Inhibition of Mitogen-activated Protein Kinase Phosphatase 3 Activity by Interdomain Binding*

Mitogen-activated protein (MAP) kinase phosphatase 3 (MKP3) is a cytoplasmic dual specificity phosphatase that functions to attenuate signaling via dephosphorylation and subsequent deactivation of its substrate and allosteric regulator, extracellular signal-regulated protein kinase 2 (ERK2). Expression of MKP3 has been shown to be under the control of ERK2, thus providing an elegant feedback mechanism for regulating the rate and duration of proliferative signals. Previously published studies suggest that MKP3 might serve as a tumor suppressor; however, significantly elevated, rather than reduced, levels of this protein have been reported in early lesions. Because overexpression of this phosphatase is counterintuitive to a proposed tumor suppressor function, the observed cellular tolerance suggested a self-inactivation mechanism. Using surface plasmon resonance, we have provided direct evidence of physical interaction between the N- and C-terminal domains. Kinetic analysis using dimethyl sulfoxide to activate the C-terminal fragment in the absence of ERK2 showed that the isolated C-terminal domain had higher catalytic efficiency than the similarly activated full-length protein. Furthermore, when the isolated N-terminal domain was added to the activated C-terminal domain, a dose-dependant inhibition of catalytic activity was observed. The similarity between the KI and KD values obtained indicate that interdomain binding stabilizes the inactive conformation of the catalytic site and implies that the N-terminal domain functions as an allosteric inhibitor of phosphatase activity. Finally, we have provided evidence for oligomerization of MKP3 in pancreatic cancer cells expressing elevated levels of this phosphatase.

Feedback regulation of cellular kinases and their cognate phosphatases is a fundamental component in many signal transduction cascades (1)(2)(3). In mitogen-activated protein (MAP) 2 kinase signaling, an initiating stimulus from extracellular growth factors, stressors or damage, leads to a hierarchical signaling cascade that involves the serial activation of MAP-extracellular signal-regulated protein kinase/extracellular signal-regulated protein kinases (MEK/ERK), followed by negative regulation of the latter by an antagonistic phosphatase (4 -7). In the kinase/phosphatase pair that regulates the classical RAS-MAP kinase pathway, activation of extracellular signal-regulated kinase 1 or 2 (ERK1 or ERK2) leads to the concomitant up-regulation and activation of MAP kinase phosphatase 3 (MKP3), which limits or abrogates the ERK signal (8 -10).
Ultimately, the consequence of ERK dephosphorylation by MKP3 restrains cell cycling, which is an absolute requirement during the critical period of early embryonic development when growth factor stimulation leads to RAS-MAPK signaling, increased metabolic load, and cellular expansion at key developmental loci. These loci, which include the neural plate and limb buds, then develop through synchronized periods of increased cell proliferation and apoptosis (11)(12)(13), a cycle that is believed to depend on the synchronization of ERK signaling and abrogation through ERK-dependant MKP3 protein synthesis (14 -16). As a result, elevated levels of MKP3 generate the feedback signal needed for coordinated growth factor-induced cell division and, later, proper cell specification and differentiation (11-13, 15, 17, 18).
Akin to its role in regulating the cell cycle via regulation of ERK signaling, MKP3 may also be involved in RAS/MEK/ERKmediated oncogenesis. Indeed, MKP3 levels appear to correlate with the severity and outcome of pancreatic adenocarcinoma (19 -21). The proposal for a role for MKP3 in oncogenesis arises from the observation that some pancreatic adenocarcinomas, as well as other solid tumors, show an allelic loss at the Mkp3 gene locus (22)(23)(24). Interestingly, precancerous intraepithelial neoplasias have been observed to overexpress MKP3, suggesting that MKP3 may serve a protective or tumor suppressive function through regulation of cell growth prior to transformation into adenocarcinoma (25). This concept further appears to be supported by histological observations that MKP3 protein levels appear to be elevated in both mildly and severely dysplastic lesions but appear reduced in advanced carcinomatous lesions that overexpress phosphorylated ERK2 (19,20,23).
