Mammalian mitogen-activated protein kinase pathways are regulated through formation of specific kinase-activator complexes.

Mammalian cells contain at least three signaling systems which are structurally related to the mitogen-activated protein kinase (MAPK) pathway. Growth factors acting through Ras primarily stimulate the Raf/MEK/MAPK cascade of protein kinases. In contrast, many stress-related signals such as heat shock, inflammatory cytokines, and hyperosmolarity induce the MEKK/SEK(MKK4)/SAPK(JNK) and/or the MKK3 or MKK6/p38hog pathways. Physiological agonists of these pathway types are either qualitatively or quantitatively distinct, suggesting few common proximal signaling elements, although past studies performed in vitro, or in cells using transient over-expression, reveal interaction between the components of all three pathways. These studies suggest a high degree of cross-talk apparently not seen in vivo. We have examined the possible molecular basis of the differing agonist profiles of these three MAPK pathways. We report preferential association between MAP kinases and their activators in eukaryotic cells. Furthermore, using the yeast 2-hybrid system, we show that association between these components can occur independent of additional eukaryotic proteins. We show that SAPK(JNK) or p38hog activation is specifically impaired by co-expression of cognate dominant negative MAP kinase kinase mutants, demonstrating functional specificity at this level. Further divergence and insulation of the stress pathways occurs proximal to the MAPK kinases since activation of the MAPK kinase kinase MEKK results in SAPK(JNK) activation but does not cause p38hog phosphorylation. Therefore, in intact cells, the three MAPK pathways may be independently regulated and their components show specificity in their interaction with cognate cascade members. The degree of intermolecular specificity suggests that mammalian MAPK signaling pathways may remain distinct without the need for specific scaffolding proteins to sequester components of individual pathways.

Genetic studies in yeast first revealed the existence of multiple distinct mitogen-activated protein kinase (MAPK) 1 related signal transduction pathways containing structurally similar protein kinase cascades mediating responses to mating factor, hyperosmolarity, and cell wall integrity (1). Pathways in mammals, which likely have different physiologic significance, have been discovered which utilize related protein kinase cascades (2)(3)(4). The archetypal MAPK pathway is activated in response to Ras-GTP loading as well as other processes such as phosphatidylinositol turnover (5)(6)(7). The pathway comprises a series of protein kinases such that activation of the Raf protooncogene causes phosphorylation and activation of MEK which, in turn, phosphorylates and activates the MAPKs, Erk1, and -2 (4,8). The targets of these kinases include other protein kinases and transcription factors such as p62 tcf (9 -11). While mitogens and growth factors commonly stimulate this MAPK pathway, cells respond to cellular stress agents by induction of two structurally related but distinct pathways (2,(12)(13)(14)(15)(16). Stress stimuli such as the inflammatory cytokines, tumor necrosis factor ␣, and interleukin-1, thermal shock, ischemia/ reperfusion, and hyperosmolarity cause activation of two recently described groups of MAPK-related proteins, the SAPK/ JNK family and p38 hog kinase (2,12,16,17). SAPK is phosphorylated by a MEK-like kinase termed SEK1 (MKK4, JNKK), while p38 hog is phosphorylated by the related kinases MKK3 and MKK6 (18 -20). SEK1 itself is phosphorylated and activated by MEKK (21) while MKK3 and MKK6 activators are being defined.
Activation of these three pathways is not mutually exclusive. For example, heat shock and tumor necrosis factor-␣ partially activate the ERK1/ERK2 cascade and some mitogens such as epidermal growth factor partially activate the SAPK pathway (2,12). The mitogenic and stress pathways are, however, discriminated by agonist profiles when efficiencies of activation by a broad range are compared (12). The functional selectivity of varying agonists may not be manifest in vitro or when signaling proteins are overexpressed in cells by transient transfection. For example, MEKK was first characterized as a protein kinase acting on MEK (22). At high levels, this enzyme does phosphorylate and activate MEK but at physiological levels MEKK specifically stimulates the SAPK pathway through SEK phosphorylation (21). Recently, SEK has been shown to phosphorylate p38 hog in vitro and in transfected COS cells raising the possibility that SAPK and p38 hog are co-regulated by this kinase (23). To address the discrepancy between agonist discrimination and apparent in vitro promiscuity, we have screened a variety of conditions for induction of these pathways and have probed for both physical association between the components in intact eukaryotic cells or in yeast cells, and for functional relationships using a spectrum of dominant negative MAP kinase kinase mutants. These results indicate that the three pathways operate independently in cells due to effective compartmentalization of their components via specific proteinprotein interactions.

