Mechanistic Basis for Catalytic Activation of Mitogen-activated Protein Kinase Phosphatase 3 by Extracellular Signal-regulated Kinase*

The dual specificity mitogen-activated protein kinase phosphatase MKP3 has been shown to down-regulate mitogenic signaling through dephosphorylation of extracellular signal-regulated kinase (ERK). Camps et al. (Camps, M., Nichols, A., Gillieron, C., Antonsson, B., Muda, M., Chabert, C., Boschert, U., and Arkinstall, S. (1998) Science 280, 1262–1265) had demonstrated that ERK binding to the noncatalytic amino-terminal domain of MKP3 can dramatically activate the phosphatase catalytic domain. The physical basis for this activation has not been established. Here, we provide detailed biochemical evidence that ERK activates MKP3 through the stabilization of the active phosphatase conformation, inducing closure of the catalytic “general acid” loop. In the closed conformation, this loop structure can participate efficiently in general acid/base catalysis, substrate binding, and transition-state stabilization. The pH activity profiles of ERK-activated MKP3 clearly indicated the involvement of general acid catalysis, a hallmark of protein-tyrosine phosphatase catalysis. In contrast, unactivated MKP3 did not display this enzymatic group as critical for the low activity form of the enzyme. Using a combination of Brönsted analyses, pre-steady-state and steady-state kinetics, we have isolated all catalytic steps in the reaction and have quantified the specific rate enhancement. Through protonation of the leaving group and transition-state stabilization, activated MKP3 catalyzes formation of the phosphoenzyme intermediate ∼100-fold faster than unactivated enzyme. In addition, ERK-activated MKP3 catalyzes intermediate hydrolysis 5–6-fold more efficiently and binds ligands up to 19-fold more tightly. Consistent with ERK stabilizing the active conformation of MKP3, the chemical chaperone dimethyl sulfoxide was able to mimic this activation. A general protein-tyrosine phosphatase regulatory mechanism involving the flexible general acid loop is discussed.

The dual specificity phosphatases (DSPs) 1 have emerged as important regulators of cell cycle control and mitogenic signal transduction. The ability to hydrolyze phosphotyrosine and phosphothreonine/serine residues efficiently and the lack of sequence homology outside the conserved active-site motif HCXXGXXRS(T) distinguish the DSPs from the protein-tyrosine phosphatases (PTPs). Several DSPs act as central regulators of the mitogen-activated protein (MAP) kinases (1). MAP kinases play a central role in mediating intracellular signaling events triggered by mitogens, growth factors, and stress (2,3). The MAP kinases are activated by specific upstream dual specificity kinases (MAP kinase kinases, MKKs) through phosphorylation on both threonine and tyrosine residues in the TXY motif. Extracellular stimuli activate MAP kinases leading to the phosphorylation of an array of cellular substrates and nuclear transcription factors. In mammals, three distinct MAP kinase families have been studied extensively. The extracellular signal-regulated kinase (ERK) family is activated by growth and differentiation factors and by phorbol esters (4). The c-Jun NH 2 -terminal kinase (JNK) family (or stress-activated protein kinase (SAPK) and p38 MAP kinase family are activated by proinflamatory cytokines and environmental stress. Specific DS-PTPs have been shown to dephosphorylate both Thr and Tyr in the TXY motif of MAP kinases (5)(6)(7)(8). Studies have shown that these MAP kinase phosphatases (MKPs) display distinct in vivo substrate preferences for the various MAP kinases (8 -11). Of the proposed MKPs, one of the most selective enzymes is MKP3 (9) (also named rVH6 (12) and Pyst1 (8)). The NH 2 -terminal "noncatalytic" region of MKP3 was shown to display tight binding to its in vivo substrate, ERK (13). This NH 2 -terminal domain is distinct from the common catalytic core that resides at the COOH-terminal half of the protein. Not only does the NH 2 -terminal domain bind ERK protein independently, but in the context of the full-length protein, ERK binding activates the phosphatase activity by at least 35-fold (14). ERK phosphorylation was not required for either binding or activation. The physical basis for this dramatic activation has not been established.
