Characterization of the structure and function of a novel MAP kinase kinase (MKK6).

Mitogen-activated protein (MAP) kinases require dual phosphorylation on threonine and tyrosine residues in order to gain enzymatic activity. This activation is carried out by a family of enzymes known as MAP kinase kinases (MKKs or MEKs). It appears that there are at least four subgroups in this family; MEK1/MEK2 subgroup that activates ERK1/ERK2, MEK5 that activates ERK5/BMK1, MKK3 that activates p38, and MKK4 that activates p38 and Jun kinase. Here we describe the characteristics of a new MKK termed MKK6. The clones we isolated encode two splice isoforms of human MKK6 comprised of 278 and 334 amino acids, respectively, and one murine MKK6 with 237 amino acids. Sequence information derived from cDNA cloning indicated that MKK6 is most closely related to MKK3. The functional data revealed from co-transfection assays suggests that MKK6, like MKK3, selectively phosphorylates p38. Unlike the previously described MKKs (or MEKs), MKK6 exists in a variety of alternatively spliced isoforms with distinct patterns of tissue expression. This suggests novel mechanisms regulating activation and/or function of various forms of MKK6.

The signal transduction pathways that utilize mitogen-activated protein (MAP) 1 kinases have an important role in a variety of cellular responses including growth factor-induced proliferation, gene expression, and compensation for alterations in the extracellular milleau induced by heat shock, UV light, increased extracellular osmolarity etc. (1)(2)(3)(4). Four MAP kinase pathways have been defined in yeast (5); these pathways are functionally independent and are regulated by distinct protein kinase cascades. Each pathway results in activation of a separate MAPK by a unique MAPK kinase (MKK or MEK). It is now clear that higher eukaryotes also have distinct MAP kinase pathways; such pathways are comprised of kinase cascades leading to the activation of discrete MKKs and MAPKs (5)(6)(7)(8)(9). Specifically four subgroups of MAPKs have been identified (5): the ERKs (extracellular signal-regulated kinase) (10,11), Jun kinases (c-Jun amino-terminal kinases (JNK) or stress-activated protein kinase (SAPK)) (12,13), p38 MAP kinase (14), and BMK1/Erk5 (15,16). The signal transduction pathway leading to p38 activation is related, in part, to a pathway in yeast leading to activation of a MAP kinase known as Hog1p. To date the activation of this yeast pathway has been shown to occur principally in response to increased extracellular osmolarity (17). Recently Saito and colleagues (18,19) have defined two distinct pathways leading to Hog1p activation in Saccharomyces cerevisiae.
In mammalian cells p38, the Hog1p homologue is activated by multiple stimuli acting through different receptors. For example Lee et al. (20) showed that p38 is involved in bacterial endotoxin (lipopolysaccharide)-induced cytokine production through the use of pharmacologic inhibitors that are specific for p38 (20). p38 is also activated by other bacterial components, proinflammatory cytokines, and physical-chemical changes in the extracellular milleau (21).
There have been several distinct MKKs/MEKs identified in mammalian cells; one type, termed MEK1/MEK2, does not phosphorylate or activate p38 or JNK while in contrast is a strong activator of ERK1/ERK2 (7,9,22). In contrast, two other MKKs known as MKK3 and MKK4 activate p38 but not ERK (6,7); MKK4 (also known as SEK1/JNKK1) also activates JNK/SAPK (6 -8). MKK3 and MKK4 are most closely related to PBS2p, the upstream activator of Hog1p (6,7). p38 and JNK are often activated in parallel (21) but independent activation of p38 also has been observed (23). Simultaneous activation of ERK and p38 also occurs when cells are exposed to a stimulus such as lipopolysaccharide or increased extracellular osmolarity (14,24). Unlike yeast MAP kinase pathways which appear to operate independently, cross-talk seems to occur between the mammalian MAP kinase pathways (6,7,25,26). Emerging data suggest that the MAPK signal transduction pathways in mammalian cells are much more complex than homologous systems in yeast. One level of such complexity might exist at the level of the MAPK kinases. Indeed multiple activators of p38 or JNK can be detected by fractionation of the fibroblasts activated by hyperosmolar media (27). Thus there is a need to identify all of the p38 upstream activators in order to fully understand the regulation of p38. To isolate additional MKKs which function as activators of p38, we have employed a polymerase chain reaction (PCR)-based strategy with degenerate oligonucleotides based on conserved kinase domains present in the known members of the MKK family. This approach resulted in the isolation of two human cDNAs and one murine cDNA encoding closely related proteins of the MKK family. These clones are the different splice forms of one gene which we have designated MKK6, which is selectively expressed in different tissues. Amino acid sequence comparison revealed that the proteins encoded by these cDNAs are most closely related to MKK3, and co-transfection studies show that MKK6, like MKK3, activates p38 but fails to activate either ERK or JNK. In this article we also present data showing that unlike MKK4, MKK3 and MKK6 are not regulated by Rac or Cdc42.

