Dual leucine zipper-bearing kinase (DLK) activates p46SAPK and p38mapk but not ERK2.

Because the catalytic domain of dual leucine zipper-bearing kinase (DLK) bears sequence similarity to members of the mitogen-activated protein (MAP) kinase kinase kinase subfamily, this protein kinase was investigated for its ability to activate MAP kinase pathways. When transiently transfected and overexpressed in either COS 7 cells or NIH3T3 cells, wild type DLK potently activated p46SAPK (SAPK/JNK) but had no detectable effect in activating p42/44MAPK. DLK also activated p38mapk when overexpressed in NIH3T3 cells. A catalytically inactive point mutant of DLK had no effect in these experiments. Consistent with its specificity in activating SAPK, DLK activated Elk-1 but not Sap1a-mediated transcription. In NIH3T3 cells, activation of SAPK by v-Src was markedly attenuated by coexpression of K185A, a catalytically inactive mutant of DLK, suggesting that this mutant could function in a dominant negative fashion in a pathway that leads from v-Src to SAPKs. In a series of co-transfection experiments, activation of p46SAPK by DLK was not inhibited by dominant negative mutants of Rac1 and Cdc42Hs, PAK65-R, or PAK65-A, but was attenuated by MEKK1(K432M). DLK(K185A) did not inhibit the ability of constitutively active MEKK1 to activate SAPK. Moreover, K185A significantly inhibited the activation of SAPK by constitutively active V-12 Rac1 and V-12 Cdc42Hs. These results suggest that DLK lies distal to Rac1 and/or Cdc42Hs but proximal to MEKK1 in a pathway leading from v-Src to SAPKs activation.

A large body of work has focused on signal transduction via cytoplasmic protein kinase cascades generically named the mitogen-activated protein kinases or MAP 1 kinases (MAPK) (re-viewed in Ref. 1). The protypical mammalian MAPKs, p42/ 44 MAPK (ERK1 and ERK2) are activated by numerous mitogens and differentiation-inducing stimuli (reviewed in Refs. [2][3][4]. Biochemical and genetic evidence demonstrates that signaling by receptor tyrosine kinases (RTKs) and seven transmembrane receptors involves the activation of a membrane-bound, small guanine nucleotide-binding GTPase called Ras. Once GTPbound, Ras activates a multiple component protein kinase cascade, culminating in the activation of ERK1/2. ERKs serve as the major effector of this cascade, exerting control over several other classes of signal-transducing proteins including other protein kinases, at least one phospholipase A 2 , and a variety of transcriptional regulatory proteins (5).
While work in mammalian systems established the importance of the ERK pathway in signal transduction from RTKs, it has become clear from studies in yeast that multiple mammalian MAPK pathways exist in parallel (50). Using both genetic and biochemical approaches in mammalian cells, the components of several additional MAPK pathways have now been identified. Best characterized is the stress-activated protein kinase (SAPK) pathway. This pathway is thought to lead from the activated Rho subfamily small GTPases Rac1 and Cdc42Hs, to activation of the MAP kinase kinase kinase kinase, p65 PAK , to the MAP kinase kinase kinase, MEKK1, to the dual specificity MAP kinase kinase, MKK4/SEK1, and, finally, to activation of the MAP kinases p46/p54 SAPK . SAPKs were discovered as the principal c-Jun NH 2 -terminal phosphorylating kinases and therefore have also been termed JNKs (6). Distinct from the ERK cascade, the SAPK pathway is predominantly activated by stress-inducing signals such as heat shock, ultraviolet irradiation, anisomycin, proinflammatory cytokines (tumor necrosis factor ␣ and interleukin 1␤), and hyperosmolarity (7). Although G-protein-coupled receptors can signal through pathways leading to the activation of SAPK (8), the upstream signaling events by which the SAPK pathway becomes activated are largely unmapped.
The mixed lineage kinase or MLK subfamily of protein kinases is a recently described subfamily of protein kinases that share two common structural features (9,10). First, each has a distinctive kinase catalytic domain whose primary structure is hybrid between those found in serine/threonine and tyrosine protein kinases. Second, closely juxtaposed COOH-terminal to the catalytic domain, each MLK protein has a domain that is predicted to form two leucine/isoleucine zippers separated by a short spacer region. Additionally, each has both NH 2 -and COOH-terminal motifs suggestive of protein-protein interaction domains. Despite the hybrid structure of the catalytic domains, two members of the family (DLK and MLK3/SPRK) have been shown to exhibit serine/threonine-specific kinase autocatalytic activity in vitro (10 -12).