Recently, the x-ray structure for the catalytic domain of Pyst1 was solved. The overall structure was remarkably similar to the previously published x-ray structure of the DSP VHR (15) and that predicted from modeling studies based on the VHR structure (16). The full-length VHR structure (15) represents the minimal catalytic domain shared with larger DSPs such as MKP3, MKP1, PAC1, which share 30% sequence identity to VHR within the catalytic domain but contain a second domain that is predicted to extend from the amino terminus. VHR has served as the archetypal DSP in detailed biochemical and structural studies (15,(17)(18)(19)(20)(21), and the deduced catalytic mechanism has served as a model for all DSPs. The thiolate of cysteine 124 in the signature motif HCXXGXXRS(T) is the nucleophile that attacks the phosphorus of substrate, transferring the phosphate group to the enzyme, forming a cysteinylphosphate enzyme intermediate. Catalysis is enhanced by conserved aspartic acid 92, which is positioned to act as a general acid by protonating the leaving group oxygen. Aspartic acid 92 is found on a separate loop structure that was shown to be flexible in several PTPs. In these PTPs, ligand binding induces the movement of the "general acid loop" toward the active site such that the conserved aspartic acid is in position to transfer a proton to the leaving group. In the recent Pyst1 structure, the proposed general acid loop was flipped 20 Å away from the active-site cleft. In this conformation, the proposed general acid (Asp-262) would not be expected to contribute to catalysis. Consistent with this idea, mutation of Asp to Asn (D262N) resulted in an enzyme that displayed activity similar to that of the wild type enzyme (22). More interestingly, the D262N mutant could no longer be activated by ERK protein. Collectively, these data suggested that ERK binding to MKP3/Pyst1/rVH6 induces general acid loop closure and the subsequent involvement of Asp-262 acting as the general acid catalyst.
To elucidate the mechanistic basis for the dramatic catalytic enhancement and provide direct evidence that ERK binding induces general acid catalysis by wild type MKP3, we have performed an extensive enzymatic comparison of the activated and unactivated form of MKP3. Using steady-state and rapidreaction kinetics, pH studies, and Brönsted analysis, we demonstrate that ERK-activated MKP3 utilizes general acid catalysis, enhancing the rate of P-O bond cleavage by 100-fold. Also, the catalytic activation on all reaction steps is evaluated.

MATERIALS AND METHODS
Reagents and Enzymes-All chemicals were of the highest grade commercially available. MKP3 was expressed and purified as described in (16), and ERK was purified as described previously in (23). The 8-fluoro-4-methylumbelliferyl phosphate substrate (24) was from Molecular Probes (Eugene, OR).
Assays-The assay buffer was a three-component system consisting of 0.1 M acetate, 0.05 M Tris, and 0.05 M bis-Tris). This buffer system maintains a constant ionic strength of 0.1 M throughout the entire pH range. To determine the kinetic parameters k cat and k cat /K m , the initial velocities were measured at various substrate concentrations, and the data were fitted to Equation 1. For the construction of the pH profiles, k cat and k cat /K m were determined at various pH values. The pH data were fitted to Equations 2-5 depending upon the shape of the profile. Data were fitted to Equation 1 using KinetAsyst (IntelliKinetics, State College, PA). Fitting of the pH-dependent data to Equations 2-5 was accomplished with nonlinear least squares fitting using the computer program Kaleidagraph (Abelbeck Software) for Macintosh. In Equations 2-5, C (or C 1 , C 2 ) is the pH-independent value of either k cat or k cat /K m ; H is the proton concentration; K a , K b , and K c are the ionization constants.
Inhibition by Phosphate-The inhibition constant for phosphate, K i , was determined for MKP3 in the following manner. At various fixed concentrations of inhibitor, the initial velocity at different pNPP concentrations was measured as described previously (18). The inhibition was competitive with respect to substrate, and the data were fit using to Equation 6 to yield the inhibition constant.
Rapid-reaction Kinetics-Enzyme and substrate were rapidly mixed at 25°C in a temperature-controlled SF-61 Hi-Tech Scientific (Hi-Tech Ltd., Salisbury, U. K.) stopped-flow spectrophotometer. Product formation of p-nitrophenol was monitored at 410 nm, 3-o-methylfluorescein at 477 nm, 6,8-difluoro-4-methylumbelliferone at 360 nm, 8-fluoro-4methylumbelliferone at 360 nm, and 4-methylumbelliferone at 370 nm. The data were fitted to Equation 7 using the nonlinear least squares fitting capability of the kinetics software (KinetAsyst), where A is the amplitude of the burst, k is the first-order rate constant of the burst, B is the slope of the linear portion of the curve, C is the intercept of the line, and t is time.
