One critical component in determining the specificity, and efficiency of MAPK (mitogen-activated protein kinase) substrate phophorylation is the presence of distinct docking domains in the substrate proteins. Docking domains have been shown to be important for the activities of members of the ERK (extracellular-signal-regulated kinase), JNK (c-Jun N-terminal kinase) and p38 subfamilies of MAPKs towards their substrates. Here, we demonstrate that docking domains also play an important role in ERK5-mediated substrate phosphorylation. The presence of a docking domain promotes both phosphorylation of myocyte enhancer factor, MEF2A, in vitro and its activation in vivo by ERK5. Mutational analysis of the MEF2A docking domain demonstrates that the specificity determinants for ERK5 are similar to those observed with members of the p38 subfamily. A docking domain recognized by ERK5 can direct ERK5 to activate heterologous substrates. Deletion analysis demonstrates that as with other MAPKs, it is the catalytic domain of ERK5 that recognizes the docking domain. Our data therefore extend previous observations on other MAPKs and demonstrate that the requirement for specific docking domains in promoting MAPK action towards substrates is a general property of MAPKs.
Several MAPK (mitogen-activated protein kinase) pathways exist in mammals. The best-studied pathways are the ERK (extracellular-signal-regulated kinase), JNK (c-Jun N-terminal kinase) and p38 pathways although additional pathways, exemplified by the ERK5 pathway, have been identified that are not functionally grouped with one of these subfamilies (reviewed in [1,2]). Docking interactions amongst members of the ERK, JNK and p38 MAPKs have been identified as important components in their specificity and efficiency of action, and their regulation. These docking interactions involve both the substrates and regulators of MAPKs (reviewed in [3,4,5]). One type of docking site found in transcription factors (referred to as D-domains) is characterized by a region rich in basic amino acids followed by either a LxL motif and/or a triplet of hydrophobic amino acids. These domains are typically less than 20 amino acids long but show only limited sequence similarity. The D-domains are portable and can confer signalling specificity on a heterologous protein towards a particular pathway. D-domains have been identified that permit specific responses to either the ERK, JNK or p38 families or to various combinations of these. However, to date, it is unclear whether other MAPKs, such as ERK5, also use docking domains on their target proteins.
ERK5 differs from other MAPKs due to the presence of an extended C-terminal domain. This C-terminal region acts as a transcriptional activation domain that can activate myocyte enhancer factor (MEF2) transcription factors through interactions with their MADS (MCM1, agamous, deficiens and SRF)-MEF2 DNA binding domain [6,7]. The C-terminal region also contains a nuclear localization signal . The N-terminal region contains the catalytic kinase domain and the phosphorylation site for the upstream activating MEK5 (MAPK/ERK kinase). ERK5 can respond to both mitogenic signals  and stress stimuli . c-Myc, SAP-1 (SRF accessory protein 1) and MEF2 subfamily members have been identified as ERK5 substrates [6,11–14]. Different phospho-acceptor sites in MEF2A, -2C and -2D appear to be important for ERK5-mediated phosphorylation and activation. In MEF2A, three sites were shown to be important that are all located in the transcriptional activation domain (TAD) (; residues T304, T312 and S347 in our constructs).
The MEF2 subfamily of MADS-box transcription factors have been shown to be targeted by a subset of p38 MAPKs through docking interactions with the MEF2 D-domain . The D-domain is essential for efficient phosphorylation in vitro and activation by the p38 cascade in vivo. The structure of the docking domain complexed with p38 has been solved , and in combination with mutagenic data , a detailed picture has been produced for how the docking domain peptide binds in an extended conformation to the surface of the MAPK. This binding surface is on the opposite side to the catalytic cleft of p38.
In this study we have investigated whether docking interactions with MEF2A are important for its phosphorylation and activation by ERK5. We demonstrate, by deletion and mutagenic analysis and by using chimeric transcription factors, that the D-domain of MEF2 plays an important role in permitting efficient substrate phosphorylation and activation by ERK5. This substrate targeting is mediated by the catalytic domain of ERK5. Thus, in addition to the members of the major classes of MAPKs, ERK, JNK and p38, ERK5 also uses docking interactions to facilitate its ability to phosphorylate and activate transcription factor substrates.
