Engagement of Tumor Necrosis Factor (TNF) Receptor 1 Leads to ATF-2- and p38 Mitogen-activated Protein Kinase-dependent TNF-α Gene Expression*

Engagement of the tumor necrosis factor-α (TNF-α) receptors by the TNF-α ligand results in the rapid induction of TNF-α gene expression. The study presented here shows that autoregulation of TNF-α gene transcription by selective signaling through tumor necrosis factor receptor 1 (TNFR1) requires p38 mitogen-activated protein (MAP) kinase activity and the binding of the transcription factors ATF-2 and Jun to the TNF-α cAMP-response element (CRE) promoter element. Consistent with these findings, TNFR1 engagement results in increased p38 MAP kinase activity and p38-dependent phosphorylation of ATF-2. Furthermore, overexpression of MADD (MAP kinase-activatingdeath domain protein), an adapter protein that binds to the death domain of TNFR1 and activates MAP kinase cascades, results in CRE-dependent induction of TNF-α gene expression. Thus, the TNF-α CRE site is the target of TNFR1 stimulation and mediates the autoregulation of TNF-α gene transcription.

The tumor necrosis factor (TNF) 1 -␣ gene encodes a pleiotropic cytokine involved in multiple immunological responses (1). Furthermore, TNF-␣ has been implicated in the pathogenesis of a variety of infectious and autoimmune diseases where host production of too much or too little TNF-␣ is associated with variable patterns of disease pathogenesis (1). The biological actions of TNF-␣ are initiated by its binding to a 55-kDa receptor (TNFR1) and/or to a 75-kDa receptor (TNFR2). Although these two receptors induce both distinct and overlapping responses (see Ref. 2 for review), the majority of TNF-␣ effects, including the initiation of cell death cascades and host responses against a variety of pathogens, appear to be mediated by TNFR1 (3)(4)(5)(6)(7)(8).
The ability of both TNFR1 and TNFR2 to transduce signals is dependent upon the interaction of their cytoplasmic tails with intracellular proteins (see Ref. 9 for review). The intracellular domain of TNFR1, in contrast to TNFR2, contains a region called the "death domain," which binds adapter proteins such as TRADD (TNFR-associated death domain protein) (10). TRADD binds two additional transducers, TRAF2 (TNFR-associated factor-2) and receptor-interacting protein (11). These proteins, in turn, induce the kinase cascades ultimately resulting in the activation of the transcription factor NF-B (reviewed in Ref. 12) and of the cell death pathway (10). MADD (MAP kinase-activating death domain protein) is another protein that binds to the death domain of TNFR1. However, by contrast to TRADD, MADD does not cause cell death or NF-B activation, but specifically stimulates the c-Jun NH 2 -terminal (JNK) and extracellular signal-regulated kinase (ERK) members of the MAP kinase family of protein kinases (13). Notably, the protein partners with which MADD interacts and the genes activated through the MADD pathway remain to be elucidated.
Induction of gene expression is an important consequence of TNFR1 and/or TNFR2 engagement and is essential for many of the biological responses of TNF-␣. However, gene regulation secondary to the selective signaling of TNF-␣ through the individual TNF receptors remains poorly understood. Studies using soluble TNF-␣, which stimulates both TNF receptors, have identified E-selectin, interleukin-6, and TNF-␣ itself as TNF-␣-inducible genes (14 -16).
Signaling through TNFR1 leads to gene induction via activation of the IB kinase pathway and NF-B translocation and by activation of the MAP kinase family, which results in the phosphorylation and transcriptional activation of AP-1 family members (reviewed in Ref. 17). Although NF-B activation was shown to be critical in the regulation of the E-selectin and interleukin-6 genes by TNF-␣ (14,15), an early study implicated a cAMP-response element (CRE) in the autoregulation of TNF-␣ gene expression (18). This TNF-␣ CRE site is critical in the regulation of TNF-␣ by multiple signal transduction pathways and binds ATF-2 and Jun proteins (19,20), which become transcriptionally active upon phosphorylation by the p38 and JNK members of the MAP kinase family of protein kinases (17).
