Characterization of the Mitogen-activated Protein Kinase Kinase 4 (MKK4)/c-Jun NH2-terminal kinase 1 and MKK3/p38 Pathways Regulated by MEK Kinases 2 and 3

We previously reported the isolation of cDNAs encoding two mammalian mitogen-activated protein kinase (MAPK)/extracellular-regulated kinase (ERK)kinase kinases, designated MEKK2 and MEKK3 (Blank, J.L., Gerwins, P., Elliott, E.M., Sather, S. and Johnson, G.L. (1996) J. Biol. Chem. 271, 5361–5368). In the present study, cotransfection experiments were used to examine the regulation by MEKK2 and MEKK3 of the dual specificityMAP kinase kinases, MKK3 and MKK4. MKK3 specifically phosphorylates and activates p38, whereas MKK4 phosphorylates and activates both p38 and JNK. Coexpression of MEKK2 or MEKK3 with MKK4 in COS-7 cells resulted in activation of MKK4, as assessed by enhanced autophosphorylation and by its ability to phosphorylate and activate recombinant JNK1 or p38 in vitro. MKK3 autophosphorylation and activation of p38 was also observed following coexpression of MKK3 with MEKK3, but not with MEKK2. Consistent with these observations, immunoprecipitated MEKK2 directly activated recombinant MKK4 in vitro but failed to activate MKK3. The sites of activating phosphorylation in MKK3 and MKK4 were identified within kinase subdomains VII and VIII. Replacement of Ser189 or Thr193 in MKK3 with Ala abolished autophosphorylation and activation of MKK3 by MEKK3. Analogous mutations in MKK4 indicated that Ser221 and, to a lesser extent, Thr225 were necessary for MKK4 activation by MEKK2 and MEKK3. These data indicate that MKK3 is preferentially activated by MEKK3, whereas MKK4 is activated both by MEKK2 and MEKK3. Consistent with these observations, MEKK2 and MEKK3 also activated JNK1 in vivo. However, MEKK3 failed to activate p38 when coexpressed in either the absence or presence of MKK3, indicating that MEKK3 is not coupled to p38 activation in vivo. These observations suggest that regulation of p38 and JNK1 pathways by MEKK3 may involve distinct mechanisms to prevent p38 activation but to allow JNK1 activation.

Activation of MEK1 involves phosphorylation on serines 218 and 222 within a conserved regulatory region between kinase subdomains VII and VIII (38 -41). Raf-1 directly phosphorylates these sites and is probably the primary MEK activator in vivo. In most cases, Raf-1 activation by receptor tyrosine kinases and G protein-coupled receptors involves Ras, which interacts directly with Raf-1, causing translocation and activation of Raf-1 at the plasma membrane (2)(3)(4)38). A-Raf and B-Raf have also been shown to activate the ERK pathway (42), although their pattern of expression is more restricted than that of Raf-1 (43) and their mechanisms of activation are less well understood. However, B-Raf appears to be the major MEK activator in brain (44,45) and nerve growth factor-treated PC12 cells (46), and A-Raf has been shown to activate MEK1 in cardiac myocytes following endothelin-1 stimulation (47).
Whereas Raf-1 shows no MKK4-stimulating activity (37), members of the mammalian MEKK family of serine/threonine kinases have been demonstrated to activate MKK4 (37,48,49). The cDNAs corresponding to three MEKK isoforms were isolated by virtue of their sequence homology with Ste11 and Byr2 (49,50), protein kinases involved in the pheromone mating response pathway in Saccharomyces cerevisiae and Schizosaccharomyces pombe, respectively. The cDNA corresponding to MEKK1 encodes a protein of 161 kDa (51), whereas those for MEKK2 and MEKK3 encode proteins of 70 and 71 kDa, respectively (49). MEKK2 and MEKK3 are also more closely related in amino acid sequence, being 94% identical to each other and ϳ50% identical to MEKK1 through their respective catalytic domains (49). When transiently overexpressed, MEKKs induce constitutive activation of JNK and ERK pathways (37, 41, 49 -53), but not p38 (37,49); both immunoprecipitated MEKK1 and MEKK2 can phosphorylate recombinant MKK4 and MEK1 in vitro (37, 48 -51), whereas MEKK3 is unable to phosphorylate either substrate (49). In NIH3T3 cells, expression of inducible MEKK1 catalytic domain results in maximal JNK activation in the absence of detectable ERK stimulation (48), suggesting that MEKK1 may preferentially regulate the JNK pathway. Current evidence suggests that MEKK2 and MEKK3 show little or no preference for activating either JNK or ERK pathways (49).