The physiological importance of MKP3 regulation of the RAS-MAPK signaling cascade is reflected in its unusually tight substrate specificity for ERK1 and ERK2. This specificity, which prevents binding to, and thereby cross-reactivity with, the stress and oxidative damage-induced c-Jun N-terminal kinase (JNK) and p38 MAP kinases, appears to be governed by the ERK-specific N-terminal binding domain of MKP3 (26) and by the allosteric activation of the MKP3 C-terminal catalytic domain (27,28). This allosteric activation of MKP3 has been well documented in vitro using p-nitrophenol phosphate (pNPP), a small-molecule phosphotyrosine analogue of its normal substrate, which allows for colorimetric determination of phosphatase activity. In the absence of ERK, MKP3 has little activity, but in the presence of ERK, the ability of MKP3 to dephosphorylate pNPP increases by nearly 100-fold. This increase in catalytic activity upon ERK binding has been attributed to a structural reorganization that triggers the closure of a flexible loop region near the MKP3 active site (26,27,29,30). In this model, loop closure repositions an aspartic acid (Asp-262), placing it in proximity with the active site cysteine and arginine residues (Cys-293 and Arg-299), which enables the acid-catalyzed hydrolysis of phosphoamino acids. Unfortunately, although the structures of MKP3 binding and catalytic domains have been determined (26, 30 -32), the structure of the full-length protein, in either its activated or nonactivated form, has not. However, a recent magnetization transfer NMR study using the MKP3 N-terminal binding domain and the PAC1 C-terminal catalytic domain (an MKP3 catalytic domain homologue) indicates that the MKP3 N-terminal domain interacts with an MKP3 catalytic domain homologue (31), suggesting binding between N-and C-terminal domains in the native, nonactivated form of full-length MKP3. The effect of this putative interdomain binding on the function of MKP3 has not yet been studied.
This article provides direct evidence of MKP3 self-association and determines the effect of this phenomenon on MKP3 function. Specifically, we report the determination of the dissociation constants for the binding between full-length MKP3 and truncated analogues (corresponding to the MKP3 binding and catalytic domains) using surface plasmon resonance (SPR). An ERK-free activity assay was developed to enable the kinetic analysis of the activated C-terminal catalytic domain in the absence of ERK and in the presence and absence of the N-terminal ERK-binding domain. These data show that the addition of the MKP3 N-terminal domain to its catalytic C-terminal domain leads to a concentration-dependant decrease in enzyme activity. These experiments further demonstrate a relationship between MKP3 self-association and catalytic function, indicating that in the absence of ERK, the MKP3 N-terminal domain binds and inhibits the activity of its C-terminal domain. In principle, this interaction improves the fidelity of MKP3 and prevents nonspecific dephosphorylation in the absence of ERK. Taken together, these data suggest an additional layer of functional control for MKP3. A model of MKP3 post-translational regulation, in which overexpression leads to the formation of inactive oligomers via interprotein interaction involving N-and C-terminal domains, is presented as an explanation for the maintenance of high levels of phosphorylated ERK in tumor cells (20).

MKP3 Recombinant Protein Production and Purification-
Wild-type MKP3 protein (MKP3:WT) for this study was generated as described previously (33). Similar techniques were used to produce a truncated MKP3 N-terminal domain (corresponding to amino acids 1-154; MKP3:NT) using forward (5Ј-GGA ATT CCA TAT GAT AGA TAC GCT CAG AC-3Ј) and reverse primers (5Ј-TAG TTT ACA TAT GTC ACG AGC  CGT CTA GAT TGG TC-3Ј). A truncated C-terminal MKP3 domain (corresponding to amino acids 155-381; MKP3:CT) was then generated using forward (5Ј-GGA ATT CCA TAT  GTG TAG CAG CAG CTC GCC GCC GTT G-3Ј) and reverse  primers (5Ј-TAG TTT ACA TAT GTC ACG TAG ATT GCA  GAG AGT CC-3Ј). Underlined sections show EcoRV cleavage sites that were used to clone the insert into a pET15b vector (Invitrogen). Cloning and sequence verification was carried out as previously described (33).