Here, we show that selective signaling through TNFR1 results in the induction of TNF-␣ gene transcription. Strikingly, this induction is dependent upon p38 MAP kinase activity and the binding of ATF-2/Jun proteins to the TNF-␣ CRE. Consistent with these findings, overexpression of MADD results in TNF-␣ gene induction, thus linking signaling through TNFR1 to MAP kinase activation and TNF-␣ gene induction. Moreover, this study establishes the TNF-␣ CRE site as the target of TNFR1 signaling and demonstrates the importance of the MAP kinase signal transduction pathway in TNFR1-mediated TNF-␣ gene transcription.
Electrophoretic Mobility Shift Assay-L929 cells were incubated for 2 h with or without 100 units/ml rhTNF-␣ where indicated; nuclear extracts were prepared; and an electrophoretic mobility shift assay was performed as described previously (22). Antibody competition assays were performed using an anti-ATF-2 antibody (a gift from Dr. D. Thanos); antibodies to Jun family members (Santa Cruz Biotechnology, Inc., Santa Cruz, CA); or antibodies to p50, p65, and c-Rel (gifts from Dr. N. Rice) as described previously (20). The synthetic oligonucleotides used in the electrophoretic mobility shift assay were as follows: 3(L), FIG. 1. Engagement of TNFR1 stimulates TNF-␣ gene expression in L929 fibroblasts. A, induction of TNF-␣ gene transcription by TNFR1 engagement in L929 cells. L929 cells were stimulated with 100 units/ml rhTNF-␣ for 2 h, and RNA was isolated. Subsequently, an RNase protection assay was performed, and murine TNF-␣ mRNA levels were measured. The ␥-actin riboprobe was used to control for equal loading and processing of RNA. B, TNFR1 engagement induces human TNF-␣/CAT reporter genes in L929 cells. L929 cells were transiently transfected with TNF-␣/CAT reporter gene constructs containing Ϫ200 or Ϫ1045 nucleotides 5Ј to the human TNF-␣ transcription start site. Twenty-four hours after transfection, cells were treated for 16 h with 100 units/ml rhTNF-␣. To control for transfection efficiency, all cells were cotransfected with CMV-␤-galactosidase, and extracts were normalized to ␤-galactosidase activity. The experiment shown is representative of three independent experiments.

FIG. 2. MADD stimulates TNF-␣ gene expression.
A, MADD, but not TRADD, induces TNF-␣ gene expression. L929 cells were transiently transfected with the human Ϫ200TNF-␣/CAT reporter gene and cotransfected with MADD, TRADD, or control vector and CMV-␤-galactosidase. Thirty-six hours after transfection, cells were harvested, and CAT assays were performed after normalization of extracts for ␤-galactosidase expression. The percent conversion of [ 14 C]chloramphenicol to its acetylated forms was quantified using a Betagen Betascope and normalized to the control levels (100%) and then averaged and plotted. The results displayed are an average from three independent experiments, and the error bars represent S.E. B, the MADD death domain is sufficient for activation of TNF-␣ gene expression. L929 cells were transiently transfected with the human Ϫ200TNF-␣/CAT reporter gene and cotransfected with pFLAG-15TU (encoding a 320-amino acid protein containing the MADD death domain), a mutant form of 15TU containing a mutation in the death domain (15TU⌬DD), or control vector and CMV-␤-galactosidase. Thirty-six hours after transfection, cells were harvested; CAT assays were performed after normalization of extracts for ␤-galactosidase expression; and results were plotted as described above. The experiment shown is representative of three independent experiments.
FIG. 3. The TNF-␣ CRE is required for TNFR1-or MADD-stimulated TNF-␣ gene expression. L929 cells were transfected with the wild-type (WT) human Ϫ200TNF-␣/CAT reporter gene or TNF-␣/CAT reporter genes bearing mutations in the CRE (C1M) or the 3 site (3ЈM) of the TNF-␣ promoter (see diagram at the bottom). Activation of the promoter by treatment with rhTNF-␣ or by overexpression of MADD was assessed. A, The wild-type, C1M, or 3ЈM TNF-␣/CAT reporter gene and CMV-␤-galactosidase were transiently transfected into L929 cells. Twenty-four hours after transfection, cells were either mock-stimulated (white bars) or stimulated with rhTNF-␣ (black bars), and CAT assays were performed and analysed as in Fig. 2. The error bars represent S.E. The results displayed are an average from three independent experiments. B, MADD stimulation TNF-␣ gene expression requires the CRE. The wild-type, C1M, or 3ЈM TNF-␣/CAT reporter gene and CMV-␤galactosidase were cotransfected with a MADD expression vector (black bars) or a control vector (white bars), and 36 h after transfection, cells were harvested, and CAT assays and analysis were carried out as described for A.