EXPERIMENTAL PROCEDURES
Mutagenesis-Mammalian SEK was mutated at serines 204 and 207, the sites of activating phosphorylation, using the pSelect system (Promega) and the protocol supplied by the manufacturer. The alanine 220, leucine 224 mutant (SEK-AL) was generated with the oligonucleotide GCTTGTGGACGCTATTGCCAAGCTTAGAGATGC. MKK3 was mutated at serine 231 to alanine and at threonine 235 to leucine using the oligonucleotide TGGTGGACGCTGTGGCCAAGCTTATGGATGCCGG, producing MKK3-AL. Serine 207 and threonine 211 of MKK6 were mutated to alanine and leucine to produce MKK6-AL using the oligonucleotide ACTTGGTGGACGCTGTTGCTAAACTAATCGATGCAGGT-TGCAAACC.
Tissue Culture, Gene Expression, and Cell Stimulation-Calcium phosphate transfections were performed as described previously (4). RIF-1 murine fibroblasts (gift of Dr. G. Hahn, Stanford University) were transfected with pcDNA3 HA/SEK-AL and selected in 300 g/ml G418. Individual drug-resistant colonies were selected after 2 weeks of incubation and analyzed for expression of tagged proteins by immunoprecipitation followed by Western blot analysis using 12CA5 tissue culture supernatant. MKK6-AL or (wt)MKK6 were transiently transfected into COS cells with either pEBG SAPK or pEBG p38 hog using the same technique. COS cells, NIH 3T3 fibroblasts, or Jurkat lymphoblasts were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 10% fetal calf serum (Life Technologies, Inc.). On the day prior to stimulation cells were serum starved for 16 h in the same medium containing 2.0% fetal calf serum. Stimulation, when performed, was with a 10-min exposure to germicidal UV light (254 nm (General Electric)) or 30 min exposures to 500 g/ml anisomycin, 0.5 mM sodium arsenite, 400 mM sorbitol, 10 nM phorbol 12-myristate 13-acetate (all from Sigma) or to heat shock for 30 min at 42.0°C.
Antibody Generation, Acrylamide Electrophoresis, and Western Blotting-Anti-p38 hog antibodies were raised in rabbits immunized with 200 g of prokaryotically expressed GST/p38 hog which was injected subcutaneously with Freund's complete adjuvant. Rabbits were boosted twice at 4-week intervals with 100 g of protein and Freund's incomplete adjuvant. Specificity for p38 hog on Western analysis and in immunoprecipitation was demonstrated by successful competition with the immunizing protein. Rabbit SAPK␣1 antisera was raised to prokaryotic thrombin-cleaved GST fusion protein as described (2). For analysis of proteins, electrophoresis was performed through an 8% polyacrylamide gel and dried onto 3MM paper for subsequent autoradiography, or electrophoretically transferred onto polyvinylidene difluoride membranes (Hybond N, Amersham) for Western analysis. Western immunoblotting with anti-GST and anti-HA antibodies was performed using a blocking buffer of 5% non-fat dried milk, phosphate-buffered saline, and 0.1% Nonidet P-40 and horseradish peroxidase-conjugated goat anti-mouse or anti-rabbit secondary antibodies (Sigma). Immobilized immune complexes were detected by Enhanced Chemiluminescence (ECL, Amersham).
