Tumor Necrosis Factor Signaling to Stress-activated Protein Kinase (SAPK)/Jun NH2-terminal Kinase (JNK) and p38

Tumor necrosis factor (TNF) elicits a diverse array of inflammatory responses through engagement of its type-1 receptor (TNFR1). Many of these responses require de novogene expression mediated by the activator protein-1 (AP-1) transcription factor. We investigated the mechanism by which TNFR1 recruits the stress-activated protein kinases (SAPKs) and the p38s, two mitogen-activated protein kinase (MAPK) families that together regulate AP-1. We show that the human SPS1 homologue germinal center kinase (GCK) can interact in vivo with the TNFR1 signal transducer TNFR-associated factor-2 (TRAF2) and with MAPK/ERK kinase kinase 1 (MEKK1), a MAPK kinase kinase (MAPKKK) upstream of the SAPKs, thereby coupling TRAF2 to the SAPKs. Receptor interacting protein (RIP) is a second TNFR signal transducer which can bind TRAF2. We show that RIP activates both p38 and SAPK; and that TRAF2 activation of p38 requires RIP. We also demonstrate that the RIP noncatalytic intermediate domain associates in vivo with an endogenous MAPKKK that can activate the p38 pathway in vitro. Thus, TRAF2 initiates SAPK and p38 activation by binding two proximal protein kinases: GCK and RIP. GCK and RIP, in turn, signal by binding MAPKKKs upstream of the SAPKs and p38s.

Tumor necrosis factor (TNF) elicits a diverse array of inflammatory responses through engagement of its type-1 receptor (TNFR1). Many of these responses require de novo gene expression mediated by the activator protein-1 (AP-1) transcription factor. We investigated the mechanism by which TNFR1 recruits the stress-activated protein kinases (SAPKs) and the p38s, two mitogen-activated protein kinase (MAPK) families that together regulate AP-1. We show that the human SPS1 homologue germinal center kinase (GCK) can interact in vivo with the TNFR1 signal transducer TNFR-associated factor-2 (TRAF2) and with MAPK/ERK kinase kinase 1 (MEKK1), a MAPK kinase kinase (MAPKKK) upstream of the SAPKs, thereby coupling TRAF2 to the SAPKs. Receptor interacting protein (RIP) is a second TNFR signal transducer which can bind TRAF2. We show that RIP activates both p38 and SAPK; and that TRAF2 activation of p38 requires RIP. We also demonstrate that the RIP noncatalytic intermediate domain associates in vivo with an endogenous MAPKKK that can activate the p38 pathway in vitro. Thus, TRAF2 initiates SAPK and p38 activation by binding two proximal protein kinases: GCK and RIP. GCK and RIP, in turn, signal by binding MAPKKKs upstream of the SAPKs and p38s.
Tumor necrosis factor (TNF) 1 is a multifunctional cytokine that induces a broad spectrum of responses, both at the cellular and organismal level. These include fever, shock, cachexia, tumor necrosis, leukocyte adhesion, and extravasation, induction of other cytokines, cell growth, and apoptosis (1). TNF is important in the regulation of normal immune development where it governs in part the ablation of autoreactive lymphocytes during immune cell selection. In addition, TNF has been implicated in the pathogenesis of noninsulin-dependent diabetes, as well as acute and chronic inflammatory diseases such as endotoxic shock and arthritis. TNF exerts its effects by binding to one of two receptors, the 55-kDa TNF receptor (TNFR)-1/ CD120a or the 75-kDa TNFR2/CD120b (1,2). TNF binding results in receptor trimerization and the consequent initiation of signal transduction (2). The protein recruitment model for TNF signaling posits that ligand-induced TNFR trimerization results in the binding, to the TNFR intracellular extensions, of signal transducing proteins which then relay signals to downstream effectors. The intracellular extension of TNFR1 contains a death domain. Death domains mediate homotypic and heterotypic proteinprotein interactions (2,3) and are critical for nucleating receptor-effector complexes and implementing several signaling programs including apoptosis (2,3). Among the first death domain-containing proteins shown to bind TNFR1 was TNFRassociated death domain protein (TRADD). The binding of TRADD and TNFR1 requires the death domains of both polypeptides (4). TRADD can also bind two additional signal transducers: TNFR-associated factor-2 (TRAF2), a member of the TRAF family of signal transducers, and receptor interacting protein (RIP), a death domain-containing Ser/Thr kinase. RIP can also bind TRAF2 and, accordingly, TNF treatment is thought to result in the formation of a TRADD⅐RIP⅐TRAF2 complex (5)(6)(7)(8)(9)(10)(11)(12)(13).
The activation of gene expression is an important consequence of TNFR1 engagement and is essential for many of the biological responses to TNF. TNF can recruit the multimeric transcription factors activator protein-1 (AP-1) and nuclear factor B (NF-B) (14,15). TNF recruitment of AP-1 is pivotal to the inflammatory response. AP-1 is a heterodimer consisting of the c-Jun transcription factor and either a member of the Fos or activating transcription factor (ATF) family of transcription factors (16). Upon activation, AP-1 binds to cis elements in the genes for cytokines such as interleukin-2 and TNF itself. In addition, AP-1 is necessary for the expression of inflammatory proteases such as collagenase, and cell surface adhesion molecules such as E-selectin, which promote leukocyte adhesion and extravasation (16,17).
TNF can activate two subfamilies of the mitogen-activated protein kinase (MAPK) family of Ser/Thr kinases that together are largely responsible for the regulation of AP-1 in response to inflammatory stimuli (18 -20): the stress-activated protein kinases (SAPKs, also referred to as Jun NH 2 -terminal kinases, JNKs (18,21)) and the p38s. The SAPKs and p38s regulate AP-1 both by directly phosphorylating AP-1 components, and through the phosphorylation of transcription factors that participate in regulating the expression of c-jun and c-fos (18 -25).
Mammalian protein kinases homologous to S. cerevisiae SPS1 have been implicated in the selective activation of the SAPKs, possibly through the direct recruitment of MAPKKKs (40 -45). Germinal center kinase (GCK) was the first mammalian SPS1 shown to activate the SAPK pathway (40). Endogenous GCK is activated by TNF (40). Subsequently, four additional SPS1 homologues were shown to activate co-expressed SAPK. None of these can activate p38 or ERK1/2 in vivo (40 -45).