RNase Protection Assay-RNA was prepared from L929 cells stimulated with TNF-␣ as described above, and an RNase protection assay was performed using murine TNF-␣ and ␥-actin probes as described previously (22).
Immunoprecipitation and Western Blot Analysis-Nuclear extracts and whole cell extracts were prepared from L929 cells after a 15-min stimulation with 1000 units/ml rhTNF-␣ in the presence or absence of the p38 inhibitor SB203580 (2.5-20 M) as indicated, which was added 1 h prior to stimulation with rhTNF-␣. Nuclear extracts (10 g) were subjected to Western blot analysis, and phosphorylated ATF-2 and c-Jun proteins were detected using a phospho-specific antibody kit (New England Biolabs Inc., Beverly, MA) as directed by the manufacturer. p38 MAP kinase was immunoprecipitated from whole cell extracts and used to perform a kinase assay using GST-ATF-2 as a substrate. Both the immunoprecipitation and kinase assays were performed using the p38 MAP kinase assay kit from New England Biolabs Inc.

Engagement of TNFR1 Induces TNF-␣ Gene Expression in L929
Cells-Human TNF-␣ binds to murine TNFR1, but not to murine TNFR2 (23), and as a result, treatment of murine cells with rhTNF-␣ results only in TNFR1-mediated effects. Therefore, to determine whether signaling through TNFR1 alone resulted in the induction of TNF-␣ gene expression, we stimulated murine L929 fibroblast cells with rhTNF-␣ and measured TNF-␣ mRNA levels. As shown in Fig. 1A, L929 cells constitutively expressed a very low level of TNF-␣ mRNA (lane 1), which was induced after stimulation with rhTNF-␣ (lane 2). Thus, selective signaling through TNFR1 results in the induction of TNF-␣ gene expression.
To identify the TNF-␣ promoter sequences required for the transcriptional activation of the TNF-␣ gene via TNFR1 signaling, we first transfected L929 cells with a TNF-␣/CAT reporter gene containing either Ϫ1045 or Ϫ200 nucleotides upstream of the TNF-␣ mRNA cap site. Consistent with studies in multiple other cell types and using a variety of inducers (19,20,22), Ϫ200 nucleotides upstream of the TNF-␣ transcription start site are sufficient for maximal inducibility of the gene by TNFR1 engagement in L929 cells (Fig. 1B). In fact, the Ϫ200TNF-␣/CAT reporter gene construct was consistently more inducible by rhTNF-␣ than the Ϫ1045TNF-␣/CAT construct. Taken together, these experiments demonstrate that L929 cells are a physiological system in which to characterize TNF-␣ autoregulation secondary to TNFR1 signaling and that Ϫ200 nucleotides are sufficient for maximal TNFR1-mediated induction of the TNF-␣ gene.
MADD, but Not TRADD, Activates TNF-␣ Gene Expression-MADD recruitment to TNFR1 results in the activation of JNK and ERK MAP kinase pathways, but does not activate NF-B (13). By contrast, TRADD interaction with TNFR1 initiates a pathway resulting in NF-B activation (10). To determine whether the signal transduction pathways initiated by MADD and/or TRADD resulted in the activation of TNF-␣ gene expression, we cotransfected MADD or TRADD expression vectors with the TNF-␣/CAT reporter gene. As shown in Fig. 2A, overexpression of MADD, but not TRADD, increased TNF-␣/CAT reporter activity. Thus, MADD-stimulated MAP kinase activity activates TNF-␣ transcription.
