Tumor Necrosis Factor Q! Stimulates AP-1 Activity through Prolonged Activation of the c-Jun Kinase*

Tumor necrosis factor a (TNFa) has multiple biologi- cal functions including the prolonged activation of the collagenase and c;iun genes, which are regulated via their AP-1 binding sites. We show that incubating human fibroblasts with TNFa induces prolonged activation of JNK, the cJun kinase, which phosphorylates the transactivation domain of c-Jun. Furthermore, an immune complex kinase assay specifically demonstrates that TNFa stimulates the activity of JNK1, the recently described predominant form of JNK. TNFa also pro- duces a small and transient increase in extracellular signal-regulated kinase (ERK) activity and no measured increase in Raf-1 kinase activity. On the other hand, epi- dermal growth factor causes a prolonged activation of Raf-1 kinase and ERK activity and a smaller, more tran- sient activation of JNK, whereas the phorbol ester phor-bo1 12-myristate 13-acetate causes a small stimulation of Raf-1 kinase and a pronounced stimulation of ERK activity. The activation of JNK by TNFa does not correlate with Raf-1 or ERK activity. The kinetics of Raf-1, ERK, and JNK induction by epidermal growth factor, phorbol 12-myristate 13-acetate, or TNFa

Tumor necrosis factor a (TNFa),' a potent cytokine produced mainly by activated monocytes, has multiple biological functions (1). TNFa actions are initiated by binding to two distinct TNFa receptors of 55-60 kDa (TNF R-55 or TNFR1) and 75-80 kDa (TNF R-75 or TNFRB), which have different affinities for TNFa and may mediate different downstream responses (2)(3)(4)(5)(6)(7)(8). The binding of TNFa to its receptors triggers the activation of several second messengers through stimulation of protein kinase C, sphingomyelinase, and phospholipase 4 (9)(10)(11)(12). Exposure to TNFa also results in NFKB activation (13,14) and stimulation of AP-1 activity (9). The final nuclear targets for TNFa include genes containing AP-1 binding sites, such as collagenase and c-jun, and genes containing NFKB binding sites, such as the human immunodeficiency virus-long terminal repeat and a variety of cytokine genes (9,14,15). Incubating cultured fibroblasts with TNFa results in prolonged activation of the collagenase and c-jun genes as compared with incubating with the phorbol ester PMA (9). The stimulation of c-jun transcription by TNFa is independent of de nouo protein synthesis (9). TNFa also causes a transient activation of the c-fos gene. The products of the c-fos and c-jun genes are components of the transcription factor AP-l(l6, 17).
The transcriptional activity of the c-Jun protein is regulated by its phosphorylation status. Phosphorylation of c-Jun at Thr-231, Ser-243, and Ser-249 near its DNA binding domain inhibits DNA binding and therefore reduces transcriptional activity (17,18). Phosphorylation of two of these sites (Thr-231 and Ser-249) is catalyzed in. vivo and in vitro by casein kinase I1 (18). Ser-243 is the major site of phosphorylation by E R G (19)(20)(21). O n the other hand, phosphorylation of c-Jun in its activation domain at Ser-63 and Ser-73 increases its transcriptional activity (22-24) but does not affect its DNA binding activity (25). Phosphorylation of the amino-terminal activation domain is catalyzed by a specific Jun nuclear kinase (JNK), which has been recently purified, and one of its constituents, JNK1, has been cloned (26,27).
In this study, the pathways in human diploid fibroblasts by which TNFa stimulates c-Jun are compared with those stimulated by EGF and the phorbol ester PMA. PMA functions in part by causing dephosphorylation of residues adjacent to the DNA binding domain in the carboxyl-terminal half of c-Jun (17). EGF functions through multiple signal transduction pathways, which include the activation of Ras, Raf, and extracellular signal-regulated kinases (ERK1 and ERK2) (reviewed in Refs. 19 and 28). Cell Culture-MRC-5 human fibroblast cells at the 14th passage were obtained from ATCC (CCL 171) and cultured in minimum essential medium with Earle's salts (Life Technologies, Inc.) containing 10% fetal bovine serum (Gemini), 0.5 mg/ml L-glutamine, 180 unitdml penicillin G, and 180 pg/ml streptomycin (Irvine Scientific).

EXPERIMENTAL PROCEDURES
Materials-Receptor grade EGF (Sigma) and TNFa (R & D Systems) were used at a final concentration of 10 ng/ml. PMA (Sigma) was used at a final concentration of 60 ng/ml.
Dansfections and Luciferase Assays-MRC-5 cells were transfected by the calcium phosphate coprecipitation technique as previously described (32). Briefly, 5 pg of reporter gene were cotransfected with 0.5 ! -% of expression vector where indicated. Cells were incubated in complete media for 4 days after removing the precipitates, and then the indicated TNFa Stimulates c-Jun Kinase Activity factors were added. 5 h after factor addition, extracts for luciferase and protein assays were prepared as described (32). All luciferase values were normalized for extract protein concentration.