Despite the clear cut activation of AP-1, the SAPKs, p38s, and GCK by TNF, and the recent finding that TRAF2 is required for TNF activation of SAPK (46 -49), the molecular mechanisms coupling the TNFR signaling complex to SAPK and p38 have not been well characterized. We report herein that the GCK COOH-terminal regulatory domain can bind both TRAF2 and MEKK1 and that MEKK1 is a likely physiologic target of GCK. The TRAF2-GCK and GCK-MEKK1 interactions effectively couple one SAPK regulatory pathway to TNFR1. We also show that RIP can activate both the SAPK and p38 pathways in vivo; and our results indicate that RIP is required for TRAF2 activation of p38. The RIP intermediate domain is both necessary and sufficient for SAPK and p38 activation; and we demonstrate that the RIP intermediate domain, when overexpressed, can associate in vivo with an endogenous MAPKKK activity. This complex can be employed in assays to reconstitute the p38 pathway in vitro. We conclude from these results that GCK and RIP are proximal components in redundant, bifurcating mechanisms for TRAF2-mediated SAPK activation; by contrast, RIP is a dominant effector for TRAF2 activation of p38. In addition, we propose that both GCK and RIP elicit SAPK and p38 activation by binding, and possibly triggering activation of MAPKKKs.

EXPERIMENTAL PROCEDURES
Plasmids and Constructs-We used the following vectors: pEBG, which expresses a GST-tagged polypeptide (27), pCMV5, which expresses an M2-FLAG-tagged polypeptide, pMT3, which expresses an HA-tagged polypeptide, pCDM12, which expresses a Myc-tagged polypeptide and pEBV, which expresses an untagged polypeptide. pMT3-SAPK-p46␤1 and p38␣ have been described (18,40). GCK and MEKK1 constructs were prepared by polymerase chain reaction amplification and cloning according to standard methods (50). RIP constructs were prepared as described (51). Human TRAF2 was amplified by polymerase chain reaction from human T cell cDNA.
Kinase Assays and in Vivo GCK or RIP Activation of SAPK and p38 -293 cells were cultivated in 10-cm dishes and, as shown in the figures, were co-transfected (by the CaPO 4 method) with either 0.3 g of pCMV5-GCK, 0.3 g of pEBG-GCK (unless indicated), 7 g of the indicated pCDM12-Myc-RIP or 7 g of pEBV-TRAF2 and 1 g of either pMT3-SAPK-p46␤1 or p38␣. As necessary, transfected DNA levels were balanced with empty plasmid. After 18 h, SAPK and p38 were immunoprecipitated with anti-HA and assayed, respectively, for c-Jun or ATF2 kinase activity as described (18). GCK was assayed as described (38,52). For SEK1 phosphorylation assays, cells were transfected with 5 g of pEBG-GCK-CTD or pEBG and either 5 g of pCMV5-MEKK1 or pCMV5. GST polypeptides were isolated as described (27). GST-SEK1-KR was purified from transfected 293 cells as described below for MKK6 and p38. Phosphorylation of GST-SEK1-KR was performed as in Yan et al. (32). For dominant inhibitory experiments, 3 g of pCMV5-MEKK1 (817-1340) were co-transfected along with 0.3 g of pEBG-GCK and 1 g of pMT3 SAPK.
Coimmunoprecipitation Assays-293 cells were transfected by the CaPO 4 method. Unless indicated, 1-5 g of plasmid was used. As necessary, transfected DNA levels were balanced with empty plasmid. Immunoprecipitations and GST pulldowns were performed as described (18,27) with the following modifications. Lysis buffer was 20 mM Tris, pH 7.4, 2 mM EGTA, 10 mM MgCl 2 , 0.1% (v/v) ␤-mercaptoethanol, 1% (w/v) Triton X-100, 100 M phenylmethylsulfonyl fluoride, 10 kallekrein inhibiting units/ml aprotinin, 2 M leupeptin, 2 M pepstatin. Immunoprecipitates were washed twice with lysis buffer, twice with high stringency wash buffer (lysis buffer prepared with 0.1% (w/v) Triton X-100 and containing 1 M LiCl) and twice with wash buffer (no LiCl). Immunoblotting was performed using the enhanced chemiluminescence method (Amersham) according to the manufacturer's instructions. Anti-FLAG antibody was from Kodak, anti-GST and MEKK1 antibodies were from Upstate Biotechnology, anti-TRAF2 antibody was from Santa Cruz.
For in vitro binding of GCK to MEKK1, 293 cells were transfected with 5 g of pCMV5-M2-FLAG-MEKK1 (817-1221). After 20 h, the MEKK1 was immunoprecipitated and washed under stringent conditions as described above. To the beads were added 10 ng of GST or GST-GCK which had been purified from transfected cell extracts as described below. As controls, mock immunoprecipitations were prepared using extracts of cells transfected with empty vector. The GCK was allowed to incubate with the MEKK1 at which time the beads were washed under stringent conditions as described above. To the beads were added kinase assay buffer (20 mM Tris, pH 7.4, 1 mM EGTA, 1 mM dithiothreitol, 0.1% (w/v) Triton X-100) containing MgCl 2 (10 mM) and [␥-32 P]ATP (100 M). Autophosphorylation/phosphorylation was allowed to proceed for 20 min at 30°C. For mock immunoprecipitations, the supernatants containing GST-GCK were removed and allowed to autophosphorylate as above.
Purification of GST-GCK, -SEK1-KR, -MKK6, and -p38 from Transfected Cells-GST, GST-p38, -SEK1-K129R, and -MKK6 were purified from 7-10 plates of 293 cells transfected with the relevant pEBG constructs. GST-GCK was purified from 20 plates of transfected cells. The purification protocol was the same for all four proteins. Cells were lysed by Dounce homogenization in 5 ml of lysis buffer (20 mM Hepes, pH 7.4, 2 mM EGTA, 1 mM dithiothreitol, 250 mM sucrose, 200 M phenylmethylsulfonyl fluoride, 2 M pepstatin, 10 kallekrein inhibiting units of aprotinin). Lysates were cleared by centrifugation (100,000 ϫ g, 30 min) and Triton X-100 (0.1% w/v) was added to the supernatants. Supernatants were then loaded onto 250-l glutathione-agarose columns pre-equilibrated with column buffer (lysis buffer prepared without sucrose and with 0.1% (w/v) Triton X-100). Columns were washed twice with column buffer, three times with high stringency column wash buffer (column buffer containing 1 M LiCl), and twice again with column buffer. Bound proteins were eluted with 100 mM glutathione in column buffer and the purified proteins dialyzed into storage buffer (column buffer containing 50% (v/v) glycerol). Proteins prepared in this manner were stable for up to 6 months at Ϫ20°C. GST-ATF2(4 -94) or GST-c-Jun(1-135) were expressed in bacteria and purified as described (18).
