Characterization of the Atypical MAPK ERK4 and Its Activation of the MAPK-activated Protein Kinase MK5*

The extracellular-regulated kinase (ERK) 4 (MAPK4) and ERK3 (MAPK6) are structurally related atypical MAPKs displaying major differences only in the C-terminal extension. ERK3 is known as an unstable mostly cytoplasmic protein that binds, translocates, and activates the MAPK-activated protein kinase (MK) 5. Here we have investigated the stability and expression of ERK4 and have analyzed its ability to bind, translocate, and activate MK5. We show that, in contrast to ERK3, ERK4 is a stable protein that binds to endogenous MK5. Interaction of ERK4 with MK5 leads to translocation of MK5 to the cytoplasm and to its activation by phosphorylation. In transfected HEK293 cells, where overexpressed catalytically dead ERK3 is able to activate MK5, catalytic activity of ERK4 is necessary for activation of MK5, indicating that ERK4 directly phosphorylates MK5. Interestingly, ERK4 dimerizes and/or oligomerizes with ERK3, suggesting that overexpressed inactive ERK3 recruits active endogenous ERK4 to MK5 for its activation. Hence, ERK3 and ERK4 cooperate in activation of MK5.

Mitogen-activated protein kinases (MAPKs) 2 represent a family of evolutionary conserved enzymes with a central role in the well characterized MAPK signaling cascades. A wide variety of extracellular stimuli serve as activators of MAPK pathways leading to appropriate responses of cells, such as proliferation, differentiation, growth, and migration. MAPK pathways generally have a three-kinase module architecture by which the signal is transmitted from an upstream kinase to a downstream kinase by sequential phosphorylation. MAPKs comprise four well defined groups (ERK1/2 (1, 2), c-Jun N-terminal kinases, p38s, and ERK5 (BMK) (3)), but additional members including ERK3 (1,4), ERK4 (p63 MAPK, ERK3-related, ERK3␤, MAPK4, Prkm4) (5), and ERK8 (6) have been identified. ERK4 (p63 MAPK) was described in humans (5), soon after ERK1, ERK2, and ERK3 were identified (1). Among MAPKs, ERK4 is most closely related to ERK3 displaying 62% overall amino acid sequence identity and 73% within the predicted kinase domain. Both kinases do not contain the highly conserved activation loop ("a-loop") motif TXY between kinase subdomains VII and VIII that is found in all other MAPKs but possess a SEG sequence instead (Fig. 1A). Even the APE motif of subdomain VIII, which is extremely conserved in other MAPKs, is replaced by an SPR motif in ERK3 and ERK4 (Fig. 1A). ERK4 and ERK3 carry long C-terminal extensions (Fig. 1B). Human Erk4 was mapped on chromosome 18q12-21 (7), and a cDNA for the rat homolog rMNK2 was isolated (8). Stimuli, activators, or relevant substrates of ERK4 have remained elusive, and enzymatic activities of the atypical ERKs have not been well defined so far.
Initially the MAPK-activated protein kinase MK5 (9, 10), also known as p38-regulated and -activated kinase (PRAK), was described as a member of the MK family and a downstream target of p38 (for recent reviews see Refs. 11 and 12). Previous data suggested that MK5 is not a physiological substrate for p38 in vivo (13), because the stimuli that activate the p38 pathway fail to activate MK5, and binding of endogenous p38 to MK5 is weaker than interaction of p38 with other established substrates, such as MK2 or MK3. Interestingly, it has been recently demonstrated that MK5 strongly interacts with and is activated by ERK3 (14,15).
In this study, we characterize expression, stability, and protein interaction of the ERK3-related kinase ERK4 and analyze its influence on subcellular localization and activity of MK5.
In Vitro Pull-down-Hexahistidine-tagged ERK4 was expressed in bacteria, or the protein isolated from different tissues of mouse. One mg of bacterial or tissue lysate protein was incubated with either 0.1 nmol of recombinant GST, GST-MK2, or GST-MK5 bound to glutathione-Sepharose 4B (Amersham Biosciences). After five washes with IP buffer (1ϫ PBS, 50 mM NaF, 1% Triton X-100, 1 mM Na 3 VO 4 ), the beads were resuspended in 4ϫ Laemmli buffer, and proteins from the beads were separated by SDS-PAGE, used for Western blotting, and developed with anti-His antibody (Penta-His; Qiagen) or anti-ERK3 (D23; Santa Cruz) antibody, which cross-reacts with ERK4.
