INTRODUCTION

Eukaryotic cells have developed specific signal transduction pathways for response to and integration of extracellular stimuli. Mitogen-activated protein kinases (MAPKs)¹ represent a family of kinases that respond to diverse stimuli and are composed of sequential protein kinase cascades (1, 2). MAPKs are activated through phosphorylation of a specific threonine and tyrosine by dual specificity MAPK kinases referred to as MKKs or MEKs. MKK/MEKs are phosphorylated and activated by MKK/MEK kinases (MKKK/MEKKs). Homologous kinases in several sequential protein kinase cascades have been identified in yeast and mammalian cells, indicating conserved MAPK modules for signal transduction in eukaryotes (3, 4). There are three well defined MAPK pathways: ERK1/ERK2, also referred to as p42/p44 MAPKs (5, 6); the p38/HOG1 kinases (7-9); and the c-Jun NH₂-terminal kinases/stress-activated protein kinases (JNK/SAPKs) (10-13). In addition, ERK3, which shares 50% sequence homology with ERK1/ERK2 (5), and MAPK5 (14) have been cloned but not characterized in terms of regulation and substrate recognition.

Activation of growth factor receptor tyrosine kinases, heterotrimeric G-protein coupled receptors and specific cytokine receptors activate the ERKs (1, 15). The p38 protein kinases (p38 and p38) are activated by exposure of cells to lipopolysaccharide, proinflammatory cytokines, and cellular stresses such as osmotic imbalance (7-9, 16-18). JNKs include three different isoforms with 2-4 splice variants of each (10-13). JNKs are activated by diverse stimuli, including cellular stresses such as UV and protein synthesis inhibitors, proinflammatory cytokines, and G-protein coupled and tyrosine kinase growth factor receptors (2, 10-13, 19-21).

Each MAPK group is phosphorylated and activated by specific MKKs/MEKs. MEK1 and MEK2 activate ERKs (22-24), MKK3 and MKK6 activate p38 kinases (25-28), and MKK4 activates JNKs (26, 29). MEK5 is presumed to activate MAPK5, but this has not been demonstrated biochemically (14, 30). The activators of MKK/MEKs comprise a group of diverse kinases allowing integration of upstream inputs to regulate MAPK pathways. Raf phosphorylates and activates MEK1/MEK2 (31, 32). MEKKs 1, 2, and 3 (33-35) and tumor progression locus 2 (Tpl-2) (36) activate both the ERKs and JNKs, and immunoprecipitates of these kinases have been shown to phosphorylate and activate MEK1/MEK2 and MKK4. Germinal center kinase (GCK) (37), the mixed lineage kinases (MLK) MLK3/SPRK, DLK/MUK (38-40), and TGF--activated protein kinase (TAK1) (41) show selectivity for activation of the JNK pathway, and immunoprecipitates of these kinases from transiently transfected cells stimulate the phosphorylation of MKK4 in vitro. The multiple kinases that can apparently function as MKKK/MEKKs for the JNK pathway indicates that many different extracellular stimuli will regulate this pathway, emphasizing its importance in the control of cell function. In this report, we present the cloning of a novel MEK kinase, MEKK4. The properties of MEKK4 differ from those of MEKKs 1, 2, and 3 in that it is highly selective for the activation of the MKK4/JNK pathway. MEKK4 has interesting structural motifs suggesting regulatory mechanisms for the control of MEKK4 that are not encoded in some of the previously characterized MEKKs.