Although the mechanisms of MEKK activation are unknown, MEKK activities have been implicated in MAPK pathways regulated by a number of extracellular stimuli in several cell types. For example, an endogenous MEKK1 is activated in epidermal growth factor-stimulated PC12 cells via a pathway involving Ras (54). MEKK1 is also activated in mast cells following aggregation of high affinity IgE receptors (55) and in tumor necrosis factor-␣-stimulated macrophages (56). Using a kinase-deficient MEKK1 mutant, evidence for MEKK1 involvement in tumor necrosis factor-␣-induced NF-B activation in NIH3T3 cells has also been reported (57). Another serine/threonine kinase shown to activate MKK4 in vitro is the MEKKrelated TAK-1 enzyme, which is activated in murine osteoblast cells following transforming growth factor-␤ stimulation (58). TAK-1 has also recently been shown to activate MKK3 and MKK6 in vitro and in vivo (35). A fifth MEKK homologue, designated Tpl-2, has also recently been shown to activate the ERK and JNK pathways in vivo and to directly phosphorylate MEK1 and MKK4 in vitro (59).
Using transient transfection experiments, we now show that MEKK3 activates MKK3 in vivo and identify Ser 189 and Thr 193 in MKK3 as essential activating phosphorylation sites. Furthermore, MKK3 is not significantly activated by MEKK2 in vivo or in vitro, indicating that MKK3 is differentially regulated by MEKK2 and MEKK3. We also show that MEKK2 and MEKK3 can activate the MKK4/JNK pathway in vivo and that Ser 221 and, to a lesser extent, Thr 225 in MKK4 are necessary for activation. However, MEKK3 fails to activate p38 when coexpressed in either the absence or presence of MKK3, suggesting that regulation of p38 and JNK involves distinct mechanisms to oppose sustained p38 stimulation in the presence of MEKK3.

EXPERIMENTAL PROCEDURES
DNA Constructs-Constructs for bacterial expression of MKK3, MKK4, JNK1, p38, ATF2, and c-Jun as translational fusions with glutathione S-transferase (GST) were a kind gift from Dr. R. J. Davis and have been described previously (6,12,32). GST-MEKK3 was prepared with pGEX-3T (Pharmacia Biotech Inc.) and a polymerase chain reaction fragment encoding amino acids 1-301 of MEKK3. The GST fusion proteins were purified by affinity chromatography (70) on glutathione-Sepharose (Pharmacia). Purified fusion proteins were resolved by SDS-polyacrylamide gel electrophoresis using 10 or 12% polyacrylamide in the presence of 0.1% SDS (71) and quantified by comparative Coomassie Blue staining using bovine serum albumin to construct a standard curve. The Flag epitope tag DYKDDDDK was introduced between codons 1 and 2 of MEKK2, MEKK3, MKK3, MKK4, and p38 by insertional overlapping polymerase chain reaction (72). The cDNA for each epitope-tagged kinase was subcloned into the polylinker of pCMV5 (73) for mammalian expression. The sequence of each of these Flagtagged constructs was confirmed by DNA sequencing. JNK1 cDNA was transiently expressed as a hemagglutinin (HA) epitope-tagged protein from pSR␣ (52) and was kindly provided by Dr. G. L. Johnson.
The cDNAs containing the entire coding region of Flag epitopetagged MKK3 and MKK4 were amplified by polymerase chain reaction and subcloned into pALTER (Promega). Site-directed mutagenesis of MKK3 and MKK4 was performed using the Altered Sites II mutagenesis system (Promega) according to the manufacturer's instructions and confirmed by DNA sequencing. Flag-MKK3 mutants containing single and double point substitutions of Ser 189 and Thr 193 with alanine (S189A, T193A, S189A/T193A) were subcloned into pCMV5 for mammalian expression. Corresponding single and double point mutants of Flag-MKK4 (S221A, T225A, and S221A/T225A) were similarly expressed from pCMV5.