Verification of Recombinant Protein Identity-Samples of the total cellular protein from recombinant Escherichia coli cells expressing the MKP3:WT, MKP3:NT, or MKP3:CT proteins were separated by SDS-PAGE and transferred to Immobilon-P membrane (Millipore, Danvers, MA). To verify the identities of the immunodetected bands, MKP3-containing cell lysates were separated on a duplicate SDS-PAGE and stained with Sypro Ruby Red (Invitrogen). In-gel tryptic digests were performed on bands of interest, which were excised from the gel, diced to a uniform size (ϳ1 mm 3 ), and then mixed with acetonitrile. The gel slurry was dried and resuspended in 1 ml of 100 mM ammonium bicarbonate containing 10 mM dithiothreitol to cleave disulfide bonds. The cysteine residues were then capped by the addition of iodoacetamide (50 mM final concentration). Gel pieces were washed and rehydrated in 1 ml of 50 mM ammonium bicarbonate, and the proteins were digested by the addition of trypsin (20 g). Peptide fragments were identified using an LTQ-FT mass spectrometry system (Thermo Finnigan, Waltham, MA) coupled to an Agilent 1100 series HPLC using a Zorbax 300SB-C18 capillary HPLC column (Agilent Technologies, Santa Clara, CA). PEAKS software (Bioinformatics Solutions, Waterloo, Ontario, Canada) was used to process the mass spectrometry data. The purity of MKP3 preparations after isolation and refolding was estimated by 12% SDS-PAGE (34) using a Mini-Protean 3 cell (Bio-Rad) and staining by Coomassie Blue. The MKP3 protein band density was assessed as a percentage of the total lane density using a Gel-doc XR with Quantity One 4.5.2 software (Bio-Rad).
CD Spectroscopic Verification of Refolding-CD spectroscopy experiments were performed using a J-810 spectrometer (Jasco, Easton, MD). Protein solutions in TBS, pH 7.6, were adjusted to concentrations between 0.5 and 1 mg/ml and then placed in a round cuvette with a 0.02-cm path length. Data were collected over a wavelength range of 178 -260 nm. A total of five scans were acquired, and the average was used for final analysis. Because of interference at lower wavelengths, the CD spectra were cropped to include only the region encompassing 190 -260 nm before analysis using DICHROWEB (35,36) and CDSSTR (37).
Enzyme Kinetic Analysis Using pNPP-Specific activities of MKP3:WT and MKP3:CT were assessed by nitrophenol phosphate assay in a 96-well plate. Assay buffer was prepared containing 20 mM pNPP in a 10 mM Tris, 50 mM NaCl solution at pH 8.0. A 75-l aliquot of assay buffer was added to each well together with 14 g of MKP3 in 25 l of TBS. MKP3 was assayed in the presence of dimethyl sulfoxide (DMSO), which has been shown to induce an activity in wild-type MKP3 similar to the effect of ERK2 binding (27,33). Concentrations of DMSO ranging from 0 to 50% (v/v) were tested to determine an optimum concentration. Steady state kinetic analysis was performed using TBS buffer containing 0 Inhibition was determined using MKP3:NT added to MKP3:WT and MKP3:CT in 33% DMSO containing 10 mM Tris and 50 mM NaCl with 0.6 mM pNPP as a substrate. Each assay contained 1 M MKP3:WT or MKP3:CT in a total assay volume of 150 l. The amount of MKP3:NT was increased with concentrations ranging from 0 to 2.2 M. IC 50 , the concentration of inhibitor that led to a 50% decrease in activity, was determined by curve fitting to a sigmoidal dose-response curve. K I was determined by the equation K I ϭ IC 50 /[1 ϩ (substrate concentration)/(substrate K D )] using GraphPad Prizm. Statistical analysis on MKP3 kinetic data were performed using an average of three independent experiments, with each experiment comprising three determinations. Student's t test was used to determine significance at a p value of Ͻ0.05.
Detection of MKP3 Oligomerization-Human HPDE6 immortalized duct and pancreatic adenocarcinoma cell lines were lysed using a nondenaturing radioimmunoprecipitation assay buffer to retain protein complexes. Cell lysates and purified recombinant MKP3:WT and MKP3:CT were diluted in a modified Laemmli loading buffer, which contained reducing agent but not SDS. Samples (10 g) were loaded, without prior heating, onto 5% stacking, 7.5% separating polyacrylamide mini-gels. Electrophoresis, Western transfer, and detection were performed as described above.
