Isolation of MEK5 and differential expression of alternatively spliced forms.

The prototype mitogen-activated protein (MAP) kinase module is a three-kinase cascade consisting of the MAP kinase, extracellular signal-regulated protein kinase (ERK) 1 or ERK2, the MAP/ERK kinase (MEK) MEK1 or MEK2, and the MEK kinase, Raf-1 or B-Raf. This and other MAP kinase modules are thought to be critical signal transducers in major cellular events including proliferation, differentiation, and stress responses. To identify novel mammalian MAP kinase modules, polymerase chain reaction was used to isolate a new MEK family member, MEK5, from the rat. MEK5 is more closely related to MEK1 and MEK2 than to the other known mammalian MEKs, MKK3 and MKK4. MEK5 is thought to lie in an uncharacterized MAP kinase pathway, because MEK5 does not phosphorylate the ERK/MAP kinase family members ERK1, ERK2, ERK3, JNK/SAPK, or p38/HOG1, nor will Raf-1, c-Mos, or MEKK1 highly phosphorylate it. Alternative splicing results in a 50-kDa α and a 40-kDa β isoform of MEK5. MEK5β is ubiquitously distributed and primarily cytosolic. MEK5α is expressed most highly in liver and brain and is particulate. The 23 amino acids encoded by the 5′ exon in the larger α isoform are similar to a sequence found in certain proteins believed to associate with the actin cytoskeleton; this alternatively spliced modular domain may lead to the differential subcellular localization of MEK5α.

The prototype mitogen-activated protein (MAP) kinase module is a three-kinase cascade consisting of the MAP kinase, extracellular signal-regulated protein kinase (ERK) 1 or ERK2, the MAP/ERK kinase (MEK) MEK1 or MEK2, and the MEK kinase, Raf-1 or B-Raf. This and other MAP kinase modules are thought to be critical signal transducers in major cellular events including proliferation, differentiation, and stress responses. To identify novel mammalian MAP kinase modules, polymerase chain reaction was used to isolate a new MEK family member, MEK5, from the rat. MEK5 is more closely related to MEK1 and MEK2 than to the other known mammalian MEKs, MKK3 and MKK4. MEK5 is thought to lie in an uncharacterized MAP kinase pathway, because MEK5 does not phosphorylate the ERK/MAP kinase family members ERK1, ERK2, ERK3, JNK/SAPK, or p38/HOG1, nor will Raf-1, c-Mos, or MEKK1 highly phosphorylate it. Alternative splicing results in a 50-kDa ␣ and a 40-kDa ␤ isoform of MEK5. MEK5␤ is ubiquitously distributed and primarily cytosolic. MEK5␣ is expressed most highly in liver and brain and is particulate. The 23 amino acids encoded by the 5 exon in the larger ␣ isoform are similar to a sequence found in certain proteins believed to associate with the actin cytoskeleton; this alternatively spliced modular domain may lead to the differential subcellular localization of MEK5␣.
A common element in many eukaryotic regulatory pathways is a three-kinase cascade, known as a MAP 1 kinase module. A module consists of three protein kinases that act sequentially within a pathway: a MAP kinase kinase kinase or MEKK (a MEK activator), a MAP kinase/ERK kinase or MEK (a MAP kinase activator), and a MAP kinase or ERK (extracellular signal-regulated protein kinase) homolog. First recognized in yeast (1), several MAP kinase modules have now been identified in mammalian systems (2,3). This kind of three-kinase regulatory cascade conveys information to target effectors, co-ordinates incoming information from parallel signaling pathways, confers a vast potential for amplification and specificity, and incorporates multiple inactivation mechanisms. The first and best studied is the MAP kinase pathway, made up of Raf-1 or B-Raf, MEK1 or MEK2, and ERK1 or ERK2 (4).
