Phosphorylation of MAP Kinases by MAP/ERK Involves Multiple Regions of MAP Kinases*

Mitogen-activated protein (MAP) kinases are activated with great specificity by MAP/ERK kinases (MEKs). The basis for the specific activation is not understood. In this study chimeras composed of two MAP kinases, extracellular signal-regulated protein kinase 2 and p38, were assayed in vitro for phosphorylation and activation by different MEK isoforms to probe the requirements for productive interaction of MAP kinases with MEKs. Experimental results and modeling support the conclusion that the specificity of MEK/MAP kinase phosphorylation results from multiple contacts, including surfaces in both the N- and C-terminal domains.

300 residues encoding the two-domain structure (13). Protein kinases possess 12 conserved stretches of amino acids within their catalytic domains known as subdomains (13)(14)(15). These conserved elements as well as unique structures contribute to catalysis by and regulation of protein kinases. The functions of several of the conserved and unique structural motifs are known. A glycine-rich loop in the N-terminal domain, termed the phosphate anchor ribbon, has a role in binding ATP. Also in the N-terminal domain, subdomain III encodes the C helix, which contains an invariant Glu involved in binding MgATP (16,17). This helix is important for maintaining an open domain conformation in unphosphorylated ERK2 (18), aligning catalytic residues in Src (19) and Cdk2 (20), and binding to the cyclin regulatory subunit in Cdk2 (21). The activation loop, known as the phosphorylation lip in MAP kinases, is a poorly conserved element in the C-terminal domain and contains the two MAP kinase phosphorylation site residues within a TXY sequence. Also in the C-terminal domain is the MAP kinase insert. It is composed of 32 residues and is only found in the MAP kinases and the Cdks. The MAP kinase insert and a loop consisting of residues 199 -205 of ERK2 interact with the phosphorylation lip in the unphosphorylated form of ERK2. Finally, a loop at the C terminus of the MAP kinases, L16 or the C-terminal tail, wraps around the back of the structure and interacts with the N-terminal domain (18).
The crystal structures of unphosphorylated and phosphorylated ERK2 and unphosphorylated p38 have been solved (18,(22)(23)(24). Several important differences exist between the structures of the unphosphorylated forms of ERK2 and p38. There is a wider domain separation in p38 than in ERK2. The phosphorylation lip in p38 is six residues shorter and has a different conformation. In p38 the lip is folded up between the two domains, and the C-terminal portion of the lip forms a turn of helix that blocks the Pϩ1 specificity pocket. This pocket directs the proline specificity of MAP kinases. The axis of the C helix is rotated, and the N terminus of this helix is shifted by 6 Å relative to the helix in ERK2. The helix at the end of L16 is extended by 7 Å, and there is an increase in the hydrophilicity of residues that form contacts between L16 and the N-terminal domain. This results in a less intimate interaction between L16 and the N-terminal domain in p38 (22,24).
Within MAP kinase cascades, the MEKs are the most specific enzymes. These dual specificity kinases activate their respective MAP kinase substrates by phosphorylating the threonine and tyrosine of the specific TXY sequence located in the phosphorylation lip. The only known substrates of MEK1 and MEK2 are ERK1 and ERK2 (25). Other MEKs also phosphorylate only a small subset of the MAP kinase family. For example, MEK3 and MEK6 will only phosphorylate p38 isoforms, whereas MEK5 will only phosphorylate ERK5 (26 -28). In addition, the MEKs require native MAP kinases as substrates; they will not phosphorylate denatured proteins or pep-tides derived from the phosphorylation lip (25). Little is known about the structural basis of the specificity of interactions between MEKs and MAP kinases despite information from the crystal structures of ERK2 and p38.
Previous in vitro studies concluded that neither the phosphorylation lip length nor the residue between the phosphorylation sites is a critical element for directing MEK specificity (29,30). Brunet and Pouysségur (31) measured the in vivo activities of chimeras of p38 and ERK1 after treatment of transfected cells with stimuli that activated either p38 or ERK1. Based on these findings they proposed that a 40-residue stretch of amino acids within subdomains III and IV of p38 was important for directing the specificity of activation. In addition, they suggested that the C helix within subdomain III was the key element for recognition of MAP kinases by MEKs.