BioEase Pull-down-For BE pull-down HEK293 cells were co-transfected either with pcDNA6/BioEase-ERK4 or -ERK3 and pDEST27-ERK3 or -ERK4 vectors, after 16 h the cells were lysed, and supernatant was applied to pull-down with 25 l of Streptavidin-agarose beads (Invitrogen). Western blot was developed with anti-GST antibody.
GST-ERK4 Pull-down of Endogenous MK5-5 ϫ 10 6 WT and MK5 Ϫ/Ϫ mouse embryonic fibroblast (MEF) cells were transfected with either GST or GST-ERK4 expressing vector. After 16 h cells were lysed, and GST pull-down was performed by using 25 l of glutathione-Sepharose beads. Western blot was developed with anti-MK5 antibody (a kind gift from Dr. Sir Philip Cohen).
Immunodetection of Endogenous MK5/ERK4 Complexes-5 ϫ 10 6 WT, MK5 Ϫ/Ϫ , or ERK3 Ϫ/Ϫ MEF cells were grown in culture and lysed with kinase lysis buffer for 30 min on ice. IP of endogenous MK5 was performed by incubation of the 1.5 mg of protein lysate (150 l) with 4 l of anti-MK5 antibody overnight followed by incubation with 15 l of protein G-Sepharose (Amersham Biosciences) for 1 h at 4°C. After five washings of the Sepharose beads with IP buffer, the beads were resuspended and boiled in SDS loading buffer. Western blot was developed using anti-ERK4 antibody (a kind gift from Dr. Ole Morten Seternes). Western Blot-The protein concentration was measured using the Bradford assay (Bio-Rad). To equalize amounts of protein 4ϫ Laemmli's SDS sample buffer (40% glycerol, 4% SDS, 4% ␤-mercaptoethanol, 0.4 M Tris-HCl, pH 6.7, and 2 mg/ml bromphenol blue) was added. The samples were boiled and centrifuged. Soluble protein extract was subjected to 10% polyacrylamide-SDS gels and transferred to Hybond ECL membranes (Amersham Biosciences). The blots were incubated for 1 h in PBS with 0.1% Tween 20 containing 5% powdered skim milk. After three washes with PBS with 0.1% Tween 20, the membranes were incubated for 16 h with the primary antibody at 4°C or 1 h at room temperature and for 1 h with horseradish peroxidase-conjugated secondary antibodies at room temperature. Antigen-antibody complexes were detected with an ECL detection kit (Santa Cruz Biotechnology), and the digital chemiluminescence images were taken by a Luminescent Image Analyzer LAS-3000 (Fuji Film).
In Vitro Pull-down and IP Kinase Assays-Kinase assays were performed after pull-down as described above and elsewhere (13) using 15 l of 50% glutathione or protein G-Sepharose suspension (Amersham Biosciences), 2.5 l of buffer (500 mM sodium ␤-glycerophosphate, 1 mM EDTA, pH 7.4), 10 g of substrate recombinant Hsp25 in a final volume of 20 l. Then 5 l of hot ATP mixture (20 mM MgCl 2 , 0.5 mM ATP, 0.1 l of [␥-33 P]ATP) was added, and the reaction mix was incubated for 10 min at 30°C. Radioactivity incorporated into Hsp25 was quantified by phosphorimaging using a Fuji Bas-1500 and TINA 2.09 software.
Expression of Fusion Proteins in HEK293 and Detection of Subcellular Localization of GFP-tagged Proteins-5 ϫ 10 6 HEK293 cells were transiently transfected by Lipofectamine in accordance to the manufacturer's protocol (Invitrogen). An equimolar amount was used for each vector. For analysis of subcellular localization, GFP and cyan fluorescent protein expression vectors were transfected, the cells were replated in Chambered cover glass (Labtek, Nunc) and analyzed using a Leica DM IRBE microscope with the Leica TCS confocal systems program or Visitron Systems and SPOT Advanced Programme. Staining of nuclei was performed by adding TO-PRO3 (Molecular Probes) to mounting medium (1:1000).