EXPERIMENTAL PROCEDURES

The degenerate primers 5-GA(A or G)(C or T)TIATGGCIGTIAA(A or G)CA-3 (sense) and 5-TTIGCICC(T or C)TTIAT(A or G)TCIC(G or T)(A or G)TG-3 (antisense) were used in a polymerase chain reaction (PCR) using first strand cDNA generated from polyadenylated RNA prepared from NIH 3T3 cells as template. Taq DNA polymerase (Boehringer Mannheim) was used in a PCR of 30 cycles (1 min, 94 °C; 2 min, 37 °C; 3 min, 72 °C), followed by a 10-min cycle at 72 °C. A band of approximately 300 bp was recovered from the PCR mixture, and the products were cloned into pGEM-T (Promega). The PCR cDNA products were sequenced and compared with the MEKKs 1, 2, and 3 sequences. A unique cDNA sequence of 262 bp, having significant homology to the catalytic domains of MEKKs 1, 2, and 3, was identified and used to screen cDNA libraries from mouse brain, liver, and NIH 3T3 cells (Stratagene). The  phage libraries were plated, and DNA from plaques were transferred to Hybond-N filters (Amersham Corp.) followed by UV cross-linking of DNA to the filters. Filters were prehybridized for 2 h and then hybridized overnight in 0.5 M Na₂H₂PO₄, pH 7.2, 10% bovine serum albumin, 1 mM EDTA, 7% SDS at 68 °C. Filters were washed 2 times at 42 °C with 2 × SSC, once with 1 × SSC, and once with 0.5 × SSC containing 0.1% SDS (1 × SSC is 0.15 M NaCl, 0.015 M sodium citrate, pH 7.0). Positive hybridizing clones were purified and sequenced. To resolve GC-rich regions, cDNAs were subcloned into M13 vectors (New England Biolabs), and single strand DNA was sequenced. In all cases, both strands of DNA were sequenced. To isolate the 5 end of the gene, we utilized the rapid amplification of cDNA ends (RACE) procedure using the mouse brain Marathon-Ready cDNA kit (Clontech). The first PCR was performed using the Vent exo(-) DNA polymerase (New England Biolabs) and the sense anchor primer supplied with the kit in combination with the MEKK4-specific antisense primer 5-AGTCCAACATGAATGAGCACTGTGCAT-3. The PCR reaction involved 25 cycles (1.5 min, 99 °C; 1 min, 68 °C; 2 min, 74 °C) in a final volume of 50 µl. A second PCR was performed using 5 µl of the reaction mixture from the first PCR as template with the nested anchor primer supplied with the kit in combination with the nested MEKK4-specific antisense primer 5-GCTCCGTTGTTCTCAGAGTTGCTCGAA-3. The second PCR generated a 650-bp fragment that was subcloned into pGEM-T and 21 clones sequenced. The identity of the RACE product was confirmed by PCR using the sense primer 5-AAAATCTAGACCTGCGGCGGGCTAGAGGCGGAGG-3 that encodes the extreme 5 end of the RACE product and the antisense primer 5-GCTCCCGTAGTTAACTTTGAAGGTGA-3 that is based on the cDNA encoded in the original clone that was isolated from the brain cDNA library. Vent exo(-) DNA polymerase was used in a PCR consisting of 30 cycles (1.5 min, 99 °C; 1.5 min, 66 °C; 3 min, 74 °C) in a final volume of 50 µl with first strand cDNA generated from mRNA isolated from NIH 3T3 cells as template. A second PCR was performed using 2 µl of the first reaction as a template and the nested MEKK4-specific antisense primer 5-CTGGAATCGATTTTTTTGGCAAAGACC-3 that is also encoded in the original brain cDNA clone. PCR conditions and the sense primer were the same as in the first PCR reaction. The resulting 645-bp PCR product was purified, cloned into pGEM-T, and five cDNAs from each of two separate PCRs were sequenced, and all confirmed the 5 sequence of MEKK4 that was obtained through the RACE procedure.

Poly(A)⁺ RNA (2 µg) from eight different mouse tissues were separated by denaturing formaldehyde, 1.2% agarose gel electrophoresis, transferred to a charge-modified nylon membrane by Northern blotting, and fixed by UV irradiation (mouse multiple Northern blot, Clontech). The membrane was hybridized with either a 300-bp cDNA fragment derived from the catalytic domain of MEKK4 (recognizes both splice forms of MEKK4) or a -actin control probe. The probes were randomly primed and labeled with [³²P]dCTP (Prime it II, Stratagene), and hybridization was performed as described for screening of cDNA libraries.

All cDNAs encoding MEKK4 used in transient transfection assays were subcloned into the mammalian expression plasmid pCMV5. Where indicated, MEKK4 and MEKK4 kin were epitope-tagged using a PCR at their carboxyl terminus with the hemagglutinin (HA)-tag sequence YPYDVPDYA. The insertion of a carboxyl-terminal epitope tag was performed using the sense oligonucleotide 5-TCTAAGCAGGGGCCCATAGAAGCTATC-3, which encodes an ApaI restriction site encoded in MEKK4, and the antisense oligonucleotide 5-CTCTCTAGAGGTACCTCATTAAGCATAATCTGGAACATCATATGGATACTCTTCATCTGTGCAAACCTTGAC-3, which encoded an XbaI restriction site, two sets of termination codons, the HA-tag sequence, and the MEKK4 sequence. These primers were used in a PCR using Vent exo(-) DNA polymerase with MEKK4 or MEKK4 kin (see below) as templates. The PCR reaction involved 30 cycles (1 min, 99 °C; 2 min, 68 °C; 3 min, 74 °C). The PCR products were purified and digested with ApaI, blunt ended, and then digested with XbaI and cloned into pCMV5. This gives a catalytic domain of MEKK4 with an initiation methionine at position 1301. The sequences were confirmed by DNA sequencing.