In Vitro Kinase Reaction-In vitro kinase reactions were performed using immunoprecipitated bead immobilized native p38 hog or SAPK or affinity purified bead immobilized GST tagged versions of the same proteins. Each kinase reaction was performed in a final volume of 50 l containing 25 l of 2 ϫ kinase buffer (100 M [␥-32 P]ATP, 10 mM MgCl, 50 mM Tris-HCl (pH 7.5), 1 mM EDTA), 5 l of myelin basic protein (1 mg/ml) or c-Jun (1 mg/ml) dissolved in distilled water and 20 l of immobilized kinase slurry. Samples were incubated at 37°C for 30 min. Reactions were terminated by the addition of 50 l of 2 ϫ SDS protein loading buffer followed by boiling for 10 min. Products were separated on an 8% polyacrylamide protein gel followed by transfer to 3MM paper and autoradiography.
Image Analysis-Radiolabeled proteins were detected post-electrophoresis by PhosphorImager and the ImageQuant analysis software (Molecular Dynamics Inc., Sunnyvale CA).
Yeast 2-Hybrid Analysis-The complete coding sequences of SAPK␣, p38 hog , or p44 erk were expressed as GAL4 DNA-binding domain fusions in the plasmid pAS1 (Clonetech Inc.) while cDNAs coding SEK1, MEK, MKK3, and MKK6 were expressed as GAL4 trans-activating domain fusions using the vector pACTII (Clonetech Inc.). Expression of each fusion protein was confirmed by plasmid transfection into yeast strain HF7c followed by Western analysis of nutritionally selected colonies using antibodies specific for either the GAL4 DNA binding or transactivating domain. Co-transformation of HF7c with each pAS1 construct and each of the pACTII constructs was then performed. Control cotransformations of the same yeast strain was performed using each of the pAS1 kinase constructs with empty pACTII. Co-transformed yeast were maintained prior to testing on minimal growth medium lacking amino acids tryptophan and leucine. Filter lift assays for ␤-galactosidase production, as described by the manufacturer, were performed for each transfected yeast strain after growth on tryptophan/leucine-deficient medium. Yeast growth and the pigmentation of lifted filters were documented by photography. All experiments were done at least 5 times.

RESULTS
Differential Activation of MAPK Homologues-By assaying the activation state of the ERKs and SAPKs we have previously shown that agonists of these pathways are largely distinct (19). These studies did not evaluate the activation of p38 hog , which has been shown by others to be stimulated by agonists similar to those activating the SAPKs (16,17), suggesting that the p38 hog pathway may share upstream components with the SAPK pathway. To test this we first directly compared activation of SAPK and p38 hog which were immunoprecipitated after stimulation of either NIH 3T3 cells or Jurkat lymphocytes. Activity of immunoprecipitated SAPK was assayed using its physiologic substrate c-Jun, while the artificial substrate myelin basic protein was used to assess the activation state of immunoprecipitated p38 hog . As shown in Fig. 1, activation of each of these kinases was both agonist and cell type dependent. p38 hog activity was stimulated only by UV irradiation and hyperosmolar shock (sorbitol) in Jurkat cells and by these agonists and phorbol 12-myristate 13-acetate, heat shock, and sodium arsenite in NIH 3T3 cells. In contrast, SAPK activity was stimulated by hyperosmolarity, heat, and anisomycin in Jurkat cells and by all the agonists tested in NIH 3T3 cells. Of note, anisomycin stimulated SAPK but not p38 hog in both cell types. Sodium arsenite stimulated p38 hog in both cell types but stimulated SAPK only in NIH 3T3 cells while UV irradiation stimulated SAPK only in NIH 3T3 cells but stimulated p38 hog in both cell types tested.

