis of ERK2/pTpY during H/DX-MS analysis.
Binding Affinity of MKP3/C293S for ERK2/pTpY-The affinity of MKP3/C293S for ERK2 (K d ϭ 0.13 Ϯ 0.01 M) is similar to that of wild-type MKP3 (K d ϭ 0.17 Ϯ 0.04 M) (14). In addition, GST-MKP3 and MKP3 display identical affinity for ERK2 (17). To determine the K d for MKP3/C293S and ERK2/pTpY, we carried out GST-MKP3/C293S pulldown experiments to measure ERK2/pTpY binding directly. As a control, we also performed GST-MKP3/C293S pulldown of unphosphorylated ERK2. The amount of ERK2 or ERK2/pTpY associated with GST-MKP3/C293S was visualized by anti-ERK2 or anti-phospho-ERK2 antibody and quantified by densitometry (Fig. 1). The dissociation constant (K d ) for ERK2 binding to GST-MKP3/C293S is 0.19 Ϯ 0.06 M, which is similar to that measured previously by a competition-based assay (14). Likewise, the K d determined by the pulldown experiments for GST-MKP3/C293S and ERK2/pTpY is 31 Ϯ 5 nM, which is comparable with the K m of MKP3 for ERK2/pTpY (22 Ϯ 5 nM) (12). Thus, GST-MKP3/C293S binds doubly phosphorylated ERK2 6-fold tighter than unphosphorylated ERK2, consistent with the expectation that the interactions between the MKP3 active site and the phosphoamino acids in ERK2 should further enhance the binding affinity between the two proteins. Obviously, characterization of the interfaces between MKP3/C293S and ERK2/pTpY should enhance our understanding of ERK2 recognition by MKP3.
Global Hydrogen/Deuterium Exchange in Intact MKP3/ C293S and ERK2/pTpY-In a typical H/DX-MS experiment, MKP3/C293S and ERK2/pTpY, either alone or in complex, were incubated in D 2 O to allow exchange of protons with solvent deuterium, and mass spectrometry was used to monitor the in-exchange rates. The final concentration for both proteins was 10 M. Given the K d (31 nM) for complex formation, 95% of MKP3/C293S and ERK2/pTpY should be in the bound state under our experimental conditions. C-terminally His 6tagged MKP3/C293S contains 387 amino acids (43,117.5 Da) with 365 exchangeable amide hydrogens, whereas N-terminally His 6 -tagged ERK2/pYpT has 364 residues (42,328.0 Da) with 343 exchangeable amide hydrogens. Fig. 2 shows the total deuterium incorporation into MKP3/C293S and ERK2/pTpY in either the free or bound form. Within 40 min, a total of 190 amide hydrogens in MKP3/C293S were replaced with deuterium in the absence of ERK2/pTpY, whereas 136 of the MKP3/ C293S amide protons were exchanged with deuterium in the presence of ERK2/pTpY, indicating that 54 amide protons in MKP3/C293S were protected from deuterium exchange upon complex formation. Compared with ERK2/pTpY alone, 30 of the ERK2 amide protons were protected from exchange in the ERK2/pTpY-MKP3/C293S complex. These results indicate that an overall decrease in solvent accessibility occurred in MKP3/C293S and ERK2/pTpY as a result of protein-protein interaction. By determining the changes in solvent accessibility of MKP3/C293S and ERK2/pTpY resulting from complex formation, one can identify the ERK2-MKP3 binding interfaces and the structural changes that accompany the complex formation.
Changes in Solvent Accessibility in Specific Segments of MKP3/C293S and ERK2/pTpY upon Complex Formation-H/DX-MS combined with peptidic mapping permits segmentspecific identification of solvent-accessible exchange sites in proteins. The principle for this approach is that ligand binding or protein-protein interaction perturbs protein structure and thus solvent accessibility of the contact regions. The binding interfaces can be localized by comparing the rates of hydrogen/ deuterium exchange on proteins in bound and unbound states. To identify the segments of MKP3/C293S and ERK2/pTpY that displayed altered solvent accessibility, deuterium uptake in MKP3/C293S and ERK2/pTpY was locked in place at various times by rapidly lowering the pH and temperature; the proteins were digested with pepsin; and the resulting peptides were separated by HPLC and analyzed by electrospray ionization mass spectrometry. This measurement reports time-dependent changes in weighted average peptide masses, which yields rates of hydrogen/deuterium exchange within different regions of the protein (27). Overall, 49 peptides (including those with overlapping sequences), covering 92% of the MKP3 amino acid sequence, were identified by MS/MS as shown by the coverage map (Fig. 3). Of the 49 peptides, 30 displayed reduced exchange rates, 2 exhibited an increase in exchange rates, and 17 had no change in deuterium incorporation upon binding to ERK2/pTpY. For ERK2/pTpY, 47 peptides were identified, effectively covering 92.5% of the primary structure (Fig. 4). Upon complex formation with MKP3/C293S, there were 22 peptides displaying a decrease in deuterium incorporation, 22 showing no change, and 3 exhibiting an increase in deuterium incorporation.