Kinase inactive variants of MEKK4 were generated by changing lysine at position 1361 to a methionine using the Altered Sites II in vitro Mutagenesis Systems (Promega) with the mutagenisis oligonucleotide 5-ACAGGGGAGCTGATGGCCATGATGGAGATTCGATTTCAGCCTAAC-3. The mutation was confirmed by DNA sequencing.

The MEKK4 catalytic domain used in polyhistidine tagged or GST-fusion proteins was generated by PCR using the sense primer 5-AAGCTTGGATCCGAATTCAGGAGAAAGAATATCATCGGCCAA-3, which encodes EcoRI and BamHI restriction sites, followed by MEKK4 sequence starting at amino acid 1302 in combination with the T7 primer encoded in pBluescript II SK^-. MEKK4 in pBluescript II SK^- was used as template for a PCR of 30 cycles (1 min, 92 °C; 2 min, 58 °C; 3 min, 72 °C) using Taq DNA polymerase (Boehringer Mannheim). The PCR product was digested with restriction endonucleases directed toward sites encoded in the sense primer and in the multiple cloning site of pBluescript SK II and cloned into pRSET (BamHI and XhoI; New England Biolabs) or pGEX (EcoRI and XhoI; Pharmacia Biotech In.). Growth, induction, and purification was performed according to the manufacturer instructions.

Wild type and kinase inactive (K1361M), non-epitope tagged, catalytic domains of MEKK4 were constructed using the same procedure as for the fusion protein constructs except that cDNAs were cloned into pCMV5.

The plasmids encoding HA-tagged JNK1 (pSR) and JNK2 (pCMV5) have been described previously (10, 11). Rac (Q61L) and Cdcd42 (Q61L) were in pCMV5.

HEK293 cells were maintained in Dulbecco's modified Eagle's medium that contained 10% bovine calf serum, 100 units/ml penicillin, and 100 µg/ml streptomycin at 37 °C in 5% CO₂, 95% air. COS cells were maintained under the same conditions except that 5% bovine calf and 5% newborn calf serum was used. Cells were grown and transfected in 100-mm tissue culture dishes and transfected using the calcium phosphate (HEK293) or DEAE-dextran (COS) methods (42).

Peptides corresponding to the COOH-terminal sequences of MEKK4 (CLESDPKIRWTASQLLD) and p38 (CFVPPPLDQEEMES) were conjugated to KLH and used to immunize rabbits. Antisera were characterized for specificity by immunoblotting of lysates prepared from appropriately transfected HEK293 cells.

JNK activity was measured by immunoprecipitation of endogenous JNK1 and JNK2 using specific antibodies (Santa Cruz Biotechnology), transiently overexpressed HA-tagged JNK1 and JNK2 using a monoclonal antibody directed against the hemagglutinin epitope (the 12CA5 antibody from Berkeley Antibody Co.), or by a solid phase assay using glutathione S-transferase (GST)-c-Jun _(1-79) coupled to glutathione-Sepharose-4B. Transfected cells were lysed in 0.5% Nonidet P-40, 1% Triton X-100, 20 mM Tris-HCl, pH 7.6, 0.25 M NaCl, 3 mM EDTA, 3 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 2 mM sodium vanadate, 20 µg/ml aprotinin, and 5 µg/ml leupeptin. Nuclei were removed by centrifugation at 15,000 × g for 10 min, and the supernatants were used for immunoprecipitation of JNKs. For immunoprecipitations, 400 µg of protein was mixed with the appropriate antibody (1:100 dilution for endogenous JNK1 and JNK2 antibodies and 1:250 dilution for the 12CA5 antibody) in a final volume of 400 µl of lysis buffer and rotated at 4 °C for 1 h. Immune complexes were captured by adding 15 µl of a 1:1 slurry of protein A-Sepharose (Sigma). The mixture was rotated at 4 °C for an additional hr and washed 2 times in lysis buffer and once in kinase buffer (20 mM Hepes, pH 7.5, 10 mM MgCl₂, 20 mM -glycerophosphate, 10 mM p-nitrophenyl phosphate, 1 mM dithiothreitol, 50 µM sodium vanadate). Beads were suspended in 40 µl of kinase buffer containing 10 µCi of [³²P]ATP and 2 µg of GST-c-Jun as substrate. For solid phase JNK assays, 400 µg of protein was mixed with 10 µl of a 1:1 slurry of GST-c-Jun-(1-79)-Sepharose (3-5 µg of GST-c-Jun-(1-79)). The mixture was rotated at 4 °C for 1 h and washed as described above, and kinase reactions were performed by adding 10 µCi of [³²P]ATP. Kinase reactions were performed at 30 °C for 20 min, and reactions were terminated by addition of Laemmli sample buffer. Samples were boiled, and phosphorylated proteins were resolved on SDS, 10% polyacrylamide gels and visualized by autoradiography.

