Molecular cloning of mouse ERK5/BMK1 splice variants and characterization of ERK5 functional domains.

The mitogen-activated protein kinases (MAPKs) play important roles in regulation of cell growth and survival. Human MAPK 5 (ERK5) or Big MAP kinase 1 (BMK1) is a recently cloned member of the MAPK family. To identify ERK5-related kinases, we searched the GenBanktrade mark expressed sequence tag (EST) data base for mouse cDNAs with homology to human ERK5. A full-length mouse cDNA that was highly homologous to the human ERK5 was identified. Further analysis of ERK5 polymerase chain reaction products generated from mouse embryo cDNA yielded three mouse ERK5 cDNAs (mERK5a, mERK5b, and mERK5c). Sequence analysis showed that these cDNAs are alternative splice products of the mouse ERK5 gene. Interestingly, expressed mERK5b and mERK5c act as dominant negative inhibitors based on inhibition of mERK5a kinase activity and mERK5a-mediated MEF2C transactivation. However, the physiological significance of mERK5b and mERK5c is not fully understood. Further investigation using these mouse ERK5 splice variants and other constructed mutants identified functional roles of several regions of mERK5, which appear to be important for protein-protein interaction and intracellular localization. Specifically, we found that the long C-terminal tail, which contains a putative nuclear localization signal, is not required for activation and kinase activity but is responsible for the activation of nuclear transcription factor MEF2C due to nuclear targeting. In addition, the N-terminal domain spanning amino acids (aa) 1-77 is important for cytoplasmic targeting; the domain from aa 78 to 139 is required for association with the upstream kinase MEK5; and the domain from aa 140-406 is necessary for oligomerization. Taken together, these observations indicate that ERK5 is regulated by distinct mechanisms determined by its unique structure and presumably the presence of multiple splice variants.

Human ERK5 was recently cloned by two groups (8,9). Human ERK5 contains 816 amino acid residues with a primary structure distinct from other MAPK members. ERK5 has a unique long C-tail and a distinct loop-12 domain. ERK5 is activated by reactive oxygen species (15), hyperosmolarity (16), and fluid shear stress (16). Most recently, it has been shown that ERK5 is required for EGF-induced cell proliferation and progression through the cell cycle (17). Although ERK5 has a TEY motif in its dual phosphorylation site similar to ERK1/2 and ERK7, several studies have shown that ERK5 has different upstream activators and downstream substrates compared with other MAPKs.
In the present study, we report the identification of three alternatively spliced mouse ERK5 cDNAs, termed mERK5a, mERK5b, and mERK5c. The putative protein sequences deduced from these three cDNAs are identical except in their N-terminal regions. mERK5b and mERK5c lack N-terminal 69 amino acids and 139 amino acids, respectively, compared with mERK5a. It is likely that the variations in N-terminal sequences are produced by alternatively choosing different splicing donors and acceptors within a single intron. More interestingly, mERK5b and mERK5c, which are catalytically inactive, act as dominant negative kinases. Finally, by using N-or C-terminal truncated mERK5, we found that three N-terminal domains spanning aa 1-77, aa 78 -139, and aa 140 -406 are important for cytoplasmic targeting, association with the upstream kinase MEK5, and oligomerization, respectively. The C-terminal tail is essential for the biological activity of ERK5 in vivo by mediating nucleus translocation, which is dependent upon the NLS in the C-terminal region. These observations should help us understand the biological reasons for this diversity in ERK5 and may provide new approaches to modify ERK5 signaling.

MATERIALS AND METHODS
Isolation of Mouse ERK5 Isoforms-The GenBank expressed sequence tag (EST) data base was searched using the program BLAST with amino acid sequences corresponding to the human ERK5. Several EST fragments displayed high degrees of amino acid homology. Among them, a full-length mouse clone was obtained from Genome Systems Inc (GenBank accession number AA288345). A manual sequencing method with the Sequetherm Cycle sequencing kit (Epicentre Technologies Corp.) was used to sequence this clone according to the manufacturer's protocols. Several PCR fragments were found in the first-strand cDNAs of a mouse embryo (purchased from CLONTECH) and were subcloned into the TA vector (Invitrogen) for sequencing. Combining analysis of mouse ERK5 RT-PCR clones with mouse EST clones, three mouse ERK5 cDNAs were identified. To determine the mechanism for generating three mouse ERK5 isoforms, a mouse genomic DNA fragment encompassing the mouse ERK5 isoforms splicing junction was isolated by PCR amplification using oligonucleotides 5Ј-ACGAGTAC-GAGATCATCGAGACC-3Ј and 5Ј-GGTCACCACATCAAAAGCATT-AGG-3Ј.