Detection of Mkp3 RNA-Ten micrograms of total RNA isolated and purified using TRIzol (Invitrogen) were fractionated on a 1% agarose formaldehyde-MOPS gel and processed for Northern blot hybridization and autoradiography according to standard procedures. The blot was first probed with a 32 P-radiolabeled 1.25-kbp human MKP3 HindIII/XhoI cDNA fragment, stripped, and reprobed with a human ␤-actin cDNA.
Size Exclusion Chromatography-MKP3 monomer and oligomers were separated on an AKTA-FPLC system (GE Healthcare) using a Superdex G200 10/300 gel filtration column (GE Healthcare). 1 ml of 1 mg/ml purified recombinant MKP3 was injected at a flow rate of 0.5 ml/min. Transient Transfection and Inducible Expression of Epitopetagged MKP3-Log phase human pancreatic adenocarcinoma cell lines CRL1469 and CRL1682 (5 ϫ 10 5 viable cells/60-mm diameter dish) were co-transfected with 2.5 g each of form I pVgRXR and pInd/GS/MK3/V5 or pInd/V5-his/LacZ (control) plasmids (all constructs from Invitrogen) using Lipofectamine (Invitrogen). V5 epitope-tagged MKP3 or ␤-galactosidase expression was induced at 16 h post-transfection with 5 M pronasterone A. Cell lysates were prepared 48 h later and processed for Western blot hybridization and laser densitometry as described above. Epitope-tagged MKP3 was detected with a mouse monoclonal anti-V5 antibody (Invitrogen). Transfected cells were photographed under phase contrast (Zeiss Axiovert 10) using Ilford FP-4 black and white film. Photographs were scanned and sharpened using Adobe Photoshop.

RESULTS
Production of MKP3 Proteins-The integrity of the MKP3 proteins produced for use in this study was verified in several ways. Initial assessment was by Western blot and SDS-PAGE (Fig. 1, A  and B). The apparent molecular masses of the protein bands Secondary Structure Analysis Using CD Spectroscopy-Once the identity of the recombinant MKP3 proteins had been verified by size and mass spectroscopic analysis, protein folding was assessed by circular dichroism (CD) spectroscopy. The secondary structural features were then compared with previously published data for the N-terminal ERK-binding domain (amino acids 1-154; Protein Data Bank ID code 1HZM) and the C-terminal catalytic domain (amino acids 204 -347; Protein Data Bank ID code 1MKP) (30,31). CD of MKP3:NT and MKP3:CT showed well structured domains with helical and sheet contents generally consistent with published data (30, 31) (Fig. 2). These data show that the MKP3 proteins were successfully produced and refolded and thus suitable for use in subsequent experiments.
Surface Plasmon Resonance-The affinities between MKP3 and ERK were determined using surface SPR as further verification that the MKP3 protein produced was correctly folded and functional. Experiments were performed using sensor chip-immobilized MKP3:WT with the in vivo substrate, ERK2, to test for protein-protein interactions. JNK1 was used as a negative control. SPR data indicated that MKP3:WT and ERK2 associated with k a ϭ 6.0 ϫ 10 6 M Ϫ1 s Ϫ1 and K D ϭ 2.63 ϫ 10 Ϫ7 M (Table 1), which was comparable to the previously reported value obtained via enzyme assay (K D ϭ 1.7 ϫ 10 Ϫ7 Ϯ 0.4 ϫ 10 Ϫ7 M) (38). In contrast to MKP3-ERK2 binding, JNK1 showed no detectable association (k a and K D could not be determined using a 1:1 Langmuir binding curve), consistent with previous . Secondary structure analysis by circular dichroism spectroscopy. Folding of the recombinant MKP3 proteins was assessed using circular dichroism spectroscopy to determine the secondary structure composition. Data shows the residue-weighted CD spectra traces from MKP3:NT, MKP3:CT, and MKP3:WT. The spectra were deconvoluted using DICHROWEB as described under "Experimental Procedures" to obtain the average secondary structure content from each recombinant protein. reports that MKP3 and JNK1 do not interact (10). These SPR data represent the first direct demonstration of MKP3:WT-ERK2 binding and confirm previous reports that assessed binding indirectly using enzymatic assays (26). Taken together with the CD data, these results provide further confirmation that MKP3:WT produced for this study is correctly folded and functional.