To directly visualize the hydrogen/deuterium exchange results, we depicted changes in solvent accessibility in MKP3 upon complex formation by color-coding segments of the solution NMR structure of the N-terminal domain (residues 1-154) (20) and the x-ray crystal structure of the C-terminal catalytic domain (residues 204 -347) (18) of MKP3 (Fig. 3C).   DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50
Identification of the ERK2/pTpYbinding Sites in MKP3-The H/DX-MS technique cannot differentiate between changes in solvent accessibility as a result of direct binding and those due to conformational or dynamic perturbations. Thus, peptide segments showing altered hydrogen/deuterium exchange after complex formation could be involved in either ERK2-MKP3 binding or conformational/dynamic changes. Additional biochemical experiments in combination with the H/DX-MS data were required to identify the ERK2binding sites in MKP3. To this end, we had suggested previously that KIM ( 64 RRLQKGNLPVRAL 76 ) in MKP3 plays a major role in high affinity ERK2 binding (14). To provide further evidence that KIM is directly involved in ERK2 binding, we performed GST pulldown assays. Purified His 6 -tagged MKP3/C293S proteins were incubated with GST-ERK2/pTpY immobilized on glutathione-Sepharose beads. As expected, replacement of Arg 65 with Ala (MKP3/R65A) completely disrupted ERK2 binding (Fig. 5). Finally, the crystal structure of ERK2 in complex with the KIM peptide from MKP3 provides a direct visualization of the docking interactions between KIM and ERK2 (22). Together, these findings are fully consistent with the strong solvent protection observed in Val 56 -Leu 76 (N-␤ 3 -␤ 3Ј -␣ 3 ) upon ERK2-MKP3 complex formation (Fig. 3).
Aside from KIM, no other regions in the MKP3 N-terminal domain have been implicated previously in ERK2 binding. The H/DX-MS experiments revealed that, in addition to KIM, the Tyr 98 -Leu 115 (N-␤ 4 -␣ 4 ) segment also displayed a significant decrease in deuterium incorporation, whereas the Thr 87 -Leu 97 (N-␣ 3 -␤ 4 ) and Gln 140 -Phe 143 (N-␣ 5 -␤ 6 ) segments showed increased deuterium uptake in the complex (Fig. 3). To evaluate their potential involvement in ERK2 binding, a number of residues in these regions were mutagenized, and the effect on ERK2 binding was determined by the GST pulldown assay. As shown in Fig. 5, substitution of Leu 97 , Asn 106 , and Glu 142 with Ala abrogated the ability of MKP3/C293S to bind GST-ERK2/pTpY, suggesting that these segments in the N-terminal domain of MKP3 may also be important for ERK2 binding.
The existing MKP3 three-dimensional structures do not include residues 155-203 between the N-and C-terminal domains. A previous study suggested that residues 161-177 might contribute to ERK2 binding (14). These residues may also serve as a nuclear export signal for cytoplasmic retention of ERK2 (32). Furthermore, ERK2 can phosphorylate two residues in this region, Ser 159 and Ser 197 , promoting MKP3 degradation by the proteasomal pathway (33). Although no change in solvent accessibility was detected for residues 151-193 upon complex formation, a decrease in deuterium uptake was observed for residues 194 -206 (Fig. 3). Surprisingly, replacement of Asp 195 or Leu 199 with Ala had no significant effect on MKP3/C293S binding to ERK2/ pTpY (Fig. 5). Thus, based on the present data, it is not clear whether this region is directly involved in ERK2 binding.