Construction of Expression Plasmids-Clones for mouse ERK5a (nucleotides 10 -2776), ERK5b (nucleotides 280 -2826), and ERK5c (nucleotides 837-3196) were subcloned into the multiple cloning regions of the pcDNA3.1/His vectors (Invitrogen). All isoforms were fused in-frame with an N-terminal Xpress tag. mERK5a(-tail) (aa 1-406) was constructed by removing aa 407-806 with PstI digestion. GFP-tail and GFP-NLS were generated by ligation of the PCR-amplified C-tail (aa 407-806) and NLS fragment (aa 505-539), respectively, containing the artificial restriction sites BamHI and EcoRI to the pEGFP-C3 (CLON-TECH) cut with BamHI and EcoRI. mERK5a(-NLS), lacking aa 505-539, was generated by PCR with the Expand High-Fidelity PCR system (Roche Molecular Biochemicals). Briefly, an antisense primer matching the DNA sequence upstream of the codon aa 505 and a sense primer downstream of the codon aa 539 were generated. PCR reaction was performed using the entire mERK5a in the PcDNA3.1/His vector as the template, and the PCR product was treated with T4 DNA polymerase to create a blunt end. Purified PCR product was further treated with polynucleotide kinase and T4 ligase and subsequently transformed into the competent bacterial DH5␣.
Cell fractionation was performed by extraction with different buffer and sequential centrifugation. First, cells were lysed in the TME lysis buffer contain 0.1% Triton X-100, and the cell lysate was centrifuged at 10,000 ϫ g for 1 h. The supernatant was collected as the cytoplasmic fraction. The pellet was then resuspended in the lysis buffer containing 1% Triton X-100 and centrifuged at 10,000 ϫ g for 1 h. The supernatant was collected as the nuclear fraction. The pellet was further resuspended in the RIPA buffer (20 mM Tris-HCl, pH 7.5, 2.5 mM EDTA, 1% Triton X-100, 10% glycerol, 1% deoxycholic acid, 0.1% SDS, 10 g/ml leupeptin, and 10 g/ml aprotinin) and centrifuged at 10,000 ϫ g for 1 h. The supernatant was collected as the cytoskeleton fraction, and the pellet was resuspended in sample buffer as the nuclear fraction.
Cell lysates, cellular fractionates, or tissue extracts were boiled in the presence of 1ϫ sample buffer (50 mM Tris-HCl, pH 6.8, 10% glycerol, 2% SDS, 4% ␤-mercaptoethanol, 0.02% bromphenol blue) and subjected to SDS-PAGE, and proteins were then transferred to nitrocellulose. The membrane was blocked for 1-2 h at room temperature with a commercial blocking buffer (Life Technologies, Inc.). The blot was incubated for 1 h at room temperature with the primary antibody (anti-Xpress antibody from Invitrogen), followed by incubation for 1 h with secondary antibody (horseradish peroxidase-conjugated). Immunoreactive bands were visualized by chemiluminescence (ECL, Amersham Pharmacia Biotech).
Immunoprecipitation and Immune Complex Kinase Assay-Immune complex kinase assays were performed with ectopically expressed tagged ERK5 proteins from CHO-K1 cells as previously described (19), except 2 g of MBP was used per reaction in kinase buffer. Proteins were separated by 15% SDS-PAGE, transferred to a nitrocellulose, and subjected to autoradiography. The presence of epitope-tagged proteins in immunoprecipitates was verified by Western analysis with antibody against the tag.
RNA Isolation and RT-PCR-Total RNA was prepared from multiple adult mouse tissues and mouse embryos (15 days) using an RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Two degenerate oligonucleotide primers were designed corresponding to splicing junction sequences of three cDNAs: the sense primer 5Ј-AC-GAGTACGAGATCATCGAGACC-3Ј and antisense primer 5Ј-GGTCAC-CACATCAAAAGCATTAGG-3Ј. The first-strand cDNA was synthesized by Superscript II reverse-transcriptase (Life Technologies, Inc.) with antisense primer. The amplification was carried out in a 100-l mixture containing 2 l of the first-strand cDNA product, 10 M each of the sense and antisense primer, and 5 units of Taq DNA polymerase (Life Technologies, Inc.). The PCR reaction was performed as follows: initial denaturation at 94°C for 3 min and 30 cycles of amplification (denaturation at 94°C for 1 min, annealing at 55°C for 2 min, extension at 72°C for 2 min), followed by a final extension step of 10 min at 72°C. Reactions were electrophoresed on a 1.5% agarose gel.