Previous studies have shown that the catalytic domain of PAC1 (a close structural homologue of the MKP3 C-terminal domain) binds to the N-terminal domain of MKP3 (31). Thus, this series of experiments was designed to examine directly the putative binding between the MKP3 N-and C-terminal domains. Samples of MKP3 proteins (MKP3:WT, MKP3:NT, MKP3:CT) were immobilized on the sensor chip, and MKP3:CT was added as an analyte. These data showed that although the MKP3 C-terminal domain possesses little ability to self-associate, it can bind both the full-length MKP3:WT and the N-terminal MKP3:NT proteins (Fig. 3, Table 2). These results show that the MKP3 C-terminal domain directly binds the N-terminal domain in the full-length MKP3 protein. Furthermore, the data indicating low micromolar values for K D suggest that, under normal physiological conditions, the N-and C-terminal domains exist in the bound conformation.
DMSO-induced MKP3 Phosphatase Activity-To measure MKP3:CT activity in the absence of ERK, a phosphatase assay was developed in which MKP3:CT was assayed in the presence of DMSO, a compound that has been shown previously to induce activity in wild-type MKP3 similar to that seen with ERK2 binding (27,33). pNPP was used as a colorimetric substrate. MKP3:WT and MKP3:CT both showed a dose-dependant enhancement of activity with increasing concentrations of DMSO (Fig. 4). The MKP3:CT activity peaked at ϳ40% DMSO and then declined, probably because higher concentrations of DMSO resulted in protein denaturation. These data are consistent with previous studies of full-length MKP3 that show maximal activation at 33% DMSO (27,33). Thus, a DMSO concentration of 33% was chosen for this assay, as it results in significantly increased activity with no protein denaturation.
After confirming that MKP3:CT could be activated using DMSO, steady state kinetic analyses were performed to compare the effects of ERK and DMSO on MKP3 activity. The results showed the expected low activity in the nonactivated These data indicate that increasing concentrations of MKP3:CT lead to an increase in the SPR plateau phase response expressed as resonance units (RU). These data were analyzed by curve fitting to a 1:1 Langmuir association to determine k a and K D values (Table 1).   MKP3 proteins were assayed using pNPP as a colorimetric substrate in the presence or absence of ERK2 and DMSO, which were used to activate the MKP3 phosphatase function. MKP3 proteins were tested using 0 -50 mM substrate (for nonactive MKP3 tests) or 0 -10 mM substrate (for ERK-or DMSO-activated tests). To detect MKP3 phosphatase activity, the hydrolysis of substrate was measured at 405 nm following a 60-min incubation at 25°C. To determine k cat and K m , the data were analyzed using GraphPad Prizm 4.0. The k cat for MKP3:WT increased 3-fold with the addition of ERK2 and 4-fold with the addition of DMSO (Table 2). In contrast, MKP3:CT showed no increase in k cat with the addition of ERK2 but a marked increase with the addition of DMSO. The K m values of MKP3:WT and MKP3:CT also showed differing responses with the addition of ERK2: K m of MKP3:WT decreased 5-fold with the addition of ERK2 but no further with DMSO, in comparison with MKP3:CT, which decreased slightly with the addition of ERK2, and decreased further with DMSO. These data are consistent with previous reports on the MKP3 C-terminal catalytic domain that show a similar 2-fold decrease in K m when the MKP3 catalytic domain was tested in the presence of ERK2 (30). Our current data also demonstrate, for the first time, that DMSO is able to induce activity in MKP3:CT similar to ERK or DMSO-based activation of the full-length MKP3.

Enzyme/Activator
Inhibition of MKP3:CT by MKP3:NT-After confirming our ability to assay the activity of activated MKP3:CT in an ERKfree system (using DMSO), the activity of MKP3:CT was determined in the presence of increasing concentrations of MKP3: NT. These data show that the addition of MKP3:NT led to a dose-dependant decrease in the specific activity of the MKP3:CT (Fig. 5). A similar addition of increasing amounts of MKP3:NT to DMSO-activated MKP3:WT led to a similar dosedependent decrease in activity. When these data were subsequently analyzed using a nonlinear regression fit to a sigmoidal dose response, the half-maximal inhibition of MKP3:CT by MKP3:NT (i.e. IC 50 ) was estimated to be approximately ϳ0.97 M. This IC 50 value was then used to calculate K I , which was determined to be 0.68 Ϯ 0.37 M. The value determined for the K I of MKP3:NT compares well with the SPR-determined K D for MKP3:NT and MKP3:CT binding. This indicates that although DMSO is capable of stabilizing the active structure of the MKP3:CT protein, MKP3:NT acts as an inhibitor of the C-terminal catalytic domain through interdomain binding.