The burst rate and the linear rate of the stopped-flow trace were determined as a function of substrate concentration. Each kinetic trace was fitted to Equation 7. The resulting burst rate and the slope rate were then fitted to Equations 8 and 1, respectively. The resulting (k 3 ϩ k 5 ) value from Equation 8 is essentially the maximal rate constant obtained from the plots of k burst versus [S]. Using the maximal (k 3 ϩ k 5 ) value, the fit to Equation 9 yields the actual k 3 value. Equation 9 describes the observed amplitude (B) of the burst phase as a function of [E] 0 , k 3 , (k 3 ϩ k 5 ), K m , and varied [S].

RESULTS
pH Rate Analysis-In initial experiments, the amount of ERK required to activate MK3 maximally was determined by measuring the initial velocities at increasing concentrations of ERK and varying concentrations of the phosphatase substrate oMFP. Consistent with previous observations (14), we found that maximal activation resulted from an approximate 1:1 stoichiometry of MKP3 to ERK concentration (data not shown). To ensure maximal activation, the concentration of ERK in all experiments was 2-fold higher than the MKP3 concentration. To begin to probe the mechanism of activation, the k cat and k cat /K m pH profiles were generated for MKP3 (both in the presence and absence of the activator ERK), and the critical ionizations were determined. The apparent second-order rate constant k cat /K m describes the reaction between free enzyme and free substrate. Reflecting both substrate binding and catalysis, the k cat /K m pH profile yields the intrinsic pK a values of critical ionizations. The k cat parameter describes the unimolecular catalytic steps subsequent to enzyme-substrate complex formation and will therefore yield apparent pK a values of the complex.
Previously, pH studies of unactivated MKP3 had not revealed an ionization that must be protonated for activity, suggesting the lack of general acid catalysis (16). Based upon the precedent for general acid catalysis by the DSPs (20), we predicted that the pH profiles of activated MKP3 would display the ionization for general acid catalysis. The commonly employed phosphatase substrate pNPP was used to generate the pH profiles ( Fig. 1), allowing direct comparison with previously published data using this substrate (14,16).
Two dramatic features distinguished the pH profiles of activated and unactivated MKP3 (Fig. 1). The k cat /K m pH profile of activated MKP3 displayed an optimum value that was 48-fold higher than the optimum value observed in the absence of ERK ( Fig. 1a and Table I). Second, activated MKP3 displayed a critical ionization (pK a ϭ 7.6) from an amino acid residue that must be protonated for activity. In the k cat /K m profile generated by MKP3 alone, a critical ionization was not observed (Fig. 1a). Instead, there was a small drop in rate at high pH caused by a noncritical ionization with an apparent pK a value of 7.6. This ionization did not decrease the rate to zero, but rather only dropped the rate to a new plateau value. These data were fitted to Equation 3, which describes this profile. Below the pH optimum of ϳ7, the shape of the two k cat /K m profiles was nearly identical. Both k cat /K m profiles displayed two ionizations (pK a values of 5.1 and ϳ6.5) that must be unprotonated for activity. The acidic limb of these pH profiles was consistent with the results reported by Wiland et al. for MKP3 (16) and by Denu et al. for the DS-PTP VHR (20). The pK a value of 5.1 was assigned to the ionization of the substrate pNPP, which reacts with DSPs and PTPs as the dianion (26,27). The pK a of 6.5 represents the pK a of the cysteine nucleophile Cys-293 of MKP3 and is consistent with the low pK a value observed for the catalytic Cys-124 of VHR (20).
The k cat pH profiles (Fig. 1b) displayed the same general trend as the k cat /K m profiles; however, the magnitude of the activation was less pronounced. With activated MKP3, the k cat value was 10-fold higher at the optimum and displayed an ionization with a pK a value of 5.8, which must be unprotonated for activity, and an ionization with a pK a value of 8.1, which must be protonated for activity (Table I). Although the group that must be unprotonated (pK a value ϭ 6.0 Ϯ 0.1) was still observed, unactivated MKP3 displayed no steep drop in activity at high pH due to a critical ionization. Instead, the k cat values fell only slightly at high pH, and therefore the data were fitted to Equation 5. As concluded previously (16), the group with a pK a value of ϳ6 is the catalytic Cys-293. In activated MKP3, the group that must be protonated is consistent with the involvement of a general acid catalyst.