Reporter Gene Expression-The MEF2C fusion activator vector, encoding the GAL4-binding domain fused to MEF2C activation domain (14), was cotransfected into CHO-K1 cells along with the GAL4-responsive reporter plasmid pG5E1bLUC, which contains five GAL4 sites cloned upstream of a minimal promoter driving a luciferase gene (14). For transfection, CHO-K1 cells (0.2 ϫ 10 6 cells per well) were seeded into 24-well plates the day before transfection. Cells were transfected with 0.5 g of DNA in total per well using LipofectAMINE Plus (Life Technologies). After 5 h, the transfection was stopped by adding equal volume of DMEM/F-12 (10% fetal bovine serum). After 24 h, the medium was changed to a serum-free DMEM/F-12 for an additional 24 h. Then cells were collected for luciferase assay. In the cases of testing the endogenous ERK5, 10% serum was added and the cells were incubated for an additional 4 h before harvesting. A green fluorescence protein (GFP) expression vector (pEGFP-N1, from CLONTECH) was used to control for transfection efficiency. The total amount of DNA for each well was kept constant using the empty vector pcDNA3.1/His (Invitrogen). Luciferase assays were performed with a Luciferase Reporter Gene Assay kit (Roche Molecular Biochemicals) as instructed. Briefly, cells were washed twice with PBS and lysed in 200 l of lysis buffer at room temperature for 15 min with shaking. 50 l of cell extracts was transferred into a 96-well microtiter plate. The fluorescence intensity of GFP was measured using a Wallace multicounter (Wallace). 50 l of luciferase substrate were then added to the cell lysates, and the luciferase activities were determined by measuring luminescence intensity using the same Wallace multicounter. To correct for transfection efficiency, the luciferase activity was divided by the green fluorescence intensity.
Immunocytochemistry-CHO-K1 cells, grown on LabTek II chamber slides, were cotransfected with Xpress-tagged ERK constructs in the presence of either pCDNA3 vector or MEK5(D). Following the transfection for 48 h, the cells were then fixed in 10% formalin in PBS for 15 min, washed, blocked, and incubated with anti-Xpress antibody at a 1:2000 dilution. After a 1-h incubation at room temperature, the cells were washed and then incubated with fluorescein-conjugated antimouse antibody (Vector) at a 1:200 dilution. The stained cells were analyzed under an Olympus Fluoview confocal microscope.

Cloning of Three Different Mouse ERK5 Splice Variants and
Genomic Sequence Analysis-To identify mouse ERK5 cDNAs, the EST data base was searched, and several EST clones with high homology to human ERK5 were found. Among these, a full-length EST cDNA clone (AA288345) was purchased and sequenced. Comparing this sequence with the published human ERK5 sequence (8,9), we found that this cDNA clone contained an insert as well as an in-frame stop codon within the insert. This suggested that multiple ERK5 cDNAs might exist.
To confirm the presence of diverse species of ERK5 cDNAs in the mouse, mouse embryo first-strand cDNA (from CLON-TECH) was used for PCR with primers from conserved regions of human and mouse ERK5. Sequence analysis showed that at least three different ERK5 cDNAs were present, which we will designate as mERK5a, mERK5b, and mERK5c (Fig. 1A). mERK5a is the mouse cDNA most homologous to human ERK5. mERK5b and mERK5c cDNAs contain one or two inserts compared with mERK5a, respectively. Although mERK5b has a 50-bp insert (I-1) (Fig. 1A), mERK5c, which corresponds to the mouse EST clone mentioned above, contains both the 50-bp (I-1) insert found in mERK5b and a 91-bp (I-2) insert (Fig. 1A). Both inserts start with consensus splice donor "gt," and the second insert (I-2) ends with an acceptor sequence "ag," suggesting that they are likely to be intron sequences that were not spliced out.
To explore the mechanism for producing these three different mouse ERK5 cDNAs, PCR and sequencing of the mouse genomic DNA flanking the splicing region of the ERK5 gene was performed. The genomic structure of the splicing junction of mouse ERK5 gene is shown in Fig. 1B. Sequence analysis of genomic DNA showed an additional 429-bp (I-3) insert with consensus "gt/ag" just after I-2. It is likely that the variations in N-terminal sequences are produced by alternatively choosing different splicing donors and acceptors within a single intron that is composed of I-1, I-2, and I-3. mERK5a is likely generated by use of the splicing donor D1 and the acceptor A2, whereas mERK5b is generated by use of D2 and A2 and mERK5c is generated by use of D3 and A2. The I-1 intron introduces a stop codon and causes mERK5b and mERK5c to have shorter N termini than mERK5a. Protein Sequence Analysis-mERK5a contains a putative open reading frame (ORF) from nucleotide 27 to nucleotide 2447 that encodes a protein of 806 amino acids with a predicted molecular mass of 88 kDa, whereas mERK5b (putative ORF: The deduced amino acid sequence of human ERK5 (hERK5) was aligned with that of mERK5a using the BLAST program. The prolinerich region and the nuclear localization signal (NLS) are underlined. The activating phosphorylation motif (TEY) is enclosed in the box. Identical residues between the two sequences are indicated by vertical lines. Gaps, denoted by periods, are introduced into the sequence to optimize alignment. nucleotides 284 -2497) contains 737 amino acid with a predicted molecular mass of 80 kDa and mERK5c (putative ORF: nucleotide 864 -2867) consists of 667 amino acid residues with a predicted molecular mass of 73 kDa. The predicted N-terminal amino acid sequences for the three isoforms of ERK5 are shown in Fig. 2A. It is important to note that mERK5b and mERK5c lack the GXGXXG domain for ATP binding, which is present in mERK5a (underlined in Fig. 2A).
