Identification of a Tumor Necrosis Factor-responsive Element in the Tumor Necrosis Factor a Gene*

The regulation by tumor necrosis factor a (TNF) of its own promoter has been investigated by transient transfection and nuclear protein binding assays. In human K652 erythroleukemia cells TNF produced an 8-lo-fold activation of the human TNF promoter linked to the chloramphenicol acetyltransferase gene. The TNF-responsive element was localized to the -125 to -82 region by examining the TNF activation in 5’-deletion or site-directed mutants of the TNF promoter and by demonstrating that the -125 to -82 fragment confers TNF responsiveness to the thymidine kinase promoter. This region contains a palindrome, 5‘ TGAGCTCA 3‘, that resembles the consensus binding sequences for the transcription factors, activator pro- tein- 1 (AP- l), cyclic AMP-responsive element binding protein (CREB), and activation transcription factor (ATF). An internal deletion in the palindrome abol- ished the TNF responsiveness, whereas known AP-1 and CREB/ATF elements were unresponsive to TNF. In band shift analyses a nuclear factor from U937 cells specifically bound to the -125 to -82 TNF-responsive fragment in or near the palindromic sequence. Oligo- nucleotides containing AP-1 or CREB/ATF sites did not effectively compete for the binding, indicating that the U937 cell factor is different from these factors. did not affect binding of the cell factor, anti-c-jun antiserum did block its binding, that c-jun or a protein These results suggest that the TNF-responsive element not that a novel DNA binding factor is important for con- stitutive and inducible TNF gene expression.

The regulation by tumor necrosis factor a (TNF) of its own promoter has been investigated by transient transfection and nuclear protein binding assays. In human K652 erythroleukemia cells TNF produced an 8-lo-fold activation of the human TNF promoter linked to the chloramphenicol acetyltransferase gene. The TNF-responsive element was localized to the -125 to -82 region by examining the TNF activation in 5'deletion or site-directed mutants of the TNF promoter and by demonstrating that the -125 to -82 fragment confers TNF responsiveness to the thymidine kinase promoter. This region contains a palindrome, 5' TGAGCTCA 3', that resembles the consensus binding sequences for the transcription factors, activator protein-1 (APl), cyclic AMP-responsive element binding protein (CREB), and activation transcription factor (ATF). An internal deletion in the palindrome abolished the TNF responsiveness, whereas known AP-1 and CREB/ATF elements were unresponsive to TNF. In band shift analyses a nuclear factor from U937 cells specifically bound to the -125 to -82 TNF-responsive fragment in or near the palindromic sequence. Oligonucleotides containing AP-1 or CREB/ATF sites did not effectively compete for the binding, indicating that the U937 cell factor is different from these factors. Anti-c-fos antiserum did not affect binding of the U937 cell factor, whereas anti-c-jun antiserum did block its binding, indicating that either c-jun or a protein antigenically related to c-jun is a component of the factor. These results suggest that the TNF-responsive element is not activated by AP-1 or CREB in U937 cells and that a novel DNA binding factor is important for constitutive and inducible TNF gene expression.
Tumor necrosis factor a (TNF),' also known as cachectin,  is a pleiotropic cytokine that is released from macrophages in response to tissue damage, bacterial endotoxin, viruses, and other cytokines (1)(2)(3)(4). It is one of the principal cytokines participating in immune and inflammatory processes that protects the host against tissue destruction and infections and facilitates tissue remodeling. T N F is selectively cytostatic or cytotoxic to some tumor cells, suggesting that it may function as an endogenous antitumor agent (1)(2)(3)(4). Despite these beneficial roles, excessive T N F production is associated with several pathological conditions (1)(2)(3)(4). Chronic exposure to T N F leads to cachexia, which is manifested by severe wasting of body tissue, anorexia, and anemia and is observed in patients with cancer, acquired immunodeficiency syndrome (AIDS), or chronic infections (1)(2)(3)(4) and in animals implanted with tumor cells that secrete TNF ( 5 ) . T N F also mediates endotoxic shock following infections with certain bacteria (1)(2)(3)(4) and may contribute to the pathogenesis of AIDS (6,7) and adult respiratory distress syndrome (8).
The cellular actions of TNF that mediate the diverse physiological responses and contribute to the pathogenesis of cachexia, endotoxic shock, and other conditions are largely unknown. However, a major site where T N F probably acts is at the level of gene transcription. Torti et al. (9) first reported that TNF inhibits the transcription of adipocyte genes. Since this discovery TNF has been shown to alter the transcription of numerous vital genes that undoubtedly contribute to its physiological and pathological effects. For example, T N F increases the transcription of proto-oncogenes (10, 11) and genes for cytokines (12,13), ferritin heavy chain (14), and major histocompatibility complex antigens (15,16). TNF inhibits the expression of collagen (17), pulmonary surfactant (18), and genes involved in lipogenesis (9, 19) and muscle differentiation (20). The mechanisms for the gene regulatory effects of T N F have not been examined extensively. However, it has been reported that TNF activates the promoters for human immunodeficiency virus (6, 7), interleukin 8 (13), and the interleukin 2 receptor (21) by inducing the transcriptional factor, NFKB. T N F also stimulates the collagenase promoter by inducing AP-1 (22), and the TNF induction of the interleukin 8 gene involves a cis-regulatory enhancer binding protein-like factor (13). Thus, TNF may regulate gene expression by activating a variety of DNA binding proteins.
In this study, we investigated the regulation of the human T N F promoter by T N F for several reasons. First, TNF has been shown to increase its own synthesis (23). Therefore, agents that release T N F from macrophages may regulate T N F gene expression via autoregulation by TNF. Second, the TNF promoter may provide a general model for understanding how T N F regulates gene transcription. Third, despite the clinical importance of TNF and the need for precise control of T N F production, virtually nothing is known about the transcriptional factors that regulate the human T N F promoter. Several recent reports indicate that phorbol esters activate the transfected human TNF promoter in human cell lines (24,25), but the factors that mediate constitutive expression and phorbol ester responsiveness were not defined in these studies. Lipopolysaccharride has been shown to activate the mouse T N F promoter by activating NFKB (26), but NFKB does not participate in the lipopolysaccharide-induced activation of the human TNF promoter (27).
In the present study, we demonstrate that TNF activates its promoter in K652 and U937 cells and that the core of an eight-nucleotide palindrome sequence, 5' TGAGCTCA 3', located at -108 to -101 is essential for this activation. This palindrome resembles the consensus sequences that bind the factors AP-1 and CREB. However, both our functional and DNA binding studies suggest that the TNF activation of the T N F promoter is not mediated by these factors but may utilize a novel DNA binding protein.