In Vitro Activation of MKK6 by RIP Mutants-For activation assays, each of the Myc-RIP constructs was immunoprecipitated from at least seven 10-cm plates of 293 cells expressing the relevant pCDM12 construct. Cells were transfected with 10 g of RIP plasmid for these experiments. The lysis and immunoprecipitation procedure have been described (Ref. 18 and see above). Immune complexes were washed at high stringency as described above for the co-immunoprecipitation ex- Numbers correspond to GCK constructs transfected. B, recombinant MEKK1 interacts with the GCK-CTD. MEKK1 also interacts with SEK1 but not with PAK1. 293 cells were co-transfected with GST-tagged GCK-FL, GCK-CTD, PAK1, or SEK1 as indicated, and M2-FLAG-tagged MEKK1(817-1493). GST polypeptides were isolated on glutathione-agarose and probed with anti-FLAG antibody as indicated. In order to judge expression of the various constructs, crude extracts were also subjected to SDS-PAGE and immunoblotted with anti-GST and anti-FLAG as indicated. Numbers correspond to the GST-tagged constructs transfected. C, immobilized FLAG-MEKK1 (817-1221) can bind purified GST-GCK; the bound GCK can then phosphorylate the immobilized MEKK1. Left panel, purification of GST-GCK from 293 cells transfected with pEBG-GCK. Ten percent of the material was subjected to SDS-PAGE and Coomassie Blue staining. Right panel, 293 cells were transfected with M2-FLAG-MEKK1 (817-1221), the MEKK1 was immunoprecipitated with anti-FLAG and exposed to purified GST-GCK (full-length) as indicated. The FLAG-MEKK1 beads were washed at high stringency and incubated with [ 32 P]ATP. As a control, an identical amount of GST-GCK was incubated with blank beads and the supernatant incubated with [ 32 P]ATP. Proteins were resolved by SDS-PAGE and the gels subjected to autoradiography. D, the MEKK1 associated with the GCK-CTD is catalytically active. 293 cells were co-transfected with either GST (lane numbers 1 and 2) or GCK-CTD (lane numbers 3 and 4) and either empty FLAG vector (lanes 1 and 3) or M2-FLAG-MEKK1 (lanes 2 and 4). FLAG immunoprecipitates or GST isolates were prepared and assayed for phosphorylation of the MEKK1 substrate GST-SEK1-KR (top panels). Crude extracts were subjected to SDS-PAGE and immunoblotting with anti-FLAG or anti-GST as indicated (bottom panels). . GST pulldowns were performed and these as well as the crude extracts were subjected to SDS-PAGE and immunoblotting with either anti-FLAG antibody or anti-GST antibody as indicated (WB denotes Western blot). B, K44M-GCK is devoid of detectable MBP kinase activity. 293 cells were transfected with the indicated amounts of M2-FLAG-tagged wild type or K44M-GCK as indicated. FLAG immunoprecipitations were performed and assayed for MBP phosphorylation as described (55). Crude extracts were subjected to SDS-PAGE and immunoblotting with anti-FLAG as indicated (WB). C, kinase-inactive GCK binds MEKK1 poorly. The CT and PEST3 motifs of the GCK-CTD are necessary for MEKK1 binding. 293 cells were co-transfected with the indicated M2-FLAG-tagged-GCK mutant and deletion constructs (described in the text and shown in the top panel) and GST-MEKK1 (817-1221). GST pulldowns were performed. These periments. 20-l RIP beads or blank beads (prepared from mock immunoprecipitations performed with nonimmune serum) were suspended in a total of 40 l of assay buffer (20 mM Tris, pH 7.4, 2 mM EGTA, 1 mM dithiothreitol, 0.1% (w/v) Triton X-100). To this were added 20 l of assay buffer containing 6 ng of inactive GST-MKK6 or an equivalent amount of MKK6/p38 storage buffer. Reactions were started with the addition of 15 l of [ 32 P]ATP/MgCl 2 mixture to give final concentrations of 100 M [ 32 P]ATP and 10 mM MgCl 2 . The reactions were allowed to proceed for 30 min at 30°C. Reaction tubes were then centrifuged and 30 l of supernatant were removed to separate tubes. To each of these tubes were added 30 l of assay buffer containing 6 ng of GST-p38 or an equivalent amount of storage buffer. Additional ATP/Mg (100 M/10 mM plus that already present) was added and the reactions continued for 30 min at 30°C. At this time, 1 g of GST-ATF2(4 -94) plus additional ATP (100 M/10 mM plus that already present) was added and the reactions continued for 20 min at 30°C. Reactions were quenched with Laemmli sample buffer and the products subjected to SDS-PAGE and autoradiography.

RESULTS
The GCK COOH-terminal Regulatory Domain (CTD) Strongly Binds MEKK1-MEKK1, like GCK, selectively activates the SAPK pathway (32,40). The specificity of both GCK and MEKK1 for the SAPK pathway compelled us to ask if MEKK1 might be a target of GCK. Accordingly, we expressed in 293 cells glutathione S-transferase (GST)-tagged constructs of either full-length GCK or the GCK-CTD (residues 271-819). The GCK polypeptides were purified on glutathione-agarose. We then looked for an association between the GCK constructs and endogenous MEKK1. To do this, we immunoblotted the GCK isolates with an antibody to the MEKK1 COOH-terminal catalytic domain. Probing of the GCK isolates with this antibody indicated that a portion of the endogenous MEKK1 was selectively associated with the GCK-CTD (Fig. 1A).