The two dual specificity MAP kinase kinases, MEK1 and MEK2 (19 -24), are the only known enzymes capable of phosphorylating and activating ERK1 and ERK2. Several laboratories have recently uncovered additional MEKs, for which some substrates have been defined. A mammalian homolog of a MEK first identified in Xenopus (25) is called MAP kinase kinase 4 (MKK4) (26), SAPK/ERK kinase (SEK) (27) or JNK kinase (JNKK) (28) and activates JNK/SAPK and p38/HOG1, but not ERK1 or ERK2. Yet another newly cloned MEK, MKK3, selectively activates p38/HOG1 in transfected cells (26). Thus, the second mammalian MAP kinase module to be defined apparently consists of MEKK1 (the MEKK) (29 -31), MKK4 (the MEK), and SAPK/JNK (the ERK), and a third module contains MKK3 and p38/HOG1. It is likely that the MEK in each module will confer much specificity in signaling, as evidence to date suggests that the MEKK and ERK components are more promiscuous enzymes (3).
MAP kinase modules are thought to be critical participants in major cellular events including proliferation, differentiation, and stress responses (4,32). Thus, an important goal is to identify and characterize other parallel cascades. Because of their enzymatic specificity, we chose to search for additional members of the MEK family. A PCR-based approach yielded a new MEK family member, MEK5, which by sequence is slightly more related to MEK1, MEK2, and the fission yeast MEK homolog Byr1 than to MKK3 or MKK4. None of the known ERK/MAP kinase family members are substrates for MEK5 nor will Raf-1, c-Mos, or MEKK1 highly phosphorylate it, suggesting that MEK5 is part of a novel MAP kinase module. Consistent with this biochemical data, MEK5 failed to complement deletions of MEKs in four yeast MAP kinase pathways. MEK5 exists in at least two spliced forms; the smaller ␤ isoform is ubiquitously distributed and is primarily cytosolic, while the larger ␣ isoform is expressed most highly in liver and brain and is primarily particulate. The alternatively spliced * This work was supported by Grant I1243 from the Welch Foundation, a grant from the Tobacco Research Council, and Grant DK34128 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U37462, U37463, and U37464. 1 The abbreviations used are: MAP kinase, mitogen-activated protein kinase; ERK, extracellular signal-regulated protein kinase; MEK (also called MAP kinase kinase or MKK), MAP kinase/ERK kinase; MEKK, MEK kinase; JNK/SAPK, Jun N-terminal kinase/stress-activated protein kinase; GST, glutathione S-transferase; MBP, myelin basic protein; RACE, rapid amplification of cDNA ends; PCR, polymerase chain reaction; bp, base pair(s); kb, kilobase(s). exon in MEK5␣ contains a sequence conserved in certain proteins thought to interact with the actin cytoskeleton and is likely to account for its differential localization. This is the first report of differential intracellular targeting of proteins in MAP kinase modules directed by alternative splicing.

RNA Isolation and First
Strand cDNA Synthesis-Frozen adult rat tissues were homogenized in 4 M guanidine thiocyanate (33). Total RNA was isolated by centrifugation of the homogenate over a cushion of 5.7 M CsCl (34). Poly(A ϩ )-enriched RNA and poly(A Ϫ ) RNA were prepared by affinity chromatography using oligo(dT)-cellulose (35). First strand cDNA was synthesized using the Invitrogen cDNA Cycle Kit (San Diego, CA).
Isolation of cDNA Clones-Nested degenerate primers were utilized in sequential PCR reactions. 5Ј primers corresponding to MEK subdomains IV and V (YIVGFYG and MEHMDGG) and 3Ј primers corresponding to MEK subdomains VI and VIII (MHRDVKP and YMSPER) were used to perform the PCR on first strand cDNA synthesized from poly(A ϩ ) RNA isolated from multiple rat tissues. The PCR products were cloned into pCRII using the Invitrogen TA cloning kit.