Determination of Protein Stability-HEK293 cells were transfected with GFP-ERK3 and GFP-ERK4 or BE-ERK3 and BE-ERK4 plasmids. After 16 h the cells were treated with cycloheximide (100 g/ml; Calbiochem) for indicated the time points with or without MG132 (20 M; Biomol). The cells were lysed in kinase lysis buffer and centrifuged, and equal amounts of lysate were loaded onto a SDS gel. Protein amounts were analyzed by Western blotting against GFP or BE.

RESULTS
ERK4 Is a Stable Protein-Recent data showed that human ERK3 is a highly unstable protein and that its degradation depends on N-terminal ubiquitination and subsequent proteosomal degradation (17). To analyze whether ERK4 shows a similar tendency to rapid proteosomal degradation, mouse cDNAs of GFP-ERK3 and GFP-ERK4 were expressed in HEK 293 cells, and proteins were compared for stability. After 24 h, protein biosynthesis was inhibited by blocking translation using cycloheximide treatment for 2 and 4 h. In parallel, the proteasome inhibitor MG132 was applied together with cycloheximide to identify proteasome-dependent degradation of the protein kinases. As previously described for human ERK3 (17), in our experimental system N-terminal GFP-tagged mouse ERK3 is an unstable protein with a half-life of about 1 h that undergoes proteasome-dependent degradation (Fig. 2). In contrast, there was no significant difference in the protein level of GFP-ERK4 after 4 h of cycloheximide treatment, and no effect of MG132 was detected. This indicates that ERK3-and ERK4-protein stability are differentially regulated. To ensure that protein stability is not altered by the large N-terminal fusion, as described for C-terminal fusions of ERK3 (21), we also investigated stability of ERK4 and ERK3 using the small N-terminal BioEase tag. The stability of both proteins was independent of the respective tag (Fig. 2).  NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46 ERK4 Interacts with MK5 in Vitro as Well as in Transfected HEK293 Cells-ERK3 and ERK4 are atypical MAP kinases that possess a long C terminus and lack the conserved TXY motif. Because Western blot and in situ hybridization analysis detected co-expression of ERK4 and ERK3 with MK5 in mouse (see supplemental Fig. S1), we were interested to study whether MK5, a known partner for ERK3 (14,15), also interacts with ERK4. Recombinant His-ERK4 was expressed in bacteria, and in vitro pull-down was performed with purified GST-MK5 and GST-MK2 protein. His-ERK4 was found to interact with GST-MK5 as well as with GST-MK2 (Fig. 3A). To further study this interaction in a cellular model, GFP-ERK4 was co-transfected with GST-MK5 and GST-MK2, respectively, in HEK293 cells. In GST pull-down, GFP-ERK4 was precipitated with GST-MK5 (Fig. 3B). GFP-ERK3, which binds to MK5, was used as a positive control (Fig. 3B). Only weak interaction of GFP-ERK4 was observed with GST-MK2 (Fig. 3B). Although MK5 shows significant homology to MK2, and both kinases can bind to ERK4 in vitro, the data from co-transfected HEK 293 cells imply that a more specific interaction of ERK4 exists with MK5 in vivo. This notion is supported by the co-localization studies below.

ERK4 Interacts with and Activates MK5
Next, we studied interaction of the endogenous proteins. By using a GST-ERK4 pull-down assay, we demonstrated that endogenous MK5 interacts with the overexpressed bait in MEF cells (Fig. 3C). As negative controls, we failed to detect interaction with GST alone in WT cells and with GST-ERK4 in MK5-deficient cells. We then analyzed interaction of endogenous proteins in mouse embryonic fibroblasts of different genotypes (WT, MK5 Ϫ/Ϫ , and ERK3 Ϫ/Ϫ ) using immunoprecipitation with MK5 antibodies and subsequent Western blot detection of ERK4 in the precipitate (Fig. 3D). ERK4 was co-immunoprecipitated with MK5 in WT and ERK3deficient cells, indicating an ERK3-independent interaction between both proteins. The specificity of the co-immunoprecipitation was confirmed by the failure to precipitate ERK4 from MK5-deficient cells.
