ERK2 Shows a Restrictive and Locally Selective Mechanism of Recognition by Its Tyrosine Phosphatase Inactivators Not Shared by Its Activator MEK1*

The two regulatory residues that control the enzymatic activity of the mitogen-activated protein (MAP) kinase ERK2 are phosphorylated by the unique MAP kinase kinases MEK1/2 and dephosphorylated by several tyrosine-specific and dual specificity protein phosphatases. Selective docking interactions facilitate these phosphorylation and dephosphorylation events, controlling the specificity and duration of the MAP kinase activation-inactivation cycles. We have analyzed the contribution of specific residues of ERK2 in the physical and functional interaction with the ERK2 phosphatase inactivators PTP-SL and MKP-3 and with its activator MEK1. Single mutations in ERK2 that abrogated the dephosphorylation by endogenous tyrosine phosphatases from HEK293 cells still allowed efficient phosphorylation by endogenous MEK1/2. Discrete ERK2 mutations at the ERK2 docking groove differentially affected binding and inactivation by PTP-SL and MKP-3. Remarkably, the cytosolic retention of ERK2 by its activator MEK1 was not affected by any of the analyzed ERK2 single amino acid substitutions. A chimeric MEK1 protein, containing the kinase interaction motif of PTP-SL, bound tightly to ERK2 through its docking groove and behaved as a gain-of-function MAP kinase kinase that hyperactivated ERK2. Our results provide evidence that the ERK2 docking groove is more restrictive and selective for its tyrosine phosphatase inactivators than for MEK1/2 and indicate that distinct ERK2 residues modulate the docking interactions with activating and inactivating effectors.

Mitogen-activated protein (MAP) 6 kinases are key signaling Ser/Thr kinases, evolutionarily conserved from yeast to mammals and ubiquitously expressed in the organism, that are involved in the control of multiple and critical biological functions (1)(2)(3)(4)(5). The specificity of MAP kinase signaling is controlled by docking interactions, which provide high affinity binding between MAP kinases and their effector proteins, including substrates, activators (MAP kinase kinases), inactivators (phosphatases), and scaffolding proteins. These docking interactions improve the efficiency of the enzymatic reactions between the two molecules involved and are distinct from the transient interactions that occur during catalysis through the active sites of the enzymes. Therefore, docking interactions are crucial for the correct and specific recognition of MAP kinases by their functional partners. Two types of MAP kinase docking motifs have been identified in MAP kinase-interacting proteins. They are featured by clusters of basic and hydrophobic residues, with either an arginine-and leucine-, isoleucine-, or valine-rich motif (kinase interaction motif (KIM), D site, DEJL motif, ␦ domain) or a phenylalanine-rich motif (DEF motif). These motifs, alone or in combination, form a modular system that mediates molecular recognition by MAP kinases (6 -8).
The crystallographic analysis of several MAP kinases has shown that these proteins are compact enzymes with a two-domain structure (9 -11). The regulation of the kinase activity relies on the flexibility of the activation loop (loop L12) that, upon phosphorylation by specific MAP kinase kinases, refolds and activates catalysis (12,13). No crystal structure of any MAP kinase complexed to effectors is available. However, the crystal structures of p38␣ and JNK1 in complex with peptides that mimic docking motifs from specific interacting proteins have been determined (14,15). These studies, together with mutational analysis studies, highlighted the existence of several docking sites on different structural elements of the MAP kinase. In particular, studies with extracellular signal-regulated kinase 2 (ERK2) have defined an acidic patch on the surface-exposed loop L16 of the kinase opposite to its catalytic cleft, which acts as a MAP kinase conserved docking motif (CD site; residues 310 -325) common to substrates, activators, and inactivators (16 -21). Additional residues, located in loop L11 (ED site; residues 152-167), together with the CD site residues, have been proposed to form the ERK2 docking groove (20,22). Mutations of the Asp 319 and the Asp 160 residues within the loops L16 and L11 of the ERK2 docking groove, respectively, disrupt the docking site and are related with ERK2 gain-of-function phenotypes in Drosophila melanogaster and Saccharomyces cerevisiae (23)(24)(25). This suggests the existence of functional differences in the use of this recognition site by the distinct ERK2 effectors (17,18,26,27).
ERK2 is specifically activated by the MAP kinase kinases MEK1/2, whereas its inactivation can be achieved by several phosphatases, including serine/threonine-specific, tyrosine-specific, and dual specificity protein phosphatases (28 -33). In this report, we have investigated the involvement of ERK2 docking sites in the regulation of ERK2 by MEK1 and by the tyrosine phosphatase PTP-SL and the dual specific phosphatase MKP-3, which have high specificity toward ERK2. These three ERK2 regulators contain individual KIM-type docking motifs for ERK2 (34 -36), are able to anchor ERK2 in the cytosol (37)(38)(39), and are phosphorylated by ERK2 (36, 40 -42). The distinct and common features of MEK1, PTP-SL, and MKP-3 prompted us to compare the activity, binding, and cytosolic retention of ERK2 upon docking interactions with these three proteins. We have also studied the functional properties of a chimera of MEK1 that contains the KIM of PTP-SL, which behaved as a gain-of-function MAP kinase kinase in a docking groovedependent manner. Our results evidence a restrictive and locally selective mechanism of ERK2 interaction with its phosphatase inactivators that involves the ERK2 docking groove and is not shared by the ERK2 activator MEK1.