MEKK Overexpression Does Not Result in in Vivo
Phosphorylation of p38 hog -Since many of the agonists which stimulate the SAPK pathway stimulate p38 hog , these enzymes may share common upstream activators. SEK1, which has been shown to activate SAPK in vivo, phosphorylates both SAPK and p38 hog in vitro (23), implying that SEK1 may be common to both pathways in vivo. MEKK, the physiologic activator of SEK1, would then be expected to activate both SAPK and p38 hog by phosphorylation after transient overexpression. Phosphorylation of both by SEK1 would explain the similarity of agonist activation profiles of these pathways. pEBG/p38 hog was transiently transfected alone or with pMT2/MEKK into COS cells. Samples expressing GST/p38 hog were either left unstimulated or incubated with sorbitol, to demonstrate the ability of p38 hog to become phosphorylated under these experimental circumstances. GST/p38 hog was purified on glutathione beads from each cell sample. p38 hog phosphorylation on tyrosine, indicative of activation, was assessed by anti-phosphotyrosine immunoblotting of the purified products. Fig. 2 shows that while sorbitol stimulation results in the expected tyrosine phosphorylation of p38 hog , co-expression of MEKK had no stimulating effect. To directly compare the enzymatic activity of either SAPK or p38 hog after MEKK overexpression, pEBG/SAPK or pEBG/p38 hog were transiently expressed in COS cells alone or with pMT2/MEKK. GST/SAPK or GST/p38 hog were affinity purified on glutathione beads in each case and enzymatic activity was assessed in vitro against c-Jun for SAPK or myelin basic protein for p38 hog . Fig. 3 shows that SAPK, as expected, was stimulated by MEKK co-expression to a level equivalent to that induced by hyperosmolarity. p38 hog , however, was not stimulated by MEKK co-expression but only by the hyperosmolar stimulus.
SEK1 Co-precipitates from Cells with SAPK But Not with p38 hog -MEKK, a strong activator of SEK1, does not appear to cause enzymatic activation or tyrosine phosphorylation of p38 hog . This observation, which is in variance with the reported ability of SEK1 to phosphorylate p38 hog in vitro, suggests that SEK1 does not activate or associate with p38 hog in cells. We have previously observed a tight in vivo interaction between SEK1 and SAPK suggesting physiologic relevance (19). These observations were extended to evaluate potential in vivo association between SEK1 and p38 hog . pEBG/SAPK or pEBG/ p38 hog were transiently expressed in COS cells with pMT2/HA-SEK. The GST proteins were affinity purified, washed on glutathione beads, and divided for analysis. Relative expression of SAPK and p38 hog was determined by anti-GST immunoblotting using half of each sample (Fig. 4A). GST/SAPK appears to be under expressed relative to GST/p38 hog in this typical experiment. Associating HA-SEK1 was detected in the remaining half of each sample by immunoblotting with the 12CA5 monoclonal which is specific for the HA tag (Fig. 4B). Despite the underexpression of SAPK, under these experimental conditions, SEK1 binds to SAPK but not to p38 hog . This demonstrates specificity of SEK1 association with SAPK and suggests that p38 hog is unlikely to be activated by SEK1 in vivo although in vitro phosphorylation has been demonstrated. In a separate control experiment, pMT2/HA-MEK was expressed with pEBG/SAPK, pEBG/p38 hog , or pEBG/MAPK. Similarly, glutathione-Sepharose associating proteins were purified and analyzed by Western immunoblotting using the 12CA5 monoclonal antibody. Fig. 4C shows, as expected, that HA-MEK associates only with GST/MAPK but not with either GST/ SAPK or GST/p38 hog , thus confirming the specificity of coassociation observed under these experimental conditions. SEK-AL Expression Impairs SAPK But Not p38 hog Activation-SEK is activated by MEKK by phosphorylation on serine 204 and serine 207. We have mutated these residues to alanine and leucine, generating a dominantly active inhibitor of wild type SEK1 (19). Stable transfectants expressing SEK-AL in the murine fibrosarcoma cell line RIF-1 (24) were generated by G418 drug selection for pcDNA3/HA-SEK-AL plasmid incorporation. Two such lines, RIF/SEK-AL18 and RIF/SEK-AL19 were selected for further study. Samples of vector transfected RIF-1 and of the two SEK-AL expressing lines were left unstimulated or treated with either sorbitol or anisomycin. Endogenous SAPK or p38 hog were immunoprecipitated and activity of each kinase was directly evaluated in vitro. Fig. 5 demonstrates that while endogenous SAPK activation by ani- FIG. 1. p38hog and SAPK from two cell lines demonstrate distinct patterns of activation in response to stress agonists. NIH 3T3 cells or Jurkat lymphocytes were incubated with phorbol 12-myristate 13-acetate, sorbitol, sodium arsenite, or anisomycin or exposed to either UV irradiation or heat shock as described. Native p38 hog or SAPK were immunoprecipitated using specific rabbit antisera and purified on protein A-Sepharose beads. In vitro kinase reactions were performed to assess activity for each immunoprecipitate using the substrates myelin basic protein for p38 hog and GST-Jun for SAPK. The results were analyzed by acrylamide gel electrophoresis followed by PhosphorImaging. Densitometric determinations in each lane are provided and allow comparison within each panel.