Dual leucine zipper-bearing kinase, or DLK, has been identified as a member of the MLK subfamily (10). DLK is most abundantly expressed in nervous tissue where it is predominantly found in neurons (12). Here, DLK protein is present in the nerve terminal where it is associated with both the plasma membrane and cytosolic compartments. Like DLK isolated from nerve terminal cytosol, DLK exists in both phosphorylated and dephosphorylated states under basal conditions in aggregating neuronal glial cultures. Membrane depolarization leads to DLK dephosphorylation, an effect that is completely inhibited by cyclosporin A, a specific inhibitor of the calcium/ calmodulin-dependent protein phosphatase calcineurin (12).
When aligned to protein sequence data bases, DLK catalytic subdomains I through VII are most similar to those of the serine/threonine family of mitogen-activated protein kinase kinase kinases including CTR-1, STE11, and Byr2 (10). Recently, multiple putative Rac1-and/or Cdc42Hs-binding proteins were identified in a search of GenBank using the sequence of a minimal Cdc42Hs binding domain in a murine p65 PAK isoform (51). Several MLK proteins including DLK were identified by this search, and MLK3 was demonstrated to bind weakly to Rac1 and Cdc42Hs but not Rho in a GTP-dependent fashion in a filter binding assay.
For these reasons, and having established that wild type DLK was constitutively active when overexpressed by transient transfection (12), we tested whether DLK could activate an established MAP kinase pathway.
Cell Culture-COS 7, NIH3T3, and 293 cells (ATCC) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillin, and streptomycin. For experiments described in Figs. 1 and 4, 2 ϫ 10 5 cells plated in 35-mm tissue culture dishes were transiently transfected with a total of 2 g of the appropriate combinations of eukaryotic expression plasmid using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's protocol. For experiments described in Figs. 2 and 3, cells were transfected by the standard DNA-calcium phosphate coprecipitation procedure. After 15-20 h, cells were washed of precipitates with serum-free medium. In experiments described in Fig. 2, cells were starved for another 24 h by incubation with medium containing 0% (ERK and SAPK expression) or 0.2% (p38 mapk expression) fetal calf serum. Where indicated in figure legends, after 48 h, cells were treated with epidermal growth factor (Sigma) or anisomycin (Sigma).
Immunoprecipitations were performed as described elsewhere (12) using the indicated antibodies. For immunoblotting, immunoprecipitates or 30 l of cell lysate were separated under reducing conditions by SDS-polyacrylamide gel electrophoresis. As described previously (10), proteins were transferred to nitrocellulose or polyvinylidene difluoride membrane (Bio-Rad), blocked overnight, and incubated with the indicated primary antibody diluted in Tris-buffered saline (TBS) containing 0.1% Tween 20. Blots were developed using the ECL chemiluminescent reagent (Amersham) and subjected to autoradiography as directed by the manufacturer.
⌬v-Raf was constructed by deleting the two BstEII fragments from the coding region of 3611MSV. Forty-eight h following transfection, cells were washed, harvested in cold phosphate-buffered saline, and subsequently lysed in 150 l of buffer LL (250 mM KCl, 50 mM HEPES, pH 7.5, 0.1% Nonidet P-40, 10% glycerol, and protease inhibitors) (22) per plate. Protease inhibitors included 2 mM phenylmethylsulfonyl fluoride, 0.5 g/ml pepstatin, 0.5 g/ml leupeptin, and 1 g/ml aprotinin. Luciferase assays were performed as described previously. Values in the luciferase assays were normalized to the corresponding ␤-galactosidase activity resulting from cotransfection of 1 g of pCH110 (Pharmacia) per plate.

RESULTS AND DISCUSSION
DLK Activates p46 SAPK and p38 mapk but Not ERK2-To examine whether DLK activates a defined MAPK pathway, epitope-tagged ERK2 or p46 SAPK was coexpressed with either wild type DLK, a catalytically inactive DLK mutant (K185A) (12), or appropriate vector control. Following cell lysis, immunoprecipitated epitope-tagged MAP kinases were assayed in vitro for catalytic activity. In COS 7 cells transiently transfected with wild type DLK, SAPK catalytic activity was stimulated greater than 9-fold relative to the vector control (Fig.  1A). When overexpressed in a similar fashion, K185A did not induce SAPK activity. As previously reported, anisomycin, an inhibitor of protein synthesis, activated SAPK only 5-fold in the absence of DLK. Addition of anisomycin to cells coexpressing DLK and SAPK resulted in no further activation of SAPK relative to untreated cells overexpressing DLK. In parallel experiments, DLK overexpression in COS 7 cells had no effect on ERK2 activity (Fig. 1B). Here, epidermal growth factor was used to demonstrate that the endogenous ERK pathway was intact.