MATERIALS AND METHODS
The following plasmids were used for expressing GST (glutathione S-transferase) fusion proteins in Escherichia coli. Vectors encoding GST fusions to MEF2A(266–413)(pAS860), GST–MEF2A(283–413)(pAS861), GST–MEF2A(266–413)M2 (pAS886) containing the double mutant L273A/V275A, GST–Elk-1(310–428)(pAS545) (where Elk-1 is the Ets-like transcription factor 1) and GST–MEF2A-Elk-1(pAS865) have been described previously . GST–MEF2A(266–413) with mutations (denoted by single-letter amino acid substitution symbols) S268A (pAS1664); R269A (pAS1665); K270A (pAS1666); P271A (pAS1667); D272A (pAS1668); L273A (pAS1669); R274A (pAS1670); V275A (pAS1671); V276A (pAS1672); I277A (pAS1673); P278A (pAS1674); P279A (pAS1675) were described previously .
The following plasmids were used in mammalian cell transfections. pG5E1b reporter construct contains five of the yeast transcription factor, GAL4, DNA binding sites cloned upstream of a minimal promoter element and the firefly luciferase gene . pMEF-Luc contains three copies of the MEF2A binding sites located upstream from a luciferase reporter gene (kindly provided by Dr A. Bannister and Professor T. Kouzarides, Cambridge University, Cambridge, U.K.). Vectors pSRα-HA-ERK5 encoding HA-tagged ERK5, and pSRα-MEK5(D) encoding constitutively active MEK5 were kindly provided by Professor E. Nishida (Kyoto University, Kyoto, Japan) and were described previously . pAS1131 [encoding pCMV (cytomegalovirus)-driven mouse Flag-tagged ERK5 N-terminal amino acids 1–440; ERK5-kin] and pAS1132 (encoding pCMV-driven mouse Flag-tagged ERK5 C-terminal amino acids 512–877; ERK5-TAD) were constructed by inserting HindIII/BamHI fragments from pAS1129 and pAS1130 respectively into the same sites in pCDNA3. pAS1129 and pAS1130 were constructed by inserting BamHI/XbaI cleaved PCR fragments (primers ADS 955/957 and ADS 956/982) into the same sites of pAS801 and pAS788 respectively. pGAL-c-Jun [encoding GAL4 DNA binding domain (DBD) fused to c-Jun], pAS883 (pCMV5-GAL4-MEF2A), pAS884(pCMV5-GAL4-MEF2AΔD), pAS900 (encoding GAL4 DBD fused to the Elk-1 TAD), pAS889 and pAS888 (encoding GAL4 DBD fused to the MEF2 docking domain and the c-Jun and Elk-1 TADs respectively) have been described previously . pAS2076 [encoding CMV-driven, Flag-tagged wild-type (WT) full-length MEF2A] and pAS2077 [encoding CMV-driven, Flag-tagged full-length MEF2A(M3); mutations I277A/P278A] were constructed by inserting KpnI/EcoRI cleaved fragments from pAS2074 and pAS2075 respectively into the same sites in pCDNA3. pAS2074 and pAS2075 were constructed by inserting NcoI/EcoRI cleaved fragments from pAS1203  and pAS2073 respectively into the same sites in pRSETB. pAS2073 was constructed by inserting the NcoI/XhoI fragment from pAS1397  into the same sites in pET-nef-PFH. pAS1203 and pAS2073 were also used for expressing His-tagged MEF2A derivatives in E. coli.
All plasmids made by using PCR were subjected to automated dideoxy sequencing.
Tissue culture, cell transfection and reporter assays
293 and COS7 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (Gibco BRL). Transfections were performed using Superfect transfection reagent (Qiagen) as described previously .