Leaving Group Effects: Brönsted Analysis-To provide evidence for general acid involvement induced by the binding of ERK to MKP3, a Brönsted analysis was carried out. It has been shown previously that the pK a value of the substrate leaving group (R-O Ϫ ) has little effect on the ability of wild type PTPs to catalyze the first chemical step, P-O bond cleavage and concomitant formation of the thiol-phosphate intermediate (Scheme 1). Because expulsion of the leaving group is facilitated by proton donation from the conserved aspartic acid to the leaving group oxygen, the leaving group pK a has little effect on catalysis (26 -28). However, when aspartic acid mutants were analyzed, large linear free energy relationships between leaving group pK a values and the rates of catalysis were observed (28).
If general acid catalysis is not employed, one would predict that activity will decrease as the pK a of the leaving group increases. On the other hand, if MKP3 utilizes general acid catalysis, the leaving group dependence would be greatly diminished. The k cat and k cat /K m values were determined for substrates whose leaving group pK a values ranged from 4.6 to 10. The k cat data are plotted as the log of k cat versus the pK a value of the leaving group (Fig. 2a). In the absence of activation by ERK, the k cat value displays a small dependence on the leaving group pK a value, with a slope value of Ϫ0.08 Ϯ 0.03. In contrast, the k cat value of activated MKP3 displays no significant pK a effect (slope ϭ 0.013 Ϯ 0.017). It should be noted that leaving group dependence on k cat values can be kinetically masked because k cat values are a function of all first-order steps in the reaction. For DS-PTPs and PTPs, k cat is composed of the individual rate constants of intermediate formation, k 3 and intermediate hydrolysis, k 5 (Scheme 1), and is equal to k 3 *k 5 /(k 3 ϩ k 5 ). The lack of a substantial pK a effect in the activated enzyme could be caused by a change in the rate-   Fig. 2b.
To correct for any inherent differences in K m values between different substrates, the ratio of activated to unactivated k cat /K m values is plotted. In essence, the ratio reflects the magnitude activation (fold increase) as a function of leaving group pK a . The resulting plot displays a substantial linear relationship (slope ϭ 0.34 Ϯ 0.9) between pK a values 4.6 and 7.8. However, above pK a values of 7-8, there is a break in the plot where the magnitude of the activation reaches a plateau value of 70-fold. With very good leaving groups (i.e. low pK a values), ERK-induced activation results in only ϳ6-fold catalytic rate enhancement. However, with poorer leaving groups, activated MKP3 catalyzed P-O bond cleavage 2 orders of magnitude more efficiently than unactivated enzyme. These data suggest that ERK functions to enhance the rate of phosphoenzyme intermediate formation and that this may be the direct result of general acid catalysis. To elucidate the individual rate enhancement on each discrete step in catalysis and to provide evidence that ERK induces general acid involvement during the event of P-O bond cleavage, rapid-reaction kinetic analyses were performed.
Rapid-reaction Kinetics-With the DSPs VHR and CDC25, rapid-reaction kinetic experiments have proven extremely useful in resolving the rates for both intermediate formation and hydrolysis (19,29). When rapid biphasic kinetics were observed, it was demonstrated that the rapid exponential rate (burst phase) corresponds to intermediate formation (k 3 in Scheme 1), whereas the slower linear phase corresponds to intermediate hydrolysis (k 5 ). Here, kinetic isolation of these steps in MKP3 catalysis would provide a direct measure of the intrinsic rate enhancement upon each chemical step in catalysis. Therefore, MKP3 and substrate pNPP were mixed rapidly in a stopped-flow spectrophotometer, and the production of pNP was monitored at 410 nm. The resulting kinetic traces displayed biphasic characteristics, although the two phases were not well resolved (Fig. 3a). A slow exponential burst phase was followed by a slower linear steady-state rate. The kinetic traces were fitted to Equation 7 to yield the rate constants for the first-order burst and the steady-state phase. Because the two rates were not well resolved, both phases are a function of both individual rate constants k 3 and k 5 . To obtain a complete complement of individual rate constants, the rapid-reaction rates were determined as a function of substrate concentration (Fig. 3b). The linear portion (steady state) of the traces were fitted to Equation 1 to yield the steady-state parameters k cat , k cat /K m , and K m . These values were in excellent agreement with those values obtain from the steady-state analysis (Figs. 1  and 2). The calculated k burst values were fitted to Equation 8 to yield (k 3 ϩ k 5 ) and K d values. When k 3 Ͼ Ͼ k 5 , then the (k 3 ϩ k 5 ) term simplifies to k 3 , the rate constant for intermediate formation. The results are listed in Table II. From the analysis, k 3 ϩ k 5 ϭ 0.08 Ϯ 0.01 s Ϫ1 , and the K d value for pNPP was 34 Ϯ 10 mM. Using the (k 3 ϩ k 5 ) value, a fit to Equation 9 yielded the actual k 3 value of 0.07 s Ϫ1 (Table II) The rapid kinetics of MKP3 and pNPP were repeated in the presence of ERK protein. In these experiments, a solution of MKP3 and ERK was mixed rapidly with pNPP, and product formation was followed as discussed above. The resulting kinetic traces displayed a rapid exponential burst phase followed by a well resolved slower linear phase (Fig. 3a). To obtain a complete set of kinetic constants, the kinetic traces were determined as a function of pNPP concentration (Fig. 3b). The subsequent fitting analysis yielded a first-order rate constant of 4.46 s Ϫ1 and a k cat value of 0.45 s Ϫ1 . Amazingly, the first-order rate was 57-fold higher, and the linear rate was 14-fold higher when MKP3 was activated by ERK. Because the k burst and the k cat values were well resolved with activated MKP3, k burst ϭ k 3 and the k cat ϭ k 5 at saturating levels of substrate (Table II). The calculated k cat values were in good agreement with the k cat values obtained in the steady-state analysis (Table I).