To check whether the C-terminal region of ERK4, which does not show significant sequence homology to ERK3, or the region homologous to the MK5 binding site in ERK3 between amino acids 330 and 340 of ERK3 is important for specific interaction with MK5, we co-expressed GFP-tagged C-terminal deletion mutants (GFP-ERK4⌬C1-3, cf. Fig. 1B) together with GST-MK5 in HEK293 cells and performed GST pull-down. The result demonstrates that GFP-ERK4⌬C1 and 2 interact with GST-MK5, whereas GFP-ERK4⌬C3, which lacks the region between amino acids 326 and 340, fails to bind (Fig. 3E). Hence, the region in ERK4 homologous to amino acids 330 -340 in ERK3, which is essential for ERK3 to bind MK5 (15), is also important for MK5 binding of ERK4. The interaction between GFP-ERK4⌬C1 and 2 and GST-MK5 was almost quantitative, and GFP-ERK4⌬C1 and 2 could even be detected as Ponceau-stained protein bands on the filter (not shown).
ERK4 Translocates MK5 from the Nucleus to the Cytoplasm-We next analyzed the consequence of ERK4-MK5 interaction for the subcellular localization of both proteins using confocal fluorescence microscopy after transfection of ERK4 and/or MK5 expression plasmids into HEK293 cells. Nuclear counterstain was performed with the dye TO-PRO3. Overexpressed MK5 alone showed predominantly nuclear localization as expected by the presence of a nuclear localiza-tion signal (16). In contrast, GFP-ERK4 was mainly present in the cytoplasmic compartment of HEK293 cells (Fig. 4, left panel), similar to ERK3 (14), which carries a functional nuclear export signal in a region that is homologous to ERK4 (18). Cytoplasmic location of ERK4 results from CRM1-dependent nuclear export because treatment with leptomycin B (10 ng/ml; Sigma) for 8 h caused equal distribution of GFP-ERK4 between the nuclear and cytoplasmic compartments of HEK293 cells (not shown).
To test the physiological relevance of the interaction between ERK4 and MK5, GFP-MK5, or GFP-MK2 as a control, was co-expressed with GST-ERK4 or His-ERK4 in HEK293 cells. MK5, but not MK2, was almost completely translocated to the cytoplasmic compartment in the presence of both GST-or His-tagged ERK4 as well as in the presence of the ERK3 as positive control (Fig. 4, right panel). In contrast, MK2 was mainly present in the nucleus of HEK293 cells and was not affected in its localization by co-expression of neither GST-or His-ERK4 nor ERK3. Hence, ERK4 specifically translocates MK5 into the cytoplasm. This also supports the notion that ERK4 binding is independent of the ERK4 fusion tag and specific for MK5 in vivo. ERK4, but Not the ATP-binding Site Mutant ERK4-K49A, K50A, Activates MK5 in HEK293 Cells-ERK3 activates MK5 in HEK293 cells by binding and translocation independent of ERK3 catalytic activity (14). We were interested in whether ERK4 regulates MK5 kinase activity in a similar manner. Protein kinases were expressed in HEK293 cells, and an in vitro kinase assay was performed after pull-down of GST-MK5 using the known in vitro substrate Hsp25. We detected Hsp25 phosphorylating kinase activity and phosphorylated MK5 but also phosphorylated ERK4 and in controls phosphorylated ERK3. This indicates the existence of a productive complex between MK5 and ERK4 and ERK3, respectively. Both co-expressed GFP-ERK4 as well as GFP-ERK4⌬C1 activated MK5 (Fig. 5A) and were subsequently also phosphorylated by MK5 (see below). The activation of MK5 by ERK4 was comparable with the activation by ERK3 and its C-terminal deletion mutants ERK3⌬C1 and ERK3⌬C2. Only ERK3⌬C3, which is no longer able to bind and to translocate MK5 into the cytoplasm (14), failed to activate MK5 in this assay (Fig. 5A).
To see whether endogenous ERK4 is able to active MK5 in vitro, we immunoprecipitated ERK4 from MK5-deficient MEFs and assayed the immunoprecipitate in a coupled kinase assay using recombinant MK5 and Hsp25. The IP of ERK4-specific antibodies but not of a preimmune serum (control) was able to phosphorylate and activate MK5 (Fig. 5B).