For reporter gene assays a luciferase reporter construct controlled by a GAL4-driven promoter pG5E1bLUC (0.4–1 μg) was co-transfected with pCMV promoter-driven vectors encoding various GAL4-MEF2A fusion proteins (40–100 ng) and, where indicated, 40–300 ng of MEK5(D) and ERK5 derivatives were transfected.
Transfection efficiencies were normalized by co-transfecting pCH110 plasmid (Pharmacia) and measuring β-galactosidase activities. Cell extracts were prepared and luciferase and β-galacto-sidase assays were performed as described previously .
Protein kinase assays
Recombinant protein kinases, HA-tagged ERK5 and Flag-tagged ERK5-kin, were immunoprecipitated from transfected COS7 cells essentially as described previously, using the appropriate immobilized antibodies , except that cells were lysed in 100 μl triton lysis buffer (Figures 2 and 4). An alternative method of isolating the kinase was also employed for Figures 3 and 5 as described previously . To produce activated ERK5 derivatives, cells were co-transfected with MEK5(D). Protein kinase assays were carried out essentially as described previously (Figures 2 and 4, ; Figures 3 and 5, ) however, the ERK5 kinase was not eluted from the beads prior to carrying out the kinase reaction. Substrate His-tagged and GST fusion proteins were expressed in E. coli BL21 or JM101 and purified as described previously .
GAL4-fusion proteins were detected in total 293 cell extracts using anti GAL4 antibody against the N-terminal DNA-binding domain (Santa Cruz Biotechnology, Santa Cruz, CA, U.S.A.). Full-length MEF2A derivatives were detected in COS7 cells using an anti-MEF2 antibody (SC10794x, Santa Cruz). Immunocomplexes were detected by using horseradish peroxidase conjugated secondary antibody followed by enhanced chemiluminescence (Pierce).
Figure generation and data quantification
Figures were generated electronically by using Adobe Photo-deluxe Business Edition version 1.0 and Powerpoint version 7.0 (Microsoft) software. Data from Western blots are computer generated images (FluorS Max and Quantity One-BIO-RAD). Phosphorimager data from kinase assays were quantified using Quantity One software (BIO-RAD).
Requirement of the D-domain of MEF2A for efficient ERK5-mediated activation of MEF2A
The D-domain of MEF2A has been shown previously to act as a kinase-docking domain and to be important for the efficient phosphorylation of the MEF2A TAD in vitro and activation in vivo by a subset of p38 MAPKs . MEF2A and the closely related protein MEF2C have also been shown to be targets of the ERK5 pathway [13,14]. Therefore, we wished to examine whether the D-domain of MEF2A played a role in its activation by ERK5.
GAL4 fusion proteins were tested that contained the trans-activation domain of MEF2A, but differed by the inclusion or omission of the upstream D-domain (Figure 1A). In the presence of co-transfected ERK5 and a constitutively active form of its upstream activator MEK5, the transcriptional activity of MEF2A was greatly enhanced (15-fold; Figure 1B). In contrast, in the absence of the D-domain in MEF2AΔD, this ERK5-mediated activation was severely ablated (2-fold; Figure 1B). Western blotting demonstrated that this differential increase was not due to changes in protein levels, as both wild-type and D-domain deleted versions of MEF2A were moderately enhanced by transfection of ERK5/MEK5 (Figure 1C).
The D-domain of MEF2A is required for ERK5-mediated transcriptional activation
The D-domain of MEF2A is therefore an important determinant of ERK5-mediated activation of its TAD.
Importance of the D-domain of MEF2A for efficient ERK5-mediated phosphorylation of MEF2A
By analogy with other MAPK-mediated activation events, the most likely explanation for the effect of the D-domain on the TAD activity of MEF2A is by directly increasing the ability of ERK5 to phosphorylate its substrate. To establish whether the D-domain of MEF2A plays a direct role in ERK5-mediated phosphorylation, we analysed the ability of ERK5 to phosphorylate GST–MEF2A fusion proteins.