Further characterization of the mouse ERK5 protein sequence with currently available profile data bases resulted in the identification of a proline-rich region and a bipartite nuclear localization signal (NLS) in the C-terminal domain (Fig.  2B). The proline-rich region and the NLS are located at aa 578 -690 and aa 505-539, respectively. Comparison of the deduced amino acid sequence of mouse ERK5a with human ERK5 showed 91% homology to human ERK5 (Fig. 2B). Many functional domains important for kinase activity, including the TEY phosphorylation site, are conserved between human and mouse ERK5. The major differences between human and mouse ERK5 occur in a small portion of the N terminus and the proline-rich region in the C terminus.

Presence of Mouse ERK5 Splice Variant Expression in Mouse
Embryo and Adult Mouse Tissues-To confirm the existence of the three forms of mouse ERK5 mRNAs, we performed RT-PCR with a different source of mouse embryo mRNA. Three bands corresponding to PCR products from the three recombinant cDNAs were detected (Fig. 3A). To determine the presence of three mouse ERK5 mRNAs in adult mouse tissues and examine tissue-specific expression, RT-PCR was performed for multiple adult mouse tissues. PCR products encoding the three mouse ERK5 mRNAs were detected in all mouse adult tissues examined (Fig. 3B).
To confirm the presence of the endogenous protein products corresponding to the three splice variants, we used a polyclonal antibody against ERK5, which was made using amino acids EGHGMNPADIESLQREIQMDSPML of the human ERK5 as antigen. This human ERK5 peptide is 100% similar to the corresponding amino acids of the three mouse isoforms. Preliminary data showed that it recognized mouse and human ERK5 equally well. Immunoblotting of mouse embryonic proteins revealed three distinct bands (Fig. 3C) with the molecular weights corresponding to mERK5a, mERK5b, and mERK5c, respectively. Relative protein levels of three splice variants are consistent with their relative mRNA levels, mERK5a Ͼ mERK5c Ͼ mERK5b.
Kinase Activities and Effects of Different Mouse ERK5 Isoforms on Transactivation of the Transcription Factor MEF2C-Because mERK5b and mERK5c lack the GXGXXG domain required for ATP binding, it is very likely that mERK5b and mERK5c have no kinase activity. To further explore the function of mERK5b and mERK5c, phosphorylation of myelin basic protein (MBP) was evaluated by an in vitro kinase assay. Immunoprecipitated mERK5a, isolated from the cells coexpressing constitutively active MEK5(D), rapidly phosphorylated MBP (Fig. 4). In contrast, mERK5b and mERK5c failed to phosphorylate MBP, which is similar to results using human dominant negative ERK5 (DN-hERK5) (14). To investigate the possibility that mERK5b and mERK5c act as dominant negative isoforms, mERK5b or mERK5c were cotransfected with mERK5a and kinase assays were performed. As shown in the last two lanes of Fig. 4, mERK5b inhibited mERK5a kinase activity. mERK5c behaved in a similar manner (data not shown).
Because the transactivation of the transcription factor MEF2C is stimulated by human ERK5-induced phosphorylation (14), we measured MEF2C activity as a means to determine the kinase activities of different ERK5 isoforms. Utilizing fusion proteins containing the transactivation domain of MEF2C fused to the DNA binding domain of the yeast tran- . Lane 1 is a negative control. Lanes 3, 4, and 5 represent three mERK5 PCR products from three mERK5 cDNAs, using the same primer pairs, which were used as size markers. B, mRNA expression in multiple adult tissues. RT-PCR for multiple tissue mRNAs were carried out as described under "Materials and Methods." Lanes 1, 2, and 3 are PCR products from three mERK5 cDNAs used as size markers. Lanes 5,7,9,11, and 13 are RT-PCR products performed in the absence of reverse transcriptase, used as a negative control for contamination. Lanes 4, 6, 8, 10, and 12 are RT-PCR products for multiple tissues as indicated. C, protein expression detected by Western blotting. Lane 1 is total mouse embryo protein; lanes 2, 3, and 4 are three expressed recombinant mouse ERK5 proteins, which were used as size markers. scription factor GAL4, we were able to assess the effects of the mouse ERK5 isoforms on the activity of transcription factor MEF2C fusion protein. This was done by measuring the luciferase activity from CHO-K1 cells cotransfected with a construct containing five copies of the GAL4-binding site upstream of a luciferase reporter gene. As expected, MEF2C-dependent reporter gene expression was enhanced dramatically when mERK5a and MEK5(D) were cotransfected into CHO-K1 cells (Fig. 5A). mERK5b and mERK5c, similar to DN-hERK5, did not stimulate MEF2C activity (Fig. 5A). The same results were observed after extracellular stimulation with 10% serum (Fig.  5B). To further demonstrate that mERK5b and mERK5c behave as dominant negative forms, we coexpressed mERK5b, mERK5c, or DN-hERK5 with mERK5a. Introduction of mERK5b (Fig. 5C) and mERK5c (Fig. 5D), similar to DN-hERK5 (Fig. 5E), inhibited mERK5a-induced MEF2C-dependent reporter gene expression in a dose-dependent manner. mERK5b and mERK5c also dose-dependently inhibited MEF2C activation by endogenous ERK5 after stimulation with 10% serum (Fig. 5F). These results suggest that mERK5b and mERK5c may function as dominant negative inhibitors of the ERK5 signaling pathway.