Immunodetections and primary and secondary antibodies used in this study have been described (20,36,38). The anti-GST rabbit polyclonal antiserum was obtained by immunization of rabbits with purified recombinant GST. To generate a polyclonal antiserum (anti-LexA) recognizing two-hybrid bait proteins expressed from pEG202-derived plasmids, a 640-bp LexA DNA binding domain-encoding HindIII-BamHI fragment was excised from pEG202. The fragment was subjected to a Klenow fill-in reaction to blunt the ends and cloned into SmaI-digested pGEX-1N. The resulting pGEX-LexA plasmid was introduced into DH5␣ strain, and GST fusion proteins were induced and isolated as described (48). Purified GST or GST-LexA proteins were used to immunize rabbits following standard protocols. For kinase assays, GST fusion proteins from bacteria were precipitated with glutathione-Sepharose; after washing with phosphate-buffered saline, the pellets were washed once with kinase reaction buffer and used as substrates in the kinase assays. GST fusion proteins expressed in HEK293 mammalian cells were precipitated with glutathione-Sepharose for 2 h; then the glutathione-Sepharose beads were pulled down by centrifugation and washed four times with HNTG buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 10% glycerol, and 0.1% Triton X-100). After a final washing with phosphate-buffered saline or with the appropriate reaction buffer, pellets were used as GST pull-down reagents ( Samples were resolved on 1% agarose gels. All primers are available upon request. Quantitative Yeast Two-hybrid Interaction Assays-The yeast strain EGY48 was cotransformed with pMW107 (lacZ reporter plasmid), pEG202-PTP-SL or pEG202-MKP-3 (lexA-DNA binding domain fusion proteins), and the empty pJG4-5 plasmid or pJG4-5-ERK2 wild type or mutations (B42 activation domain-HA fusion proteins) using the lithium acetate method and plated on appropriate selection medium. To determine the relative strength of the two-hybrid bait-prey interactions, ␤-galactosidase activity measurements were performed on crude yeast protein extracts obtained from liquid overnight cultures. Briefly, yeast cells were collected by centrifugation and resuspended in LacZ buffer (100 mM sodium phosphate, pH 7, 10 mM KCl, 1 mM MgSO 4 ). Lysates were obtained by adding 0.1% SDS and chloroform to the resuspended cells, followed by vortexing. One fraction was used to determine the protein concentration according to the Bradford method and to determine the relative expression levels of the various ERK2, PTP-SL, and MKP-3 bait and prey fusion proteins by immunoblot. Another fraction was processed in 96-well microtiter plates to assay for ␤-galactosidase activity. The enzymatic reaction was initiated by the addition of o-nitrophenyl ␤-D-galactopyranoside (ONPG) substrate (2 mg/ml final concentration) plus ␤-mercaptoethanol (25 mM final concentration). The hydrolysis of ONPG was monitored for 1 h at 30°C by measuring absorbance at 405 nm every minute. The ␤-galactosidase activity (arbitrary units) was calculated as the quotient between the slope of the linear part of the absorbance versus time plot and the total protein concentration. Experiments were performed analyzing simultaneously five separate transformant colonies for each bait and prey plasmid combination.
In Vitro Phosphatase and Kinase Assays-For in vitro phosphatase assays, EGF-activated HA-ERK2, immunoprecipitated from HEK293 transfected cells, was used as the substrate. HA-ERK2 pellets were washed with HNTG and subsequently once with phosphatase reaction buffer (25 mM HEPES, pH 7.5, 5 mM EDTA, 10 mM dithiothreitol). Pellets were resuspended in phosphatase reaction buffer, divided into three equal fractions, and incubated with increasing amounts of GST-PTP-SL or GST-MKP-3 (from transfected HEK293 mammalian cells) bound to glutathione-Sepharose under constant shaking. Phosphatase reactions and detection of phosphorylated HA-ERK2 was performed as described (20). For in vitro kinase assays, MBP or GST fusion proteins bound to glutathione-Sepharose, purified from bacteria, were used as the substrates. Activated HA-ERK2, GST-ERK2, and HA-MEK1, precipitated from transfected HEK293 cells, were used as the enzymes in the kinase assays. Assays with HA-ERK2 and GST-ERK2 were performed in kinase reaction buffer A (20 mM HEPES, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, 2 mM Na 3 VO 4 , and 0.3 M ATP) as described (20). Assays with HA-MEK1 were performed for 30 min at 30°C, with constant shaking, in kinase reaction buffer B (50 mM HEPES, pH 7.5, 10 mM MgCl 2 , 1 mM dithiothreitol, and 0.5 M ATP), in the presence of 5 Ci of [␥-32 P]ATP. All of the reaction mixtures were stopped by adding SDS sample buffer and boiling, and samples were resolved on SDS-polyacrylamide gels. Where indicated, phosphorylated bands were excised from the gel, and radioactivity was quantified using a scintillation counter.