FIG. 2. MEKK overexpression does not result in p38hog phosphorylation. COS cells were transiently co-transfected with both pEBG/p38 hog and pMT2/MEKK expression plasmids. Similar cell samples were transfected with pEBG/p38 hog alone and left unstimulated or were treated with 400 mM sorbitol for 30 min. GST/p38 hog was purified from each set of cells on glutathione-conjugated Sepharose. Tyrosine phosphorylation of the product, a measure of activation, was analyzed by acrylamide gel electrophoresis followed by anti-phosphotyrosine Western blotting. Data are shown in duplicate.

FIG. 3. MEKK overexpression enhances the activity of SAPK
but not of p38hog. pMT2/MEKK and either pEBG/p38 hog or pEBG/ SAPK were transiently transfected into COS cells and either left unstimulated or stimulated with 400 mM sorbitol for 30 min. Control COS cells were transfected with either pEBG/p38 hog or pEBG/SAPK alone and left unstimulated. GST/p38 hog or GST/SAPK were purified on glutathione-conjugated Sepharose. In vitro kinase reactions were performed to assess activity for each immunoprecipitate using the substrates myelin basic protein for p38 hog and GST-Jun for SAPK. The results were analyzed by acrylamide gel electrophoresis followed by autoradiography.
somycin is impaired in the SEK-AL expressing cell lines, no impairment in the activation of p38 hog is observed. The degree of substrate phosphorylation in Fig. 5A, an indication of activation, is represented graphically by densitometry in Fig. 5B.
MKK3-AL and MKK6-AL Expression Impair the Activation of p38 hog But Not SAPK-MKK3 and MKK6 can phosphorylate p38 hog in vitro and have been proposed as potential physiologic p38 hog activators (18 -20). MKK3 is regulated by phosphorylation on serine 189 and threonine 193, while MKK6 is phosphorylated on serine 207 and threonine 211, sites that are conserved among the other MAPK kinases SEK1 and MEK. Sites of serine phosphorylation were mutated to alanine and threo-nine phosphorylation sites were mutated to leucine in MKK3 and MKK6 using site-directed mutagenesis, forming MKK3-AL and MKK6-AL. MKK3-AL, MKK6-AL, and their wild type forms were HA epitope-tagged and each inserted into the vector pMT2. COS cells were transiently transfected with pEBG/ p38 hog or pEBG/SAPK and with empty pMT2 vector, pMT2/ MKK3-AL, pMT2/MKK3-WT, pMT2/MKK6-AL, or pMT2/ MKK6-AL. Samples of transfected cells were either unstimulated or incubated with anisomycin or sorbitol. GST/ p38 hog or GST/SAPK were affinity purified on glutathione beads and enzymatic activity was evaluated by in vitro kinase reaction. Figs. 6 and 7 show that GST/SAPK activity was markedly stimulated by anisomycin or by sorbitol which was not changed by the co-expression of MKK3-WT, MKK3-AL, MKK6-WT, or MKK6-AL. GST/p38 hog activation by sorbitol was observed with co-expression of MKK3-WT and with MKK6-WT but was totally eliminated by the co-expression of MKK3-AL or MKK6-AL, which acted as dominantly acting inhibitors.