We predicted that if general acid catalysis is employed by the activated phosphatase, then substrates with very low leaving group pK a values should display more comparable P-O cleavage (k 3 ) rates between activated and unactivated enzyme. To explore this and to analyze fully the rate enhancement on all catalytic steps in the reaction, oMFP (3-o-methylfluorescein pK a ϭ 4.6) was rapidly reacted with either MKP3 alone or MKP3 in the presence of ERK. At saturating levels of oMFP, representative traces are displayed in Fig. 4a. Simple inspection of the two reactions revealed that the burst rates were more comparable between activated and unactivated MKP3 than those observed with pNPP as subtrate (pNP pK a ϭ 7.1). Also, burst and linear phases were well resolved even with unactivated MKP3. This is because hydrolysis of the phosphoenzyme intermediate is rate-limiting with oMFP as substrate, allowing clear kinetic isolation of the two chemical steps in the reaction. The rapid kinetic traces were determined as a function of oMFP concentrations, and the resulting burst rate constants at increasing concentrations of oMFP are plotted in Fig. 4b. With this complete analysis, all of the pertinent kinetic constants (K d , k 3 , and k 5 ) could be determined and compared between activated and unactivated MKP3. The results are summarized in Table III. There was only a 2.2-fold rate enhancement on intermediate formation (k 3 ) when oMFP was employed as substrate. A significantly larger activation of 5-fold was observed on the rate of intermediate hydrolysis, k 5 . In excellent agreement between the two substrates pNPP and oMFP, the activated k 5 values of 0.72 s Ϫ1 and 0.66 s Ϫ1 , respectively, are consistent with this substrate-independent step. It should be noted that this slow step (k 5 ) in catalysis may not reflect intermediate hydrolysis, but rather may reflect the release of enzyme-bound phosphate. Although we do not have direct evidence, two observations suggest that intermediate hydrolysis is rate-limiting in MKP3. First, the K i for phosphate (competitive versus substrate) inhibition of MKP3 ranges from 1.5 to 28 mM. This high value indicates that MKP3 has a low affinity for phosphate and therefore suggests a rapid dissociation of phosphate from the enzyme. Also, in the related enzyme VHR, k 5 was shown to correspond to the slow hydrolysis of the intermediate (28). Because the rate of intermediate hydrolysis is independent of substrate, we could use the unactivated k 5 value of 0.13 s Ϫ1 from oMFP and the unactivated k cat value from pNPP to solve for the unactivated k 3 value for pNPP (using Equation 10). k cat ϭ k 3 *k 5 /͑k 3 ϩ k 5 ͒ (Eq. 10) With pNPP, the unactivated value for k 3 was calculated to be 0.044 s Ϫ1 . Compared with the ERK-activated k 3 value of 4.46 s Ϫ1 , this represents a rate enhancement of 101-fold. Utilizing the rapid-reaction approach to obtain the rate of P-O bond cleavage (k 3 ) for several different substrates, the k 3 value was determined as a function of leaving group pK a value (Fig. 5). Consistent with the steady-state Brönsted analysis of k cat /K m values, the k 3 value displayed a large dependence (slope ϭ Ϫ0.58 Ϯ 0.14), whereas the k 3 value of activated MKP3 displayed no significant effect (slope ϭ Ϫ0.13 Ϯ 0.13) on leaving group pK a value. Because of the lack of significant absorbance at pH 7, substrates with leaving group pK a values above 7.8 could not be analyzed by this stopped-flow approach.