One 150-bp PCR product was used to screen an oligo dT-primed rat brain cDNA library (Stratagene). Two hybridizing clones were isolated, neither of which contained a full-length cDNA. To obtain a full-length clone, one of the partial clones was used to screen a random-primed rat forebrain cDNA library (kindly provided by J. Boulter, Salk Institute) with 32 P-labeled, randomly primed probes. Sequencing of six isolates was performed using the dideoxynucleotide chain termination method (36) and an Applied Biosystems automated sequencer.
To confirm the 3Ј end of the clone, 3Ј nested RACE was performed using oligo(dT)-primed first strand cDNA from rat brain and testes as template. MEK5 5Ј primers were CAGGATTTCGATTTGTACAG, GTG-AAGCCTTCCAACATGC, and GAGGTTCTCGGAGCCGTTTG; 3Ј primers were GACTCGAGTCGACATCGAT 17 and GACTCGAGTCGACATC-GAT.
Northern Analysis-Poly(A ϩ ) RNAs were denatured in glyoxal and dimethyl sulfoxide, size-fractionated on 0.8% agarose gels, and transferred to Biotrans Plus membrane (ICN, Costa Mesa, CA). The membrane was hybridized as described (35).
Expression and Purification of Recombinant GST-MEK5-MEK5␤ was subcloned into pGEX-KG using Vent polymerase (New England Biolabs, Beverly, MA) and PCR primers that added NcoI and HindIII restriction sites. The antisense primer also added 13 amino acids and a stop codon, because these were missing from the clone used as template. To construct MEK5␤ containing the subdomain IX-X exon, a StuI-HindIII fragment from a clone containing the exon was subcloned into the MEK5␤ pGEX-KG construct. The MEK5␣ construct was made by subcloning an NcoI-BglII fragment into the MEK5␤ pGEX-KG construct. Constructs were sequenced to confirm that no changes in coding sequence had been introduced. Expression of GST-MEK5 was induced for 4 h with 40 M isopropyl-1-thio-␤-D-galactopyranoside. GST-MEK5 was purified on glutathione-agarose using standard protocols (37) and then on Mono Q to remove the remaining impurities.

RESULTS
Isolation of cDNA Clones Encoding MEK5-A PCR product derived from rat brain and kidney first strand cDNA was used to screen an oligo(dT)-primed cDNA library from adult rat brain. Two partial clones representing a novel MEK family member were isolated. One, missing the C terminus, and the other, missing the N terminus, were used to screen a randomprimed rat cDNA library; multiple independent clones were isolated, none of which included a 3Ј stop codon. To confirm the sequence of the 3Ј end of the clone, seven clones derived from 3Ј RACE were sequenced and encoded the same C terminus as the clone from the oligo(dT)-primed library. This complete cDNA is designated MEK5␤ (Fig. 1, A and B). The existence of a 3Ј alternatively spliced, 10-amino acid exon corresponding to the region between kinase subdomains IX and X (Fig. 1A) was confirmed within multiple RACE products.
A second alternative splicing event was discovered when two clones were isolated from the random-primed library that contained a second alternatively spliced exon of 68 bp near the 5Ј 2 M. Cheng, T. Boulton, and M. Cobb, manuscript in preparation. end of MEK5. The inclusion of this alternatively spliced 5Ј exon generates an initiation site 5Ј to the spliced exon and to the initiation site in MEK5␤. Initiation at the more 5Ј site produces a predicted protein, MEK5␣ (Fig. 1A), that is 89 amino acids longer at the N terminus than MEK5␤ (Fig. 1B). This 68-bp exon (Fig. 1A) causes an upstream start codon to be in frame with the kinase domain and contains no intervening in-frame stop codons. Thus, the inclusion of an alternatively spliced 5Ј exon generates a distinct initiation site in MEK5␣. Alternative splicing of the 5Ј and 3Ј exons may generate four forms of MEK5. We have isolated MEK5␣ clones both with and without the subdomain IX-X exon. It is not clear whether MEK5␤ exists in both forms.