Cell Culture and Transfection-COS-7 cells were maintained in Dulbecco's modified Eagle's medium at 37°C in a 5% CO 2 , 95% air mixture. HEK293 cells were maintained under similar conditions in minimum essential medium (Life Technologies, Inc.). Media were supplemented with 10% fetal bovine serum, 50 IU/ml penicillin, 50 g/ml streptomycin, 2.5 g/ml amphotericin B, and 2 mM L-glutamine (Life Technologies, Inc.). COS-7 cells were transfected with 2 g of each plasmid DNA/plate using the DEAE-dextran/cholorquine method (68). HEK293 cells were transfected with 5 g of each plasmid DNA/plate by calcium phosphate precipitation (69). Where necessary, the total amount of DNA used for separate transfections was made equivalent with empty pCMV5 vector. 48 h after transfection, COS-7 and HEK293 cells were made quiescent for 16 -18 h in the appropriate serum-free medium containing 0.1% bovine serum albumin.
Protein Kinase Assays-Immune complex protein kinase assays were performed in a final volume of 40 l of kinase buffer. Protein kinase reactions were initiated by the addition of the appropriate recombinant substrate proteins (5 g each of GST-c-Jun and GST-ATF2; 1 g each of GST-JNK, GST-p38, GST-MKK3, and GST-MKK4) and kinase buffer containing 250 M ATP and 5 Ci of [␥-32 P]ATP. Reactions were incubated for 20 min at 30°C and terminated by the addition of an equal volume of Laemmli sample buffer. Samples were boiled for 5 min, and proteins were resolved by SDS-polyacrylamide gel electrophoresis through 10% polyacrylamide. Gels were dried under vacuum, and phosphorylated proteins were visualized by autoradiography. Data shown in each figure are representative of 2-5 independent experiments.
Antisera-Rabbit antiserum B128 was raised against a synthetic peptide corresponding to amino acids 602-619 in the COOH-terminal sequence of MEKK2 (49) as described previously (74). Rabbit antiserum B130 was raised against a purified recombinant GST fusion protein containing amino acids 1-301 of MEKK3 (49).

RESULTS
Differential Activation of MKK3 and MKK4 by MEKK2 and MEKK3-Epitope-tagged MKK3 and MKK4 were separately expressed in COS-7 cells either alone or in the presence of MEKK2 or MEKK3. MKK3 and MKK4 were then isolated by immunoprecipitation, and their protein kinase activities were assessed in vitro. Fig. 1A shows that MKK4 autophosphorylation was enhanced by coexpression with either MEKK2 or MEKK3. MKK3 autocatalytic activity was also enhanced by coexpression with MEKK3, whereas MEKK2 failed to cause a significant increase in MKK3 autophosphorylation (Fig. 1A). MKK3 and MKK4 activity was also directly assessed in the immune complex kinase assay using purified recombinant p38 or JNK1 as substrates, respectively. Fig. 1B indicates that MKK4-catalyzed phosphorylation of JNK1 was enhanced by coexpression of MKK4 with either MEKK2 or MEKK3. By contrast, phosphorylation of p38 by MKK3 was only stimulated following coexpression of MKK3 with MEKK3. Western blot analysis using a monoclonal antiserum to the Flag epitope showed that expression of MKK3 or MKK4 was essentially unaffected by MEKK coexpression (Fig. 1C).
To determine whether MKK-catalyzed phosphorylation of JNK1 and p38 resulted in their activation, immune complex kinase assays of MKK3 and MKK4 activity were performed in which ATF2 was provided as an in vitro substrate for JNK1 or p38. Fig. 2A shows that ATF2 phosphorylation by JNK1 was stimulated by immunoprecipitates of MKK4 obtained from COS-7 cells coexpressing either MEKK2 or MEKK3. Similar data were obtained in parallel experiments in which p38 and ATF2 were the sequential substrates (Fig. 2B), confirming that p38 is also an effective substrate for MKK4 in vitro (32,37). Thus, MEKK2 and MEKK3 activated MKK4, enhancing its ability to autophosphorylate and to phosphorylate and activate JNK1 and p38 in vitro.