Collectively, the results are fully consistent with activated MKP3 utilizing general acid catalysis during P-O bond cleavage. In several PTP x-ray structures, it has been observed that the general acid loop (i.e. the loop structure harboring the general acid) can occupy two distinct conformations. Ligand binding is thought to induce the "closed" or active conformation, bringing the conserved aspartic acid into position to act as the general acid. In the "open" conformation (inactive), the loop is flipped out away from the active-site cleft, and the aspartic acid cannot participate in the reaction. Interestingly, the x-ray structure of the catalytic domain of Pyst1 (ortholog to MKP3) revealed a markedly open conformation. Presumably, this conformation would render the enzyme incapable of general acid catalysis. Logically, we proposed that ERK induces (or stabilizes) general acid loop closure, thereby permitting efficient general acid catalysis. To test this hypothesis further, we attempted to mimic artificially the effect of ERK activation by stabilizing the closed conformation. This was accomplished by the addition of the organic solvent dimethyl sulfoxide (Me 2 SO). Reagents such as Me 2 SO have been referred to as "chemical chaperones" because of their ability to stabilize certain protein conformational states.
Substrate saturation curves were generated in increasing amounts of Me 2 SO (Fig. 6). Me 2 SO had a dramatic stimulatory effect on the kinetics of pNPP hydrolysis, with a maximal effect occurring by 25% Me 2 SO. The k cat /K m and k cat values were 42-fold and 2.2-fold higher, respectively, than those values obtained in the absence of Me 2 SO. This level of activation was very similar to that observed with ERK. To examine whether the same mechanism of activation was utilized by both ERK and Me 2 SO, ERK was added to samples containing either par-  Table II tially (15%) or fully (30%) Me 2 SO-activated MKP3. Addition of ERK to fully activated MKP3 had no further effect on the kinetics, but, the addition of ERK to partially (Me 2 SO 15%) activated MKP3 resulted in additional rate enhancement that reached the same overall activation as 30% Me 2 SO alone. Rapid kinetic studies in Me 2 SO confirmed that the dramatic rate acceleration occurs at the step of intermediate formation. 2 Taken together, these data strongly suggest that Me 2 SO and ERK activate MKP3 through the same mechanism of inducing the general acid loop closure.
To provide additional evidence for this proposal, the ability of MKP3 to bind the competitive inhibitor phosphate was assessed. In the closed conformation observed in the VHR crystal structure, the general acid loop makes two key interactions that allow efficient oxyanion binding in the active-site cleft. First, the general acid Asp-92 hydrogen bonds to one of the oxygens of ligand. Second, the backbone carbonyl of Ala-90 hydrogen bonds to the catalytic Arg-130, positioning this critical arginine toward direct electrostatic interactions with two of oxygens of phosphate. Therefore, we predicted that induced loop closure by ERK or Me 2 SO would optimize the active site for binding oxyanions, such as the competitive inhibitor phosphate. The inhibition constant K i of phosphate was determined for MKP3 alone, ERK-activated MKP3, and Me 2 SO-activated MKP3. With MKP3 alone, the K i value was 28 Ϯ 2 mM. The K i values with ERK-activated MKP3 and Me 2 SO-activated MKP3 were 1.5 Ϯ 0.1 mM and 4.8 Ϯ 0.7 mM, respectively. Both ERK and Me 2 SO significantly increased the affinity for ligand binding to MKP3. Consistent with these findings, both K m and K d values for substrates were significantly lower in ERK-activated MKP3 (Tables II and III). DISCUSSION When the enzymatic activity of MKP3 was first characterized (16), it was quite surprising that the rates of catalysis were almost 3 orders of magnitude lower than the related DSP VHR. It was also curious that MKP3 lacked the critical ionization for a group that must be protonated for activity (Ref. 16 and this paper). The pH profiles of unactivated and activated MKP3 bore a remarkable resemblance to the pH profiles of wild type VHR and the general acid mutant VHR D92N (20), respectively, both in the magnitude of the general acid rate enhancement and the loss of this ionization in the pH profiles. The D92N mutant of VHR was ϳ100-fold less active than native enzyme and exhibited the loss of the basic limb in the pH profiles. We had suggested that MKP3 was not employing general acid catalysis by the proposed general acid Asp-262 (16). We reasoned that Asp-262 was not positioned properly to aid in catalysis; however, why this was the case was not clear. In 1998, Camps et al. (14) discovered that MKP3 could be activated catalytically by binding to ERK, MKP3's proposed authentic substrate. The noncatalytic amino-terminal domain was believed to play a role in this activation because this domain could bind ERK independently and was necessary to induce the dramatic activation of MKP3. The recent x-ray structure of the catalytic domain of Pyst1 (MKP3) revealed an overall structure (22) that was very similar to VHR (15). However, the proposed general acid loop was flipped 20 Å away from the active-site cleft. In this conformation, the proposed general acid Asp-262 would not be expected to contribute to catalysis. Consistent with this idea, mutation of Asp to Asn (D262N) resulted in an enzyme that displayed activity similar to that of the wild type enzyme (22). Moreover, the D262N mutant could no longer be activated by ERK protein. Collectively, these observations suggested that ERK binding to MKP3 may induce general acid loop closure and the subsequent involvement of Asp-262 acting as general acid catalyst. The rationale for the present study was to pinpoint which steps in catalysis undergo activation and to provide evidence for direct general acid catalysis by wild type activated MKP3.