First we compared the ability of ERK5 to phosphorylate GST–MEF2A fusion proteins in the presence and absence of the D-domain. While MEF2A is efficiently phosphorylated by activated ERK5, phosphorylation of MEF2AΔD is severely compromised (Figure 2B, lanes ‘M2’ and ‘ΔD’ respectively).
Identification of the MEF2A D-domain residues required for targeting of ERK5 by alanine scanning
To establish which amino acids were important for recognition by the ERK5 MAPK, we analysed an alanine scanning series across the MEF2A D-domain. With the exception of Asp-272 and Pro-279, all the point mutations resulted in at least a 40% decrease in phosphorylation by ERK5 (Figure 2B, lanes 1–12). The greatest decrease was seen with Leu-273 and Val-275 mutants that were reduced to virtually the same level as complete deletion of the D-domain (Figure 2B, lanes L273 and V275). The phosphorylation of the M2 mutant (containing the double L273A and V275A mutations) was similarly reduced to basal levels (Figure 2B, lane M2). Leu-273 and Val-275 have been implicated previously as important determinants for p38 MAPK-mediated MEF2A phosphorylation [16,18].
Taken together, therefore, these results demonstrate that the D-domain is an important determinant for phosphorylation of MEF2A by ERK5 and they identify the amino acids in the D-domain important in this process.
The MEF2A D-domain is a portable ERK5-docking motif
One characteristic of MAPK docking domains is that they are portable motifs that can confer ERK5-responsiveness on heterologous substrates. Therefore, we tested the activity of the MEF2A D-domain in the context of two different potential substrates, c-Jun and Elk-1. The docking domain of c-Jun usually specifies this as a JNK substrate , whereas the analogous domain in Elk-1 specifies it as a JNK and ERK2 substrate .
The MEF2A D-domain was fused to the TAD of c-Jun and Elk-1 in place of their normal MAPK docking domains and the resulting chimeric proteins (Figure 3A) were tested as GAL4 fusion proteins for activation by ERK5 (Figure 3B). Neither c-Jun nor Elk-1 were greatly activated by ERK5 in vivo (Figure 3B). The latter observation is consistent with previous data that demonstrated that Elk-1 is not a target of the ERK5 pathway . Fusion of the MEF2 docking domain to the c-Jun TAD resulted in a small increase in ERK5 responsiveness (3-fold; Figure 3B). In contrast, fusion of the MEF2 D-domain to the Elk-1 TAD caused a large increase in ERK5-responsiveness (85-fold; Figure 3B). This enhancement was not due to an increase in the levels of the MEF2-Elk-1 fusion protein (Figure 3B, insert).
The MEF2A D-domain acts as a portable ERK5 docking motif
To correlate these changes in the transactivation capacity of the Elk-1 TAD with enhanced propensity for phosphorylation by ERK5, analogous GST fusion proteins were tested as substrates for ERK5 in vitro. While Elk-1 is a poor ERK5 substrate, the MEF2–Elk-1 fusion protein shows enhanced phosphorylation by ERK5 [Figure 3C, compare lanes 1–4 with 5–8 (upper panel); summarized in graph (lower panel)].
Collectively, these data demonstrate that the MEF2A D-domain can act as a portable docking motif and can confer ERK5-responsiveness on a heterologous substrate.
MAPK docking specificity is mediated by the catalytic domain of ERK5
In contrast to other MAPKs, ERK5 contains a long C-terminal extension to the catalytic domain (Figure 4A). Several activities have been attributed to this motif, including binding to the DNA binding domain of MEF2A . As other MAPKs use their catalytic domains to bind to the docking domains on substrates, we analysed whether the same was true for ERK5. Therefore, we constructed an expression vector encoding the N-terminal catalytic kinase domain of ERK5 (ERK5-kin; Figure 4A).