Mutations at Different Structural Elements within ERK2 Confer Resistance to Dephosphorylation by Specific MAP Kinase Phosphatases-
Our previous studies on ERK2 regulation by the tyrosine phosphatases PTP-SL and STEP (20) revealed the importance of key charged residues at loops L11 and L16 (Asp 160 and Asp 319 , respectively; see Fig. 1A) of the docking groove in ERK2 binding and dephosphorylation by MAP kinase inactivators. However, only Asp 319 mutations (D319N and D319E) displayed a gain-of-function phenotype in terms of an increased kinase activity toward MBP, in comparison with ERK2 wild type. Intriguingly, mutation of the Asp 160 residue (D160N and D160E) also resulted in a defective recognition by the tyrosine phosphatase PTP-SL, but not in a gain-of-function phenotype (20). To further investigate this phenomenon, the Asp 319 and Asp 160 mutations were overexpressed in HEK293 cells as HA epitopetagged proteins, followed by EGF activation and cell lysis in the presence or in the absence of the tyrosine phosphatase inhibitor sodium vanadate. Then the recombinant proteins were immunoprecipitated with an anti-HA antibody and subjected to immunoblot with an anti-phospho-ERK1/2 antibody ( Fig. 2). In the absence of sodium vanadate (Fig. 2, minus lanes), dephosphorylation of wild type ERK2 by phosphatases present in the cell extract was readily achieved, whereas ERK2 dephosphorylation was not observed in the presence of sodium vanadate. All of the mutations tested were efficiently phosphorylated in intact cells by MEK1/2 (Fig. 2, plus lanes (containing sodium vanadate in the lysis buffer)). However, dephosphorylation of the D319N or D319E ERK2 mutations was not observed even in the absence of sodium vanadate (Fig. 2, lanes 1-6). These results indicate that sodium vanadate-sensitive phosphatases from the cell extract account for the dephosphorylation of the ERK2 activation loop in a docking interactiondependent manner. Interestingly, dephosphorylation of the mutation D160N, but not that of the charge-conservative mutation D160E, was also prevented in the absence of sodium vanadate. This suggests that a negative charge at position 160 may be important for the recognition of ERK2 by its phosphatase inactivators (Fig. 2, lanes 7-12). Thus, dephosphorylation of ERK2-specific mutations does not take place in cell lysates in the absence of sodium vanadate.
To identify additional residues in ERK2 involved in its dephosphorylation by tyrosine phosphatases, the same experiments were performed FIGURE 1. ERK2 single amino acid substitutions analyzed in this study. A, depiction of the crystal structure of ERK2 (9). The docking groove and the catalytic regulatory site (TEY motif) are indicated within circles. The mutated residues (pink) are indicated with arrows. B, mutations analyzed in this study. The structural elements where the mutated residues are located are indicated. A summary of the sensitivity of the mutations to MEK1/2 phosphorylation and to tyrosine phosphatase dephosphorylation is presented, compiled from the results from Fig. 2. ϩ, sensitivity similar to wild type; Ϫ, sensitivity diminished with respect to wild type. Amino acids are indicated using one-letter codes, and amino acid numbering is according to mouse ERK2 sequence (X58712). All of the ERK2 mutations from the present work, except the V171L mutation, did not disturb the kinase catalytic activity of the kinase upon EGF activation, using MBP as substrate, indicating that these mutations do not introduce gross conformational changes in the ERK2 catalytic site or in its overall structure (data not shown) (20).
using selected ERK2 mutations that targeted key charged residues at several structural elements within the docking groove, including Asp 316 and Glu 320 (loop L16), Glu 79 (loop L5), and Arg 133 (helix ␣e) (18, 20, 21) (see Fig. 1). Remarkably, phosphorylation of the ERK2 mutations E79A, R133A, D316N, E320A, and E320Q was preserved in the absence of sodium vanadate (Fig. 2, lanes [13][14][15][16][17][18][19][20][21][22][23][24][25][26][27][28]. Experiments were also performed using ERK2 mutations that targeted residues outside of the docking groove, including Gln 58 (loop L4), Val 171 (strand ␤9) and His 230 (loop L13). These residues are homologues to those that upon mutation cause a gain-of-function phenotype in S. cerevisiae. The mutations included E58Q, V171L, and H230N, which have also been proposed as putative gain-of-function mutations in mammalian ERK2 (29). Our previous studies did not detect an increase in kinase activity toward MBP in these mutations, and they were found to bind normally to PTP-SL (20). However, we have now found that the H230N mutation causes partial resistance to dephosphorylation by the pool of sodium vanadate-sensitive phosphatases from HEK293 cell extracts. This indicates a defective recognition of this mutation by specific phosphatases (Fig. 2, lanes 35  and 36). On the other hand, the E58Q and V171L mutations were efficiently dephosphorylated under the same conditions (Fig. 2, lanes 29 -34). A summary of the activation by MEK1/2 and the inactivation by endogenous tyrosine phosphatases of the different ERK2 mutations is shown in Fig. 1B. Together, these results demonstrate that mutations at different structural elements within ERK2 confer resistance to dephosphorylation by sodium vanadate-sensitive phosphatases without affecting their phosphorylation by MEK1/2 and suggest a potential role for these mutations in gain-of-function ERK2 phenotypes.
To ascertain the identity of the MAP kinase phosphatases that could be responsible of ERK2 dephosphorylation in our assays using HEK293 cell lysates, we performed a comprehensive analysis on HEK293 cells for mRNA expression of dual specificity and tyrosine-specific MAP kinase phosphatases. Semiquantitative RT-PCR analysis was performed using HEK293 RNA and specific primers for the 13 identified MAP kinase phosphatases (MKP-1, MKP-2, MKP-3, MKP-4, MKP-5, MKP-7, DUSP5, DUSP2, DUSP8, DUSP7, PTP-SL, HePTP, and STEP) (49). As shown in Fig. 3, mRNA message was detected for the dual specificity phosphatases MKP-2, MKP-3, MKP-7, DUSP2, and DUSP8, as well as for the tyrosine-specific, PTP-SL-related phosphatase, HePTP. Thus, this subset of MAP kinase phosphatases could account for the dephosphorylation of ERK2 in the HEK293 cell lysates in the absence of sodium vanadate.