A truncated form of MADD containing the death domain, called 15TU, also activates the JNK and ERK pathways (13). We tested whether the MADD death domain was sufficient for activation of TNF-␣ by cotransfection of 15TU with the TNF-␣ reporter. As shown in Fig. 2B, 15TU activated the TNF-␣ reporter to levels comparable to MADD. However, a mutant form of 15TU, with a 3-amino acid substitution in the death domain (15TU⌬DD), was not capable of inducing the TNF-␣/ CAT reporter (Fig. 2B). Taken together, our results establish the TNF-␣ gene as the first identified gene target of MADD and implicate the involvement of MAP kinase activation in TNFR1 stimulation of TNF-␣. We note that overexpression of 15TU⌬DD did not abrogate rhTNF-␣ induction of the TNF-␣/ CAT reporter gene (data not shown), consistent with the involvement of other TNFR1-stimulated pathways not transduced by MADD.
The TNF-␣ CRE Is Required for Activation of the Gene via TNFR1-The TNF-␣ composite CRE/3 promoter element, which is critical in the regulation of TNF-␣ by a variety of stimuli, binds ATF-2/Jun proteins (19,20). Given the facts that ATF-2/Jun become transcriptionally active upon phosphorylation by JNK and that ATF-2 is also phosphorylated by p38 MAP kinase (see Ref. 17, for example), we tested whether the CRE was required for TNFR1-mediated TNF-␣ gene expression. As shown in Fig. 3A, a mutation in the TNF-␣ CRE, called C1M, inhibited TNFR1-stimulated TNF-␣/CAT activity in L929 cells, whereas a mutation in the adjacent 3-nuclear factor of activated T cells (NFAT)-binding site, called 3ЈM, had no effect upon TNF-␣ gene expression. Similarly, MADD-stimulated TNF-␣/CAT reporter activity was also inhibited by mutation of the CRE, but not by mutation of the 3-NFAT site (Fig. 3B). Thus, the TNF-␣ CRE site is required for activation of TNF-␣ gene expression stimulated by TNFR1 or by MADD.
ATF-2/Jun Proteins Constitutively Bind to the CRE in L929 Cells-Given the functional importance of the CRE in TNFR1mediated TNF-␣ gene transcription, we next investigated which activators bind to the CRE in TNFR1-stimulated L929 cells. We prepared nuclear extracts from L929 cells stimulated with rhTNF-␣ and performed an electrophoretic mobility shift assay using an oligonucleotide probe containing the composite CRE/3 site (3(L)). As shown in Fig. 4A (lanes 1 and 2), the 3(L) probe bound three constitutive complexes, which were not inducible by rhTNF-␣. Using antibodies to ATF-2 and Jun proteins, we showed that ATF-2 proteins were contained in the upper two complexes (Fig. 4A, lanes 3-5, labeled 1 and 2) and that Jun proteins were contained in the two lower complexes (labeled 2 and 3). In contrast to ATF-2 and Jun, antibodies to the NF-B proteins p50, p65, and c-Rel did not react with the 3(L)-binding complexes (Fig. 4A, lanes 8 -10). As a positive control for TNF-␣ stimulation of L929 cells and NF-B binding, we used the high affinity NF-B-binding site, PRDII (24), as a probe. PRDII bound a TNF-␣-inducible complex (Fig. 4B, compare lanes 1 and 2), which reacted with antibodies to p50/p65 NF-B proteins in a supershift assay (lanes 3 and 4). From these experiments, we conclude that ATF-2/Jun proteins bind constitutively to the TNF-␣ CRE site in L929 cells.
TNFR1-mediated Phosphorylation of ATF-2 Is Dependent upon p38 MAP Kinase-Since ATF-2 becomes transcriptionally active upon phosphorylation by p38 MAP kinase, we studied the role of p38 MAP kinase activity in TNFR1-stimulated TNF-␣ gene expression and ATF-2 phosphorylation in L929 cells. As shown in Fig. 5A, TNFR1-mediated induction of the TNF-␣/CAT reporter was inhibited in a dose-dependent manner by SB203580, a pyridinylimidazole that specifically binds to and inhibits p38 MAP kinase (25).