In vitro activation of JNK was determined essentially as described (33). HEK293 cells were transfected with the expression vector pCMV5 or pCMV5 containing HA-tagged wild type or kinase inactive MEKK4. Cells were lysed, and MEKK4 was immunoprecipitated using the 12CA5 antibody as described for the JNK assay above. In addition, recombinant, bacterially expressed, and purified polyhistidine-tagged MEKK4 was used in an in vitro reconstituted coupled kinase assay, where immunoprecipitates or recombinant MEKK4 were mixed with recombinant, bacterially expressed and purified, wild type or kinase inactive (K116M) MKK4 in combination with wild type or kinase inactive (K55M) JNK in the presence of 50 µM ATP (40 µl final volume) and incubated at 30 °C for 20 min. GST-c-Jun-(1-79)-Sepharose beads (3-5 µg) were then added, and samples were rotated at 4 °C for 20 min. Beads were washed, resuspended in 40 µl JNK kinase buffer containing 10 µCi of [³²P]ATP, and incubated for 20 min at 30 °C. Direct phosphorylation of MKK4 was demonstrated by incubating recombinant purified GST-MEKK4 and kinase inactive (K116M) MKK4 together with 10 µCi of [³²P]ATP in 40 µl of kinase buffer (20 mM Pipes, pH 7.0, 10 mM MnCl₂, 20 µg/ml aprotinin) for 20 min at 30 °C. Reaction mixtures were added to Laemmli sample buffer and boiled, proteins were resolved on SDS, 10% polyacrylamide gels, and phosphorylated proteins were visualized by autoradiography.

HEK293 cells transiently transfected with empty vector (pCMV5), the indicated MEKK4 constructs, or treated with EGF (cells starved in 0.1% bovine serum albumin overnight and then activated with 30 ng/ml of EGF for 10 min) were lysed, and ERK1 and ERK2 were immunoprecipitated using specific antibodies (1:125 dilution; Santa Cruz Biotechnology) as described above for immunoprecipitation of JNK. Kinase reactions were performed as described (43) using the EGF_662-681 receptor peptide as a substrate. The amounts of ³²P incorporated into EGF_662-681 receptor peptide was quantitated by spotting samples on P81 phosphocellulose paper followed by scintillation counting.

HEK293 cells transiently transfected with empty vector (pCMV5), the indicated MEKK4 constructs, or cells treated with sorbitol (0.4 M, 20 min) were lysed, and p38 was immunoprecipitated from cell lysates (400 µg) as described for the JNK assay using rabbit antiserum (1:250 dilution) raised against the COOH-terminal peptide sequence of p38 (described above). In vitro kinase assays were carried out as described for the JNK assay with the exception that recombinant ATF-2 was used as a substrate. Proteins were separated through SDS, 10% polyacrylamide gel electrophoresis, and phosphorylated ATF-2 was visualized by autoradiography.