Single Amino Acid Substitutions in ERK2 That Selectively Affect the Recognition by PTP-SL and MKP-3 Do Not Affect Recognition by the MAP Kinase Kinase MEK1-
The above results outline the important contribution of residues in the ERK2 docking groove in dephosphorylation by sodium vanadate-sensitive phosphatases. PTP-SL and MKP-3 are two ERK2-specific phosphatases that inactivate ERK2 upon binding to its docking groove (20,50). However, due to their distinct substrate specificity, PTP-SL only dephosphorylates the ERK2 regulatory tyrosine, whereas MKP-3 dephosphorylates both the threonine and the tyrosine regulatory residues (30,31). To compare the association of ERK2 with PTP-SL and MKP-3, we performed a quantitative yeast twohybrid analysis of the ERK2/PTP-SL and ERK2/MKP-3 interactions (Fig. 4), using the panel of ERK2 docking groove mutations shown above. To validate the quantitative and comparative analysis of the yeast two-hybrid interactions, the expression levels of the various hybrid bait and prey proteins in the yeast were monitored by immunoblot with anti-LexA and anti-HA antibodies, respectively (Fig. 4A). As shown, similar amounts of the different ERK2 mutations, as well as of the two FIGURE 2. Sensitivity of ERK2 wild type and mutations to endogenous tyrosine phosphatases. HEK293 cells were transfected with pCDNA3-HA-ERK2 wild type (wt) or mutations, as indicated. After 48 h, the cells were treated with EGF for 5 min, and then the cells were lysed with (ϩ) or without (Ϫ) 2 mM of Na 3 VO 4 (NaV) in the lysis buffer type A. HA-ERK2 wild type or mutations were immunoprecipitated (i.p.) from equal amounts of cell lysates using the anti-HA 12CA5 mAb. The samples were resolved on 10% SDS-PAGE and subjected to immunoblot with anti-phosphoactive ERK1/2 mAb (anti-ppERK1/2; upper panels). As controls of expression, total lysate samples were processed in parallel for immunoblot with the anti-HA 12CA5 mAb (lower panels). To allow comparison, HA-ERK2 wild type was included in each set of independent experiments. A representative experiment is shown of at least two experiments with similar results. phosphatases, were expressed in yeast. The relative strength of the ERK2/PTP-SL and ERK2/MKP-3 interactions was quantified by measuring the lacz gene reporter-dependent ␤-galactosidase activity in the yeast cultures (Fig. 4B). Most of the ERK2 mutations displayed greatly diminished association with both PTP-SL and MKP-3 (less than 20% with respect to ERK2 wild type), whereas the association of these two phosphatases with some ERK2 mutations, including D319N and E320A at loop L16, was undetectable. Although the pattern of association of PTP-SL and MKP-3 to the ERK2-docking groove mutations was similar, significant differences were found between the two phosphatases when the effect of particular ERK2 docking groove mutations was compared. In summary, mutations in loop L16, including conservative amino acid substitutions, were highly detrimental for association of ERK2 with both PTP-SL and MKP-3. However, the D160E conservative mutation at loop L11 affected only partially the association of ERK2 with both phosphatases in comparison with the nonconservative mutation D160N. Finally, mutations at the helix ␣e (R133A) and loop L5 (E79A) affected more the association of ERK2 with PTP-SL (1 and 9% of interaction, respectively) than that with MKP-3 (19 and 47% of interaction, respectively).
Since the strength of the association of the ERK2 E79A mutation with PTP-SL or MKP-3 was substantially different (9 versus 47% of interaction, respectively; see Fig. 4B), comparative in vitro dephosphorylation assays of this ERK2 mutation were performed using GST-PTP-SL and GST-MKP-3 recombinant proteins purified from HEK293 cells (Fig. 5). Activated HA-tagged wild type ERK2 or the E79A mutation were immunoprecipitated from HEK293 cell lysates and incubated with increasing amounts of the phosphatases, and samples were subjected to immunoblot with anti-phospho-ERK1/2 antibody. Remarkably, dephosphorylation of the E79A mutation by MKP-3 was similar to that of ERK2 wild type, whereas dephosphorylation of the E79A mutation by PTP-SL was severely impaired as compared with ERK2 wild type. Together, these results demonstrate that discrete residues within the different structural elements that configure the ERK2 docking groove (helix ␣e and loops L5, L11, and L16) contribute differentially to the recognition of ERK2 by its inactivators PTP-SL and MKP-3 and outline the importance of Glu 79 in loop L5 in the selective ERK2 binding and dephosphorylation by these two phosphatases.