The preferential interaction between MEKK1 and the GCK-CTD was somewhat counterintuitive. Both the GCK-CTD and full-length GCK, when overexpressed, can activate the SAPKs in vivo; however, full-length GCK activates coexpressed SAPK 5-10-fold more effectively than does the GCK-CTD (40). Thus we wished to explore in detail the molecular basis of this interaction and clarify its physiologic significance. First, in order to determine if recombinant MEKK1 behaved in a manner similar to that of the endogenous protein, we coexpressed an M2-FLAG-tagged form of MEKK1 (residues 817-1493) and either full-length, GST-tagged GCK, or a GST-tagged GCK-CTD construct. As specificity controls, we co-transfected parallel 293 cell cultures with a GST-tagged form of the mammalian STE20 homologue p21-activated kinase (PAK)-1 (26) and M2-FLAG-tagged MEKK1. Extracts were prepared and GSTtagged polypeptides were purified on glutathione-agarose, washed, and immunoblotted with anti-FLAG antibody. Immunoblotting of the immobilized GCK constructs with anti-FLAG antibody demonstrated that the 110-kDa FLAG-tagged MEKK1 species bound preferentially to the GCK-CTD construct. The interaction between GCK and MEKK1 was extremely stable and was resistant to washing with 1% Triton X-100 and 1 M LiCl. Moreover, MEKK1's interaction with GCK was not a promiscuous binding event insofar as we did not detect an interaction between MEKK1 and GST-PAK1 (Fig.  1B). This, despite the fact that constitutively active PAK mutants can activate coexpressed SAPK (53), and models developed from early yeast studies had indicated that kinases of the STE20 family such as PAK1 could regulate MAPKKK 3 MEK 3 ERK/MAPK core signaling modules (54). Inasmuch as MEKK1 interacts preferentially with the GCK-CTD as compared with full-length GCK (Figs. 1, A and B, and 2), we conclude that MEKK1 (817-1493) interacts with GCK in a manner similar to that of endogenous MEKK1.
While an interaction between full-length GCK and MEKK1 is not apparent in Fig. 1, A and B, it can occur in vivo (see also Fig. 2). Moreover, it is possible to recreate this interaction in vitro to a extent sufficient to allow for the phosphorylation, by the bound GCK, of an MEKK1 construct consisting of residues 817-1221. Thus, cells were transfected with vector or M2-FLAG-MEKK1 (817-1221) (Fig. 1C). The MEKK1 was immunoprecipitated from the transfected cells with anti-FLAG, the immunoprecipitates washed with high salt buffer (1 M LiCl) and incubated with GST or GST-full-length GCK that had been purified (Fig. 1C, left) from transfected cells expressing the cognate constructs. The resulting GCK ⅐ MEKK1 complexes were again washed in high salt (1 M LiCl) buffer and incubated with [␥-32 P]ATP. As a control for the kinase activity of the purified GCK, an identical amount of purified GST-GCK was allowed to autophosphorylate in the presence of empty protein-G beads. This autophosphorylation is shown in Fig. 1C, right panel, right lanes. Incubations containing both GCK and MEKK1 (817-1221) show both autophosphorylated GCK and phosphorylated MEKK1 (817-1221) (Fig. 1C, right panel). Neither 32 P-labeled polypeptide is detected in the absence of added GST-GCK, and the phosphorylated MEKK1 polypeptide is not seen in the absence of immunoprecipitated MEKK1. Moreover, no detectable protein kinase activity is present in the MEKK1 immunoprecipitates (Fig. 1C, right panel). Thus, the immobilized MEKK1 can bind sufficient purified full-length GCK to support detectable phosphorylation in vitro of the MEKK1 polypeptide. While the GCK preparation used in Fig. 1C is highly purified, and the MEKK1 immunoprecipitate was washed at high stringency prior to incubation with GCK, it is conceivable (albeit remotely) that phosphorylation of MEKK1 by the purified GCK might be catalyzed by a contaminating kinase, present in the GCK preparation or in the MEKK1 immunoprecipitate, that is activated by the bound GCK. However, insofar as the GCK⅐MEKK complex was washed at high stringency before adding ATP and proceeding with the kinase assay, phosphorylation by a contaminating kinase would require as a prerequisite the binding of GCK to MEKK1. Thus, the results in Fig. 1C lend further support to the notion that full-length GCK can bind MEKK1.
The MEKK1 associated with the GCK carboxyl terminus is catalytically active; and the GCK-CTD-MEKK1 interaction is sufficiently stable to support MEKK1's phosphorylation of its substrate, SEK1. Glutathione-agarose isolates from cells coexpressing either empty M2-FLAG vector or M2-FLAG-MEKK1, and either GST or GST-GCK-CTD were assayed for phosphorylation of kinase-inactive SEK1 (SEK1-K129R) in vitro. SEK1-K129R phosphorylation was only observed in GST isolates containing both the GCK-COOH terminus and MEKK1 (Fig. 1D).
In addition to interacting with the GCK CTD, MEKK1 can interact in vivo with SEK1. Coexpressed GST-tagged SEK1 and M2-FLAG-tagged MEKK1 form a complex that can be detected in anti-M2-FLAG blots of GST-SEK1 isolates (Fig.  1B). Inasmuch as SEK1-SAPK in vivo interactions have been and the crude extracts were then subjected to SDS-PAGE and immunoblotting with anti-FLAG or -GST as indicated. A control (untransfected) crude extract is labeled C in the anti-GST immunoblot. D, expression of the GCK-binding site of MEKK1 blocks SAPK activation by coexpressed GCK. 293 cells were co-transfected with GST-GCK, HA-SAPK-p46␤1, and FLAG-MEKK1(817-1340) as indicated. SAPK was immunoprecipitated and assayed in immune complexes as described (18). The effect of FLAG-MEKK1 (817-1340) on GCK activation of SAPK is corrected for any SAPK activating activity over basal contributed by FLAG-MEKK1 (817-1340) itself, which is shown in the figure. Expression of the transfected constructs is shown in the bottom panels.
reported previously, GCK may form part of a multimeric SAPK signaling module.
GCK Binds MEKK1 at Amino Acids 817-1221 of the MEKK1 Amino-terminal Regulatory Domain-MEKK1 contains an extensive (1221 amino acids) amino-terminal regulatory domain (33,55). Previous studies have demonstrated that Nck-interacting kinase, a mammalian SPS1 homologue with similar properties to GCK, can bind only full-length MEKK1 (42). The Nck-interacting kinase-MEKK1 interaction requires residues 1-719 of the MEKK1 polypeptide (42). While endogenous, fulllength MEKK1 can also interact with the GCK-CTD (Fig. 1A), our recombinant MEKK1 construct, which, like endogenous, full-length MEKK1, binds the GCK-CTD strongly and preferentially (Fig. 1B), consists of residues 817-1493; thus, GCK binds MEKK1 at a region on the MEKK1 polypeptide distinct from that which interacts with Nck-interacting kinase. Fig. 1C indicates that full-length GCK can bind to a MEKK1 construct consisting of residues 817-1221, a domain just outside the MEKK1 catalytic domain (amino acids 1221-1493). To confirm that the MEKK1 catalytic domain was not necessary for GCK binding, we employed two constructs, MEKK1(817-1340) and MEKK1(817-1221) wherein either subdomains V-XI (residues 1341-1493) of the catalytic domain, or all of the catalytic domain (residues 1221-1493), respectively, were deleted. M2-FLAG-tagged MEKK1(817-1493), MEKK1(817-1340), or MEKK1(817-1221) were coexpressed with either GST fulllength GCK or GST-GCK-CTD. GSH-agarose isolates were probed with anti-M2-FLAG to detect bound MEKK1. From Fig.  2A, it is clear that all three MEKK1 constructs can interact with GCK. While binding of MEKK1 to full-length GCK is clearly detectable, these MEKK constructs preferentially bind the GCK CTD. Again, the binding of MEKK1 to either fulllength GCK or the GCK-CTD is strikingly stable and is resistant to washing in 1 M LiCl and 1% Triton X-100. Similar experiments using the MEKK1 catalytic domain (amino acids 1222-1493) showed that this region of the MEKK1 polypeptide could not bind GCK-CTD (data not shown). We conclude that the amino-terminal 817 amino acids and the catalytic domain of MEKK1 (residues 1222-1493) are not required for binding GCK, and that the GCK-binding site on MEKK1 lies between residues 817 and 1221.