Additional experiments were carried out in NIH3T3 cells (and 293 cells, results not shown) to demonstrate that the observed results were not cell type-specific (Fig. 2). Again, expression of DLK activated epitope-tagged SAPK catalytic activity but did not activate ERK2. In NIH3T3 cells, DLK activated SAPK activity with a potency similar to that observed following coexpression of a catalytically active COOH-terminal fragment of MEKK1 (20) (also see Fig. 5B).
For this reason, we tested whether DLK could activate coex-2 P. E. Shaw, unpublished data. pressed p38 mapk . As with SAPK, overexpressed DLK stimulated p38 mapk activity with a potency similar to that observed by coexpression of catalytically active MEKK1 or by treatment of cells with anisomycin. Neither empty vector control nor v-Raf induced p38 mapk activity. p38 mapk can be activated by three known MAP kinase kinases or MEKs. While activated MKK3 and MKK6 can activate only p38 mapk , MKK4/SEK1 has been shown to activate both the SAPKs and p38 mapk (36 -40). Therefore, that DLK can activate both SAPK and p38 mapk suggests that DLK might activate these MAPKs via a MEKK1-MKK4-dependent pathway.
DLK Activates Elk-1 but Not Sap1a-mediated Transactivation-Transcriptional activation of c-fos is induced by both mitogenic and cellular stress stimuli via pathways that lead to the activation of both ERK and SAPK (reviewed in Ref. 41). This transcriptional activation is mediated by transcription factor complexes that bind to the serum response element. When activated, ternary complex factors such as Elk-1 and Sap1a associate with the serum response factor (SRF) and bind to the serum response element resulting in c-fos transcriptional activation. Elk-1 and Sap1a both serve as substrates for and are activated by ERK. However, that Elk-1 and Sap1a subserve different physiological roles is suggested by the observation that only Elk-1, and not Sap1a, can be phosphorylated and activated by SAPK (42)(43)(44)(45).
Since overexpressed DLK specifically induced SAPK and not ERK activity, we sought to test whether DLK would activate the transcriptional activity of Elk-1 but not Sap1a. Expression vectors encoding chimeric gene products of Elk-1 and Sap1a were prepared in which the respective ets domains were replaced by the DNA binding domain of the yeast transcription factor Gal4. As described previously, these gene products confer mitogenic and cellular stress signal dependence upon a Gal4-CAT reporter when co-expressed by transient transfection (13,18,46,47). As expected, expression of v-Raf in this system led to potent induction of both Elk-1 and Sap1a-medi-ated transcriptional activation (Fig. 3). Serine to alanine point mutations at Ser-383 in Elk-1 and at Ser-381 in Sap1a have been demonstrated to attenuate Elk-1-and Sap1a-mediated induction of transcription by mitogenic stimuli (46,53,54). Similarly, the Gal-Elk 383A mutant has been shown to abrogate anisomycin-induced transcription in the same model (20). Consistent with these previous observations, overexpressed v-Raf failed to activate Gal-Elk 383A mediated transcription and minimally activated Gal-Sap 381A -mediated transcription.
As shown in Fig. 3, overexpression of DLK in this model led to a 7-fold induction of Gal-Elk-mediated transcription relative to vector only control, but failed to induce Gal-Sap1a-mediated transcription. DLK induced Gal-Elk-mediated transcription with less potency than that observed following v-Raf overexpression. However, the degree of induction of Gal-Elk-mediated transcription following overexpression of DLK was similar to that observed following treatment of these cells with anisomycin (20). Therefore, the finding that DLK activates Elk-1-but not Sap1a-mediated transcription is consistent with DLK's ability to induce SAPK but not ERK catalytic activity.