Subcellular Localization of Mouse ERK5 Isoforms-To deter-mine the role of the N-terminal region, which is absent in mERK5b and mERK5c, we examined the subcellular distribution of different ERK5 isoforms by cell fractionation and Western blotting of lysates from cells expressing epitope-tagged mERK5s. In the unstimulated cells, we observed that mouse mERK5a and human hERK5 were present in both the cytoplasm and the nucleus with the majority in the cytoplasm (Fig.  6). However, mERK5b and mERK5c were exclusively present in the nucleus. There was no detectable mERK5a, mERK5b, and mERK5c expression in the membrane and cytoskeleton fractions (data not shown). These results suggest that the Nterminal domain spanning aa 1-77 is important for cytoplasmic targeting of mERK5a.

The C-terminal Tail of ERK5 Is Not Required for Catalytic Activity but Is Essential for Activation of Transcription Factor
MEF2C-Because the protein sequence analysis of ERK5 uncovered several interesting domains, we further investigated the roles of these regions. ERK5 has a unique 400-amino acid long C-terminal tail whose function is not known. To determine whether the C-terminal domain of ERK5 plays a role in regulating ERK5 kinase activity, an Xpress-tagged ERK5 truncated at Gln-406, termed mERK5a(-tail), was generated. When mERK5a(-tail) and ERK2 amino acid sequences are aligned, the length of mERK5a(-tail) is comparable to that of ERK2. To test whether mERK5a(-tail) was catalytically active, an in vitro kinase assay using MBP as a substrate was performed. Using MEK5(D) to activate ERK5, mERK5a(-tail) was able to phosphorylate MBP in vitro similar to mERK5a (Fig. 7A, upper  panel). These results suggest that the kinase activity of mERK5a(-tail) is comparable to mERK5a.
Another indication of kinase activity is autophosphorylation. The anti-Xpress antibody recognized both the full-length mERK5A and the truncated mERK5a(-tail) in cell lysates from transfected CHO-K1 cells, suggesting that mERK5a(-tail) protein was stably expressed (Fig. 7A, lower panel). Upon activation by coexpression of MEK5(D), mERK5a(-tail) exhibited an electrophoretically shifted band similar to the full-length mERK5a (Fig. 7A, lower panel), suggesting that mERK5a(-tail) was phosphorylated by MEK5. The shifted band of the fulllength ERK5 is consistent with the phosphorylation of ERK5 and is generally thought to correspond to the activated form (14,15). Together, these data indicate that the C-terminal tail is not required for the full kinase activity of ERK5 stimulated with activated MEK5. This result is consistent with the previous observation that the C-terminal domain was not needed for kinase activity in vitro (9).
Finally the effects of mERK5a(-tail) on the activity of transcription factor MEF2C fusion protein were assessed by meas-  C, D, E). In the cases of testing the endogenous ERK5 (B, F), 10% serum was added and the cells were incubated for an additional 4 h before harvesting. Transfection efficiency was determined by cotransfection with a green fluorescence protein (GFP) expression vector (pEGFP-N1, from CLONTECH). The luciferase activities were normalized against cells transfected with pG5E1bLuc and GAL4 reporter plasmid alone, whose value was taken as 1. uring the luciferase activity from CHO-K1 cells cotransfected with a construct containing five copies of the GAL4-binding site upstream of a luciferase reporter gene. As expected, MEF2Cdependent reporter gene expression was enhanced dramatically when the full-length mERK5a and MEK5(D) were cotransfected into CHO-K1 cells (Fig. 7B). However, the mERK5a(-tail), which can act as an active kinase, did not stimulate MEF2C activity (Fig. 7B).