We have not only provided direct evidence that ERK induces efficient general acid catalysis during phosphoenzyme intermediate formation, but we have identified and quantified the activation on all major catalytic steps in the reaction. The largest catalytic activation occurs on intermediate formation (k 3 , Scheme 1), where ERK induces 2 orders of magnitude rate enhancement. ERK also induces 5-6-fold activation on intermediate hydrolysis (k 5 , Scheme 1). The apparent K d values are 2-3-fold lower in the presence of ERK. Consistent with higher affinity for substrate, the oxyanion phosphate binds 19-fold tighter to ERK-activated MKP3 than to MKP3.
The pH dependence of activated MKP3 clearly established the involvement of general acid catalysis (Fig. 1). In contrast, unactivated MKP3 did not display this ionization as critical for the low activity form of the enzyme. Instead, only slight decreases in rates were observed when Asp-262 is deprotonated. These results can be explained by a model in which ERK binding induces the movement of the general acid loop into its catalytically competent position. In the active conformation, the Asp must be protonated to act as a general acid. In its ionized form, Asp-262 is incapable of stabilizing the developing negative charge on the leaving group; and, if the negative charge on unprotonated Asp-262 is positioned at the active site, this will generate charge repulsion between the dianion of the binding substrate. Consistent with the conclusion that ERK binding correctly positions the general acid loop and Asp-262, there is a greater increase in K m values (with increasing pH above the pK a for Asp-262) for activated versus unactivated MKP3 (Fig. 1). In the unactivated enzyme, the loop would be predominantly flipped out of the active site, resulting in only a small relative change in activity and binding when Asp-262 becomes ionized. However, because there is a slight change in activity caused by this ionization, Asp may be partially or inefficiently contributing to binding/catalysis in the unactivated enzyme.
With excellent leaving groups such as 3-o-methylfluorescein and 6,8-difluoro-4-methylumbelliferone, the rate of intermediate formation is similar between activated and unactivated MKP3 (Fig. 2). However, as the pK a increases, there is a large dependence on leaving group pK a for unactivated enzyme compared with activated enzyme (Fig. 2a). This result was observed in both the pre-steady-state (Fig. 5) and steady-state (Fig. 2) analyses and strongly suggests that unactivated MKP3 lacks efficient general acid catalysis. In contrast, ERK-activated enzyme is able to utilize general acid catalysis, transferring a proton to the leaving group oxygen such that no significant negative charge is built up during the transition state of P-O bond cleavage. Curiously though, k cat /K m activation by ERK reaches a plateau value of ϳ70-fold with increasing leaving group pK a value (Fig. 2b). Unfortunately, the pre-steadystate analysis could not be extended to substrates whose leaving group pK a value was greater than 7.8 (Fig. 5). The plateau in k cat /K m activation would not necessarily be predicted if the unactivated enzyme were completely devoid of any general acid catalysis. One might expect that the magnitude of activation would become increasing larger as the pK a value increases. These observations suggest that the rate of loop closure in unactivated MKP3 may contribute to the observed rates of intermediate formation. The data can be explained in the following manner. With good leaving groups, unactivated MKP3 does not require general acid catalysis and thus does not require loop closure. But as the leaving group becomes poorer, proton donation becomes increasingly important for catalysis. The k cat /K m values do not continue to drop because loop closure now limits the rate at which the proton can be donated to the leaving group.