A Catalytically Inactive Mutant of DLK Inhibits v-Src-induced Activation of p46 SAPK -Transiently transfected v-Src has been shown to activate SAPK via a pathway that is inhibited by co-expression of dominant negative mutants of either Ras or Rac1 (24). Therefore, NIH3T3 cells co-expressing v-Src and epitope-tagged p46 SAPK were used to confirm that each of a series of previously characterized mutants function as dominant-negative mutants in the endogenous SAPK pathway. As predicted by others, catalytically inactive PAK65-A, the regulatory domain of hPAK65 (PAK65-R), which binds and inhibits both Rac1-and Cdc42Hs-dependent signal transduction (24 -26), and MEKK1(K432M) (23) functioned as interfering mutants in these experiments (Fig. 4A). Similar experiments FIG. 2. DLK activates p38 mapk and SAPK but not ERK2 when overexpressed in NIH3T3 cells. Cells were cotransfected with 5 g of plasmid encoding either HA-ERK2 (A), p46 SAPK (B), or HA-p38 mapk (C) and plasmids encoding either appropriate vector control, v-Raf (4 g), Myc-DLK (8 g), or MEKK1 (1 g) or were treated with anisomycin (1 g/ml for 30 min) as indicated. After 48 h, cells were lysed, and immunoprecipitated p46 SAPK , p38 mapk , or ERK2 was assayed for catalytic activity in vitro. Immunoblots from corresponding experiments demonstrate relative expression of ERK2, p46 SAPK , or p38 mapk . MBP, myelin basic protein. confirmed that N-17 Rac1 and N-17 Cdc42Hs attenuated v-Src-induced activation of p46 SAPK (Fig. 4B). Importantly, activation of SAPK by v-Src was markedly attenuated by K185A, a catalytically inactive mutant of DLK (12) (Fig. 4B). That co-expression of K185A inhibited activation of SAPK by v-Src suggested that this mutant could function in a dominant negative fashion in a pathway that leads from v-Src to SAPKs.
DLK Activates SAPK via a MEKK1-dependent Pathway-Since overexpressed MEKK1 activates SAPKs and p38 mapk (37), the epistatic relationship between DLK and MEKK1 was investigated. When coexpressed with DLK in NIH3T3 cells, the catalytically inactive MEKK1(K432M) mutant attenuated SAPK activation (Fig. 5A). Conversely, SAPK activation by the catalytically active form of MEKK1 was not inhibited by coexpression of DLK (K185A) (Fig. 5B). These results emphasize that MEKK1(K432M) does not inhibit DLK-induced SAPK activation by competing with catalytically active DLK for a common downstream substrate (presumably SEK/MKK4). Rather, DLK appears to lie proximal to MEKK1 along a pathway leading to SAPK activation. The incomplete inhibitory effect of MEKK1(K432M) on overexpressed DLKs ability to activate SAPK is consistent with results by other investigators performing similar experiments (24,48). However, it is recognized that given incomplete inhibition, allowance should be made for other potential interpretations of these experiments.
Consistent with the experiments above, neither PAK65-R nor the catalytically inactive PAK65-A inhibited DLK-induced activation of SAPK (Fig. 4A). The GTP-dependent Rac1/ Cdc42Hs binding domain of hPAK65 (named CRIB or GBD) has been localized to a 14-residue domain of hPAK65's NH 2 terminus (51,52). This domain is conserved in multiple proteins including several p65 PAK isoforms, WASP (the protein that is defective in the Wiskott-Aldrich Syndrome), several STE20 homologues, and p120 ACK . PAK65-R, which contains this domain, is expected to competitively inhibit binding of proteins possessing the CRIB/GBD binding domain to Rac1 and Cdc42Hs; when overexpressed, this truncation mutant should inhibit signal transduction requiring activation of these GTPases. Of note, the mixed lineage kinases MLK2 and MLK3 FIG. 3. Overexpressed DLK activates Elk-1-but not Sap1a-dependent transcription. SV40 enhancer-based expression vectors for Gal-Sap or Gal-Elk chimeras or their respective inactive mutants, Gal-Sap 381A or Gal-Elk 383A , were co-transfected as indicated with a luciferase reporter containing five tandem Gal4 binding sites (pG5E4 -38Luc) and expression plasmids encoding either v-Raf (4 g), Myc-DLK (8 g), or their corresponding empty vectors (Ϫ). In each case, the quantification of three independent experiments is shown as fold induction relative to vector control. wt, wild type.  2 g), or vector control (to 2 g). Cells were lysed after 48 h, and immunoprecipitated p46 SAPK was assayed for catalytic activity in vitro. Immunoblots from corresponding experiments were used to evaluate relative expression of immunoprecipitated p46 SAPK . Bar graphs represent protein kinase activity relative to vector control. Also indicated (as % total stim) are the levels of SAPK activity relative to that seen in cells coexpressing DLK and v-Src only. Each set of experiments was repeated 3 times with similar results. possess 6 of 8 residues conserved in the CRIB/GBD; however, it is not clear that DLK contains this binding domain, since it possesses only 3 of 8 conserved residues (51). Nevertheless, DLK may associate with Rac1 and/or Cdc42Hs through an alternative binding domain or via an interaction with commonly associated molecules. Should DLK interact with Rac1 and/or Cdc42Hs, it is possible that the observed attenuation by K185A of V-12 Rac1-, V-12 Cdc42Hs-, and v-Src-induced activation of p46 SAPK is due to inhibition/sequestration of the bound small GTPases in a manner similar to that of PAK65-R.