MAP Kinases Associated with MAPK Kinases Independent of Additional Eukaryotic Proteins-KSS-1/FUS3
, which is a MAPK homologue in Schizosaccharomyces cerevisiae, complexes with Ste20 and Ste11, components of the sterile signaling pathway, by binding to the common scaffolding protein, Ste5. Although scaffolding proteins have not been described as elements of MAP kinase signaling pathways in eukaryotic cells, association of MAP kinases and their activators could occur through similar mechanisms. To specifically address this possibility, we expressed SAPK and p38 hog in yeast as GAL4 transcription factor DNA-binding domain fusions with the kinase activators SEK1, MKK3, or MKK6 expressed as GAL4 DNA transactivating domain fusions. Since it is unlikely that yeast would contain scaffolding elements able to bind eukaryotic proteins, interaction of these components in a yeast 2-hybrid assay would provide evidence for direct interaction between co-expressed molecules. As demonstrated by filter lift assay for ␤-galactosidase activity, an indicator of inter-molecular association between the two GAL4 component fusions, SAPK associates with SEK1 but not with MKK3 or MKK6. p38 hog associates with all three potential activators consistent with previous data demonstrating specific activation by MKK3 and MKK6 and in vitro activation by SEK1 (Fig. 8).  2 and 4) or pEBG-p38 hog (lanes  1 and 3). Samples from lanes 1 and 2 were from untreated cells while those in lanes 3 and 4 were from sorbitol-treated cells. In A, the lysates were incubated with glutathione beads, washed, and the bead-associated proteins immunoblotted with anti-GST antibodies to determine the level of expression of the GST fusion proteins encoded by the pEBG plasmids. In B, the lysates were directly immunoblotted with the monoclonal antibody 12CA5, specific for the HA epitope to detect associating tagged SEK1. In panel C, COS cells were co-transfected with pMT2/ HA-MEK together with pEBG-SAPK (lane 1), pEBG-p38 hog (lane 2), or pEBG-MAPK (lane 3). As in panel A, the GST fusion proteins were isolated and associating proteins were probed with 12CA5 antibody to detect the presence of HA-MEK. Detection was by enhanced chemiluminescence (ECL).
FIG. 5. Expression of the dominantly active inhibitor of SAPK activation, SEK-AL, does not inhibit p38hog activation in response to hyperosmolar shock. Two independently derived RIF-1 murine fibrosarcoma cell line clones expressing SEK-AL were derived utilizing G418 selection. A, SAPK activation in response to the stress agonist anisomycin and p38 hog activation after sorbitol hyperosmolar shock was assessed by immunoprecipitating each native kinase followed by in vitro kinase reactions to assess activity as described above. Results were assessed by PhosphorImager and quantitated using the ImageQuant software package. B, graphical representation of p38 hog and SAPK activation in each cell line and in vector transfected G418resistant control cells.
FIG. 6. MKK3-AL expression in COS cells impairs hyperosmolarity induced activation of p38hog but has no effect on the activation of SAPK. pEBG/SAPK or pEBG/p38 hog were transiently transfected into COS cells with pMT2/MKK3-AL, pMT2/MKK3-WT, or the empty pMT2 vector. Transfected samples were either unstimulated or treated with sorbitol or anisomycin (aniso) for 30 min. GST/p38 hog or GST/SAPK were purified on glutathione-conjugated Sepharose. In vitro kinase reactions were performed to assess activity for each immunoprecipitate using the substrates myelin basic protein for p38 hog and GST-Jun for SAPK. The results were analyzed by acrylamide gel electrophoresis followed by autoradiography. A, GST/p38 hog activity was stimulated by sorbitol in cells expressing wild type MKK3 but was totally inhibited by co-expression of MKK3-AL. B, in contrast, SAPK activation by either anisomycin or sorbitol was unaffected by co-expression of either MKK3-WT or MKK3-AL.