Kinase-inactive GCK Binds MEKK1 Less Strongly Than Does Wild Type-The preferential association between MEKK1 and the free GCK-CTD suggested that wild type catalytically active GCK might not interact with MEKK1 with sufficient avidity to be readily detected, whereas kinase-inactive GCK might interact more stably with GCK. Precedent for stable interactions between inactive protein kinases and their downstream effectors has been demonstrated (25,56,57). To compare the MEKK1 binding properties of kinase-inactive and wild type GCK, we expressed M2-FLAG-tagged GCK-K44M (Fig. 2C), which is completely devoid of MBP kinase activity (Fig. 2B) and GST-MEKK1 in 293 cells. To our surprise, while K44M-GCK does interact with MEKK1, it does so less stably than does wild type, full-length GCK (Fig. 2C). This lesser degree of MEKK1 binding correlates with the lesser degree of SAPK pathway activation elicited by K44M-GCK. 2 Thus, MEKK1 binds most stably to the free GCK-CTD; binding of GCK to full-length GCK, while detectable, is less avid, and kinase-inactive GCK binds MEKK1 less stably than does wild type GCK.
Binding of GCK to MEKK1 Is Mediated by the COOH-terminal 141 Amino Acids and the COOH-terminal PEST Motif (PEST3) of GCK-We wished next to identify the region(s) on the GCK-CTD necessary for MEKK1 binding. The GCK-CTD consists of three PEST motifs, and a leucine-rich region followed by a short, 141-amino acid extension, the CT (CT, Fig.  2C, top) (52). The leucine-rich and CT tail regions are modestly conserved in other mammalian SPS1 family kinases, including hematopoietic progenitor kinase-1, Nck-interacting kinase, and GCK-related (GCKR), and appear necessary for the interaction between Nck-interacting kinase and MEKK1 (39,40,41,52). A series of deletion constructs of the GCK-CTD, as well as full-length GCK were prepared and cloned into the pCMV5-M2-FLAG vector (Fig. 2C, top). These constructs were coexpressed in 293 cells with GST-tagged-MEKK1. As observed above (Fig. 1C), the interaction between full-length GCK and MEKK1, while clearly detectable, is comparatively less stable than that between MEKK1 and the GCK-CTD (Fig. 2C). A construct of the GCK-CTD wherein the amino-terminal PEST domain is deleted still readily interacts with MEKK1. By contrast, deletion of the short CT tail abrogates completely MEKK binding to the GCK-CTD. Deletion of the Leu-rich region restores binding while subsequent deletion of the COOH-terminal PEST domain again abolishes MEKK1 binding to the GCK-CTD (Fig. 2C).
Thus, the structural elements of GCK that contribute to MEKK1 binding are complex. While we do not know if PEST3 and CT are sufficient for MEKK1 binding (constructs consisting of these domains alone are unstable in vivo); from the results in Fig. 2C, we conclude that the CT and COOH-terminal PEST motif (PEST3) are necessary for MEKK1 binding to GCK. Insofar as deletion of the CT domain results in a loss of MEKK1 binding to the GCK-CTD, whereas subsequent deletion of the Leu-rich domain restores MEKK1 binding, the CT domain may act to relieve an inhibition to MEKK1 binding conferred by the Leu-rich domain. While the CT and PEST3 domains and, possibly, the Leu-rich domain appear to play a direct role in MEKK1 binding, PEST1 and -2 may contribute to a conformation that allows for MEKK1 binding; however, PEST1 and -2 are insufficient per se to support MEKK1 binding.
The GCK-binding Domain of MEKK1 Can Block GCK Activation of SAPK-What is the functional significance of the GCK interaction with MEKK1? To explore this question, we expressed GST-GCK and HA-SAPK (the p46␤1 isoform) with the MEKK1 construct, M2-FLAG-MEKK1(817-1340) ( Fig. 2A), wherein the COOH-terminal catalytic domain was deleted after subdomain IV, removing the substrate binding loop. This construct is essentially devoid of SAPK activating activity in vivo (Fig. 2D). However, expression of this construct results in a dramatic 80% inhibition of GCK's ability to activate the SAPK pathway (Fig. 2D). MEKK1(817-1340) expression does not inhibit GCK's kinase activity (Fig. 2D) and likely acts by preventing the binding of endogenous MEKK1 to the expressed GCK. The results in Figs. 1 and 2 are consistent with the hypothesis that MEKK1 is a physiologic target of GCK.
The GCK-CTD Can Bind TRAF2 in Vivo-Upon immunoprecipitation from cell extracts, endogenous GCK exhibits significant basal activity; however, the enzyme can be activated substantially in vivo by TNF (40). Thus we wished to determine if GCK could interact with any of the signaling molecules known to couple the SAPKs to TNFR1. Previous studies have shown that both TRAF2 and RIP overexpression can activate the SAPKs (44,45). However, targeted disruption of TRAF2 abolishes TNF activation of the SAPKs. Thus, TRAF2 is essential to TNF recruitment of the SAPKs (48,49), while the role of RIP in SAPK regulation is less clear. In light of these findings, we wished to ascertain if TRAF2 could associate in vivo with GCK. Accordingly, we coexpressed either GST-tagged full-length GCK or GST-GCK-CTD in 293 cells along with TRAF2 (un-2 T. Yuasa and J. M. Kyriakis, unpublished observations. tagged). GST-GCK isolates were then immunoblotted with an antibody to TRAF2 to detect TRAF2 bound to GCK. From Fig.  3A, it is clear that both full-length GCK and the GCK-CTD can form complexes in vivo with TRAF2. As is the case with the GCK⅐MEKK1 complexes, the GCK⅐TRAF2 complexes are stable to washing in 1% Triton X-100 and 1 M LiCl. We do not see an association between GCK and RIP (data not shown). In addition, as with MEKK1, the binding of full-length GCK to TRAF2, while clearly detectable and resistant to high stringency washing, is less stable than that between TRAF2 and the free GCK-CTD (Fig. 3A).