The C-terminal domain of MKP3 is the catalytic domain, which adopts a typical PTP fold (Fig. 3C) (18). The active site is situated within a cleft formed by the PTP loop between C-␤ 8 and C-␣ 5 ( 291 VHCLAGISRS 300 ). In the catalytic mechanism, Cys 293 functions as a nucleophile, whereas Arg 299 coordinates the substrate phosphoryl group. Substrate turnover is facilitated by a general acid (Asp 262 in MKP3) located in a flexible surface loop (the WpD loop, C-␤ 7 -3 10 , residues 257-268) in the vicinity of the active site. However, unlike the PTPs, MKP3 exhibits an extremely low basal activity for p-nitrophenyl phosphate (17), and the three-dimensional structures of the catalytic domain of MKP3 show that Cys 293 and Arg 299 are misaligned and that the WpD loop is in the open conformation such that the general acid Asp 262 is positioned away from the active site (18,20). Interestingly, the MKP3 phosphatase activity can be  substantially increased by the presence of ERK2 (16). Kinetic evidence suggests that ERK2 binding elicits MKP3 activation by facilitating the repositioning of active-site residues (Cys 293 and Arg 299 ) and WpD loop closure, bringing the carboxyl group of Asp 262 close to the leaving group oxygen of a substrate (12,17,19).
As shown in Fig. 3, binding of ERK2/pTpY reduces the hydrogen/deuterium exchange rates throughout the MKP3 catalytic domain, including both the PTP and WpD loops. Strong solvent protection in the PTP and WpD loops is consistent with the expectation that the main chain amides and the Arg 299 side chain in the PTP loop make numerous hydrogen bonds with the bound phosphoryl oxygens and that the WpD loop should adopt the closed conformation upon substrate binding (28). However, it is unlikely that the decrease in hydrogen/deuterium exchange over the vast majority of the C-terminal domain is due to direct contact with ERK2/pTpY. To determine whether any of the segments displaying reduced deuterium incorporation are important for ERK2 binding, a number of surface-exposed residues in these regions were changed to Ala, and the effect on ERK2 binding was assessed with the GST pulldown assay. Substitution at Asp 226 in C-␣ 2 ; Tyr 235 in C-␤ 3 ; Asp 262 and Trp 264 in the WpD loop (C-␤ 7 -3 10 ); Glu 274 in C-␣ 4 ; Asn 329 , Ser 331 , and Phe 334 in the C-␣ 6 -␣ 7 loop; and Leu 341 in C-␣ 7 had little effect on MKP3/C293S binding to ERK2/pTpY (Fig. 5). The results suggest that these residues are not directly involved in ERK2 binding. On the other hand, elimination of the side chains at Tyr 215 in C-␤ 2 -␣ 2 , Leu 246 in C-␤ 3 -␤ 3Ј , and Tyr 255 in C-␤ 7 significantly reduced the binding affinity of MKP3/C293S for ERK2/pTpY (Fig. 5), indicating that these residues may play a role in ERK2 binding. Thus, ERK2/pTpY binding may not only lead to closure of the WpD loop, but also tighten up much of the structures stabilizing the PTP loop, resulting in an overall decrease in protein dynamics and flexibility in the MKP3 catalytic domain. This is consistent with the notion that an ERK2-induced conformational reorganization in the active site is required for MKP3 to achieve full activity (17)(18)(19).
No structural information is available for the C-terminal tail of MKP3 (residues 348 -381). The only segment in this region that displayed reduced deuterium incorporation upon ERK2 binding encompasses residues 363-378 (Fig. 3B). Interestingly, this segment contains a putative ERK2-docking motif ( 364 FTAP 367 ), which is also found in many ERK2 substrates (34). Similar FXFP sequences can also be found in a number of MKPs that are capable of inactivating ERK2 (14). Although deletion or mutation of 364 FTAP 367 reduces the affinity of MKP3 for ERK2 by Ͻ10-fold, this region is absolutely required for ERK2-induced MKP3 activation (14). Consistent with this finding, we found that conversion of the 364 FTAP 367 sequence to AAAA (FTAP/A4) completely abolished the association of MKP3/C293S with ERK2/pTpY (Fig. 5). Together, these results provide strong evidence that the FXFP peptide is directly involved in ERK2 binding and possibly in ERK2-induced MKP3 activation.