To define structural elements of the MAP kinases that are required to direct specific recognition by MEKs, chimeras of ERK2 and p38 were tested for phosphorylation and activation by five different MEKs in vitro. Compilation of the data reveals multiple spatially segregated contacts in the MEK/MAP kinase interface.

EXPERIMENTAL PROCEDURES
Generation of Mutants and Purification of Proteins-Restriction sites were inserted at the ends of subdomains II and IV and at residue 326 of ERK2 and residue 322 of p38 with the QuikChange kit (Stratagene, La Jolla, CA). ERK2II3 CT contains all the restriction sites as well as sites at the ends of subdomains III and V, and p38II contains the site at subdomain II. The resulting plasmids were used to create ERK2-p38 chimeras. His 6 -tagged proteins were expressed and purified as described previously (32) or by batch binding to Ni 2ϩ -nitrilotriacetic acidagarose (Qiagen). Protein concentrations were determined with the BCA assay (Pierce) and by comparison of Coomassie Blue-stained bands after SDS-polyacrylamide gel electrophoresis. Phosphorylated His 6 -MEK1, GST-MEK2, and His 6 -MEK3 were purified from a previously described coexpression system (33), and GST-MEK4 and GST-MEK6 were purified and activated by GST-MEKK-C (34) essentially as described (29).
Phosphorylation and Activity Assays-Phosphorylation and activity assays were performed essentially as described (29). Activity of ERK2, p38, or chimeras was determined using myelin basic protein (Sigma) as substrate. One-dimensional phosphoamino acid analysis was performed as described (35).
Time Courses of Phosphorylation and Activity-Assays were performed as described (29), but reactions were stopped at the times indicated in Figs. 5 and 6 for MEK2 and MEK6, respectively. For MEKs 1, 3, and 4, 50-l aliquots were removed from one large reaction mix and stopped by addition to 15 l of 5ϫ SDS sample buffer.
Western Blotting-20 ng of ERK2 or each chimera from the time course reactions with MEK2 and MEK6 were resolved by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose. Blots were probed with an antibody specific for phosphorylated ERK2, V803 (Promega), and an antibody that recognizes phosphorylated and unphosphorylated ERK2, Y691.
Phosphatase Treatment-ERK2 or chimeras were incubated with an equimolar amount of PTP1 at 30°C for 30 min. Sodium orthovanadate was added to one set of reactions at time 0 and to an identical set at 30 min. Kinase assays with MEK2 were then performed as above, but they were stopped after 1 min.

RESULTS
MEKs recognize the native MAP kinase structure with considerable selectivity. The primary sequence surrounding the TXY phosphorylation sites of the ERKs is not sufficient to determine MEK recognition. Neither peptides derived from the double phosphorylation site nor denatured proteins are phosphorylated by MEKs (25). These findings support the conclusion that MEKs require specific conformations or secondary determinants for productive interaction with their MAP kinase substrates. We wished to identify the structural features of MAP kinases that allow MEKs to distinguish among them as substrates. To do this we created chimeras between ERK2 and p38 that contained selected intact structural elements from each enzyme (Fig. 1A). The chimeras were composed of ERK2 and p38 because a related set of chimeras had been tested in transfected cells (31), and the structures of both of these MAP kinases were available for analysis (18,22,24). The chimeras were expressed and purified and then tested in vitro for activation by MEK family members, MEKs 1, 2, 3, 4, and 6, which phosphorylate ERK2 or p38 in vitro. Measurements included rate and extent of phosphorylation, phosphoamino acids, recognition by antibodies selective for phosphorylated ERK2, and activated activity relative to the wild type proteins. ERK2 and p38 proteins containing inserted restriction sites were also purified and tested as controls.
The MAP kinase catalytic core contains two domains. The N-terminal domain is composed of N-terminal sequence (white and red in Fig. 1B) and the C-terminal tail, L16 (yellow in Fig.  1B), and is primarily involved in binding ATP. In the MAP kinases L16 lies on the surface of the N-terminal domain. The C-terminal domain (blue in Fig. 1B) binds the protein substrate. To identify important structural features, we first created a chimera, PIVECTP, which contained one of its two major Subdomains from ERK2 are white, and subdomains from p38 are gray. The chimeras were named according to the subdomains that were exchanged. B, structural representation of MAP kinase domains using ERK2 as the model. The C-terminal domain is shown in blue. The N-terminal domain includes residues in white (subdomains I and II), red (subdomains III and IV, which include the C helix and ␤ strands 4 and 5), and yellow (L16). Chimeras swapped the white, red, and yellow structures as indicated in A and the text. structural domains from each kinase (Fig. 1). This chimera contained the larger C-terminal structural domain from ERK2 and the smaller N-terminal domain from p38. The p38 elements making up the N-terminal domain are the N terminus through subdomain IV and L16 (shown in white, red, and yellow in Fig. 1B).