Evaluation of MKP3, ERK1/2, and Phosphorylated ERK1/2 Expression in Pancreatic Adenocarcinoma Cell Lines-The ability of MKP3:NT domains to bind and inhibit MKP3:CT activity suggests that at high levels of MKP3 expression, the autoinhibitory effect of the MKP3 N-terminal domain results in a loss of enzymatic function. To explore the effects of MKP3 overexpression, a panel of pancreatic cells was tested for MKP3 levels using Western blot. MKP3 protein levels were evaluated in HeLa cells, normal human pancreatic tissue, immortalized (but not tumorigenic) pancreatic duct cell lines, and a series of pancreatic adenocarcinoma cell lines. This panel of cells included three adenocarcinoma cell lines used in a previously published study (20). Initial blots were probed with the same antibody (Santa Cruz Biotechnology, sc-8599) as that used in the previous study (20) and showed negligible expression of MKP3 (data not shown) as reported previously. In contrast, duplicate blots probed with a more recent, more robust and selective monoclonal antibody to MKP3 (R&D Systems, MAB3576) revealed that although MKP3 was detected at low levels in normal human pancreatic tissue, all other lines exhibited strong expression of MKP3 (Fig. 6A). We are in the process of resolving this apparent discrepancy by defining the amino acid residues involved in this interdomain binding. Preliminary evidence suggests that the epitope recognized by the sc8599 antibody may be masked as a result of interaction between the N and C termini. 3 Cellular expression of MKP3 was confirmed using Northern blot analysis to detect Mkp3 RNA transcripts in normal human pancreatic tissue, immortalized pancreatic duct cell lines, and adenocarcinoma cell lines (Fig. 6B). Human pancreas showed a lower level of Mkp3 RNA, but in seven of the nine adenocarcinoma lines, RNA levels were significantly elevated. These results correlate well with the protein levels detected in Fig. 6A. However, in cell lines CRL1420 and CRL1469, Mkp3 RNA levels had decreased relative to other pancreatic adenocarcinoma cell lines in the panel. Despite the lower RNA levels, the cells displayed high levels of MKP3 protein. These results suggest that MKP3 protein displays a long half-life within these cell lines.
In addition to MKP3, total and phosphorylated ERK1/2 was detected on parallel blots and juxtaposed with the MKP3 data (Fig. 6A). Densitometric analysis of the ERK data showed that despite up-regulated MKP3 in all pancreatic adenocarcinomas, the cells concurrently expressed high levels of total and phosphorylated ERK1/2. A consistent relationship between MKP3 levels and those of phosphorylated ERK1 and/or ERK2 could not be discerned.
Immunodetection  ERK (20) implied a loss of function. From Arkinstall and coworkers findings (26) and Hafen's model of cognate kinases and phosphatase (5), one would expect the presence of MKP3 to lead to a down-regulation of ERK expression/activity rather than to the high levels of ERK expression observed. To investigate this apparent discrepancy, MKP3 interdomain binding was examined by additional Western blots using samples of nondenatured pancreatic adenocarcinoma cell lysates and samples of purified recombinant MKP3:WT and MKP3:CT. The results in Fig. 7, A (7.5% cross-link PAGE separation) and B (immunoblot), show that significant amounts of cellular MKP3 were readily detectable in the adenocarcinoma cells. Also supporting previous surface plasmon resonance data is the finding that recombinant MKP3:CT was able to bind and form a dimer, detected as a band at ϳ60 kDa (Fig. 7B). Recombinant MKP3:WT (Fig. 7, A and B) was also able to form dimers (90 kDa) but was also found in higher molecular mass trimers (135 kDa). Presumably this occurs through interdomain binding, analogous to "three-dimensional domain swapping" (39,40) of the N and C termini from different MKP3:WT molecules. When recombinant MKP3 was compared with CRL1687 pancreatic adenocarcinoma cell lysate, an identical set of immunoreactive bands (corresponding to monomer, dimer, and trimer) was identified. This result demonstrates that MKP3 can exist in progressively oligomerized states when overexpressed in tumor cells.