Cellular Extmcts-Extracts were prepared a s described in Ref. 33. The nuclear isolation step was omitted for the preparation of whole cell extracts.
Cell Labeling, Immunoprecipitation, and Phosphopeptide Mapping-Subconfluent 100-mm plates of MRC-5 cells were labeled with 0.6 mCi/ml "P (orthophosphate) or 150 pCi/ml Tran3?3-label (ICN). After 2.5 h, EGF or TNFcu was added to the media, and cells were incubated an additional 30 min before washing with phosphate-buffered saline and lysis in radioimmune precipitation buffer. Immunoprecipitation of endogenous c-Jun protein, transfer to nitrocellulose, and two-dimensional phosphopeptide mapping were performed as described (32). No differences in c-Jun protein content between treated and untreated cells were detected on ["Slmethionine-labeled gels (data not shown). Plates were exposed to film (Kodak X A R ) for 10 days a t -85 "C.
Gel Mobility Shifts Assays-Electrophoretic mobility shift assays were carried out as previously described (29) using 10 pg of MRC-5 nuclear extract prepared a t various times following cell stimulation. An oligonucleotide representing the consensus AP-1 binding site from the human collagenase gene (34) was synthesized on a Millipore Cyclone Plus synthesizer and used as probe. Polyclonal rabbit anti-c-Jun antiserum or normal rabbit serum was added where indicated, and reactions were continued another 20 min before loading onto 4% non-denaturing polyacrylamide gels. The anti-cJun antisera had no effect on binding when assayed with a binding site for HNF-1 (data not shown).
In Vitro Kinase Assays-The solid-phase J u n kinase assays were carried out as described (26) using a GST-c-Jun (1-223) fusion protein coupled to glutathione-agarose beads as substrate. 50 pg of cellular extract were used as the kinase source. The mutated version (GST-c-J u n Ala 63,731 and GST protein alone were used as control substrates. Preparation of recombinant kinase inactive MEK-1, immunoprecipitation of Raf-1, and Raf-1 immune complex kinase assays were performed as described (35). For the immune complex JNK1 assay, 50 pg of cellular extracts were immunoprecipitated with a specific antiserum raised against recombinant JNKl(271, and the kinase assay was performed as described for Raf-1 with the addition of 20 p~ ATP, using GST-c-Jun as substrate. Phosphorylated proteins were resolved on 10% SDS-PAGE, radioactivity was visualized by autoradiography, and all gels were quantified by scintillation counting of excised bands. In-& ERK Assay-Activation of ERKl and ERK2 was determined using myelin basic protein as the substrate in an in-gel kinase assay as described, using 50 pg of each cell extract (36). The position of the ERKs was confirmed by comparison with both Rainbow molecular weight markers (Amersham Corp.) and to low molecular weight protein markers (Life Technologies, Inc.).

RESULTS
TNFa Stimulates AP-1 Activity and c-dun Phosphorylation-To assess the effect of TNFa, PMA, or EGF on AF-1 transcriptional activity, MRC-5 fibroblasts were transfected with a luciferase reporter gene driven by AP-1 binding sites (BXTRE). As assessed by this assay, treatment with TNFa, EGF, or PMA stimulates the endogenous AP-1 activity of the cells (Fig. lA), as previously demonstrated (9, (Fig. 1B). EGF treatment stimulates luciferase activity t o a lesser degree ( p < 0.05), and PMA has no significant effect. Cotransfection with the dominant negative Ras N17 mutant does not inhibit TNFa stimulation of the reporter gene (data not shown).
Although ERK activation is critical in multiple signaling pathways, its role in TNFa signaling is unknown. To functionally assess ERK activation by TNFa, EGF, or PMA, we transfected the fibroblasts with an expression construct encoding the activation domain of the transcription factor Elk-1 fused to the DNA binding domain of the Gal4 protein. Phosphorylation of Elk-1 by ERKs stimulates the Elk-1 activation domain (30). This Gal-ElkC expression vector is cotransfected with the 5XGal-Luc reporter gene. Treatment with PMA or to a lesser degree EGF enhances the Elk-1 transcriptional activity (Fig.  1C). TNFa weakly stimulates Elk-1 transcriptional activity, consistent with its minimal stimulation of ERK activity, as discussed below.