The GCK-TRAF2 Interaction Requires the PEST1 and CT Motifs of the GCK-CTD-We next wished to identify the domain(s) on the GCK-CTD necessary for TRAF2 binding. Accordingly, 293 cells were co-transfected with untagged TRAF2 and an M2-FLAG-tagged deletion series (Fig. 3B) of the GCK CTD. From Fig. 3C, it is clear that deletion of either the amino-terminal PEST motif (PEST1) or the 141-amino acid CT region of the GCK CTD eliminates the binding of TRAF2 to the GCK-CTD. It is noteworthy that the binding of MEKK1 to GCK is also inhibited upon deletion of the CT region of the GCK-CTD, and restored upon subsequent deletion of the GCK-Leurich region (Fig. 2C). By contrast, deletion of the GCK-CTD Leu-rich motif does not restore TRAF2 binding (Fig. 3C). Thus, the CT and PEST1 regions of GCK-CTD are necessary for binding TRAF2. PEST2 and -3 may contribute to stabilize the binding of TRAF2; however, these domains are insufficient, in and of themselves, to mediate TRAF2 binding.
The TRAF Domain of TRAF2 Mediates the GCK-TRAF2 Interaction-TRAF2 consists of an amino-terminal RING finger domain, a conserved TRAF domain, comprised of two tandem TRAF motifs, and an intermediate region containing five zinc fingers (Fig. 3D, top) (7)(8)(9)(10)(11). TRAF domains are thought to mediate heterotypic and homotypic protein-protein interactions (13). Deletion of the RING finger renders TRAF2 incapable of activating the SAPK pathway, indicating that the activation (but not necessarily the binding) of downstream signaling elements is mediated by this domain (46,47). We generated a series of GST-tagged TRAF2 deletion constructs (Fig. 3D, top) in order to determine the portion of TRAF2 responsible for binding GCK. These were coexpressed in 293 cells with M2-FLAG-GCK-CTD. Anti-FLAG immunoprecipitates were probed with anti-TRAF2 to detect bound TRAF2 deletion constructs. From Fig. 3D, it is clear that deletion of the RING and/or zinc finger domains does not compromise GCK binding to TRAF2. By contrast, deletion of the TRAF domains completely eliminates the interaction between TRAF2 and GCK. Interestingly, RIP also binds to the TRAF domains of TRAF2 (13). Our findings are consistent with the idea that the TRAF domains can mediate the interaction between TRAF proteins and their effectors. Once bound to TRAF2, the RING domain may promote GCK activation.
Activation of the SAPK and p38 Pathways by RIP Requires the RIP Intermediate Domain-While the GCK-TRAF2 and GCK-MEKK1 interactions provide one mechanism whereby the SAPKs are recruited to TNFR1, these interactions do not account for TNF activation of p38 inasmuch as GCK and MEKK1 are highly selective for the SAPK pathway, and neither can effectively activate p38 in vivo (26,32,40). Moreover, whereas overexpressed TRAF2 and RIP can activate the SAPKs, gene disruption studies indicate that TRAF2 is required for TNF activation of the SAPKs, while the role of RIP in SAPK activation is nebulous (44,45,48,49). Accordingly, we wished to determine the relationship between TRAF2 and RIP with regard to SAPK and p38 regulation.
RIP consists of an amino-terminal protein Ser/Thr kinase domain, a carboxyl-terminal death domain, and an intermediate domain (12,51) (Fig. 4A). In order to identify the region of RIP necessary for SAPK and p38 activation, we tested the ability of a panel of RIP mutants (Fig. 4A) to activate coexpressed SAPK and p38. Thus, 293 cells were transiently transfected with the mutant RIP constructs (Myc-tagged) and either HA-SAPK or HA-p38. SAPK and p38 were immunoprecipitated and assayed. Transient coexpression of Myc-tagged wild type (wt) RIP in 293 cells with either HA-tagged SAPK (the p46␤1 isoform) or HA-tagged p38 (the ␣ isoform) results in robust activation of both kinase pathways (Fig. 4B). From Fig. 4B, it is also clear that the RIP catalytic domain alone is incapable of SAPK or p38 activation. Moreover, D138N-RIP, which has been mutated within the phosphotransferase loop of subdomain VI, and is devoid of detectable kinase activity (51), activates both the SAPK and p38 pathways strongly. We observe no activation of the SAPKs or p38 upon overexpression of the RIP death domain (Fig. 4B).
Interestingly, deletion of the RIP intermediate domain abolishes completely SAPK and p38 activation, even if the kinase domain is intact (Fig. 4B). Thus the RIP intermediate domain appears necessary for SAPK and p38 activation while the catalytic domain is dispensable. The ⌬391-422-RIP mutant harbors a deletion of a charged motif within the intermediate domain (12,51). This mutant is unable to activate NF-B (51). However, expression of ⌬391-422-RIP results in vigorous activation of both SAPK and p38. Thus, the region of the RIP intermediate domain required for SAPK and p38 activation is distinct from that required for NF-B activation, suggesting that the SAPK, p38, and NF-B pathways emanating from RIP actually diverge at the RIP-ID (Fig. 4B).
The RIP Intermediate Domain Is Sufficient for SAPK and p38 Activation and Couples TNF to the SAPK and p38 Pathways, and TRAF2 to the p38 Pathway-The results in Fig. 4 indicate that the intermediate domain is necessary for RIP activation of coexpressed SAPK. To determine if this domain was sufficient for SAPK and p38 activation, we coexpressed the Myc-tagged RIP intermediate domain with either HA-SAPK or HA-p38 (Fig. 5A). It is clear that expression of the RIP intermediate domain results in dramatic activation of both the SAPK and p38 pathways. Thus, the RIP death domain, the region of RIP that associates with TRADD and is required for apoptosis (6,12), is not necessary for SAPK or p38 activation. Insofar as RIP and TRAF2 associate through the RIP-ID and kinase domains (6,13), RIP may couple TNFR1 to the SAPKs and p38s solely through its association with TRAF2.