Activation of p38 by immuonprecipitated MKK3 was also observed following coexpression of MKK3 with MEKK3, whereas MEKK2 failed to activate the MKK3-p38 pathway (Fig. 2C). Of note, MKK3 activation could not be measured using JNK1 as substrate (data not shown), confirming that MKK3 is specific for p38 (32)(33)(34)(35). Thus, by all criteria used to assess MKK3 activity (i.e. autophosphorylation, phosphorylation, and activation of p38), MKK3 is activated by MEKK3 but not by MEKK2. MEKK2 Directly Activates MKK4 but Not MKK3-The cotransfection studies show that MKK4 is activated by MEKK2 and MEKK3, whereas MKK3 can only be activated by MEKK3. To establish whether these effects are direct, epitope-tagged MEKK2 and MEKK3 were separately expressed in COS-7 cells and isolated by immunoprecipitation. MEKK activity was then assessed in an immune complex kinase assay using purified recombinant MKK3 or MKK4 as substrates. Immunoprecipitated MEKK2 strongly autophosphorylated in vitro, whereas MEKK3 did not (Fig. 3). This was not due to differences in MEKK expression or recovery after immunoprecipitation, both of which are quantitatively similar (data not shown). Immunoprecipitated MEKK2 caused a small but equivalent increase in MKK3 and MKK4 phosphorylation, whereas MEKK3 was without effect. In other experiments, MKK4 phosphorylation by MEKK2 exceeded that of MKK3 (not shown). Under different in vitro conditions, the murine homologue of human MKK4, termed SEK1 or JNKK, has been shown to be a good substrate for direct phosphorylation by MEKK2 (49).
To reconstitute each activation pathway in vitro using recombinant proteins, JNK and ATF2 were provided with MKK4, whereas p38 and ATF2 were provided with MKK3 (Fig. 3). Immunoprecipitated MEKK2 activated MKK4, resulting in phosphorylation of ATF2 by JNK. Omission of MKK4 prevented ATF2 phosphorylation (data not shown), indicating that activation of MKK4 by MEKK2 was direct. In this and other experiments, MEKK2 caused little or no activation of MKK3, as shown by p38-dependent ATF2 phosphorylation. Thus, MEKK2 preferentially activates MKK4 over MKK3 in vivo (Figs. 1 and 2) and in vitro (Fig. 3). By contrast, immunoprecipitated MEKK3 was inactive with either MKK3 or MKK4 as substrate under all in vitro conditions examined. As has been speculated (49), the failure of MEKK3 to show significant kinase activity in vitro may be due to nonoptimal assay conditions or to loss of an essential cofactor during immunoprecipitation. Therefore, it is not possible at present to establish whether the activation of MKK3 by MEKK3 observed in vivo is direct.
Identification of the Sites of Activating Phosphorylation in MKK3 and MKK4 -MEKs are activated by phosphorylation on two serine residues within a regulatory region between kinase subdomains VII and VIII (38 -41). These sites are necessary for MEK activation by Raf-1 and MEKK1 in vitro and in vivo. Sequence comparison of MKK3, MKK4, and MKK6 with MEK1 and MEK2 indicates that these phosphorylation sites are conserved (32)(33)(34)(35). To test whether MKK3 and MKK4 activation by MEKK2 and MEKK3 required phosphorylation at these sites, a series of epitope-tagged MKK mutants were prepared in which one or both sites were replaced by alanine. Wild-type and mutant MKKs were separately expressed in COS-7 cells either alone or in the presence of MEKK2 or MEKK3. MKKs were then isolated by immunoprecipitation, and their protein kinase activities were measured in vitro.
Replacement of Ser 189 and/or Thr 193 in MKK3 with Ala blocked MEKK3-induced autophosphorylation (Fig. 4A) and activation of MKK3, as assessed by its ability to phosphorylate and activate p38 using ATF2 as substrate (Fig. 4B). Immunoblot analysis showed that MEKK3 and MKK3 mutants were each present at similar levels in the appropriate cell lysates (Fig. 4C). These data indicate that both Ser 189 and Thr 193 in MKK3 are essential for MKK3 activation by MEKK3.