The C-terminal Tail Containing the NLS Plays a Role in Nuclear Translocation of ERK5-Like other MAPKs, ERK5 is localized in cytoplasm in the unstimulated state and translocates into the nucleus upon activation (14). A possible role for the C-terminal domain is to facilitate the translocation of activated ERK5 to the nucleus, because this domain is required for the transactivation of the transcription factor MEF2C and it has a possible nuclear localization sequence. To explore this possibility, immunocytochemistry was performed to test whether the C-terminal domain influences the nuclear translocation of mERK5a. In the absence of MEK5(D), cells expressing mERK5a or mERK5a(-tail) had predominantly cytoplasmic staining (Fig. 8, A and C). In the presence of MEK5(D), a significant amount of mERK5a localized in the nucleus (Fig.  8B). In contrast, mERK5a(-tail) was unable to translocate to the nucleus (Fig. 8D). In CHO cells, recombinant GFP was distributed in both the cytoplasm and the nucleus (Fig. 8E). However, GFP-tail, a fusion protein containing the GFP and the C-tail of mERK5a (aa 407-806), was exclusively localized in the nucleus (Fig. 8F), suggesting that the C-tail of mERK5a is able to drive GFP to the nucleus.
Searching the mouse ERK5 sequence against a currently available profile data base, a putative bipartite nuclear localization signal (NLS) (aa 505-539) was found in the C-tail. To determine whether the putative NLS is biologically important, GFP-NLS, a fusion protein containing GFP and the NLS domain of mERK5a (aa 505-539), as well as mERK5a(-NLS), an NLS-lacking mutant of mERK5a (deleting aa 505-539), were constructed. As expected, the NLS domain was able to drive GFP to the nucleus (data not shown). In contrast, mERK5a(-NLS), lacking the NLS domain, was localized in the cytoplasm under unstimulated conditions (Fig. 8G) but failed to move to the nucleus efficiently upon activation by MEK5(D) (Fig. 8H), indicating that the NLS domain is biologically active (Fig. 8,  compare B and H). Taken together, these findings suggest that FIG. 7. The effect of the C-terminal domain on ERK5 activity. A, kinase activity. CHO-K1 cells were transfected with the indicated Xpress-tagged ERK5 constructs and cotransfected with or without MEK5(D). Growth-arrested cells were harvested and lysed. ERK5 activity was analyzed using MBP as a substrate in an immune complex kinase assay (upper panel). The expression levels of recombinant mERK5 isoforms were assessed by Western blot analysis with anti-Xpress antibody (lower panel). B, activation of MEF2C. CHO-K1 cells were cotransfected with the indicated ERK5 constructs and the reporter plasmid pG5E1bLuc along with GAL4 fusion expression vectors containing MEF2C in the presence or absence of MEK5(D). After 24 h, the medium was changed to serum-free DMEM/F-12 for an additional 24 h. Then cells were collected for luciferase assay. Transfection efficiency was determined by cotransfection with a green fluorescence protein (GFP) expression vector (pEGFP-N1, from CLONTECH). The luciferase activities were normalized against cells transfected with pG5E1bLuc and GAL4 reporter plasmid alone, whose value was taken as 1. Immunofluorescence staining was performed with the anti-Xpress antibody as a first antibody and a fluorescein-conjugated anti-mouse IgG as a secondary antibody. CHO-K1 cells were transfected with pEGFP-C3 purchased from CLONTECH (E) or GFP-tail containing the C-terminal tail of mERK5a fused to the C terminus of EGFP (F). The images were analyzed and captured with an Olympus Fluoview confocal microscope.
translocation into nucleus via the NLS in the C-terminal tail is essential for biological activity of ERK5 in vivo.
The N-terminal Domain of ERK5 Is Responsible for Its Oligomerization and Association with MEK5-It has been reported that ERK1/2 oligomerizes upon phosphorylation. To determine whether ERK5 oligomerizes, we coexpressed Xpresstagged mouse ERK5a (Xp-mERK5a) and FLAG-tagged mouse ERK5a (FLAG-mERK5a) in CHO-K1 cells and performed coimmunoprecipitation assays. Immunoprecipitation of Xp-mERK5a with the Xpress antibody brought down FLAG-mERK5a (Fig. 9A) and vice versa (data not shown), suggesting that ERK5a forms oligomers in the cells. Oligomerization of ERK5 was observed in cell lysates from both activated and control cells, because coimmunoprecipitation was found in cells with or without expression of MEK5D, different from reported data for ERK1/2 (20). The observation, that the wild type FLAG-mERK5a and truncated Xp-mERK5a(-tail) could also be coimmunoprecipitated (Fig. 9B) but tail-GFP could not be coimmunoprecipitated (data not shown), suggests that the Nterminal domain but not the C-terminal domain of ERK5 is involved in the oligomerization. Furthermore, the N-terminal truncated isoforms, mERK5b and mERK5c, were able to be coimmunoprecipitated with wild type mERK5a (Fig. 9B), suggesting that the region aa 140 -406 but not the region aa 1-139 of the N terminus of ERK5 is important for the oligomerization.