EXPERIMENTAL PROCEDURES
Materials-The human eyrthroleukemia K652 and human promonocyte U937 cell lines, RPMI-1640, Dulbecco's phosphate-buffered saline (PBS), glutamine, and penicillin-streptomycin were obtained from the cell culture facility at the University of California, San Francisco. Biobrene was purchased from Applied Biosystems. Recombinant T N F was a generous gift from Dr. Jeffrey Andresen (Amgen Biologicals). Riboprobe System and rabbit reticulocyte lysate were purchased from Promega.
Plasmid Constructions-A Pstl to Aha11 fragment (-1044 to +93) was excised from the human TNF gene (pLT) (28), which was generously provided by Dr. David Goeddel (Genentech) and cloned into the PstI and AccI sites in a polylinker in a pUC19-based vector containing the rabbit @-globin polyadenylation signals cloned upstream in the Hind111 site and the chloramphenicol acetyltransferase (CAT) gene and an SV40 intron and polyadenylation signals cloned downstream in the BamHI site (29). This vector had also been deleted in the plasmid sequences from the polylinker Asp-718 site to the AatII site.2 The 5"deletions were constructed using unique restrictions sites, BstXI for the -345 deletion, ApaI for the -125 deletion, and SstI for the -100 deletion. After digestion with the restriction enzymes, the plasmids were blunted with T4 DNA polymerase and then ligated with phosphorylated HindIII linkers. The DNA fragments and double-stranded oligonucleotides shown in Fig. 3 were ligated into a blunted Sal1 site in the polylinker upstream of -32 to +45 herpes simplex virus thymidine kinase promoter, linked to the CAT-SV40 gene in the same vector used for the TNF promoter.
Cell Culture, Electroporation, and CAT Assays"K652 and U937 cells were maintained and subcultured in RPMI-1640 containing 10% fetal bovine serum, 2 mM glutamine, 50 units/ml penicillin, and 50 pg/ml streptomycin. For transfections, cells were collected by centrifugation and washed once with Dulbecco's PBS and resuspended in Dulbecco's PBS (0.5 m1/1.5 X lo' cells) containing 0.1% dextrose and 10 pg/ml Biobrene, 15 pg of reporter plasmid, and 5 pg of a plasmid that expresses @-galactosidase to control for transfection efficiency (30). The suspended cells were transferred to a cuvette and kept a t room temperature for 5 min. The cells were electroporated using a Bio-Rad gene pulser set a t 300 V and 960 microfarads and then kept at room temperature for 10 min. The electroporated cells were then transferred to 6.5 ml of RPMI-1640 containing 10% fetal bovine serum, 2 mM glutamine, 50 units/ml penicillin, and 50 pg/ml streptomycin. The cells were kept in the medium for 10 min, and then resuspended and plated a t 1 ml/dish in 6-well multiplates. Human TNF was added to triplicate wells to a final concentration of 5 ng/ml, and the cells were incubated for 18 h.
The cells were harvested by transferring them to a 1.5-ml microcentrifuge tube and pelleting by gentle centrifugation in a microcentrifuge for several seconds. After the medium was aspirated the cell pellets were lysed by the addition of 150 pl of 0.25 M Tris-HC1, pH 7.6, containing 0.1% Triton X-100. The cellular lysates were assayed for @-galactosidase activity (31) and CAT activity using a liquid scintilation method (32). The TNF promoter activity is expressed in arbitrary units as CAT activity divided by @-galactosidase activity.
Preparation of Nuclear Extracts"U937 cells were collected by centrifugation and washed 2 times with ice-cold PBS. The cell pellet was washed once with 10 ml of buffer A (10 mM Hepes, pH 7.9, 0.5 mM spermidine, 0.15 mM spermine, 10 mM potassium chloride, 1.5 mM EGTA, and 0.3 M sucrose). After centrifugation, the cell pellets were lysed in 1.25 ml of buffer A containing 0.5% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 5 pg/ml leupeptin, 5 pg/ml pepstatin, and 5 pg/ml chymotrypsin. The cells were incubated on ice for 10 min and then centrifuged at 4,000 rpm for 10 min a t 4 "C. The supernatant fraction was discarded, and the ' B. L. West, unpublished data.  1. TNF activates the human TNF promoter in K652 cells. K652 cells were cotransfected by electroporation with a series of 5"deletion mutants of the TNF promoter and a plasmid that expresses @-galactosidase to control for transfection efficiency. Triplicate cell cultures were treated without or with 5 ng/ml T N F for 18 h. Cellular lysates were assayed for CAT activity and @-galactosidase activity. T N F promoter activity is expressed as CAT activity divided by @-galactosidase actvity. nuclear proteins were extracted by resuspending the pellet in 0.5 ml of buffer B (10 mM Hepes, p H 7.9, 0.4 M potassium chloride, 0.1 mM EGTA, 1.5 mM magnesium chloride, 10% glycerol, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 5 Fg/ml leupeptin, 5 pg/ml pepstatin, and 5 pg/ml chymotrypsin). The nuclear extract was centrifugedat 18,000 rpm for 30 min at 4 "C. The supernatant fraction was collected and centrifuged for 15 min in a microcentrifuge a t 4 "C, and then aliquoted and stored at -70 "C.