The ERK5 catalytic domain retains targeting activity towards the MEF2 D-domain
First we tested the ability of ERK5-kin to activate the MEF2A TAD in vivo. In the absence of co-transfected MEK5, very little effect was seen on the reporter; however, co-transfection of MEK5 and ERK5-kin led to enhancement of the activity of the MEF2A TAD (12-fold; Figure 4B).
To establish the D-domain dependence of this activation, the activity of MEF2AΔD, lacking the D-domain, was analysed. In the absence of the D-domain, no enhancement of the MEF2A TAD was seen by ERK5-kin (Figure 4C). In contrast, the ERK5 C-terminal region (ERK5-TAD) is unable to enhance the activity of the MEF2A TAD, irrespective of the presence of the D-domain (Figure 4D). To confirm that the differences in transactivation we observe are due to a change in the ability of ERK5-kin to phosphorylate MEF2A, we compared the activity of ERK5-kin towards GST–MEF2A fusion proteins in vitro. In the absence of the MEF2A D-domain, little phosphorylation of MEF2A was observed. However, phosphorylation of MEF2A by ERK5-kin was greatly enhanced in the presence of the D-domain (Figure 4E, lanes 3 and 4).
Collectively, these data demonstrate that substrate targeting through docking domains is mediated by the catalytic domain of ERK5.
Efficient ERK5-mediated activation of MEF2A-driven gene expression requires the MAPK docking motif
The MAPK docking motif in MEF2A clearly plays an important role in regulating the phosphorylation of the MEF2A TAD and its response to the ERK5 pathway (Figures 1 and 2). To investigate whether the docking motif also plays an important role in the response of the full-length protein to ERK5, we first tested the phosphorylation of WT and a D-domain mutant (M3) versions of MEF2A by ERK5 in vitro. In these experiments, we used the M3 mutant version of MEF2A that contains the mutations I277A and P278A. Individually these mutations cause large decreases in phosphorylation by ERK5 in the context of the isolated MEF2A TAD (Figure 2), and together play an important role in p38-mediated phosphorylation of MEF2A . MEF2A(WT) is phosphorylated efficiently by ERK5 after 10 min incubation, however little phosphorylation of MEF2A(M3) is observed, even after 40 min (Figure 5A).
The MEF2A D-domain plays an important role in the response of MEF2-dependent genes to ERK5 activation
Next, we tested the activity of a reporter gene driven by MEF2 binding sites in the presence of full-length versions of WT and mutant MEF2A (Figure 5B). The MEF2-driven reporter gene was efficiently activated in the presence of WT MEF2A and activated ERK5. However, in the presence of the mutant version of MEF2A, MEF2A(M3), the activity of the reporter in response to activated ERK5 was much reduced (14% of WT protein). This difference was not due to differing amounts of MEF2A protein in the cells (Figure 5B, bottom panel). Thus, the MAPK docking domain in MEF2A plays an important role in the response of MEF2A to the ERK5 pathway in vivo when investigating either the isolated transcriptional activation domain or the full-length protein.
Docking domains play a major role in dictating the specificity of substrate phosphorylation by MAPKs [3,4,5]. Here we demonstrate that in addition to members of the major classes of MAPKs, ERK, JNK and p38, ERK5 also uses docking interactions to facilitate its ability to phosphorylate and activate transcription factor substrates.