EXPERIMENTAL PROCEDURES
cDNA Cloning-Degenerate oligonucleotides (AARYTNTGYGAYT-TYGGNGT and ATNCKYTCNGGNGCCATRTANGG) were designed using published information from the conserved kinase subdomain of known MKKs (7,8,22); these oligonucleotides were employed as PCR primers to isolate fragments of MKK-related cDNAs from a human placenta cDNA library. Comparison of the sequence of 48 clones with the GenBank data base (Blast Fileserver, National Center of Biotechnology Information) allowed identification of one clone that is novel and exhibits a high level of homology with the MKK group. The PCR fragment was labeled with 32 P by random priming and used as a probe to screen a ZAPII human placenta library (Stratagene, La Jolla, CA) and a ZAPII murine jejunum library (provided by Dr. Z. Chen, The Scripps Research Institute, La Jolla, CA). Two positive clones were obtained from the human library after screening 2 ϫ 10 6 phage. DNA sequencing of both strands of each clone was performed by the Microchemistry Core Facility at The Scripps Research Institute using a model 373A automated sequencer (Applied Biosystems, Foster City, CA). The sequence data indicated these two clones are identical and both contain a complete open reading frame. One positive clone was isolated from the murine library after screening 5 ϫ 10 5 phage. The sequencing of this clone was done as noted above. The murine clone shows strong homology to the human clones except at the 5Ј end. Nucleotide sequence comparison of the 5Ј end sequence of the murine clone with the GenBank Data base reveals four sets of nucleotide sequences (GenBank access number H08016, H06765, T66783, and R15387) that have nearly identical sequence. These clones are human fetal brain cDNAs that were sequenced by the Washington University-Merck EST project. The clone R15387 was requested and completely sequenced as noted above.
Northern Blot Analysis-An RNA tissue blot was purchased from ClonTech (San Francisco, CA). The blot contains 2 g of poly(A) ϩ RNA isolated from different human tissues, fractionated by denaturing agarose gel and transferred onto a nylon membrane. The blot was hybridized to a probe that was prepared by labeling the 5Ј end fragment (Ϫ404 to Ϫ199 bp) of MKK6 with [␥-32 P]dATP by random priming. The blot was stripped by incubation for 2 min at 100°C in water and reprobed with a probe corresponding to the 5Ј end sequence (Ϫ453 to Ϫ159 bp) of MKK6b/6c. The blot was stripped again and reprobed with a probe containing the coding region and 3Ј-untranslated region (Ϫ87 to 1560) of MKK6. Hybridization was performed overnight at 50°C using 50% formamide, 5 ϫ SSPE (1 ϫ SSPE, 150 mM NaCl, 15 mM sodium citrate, pH 7.2), 5 ϫ Denhardt's, 1% SDS, and 200 g/ml single-stranded fish sperm. The blot was washed two times with 1 ϫ SSC (1 ϫ SSC, 150 mM NaCl, 10 mM NaH 2 PO 4 , 1 mM EDTA, pH 7.2), 0.1% SDS, and 1 mM EDTA and one time with 0.1 ϫ SSC, 0.1% SDS, 1 mM EDTA at 65°C prior to autoradiography.