DISCUSSION
Lower eukaryotic cells have MAPK related signaling cascades which control diverse response to extracelluar stimulation. In the budding yeast S. cerevisiae at least 5 physiologically distinct MAPK-related signaling pathways are known to exist (25) including two with homologues found in mammalian cells: the Fus3/Kss1 pathway which controls the mating response to extracelluar pheromones (26,27), and the HOG-1 pathway which results in the synthesis of intracellular glycerol to adapt to hyperosmolar stress (1). These kinases have highly similar sequences and are part of homologous activational pathways. To segregate components of distinct signaling cascades, members must have strict specificity for one another or alternatively, be physically compartmentalized. In the yeast mating factor cascade, Ste5 functions to bind together and compartmentalize three protein kinase components, Ste11, Ste7, and Fus3/Kss1 which have sequence similarity to mammalian MEKK, MEK, and MAPK, respectively (26,28). Ste5 may therefore act as an insulator for this pathway by chaperoning these proteins and preventing association with other transduction cascades. Since no Ste5 homologues have been found in mammalian cells to date, similar signaling specificity might be provided by highly specific direct interactions between the mammalian protein kinases such as we have observed.
In mammalian cells three highly homologous MAPK signal transduction pathways have been identified, the ERK mitogenic cascade and the 2 stress activated cascades resulting in SAPK or p38 hog activation. These kinase pathways fulfil distinct roles within cells and must be individually insulated or integrated in a coordinated way to maintain the specificity and diversity of extracelluar signals. Since cells contain these and hundreds of other protein kinases which are often structurally highly related, there must be mechanisms to allow specificity. Here, we provide evidence for direct in vivo protein-protein interactions occurring between cognate components of the two most closely related of the three known mammalian MAPK related pathways, the SAPK and the p38 hog pathways. The importance of these interactions in defining specificity is illustrated by the lack of specificity between the enzymes when mixed independently in vitro (22).
We have demonstrated that SAPK and p38 hog are activated with both quantitative and qualitative differences after a variety of stress stimuli in both Jurkat lymphocytes and in NIH 3T3 cells. Such differential activation must reflect a divergence in activating pathways immediately upstream of these kinases. We show here that SEK1, a kinase which physiologically activates and associates with SAPK, is unable to bind to p38 hog in vivo suggesting its independence from p38 hog activation. MEKK, a kinase upstream of SEK1, while able to activate SAPK, is unable to activate p38 hog even when overexpressed, suggesting a divergence of the SAPK and p38 hog pathways at the level of MEKK. Furthermore, SEK-AL, a dominantly acting inhibitor of SAPK, while efficiently inhibiting SAPK activation, is unable to affect p38 hog activation after hyperosmolar stress in COS cells. These observations imply that p38 hog is not activated by SEK1 after stress stimuli in vivo, but becomes activated by a distinct kinase acting immediately upstream.
MKK3 and MKK6 have been identified as mammalian homologues of SEK1 and likely induce activation of MAPK family kinases in response to extracelluar stimuli (18 -20). Both molecules can activate p38 hog in vitro suggesting possible physiologic function. We have demonstrated that dominantly acting inhibitors of MKK3 and MKK6, MKK3-AL, and MKK6-AL respectively, prevent p38 hog activation in response to hyperosmolar stress in COS cells but have no effect on SAPK activation. This is consistent with a role for each of these molecules as specific p38 hog activators in vivo and demonstrates that the SAPK and the p38 hog signaling pathways may be segregated in cells by specific intramolecular associations which have not been observed by in vitro testing.