ERK5 has previously been shown to bind to MEF2 transcription factors though interactions between the MADS-MEF2 DNA binding domain and the C-terminal extension in ERK5 [6,7]. This interaction is thought to bring the TAD located in the C-terminus of ERK5 into a DNA-bound complex with MEF2 proteins and hence activate transcription. However, this interaction appears not to affect MEF2 phosphorylation itself . Here we have shown that the kinase domain of ERK5 is targeted to the D-domain of MEF2A. This interaction is clearly important in enhancing the ability of ERK5 to potentiate the activity of the MEF2A TAD in response to ERK5 pathway activation (Figures 1B, 4B and 4C). Mutational analysis of the D-domain in the context of full-length MEF2A also reduces its transcriptional activation response to ERK5 activation (Figure 5B). This enhancement is mirrored by the observation that the MEF2A D-domain is also required for efficient phosphorylation of the TAD (Figure 2B and 4E) and the full-length protein (Figure 5A) by ERK5 in vitro. Thus a model emerges in which ERK5 can potentially interact with two different regions of MEF2A (Figure 5C). Interactions between the C-terminal end of ERK5 and the N-terminus of MEF2A can cause transcriptional enhancement of MEF2 activity by recruitment of the ERK5 TAD. Further interactions between the kinase domain of ERK5 and the D-domain of MEF2A can then lead to phosphorylation and activation of the MEF2A TAD in a signal-dependent manner. Together, this permits a high level of activation to be achieved through MEF2A via the ERK5 pathway.
Our data further extend the paradigm that docking domains play an important role in mediating the downstream effects of MAPK signalling. Previous studies on MEF2A demonstrated that two p38 family members, p38α and p38β2, are specifically targeted to MEF2A through the D-domain . Alanine scanning revealed the residues in MEF2A that are important in this process  and structural studies confirmed this mutagenic analysis . A comparison of the alanine scanning data for ERK5-mediated MEF2A phosphorylation (Figure 2B) and previous data with p38 MAPKs, demonstrates an almost identical profile. In particular, Leu-273 and Val-275 were particularly sensitive to mutation in both cases. Thus, the docking requirements for a subset of p38 MAPKs and ERK5 appear very similar. This suggests that the interaction surface on the kinase is also likely to be similar. Indeed, a comparison of the residues located at the analogous positions in ERK5 to those in p38α that form the docking groove for MEF2A  reveals that of the 15 residues involved, seven are identical and seven are highly conserved between the two proteins. Consistent with this conclusion is the observation that SAP-1 is also a p38 and ERK5 substrate. Substrate recognition in this case is also via a docking domain that is recognized specifically by the p38α and p38β2 MAPKs . It remains to be seen how far this overlap between substrate specificity through MAPK docking domains extends.
Neither Elk-1 (Figure 3A; ) nor c-Jun (Figure 3A) represent ERK5 targets in vivo. However, by inserting the MEF2 docking domain in place of the natural docking domains of these substrates, ERK5 responsiveness could be conferred on these transcription factors (Figure 3). However, while MEF2 D-domain fusions to Elk-1 were efficiently activated by ERK5, the analogous fusions to c-Jun were only slightly responsive. Thus, in addition to the docking interaction, the local context of the phospho-acceptor motifs clearly plays an important part in the response of transcription factors to ERK5.
In summary, our data further extend the paradigm of MAPK docking domains being important for generating substrate specificity to ERK5. It appears likely that other MAPKs, not studied to date, will also use such interactions.
We thank Linda Shore for excellent technical assistance. We also thank Dr Alan Whitmarsh and Dr Shen-Hsi Yang (University of Manchester, Manchester, U.K.) for comments on the manuscript. We are grateful to Dr A. Bannister, Professor T. Kouzarides (Cambridge University, Cambridge, U.K.) and E. Nishida (Kyoto University, Kyoto, Japan) for reagents. This work was supported by grants from Cancer Research UK, the Wellcome Trust and a Lister Institute of Preventative Medicine Research Fellowship to A. D. S.
DNA binding domain
MAPK docking domain
Ets-like transcription factor 1
N-terminal catalytic kinase domain of ERK5
ERK5 C-terminal region
c-Jun N-terminal kinase
(box), domain found in MCM1, agamous, deficiens and SRF
mitogen-activated protein kinase
myocyte enhancer factor 2
MEF2A lacking a D-domain
a mutant form of MEK that is constitutively active
etc., substitution of Ser-268 with alanine etc. (denoted by single-letter amino acid coding)
SRF accessory protein 1
transcriptional activation domain
Present address: Ontario Cancer Institute, University of Toronto, 610 University Avenue, Toronto, Ontario, M5G 2M9, Canada.