cDNA Constructs-The A, E, or M mutants of MKK6 were created by the substitution of Ser 151 and Thr 155 with Ala or Glu and Lys 25 with Met using a PCR-based procedure as described (28). The MKK6 cDNA and the mutants were cloned into the expression vector pCDNA3 (Invitrogen, San Diego, CA) with the HA epitope tag added so that the protein contained an HA-tag in the amino-terminal region. The cloning sites used for these constructs are HindIII and KpnI. The HA-tag sequence is YPYDVPDYAGYPYDVPDYAGSYPYDVPDYAAA. The tag was added by a PCR recombination as described (28). The sequence of all constructs were confirmed by DNA sequencing in the The Scripps Research Institute Microchemistry Core Laboratory. p38 or JNK1 cDNA containing a FLAG-epitope tag were prepared using a pCDNA3 vector (29). The constructs of Rac1(Q61L), Cdc42(Q61L), RhoA(Q63L), Ras(Q61L), Raf * (22W), and Myc-tagged Erk1 have been described (29). The truncated form of MEKK1, here termed MEKK1⌬, is in pCMV5 (25). The HA-tagged MKK3, MKK4(SEK1/JNKK1) were prepared as described (6). The expression plasmid of GST-ATF2 was made by removing the 3Ј end part of the cDNA with XbaI. The cDNA was then blunted and religated into pGEX-2T vector (Pharmacia, Uppsala, Sweden). The recombinant GST-ATF2 protein contains 1-109 amino-terminal amino acids of ATF2.
Transient Expression of Various cDNAs-COS-7 cells were main-tained in Dulbecco's modified Eagle's medium supplemented with 5% fetal bovine serum. Cells on 35-mm plates were transiently transfected with 1 g (total) of plasmids DNA using lipofectamine reagent (Life Technologies, Inc., Grand Island, NY) according to the manufacturer's recommendations. After 48 h, the cells treated with or without UV radiation as described (13). The transfection efficiency was evaluated by co-transfection with plasmid pCMV␤ (ClonTech). The cell lysates were normalized according to ␤-galactosidase activity prior to immunoprecipitation. ␤-Galactosidase activity was measured as described (32). Immunoprecipitation and Western Blot Analysis-The cells were suspended in lysis buffer (25 mM HEPES (pH 7.5), 1% Triton X-100, 130 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM Na 3 VO 4 , and 1 mM phenylmethylsulfonyl fluoride) for 20 min with shaking at 4°C and the lysate cleared by centrifugation at 130,000 ϫ g for 10 min. Monoclonal antibodies against the flag-epitope, M2(IBI-Kodak), the HA-epitope, 12CA5 (kindly provided by Dr.Ian Wilson, TSRI), or the c-Myc epitope, 9E10 (Santa Cruz Biotechnology, Santa Cruz, CA), were prebound to protein G-or A-Sepharose beads, respectively. 20 l of 50% suspension of beads were added to 300 l of cell lysates and gently shaken for 3-10 h at 4°C. The beads were washed six times with 1 ml of washing buffer (25 mM HEPES (pH 7.5), 50 mM NaCl, 0.1 mM EDTA, 0.05% Triton X-100, 0.1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride). The immunoprecipitates were used in the kinase assay within 2 days. The immunoprecipitates were always analyzed by Western blotting to ensure the precipitates contained the targeted proteins (29). The monoclonal antibodies M2, 12CA5, or 9E10 were used to detect epitopetagged proteins.