In vitro binding was performed as described (44). COS cells were transiently transfected with MEKK4 and lysed in extraction buffer (10 mM Tris-HCl, pH 7.4, 1% Triton X-100, 5 mM EDTA, 50 mM NaCl, 50 mM NaF, 20 µg/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 2 mM Na₃VO₄) and 200 µg of lysates incubated with 3-5 µg of GST, GST-Rac, or GST-Cdc42 fusion proteins (coupled to Sepharose beads). The GST-fusion proteins were preloaded with either GDP or GTP--S through incubation with 1 mM GDP or GTP--S in PBS at 30 °C for 20 min followed by addition of 12 mM MgCl₂ to terminate loading reactions. Binding was performed for 2 h at 4 °C in a final volume of 0.5 ml. Beads were then washed 3 times in extraction buffer and resuspended in sample buffer, and proteins were separated by SDS, 10% PAGE. MEKK4 proteins were visualized through Western blotting using a primary rabbit anti-MEKK4 antibody (described above) and horseradish peroxidase-linked protein A followed by chemiluminescence (DuPont NEN).

T47D human breast carcinoma cells or HEK293 cells on glass coverslips were fixed in 3% paraformaldehyde in PBS. Cells were permeabilized in 0.2% Triton X-100 in PBS and incubated for 10 min in Dulbecco's modified Eagle's medium, 10% calf serum. The rabbit antibody raised against the MEKK4 COOH terminus was incubated with the cells for 1 h, and cells were then washed extensively in PBS. A Cy³ donkey anti-rabbit antibody (Jackson Immunological Laboratories) was then used as a secondary antibody for visualization. The MEKK4 anti-COOH-terminal peptide antibody has been extensively characterized and shown not to recognize other MEKK proteins. Characterization of the subcellular localization of the other MEKK proteins is being described elsewhere.² Bodipy-ceramide bound to bovine serum albumin was incubated with T47D cells for 15 min prior to fixation for specific staining of the Golgi (45).

RESULTS

Degenerate primers were used in a PCR with cDNA, synthesized from RNA isolated from NIH 3T3 cells, as a template. Of 185 PCR products that were sequenced, 155 encoded MEKK1, 15 encoded MEKK2, and 5 encoded a novel MEKK-like sequence. Using this novel MEKK cDNA fragment, cDNA libraries from mouse brain, liver, and NIH 3T3 cells were screened. A 5.1-kb cDNA was isolated from the brain library that contained a polyadenylated 3 tail and encoded a 1536-amino acid open reading frame. No 5 in-frame stop sites were apparent, indicating that the full coding sequence was not within the 5.1-kb cDNA. The RACE procedure was used to isolate the remainder of the cDNA clone with a 5 in-frame stop codon. The identity of the 5 end was confirmed by PCR with oligonucleotide primers that were designed so that the sense primer started at the extreme 5 end of the RACE product and the antisense primer started in the original clone obtained from cDNA library screens and hence should cover the junction between the original clone and the RACE product. PCR, using these primers, and cDNA, prepared from NIH 3T3 cells, gave a cDNA fragment that when sequenced confirmed the junction between the original clone and the RACE product. The isolated cDNA encoding MEKK4 is 5426 base pairs, which encodes a 1597-amino acid protein with a deduced molecular mass of 180 kDa (Fig. 1A).

An alternative splice form was also isolated from the brain library with a deletion of 52 amino acids between residues 1162 and 1213 in the reported sequence (Fig. 1A). The alternatively spliced mRNAs were confirmed by reverse transcriptase-PCR using oligonucleotides flanking the splice region and cDNA prepared from mRNA from PC12 cells. The cDNAs for both splice forms could be detected; the larger variant expressing the additional 52 residues has been designated MEKK4 and the smaller alternatively spliced form without the 52 amino acids as MEKK4. Northern blot analysis revealed an mRNA for MEKK4 of approximately 6 kb that is expressed in several different mouse tissues (Fig. 2).

The COOH-terminal moiety of MEKK4 contains the 11 consensus subdomains for the catalytic domain of protein kinases (46) and has approximately 55% amino acid homology with MEKKs 1, 2, and 3 (Fig. 1B). The NH₂-terminal region of MEKK4 shares little or no homology with MEKKs 1, 2, or 3, but several sequences within this moiety suggests specific regulatory functions (Fig. 1A). Near the NH₂ terminus is a proline-rich region suggesting a possible interaction with SH3 domain encoded proteins (47-49). A predicted pleckstrin homology (PH) domain (50, 51) is also encoded in the MEKK4 NH₂-terminal region. In addition, a modified consensus Cdc42/Rac interactive binding (CRIB) domain (52) is encoded in the MEKK4 sequence just upstream of the catalytic domain.