In summary, the activation of p46 SAPK by DLK is not inhibited by dominant negative mutants of Rac1 and Cdc42Hs, PAK65-R, or PAK65-A, but is attenuated by MEKK1(K432M). Moreover, K185A can significantly attenuate the activation of SAPK by v-Src and constitutively active V-12 Rac1 and V-12 Cdc42Hs. Therefore, in the pathway leading from v-Src to SAPKs activation, our results suggest that DLK lies distal to Rac1 and/or Cdc42Hs but proximal to MEKK1.
Recent effort in several laboratories has focused on identifying the protein kinase or kinases that are responsible for linking the activation of Rac1 and Cdc42Hs to activation of the MEKK1-MKK4-SAPKs module. By homology to yeast, it has been presumed that the linking kinase is a STE20 homologue or a MAPKKKK. A large number of kinases bearing sequence similarity in their catalytic domain to STE20 have now been identified. These kinases can be divided into two subfamilies on the basis of structure. The first subfamily including Cla4, rat p65PAK, hPAK65, hPAK1, MST1, and MST2, are most closely related in overall structure to STE20 (25)(26)(27)(28)(29)(30)(31). Members of this first subfamily all contain CRIB/GBD domains in their NH 2terminals. To date, some PAKs have been shown to become activated by GTP-bound Rac1 and Cdc42Hs (25,26) and several have been shown to activate p46 SAPK (32)(33)(34). Taken together, these observations suggest that the PAKs participate in linking activation of Rac1 and Cdc42Hs to activation of the MEKK1-MKK4-SAPKs module. Importantly, it has not yet been demonstrated that MEKK1 serves as the direct substrate of these PAKs nor that the kinases interact directly in vivo.
This picture has been made more complex by the identification of a second subfamily of STE20-related kinases which includes germinal center kinase (GCK) and upstream kinase (49). These kinases are more distantly related to STE20; they contain a distinctly different modular structure and lack CRIB/ GBD binding domains. When overexpressed by transient transfection, GCK can also activate p46 SAPK via a MKK4-dependent pathway (49). Further mapping of the GCK-dependent SAPK pathway has not been reported.
The findings reported herein introduce a third class of pro-  2 g), or vector control (to 2 g). Cells were lysed after 48 h, and immunoprecipitated p46 SAPK was assayed for catalytic activity in vitro. Immunoblots from corresponding experiments were used to evaluate relative expression of immunoprecipitated p46 SAPK . Bar graphs represent protein kinase activity relative to vector control. Also indicated (as % total stim) are the levels of SAPK activity relative to that seen in cells coexpressing DLK and SAPK only (A and D), catalytically active MEKK1 and SAPK only (B), and V-12 Cdc42Hs and SAPK only (C). Each set of experiments was repeated 3 times with similar results.  1.2 g), or vector control (to 2 g). B, NIH3T3 tein kinase that serves as a proximal regulator of the SAPK pathway. Our results suggest that DLK lies in a signaling pathway between Rac1/Cdc42Hs and MEKK1. Given the number of already identified protein kinases that might participate in linking GTPase activation to activation of MEKK1 or SAPKs, the epistatic relationship between DLK, the Rho-like GTPases, MEKK1, and the components of the various STE20 homologue subfamilies will require careful investigation. Our present understanding is insufficient to allow an immediate reconciliation between previously reported observations pertaining to the PAKs described above and those reported herein regarding DLK. However, we speculate that DLK represents a component of a pathway independent of the PAKs that links GTPase activation with activation of the MEKK1-MKK4-SAPK module. This pathway is likely to be cell type-and possibly subcellular compartment-specific.