Our results revealed that the single amino acid substitutions in ERK2 used in this study did not affect phosphorylation by endogenous MEK1/2 from HEK293 cells (see Fig. 2), suggesting that the recognition of these ERK2 mutations by MEK1/2 was not impaired. Under our experimental conditions, the binding of MEK1 to ERK2, as detected by the yeast two-hybrid technique, was too weak to allow a comparison with ERK2 mutant binding by PTP-SL and MKP-3 (data not shown). However, upon cotransfection of HA-ERK2 and GST-MEK1 on mammalian COS-7 cells, HA-ERK2 was efficiently retained in the cytosol by GST-MEK1, as observed with GST-PTP-SL and GST-MKP-3 ( Fig. 6 and TABLE ONE). ERK2 mutations D319N and E320A, which showed no residual binding to PTP-SL or MKP-3, were not retained in the cytosol by these phosphatases and were found in the cell nucleus. The mutation D319E was not retained in the cytosol by PTP-SL but was partially retained by MKP-3 in a significant number of cells. Also, the mutations E320Q, E79A, and D160N displayed partial cytosolic retention by both PTP-SL and MKP-3, whereas the mutation R133A was partially retained by PTP-SL and fully retained by MKP-3 ( Fig. 6 and TABLE ONE). In the majority of cells showing partial cytosolic reten-

. Comparative quantitative study of the binding of PTP-SL and MKP-3 to ERK2 docking groove mutations.
A, analysis of the co-expression of B42-HA-ERK2 wild type (wt) or mutations with LexA-PTP-SL or with LexA-MKP-3 in the yeast lysates of a representative experiment from data shown in B. The same amount of each cell lysate was resolved on 10% SDS-PAGE and subjected to immunoblot with anti-HA 12CA5 mAb (upper panels) or anti-LexA polyclonal antibody (lower panels). The empty vector lanes indicate lysates from yeast cotransfected with the lacZ reporter plasmid, the pJG4 -5 empty vector, and pEG202-PTP-SL (left panels) or pEG202-MKP-3 (right panels). B, quantitative yeast two-hybrid analysis of the binding of PTP-SL (left diagram) or MKP-3 (right diagram) to ERK2 wild type and mutations at the distinct structural elements (L16, ␣e, L5, and L11) of the ERK2 docking groove. Data are presented as the percentage of arbitrary units of ␤-galactosidase activity (the values are indicated on the top of each bar) with respect to the interaction with ERK2 wild type (100%). The values represent the mean of at least three separate experiments Ϯ S.D. FIGURE 5. Differential dephosphorylation of ERK2 E79A mutation by PTP-SL and MKP-3. HEK293 cells were transfected with pCDNA3-HA-ERK2 wild type (wt) or E79A mutation, as indicated. After 48 h, the cells were treated with EGF for 5 min and lysed with lysis buffer type B, and HA-ERK2 was immunoprecipitated (i.p.) from equal amounts of cell lysates with the anti-HA 12CA5 mAb. The immune complexes were subjected to in vitro phosphatase assay in the presence of increasing amounts of GST-PTP-SL or GST-MKP3 bound to glutathione-Sepharose, as indicated. GST fusion proteins were obtained by pull-down from lysates of HEK293 cells (lysis buffer type A, without Na 3 VO 4 ) transfected with pRK5-GST-PTP-SL or pRK5-GST-MKP-3. The reaction mixtures (final volume 50 l) were stopped by adding SDS sample buffer and boiling and resolved on 10% SDS-PAGE. The phosphoactive content of HA-ERK2 was measured by immunoblot with anti-phosphoactive-ERK1/2 mAb (anti-ppERK1/2; upper panel). The input of HA-ERK2 and GST fusion proteins was monitored by immunoblot with anti-HA 12CA5 mAb and with anti-GST polyclonal antibody, respectively (middle and lower panels). tion of ERK2, ERK2 staining was observed in both the cytosol and the nucleus (Fig. 6, K, P, Q, S, and T, and TABLE ONE). The rest of the ERK2 mutations (D316N, D160E, E58Q, V171L, and H230N) were kept in the cytosol by PTP-SL and MKP-3 in the majority of the cells. Remarkably, all the ERK2 mutations used in this study, including D319N and E320A, were efficiently retained in the cytosol by MEK1 (Fig. 6 and TABLE ONE). This indicates that single amino acid substitutions at the ERK2 docking groove that have a profound effect on ERK2 binding to inactivating tyrosine phosphatases are not detrimental for ERK2 binding to its activator MEK1 in intact cells.