The ability of the N-terminal or C-terminal truncated ERK5 mutants to bind to MEK5 was also examined by coimmunoprecipitation. The observation that mERK5a(-tail) and mERK5b but not mERK5c were able to bind to MEK5 (Fig. 10) indicates that the region aa 78 -139 in the N-terminal domain is important for the association of ERK5 with MEK5. DISCUSSION In this report we have identified three differentially spliced mouse ERK5 cDNAs (mERK5a, mERK5b, and mERK5c), which appear to play unique functional roles in regulating ERK5 and MEF2C activity. mERK5b and mERK5c function as dominant negative kinases blocking mERK5a activity and ERK5-mediated MEF2C activation. In addition, we have investigated the functional roles of several regions of mouse ERK5, which appear to be important for protein-protein interactions and intracellular localization. Specifically, we found that the N-terminal domain aa 1-77 is important for the cytoplasmic targeting; domain aa 78 -139 is required for association with the upstream kinase MEK5; and domain aa 140 -406 is necessary for oligomerization. The C-terminal tail, which contains a putative NLS, was found to be required for nuclear translocation of mERK5a upon activation (Fig. 11).
We identified three mouse ERK5 cDNAs by homology analysis, mERK5a (a mouse homologue of human ERK5), and two truncated mouse ERK5 isoforms termed mERK5b and mERK5c. Analysis of mouse genomic DNA sequences adjacent to the splicing junctions suggests that the three mERK5 cDNAs are generated by alternative splicing using different splicing donors and acceptors from a single gene encoding mouse ERK5. Several different ERK5 transcripts have been shown in human tissue (8), but these were generated by alternative splicing occurring at the 5Ј-noncoding region. The sequence of mERK5a is the same as that of the mouse ERK5 reported previously by Kamakura et al. (21) except for six scattered amino acid mismatches within the entire open reading frame. Possible explanations are a sequencing error or a DNA polymorphism causing the sequence variation.
The most abundant cDNA was mERK5a, which shares 91% identity with human ERK5. Compared with mERK5a, mERK5b and mERK5c lacked 69 and 139 amino acids at their N terminus, respectively. It appears likely that mERK5b and mERK5c perform functions different from mERK5a, because these proteins are unable to bind ATP and therefore are not active kinases. Our observations, that mERK5b and mERK5c lack kinase activity, inhibit mERK5a kinase activity, and inhibit mERK5a-mediated MEF2C transactivation, suggest that mERK5b and mERK5c may act as endogenous dominant negative kinases if mERK5b and mERK5c are expressed endogenously to a significant extent under some conditions. Immunoblotting results indicate that mERK5a is expressed to a greater extent than mERK5c and much greater than mERK5b. It is possible that the proteins encoded by mERK5b and mERK5c mRNAs could be expressed to different extents under some conditions or selectively expressed in some cell types. It is also possible that the mRNAs of mERK5b and mERK5c are not efficiently translated to the protein products in vivo as predicted, because the ribosome my initiate and terminate the translation early by the stop codon in the first insert. Thus, the expression of alternatively spliced mERK5b and mERK5c mRNAs may provide a mechanism to regulate ERK5 protein expression by preventing the translation of ERK5 in some cells at specific developmental stages or pathological conditions. Future studies are necessary to clarify the biological signifi- cance of mERK5b and mERK5c mRNAs in the regulation of mERK5a function.
Several other kinases exhibit independent expression of noncatalytic domains resulting from alternative splicing, which function as endogenous dominant negative inhibitors. For example, the primary transcript of a calmodulin-dependent protein kinase is alternatively spliced to generate mRNAs encoding either the full-length kinase or the calmodulin binding domain alone (22). Focal adhesion-associated protein tyrosine kinase (FAK) has an independent, C-terminal, noncatalytic domain (FRNK, FAK-related nonkinase) (23). Finally, the C terminus of the smooth muscle myosin light chain kinase is also expressed as an independent protein, telokin (24). Thus, the truncated, catalytically inactive forms of the mouse ERK5, mERK5b, and mERK5c may function as endogenous dominant negative inhibitors if they are expressed endogenously to a significant extent under some conditions.