Although one S. cerevisiae MAPK signaling pathway is integrated through the assembly of individual components on scaffolding proteins, to date proteins with similar function have not been identified in eukaryotic cells. We have demonstrated coimmunoprecipitation of members of cognate eukaryotic MAPK FIG. 7. MKK6-AL expression in COS cells prevents activation of p38hog after hyperosmoloar stress but does not affect SAPK activation. pEBG/SAPK or pEBG/p38 hog were transiently transfected into COS cells with pMT3/MKK6-AL, pMT3/MKK3-WT, or the empty pMT3 vector and stimulated with sorbitol for 30 min. Control cell samples, transfected with empty pMT3 and either pEBG/SAPK or pEBG/p38 hog , were left unstimulated. GST/p38 hog or GST/SAPK were purified on glutathione-conjugated Sepharose. In vitro kinase reactions were performed to assess activity for each immunoprecipitate using the substrates myelin basic protein for p38 hog and GST-Jun for SAPK. The results were analyzed by acrylamide gel electrophoresis followed by autoradiography. Densitometry values for each gel are graphically represented in each panel. GST/p38 hog stimulation by sorbitol was inhibited by coexpression of MKK6-AL (A) while SAPK stimulation was unaffected (B).

FIG. 8. Association of eukaryotic MAP kinases with MAPK kinases in yeast.
Fusion DNA constructs of pAS1 GAL4 DNA-binding domains and mammalian p38 hog or SAPK were individually transfected into HF7c yeast with pACTII GAL4 DNA transactivating domain fusion constructs of MKK3, MKK6, or SEK-1. Yeast were selected on tryptophan/leucine-deficient medium, and expression of each fusion partner in individual yeast clones was confirmed by Western analysis as described. A, yeast colonies were plated uniformly on double minus medium and grown for 48 h. B, filter lift assay for ␤-galactosidase production demonstrating the association of SAPK with SEK-1 and the association of p38 hog with all three MAPK kinases. signaling pathways from intact cells. These observations do not exclude the involvement of co-precipitating eukaryotic scaffolding proteins which may specifically assemble components of individual MAPK pathways. The yeast 2-hybrid assay has been used to demonstrate specific interaction between eukaryotic proteins when expressed as fusions with DNA transcription factor functional domains. To address the potential role of eukaryotic scaffolding proteins, individual MAPKs were expressed as GAL4 transcription factor DNA-binding domain fusions while their activators were expressed as GAL4 transactivating fusions. Our data demonstrate that interaction occurs between the MAPK kinases and activator molecules in yeast without the participation of additional eukaryotic molecules. These observations, while not excluding the substitution of endogenous yeast molecules for eukaryotic scaffolding proteins, makes the participation of such molecules in eukaryotic MAPK signaling pathways less likely.
Why are two highly related stress kinases, triggered by similar agonists, activated through seemingly distinct mechanisms? The physiologic consequence of SAPK activation may be quite distinct from that of p38 hog . MEKK induction in fibroblasts results in growth arrest suggesting that SAPK may lie on a growth inhibitory pathway, triggered by cellular stress and damage (21). This pathway may have evolved to allow damaged cells to repair themselves prior to division or to induce apoptotic death in severely damaged cells as suggested by recent work in NIH 3T3 and REF52 fibroblasts (29). Apoptosis as a consequence of SAPK activation has been directly demonstrated in the neuroectodermal cell line PC12 (30). After serum withdrawal, these cells, when differentiated by nerve growth factor, will undergo apoptosis which can be inhibited by SAPK blockade. We have also shown that SAPK inhibition prevents cell death in murine fibrosarcoma cells in response to heat shock and to the cytotoxic chemotherapeutic drug cis-platinum (31) which also is consistent with SAPK's role as a mediator of programmed cell death. While the consequences of specific activation of p38 hog are unknown, this protein kinase is specifically inhibited by pyridinyl-imidazole compounds, which act as powerful anti-inflammatory agents in vivo (32). p38 hog activation may therefore generate protective mechanisms in response to cellular stresses. This is consistent with data showing that p38 hog , as an activator MAP kinase-associated protein kinase II (MAPKKAP kinase II), phosphorylates, and activates small molecular weight heat shock proteins providing protection from heat shock (33)(34)(35)(36). A subtle balance may thus be struck between repair and growth arrest after stress stimuli that may be mediated differentially through each of these two pathways.
In vitro promiscuity of protein kinases can lead to artifactual connections between signaling elements. We have shown independence between SAPK and p38 hog phosphorylation implying a different physiologic role for each kinase and predict further downstream divergence of these pathways.