To assess the role of the c-Jun protein in stimulated AP-1 activity, we investigated the effect of agonist treatment on its post-translational modification. Since TNFa induces c-jun transcription in the absence of de nwo protein synthesis (9), we assessed the effect of TNFa on c-Jun phosphorylation, the only known modification that affects c-Jun activity (16). Treatment of fibroblasts with either TNFa or EGF for 30 min increased I "Plphosphate incorporation into c-Jun relative to untreated cells (Fig. U ) . Previous studies have demonstrated that the phosphorylated forms of c-Jun are present in both bands seen in SDS-PAGE gels (38). Two-dimensional phosphopeptide maps demonstrated that treatment with TNFa increased phosphorylation of Ser-63 and Ser-73 of c-Jun as indicated by the increased intensity of the X and Y phosphopeptides (Fig. 2B). Previous studies demonstrated thatXand Y contain Ser-73 and Ser-63 (22, 23). No effect on the carboxyl-terminal sites was detected. In comparison, treatment with EGF (Fig. 2B)  phopeptides compared with untreated and TNFa-treated cells. This is consistent with activation of ERKs 1 and 2 by EGF (39). The c-Jun phosphorylation pattern of PMA-treated fibroblasts has been previously shown to consist of decreased phosphorylation of sites adjacent to the DNA binding domain (17).
PMA Zncreases AP-1 DNA Binding Activity-Next, we performed gel mobility shift assays with a radiolabeled AP-1 binding site probe incubated with nuclear extracts prepared from untreated and treated fibroblasts. Treatment with the phorbol ester PMA causes a rapid and prolonged increased DNA binding (Fig. 3A), which has previously been demonstrated to result from dephosphorylation of c-Jun serine residues near the DNA binding domain as well as increased Fos synthesis (17). In comparison, treatment with EGF or TNFa results in a smaller increase in the level of the AP-1 binding complexes (Fig. 3, A  and B ) . To demonstrate that c-Jun contributes to the AP-1 complex binding to this probe, extracts from TNFa-treated cells a t a variety of time points were used in mobility shift assays in the presence or absence of c-Jun-specific antiserum. The disruption of the AP-1 DNA complex by c-Jun antiserum (Fig. 3C) demonstrates that c-Jun is a major constituent of the AP-1 complex in control and TNFa-treated cells.
TNFa Induces a Prolonged Activation of JNK-To determine whether TNFa-induced phosphorylation of c-Jun results from either increased JNK activity or decreased phosphatase activity, we used a previously described solid-state assay to measure JNK activity (26). Immobilized GST-c-Jun protein (consisting of the activation domain of c J u n linked to GST) was incubated with fibroblast extracts prepared at different time points following stimulation with EGF, TNFa, or PMA. TNFa treatment results in a biphasic induction of c-Jun kinase activity with an early maximal stimulation of greater than 23-fold, followed by a prolonged plateau stimulation persisting for at least 19 h ( Fig. 4A and data not shown). On the other hand, treatment with PMA or EGF results in a smaller, transient stimulation of JNK activity with rapid return to basal levels (Fig. 4A). Quan- three separate extract preparations withp < 0.04 for TNFa versus control or EGF for all time points. B, quantitation ofA expressed as -fold activation over unstimulated levels. C, an immune complex kinase assay using JNK1-specific antiserum was performed with cell extracts from control cells or cells treated for the indicated times with TNFa, EGF, or PMA. titation of a representative experiment is graphically depicted (as -fold increase over base-line levels) in Fig. 4B.
To directly assess the effect of agonist treatment on JNKl activity, an immune complex kinase assay was performed using JNK1-specific antiserum. Although all three agonists stimulate JNKl activity compared with control cells, TNFa treatment results in the greatest stimulation, followed by EGF and then PMA (Fig. 4C). After 10 min of incubation, JNKl activity is stimulated 9.4-fold by TNFa, 5.8-fold by EGF, and 3.1-fold by PMA.

TNFa Does Not Induce Raf-1 Kinase and Dansiently Induces
ERKs-Although recent studies have demonstrated a signal transduction pathway in which EGF receptor binding to its ligand sequentially activates Ras, Raf-1, and E R G (19,28), the signal transduction pathways for TNFa are largely unknown. Therefore, we assessed Raf-1 kinase activity in MRC-5 fibroblasts after incubation with EGF, TNFa, or PMA. As expected, EGF induces a large and prolonged activation of Raf-1 kinase (Fig. 5 , A and B). PMA produces a lesser stimulation while TNFa does not stimulate Raf-1 kinase activity at any measured time point.
Finally, we examined whether ERKs are activated in parallel with JNK. Using an in-gel myelin basic protein kinase assay with the same extracts assayed for JNK and Raf-1 kinase activity, a completely different temporal pattern of stimulation is observed. TNFa-treated extracts contain elevated ERK activity only at the earliest time points, while PMAor EGF-treated extracts demonstrate markedly elevated ERK activity at early time points followed by prolonged activity that remained greater than 2-fold over base-line levels for 3 h (Fig. 6, A and  B). Improved resolution by a longer electrophoresis time revealed the appearance of two bands consistent with the stimulation of ERKl and ERK2 activity (data not shown). The different effects of TNFa, PMA, and EGF on JNK and ERK activity were seen in several experiments using different extract preparations.