Searching the mouse ERK5 sequence against a currently available profile data base to identify known functional regions, a proline-rich region and a NLS were found in the Cterminal domain. Proline-rich regions exist widely in both prokaryotes and eukaryotes. Studies have shown that proline-rich regions may act as Src-homology 3-binding motifs. These regions can directly interact with other proteins containing Srchomology 3 domains to regulate cellular localization and/or modulate enzymatic properties (1,2). A proline-rich sequence unique to MEK1 and MEK2 is required for Raf binding and MEK function (25). The function of proline-rich regions remains largely unclear. Interestingly, a significant difference between mouse ERK5a and human ERK5 sequences is present in the proline-rich region. Thus, the cloned mouse ERK5a, which differs significantly only in the proline-rich region compared with the human ERK5, may be a useful gene to determine the function of proline-rich regions. Differences between human and mouse with regard to intracellular signal transduction by highly related proteins have been reported (26,27). For example, the BAS-like Fas-associated phosphatase-1 interacts with the human Fas receptor, but not with the mouse Fas receptor. The C terminus of the Fas receptor, which is required for this interaction, is not conserved between mouse and human. Elucidation of the role of the mouse ERK5 proline-rich region may also reveal differences in signaling mechanisms between human and mouse.
The putative bipartite NLS in ERK5 located in the C-terminal tail is very likely to be important for nuclear translocation of ERK5. A bipartite NLS was described initially in the nucleosome assembly factor nucleoplasmin, which consists of two basic amino acids, a spacer region of any 10 -12 amino acids, and a basic cluster in which at least three out of the next five amino acids must be basic (28). Although the bipartite motif is a considerably more reliable indicator of nuclear localization, because less than 5% of non-nuclear proteins have a sequence that fits this motif (29), it is important to demonstrate that the putative NLS is necessary for nuclear targeting of the parent protein and sufficient to direct a non-nuclear protein to the nucleus. Our observations, that the NLS of mERK5 is required for the nuclear targeting of mERK5 upon activation and that this NLS itself is sufficient to drive GFP to the nucleus, indicate that the mERK5 NLS is biologically functional.
The mechanisms for the nuclear translocation and cytoplasmic anchoring of ERK1/2 must be different from ERK5. It has been suggested that ERK1/2 could cross the nuclear envelope by passive diffusion if it does not have either a NLS or a nuclear export signal. The cytoplasm retention of ERK1/2 in unstimulated cells likely involves specific association with MEK1/2, and nuclear translocation of ERK1/2 upon stimulation is accompanied by dissociation from MEK1/2 (30). A nuclear export signal in the N terminus of MEK1/2 has also been identified (31). These data indicate that MEK1/2 is a cytoplasmic anchoring protein for ERK1/2. The molecular size of ERK5 is beyond the limit for passive diffusion through the nuclear envelope pore. The nuclear targeting of ERK5 seems to be mediated by the NLS. The observation that mERK5b or mERK5c was present in the nuclear portion suggest that the N-terminal domain aa 1-77 is responsible for the cytoplasmic localization. The role for the N-terminal domain aa 1-77 in cytoplasmic localization of mERK5 is not clear. It is possible that the N-terminal domain aa 1-77 associates with other cytoplasmic components other than MEK5.
ERK5, unlike ERK1/2, exists as oligomer in unstimulated cells. ERK1/2 oligomerizes in a phosphorylation-dependent manner (32). In addition, ERK1/2 dimers are composed of either two phosphorylated molecules or one phosphorylated and one unphosphorylated molecule. In contrast, ERK5 expressed from transfected plasmids was able to form oligomers in both stimulated and unstimulated conditions, suggesting that the oligomerization of ERK5 is not dependent on its phosphorylation status. We found that the region aa 140 -406 in ERK5 is important for oligomerization, homologous to the dimer interface of ERK1/2, which is localized to amino acids 170 -359 (32).
Among the MAPKs, ERK1/2, ERK5, and ERK7 share the same signature TEY activation motif. However, these three MAPKs are completely different in terms of their structure, activation, and mechanisms of regulation, mediated in part by the fact that ERK5 and ERK7 have unique C-terminal tails compared with ERK1/2. Activation by TEY phosphorylation of ERK2 leads to nuclear import even though ERK1/2 does not have any classic nuclear localization signal. It has been suggested that interaction of ERK1/2 with upstream kinases may control nucleocytoplasmic transport (30). ERK7 is a constitutively activated and permanently nuclear-localized enzyme. Its activation and kinase activity seem to be dependent on nuclear targeting, which requires the C-terminal tail (11). In the case of ERK5, nuclear translocation occurs upon activation. However, its activation and kinetic activity are not dependent on its C-terminal tail.
In summary, we have identified three differentially spliced mouse ERK5 cDNAs whose unique structures result in different roles in the regulation of ERK5 and MEF2C activity. Using these mouse ERK5 splicing variants and other constructed mutants, we have also located regions of ERK5 that are responsible for cytoplasmic targeting, nuclear translocation, oligomerization, and MEK5 binding. Further studies are required to characterize in detail the precise role of these three isoforms in the specific function of ERK5.