Characterization of MEKK4-mediated regulation of the JNK pathway was performed by transient transfection of HEK293 cells. Endogenous JNK1 and JNK2 and transfected HA-tagged JNK1 and JNK2 were characterized for their regulation in response to expression of full-length and the truncated catalytic domains of MEKK4 using immunoprecipitation of JNKs followed by an in vitro kinase assay using GST-c-Jun as a substrate (Fig. 3). Expression of the catalytic domain of MEKK4 (MEKK4) resulted in the activation of the endogenous JNK1 and JNK2 proteins. The mutant kinase inactive MEKK4 (MEKK4 kin) did not increase JNK activity. To verify the regulation of JNK1 and JNK2 by MEKK4, HA-tagged forms of JNK1 and JNK2 were coexpressed with MEKK4 and selectively immunoprecipitated from HEK293 cell lysates. Similar to the regulation of total endogenous JNK activity, measured by the binding to GST-c-Jun beads or immunoprecipitation with specific JNK antibodies, the HA-tagged JNKs were activated in cells expressing MEKK4 (Fig. 3). Expression of either full-length MEKK4 splice variant did not significantly activate JNK. The transfected HA-tagged JNKs show a very modest stimulation relative to endogenous JNK proteins when the full-length MEKK4 is expressed, probably related to the amplification of the signal resulting from overexpression of the kinases. Using ultraviolet irradiation as a control for stimulating JNK activity, it is clear that MEKK4 is a constitutively active mutant and that the full-length MEKK4 does not show significant activity when transiently overexpressed in HEK293 cells. This finding suggests that full-length MEKK4 is most likely regulated by several upstream effectors.

To verify that MEKK4 directly regulated the MKK4/JNK pathway, the sequential kinase cascade was reconstituted in vitro (Fig. 4). Immunoprecipitates of HA-tagged MEKK4 from HEK293 cell lysates or bacterially expressed recombinant MEKK4 were mixed with combinations of wild type and kinase inactive mutants of MKK4 and JNK. JNK activation was assayed by capture of JNKs onto GST-c-Jun beads, followed by a kinase assay using [³²P]ATP. MEKK4 expressed in HEK293 cells or recombinantly in Escherichia coli required functional MKK4 and JNK proteins for stimulation of GST-c-Jun phosphorylation (Fig. 4A). Recombinant MEKK4 also directly phosphorylated kinase inactive MKK4 (Fig. 4B). Thus, MEKK4 phosphorylates and activates MKK4, which regulates JNK activity both in vitro and in vivo. Expression of MEKK4 in HEK293 cells neither activates the ERK1/ERK2 pathway nor the p38 pathway (Fig. 5, A and B). The selective MEKK4-induced activation of the JNK pathway distinguishes it from MEKKs 1, 2, and 3, which are able to activate the ERK pathway when overexpressed in different cell types (33-35).

recomb. MEKK4

MEKK4

Cdc42 and Rac are GTP-binding proteins of the Rho superfamily; they are well characterized for their ability to regulate cytoskeletal functions, including the formation of filopodia and lamellopodia (53). Cdc42 and Rac have also been shown to regulate pathways leading to the activation of JNKs (54-56). Recently, a CRIB motif was proposed having the sequence ISXP(X_2-4)FXH(X₂)HVG (52). A related sequence in MEKK4 is CDTKSDNVM (identical or conserved residues underlined within residues 1311-1324, Fig. 1A). Fig. 6A shows that MEKK4 binds GST-Cdc42 and GST-Rac but not GST alone. The binding of MEKK4 to Cdc42 was significantly GTP-dependent. In contrast, the binding of MEKK4 to GST-Rac was significant when GDP was bound. We have found that GST-Rac has a lower specific activity for GDP/GTP binding relative to Cdc42 (not shown), suggesting not all the GST-Rac is functional. This result may be in part due to altered properties of Rac when fused to GST. Fig. 6B shows that a kinase inactive mutant of MEKK4 inhibits activated Rac and Cdc42 stimulation of JNK activity. Combined results indicate that MEKK4 interacts with Cdc42 and probably Rac in a GTP-dependent manner and activates the JNK pathway. We have been unable to express enough recombinant MEKK4 protein in E. coli with sufficient native folding and specific activity to perform similar binding studies as was done with COS cell-expressed MEKK4. Hence, the possibility exists that the interaction of Cdc42/Rac with MEKK4 involves additional proteins. Further analysis of the interactions of MEKK4 and Cdc42/Rac will require expression of high specific activity MEKK4 using Sf9/baculovirus or other systems for protein purification.