We either HA-SAPK or HA-p38 and either empty vector or Myc-RIP-⌬ID. From Fig. 5A, it is clear that expression of the RIP-⌬ID construct completely blocks TNF activation of the SAPK and p38 pathways. These results combined with the results in Fig. 4 suggest that the intermediate domain mediates RIP signaling to the SAPKs and p38s, and that RIP is important for both SAPK and p38 activation by TNF. However, inasmuch as deletion of the TRAF2 gene abrogates TNF activation of SAPK (48,49), and the RIP-⌬ID, upon overexpression, might compete with endogenous RIP and/or TRAF2 for access to TRADD and TNFR1, the relationship between, and relative contribution of TRAF2 and RIP to SAPK and p38 signaling cannot be assessed from these experiments.
Thus, we wished to determine if RIP was an effector for TRAF2. Overexpression of TRAF2 has been reported to activate the SAPKs (46,47). We too see activation of HA-SAPK and HA-p38 by coexpressed TRAF2 (Fig. 5B). To our surprise, however, RIP-⌬ID could completely block activation of p38 by TRAF2, but failed to inhibit TRAF2 activation of coexpressed SAPK (Fig. 5B). We conclude from these results that RIP acts downstream of TRAF2 in the p38 activation pathway. By contrast, although this same pathway may be responsible for RIP activation of SAPK, the TRAF2 3 GCK 3 MEKK1 mechanism, or similar mechanisms (43) may bypass RIP and form a parallel pathway for SAPK activation by TNF. Such a redundant pathway would protect TRAF2 activation of SAPK from inhibition by RIP-⌬ID.
Association of an Endogenous MAPKKK Activity with the RIP Intermediate Domain-RIP, like GCK, is a putative effector for TRAF2 (6,13). The RIP-ID-dependent activation of p38 by TRAF2 impelled us to investigate the molecular basis of p38 activation by RIP. We sought to determine if RIP, through its ID, could associate with MAPKKKs in vivo, just as GCK does. Accordingly, we immunoprecipitated wild type RIP and a series of RIP mutants from transfected cells and examined them for the presence of an associated endogenous MAPKKK able to activate in vitro purified MKK6, the major MEK upstream of p38.
For these experiments, MKK6, and p38 were expressed as GST fusion proteins in 293 cells and purified to apparent homogeneity. Fig. 6A is a Coomassie Blue-stained gel of the purified MKK6 and p38. RIP immunoprecipitates were incubated with purified, inactive MKK6 and Mg-ATP. A portion of the activated MKK6 was then removed and used in assays to activate purified p38. Consequent activation of p38 was measured using GST-ATF2(8 -94) as a substrate. As indicated in the figures, p38, MKK6, or RIP immunoprecipitates were replaced with buffer or nonimmune serum immunoprecipitates in control incubations. Fig. 6B shows that both wild type and kinase-inactive (K45R) mutant RIP immunoprecipitates can activate MKK6 in vitro. Overall, the activation of MKK6 is between 6-and 10fold, and the resulting activation of p38 is up to 30-fold. Fig. 6C demonstrates that all of the RIP mutants which possess p38 activating activity in vivo can also reconstitute MKK6 and p38 activation in vitro. Notably, the RIP intermediate domain alone can associate with an endogenous MAPKKK activity capable of MKK6 activation in vitro. The identity of this MAPKKK is not yet known. We performed similar experiments to detect activation of SEK1 and MKK7 in vitro. We did not observe activation of SEK1 or MKK7 in these experiments. Thus the MAP-KKK associated with RIP may target SAPK-specific MEKs other than SEK1 or MKK7. Alternatively, the conditions employed may not favor in vitro SEK1/MKK7 activation, or RIP may not form a stable association with a MAPKKK upstream of SAPK.

FIG. 5. The RIP intermediate domain is necessary and sufficient for SAPK and p38 activation.
RIP is required to couple TRAF2 to the p38 pathway. A, activation of SAPK and p38 by the RIP-ID, inhibition of TNF activation of SAPK and p38 by a RIP dominant inhibitory construct devoid of the ID (RIP-⌬ID). 293 cells were cotransfected with the indicated Myc-tagged RIP constructs and either HA-SAPK or HA-p38. Cells were treated with 50 ng/ml human TNF or vehicle (phosphate-buffered saline, 1 mg/ml bovine serum albumin, labeled none) as indicated. Crude extracts were subjected to SDS-PAGE and immunoblotting with anti-HA and Myc immunoprecipitates with anti-Myc antibodies to judge expression of the transfected constructs. SAPK and p38 were assayed in anti-HA immunoprecipitates using c-Jun or ATF2, respectively, as substrates. B, RIP is required for TRAF2 activation of p38, but not for TRAF2 activation of SAPK. 293 cells were co-transfected with HA-SAPK, HA-p38, untagged TRAF2, and Myc-RIP-⌬ID as indicated. Crude extracts were subjected to SDS-PAGE and immunoblotting with anti-HA and anti-TRAF2. Myc immunoprecipitates were subjected to SDS-PAGE and immunoblotting with anti-Myc. Anti-HA immunoprecipitates were assayed for c-Jun kinase (SAPK transfections) or ATF2 kinase (p38 transfections) as indicated.

DISCUSSION
The findings presented herein expand upon the protein recruitment model of TNFR signaling and point to a molecular mechanism by which the TNFR1 signaling complex couples to the SAPKs and p38s. Our results show that GCK can form a molecular bridge linking TRAF2 and, by extension, the TNFR1 signaling complex to MEKK1, SEK1, and the SAPKs. GCK is likely not the only mammalian SPS1 that can serve downstream of TRAF2. GCKR, a kinase highly similar to GCK, can be activated in vivo by TNF and, like GCK, is an effector for TRAF2 (41). Likewise, at least two additional members of the TRAF family, TRAF6 and TRAF5, can activate the SAPKs (58), and, insofar as TRAF2 is critical to TNF signaling to the SAPKs (48,49), activation of the SAPKs by TRAFs-5 and -6 might represent mechanisms by which other ligands in the TNF family such as interleukin 1, CD40, CD27, or lymphotoxin-␤ (8 -11) recruit the SAPKs. We propose that GCK, and at least a subset of other members of the SPS1 family upstream of the SAPKs, are effectors for TRAFs.