A Chimeric MEK1 Protein with High Avidity toward ERK2 Displays Gain-of-function Properties-Both MEK1 and PTP-SL possess a KIM that contains basic and hydrophobic residues important for association with ERK2 (18,37,38). A comparison of the amino acid sequence of the MEK1 and PTP-SL KIM, however, reveals remarkable differences (Fig.  7A), suggesting that the distinct functional binding of ERK2 with these two effectors could be explained, at least in part, by differential recognition of their docking motifs. To investigate this, chimeric MEK1 molecules were engineered by substitution of the N-terminal KIM of MEK1 (residues 1-16) with the KIM of PTP-SL (residues 224 -239) (Fig. 7B) and tested for physical and functional association with ERK2. Remark-ably, the GST-MEK1 fusion protein containing the KIM of PTP-SL (GST-(KIM-SL)MEK1; the chimera lacks the first 16 residues of MEK1; see Fig. 7B) associated strongly with ERK1/2, as detected by pull-down assays, whereas the binding to ERK1/2 of the GST-MEK1 fusion protein (GST-MEK1; MEK1 as a full-length protein; see Fig. 7B) was undetectable (Fig. 8A, upper panel). Also, a GST-ERK2 fusion protein associated tightly, by pull-down assays, to an HA-tagged MEK1 molecule containing the KIM of PTP-SL (HA-(KIM-SL)MEK1), whereas its association to HA-tagged MEK1 (HA-MEK1) was undetectable (Fig. 8B). PTP-SL also binds to the MAP kinase p38␣ in a KIM-dependent manner (38). Interestingly, the chimeric protein GST-(KIM-SL)MEK1 did not bind to p38␣ (Fig. 8A, lower panel), indicating that the recognition specificity of MEK1 toward ERK1/2 does not solely depend on its N-terminal KIM motif. Also, the phosphorylation of MEK1 and (KIM-SL)MEK1 by active ERK2 (wild type or D319N mutation) was tested by in vitro kinase assays and compared with the phosphorylation of the docking-independent substrate MBP and the docking-dependent substrate PTP-SL (Fig.  8, C and D). As shown, ERK2 wild type phosphorylated (KIM-SL)MEK1 more efficiently than MEK1 (Fig. 8C), in correlation with the increased binding previously observed. The ERK2 D319N mutation, as expected from its gain-of-function catalytic activity (20), displayed more kinase activity toward the docking-independent substrate MBP than ERK2 wild type (Fig. 8D). Moreover, ERK2 D319N also phosphorylated MEK1 more efficiently than ERK2 wild type, whereas the phosphorylation of (KIM-SL)MEK1 by ERK2 D319N was slightly attenuated when compared with ERK2 wild type. Finally, the docking-dependent PTP-SL substrate was poorly phosphorylated by ERK2 D319N (Fig. 8D), as expected from the lack of association between PTP-SL and the ERK2 D319N mutation. Together, these results demonstrate that the KIM of PTP-SL confers high avidity to MEK1 for binding and being phosphorylated by ERK2 and indicate that Asp 319 is crucial for ERK2 recognition through the high avidity KIM of PTP-SL but dispensable for recognition by the low avidity KIM of MEK1.

ERK2 mutation Cytosolic retention by a PTP-SL MKP-3 MEK1
Wild type ϩ ϩ ϩ a Estimation was made by double immunofluorescence analysis, as in Fig. 6. ϩ, ERK2 excluded from the nucleus (Ͻ10% of cells coexpressing ERK2 and the effector show nuclear localization of ERK2); Ϫ, ERK2 not excluded from the nucleus (Ͼ90% of cells coexpressing ERK2 and the effector show nuclear localization of ERK2); Ϫ/ϩ, ERK2 partially excluded from the nucleus (30 -60% of cells coexpressing ERK2 and the effector show nuclear localization of ERK2). cally inactive MAP kinases (ERK2 K52A and p38␣ K53A) were used to avoid interference with their intrinsic catalytic activity. EGF-activated HA-(KIM-SL)MEK1 phosphorylated ERK2 K52A more efficiently than EGF-activated HA-MEK1 (Fig. 9, A(left panel) and B(upper diagram)). This effect was not due to a higher enzymatic activity of (KIM-SL)MEK1 as compared with MEK1, since the in vitro autophosphorylation that was displayed by both kinases was similar (data not shown). Furthermore, constitutively active HA-(KIM-SL)MEK1 SESD also phosphorylated ERK2 K52A more efficiently than HA-MEK1 SESD (Fig. 9, A (left panel) and B (lower diagram)), ruling out that the effect was due to a higher activation by Raf of (KIM-SL)MEK1 compared with MEK1. Importantly, the phosphorylation of the ERK2 K52A/D319N mutation by the (KIM-SL)MEK1 kinases was greatly diminished when compared with the phosphorylation of ERK2 K52A. By contrast, the phosphorylation of ERK2 K52A/D319N and ERK2 K52A by the MEK1 kinases was similar (Fig. 9B). Finally, p38␣ K53A was not phosphorylated by the wild type or the constitutively active MEK1 chimeric molecules (Fig. 9A). These results demonstrate that MEK1 phosphorylation of the ERK2 D319N mutation is not compromised and indicate that the KIM of PTP-SL provides MEK1 with an increased capability to phosphorylate ERK2 in an Asp 319 -dependent manner.
The above results suggest that the (KIM-SL)MEK1 chimeric kinase could have gain-of-function properties in intact cells with respect to the activation of ERK1/2. To test this possibility, GST-ERK2 was co-expressed with HA-MEK1 SESD or HA-(KIM-SL)MEK1 SESD in HEK293 cells, and the catalytic activity of purified GST-ERK2 was tested by an in vitro kinase assay, using MBP as the substrate (Fig. 9C). As shown, ERK2 that was co-expressed with (KIM-SL)MEK1 SESD displayed a much higher catalytic activity, as compared with ERK2 that was co-expressed with MEK1 SESD. These results demonstrate that the (KIM-SL)MEK1 chimera is indeed able to hyperactivate ERK2 in intact cells.