In the absence of MEKK coexpression, wild-type MKK4 displayed a low basal autocatalytic activity that stimulated JNK1 in vitro (Figs. 5 and 6). This was blocked by the substitution in MKK4 of Ser 221 with Ala and greatly attenuated by replacement of Thr 225 with Ala. Coexpression of these mutants with MEKK2 (Fig. 5) and MEKK3 (Fig. 6) demonstrated that Ser 221 was essential for autophosphorylation and activation of MKK4. Although Thr 225 in MKK4 was also necessary for full activation by MEKK2 and MEKK3, a low level of MKK4 activation was observed when this site was replaced by Ala; substitution of both Ser 221 and Thr 225 with Ala blocked MEKK-induced activation of MKK4 (Figs. 5 and 6). Western blot analysis using selective antisera indicated that the expression level of both MEKK2 (Fig. 5C) and MEKK3 (Fig. 6C) in the appropriate cell lysates was constant. Antiserum to the Flag epitope also showed that wild-type and mutant forms of MKK3 or MKK4 were present at comparable levels in either the absence or presence of coexpressed MEKK (Figs. 5C and 6C). These data, therefore, identify Ser 221 and, to a lesser extent, Thr 225 in MKK4 as necessary sites for basal and MEKK-induced autophosphorylation and activation of MKK4.
MEKK2 and MEKK3 Activate JNK1 but Fail to Activate p38 in Vivo-Consistent with the ability of MEKK2 and MEKK3 to activate MKK4 in vivo, coexpression of either MEKK2 or MEKK3 in HEK293 cells with epitope-tagged JNK1 resulted in comparable levels of JNK1 activation when either ATF2 or c-Jun was provided as substrate (Fig. 7). Previous work has also shown that MEKK2 and MEKK3 activate endogenous JNK in HEK293 cells but have no effect on endogenous p38 activity (49). To extend these observations, epitope-tagged p38 was expressed in HEK293 cells, isolated by immunoprecipitation, and assayed for in vitro kinase activity using ATF2 as substrate. Unlike JNK1, p38 autophosphorylated under these conditions (Fig. 8). Treatment of cells with anisomycin (Fig. 8) or sorbitol (not shown) activated p38, indicating that the epitope-tagged enzyme was functionally coupled. However, co-  7 and 8). The ability of immunoprecipitated Flag-MKK3 to activate GST-p38 was assessed in vitro using GST-ATF2 as substrate (lanes 2, 4, 6, and 8).
Only those regions of the autoradiograms representing the position of GST-ATF2 are shown. expression of p38 with either MEKK2 or MEKK3 failed to cause any detectable p38 activation, whereas MKK3 expression produced a small stimulatory effect on p38 activity. We surmised that the lack of effect of MEKK3 on p38 activity was due to the absence of an MKK3-type enzyme in HEK293 cells and that anisomycin-stimulated p38 activity was due to an MKK3independent pathway. We therefore coexpressed MEKK2 or MEKK3 together with MKK3 and p38 in these cells to reconstitute the pathway in vivo. Surprisingly, under these conditions where MKK3 is activated by MEKK3 (Figs. 1, 2, and 4), MEKK3 did not cause activation of p38 (Fig. 8). These data have been reproduced using transient expression in COS-7 cells (not shown) and suggest that regulation of p38 and JNK1 by MEKK3 involves a mechanism to prevent p38 activation but to allow JNK1 activation. The stress-activated pathways regulated by MEKK2 and MEKK3 are summarized schematically in Fig. 9.

DISCUSSION
The work of Johnson and colleagues (49,50) has identified three structurally and functionally related mammalian MEK kinase isozymes whose activities are clearly distinct from those of Raf kinases. Transient overexpression studies indicate that MEK kinases can regulate at least two separate MAPK signaling pathways involving the ERK and JNK subgroups (37, 41, 49 -53). Although several of these reports suggest that MEKK1 activates MKK4 and/or JNK in preference to MEK and/or ERK (48,52,53), MEKK2 and MEKK3 do not appear to show a significant preference for activation of JNK relative to ERK (49). By contrast, Raf-1 is a specific activator of the ERK pathway and does not effect JNK activation (37,52). This selectivity is attributed to the observation that the specific ERK activators, MEK1 and MEK2, are phosphorylated and activated by Raf-1 (38 -41), whereas the JNK activator, MKK4, is not a substrate for Raf-1 (37).