Immunostaining of MEKK4 demonstrates that it is localized in perinuclear vesicular-like structures (Fig. 7). Staining with ceramide, a marker that stains Golgi and Golgi-derived vesicles shows an overlap in staining with MEKK4. Thus, MEKK4 appears associated with Golgi-associated vesicles. Interestingly, Cdc42 was recently shown to associate with Golgi vesicles (57). Thus MEKK4 and Cdc42 are in a common location where interaction and regulation might occur.

DISCUSSION

A number of MKKK/MEKK-like kinases have been postulated to regulate the JNK pathway. These include the MEKKs, GCK, MLK3/SPRK, DLK/MUK, PAKs, TAK1, and Tpl-2 (33-41, 58-60). Immunoprecipitates of these kinases (with the exception of MEKK 3 and PAKs) from transiently transfected cells have all been shown capable of phosphorylating and activating MKK4. This suggests that either many different kinases at the level of MKK4 are capable of regulating MKK4 and the JNK pathway or that the transient transfection and in vitro kinase analysis may not maintain the fidelity of regulation observed in oligomeric assemblies of these kinase modules. To sort out these potential multiple regulatory inputs versus "cross talk" of normally parallel pathways, the analysis of the endogenous proteins and their regulation in addition to gene inactivation studies will probably be required. It is also probable that additional MKK genes are yet to be identified. It should be noted that only MEKK4 of all these kinases has been expressed as a recombinant protein in E. coli and shown to directly phosphorylate and activate MKK4. Recombinant forms of the other kinases will have to be used to define if they are capable of directly phosphorylating and activating MKK4. In this regard, it has been suggested that PAKs are upstream of MEKKs in the activation of the JNK pathway (58-60). Our unpublished observations suggest that PAKs are poor activators of the JNK pathway and do not recognize MEKKs or MKK4 as substrates, suggesting their actions may be indirect.

MEKK4 binds to Cdc42/Rac, and a kinase inactive mutant of MEKK4 inhibits the ability of GTPase deficient, activated, Cdc42 and Rac to stimulate the JNK pathway. Thus, MEKK4 is a strong candidate for being a Cdc42/Rac-regulated MEKK. Testing of this hypothesis will require the generation of antibodies that efficiently immunoprecipitate the full-length MEKK4 protein from cell lysates. These antibodies are currently being generated using fusion proteins as antigens. The fact that MEKK4 has a potential CRIB-like domain and that GST-Cdc42 binds MEKK4 encoding this sequence in a GTP-dependent fashion suggests MEKK4 is in the JNK pathway regulated by Cdc42 and Rac.

MLK3/SPRK was recently demonstrated to be regulated by Cdc42 and Rac and to selectively activate the JNK pathway similar to MEKK4 (38, 39). For MLK3, six of the eight consensus CRIB domain residues are conserved, whereas five of the eight conserved residues are found in MEKK4. The other signature sequences for MLK3 and MEKK4 are different. MLK3 has a leucine zipper-like sequence suggesting it may form homo- or hetero-dimers with other proteins, MEKK4 has a putative pleckstrin homology domain, and both MLK3 and MEKK4 have proline rich sequences that may be involved in SH3 domain or other protein-protein interactions. This suggests that MLK3 and MEKK4 are regulated by different upstream inputs and potentially different extracellular stimuli even though they may both interact with Cdc42/Rac GTP-binding proteins.

Our analysis of the four MEKK proteins begins to suggest one regulatory scenario for having multiple MKKKs regulating the JNK pathway. Subcellular localization studies using antibodies to the COOH terminus of MEKKs 1, 2, 3, and 4 indicate they are localized in different places in the cell.² MEKK4 is shown to be in a perinuclear, Golgi-like localization, whereas MEKK1 is in a post-Golgi vesicle-like compartment. Differential subcellular localization suggests different regulation of the MEKKs by different extracellular stimuli and intracellular signal transduction proteins. If this is combined with localized or differential regulation of JNKs, then each MEKK could indeed have very specific regulatory functions that are not obvious by transient transfection analysis. We are currently pursuing activation, subcellular localization, and the potential redistribution of the MEKKs and other MKKKs to define their respective roles in the control of cell function and responses to extracellular stimuli.