Identification of the MKP3-binding Sites in ERK2/pTpY-The ERK2 structure includes an N-terminal lobe (␤ 1 -␤ 5 and ␣ c ) responsible for ATP binding and a larger C-terminal lobe (␣ d -␣ i and ␤ 6 -␤ 9 ) involved in protein substrate recognition (25,35). The catalytic cleft is situated between the two lobes, which mediates the phosphoryl transfer reaction (Fig. 4C). Phosphorylation of Thr 183 and Tyr 185 in the activation lip (␤ 9 -␣ f ) is important for ERK2 kinase activity and substrate recognition. Phosphorylated ERK2 is a substrate of MKP3, and MKP3 is capable of dephosphorylating both phospho-Thr 183 and phospho-Tyr 185 (12). A common docking domain in ERK2 (residues 311-324 in L 16 ), decorated with several acidic residues (e.g. Asp 316 and Asp 319 ), was proposed to interact with the positively charged residues in KIMs (36).
On the basis of the ability of ERK2 to induce MKP3 activation, we developed a biochemical assay that provides a quantitative assessment of the importance of structural features in ERK2 for MKP3 recognition in terms of both the MKP3 binding affinity and the propensity to activate MKP3 (14). Our previous mutational studies suggested that MKP3 binding and activation likely involve two distinct surface areas in ERK2 (15). One area, termed the common docking (CD) site, is situated opposite the kinase catalytic cleft and includes Glu 79 , Tyr 126 , Arg 133 , Asp 160 , Tyr 314 , Asp 316 , and Asp 319 . The CD site is important for high affinity MKP3 binding, but is not essential for ERK2-induced MKP3 activation. MKP3 activation appears to require Arg 189 , Trp 190 , Glu 218 , Arg 223 , Lys 229 , and His 230 in the putative ERK2 substrate binding (SB) site, located distal to the CD site.
The crystal structure of ERK2 bound to the KIM peptide from MKP3 reveals that the KIM peptide binds to a contiguous surface area defined by the CD site, which is nested in the C-terminal domain between L 16 , L 5 , ␣ e , ␤ 7 -␤ 8 , and ␣ d (22). This KIM-docking site consists of a highly acidic patch (Glu 79 , Tyr 126 , Asp 160 , Asp 316 , and Asp 319 in L 16 , L 5 , ␤ 8 , and ␣ e ) and a hydrophobic groove (Thr 108 , Leu 110 , Leu 113 , Leu 119 , Phe 127 , and Leu 155 in ␣ e , ␤ 7 -␤ 8 , and ␣ d ), which engage the basic (Arg 64 and Arg 65 ) and hydrophobic (Leu 71 -Pro 72 -Val 73 ) residues, respectively, in the KIM sequence.
In agreement with the x-ray structure of ERK2-KIM MKP3 and mutational studies, strong protection from hydrogen/deuterium exchange was observed in Met 106 -Tyr 126 (␣ d -L 8 -␣ e ), Asn 156 -Leu 161 (␤ 7 -␤ 8 ), and Leu 306 -Glu 324 (␣ i -L 16 ) upon MKP3/C293S binding (Fig. 4), regions that are known to directly interact with the KIM sequence from MKP3. Decreased deuterium incorporation in these regions was also detected in ERK2/pTpY upon binding the KIM peptide from the transcription factor Elk1 (residues 311-327), a substrate for ERK2 (26). In addition to these peptides, Leu 292 -Gln 313 (L 15 -␣ i ) also showed a decrease in deuterium uptake upon complex formation. Substitution of Arg 299 with Ala reduced the binding affinity of ERK2 for MKP3 by 24-fold ( Table 1), suggesting that this region may also be important for MKP3 binding.
Collectively, our biochemical and structural data have defined the CD site for KIM, which is situated on a surface of the ERK2 C-terminal lobe between ␣ d , ␣ e , L 16 , L 5 , and the ␤ 7 -␤ 8 reverse turn. Additional studies indicate that the KIMs from various MAPK substrates and regulators occupy the same docking site in all MAPKs (22,26,(37)(38)(39). This begs the question, how is specificity maintained by individual MAPK-interacting proteins to ensure signaling fidelity? Although differences in the amino acid composition of the CD sites can partially explain binding selectivity, it is unlikely that the CD site interactions alone can account for the observed pathway specificity in MAPK signaling. The interaction of the ERK2 SB site with other elements in MKP3 may provide additional contacts crucial for MKP3 phosphatase activity and/or specificity (12,14,15).