Although a weak effect of MEKK1 on p38 activity in HEK293 cells has recently been reported (53,64), attempts to demonstrate significant activation of p38 kinase by MEKK1, MEKK2, or MEKK3 have not been successful (37,49). Since MKK4 can phosphorylate and activate JNK and p38 (32, 34, 37), it would not be predicted that MEK kinase isozymes that activate the MKK4/JNK cascade fail to stimulate p38 activity in the same cell type. However, the recent identification of MKK3 and MKK6 as specific activators of p38 (32-35) allows for a more direct assessment of MEKK involvement in the p38 activation pathway. In the present study, cotransfection experiments were performed to examine the regulation by MEKK2 and MEKK3 of the stress-activated protein kinase pathways involving p38 and JNK1.
We demonstrate here that MEKK2 and MEKK3 activate the MKK4/JNK pathway in vivo and that Ser 221 and, to a lesser extent, Thr 225 located in the conserved regulatory region of MKK4 are the critical sites for activating phosphorylation. Thus, replacement of Ser 221 with Ala is sufficient to block autophosphorylation and activation of MKK4, whereas this substitution at Thr 225 greatly attenuates MKK4 activity; substitution of both Ser 221 and Thr 225 with Ala blocks MEKK2and MEKK3-induced activation of MKK4. A mutant of murine SEK1 lacking both phosphorylation sites equivalent to those in human MKK4 identified in this study is similarly not a substrate for MEKK1 in vitro and blocks activation of JNK when coexpressed with MEKK1 in L929 cells (48). The differential effects of Ser 221 and Thr 225 phosphorylation on MKK4 activity observed here have not previously been described. However, analogous single mutations in MEK1 suggest that phosphorylation of either Ser 218 or Ser 222 by Raf-1 is sufficient to cause MEK1 activation (40,41). Raf-1 appears to phosphorylate both sites approximately equally, whereas MEKK1 shows a preference for Ser 218 in MEK1 (41).
We also show in this report that MEKK3 activates MKK3 in vivo and that Ser 189 and Thr 193 in MKK3 are essential activating phosphorylation sites; substitution of one of these sites with Ala is sufficient to completely block MKK3 activation by MEKK3. Furthermore, MKK3 is not significantly activated by MEKK2 in vivo or in vitro, indicating that MKK3 is differentially regulated by MEKK2 and MEKK3. A recent report has demonstrated that MEKK1 can activate MKK4 but not MKK3 or MKK6 when their corresponding cDNAs are expressed in COS-7 cells (34). These studies, therefore, indicate that MKK3  1, 2, 11, and 12), Flag-MEKK3 (lanes 2, 3, 7, and 8) or Flag-MEKK2 (lanes 5, 6, 9, and 10). Flag-MEKK2 and Flag-MEKK3 were isolated from cell lysates by immunoprecipitation and assayed in the presence of either GST-MKK3 or GST-MKK4 as indicated. Activation of GST-MKK3 was assessed using GST-p38 and GST-ATF2 as sequential substrates (lanes 7, 9, and 11), whereas GST-MKK4 activity was assessed using GST-JNK1 and GST-ATF2 substrates (lanes 8, 10, and 12).
is regulated by MEKK3 but not by MEKK1 or MEKK2. Since MEKK3 is inactive when isolated by immunoprecipitation ( Fig.  3 and Ref. 49), it is not possible to determine at present whether MEKK3 directly phosphorylates and activates MKK3.