Isolation and Characterization of New MKK cDNAs-
We designed a set of degenerate oligonucleotides utilizing known sequence information from the conserved kinase domains of previously described MKKs (7,8,22). These oligonucleotides were used as PCR primers to isolate fragments of MKK related cDNA from human placenta cDNA; 48 clones were sequenced. When the sequences were analyzed using the GenBank data base one clone was identified with a high degree of homology to other known MKKs but nonetheless encoding what appeared to be a protein not previously described. The PCR-derived fragment was used to screen a human placenta or a murine jejunum library. Screening of the human library yielded 2 positive clones after screening 2 ϫ 10 6 phage. Sequencing revealed that the two clones were identical and contained a complete open reading frame. Moreover it was ascertained that the sequence of these clones was distinct from all other known MKKs. We therefore termed the protein encoded by the cDNA as MKK6 (GenBank U39064). The nucleotide and deduced protein sequence of MKK6 is shown in Fig. 1A. A screen of a murine cDNA library for MKK6-related clones resulted in identification of a single clone with a high degree of sequence homology to human MKK6 except in the 5Ј-region. When the 5Ј-end sequence of this murine clone was used to search GenBank we found several clones with nearly identical sequences deposited by the Washington University-Merck EST project. We obtained one of these clones from the EST project, clone R15387. Sequencing revealed a cDNA encoding a protein with some regions of complete identity with MKK6 and other regions, primarily in the amino-terminal region with marked differences compared to MKK6. Thus we believe this form to be a splice variant of MKK6 and have termed it MKK6b (Fig. 1A). The protein sequence alignments (Fig. 1B) for MKK6, MKK6b, and other MKKs reveal similarities and differences in the members of the family of enzymes. In the overlapping regions MKK6 is about 80% identical to MKK3 and 40% to MKK4.
We believe that the cDNAs for MKK6 and MKK6b contain full-coding regions because both clones contain in-frame stop codons 5Ј upstream of the first ATG sequence. The first inframe ATG of MKK6 has been designated as position 1. Since it is the first ATG codon and the typical Kozak sequence, it should be the authentic start site for MKK6. The first ATG found in MKK6b is at position Ϫ168. Although not a typical Kozak sequence, translation of MKK6b mRNA most likely begins at this position. The murine clone (GenBank U93066, Fig.  1D) which is closely related to MKK6b has the latter ATG sequence but not an equivalent ATG at position 1. Thus we predict that MKK6 contains 278 amino acids while MKK6b contains 334 amino acid residues. Of interest are differences between MKK6 and MKK6b in DNA sequence before position Ϫ154; this may be due to differential splicing. A dendrogram created by progressive pairwise alignment comparison of the MKK family is shown in Fig. 1C.
When we compared the sequence of the murine cDNA clone we isolated with the sequence of human MKK6 or MKK6b we noted that the murine clone has a 62-base pair deletion covering the region from residues 16 to 77. Due to this deletion, the ATG at position 122 is most likely to be the starting codon. We predict that this murine clone contains 237 amino acids. Although at the DNA level, this murine clone is most closely related to human MKK6b, it has a completely different 5Ј sequence beginning with position Ϫ219 and a unique 3Ј sequence starting from position 1056 when compared with MKK6b. This may result from differential splicing, although additional studies are required to firmly establish this. On the basis of these differences we term this murine cDNA, murine MKK6c. The sequence of murine MKK6c is shown in Fig. 1D. Further investigation is required to determine whether a comparable form exists in human. Amino acid sequence comparison reveals that the murine MKK6c is 97.6% identical to human MKK6/6b in the overlapping region. Conservation of the amino acid sequence between species has been found in other MKKs and may indicate the importance of this family proteins (9,22).
Tissue Distribution of MKK6 -Because of the observed differences in nucleotide sequences in the 5Ј-region and 3Ј-region of the MKK6 cDNA clones we wondered if there is tissuespecific expression of alternatively spliced forms. To do this we made specific probes based upon unique sequences in the 5Јregion of MKK6 and MKK6b/6c and probed blots containing poly(A) ϩ RNA (Fig. 2). This experiment revealed a high level of expression of MKK6 mRNA with size of 1.7 kilobase pairs in skeletal muscle (Fig. 2A). The probe encompassing the specific 5Ј-end region of MKK6b detected multiple mRNA bands with the size of 1.8, 2.4, 4.5, and 11 kilobase pairs. The 11-kilobase pair band only was noted in skeletal muscle and pancreas. We suspect that the multiple bands represent additional differential splicing forms of MKK6. The MKK6 mRNAs containing the sequence we used in this probe are enriched in heart, skeletal muscle, pancreas, and liver and detectable in brain, placenta, and lung (Fig. 2B). These data suggest that tissue-specific splicing occurs and may play a role in the control of MKK6 expression. Northern blot analysis of tissues for expression of MKK6s with a probe encompassing the coding region which is most similar in all MKK6 clones is shown in Fig. 2C. No additional bands were noted with this probe; transcripts of MKK6 gene are most abundant in the skeletal muscle.