MKP3 immunodetection was repeated on an expanded panel of pancreatic adenocarcinoma cell lines (Fig. 7C). Densitometric analysis revealed that MKP3 existed predominantly in an oligomerized state, and this oligomerized state was detected in excess of the monomeric species (Table 3).  Enzyme Activity of Oligomerized MKP3-We next wished to determine whether MKP3 oligomers were catalytically active. To this end, size exclusion chromatography was used to purify the oligomerized MKP3 for enzymatic activity assays. MKP3 monomeric and oligomeric forms were separated using gel filtration. The fractions that contained oligomerized MKP3 were tested by the pNPP hydrolysis assay. The oligomerized MKP3:WT showed no detectible phosphatase activity upon incubation with either DMSO or ERK2.
Expression of Epitope-tagged MKP3 in Pancreatic Adenocarcinoma Cells-The effect of elevated MKP3 production on pancreatic adenocarcinoma cells was tested by transient transfection and inducible expression of epitope-tagged MKP3. To confirm that the commercial Mkp3 vector construct (which contains a C-terminal V5 epitope) produced fully functional enzyme, the Mkp3/v5 cDNA cassette was excised and subcloned into a pET15b plasmid suitable for production in E. coli. The MKP3/V5 protein was then purified and tested for phosphatase activity using pNPP. The results indicated that the V5 epitope did not compromise ERK-binding or ERK-induced pNPP catalysis (Fig. 8A). Epitope-tagged MKP3 was next upregulated in the presence of pronasterone A following transient transfection into CRL1469 and CRL1682 cell lines. Despite high levels of MKP3 induction, total and phosphorylated ERK1/2 levels showed essentially no change when compared with the noninduced controls (Fig. 8B). To determine the phenotypic effect of MKP3 transgene induction, we surveyed cell morphology and general culture health by phase contrast microscopy (Fig. 8C) and noted that despite significant up-regulation of MKP3, pancreatic adenocarcinoma cell viability was unaffected. Transfection of epitope-tagged ␤-galactosidase (control) also had no effect on ERK or phosphorylated ERK expression and cell viability (data not shown).

DISCUSSION
In the currently accepted model, the regulation of MKP3 function is dependant upon the presence ERK1 or ERK2. When either ERK isoform is present, MKP3 undergoes a structural reorganization at its catalytic domain that leads to enzyme activation, ERK phosphorylation, and signal abrogation. The data presented here complement this model in an important way by suggesting an additional level of control of MKP3 activity: selfinhibition of MKP3 even in the absence of ERK (Fig. 9). Evidence suggesting interdomain binding in MKP3 has been reported previously by Farooq et al. (31) using the PAC1 C-terminal domain; they indicate that the N-terminal MKP3-binding domain was able to bind the C-terminal PAC1 catalytic domain with an approximate K D of ϳ100 M. The current study directly demonstrates that the C terminus of MKP3 can bind to the N terminus and shows that this binding is actually stronger than suggested by the PAC1 study (the dissociation constant for the MKP3:NT-MKP3:CT interaction was found to be ϳ70-fold lower than for the PAC1 interaction). This K D is still substantially higher (indicating weaker affinity) than the K D for the ERK2-MKP3:WT interaction. These data imply that although the MKP3 N-terminal domain can bind the catalytic domain when ERK is present, the ERK-MKP3 interaction is favored. Additionally, pNPP hydrolysis experiments were performed using an ERK-free assay that tested the activity of the    (27). Here we report the extension of this assay to the activation of MKP3:CT activity in the absence of either MKP3:NT or ERK. This modified assay allowed for the testing of the effect of the MKP3-binding domain (MKP3:NT) on the DMSOactivated C-terminal domain, an experiment that shows, for the first time, that the N-terminal domain plays a role in modulating MKP3 activity at its C-terminal catalytic domain. This role for the N-terminal domain in the modulation of C-terminal activity is further supported by the observation that MKP3:CT exhibits a small, but significant increase in k cat in comparison with the full-length MKP3:WT in the inactive and DMSO-activated states. These results indicate that the removal of the N-terminal noncatalytic domain is responsible for the higher rate of pNPP hydrolysis observed for basal and activated MKP3.