MKK3 and MKK6 have been established as specific physiologic regulators of p38 based on the following observations: MKK3 and MKK6 activate p38 in vitro (32,33,35); expression of an MKK3 or MKK6 mutant in which the dual activating phosphorylation sites have been replaced by acidic Glu residues causes constitutive p38 activation in vivo (33); expression of an MKK3 mutant in which Ser 189 and Thr 193 are replaced by Ala causes inhibition of UV-stimulated p38 activity (33). Therefore, the failure of MEKK3 to activate p38 when coexpressed in either the absence or presence of MKK3 is difficult to reconcile with our finding that MEKK3 clearly activates MKK3 in vivo. These observations are, however, strikingly similar to those reported by Xu et al. (51). In their study, both MEK1 and MEK2 were shown to be highly activated by cotransfection with MEKK1. Additionally, the catalytic domain of MEKK1 activated MEK1 and MEK2 in vitro and interacted with MEK in the yeast two-hybrid system. Despite these observations, MEKK1 failed to cause significant activation of endogenous or cotransfected ERK2, indicating that MEKs can be activated by MEKK1 without a concomitant activation of the ERK. The authors (51) speculate that MEKK1 expression may cause expression of an ERK phosphatase or make activated MEK inaccessible to the ERK substrate.
Very recently, Moriguchi et al. (35) have provided evidence that MKK3 and MKK6 can be regulated by transforming growth factor-␤-activated kinase, or TAK-1, a newly described member of the MEKK family. In addition, coexpression of TAK-1 and p38 in COS-7 cells leads to activation of p38 (35). Both MEKK3 and TAK-1 are capable of activating JNK, but, unlike TAK-1, MEKK3 also activates the ERK signaling path- were either singly or jointly replaced by Ala. Each immunopurified form of Flag-MKK4 was assayed in vitro by measuring its autocatalytic activity (A) and by using GST-p38 and GST-ATF2 as sequential substrates (B). C, Western blot analysis of MEKK3 expression was determined in the indicated lysates using rabbit polyclonal antiserum B130, whereas expression of wild type and mutant forms of Flag-MKK3 was detected using monoclonal anti-Flag M2 antiserum.
way (35,49,58). It is possible that the lack of activation of p38 by cotransfected MEKK3 is due to coactivation of a parallel MAPK signaling pathway, possibly involving ERK, which feeds back to inhibit p38 kinase activity. In this regard, a number of dual specificity MAPK phosphatases have been identified whose expression is induced by cellular stress and/or mitogenic stimulation (75), suggesting that these phosphatases play a role in the negative feedback regulation of MAPKs. Such a mechanism has been described for ERK regulation in NIH3T3 cells, where activation of JNK by cellular stress or induction of MEKK1 expression leads to increased expression of MAPK phosphatase-1 (76), a phosphatase that can dephosphorylate and inactivate ERK. Although there is little information regarding the substrate selectivity of different MAPK phosphatases (77), it seems likely from the specific mechanisms by which each MAPK subgroup is activated that complementary mechanisms exist to allow for their differential inactivation.
Clearly, one limitation of the cotransfection approach that we have taken to study MAPK regulation is that it provides no information regarding the temporal relationship between MEKK expression and activation of individual signaling pathways involving ERK, JNK, and p38. It seems likely that longer term regulation of these pathways by MEKKs will also involve specific changes in expression of proteins (e.g. dual specificity MAPK phosphatases (75) and nonenzymatic MAPK inhibitors such as p21 WAF1 (78,79)) that serve to attenuate individual MAPK subtypes in a heterologous or homologous fashion. Based on the data described here, we speculate that rapid induction of MEKK3 expression in model cells will lead to transient p38 activation. If correct, then this system should also provide a model in which to examine the specific mechanisms involved in p38 inactivation.  lanes 2 and 6), MEKK3 (lanes 3 and 7), or MEKK2 (lanes 4 and 8). Following immunoprecipitation, HA-JNK1 activity was assayed using either GST-Jun (lanes 1-4) or GST-ATF2 (lanes 5-8) as substrates. HEK293 cells were transfected with Flag-p38, together with MKK3, MEKK2, and MEKK3 either alone (lanes 2, 6, and 5, respectively) or in the indicated combinations (lanes 3 and 4). The total amount of DNA for individual transfections was made constant using empty vector (pCMV5). Flag-p38 was isolated by immunoprecipitation and assayed using GST-ATF2 as substrate. As a positive control, cells expressing