Northern analysis suggests that MEK5 is expressed in all the tissues examined and that alternatively spliced variants exist. A major band of 2.5 kb was detected in brain and testes, and a slightly larger band of 2.8 kb was noted in liver and spleen (Fig. 1E). In testes an additional weakly hybridizing band of 3.8 kb was detected. Immunoblotting analysis indicates tissue selective expression of MEK5 isoforms (see below).
Comparison to MEK Subfamily-MEK5 is 48% identical to MEK1 and 45% identical to Byr1, a yeast MEK. A family tree (Fig. 1D) indicates that MEK5 is more closely related to MEK1, MEK2, and Byr1 than to MKK3 and MKK4. Interestingly, the putative, alternatively spliced 3Ј exon (Fig. 1A) corresponds to a region, subdomains IX-X, expected to be involved in kinasesubstrate interactions and thus may affect substrate specificity. The comparable region of JNK/SAPK is also encoded by an alternatively spliced exon (12). The presence of sequence encoded by this exon increases the affinity of JNK/SAPK for c-Jun (40).
Bacterial Expression and Protein Kinase Activity of MEK5 and Mutants in Vitro-GST fusion proteins were expressed that contained three of the four possible MEK5 sequences, MEK5␣ and MEK␤ lacking the subdomain IX-X exon and MEK5␤ with this exon. The fusion proteins displayed roughly equal but low protein kinase activity with MBP as substrate (Fig. 2). To test whether MEK5 might participate in any of the known mammalian MAP kinase modules, recombinant ERK1, ERK2, ERK3, JNK/SAPK, and p38/HOG1 were examined as substrates. GST-MEK5␤ did not phosphorylate kinase-deficient mutants of ERK1, ERK2, ERK3, or JNK/SAPK␤, or wild type p38/HOG1 under conditions in which ERK2 was readily phosphorylated by unactivated recombinant MEK1 (not shown).
Because of the low activity of MEK5 and lack of information about its upstream activators, efforts were made to activate MEK5 by changing its putative phosphorylation sites to acidic residues. Comparable mutations resulted in biochemical and functional activation of MEK1 (41). Thus, aspartic acid was introduced into each of the phosphorylation sites singly and together (double mutants) in MEK5␤ (Fig. 2). The activity of these mutant proteins assayed with MBP was no greater than that of the wild type proteins. A similar strategy also worked poorly for MKK4. 3 Phosphorylation of MEK5 in Vitro-The other cloned MEKs are substrates for one or more of the three protein kinases, Raf-1, MEKK1, and c-Mos (39,(42)(43)(44). Thus, we tested their ability to phosphorylate MEK5 in vitro. Immunoprecipitates from cells overexpressing either Raf-BXB (45) or c-Mos (44) had little activity toward MEK5 compared to activity immunoprecipitated from control cells under conditions in which MEK1 was phosphorylated (not shown). A constitutively active, catalytic domain fragment of MEKK1 (39) also failed to catalyze significant incorporation of phosphate into MEK5 (not shown). Extracts of PC12 cells treated with nerve growth factor, bradykinin, and phorbol ester caused a slight but reproducible increase in recombinant GST-MEK5 phosphorylation (not shown).
Analysis of MEK5 in Yeast-Four MAP kinase modules have been characterized to date in the budding yeast, Saccharomyces cerevisiae (18): 1) a mating pheromone-sensitive module composed of Ste11 (MEKK), Ste7 (MEK), and Fus3 and Kss1 (ERKs); 2) a module required for proper cell wall construction, composed of Bck1 (MEKK), Mkk1 and Mkk2 (MEKs), and Mpk1 (ERK); 3) an osmotic stress-response module, composed of SSK2 and SSK22 (MEKKs), Pbs2 (MEK), and Hog1 (ERK); and 4) a pathway required for sporulation, which is not well characterized, but contains a MAPK homolog, Smk1. MEK5 was expressed from the strong S. cerevisiae ADH1 promoter in ste7, mkk1/mkk2, and pbs2 null mutants (46 -49). MEK5 failed to suppress loss of MEK function in each of these mutants. We conclude that MEK5 is not closely related in function to any of these yeast MEKs.