Enzymatic Activity of MAPK Kinases-We next investigated one aspect of the substrate specificity of MKK6 by asking if members of the MAP kinase family were activated by this enzyme. To do this we coexpressed MKK6 or MKK6b in COS-7 cells by transient transfection together with epitope-tagged forms of p38, JNK1, or ERK1. At the same time we compared the activities of MKK3 and MKK4 with MKK6. When specificities toward ERK1 were examined we used MBP as a substrate for ERK1 or p38 (Fig. 3A). In contrast, ATF2 was used as a substrate for JNK1 and p38 in a parallel set of studies (Fig.  3B). In this latter set of studies cotransfection with MKK6b was not performed (Fig. 3B). The enzyme activity of p38, JNK1, or ERK1 were determined by immunokinase assay after immunoprecipitation with the anti-epitope antibody. The overexpression of MKK6/6b in COS-7 cells caused increased activity of p38 (Fig. 3, A and B). In contrast, the co-expression of MKK6 does not enhance JNK1 (Fig. 3B) or ERK1 activity (Fig. 3A). The activation of p38 by MKK6 was similar to that of MKK3. The similar substrate specificity of MKK3 and MKK6 is consistent with the high homology between these two enzymes. In these studies we noticed that the migration of JNK1-phosphorylated GST-ATF2 is slightly slower than p38-phosphorylated GST-ATF2. This suggests differences in ATF2 phosphorylation by JNK1 and p38. Further studies are underway to verify this.
Sequence comparison of MKK6 with other MKKs suggested that Ser 151 and Thr 155 may be phosphorylation sites required for enzymatic activity. To investigate this we modified the MKK6 cDNA by replacing Ser 151 and Thr 155 with Ala (A mutant) or Glu (E mutant). An additional mutant was made by changing Lys 25 to Met to delete the ATP binding site (M mutant). Epitope-tagged versions of these mutants were made by adding an HA-tag to the amino-terminal of MKK6; the proteins encoded by these cDNAs were expressed in COS-7 cells. HAtagged proteins were immunoprecipitated using anti-HA monoclonal antibody 12CA5. The immunoprecipitates were used in a coupled MKK assay to determine the activity of MKK6 or its mutants by including recombinant His-p38 and MBP in the kinase reaction mixture (Fig. 3C). All three mutants fail to activate p38; previous studies with MEK where A and M mutants were created also resulted in a loss of activity (33). Here the E mutant also failed to activate p38 while in contrast, the analogous structural change in MEK produced an active enzyme (33).
Regulation of MAPK Kinase Activation-We and others recently reported that the low molecular weight GTP-binding proteins (29,34) Rac1 and Cdc42, possibly via activation of p21-activated kinase (PAK) (29,35), regulate the activation of p38 and JNK. Other data suggest that a MKK kinase, MEKK1, may lie downstream of these regulators and thus function as an activator of the MKKs (34). To examine whether MKK6 is regulated in a similar manner we co-expressed dominant active forms of Rac1, Cdc42, or MEKK1 with HA-tagged MKK6, MKK4, or MKK3 in COS-7 cells. The activity of MKKs were determined by immunokinase assay. Overexpression of the active-form Rac1, Cdc42, or MEKK1 led to activation of MKK4 only; we failed to observe activation of MKK6 or MKK3 as detected by our assay system (Fig. 4). This observation confirms the previous report that MKK4(JNKK1) is downstream of Rac1, Cdc42, and MEKK1 (34) and further suggests that other regulatory steps lead to activation MKK6 or MKK3. DISCUSSION Herein we describe the properties of a newly identified member of a family of enzymes known as MAP kinase kinases (MKK). Nucleotide sequence data indicates that the human cDNA clones we isolated encode two proteins generated from one gene; the distinct mRNA species are produced through alternative splicing from a single gene encoding a MAP kinase kinase, here termed MKK6. There is tissue-specific expression of the isoforms of MKK6. Functional studies indicate that MKK6 can activate p38 but not ERK or JNK.