To correlate the phosphorylation status of ATF-2 with the inhibitory effect of SB203580 upon TNF-␣ transcription, we performed a Western blot analysis using phospho-specific antibodies to ATF-2. Treatment of L929 cells with rhTNF-␣ resulted in the phosphorylation of endogenous ATF-2 (Fig. 5B,  upper panel, lane 2), whereas addition of increasing amounts of SB203580 resulted in a dose-dependent inhibition of ATF-2 phosphorylation (lanes 3-7). Notably, several groups have reported inhibition of JNK activity by higher doses of SB203580 (26,27). However, we did not observe an inhibitory effect of our preparation of the p38 inhibitor on the phosphorylation of c-Jun, even at concentrations of up to 20 M (Fig. 5B, lower  panel).
To demonstrate that TNFR1 engagement resulted in an increase in p38 MAP kinase activity, we prepared whole cell extracts of L929 cells stimulated with rhTNF-␣ and immunoprecipitated cellular proteins with an anti-p38 MAP kinase antibody. A kinase assay using GST-ATF-2 as a substrate was then performed on the immunoprecipitated samples, and they were analyzed by Western blotting. Using a phospho-specific ATF-2 antibody, we demonstrated that TNFR1 engagement resulted in an increase in p38-mediated ATF-2 phosphorylation (Fig. 5C, compare lanes 1 and 3). We demonstrated the specificity of this effect by pretreating the cells with 10 M SB203580, which, as expected, inhibited the TNFR1-mediated

FIG. 5. TNFR1 engagement of L929 cells results in p38 MAP kinase-dependent TNF-␣ gene expression and ATF-2 phosphorylation.
A, TNFR1-stimulated TNF-␣ gene expression is p38 MAP kinase-dependent. L929 cells were transfected with the Ϫ200TNF-␣/ CAT reporter gene and stimulated with rhTNF-␣ as described in the legend to Fig. 1. Where indicated, cells were pretreated with the p38 inhibitor SB203580 1 h prior to stimulation with rhTNF-␣. For all experiments, a CMV-␤-galactosidase construct was cotransfected, and extracts were normalized for ␤-galactosidase activity to control for transfection efficiency. Results of three independent experiments were normalized to the TNF-␣-induced levels of the Ϫ200TNF-␣/CAT reporter gene in each experiment in the absence of inhibitor (100%) and then averaged and plotted. An ϳ50% inhibition of TNF-␣/CAT reporter activity was achieved at a concentration of 10 M SB203580. The error bars represent S.E. B, TNFR1-mediated phosphorylation of endogenous ATF-2, but not c-Jun, is p38-dependent in L929 cells. Nuclear extracts were prepared from L929 cells that were mock-treated (untreated (UN)) or treated with the indicated concentrations of the p38 inhibitor SB203580 1 h prior to stimulation with 1000 units/ml rhTNF-␣. Western blot analysis was performed using phospho-specific antibodies recognizing ATF-2 (upper panel) and c-Jun (lower panel). The positions of the phosphorylated proteins are indicated. The results displayed are representative of three independent experiments. Quantification of the blot by densitometry revealed that a 50% inhibition of ATF-2 phosphorylation was achieved at a concentration of 10 M SB203580, which is identical to the effects of the inhibitor on TNF-␣/CAT reporter gene expression using this concentration (see A). C, TNFR1-dependent induction of cellular p38 MAP kinase activity. L929 cells were stimulated with 1000 units/ml rhTNF-␣; whole cell extracts were prepared; p38 MAP kinase was immunoprecipitated; and a kinase assay was performed using GST-ATF-2 as the substrate. The reactions were then subjected to Western blot analysis, and phosphorylated ATF-2 was visualized by chemiluminescence. The experiment displayed shows a 3.2-fold increase in phosphorylated ATF-2 levels (compare lanes 1 and  3), which was inhibited ϳ2-fold by pretreatment of the cells with 10 M SB203580 (compare lanes 3 and 4). The position of the phosphorylated protein is indicated on the left. The results displayed are representative of three independent experiments. increase in ATF-2 phosphorylation (Fig. 5C, compare lanes 3  and 4). Taken together, this series of experiments links TNFR1-mediated activation of p38 