Induced loop closure is also suggested by the Me 2 SO activation experiments. ERK's dramatic activation of MKP3 could be mimicked by the addition of 25% Me 2 SO (Fig. 6). The aprotic organic solvent Me 2 SO has been shown previously to behave as a chemical chaperone, stabilizing the cellular prion protein (30) and the E 2 conformation of the H ϩ -ATPase (31), and favoring a conformation in F 1 -ATPase which demonstrates slower catalysis and low affinity ATP binding (32). We surmised that if ERK could stabilize the active closed conformation of MKP3, then perhaps a solvent like Me 2 SO could also function as a stabilizing agent. Although Me 2 SO was functionally similar to ERK, the physical mechanism by which ERK activates the phosphatase is likely to be mediated through the specific binding to the noncatalytic NH 2 -terminal domain of MKP3. This domain has been shown to exhibit high affinity binding to ERK (13). The catalytic domain alone exhibits only small activation in the presence of ERK (14,22). Interestingly, the NH 2 terminus of MKP3 and other MKPs share significant sequence homology to regions found in the catalytic domain of the p80 cdc25 phosphatase (33). Recent x-ray analysis of the catalytic domain of p80 cdc25 revealed that these homologous regions are required to form the core domain (34) and are proposed to interact with the cyclin-dependent kinase CDK2, the in vivo substrate of p80 cdc25 .
The observed activation on both ligand binding and phosphoenzyme intermediate hydrolysis provides further evidence for ERK-induced general acid loop closure. In addition to direct general acid catalysis, the general acid loop appears to play several catalytic roles. One major role is to position the catalytic arginine. In the enzyme-substrate complex, negative charge on equatorial oxygens of the substrate phosphate are stabilized by a bidentate interaction with the conserved arginine and by the backbone amide N-H groups of the conserved active-site motif HCXXGXXRT(S) (Scheme 2). In the x-ray structure of VHR, the carbonyl oxygen of Ala-90 is hydrogenbonded to one of the guanidinium nitrogens of Arg-130 (Scheme 2). Interestingly, this interaction involving the carbonyl oxygen (two residues NH 2 -terminal of the general acid Asp) can be observed in other PTP structures where the general acid loop is in the closed conformation (15,28,35,36). This interaction helps to position the Arg for its function in phosphate binding and transition-state stabilization. The Pyst1 catalytic domain structure revealed a severely open general acid loop conformation, and the corresponding Arg-299 is positioned incorrectly to bind the oxyanion of substrate efficiently (22). Although there is no structure available for the closed conformation of Pyst1, we would predict that the carbonyl of Ile-262 will interact with and position Arg-299, creating an efficient pocket for phosphate binding. Consistent with the idea of the general acid loop positioning Arg-299, activated MKP3 displays higher affinities for oxyanions such as phosphate as well as for substrates.
We have demonstrated previously that the conserved Asp of VHR and PTP1 plays a role during intermediate hydrolysis (28). The Asp may function as a general base during intermediate hydrolysis, activating a water molecule to facilitate nucleophilic attack. Consistent with this mechanism and the involvement of Asp-262, intermediate hydrolysis is 5-6-fold faster for ERK-activated MKP3 (Tables II and III). Because the rate enhancement is significantly smaller than that observed for intermediate formation, this may indicate that the general acid loop is better able to adopt a closed conformation in the phosphoenzyme intermediate.
We provide detailed biochemical evidence that ERK activates MKP3 through the stabilization of the active phosphatase conformation, where the general acid loop efficiently participates in general acid/base catalysis, substrate binding, and transition-state stabilization. ERK binding acts as a switch in converting the low activity form to the activated form of MKP3. Regulation of PTPs through the utilization of this flexible general acid loop may be a general mechanism. Two inhibitory mechanisms have been proposed from observations of the crystal structures of SHP-2 (an SH2-domain PTP) (37) and RPTP␣ (a receptor-like PTP) (38). With SHP-2, an intramolecular interaction with its SH2 domain sterically blocks and inhibits the active site in the unliganded form. For RPTP␣, an inhibitory dimer is thought to form upon binding extracellular ligand. The structure revealed that the amino-terminal helix-turn-helix of one monomer is wedged into the active site of the dyad-related monomer. In both cases, these interactions lock the PTP domain into the open conformation by preventing closure of the general acid loop.