We also show that RIP can activate both the SAPK and p38 pathways, and that TRAF2 activation of p38 proceeds through RIP and is dependent on the RIP-ID. We conclude that TRAF2-RIP forms a second mechanism for recruiting both SAPK and p38. Our results shown in Fig. 6 indicate that at least p38 activation by RIP may involve the association of an endogenous MAPKKK activity with the RIP intermediate domain. Whether or not the TRAF2-GCK, GCK-MEKK1, or the RIP-MAPKKK interactions are direct or mediated by additional polypeptides remains to be determined. Fig. 7 illustrates these findings. The observations that TRAF2 associates with RIP (6) and GCK in vivo, and uses RIP to relay signals to p38 (and possibly SAPK), point to the pivotal position of TRAF2 in TNF signaling to SAPK and p38, and are consistent with results from recent studies of mice in which the TRAF2 gene has been disrupted. TNF activation of SAPK is completely absent in these animals, while NF-B activation is intact (48,49).
Our results demonstrate that both GCK and RIP can perform a recruiting function, coupling TRAF2 to MAPKKKs. In addition, both GCK and RIP are protein kinases. What, then, is the role of the GCK and RIP catalytic domains? Deletion or abrogation of GCK's or RIP's kinase activity compromises, but does not eliminate effector activation (Fig. 4 and Refs. 40 and 56). Conversely, the isolated GCK noncatalytic CTD interacts more stably with MEKK1 than does wild type GCK, and kinase-inactive (K44M) GCK interacts with MEKK1 less stably than does wild type GCK. The simplest explanation for these results is that activation of GCK's kinase activity both allows for enhanced MEKK1 binding and promotes efficient turnover of activated MEKK1. The RIP kinase domain may function in a similar manner.
We do yet not know how GCK and RIP are regulated. While both of these kinases can associate with elements of the TNFR1 complex (Ref. 6 and Fig. 3), simple overexpression of either results in constitutive activation (6,12,46,47,56). Inhibitors present in limiting concentrations could be overcome by overexpression of GCK or RIP. Alternatively, overexpression of FIG. 6. RIP associates in vivo with an endogenous MAPKKK activity that can be employed to reconstitute the p38 pathway in vitro. A, purification of recombinant GST-p38 and MKK6. GST-p38 and MKK6 were purified by glutathione-agarose chromatography from 293 cells expressing the cognate pEBG construct. Ten percent of the purified material was subjected to SDS-PAGE and stained with Coomassie Blue. B, activation in vitro of the p38 pathway by immunoprecipitates of wild type or K45R-RIP. The indicated Myc-RIP constructs were immunoprecipitated from transfected 293 cells and used to activate recombinant GST-MKK6. Activated MKK6 was then removed to a separate tube and used to activate p38. Activation of p38 was determined by assaying ATF2 kinase activity. Nonimmune serum immunoprecipitates and/or kinase storage buffer replaced RIP, MKK6, or p38 in control assays as indicated. C, same as B, except that a panel of RIP mutants was used in the MKK6 activation assays. The MAPKKK activity associated with RIP interacts with the RIP-ID. GCK or RIP could force homoligomerization and/or interactions with TRAF2 that would not normally occur in resting cells. The TRAF domains mediate the association between TRAF2 and both GCK and RIP (Ref. 13 and Fig. 3D). However, the RING domains are necessary for SAPK pathway activation by TRAF2 (46,47). Thus, it is possible that upon the association of GCK (or RIP) with the TRAF2 TRAF domains, the TRAF2 RING finger domains could either promote activating conformational changes, or the dissociation of GCK or RIP inhibitors. These structural changes would, in turn, permit interactions with downstream effectors. As with MEKK1, TRAF2 binds the GCK-CTD more strongly than it binds full-length GCK. Again, activation of GCK's kinase activity may promote efficient turnover of activated GCK, thereby reducing the amount of recoverable wild type GCK⅐TRAF2 complexes.
It is intriguing to compare the interaction domains on the GCK-CTD for TRAF2 and MEKK1. Fig. 2 indicates that the CT and PEST3 motifs of GCK-CTD are necessary for MEKK1 binding while the CT and PEST1 motifs are needed for TRAF2 binding. There is significant homology between the GCK CT extension and the extreme COOH terminus of GCKR (43). TRAF2 binding to PEST1 and CT on GCK, or to analogous domains on GCKR, perhaps by promoting conformational changes, autophosphorylation, or other activating events, might render GCK (specifically PEST3) or GCKR accessible to MEKK1. Deletion of the Leu-rich domain of GCK-CTD restores the binding of MEKK1 lost upon deletion of CT. This may merely reflect the fact that deletion of CT destabilizes the GCK-CTD, an effect reversed upon subsequent deletion of the Leu-rich region. Alternatively, the Leu-rich domain may inhibit MEKK1 binding, perhaps by occluding PEST3; TRAF2 binding to GCK might relieve this inhibition.
Inasmuch as GCK activation of the SAPK pathway is inhibited upon expression of the GCK-binding site of MEKK1 (Fig.  2D), it is reasonable to propose that MEKK1 is a physiologic target of GCK. However, we do not yet know how GCK might trigger MEKK1 activation, nor do we know how RIP might regulate its associated MAPKKK activity. Efforts to elucidate the mechanisms of stress-regulated MAPKKK regulation have been hampered by the fact that all mammalian MAPKKKs identified thus far as activators of the SAPKs and p38s are constitutively active upon overexpression (26,(33)(34)(35)(36)(37)(38)(39). In addition, the endogenous forms of most of these MAPKKKs emerge from cell extracts in a highly activated state (26,(33)(34)(35)(36)(37)(38)(39). GCK or RIP might simply phosphorylate and activate MAPKKKs. Indeed, we observe that GCK can phosphorylate MEKK1 (Fig.  1C); however, such a mechanism fails to take into account the observation that the kinase domains of GCK and RIP are not necessary for SAPK or p38 activation. MEKK1 or the MAP-KKK associated with RIP might be regulated negatively, or through activating structural changes arising as a consequence of binding GCK, as described above for the TRAF2-GCK interaction.
The activation of AP-1 is critical to the inflammatory responses elicited by TNF. Our results provide a framework for a plausible mechanism linking the TNFR1 signaling complex to two ERK/MAPK pathways that regulate AP-1: the SAPKs and p38s. We propose that TNFR1 couples to the SAPKs and p38s through redundant and bifurcating pathways mediated in large part by TRAF2 (Fig. 7). TRAF2 associates with GCK which, in turn, associates with MEKK1. TRAF2 can also associate with RIP which activates the SAPKs and p38s, possibly through association with a distinct MAPKKK. Future studies will focus on the identification and characterization of the RIP-associated MAPKKK, the mechanisms of GCK and RIP recruitment and activation of MAPKKKs, and the mechanisms of GCK and RIP regulation.