In contrast to the previously described members of this family, which are expressed in almost all tissues with a single transcript at the mRNA level (7)(8)(9)22), numerous splice variants of this gene exist. One splice form, MKK6, appears restricted to expression in skeletal muscle. In contrast, MKK6b is present in various forms in multiple tissues. Thus this represents the first report suggesting there are tissue-specific isoforms of the MKK family. Unique functions for tissue-specific splice forms of MKK6 is not provided by our work. However, demonstration of the presence of such forms suggests the possibility of regulation that is tissue and/or cell specific.
Here we also determined that Ser 151 and Thr 155 are likely to be important phosphorylation sites since mutation of these residues to alanine prevented activation of MKK6. This region with epitope-tagged p38, Erk1, or JNK1 together with an expression vector encoding MKK3, MKK4, MKK6, MKK6b, or control DNA represented by empty vector; or transfected with a signal expression plasmid of MKK6 or MKK6 mutated at ATP binding site (M mutants), phosphorylation sites (A mutant and E mutant), or empty vector. Some of the cultured cells transfected with control empty plasmid were exposed to UV radiation (50 J/m 2 ); p38, Erk1, or JNK1 were isolated by immunoprecipitation with anti-epitope antibody. The protein kinase activity was measured in the immunocomplex with [␥-32 P]ATP and MBP (A) or GST-ATF2 (B) as substrate. The coupled kinase assay was used to test the kinase activity of MKK6 mutants by including recombinant p38 and MBP in the kinase reaction (C). The product of the phosphorylation reactions was visualized after SDS-PAGE by autoradiography. FIG. 4. Regulation of the activation of MKK3, MKK4, and MKK6. COS-7 cells were transfected with epitope-tagged MKK6, MKK4, or MKK3 together with empty expression vector or an expression vector encoding active-form MEKK1, Rac1, Cdc42, Ras, RhoA, and Raf. Some of the cell cultures were exposed to UV radiation (50 J/m 2 ). The MKK6, MKK4, or MKK3 was isolated by immunoprecipitation with the use of anti-epitope antibody. The kinase activity was measured in the immunocomplex with [␥-32 P]ATP and recombinant p38 as substrate. The product of the phosphorylation reactions was visualized after SDS-PAGE by autoradiography.
is analogous to one established to be important by others for MEK (33). We also attempted to replace Ser 151 and Thr 155 with Glu to simulate the negative charge resulting from phosphorylation. This approach did result in a gain-of-function mutation for MEK (33). Here we failed to observe a similar effect. This may be due to the fact that the phosphorylation sites of MEK are both Ser while in MKK6 the site is determined by a Ser and Thr. The Glu may not mimic phosphothreonine in the same way that it does phosphoserine.
Insofar as the MAP kinase family is concerned it was not surprising that MKK6 and MKK3 are quite similar in substrate specificity. MKK6 and MKK3 can be classified into a subgroup distinct from others members of the MKK family. This is justified based on sequence homologies and ability to activate p38 without activating ERK or JNK. Moreover here we showed that regulators of MKK4, namely Rac1 and Cdc42, apparently do not control activation of MKK6. Thus there may be two distinct pathways leading to p38 activation; one via MKK6/3 and another involving MKK4. Interestingly the studies of Saito and colleagues (18,19) with S. cerevisiae revealed two distinct pathways leading to Hog1p activation. One involves a two-component histidine kinase system leading to activation of the MKK encoded by the PBS2 gene (18). In contrast, an alternative pathway leading to PBS2p activation was discovered involving a membrane protein termed Sho1 that directly activates PBS2 through SH3 interactions (19).
Given the distinct expression patterns of MKK6, it is likely that this MKK acts as a regulator of p38 activation depending on expression in a given tissue and/or cell type. A future challenge will be to define the specific function of individual p38 activators and how these proteins interact with other components of the signal machinery to transduce extracellular information into cellular responses. Gene targeting of individual MKKs in cultured cells and mice should shed light on this issue.
We suggest that multiple signaling pathways can control activation of p38. A model which takes into account the data presented here as well as the findings of others (7,8,26,34,35) is provided in Fig. 5. Investigations that may lead to the elucidation of alternative pathways for p38 activation via MKK3 and/or MKK6 are currently underway.