We wondered whether activation of MK5 by ERK4 depends on the catalytic activity of ERK4. Hence, we gener-ated an ATP-binding deficient mutant of ERK4 by site-directed mutagenesis. We replaced the lysine residues at positions 49 and 50 of the ATP-binding pocket of ERK4 with alanine residues (K49A,K50A or "KK/AA"). When co-transfected, this kinase ATPbinding mutant was unable to activate GFP-MK5 (Fig. 5C). In the same assay we used the appropriate ERK3 ATP-binding mutant ERK3-K49A,K50A. In contrast to ERK4 and in agreement with previously published data (14), in this experimental system the ERK3 inactive mutant activated MK5 to a certain extent (Fig. 5C), probably by an indirect mechanism (see below).
We next studied the effect of ERK4 kinase activity for the cytoplasmic translocation of MK5. The GFP-tagged ATP-binding mutant of ERK4-KK/AA was expressed in HEK293 cells. Similar to wild type ERK4, this mutant is localized in the cytoplasm and translocated cyan fluorescent protein-MK5 from the nucleus to the cytoplasm (Fig. 5D). Hence, translocation of MK5 to the cytoplasm alone is not sufficient to activate MK5.
MK5 Phosphorylates ERK4-When ERK4 and MK5 were cotransfected in HEK293 cells, additional low mobility bands of GFP-ERK4 were observed in SDS-PAGE of cell lysates (Fig. 3, B and E). Furthermore, phosphate incorporation was found in bands corresponding to ERK4 or ERK4⌬C1 in in vitro kinase assay (Fig. 5, A and C), suggesting phosphorylation of ERK4 by MK5. To test this hypothesis, GFP-ERK4 was co-expressed with GST-MK5 in HEK293 cells and subsequently treated with CIP. After 10 min of treatment, the lower migrating bands began to disappear (Fig. 6A), indicating that they were due to phosphorylation of ERK4. To determine whether MK5 can directly phosphorylate ERK4, HEK 293 cells were co-transfected with wild-type GFP-MK5 and the kinase-dead mutant GFP-MK5-K51E as a control. The additional low mobility band of ERK4 was observed in the presence of wild-type MK5 but not in the control (Fig. 6B). Thus, MK5 catalytic activity is necessary for phosphorylation of ERK4 that probably occurs in a direct manner. In a further experiment, we tested whether catalytic activity of ERK4 is required for MK5-dependent phosphorylation of ERK4. GST-MK5 was co-transfected with GFP-ERK4 or the kinase-dead mutant GFP-ERK4-KK/AA, followed by Western blot against GFP. GFP-ERK4 but not the mutant GFP-ERK4-KK/AA was phosphorylated under these conditions (Fig. 6C). Because ERK4-KK/AA is unable to activate MK5 (Fig. 5C), it is likely that phosphorylation of MK5 is a prerequisite for phosphorylation of ERK4 by MK5.

ERK4
Binds to the C Terminus of MK5-To determine the ERK4-binding site on MK5, we used the C-terminal deletion mutants MK5-1-358, which lacks the p38-binding nuclear localization signal (amino acids 361-364), and MK5-1-368, which contains the nuclear localization signal but lacks the further C terminus (Fig. 7A). After co-transfection of HEK293 cells with GST-ERK4 and GFP-MK5 and the C-terminal deletion mutants, respectively, and GST pull-down, interaction was detected by Western blot using GFP antibody. GFP-tagged MK3, a kinase that is specifically activated by p38 but not by ERK4, was used as a negative control. Both C-terminal deletion mutants of MK5, MK5-1-368 (Fig. 7B), and MK5-1-358 (not shown), did not bind to ERK4 in this assay. Additionally, ERK4 failed to translocate MK5-1-368 (Fig. 7C), confirming that ERK4 binds, similar to ERK3 (14), to the more C-terminal region of MK5.
ERK4 Can Form Protein Complexes with Itself and with ERK3-Recently, a high throughput study on protein interaction in yeast revealed multimerization of ERK3 (19). To examine whether ERK4 forms homodimers and heterodimers with ERK3, HEK293 cells were co-transfected with GST-ERK4 and GST-ERK3, respectively, and BE-ERK4 plasmids. Then BE pulldown was performed and analyzed by Western blot against GST. Specific binding of GST-ERK4 and GST-ERK3 to BE-ERK4 was observed (Fig. 8A), reflecting the ability of ERK4 to form protein complexes consisting of more than one ERK4 and/or additional ERK3 molecules. A BE pull-down experiment with BE-ERK3 gave similar results (Fig. 8B). Therefore, ERK3 and ERK4 are likely to exist in multimeric protein complexes. Because BE-ERK3 pull-down also precipitates the C-terminal deletion mutant ERK4-⌬C3 (not shown), MK5-binding and ERK3-binding regions in ERK4 are different.