We examined phosphorylation of this chimera by MEKs, the amino acids phosphorylated, and its activity after phosphorylation. This chimera was phosphorylated by MEK1 and MEK2 and by MEK3, MEK4, and MEK6, although the stoichiometry of phosphorylation by MEK3/4/6 was lower (Fig. 2). The rates of phosphorylation of the chimera by MEK2 and MEK1 were similar to the rates of phosphorylation of wild type ERK2 by these MEKs (see Figs. 5A and 7A). Likewise, the rates of phosphorylation of this chimera by MEK6 and MEK3 were similar to their rates of phosphorylation of p38 (see Figs. 6A and 7B). In each case, both tyrosine and threonine were phosphorylated (see Fig. 5C and 6C). These results indicated that interaction determinants were present in both domains. All five MEKs activated PIVECTP, and in each case the activity was proportional to the extent of phosphorylation, with MEK4 being the least effective ( Figs. 2 and 3). The activity of this unphosphorylated chimera was similar to that of wild type ERK2. Maximal activity was about 15% of the activity of wild type ERK2 (Fig. 3). This may be due to imperfect folding of the chimera.
Inclusion of residues from p38 in the N-terminal domain of the chimera allowed phosphorylation by MEK3/6 in addition to MEK1/2. We next wanted to determine which structural motifs within the N-terminal domain were responsible for this expanded interaction. Therefore, three additional chimeras containing less of the N-terminal domain of p38 were constructed and analyzed as described next.
To determine the contribution of L16 (yellow in Fig. 1B) to phosphorylation by MEK, the PIVE chimera, which had all of the N-terminal domain except L16 from p38, was generated. It contained p38 residues from the N terminus through subdomain IV and ERK2 residues from subdomain V through the C terminus (Fig. 1A). Its activity in the unphosphorylated state was similar to wild type ERK2. This chimera was phosphorylated by MEK1, MEK2, and MEK6 but not by MEK3 or MEK4 (Fig. 2). All three MEKs that phosphorylated PIVE activated it equally (Fig. 3). It had stimulated protein kinase activity about 20% of that of ERK2. The rates of phosphorylation of PIVE by MEK1, MEK2, and MEK6 were similar to the rates of phosphorylation of the wild type proteins (see Figs. 5A, 6A, and 7A). These results suggested that the C-terminal tail within the N-terminal domain contributed to activation by MEK3.
EIIPIVE had only p38 subdomains III and IV, which contained the C helix and ␤ strands 4 and 5 (red in Fig. 1B). Without phosphorylation by MEK, this chimera had slightly higher activity than wild type ERK2. EIIPIVE was phosphorylated and activated by MEK1 and MEK2 (Figs. 2 and 3). When phosphorylated, this chimera demonstrated decreased activity relative to wild type ERK2 (Fig. 3). MEK3, MEK4, and MEK6 did not phosphorylate it. Therefore, the C helix, ␤4, and ␤5 of MAP kinases were not sufficient for phosphorylation of a MAP kinase by these MEKs. EIIPIVE also was phosphorylated by MEK1 and MEK2 at rates similar to the rates of phosphorylation of ERK2 (see Figs. 5A and 7A).
To determine the role of the first two subdomains (white in Fig. 1B) in phosphorylation by MEKs, PIIE was tested. This chimera contained p38 residues in the N terminus through subdomain II and ERK2 residues in subdomain III through the C terminus (Fig. 1A). This chimera had activity comparable with that of unphosphorylated ERK2. Once again, only MEK1 and MEK2 phosphorylated and activated this chimera (Figs. 2 and 3). The rates of phosphorylation of PIIE by MEK1 and MEK2 were similar to the rates of phosphorylation of wild type ERK2 by MEK1 and MEK2 (see Fig. 5A and 7A). Results from these four chimeras indicated that many p38 residues must be present in the N-terminal domain of each chimera for phosphorylation by the p38-specific MEKs.