MAP kinase to the p38-dependent phosphorylation of ATF-2 and the induction of TNF-␣ gene expression. DISCUSSION Signaling through TNFR1 leads to at least three distinct effector functions, including MAP kinase activation, NF-B activation, and the induction of apoptosis (see Ref. 26, for example). Here, we have shown that TNFR1 engagement leads to TNF-␣ gene transcription through the p38-dependent phosphorylation and binding of ATF-2/Jun to the TNF-␣ CRE promoter element. Thus, the p38 MAP kinase pathway is the critical arm of the TNFR1 signaling pathway involved in the transcriptional autoregulation of TNF-␣. Recent work from other laboratories supports a role for p38 MAP kinase in TNF-␣-mediated gene activation. For example, disruption of the Mkk3 gene, a specific activator of p38 MAP kinase, causes a selective defect in TNF-␣-stimulated p38 MAP kinase activation and blocks TNF-␣ induction of the interleukin-1 and -6 genes (27). Consistent with earlier studies showing the importance of the CRE and the binding of ATF-2/Jun in TNF-␣ gene regulation in activated T cells (19,20), p38 has also recently been shown to be involved in the transcriptional and translational control of TNF-␣ in T cells (28).
Signaling through TNFR1 and its subsequent association with TRADD, TRAF2, and NF␤-inducing kinase result in the activation of NF-B in a wide spectrum of cell types (3)(4)(5)7). Our experiments argue against a role for NF-B in the autoregulation of TNF-␣ since overexpression of TRADD, which transduces signals resulting in the activation of NF-B, did not result in the induction of TNF-␣ gene expression. Furthermore, we did not detect inducible NF-B binding activity to the TNF-␣ CRE/3 site, which bears a weak sequence similarity to a consensus NF␤ binding site (22). Rather, we have shown that ATF-2/Jun bind to the CRE site, which is required for TNFR1-mediated TNF-␣ gene expression.
Our data also suggest a role for MADD in TNFR1-mediated TNF-␣ gene regulation. MADD signaling has previously been shown to activate the ERK and JNK MAP kinase pathways (13). Consistent with its role in MAP kinase activation, we show that MADD activation of TNF-␣ gene expression is dependent upon an intact CRE site. Given that MADD stimulates the JNK pathway, it is likely that MADD mediates phosphorylation of ATF-2 via JNK activation and thereby induces transcription of the TNF-␣ gene. We note that we could not observe MADD activation of the p38␣ and p38␤ MAP kinase forms, although we did observe inhibition of MADD-stimulated TNF-␣ gene expression by SB203580 (data not shown). Interestingly, SB203580 does not inhibit JNK1, but can inhibit JNK2␤ (29). Thus, the observation that SB203580 did not inhibit the activation of c-Jun (Fig. 5B) may reflect the activation of both JNK1 and JNK2␤ by rhTNF-␣, whereas MADD may activate only JNK2␤. Taken together with our demonstration that TNFR1-mediated activation of TNF-␣ requires p38 activation, these data indicate that TNFR1 stimulation of TNF-␣ gene expression involves both MADD-dependent activation of the JNK pathway and MADD-independent activation of p38 MAP kinase.
Mice deficient in TNFR1 or TNF-␣ itself display an overlapping set of pathologies. For example, mice deficient in TNFR1 or TNF-␣ demonstrate abnormal organization of splenic B cell follicles (30), are more susceptible to a variety of intracellular pathogens including Listeria and tuberculosis, and are relatively resistant to developing septic shock (8, 31-33). Thus, TNFR1-stimulated TNF-␣ gene expression is central to normal immunological processes as well as to host defense. The autoregulation of TNF-␣, which begins at the level of the transcriptional induction of the gene, is likely to be crucial in several of these disease processes.
The study presented here has demonstrated the required and critical roles of p38 MAP kinase and ATF-2 in TNF-␣ autoregulation. These molecular targets may thus prove useful in the therapeutic modulation of TNF-␣ gene expression in cases where too much or too little TNF-␣ is associated with variable patterns of disease pathogenesis.