Only a single MAP kinase module has been well characterized in the fission yeast, Schistosaccharomyces pombe (18). This module, composed of the protein kinases Byr2 (MEKK), Byr1 (MEK), and Spk1 (ERK) is required for mating pheromone responses (1). MEK5 was expressed from the strong S. pombe adh1 promoter in mutants carrying null mutations of the byr1 gene (50). MEK5 failed to suppress the mating defect of a byr1 Ϫ S. pombe haploid strain and the sporulation defect of a byr1 Ϫ /byr1 Ϫ diploid mutant.
Detection of Two Forms of Endogenous MEK5 by Immunoblotting-A peptide from the N terminus of MEK5␣ not contained in MEK5␤ and a peptide common to the ␣ and ␤ isoforms (Fig. 1) were used as antigens to generate antibodies to MEK5. The expected specificity of the resulting antibodies was confirmed by probing blots containing recombinant GST-MEK5␣ and GST-MEK5␤ with the antisera; antiserum N797 raised against the unique ␣ sequence detected only recombinant GST-MEK5␣ and antiserum L610 detected both GST-MEK5␣ and 3 M. Karin, personal communication.

FIG. 2. Protein kinase activity of MEK5␤ and mutants.
Aspartic acid was introduced in place of the putative phosphorylation sites, Ser-222 and Thr-226, in MEK5␤ by PCR with Vent polymerase and oligomers that spanned these amino acids. The 3Ј primer corresponding to the C terminus encoded a stop codon as well as a HindIII site. Each resulting PCR product was subcloned into the PvuII and HindIII sites of the MEK5 pGEX-KG construct lacking the subdomain IX-X exon, and then a StuI-BamHI fragment from this form was subcloned into the construct containing the IX-X exon. 30 g/ml wild type GST-MEK5␤ (with and without the subdomain IX-X exon) and the indicated mutants of it were incubated either with or without 10 g/ml MBP under phosphorylating conditions. Recombinant wild type His 6 -MEK1 was incubated with and without MBP as a control. Autoradiogram is shown.
GST-MEK5␤ (Fig. 3A). Using L610 two specific bands of 50 and 40 kDa in size, consistent with the predicted sizes of MEK5␣ (50,197) and ␤ (40,091), were detected in rat tissues (Fig. 3B). Antibody N797 also recognized the 50-kDa protein in liver (Fig.  3C), confirming that it is MEK5␣. MEK5␤ was found in all tissues examined with the possible exception of muscle, where a nonspecifically reacting protein slightly larger than MEK5␤ obscured the MEK5 band. Its abundance was highest in liver. MEK5␣ was found in liver and brain but was not detected in other tissues examined. MEK5␣ was immunoblotted in whole cell lysates but not in cytosolic fractions, whereas MEK5␤ was in both (Fig. 3, B and C). This suggests that the N terminus of MEK5␣ causes its association with particulate fractions. Of cell lines examined, MEK5 is most highly expressed in PC12 cells, and is also present in HL60, Rat1, and Chinese hamster ovary cells, but has not been detected in 293, Cos-1, and CV-1 cells (data not shown). DISCUSSION We have cloned multiple forms of a novel MEK family member, MEK5. The smaller ␤ isoform is a ubiquitous cytosolic protein, while the larger ␣ isoform is particulate and has a more limited expression in tissues and cells. The MEK5 catalytic domain has features unique to the MEK protein kinase subfamily. The putative phosphorylation sites (SIAKT) are in conserved positions in the phosphorylation lip between subdomains VII and VIII and the lip is of the same length as most other members of the family. MEK5 also displays the MEK consensus sequence, CXXK, near the C terminus of the catalytic domain, not present in other protein kinases.