We wanted to determine whether any differences existed in the pattern of phosphorylation and activation by MEKs if the other MAP kinase was at the N terminus of the chimera. Thus, we expressed a reciprocal chimera, EIIP. EIIP contained ERK2 residues from the N terminus through subdomain II (white in Fig. 1B) and p38 residues from subdomain III through the C terminus. This chimera was phosphorylated by MEK4 and MEK6 but not by MEK1, MEK2, or MEK3 (Fig. 2). The time courses of phosphorylation were slightly more rapid (see Fig.  7C). This chimera had very low basal activity, less than 1% of wild type ERK2, suggesting that it was poorly folded. After phosphorylation, this chimera was inactive (Fig. 3). Analysis of its phosphoamino acids after phosphorylation by MEK4 revealed that it was inactive in part because there was little or no threonine phosphorylated (Fig. 4). Domains I and II of ERK2 alter the position of the C helix in the chimera, which may account for the inability of MEKs to phosphorylate threonine.
The two control proteins were also tested in the in vitro kinase assays to determine whether any residues changed by the insertion of the restriction sites caused the changes in interaction with MEKs. The ERK2 control, ERK2II3 CT was phosphorylated and activated only by MEK1 and MEK2 (data not shown). Therefore, residues that were changed by the insertion of restriction sites did not affect the productive interaction of ERK2 chimeras with MEKs. The p38 control, p38II, was phosphorylated by MEK3, MEK4, and MEK6 (data not shown). p38II was activated normally by MEK4/6 but less well by MEK3 (data not shown). The phosphoamino acid analysis (Fig. 4) indicates that p38II is not fully activated by MEK3 because it is poorly phosphorylated on threonine.
The maximum stoichiometries of phosphorylation of the chimeras by MEKs 1-4 and 6 were evaluated (Table I)  To determine whether phosphorylation occurred on the predicted residues (Thr 183 and Tyr 185 ), chimeras phosphorylated by MEKs for different times were immunoblotted with antibodies that selectively recognize the doubly phosphorylated ERK2 epitope. These antibodies do not detect unphosphorylated or singly phosphorylated ERK2 at equivalent protein concentrations. The immunoblots reveal an excellent parallel between increasing phosphorylation and appearance of the phospho epitope (Figs. 5 and 6). Phosphorylation of the chimeras saturated at the same times as did blotting intensity. This is consistent with the conclusion that phosphorylation occurred largely on Thr 183 and Tyr 185 and not on previously unidentified residues.
These immunoblotting findings agree with phosphoamino acid analysis, which showed the majority of phosphate on tyrosine and threonine. However, PIVECTP, for example, apparently contained less phosphotyrosine in these analyses than did wild type ERK2. To determine whether this was due to prior autophosphorylation of the chimera, we compared the stoichiometry of phosphorylation of wild type ERK2 and three of the chimeras before and after dephosphorylation with the tyrosine phosphatase PTP1 (Fig. 8A). The stoichiometry of phosphorylation of ERK2 was unchanged by pretreatment with phosphatase; on the other hand, phosphorylation of two of the chimeras increased by 20 -30%. The increase was primarily due to increased tyrosine phosphorylation as was apparent from the increase in phosphotyrosine recovered in the phosphoamino acid analysis (Fig. 8, B and C). This is consistent with the idea that the chimeras have enhanced abilities to autophosphorylate on Tyr 185 prior to isolation from bacteria. Because this tyrosine is already partly phosphorylated when , PIVE (f), PIVECTP (ࡗ), and EIIPIVE () after incubation with MEK2 for the times listed in A. C, immunoblots of ERK2 and chimeras after incubation with MEK2 for the times listed in A. Top, blots were probed with an antibody that specifically recognizes ERK2 or chimeras phosphorylated on both threonine and tyrosine. Bottom, blots were stripped and reprobed with an ERK antibody, Y691, which recognizes both unphosphorylated and phosphorylated ERK2 or chimeras. The exposure time for the PIIE pTEpY blot is longer to allow the bands to be seen. MEK isoforms recognize members of the MAP kinase family with selectivity. The residues or structural motifs that are required for this selective recognition have not been defined. ERK1/2 and p38 isoforms share about 40 -45% sequence identity (5,36,37), and the structures of the unphosphorylated forms of ERK2 and p38 are similar (18,22). The goal of this project was to use chimeras containing swapped structural elements of ERK2 and p38 to identify regions of the MAP kinases that direct interactions with particular MEKs.