DISCUSSION
ERK2 is directly inactivated by several tyrosine-specific and dual specificity protein phosphatases, including PTP-SL and MKP-3, and its direct activation requires the specific activity of the kinases MEK1/2 (28 -32). Although all three enzymes recognize ERK2 through its common docking groove (17, 20, 21, 52), subtle differences in the molecular interactions involved exist. This is demonstrated by the finding that specific mutations that target the ERK2 common docking groove conferred a gain-of-function phenotype as a result of deficient association of ERK2 to inactivating protein phosphatases without apparently affecting its activation by MEK1 (20,53,54). In this study, we have analyzed the recognition of a panel of single ERK2 mutations by MEK1 and by PTP-SL and MKP-3, in parallel with their sensitivity to dephosphorylation by the pool of protein-tyrosine phosphatases present in HEK293 cell lysates. We have found a good correlation between resistance to dephosphorylation of ERK2 mutations and their diminished association with PTP-SL and/or MKP-3 phosphatases. Thus, single amino acid substitutions at several structural elements that configurate the ERK2 docking groove, including the loops L5, L11, L16, and the helix ␣e, abrogate binding to PTP-SL and MKP-3 and confer resistance to dephosphorylation by the sodium vanadate-sensitive phosphatases from HEK293 cells (probably MKP-2, MKP-3, MKP-7, DUSP2, DUSP8, or HePTP). These findings suggest that the array of tyrosine phosphatases that inactivate ERK2 share a common recognition mechanism, in which electrostatic and hydrophobic interactions involving the ERK2 docking groove play a major role (17,20,22,38,(55)(56)(57)(58). Our finding that the differential dephosphorylation of ERK2 mutations by the endogenous MAP kinase phosphatases is only revealed after cell lysis without sodium vanadate can be explained by the absence of the counteractive effect of the MAP kinase kinases under these conditions. Also, this result suggests the existence of a strict negative control of MAP kinase phosphatase activity in intact cells during the early moments of ERK2 activation. This is in agreement with the hypothesis that distinct waves of MAP kinase phosphatase activation would be responsible of the regulated time-and spacedependent inactivation of ERK2 after ERK2-activating cell stimulation (31,33). The quantitative measurement of the interaction of ERK2 with PTP-SL and MKP-3 revealed significant differences when particular ERK2 mutations were compared. Of relevance, the mutation E79A, at loop L5 of ERK2, abrogated almost completely the interaction with PTP-SL, whereas the interaction with MKP-3 was preserved at a significant extent. Upon binding to ERK2, a conformational change of MKP-3 activates MKP-3 phosphatase activity (50), whereas activation of PTP-SL upon ERK2 binding has not been reported. However, upon ERK5 binding the catalytic activity of PTP-SL is increased (59). In addition to the intrinsic differences between PTP-SL and MKP-3, it is plausible that a distinctive molecular docking of the PTP-SL⅐ERK2 and MKP-3⅐ERK2 complexes, as illustrated from our results, may have different functional consequences on the biological activities of PTP-SL and MKP-3. A good correlation was found between the strength of MAP kinase phosphatase binding to ERK2 mutations at the loop L16 (D316N, D319N, D319E, E320A, and E320Q) in yeast and ERK2 cytosolic retention in COS-7 cells (compare Fig. 4 and TABLE ONE). However, ERK2 mutations at other regions important for MAP kinase phosphatase binding (R133A, E79A, and D160N) were retained in the cytosol at a variable extent, suggesting the existence of additional molecular mechanisms that could regulate specifically ERK2 nuclear translocation. The cellular context-dependent assembly of multiprotein complexes containing ERK2 and its effectors, as it has been proposed for PP2A⅐HePTP⅐ERK2 complex (60), may account for such an additional level of regulation.
Mutations targeting multiple residues at loop L16 of ERK2 inhibit productive MEK1 binding and ERK2 cytosolic retention, indicating a role for this region in MEK1 recognition (16 -19). However, the single ERK2 amino acid substitutions that target loop L16 and were analyzed in our study (i.e. D316N, D319N, D319E, E320A, and E320Q) did not abrogate cytosolic retention of ERK2 by MEK1. Also, these ERK2 mutations were efficiently activated by MEK1, suggesting that functional productive binding with MEK1 was not compromised. Moreover, an ERK1 mutation (D339N) equivalent to the ERK2 D319N mutation was also retained in the cytosol by MEK1 (26). Together, our data indicate that tyrosine phosphatase inactivators of ERK2 strictly depend on molecular interactions through the ERK2 docking groove for productive binding. In contrast, recognition of ERK2 by MEK1 relies on a different binding mode involving additional ERK2 regions. Such regions could include the loop L1 at the ERK2 N terminus, the ERK2 activation loop (L12), structural elements within or close to the MAP kinase insert (L13, ␣G, and ␣2L14), and the loop L8 (19,39,(61)(62)(63). Also, the role of the MEK1 C-terminal proline-rich sequence in the regulation of MEK1 activity has been documented (64 -68), and the phosphorylation of this MEK1 region by ERK2 and p21-activated kinase has been involved in the stabilization of MEK1⅐ERK2 complexes (69,70).
In MEK1, PTP-SL and MKP-3, conserved kinase interaction motifs (KIM) have been defined, which are important for physical and functional recognition of ERK2 (18, 34 -36, 58, 71, 72). Our analysis using Data are presented as the percentage of phosphorylation of (KIM-SL)MEK1 with respect to that shown by MEK1 wild type (wt) (100%), and the values represent the means of five separate experiments Ϯ S.D. In D, HEK293 cells were transfected with pCDNA3-HA-ERK2 wild type or D319N mutation. After 48 h, the cells were treated with EGF for 5 min and lysed (lysis buffer type B), and HA-ERK2 was immunoprecipitated with the anti-HA 12CA5 mAb and subjected to in vitro kinase assays, in the presence of [␥-32 P]ATP and MBP (2 g) or GST-MEK1, GST-(KIM-SL)MEK1, or GST-PTP-SL-(147-288) bound to glutathione-Sepharose as the substrates. The samples were resolved on 15% (for MBP phosphorylation) or 10% (for GST fusion protein phosphorylation) SDS-PAGE, and radioactive bands were processed for quantification. Data are presented as the percentage of phosphorylation of the distinct substrates with respect to that shown by each substrate phosphorylated by ERK2 wild type (100%). The values represent the mean of at least three separate experiments Ϯ S.D. In all the experiments, similar levels of expression of HA-ERK2 wild type and D319N mutation were obtained (data not shown).