DISCUSSION
Here we have shown that the two atypical ERKs, ERK4 and ERK3, behave similarly with respect to MK5 binding, regulation of its subcellular localization, and catalytic activity. Furthermore, ERK4 and ERK3 phosphorylation both depend on the catalytic activity of MK5, probably being direct substrates for MK5. Both atypical ERKs show similar expression patterns during mouse embryogenesis with elevated transcript levels in brain, lung, and kidney, suggesting cooperativity of ERK3 and ERK4 in MK5 activation during development (see supplemental material). Both proteins are subject to exportin1/CRM1-dependent nuclear export and carry a conserved region C-terminal to the catalytic domain that interacts with a sequence C-terminal to the nuclear localization signal and p38-binding site of MK5. The region between amino acids 326 and 340 in ERK4, which is homologous the MK5binding region of ERK3, is also necessary for MK5 binding. Both ERK4 and ERK3 bind to the region of MK5 C-terminal to amino acid 368, probably between amino acids 423 and 472 (15). As a result of interaction with ERK4 and ERK3, nuclear MK5  is exported to the cytoplasm. After activation of MK5, both ERK3 and ERK4 are phosphorylated depending on MK5 catalytic activity. Furthermore, when co-expressed in HEK293 cells, both ERK3 and ERK4 can be detected in an oligomeric protein complex together with MK5. Binding between ERK4 and ERK3 depends neither on the C-terminal extension of ERK4 nor on the MK5-binding site, indicating direct interaction of the catalytic domains. Despite the many similarities, ERK3 and ERK4 seem to differ at least in two points. First, ERK4 protein is significantly more stable than ERK3 that is rapidly degraded in a proteasome-dependent manner (17). The differences in protein stability are independent of the size of the N-terminal tag, challenging the finding that ERK3 is degraded after ubiquitinylation of the free N terminus (20) while strengthening the notion that the C-terminal part of ERK3 is also involved in targeting degradation (21). Oligomerization of ERK3 and ERK4 may affect ERK3 and/or ERK4 stability under certain conditions. However, our preliminary experiments analyzing the effect of co-expression of ERK4 on protein stability of ERK3 in HEK293 cells 3 did not reveal significant differences so far.
Second, catalytically dead ATP pocket mutants of ERK4 and ERK3 differ in their ability to activate MK5. Upon co-transfection in HEK293 cells, ATP pocket mutants of ERK3 still activated MK5 (14), whereas ERK4 mutants failed to do so. The ability of ERK3 mutants to activate MK5 obviously depends on the experimental system applied, because both in an in vitro phosphorylation assay with recombinant proteins expressed in insect Sf-9 cells and in transfected HeLa cells MK5 activation depends on catalytic activity of ERK3 (15). Taking into account that ERK3 and ERK4 co-exist in protein complexes in transfected cells, MK5 activation by catalytically dead ERK3 may be explained by its ability to recruit active endogenous ERK4 of HEK293 cells to the MK5-containing complex. Because ERK4 is more stable than ERK3, the steady state level of endogenous ERK4 in HEK293 cells is probably higher than that of ERK3. This would also explain why the kinase-dead ERK4 mutant fails to recruit sufficient ERK3 in HEK293 cells to activate MK5. In addition, because of its molar excess, overexpressed kinasedead ERK4 can also not quantitatively recruit active endogenous ERK4. Hence, the seemingly different behavior of ERK3 and ERK4 mutants in activation of MK5 may finally be caused by complex formation and differential stability of ERK3 and ERK4 proteins. Hence, protein stability remains the only clear difference detected for these atypical kinases so far. Finally, we should mention that the physiological role of the ERK4/MK5 signaling module is still enigmatic, because extracellular agonists of MK5 have not been identified among the diverse stimuli analyzed at the cellular level (13). 4 Similar to the ERK3/MK5 module (14), the ERK4/MK5 module may play a more prominent role in embryonic or post-natal development. Compound mouse mutants of the different components of these modules will ultimately provide more detailed insight into the physiological role of these atypical ERKs.