Our findings using ERK2/p38 chimeras indicate that no single MAP kinase structural motif forms a sufficient interaction surface with MEK. The N-terminal structural domain likely forms part of the interface. However, even the entire domain is not sufficient to restrict phosphorylation by MEKs, because the chimera containing the N-terminal domain from p38 was phosphorylated by MEK1 and MEK2 in addition to MEK3/4/6. This demonstrates that there are interaction determinants in the C-terminal domain. In addition, a fairly large surface area within the N-terminal domain must be present for the interaction of the chimeras with MEK3/4/6. Swapping only subdomains I and II or subdomains III and IV was not enough to allow MEK3/4/6 to phosphorylate the chimeras. Based on these data, it appears that MEKs require multiple interacting sites within both domains to specifically phosphorylate different MAP kinases.
Our results with ERK2/p38 chimeras are in part consistent with work from Brunet and Pouysségur (31), who expressed p38/ERK1 chimeras in mammalian cells. A p38/ERK1 chimera equivalent to PIVE was activated by both growth factors and cell stress, whereas a chimera equivalent to PIIE was only activated by growth factors. We have shown with further subdivision of the swapped regions that in contrast to the suggestion from the p38/ERK1 studies, the C helix alone is not the key element determining which MEKs will phosphorylate a particular MAP kinase in vitro.
Additional findings suggest that the MAP kinase insert is required for phosphorylation by MEKs. To address the function of this insert, it was deleted from wild type ERK2. 2 Examination of the crystal structure indicated that deletion of these residues would have a minimal effect on the tertiary structure of ERK2. In the unphosphorylated state, the resulting protein had protein kinase activity equivalent to unphosphorylated wild type ERK2. However, the protein lacking the MAP kinase insert was no longer phosphorylated by MEK2 in vitro. 2 It is possible that the insert aids in the correct folding of the phosphorylation lip, for binding to MEKs, or that it may be directly involved in interaction with MEKs.
We mapped the currently available data for MEK specificity determinants onto the structure of unphosphorylated ERK2 (Fig. 9). The data reported here indicate that the C helix, ␤ strands 4 and 5, and L16 of MAP kinase are important, because when these motifs are from ERK2, phosphorylation by MEK3/ 4/6 is lost. The findings noted above suggest a role for the MAP kinase insert. In addition, previous work showed that mutating tyrosine 185 in the phosphorylation lip or deleting six residues from this loop of ERK2 resulted in significantly decreased phosphorylation by MEK1 and MEK2 (29,32). Taken together, the regions of a MAP kinase that appear to be important for the MEK/MAP kinase interaction include the vicinity of the C helix and L16 from the N-terminal domain and include the phosphorylation lip region and the MAP kinase insert from the Cterminal domain. These experimental findings are in good agreement with a visually generated model of the ERK2/MEK1 interface. The model was created by mapping MEK1 residues onto the coordinates of the cAMP-dependent protein kinase (14) and manually docking with the ERK2 coordinates. The ERK2 phosphorylation lip was placed on MEK1 in the position of the bound peptide substrate (16). The sites of contact in this hypothetical ERK2/MEK1 model include residues from each of the structural elements implicated experimentally, namely, the tip of the C helix, the loop region of L16, the phosphorylation lip, and the tip of the MAP kinase insert.
The proposed MEK/MAP kinase interface is involved in other protein-protein interactions. Residues in this putative interface overlap with the ERK2 dimerization interface (38). In particular, L16 and the phosphorylation lip are intimately involved in both. Furthermore, it is intriguing that the proposed MEK/ MAP kinase interface is similar to the interface between Cdk2 and cyclin A (21). Structural elements of Cdk2 that participate in cyclin A binding include the PSTAIRE helix region, which encompasses the helix equivalent to the C helix of MAP kinases and the ␤4 and ␤5 strands. In the C-terminal domain, contacts with cyclin A are made by residues in the region of the T loop, equivalent to the ERK2 phosphorylation lip, and a C-terminal helix, helix 7. The only motif that participates in the proposed MEK/ERK interface but not in the Cdk2/cyclin A interface is the MAP kinase insert.