MEK1 chimeras ((KIM-SL)MEK1) in which the MEK1 KIM motif was replaced by that of PTP-SL revealed that ERK2 binding and activation by MEK1, as well as MEK1 phosphorylation by ERK2, are increased upon use of the KIM of PTP-SL. Similarly, the addition of the KIM of the ERK2-specific PTP-SL-related HePTP tyrosine phosphatase (73,74) to the catalytic domain of PTP1B conferred to this tyrosine phosphatase the capability to dephosphorylate ERK2 (75). Also, specificity sequences have been defined next to the KIM that account for the differential interaction of PTP-SL and the PTP-SL-related tyrosine phosphatase, STEP, with ERK2 and p38␣ (76). In this regard, the (KIM-SL)MEK1 chimera hyperactivated ERK2 but did not associate or activate p38␣. Thus, the specificity of MEK1 toward ERK2 requires additional molec-ular determinants other than the MEK1 N-terminal KIM. This is in agreement with our observations that the contribution of the MEK1 N-terminal KIM for binding of MEK1 to ERK2 is not very robust, as revealed by pull-down assays, and can be enhanced by substitution with the stronger KIM of PTP-SL. Whether the N-terminal KIM of MEK1 is the target of regulatory mechanisms of MEK1 binding to ERK2 (as described for the KIM of PTP-SL and HePTP) (43, 77) deserves further study. Together, these observations suggest the existence of recognition mechanisms that coordinate the binding and dissociation of ERK2 to its activating and inactivating effectors in a distinctive manner. This coordination would provide both fidelity and flexibility to regulate the magnitude and duration of the activation-inactivation cycles of ERK2, as . Then the cells were lysed (lysis buffer type B), and HA-tagged proteins were immunoprecipitated with the anti-HA 12CA5 mAb. Immune complexes were subjected to in vitro kinase assays in the presence of [␥-32 P]ATP and GST-ERK2-K52A or GST-p38␣-K53A bound to glutathione-Sepharose as the substrates (left panels). The samples were resolved on 10% SDS-PAGE, and subjected to Coomassie staining (lower left panel) and autoradiography (upper left panel). The equal expression of the HA-tagged MEK1 proteins is shown in the right panel, as detected by immunoblot with anti-HA 12CA5 mAb. B, the ERK2 D319N gain-of-function mutation hampers the phosphorylation by (KIM-SL)MEK1 but not the phosphorylation by MEK1. The two diagrams represent the quantification of immune complex kinase assays using GST-ERK2-K52A or GST-ERK2-K52A/D319N as the substrates, as in A. In the upper diagram, the phosphorylation of the ERK2 substrates by HA-MEK1 or HA-(KIM-SL)MEK1 is shown. In the lower diagram, the phosphorylation of the ERK2 substrates by constitutively active HA-MEK1 SESD or HA-(KIM-SL)MEK1 SESD is shown. Data are presented as the percentage of phosphorylation with respect to that shown by ERK2-K52A phosphorylated by MEK1 or MEK1 SESD (100%). Equal amounts of the different kinases were used as shown in Fig. 9A, right panel. The values represent the mean of at least three separate experiments Ϯ S.D. C, increased activation of ERK2 by (KIM-SL)MEK1. HEK293 cells were cotransfected with pRK5-GST-ERK2 plus pCDNA3-HA-MEK1 SESD or pCDNA3-HA-(KIM-SL)MEK1 SESD. After 48 h, the cells were lysed (lysis buffer type B), and equal amounts of cell lysates containing GST-ERK2 were pulled down with glutathione-Sepharose. The complexes were subjected to in vitro kinase assays, in the presence of [␥-32 P]ATP and MBP (2 g) as the substrate. The samples were resolved on 15% SDS-PAGE and subjected to autoradiography (upper panel). The expression of HA-MEK1 and GST-ERK2 is shown in the middle and lower panels, respectively, as monitored by immunoblot with anti-HA 12CA5 mAb or anti-GST polyclonal antibody. In the lower diagram, a quantification of the MBP phosphorylated bands is shown. Data are presented as the percentage of MBP phosphorylation with respect to that shown by MBP phosphorylated by ERK2 activated by MEK1 SESD (100%). The values represent the mean of at least three separate experiments Ϯ S.D. To rule out the possibility that the MBP substrate was phosphorylated also by MEK1 SESD or (KIM-SL)MEK1 SESD, coprecipitated in the kinase complexes with ERK2, kinase assays were performed under the same conditions using MEK1 SESD or (KIM-SL)MEK1 SESD as the kinases and MBP as the substrate. No phosphorylation of MBP by MEK1 SESD or (KIM-SL)MEK1 was detected (data not shown).
well as its nuclear/cytoplasmic localization, during the adaptative cell responses to environmental changes.