Consistent with these data is a model of inhibition of MKP3:CT by MKP3:NT. This inhibition can be attributed to stabilization of the inactive enzyme structure or to blockage of the MKP3:CT active site. In the presence of ERK2, competitive binding for the MKP3:NT between MKP3:CT and the higher affinity ERK2 ligand results in the release of MKP3:NT from the MKP3 catalytic domain, enabling the subsequent structural reorganizations that are necessary for high specific activity hydrolysis (Fig. 9A).
This proposed model is also consistent with the observation that MKP3 is found in abundance in many pancreatic carcinoma cell lines that exhibit constitutive ERK phosphorylation (Fig. 6). Under pathophysiological conditions, this model suggests that high levels of potentially tumor suppressive MKP3 phosphatase are constrained through interprotein binding and oligomerization (Fig. 7) in order to maintain levels of monomeric MKP3 suitable for allowing cells to cycle.
To be catalytically active, MKP3 must be present in the monomer form. Thus, the activity of any given C-terminal domain depends on two factors: that it is itself not bound to an N-terminal domain (whether inter-or intramolecular binding); and that it is allosterically activated through the binding of ERK to FIGURE 9. Models of ERK-induced activation and function in MKP3. The results of the binding and autoinhibition studies support a model of MKP3 deactivation due to oligomerization of MKP3. A, under normal conditions, the presence of ERK (purple) leads to competitive binding and a two-step MKP3 activation: displacement of the C-terminal domain (yellow oval) from the N-terminal domain (green oval) followed by allosteric activation of the C-terminal domain (yellow square) by the N-terminally bound ERK. The allosteric activation results in dephosphorylation of phosphotyrosine and phosphothreonine residues on ERK (red circles). B, if MKP3 is overexpressed, it undergoes interdomain binding, leading to the formation of high molecular mass oligomers. C, ERK binding to the N-terminal domain of the oligomerized MKP3 can release C-terminal catalytic domain, but this catalytic domain may be unable to become allosterically activated or to bind and, consequently, dephosphorylate ERK. Thus, the presence of oligomers adversely affects the ability of MKP3 to dephosphorylate ERK.
its own N-terminal domain. As a result of oligomerization, these two conditions are unlikely to be satisfied simultaneously (Fig. 9C).
We have shown through SPR that the N-terminal domain has a higher affinity for ERK than for the C-terminal domain. ERK would normally displace the C-terminal from the N-terminal, but at high MKP3 concentrations, N to C termini binding results in oligomerization. Although the preferential binding of ERK to any particular N-terminal binding domain releases a C-terminal catalytic domain, this catalytic domain remains inactive, as its own N-terminal domain is still not bound to ERK. This model of interprotein binding and oligomerization would suggest that although ERK binds to MKP3, the release of the MKP3 catalytic domain is unlikely to occur at a domain properly positioned to dephosphorylate the ERK active site residues (Fig. 9C).
The self-inactivation mechanism proposed by our study is similar to one recently described for Src homology 2 domaincontaining tyrosine phosphatase, which possesses an autoinhibitory domain that blocks the phosphatase catalytic site (41). It is also similar to a domain-domain interaction that was recently described for protein tyrosine phosphatase ␣ (PTP␣), where interaction between internal domains favors the stabilization of a dimeric state until ligand binding disrupts this interaction and leads to activation (42). Our study now suggests that a similar scheme might be applicable to other members of the dual specificity and multidomain phosphatases (43). We propose that dimerization/oligomerization of MKP3 could provide an effective means for restricting the attenuation of RAS-MAP kinase signaling to within physiologically acceptable limits. This added level of regulation complements the feedback loop elicited by downstream ERK2 activation and nuclear translocation (26). This proposal is consistent with significantly elevated MKP3 protein levels seen in the tumor cell lines used in this study and in the dysplastic pancreatic adenocarcinoma specimens reported previously (20). Thus, elevated MKP3 does not appear to exercise intrinsic tumor suppressor activity. Self-inactivation by oligomerization becomes all the more relevant in this context, as loss-of-function mutations in MKP3 have not yet been found in cultured cancer cells or clinical specimens (22).