Val 12 -Lys 53 (spanning ␤ 2L0 , L 1 , ␤ 1 , L 2 , ␤ 2 , L 3 , and ␤ 3 ) compose the third region in ERK2/pTpY in which a decrease in deuterium incorporation upon complex formation with MKP3/C293S was observed (Fig. 4). This region includes the ATP-binding site and is adjacent to the activation lip in the three-dimensional structure. However, substitution of a number of residues in this region, including Val 19 , Thr 24 , Tyr 34 , and Lys 46 , did not significantly affect the ability of ERK2 either to bind MKP3 or to induce MKP3 activation (Table 1). Thus, residues in the N-terminal ␤-strands may not be in direct contact with MKP3, and the observed protection from hydrogen/deuterium exchange in this region is likely the result of a decrease in backbone dynamics and/or flexibility. The last region of altered hydrogen/deuterium exchange in ERK2/pTpY corresponds to Asn 199 -Gly 213 (L 12 -␣ f ), which immediately follows the activation lip (Fig. 4). Because removal of the side chain at either Lys 201 or Lys 205 had little effect on MKP3 recognition (Table 1), the increase in hydrogen/deuterium exchange in Asn 199 -Gly 213 likely reports conformational/dynamic changes within L 12 -␣ f upon MKP3 binding.
Structural Model for ERK2/pTpY-MKP3 Recognition-To define the precise interaction surfaces and to identify residues that are directly involved in ERK2/pTpY recognition by MKP3, we constructed a model for the ERK2/pTpY-MKP3 complex (Fig. 6) based on existing three-dimensional structures for ERK2 and MKP3 as well as on the H/DX-MS and mutagenesis data obtained in this study. The ERK2-KIM MKP3 structure (22) was employed as the starting point for modeling the interaction between the N-terminal domain of MKP3 and ERK2. The KIM peptide from the NMR structure of the N-terminal domain of MKP3 (20) was docked onto the ERK2 structure with molecular interactions analogous to those found in the experimentally determined ERK2-KIM MKP3 structure (Fig. 6A). Residues 101-107 from the NMR structure were extended and adjusted to optimize contacts with ERK2 in the model using program O (23). The model suggests that MKP3 residues Asn 101 , Ser 102 , and Ser 103 engage in polar interactions with ERK2 residues Asp 122 , His 123 , and Ser 120 . In addition, Asp 104 in MKP3 forms two hydrogen bonds with the side chain of Gln 117 in ERK2, and Trp 105 in MKP3 makes van der Waals contacts with Gln 117 and His 118 in ERK2. Finally, the side chains of Asn 106 and Glu 107 in MKP3 are involved in polar interactions with His 118 and Asn 121 in ERK2, respectively (Fig. 6A). This model is compatible with all available structural, H/DX-MS, and mutational data and defines the interactions between the CD site and the KIM sequence and the N-␤ 4 -␣ 4 loop in MKP3.
To model the interaction between ERK2/pTpY and the MKP3 catalytic domain, we utilized the phosphorylated ERK2 structure (25) and the crystal structure of the C-terminal domain of MKP3 (18). The model was created based on the following constraints: phospho-Tyr 185 occupies the MKP3 active site; the WpD loop adopts a closed conformation; there is no contact between MKP3 and ERK2 peptides 54 -65 and 330 -358 (based on the hydrogen/deuterium exchange results); and the MKP3 peptide 362 LYFTAPSN 369 resides in the previously implicated ERK2 SB site (15,22). As shown in Fig. 6B

Effects of ERK2 mutations on MKP3 binding and activation
The K d and extent of ERK2-induced MKP3 activation were determined using methods described previously (14).  in the ERK2 activation lip. A-C were produced using GRASP (41). D, overall model of ERK2/pTpY complexed with MKP3. ERK2/pTpY is shown in gray, the MKP3 N-terminal domain in green, and the MKP3 C-terminal domain in blue. The CD and SB sites are colored yellow and cyan, respectively, and phospho-Tyr 185 is colored pink. The KIM peptide, the N-␤ 4 -␣ 4 loop, and the FXFP peptide are highlighted in red. The MKP3 active-site residues Asp 262 , Cys 293 , and Arg 299 are also depicted as stick models in atomic colors. D was created using PyMOL (www.pymol.org) (42). It should be noted that the linker (residues 155-203) between the N-and C-terminal domains of MKP3 is of sufficient length to connect the two domains in the model.