A canonical MAP kinase module consists of three protein kinases that act sequentially within one pathway: a MEKK, a MEK, and an ERK/MAP kinase. The exquisite specificity of MEKs leads to the selective activation of a defined group of ERK-related enzymes. Pathway specificity is achieved in part by pairing of ERK-related enzymes with their MEK-like acti-vators. MEK5 apparently lies in an, as yet, undescribed MAP kinase module, as it neither phosphorylates ERK1, ERK2, ERK3, JNK/SAPK, or p38/HOG1 nor is phosphorylated by known MEKKs, Raf-1, c-Mos, or MEKK1.
Data base searches (NCBI BLAST network services) using the N-terminal sequence unique to MEK5␣ revealed a region of sequence identity to the yeast putative GTP/GDP exchange factors scd1 (51) and CDC24 (52), the human p40 phox protein (53), and the transmembrane tyrosine kinase, Ros (54). This region of identity exactly overlaps the MEK5␣ alternative exon, implying that this sequence encodes a modular domain utilized by a number of proteins to specify subcellular localization. Of proteins with sequence related to the exon, Scd1 and CDC24 are believed to encode guanine nucleotide exchange factors for the yeast Rac/Rho family member Cdc42 (55,56) in S. pombe and S. cerevisiae, respectively. The Rac/Rho family of small G proteins has been linked to control of the actin cytoskeleton in mammals (57,58). In yeast these exchange factors appear to mediate certain Ras-regulated functions, particularly cytoskeleton-dependent control of cell morphology. p40 phox is an essential component of the leukocyte NADPH oxidase system, which is directly regulated by Rac (53). The distinct localization of MEK5␣ that results from expression of the longer N terminus is consistent with a role for sequence encoded by this exon in cellular localization and subsequent signaling capabilities. Subcellular localization is recognized as an important, but poorly understood, contributor to signaling and enzymatic specificity (59,60). Most frequently, targeting of signaling molecules has been accomplished by either regulatory or dedicated targeting subunits. Well studied examples include the G subunit of phosphoprotein phosphatase I (61) and the AKAPs (A kinase anchor proteins), which serve to target cAMP-dependent protein kinase (62). Recently, distinct isoforms of phosphoprotein phosphatases, including the tyrosine phosphatase PTP1 (60, 63), expressed as a consequence of alternative splic- . Upper panels, probed with a 1:500 dilution of L610 antiserum; lower panels, probed with a 1:500 dilution of L610 antiserum preincubated with 50 g/ml of its antigenic peptide. Immunoreactive bands specifically blocked by preincubation with the peptide are indicated by the arrows and correspond to MEK5␣ and ␤. Molecular size standards are indicated between the blots. E. fat, epididymal fat; CHO, Chinese hamster ovary cells. C, MEK5␣ is not cytosolic. Blots of whole cell lysates (A) and supernatants (B) from rat liver were probed with preimmune (left) or immune (right) N797 antiserum.
ing, have been found localized to different subcellular compartments, including the cytoskeleton and the nucleus. Their concentration at these specific sites is a significant determinant of their substrate availability. A number of protein kinases are thought to exist in multiple alternatively spliced forms. However, MEK5 provides the first reported instance of differential intracellular targeting of a kinase in any MAP kinase module as a result of alternative splicing. Within MAP kinase modules, the current dogma is that the MAP kinase itself translocates from cytoplasm to nucleus under the control of the activating signal by an unknown mechanism. The MEK component, on the other hand, has been assigned a static role in the cytoplasm. Perhaps, in addition to its already highly selective enzymatic role, a MEK acts to direct and tether its respective MAP kinase module to various sites of action within the cytoplasm. In the case of MEK5 subcellular targeting also occurs in a tissue-specific manner. Targeting of the MEK might restrict substrates available to its otherwise promiscuous downstream ERK. Thus, alternative splicing may provide another mechanism for targeting signaling molecules to generate specificity in protein kinase pathways.