DISCUSSION TNFa actions are initiated by binding to two distinct monomeric TNFa receptors, TNF R-55 and TNF R-75, which are found in nearly all cells (5)(6)(7)(8). Most of the downstream signaling events initiated by TNFa receptor binding has been attributed to the TNF R-55 (2,4,40). A knockout mouse in which the TNF R-55 gene is disrupted is unable to mediate proper responses to microorganisms, including endotoxic shock and clearance of Listeria monocytogenes (4,40). The role of TNF R-75 is more controversial, but biological activities have been attributed to it, including cytotoxicity and cytokine secretion by T cells and stimulation of TGFa mRNA levels (3,(41)(42)(43). Neither TNFa receptor has intrinsic kinase activity, and our understanding of the signal transduction pathways activated by TNFa are incomplete.
TNFa increases the levels of multiple second messengers. TNFa has been demonstrated to activate phosphatidylcholine- specific phospholipase C, which in turn produces diacylglycerol (2). Diacylglycerol is a classical second messenger that activates protein kinase C, and, as expected, protein kinase C is activated by TNFa (9). In addition, TNFa activates sphingomyelinase, which releases the second messenger ceramide (21, and some of the TNFa biological activities have been attributed to elevated ceramide levels (2,44,45). Furthermore, TNFa activates phospholipase 4, which produces the second messenger arachidonic acid with the subsequent synthesis of leukotrienes and prostaglandins (2,lO). Finally, TNFa is reported to inhibit protein phosphatase activity (461, as many of the effects of TNFa are mimicked by the phosphatase inhibitor, okadaic acid (47). TNFa induces the transcription factor NFKB (13,14,44), which requires the TNF R-55 as demonstrated by the inhibition of this pathway in the TNF R-55 knockout mouse (4). TNFa activates NFKB by causing the rapid degradation of IKB, and the released cytoplasmic NFKB then translocates to the nucleus to activate gene transcription (11).
TNFa also stimulates the activity of transcription factor AF"1, which results in increased expression of the cjun and collagenase genes (9) and decreased expression of the elastin gene (48). The present study demonstrates that TNFa causes a prolonged activation of JNK, a small transient activation of ERK, and no increased activation of Raf-1 kinase. This results in the phosphorylation of the cJun activation domain with enhanced c d u n transcriptional activity and little change in c d u n DNA binding activity.
This study demonstrates that the kinetics for the activation of JNK and ERKs by TNFa, EGF, and PMA are different. Re- cent studies indicate that JNK1, the 46-kDa protein that contributes the predominant c-Jun amino-terminal kinase activity, is biochemically and immunologically distinct from ERKl and ERK2 (26). Furthermore, JNK1, although a member of the mitogen-activated protein kinase group, is distantly related to ERKs 1 and 2 (27). Finally, overexpression and activation of ERKs 1 and 2 fail to enhance the transcriptional activation of c-Jun by Ras or Raf (32). Altogether, these results imply that JNK and ERKs are distinct proteins and are regulated by different, possibly parallel, signal transduction pathways. Therefore, the activities of JNK and ERKs can be independently regulated.
The signal transduction pathway by which TNFa activates JNK is unknown. Since oncogenic Ras proteins activate JNK and enhance c-Jun transactivation activity (22,32), the role of Ras and Raf in TNFa signal transduction needs to be assessed. A dominant negative mutant Ras (H-Ras N17) has been used to demonstrate the requirement for endogenous activation in signal pathways, including insulin and platelet-derived growth factor activation of ERK2 (49) and UV induction of NFKB (50).
However, H-Ras N17 alone failed to inhibit EGF stimulation of ERK2 in murine fibroblast cell lines, which required simultaneous inhibition of additional pathways (51). Since H-Ras N17 fails to inhibit TNFa stimulation of c-Jun transcriptional activity, this implies a Ras-independent pathway for TNFa signaling. Activated Ras stimulates ERKs in a variety of cell types (49, 52-54). Ras directly interacts with the amino-terminal regula-tory domain of Raf-1, which in turn leads to phosphorylation and activation of MEK-1 and MEK-2 in a variety of cells, including fibroblasts (55-58). Recent work demonstrates that Raf activity is a necessary component for activation of ERK by oncogenes, serum, and PMA (59). Our study confirms the activation of Raf and ERKs by PMA and EGF. However, TNFa does not stimulate Raf-1 kinase activity and only transiently stimulates ERK activity, presumably through a Raf-independent pathway such as MEKK (60). Thus, this study provides evidence that TNFa activates JNK through a Ras-, Raf-, and ERKindependent pathway. Addendum-While this paper was under review, Kyriakis et al. (61) reported a family of cJun kinases that includes JNKl and is activated by TNFa.