Control TNF Fold
Band Shift Binding Assays-Both the radiolabeled and cold competitor DNA fragments were obtained by releasing the HindIII to XbaI fragments from pUC19 vectors into which the test binding sites had been cloned into a blunted Sal1 site. The sequences of the test binding sites are shown in Fig. 3. Radiolabeling was done by 5' phosphorylation using T4 polynucleotide kinase and [-y-"P]ATP. DNA binding reactions were performed in a 20-pl volume containing a 10,000-cpm probe in a final concentration of 12 mM Hepes, pH 7.6, 48 mM potassium chloride, 0.8 mM EDTA, 4 mM magnesium chloride, 10% glycerol, and 4 pg of poly(d1-dC). The binding reaction was initiated by the addition of 1 pl of nuclear extract, and the samples were then incubated for 15 min a t room temperature. The samples were placed on ice and then loaded on a 6% polyacrylamide gel and electrophoresed a t 200 mV with running buffer consisting of 25 mM Trizma base, 25 mM sodium borate, and 1 mM EDTA.
In Vitro Transcription and Translation-A plasmid containing the rat c-fos cDNA was linearized using EcoRI, and capped mRNA was synthesized (Promega Riboprobe System) with T7 RNA polymerase. The reactions were incubated for 1-2 h at 37 "C. RNA was treated with RNase-free DNase, extracted with phenol/chloroform, dissolved in H20, and stored at -70 "C. c-fos RNA (20-40 ng/pl) was translated in a rabbit reticulocyte lysate system containing 40 pM zinc chloride, a methionine-free amino acid mixture, and 20 p~ unlabeled methionine. The translation reactions were incubated at 30 "C for 60-120 min. For band shift analysis, purified bacterially expressed c-junA (33) was preincubated with translated c-fos for 30 min a t 37 "C. In the antibody experiments U937 cell nuclear extracts or AP-1 were preincubated with the antiserum for 30 min at room temperature prior to binding to the radioactive probes. The anti-c-jun antiserum was generated from bacterially expressed ~-j u n A ,~ a n d t h e anti-c-fos antiserum was generated from the N-terminal amino acids number 4 to 17 (34).

RESULTS AND DISCUSSION
T N F activated a -1044 T N F promoter by 10-fold. The activation by T N F was similar in promoters deleted to -345 and -125 but was abolished by deleting the TNF promoter to -100 (Fig. 1). These results suggest that the TNF promoter contains a TNF-responsive element (TNF-RE), located in the -125 to -101 region. Nucleotides -108 to -101 of the promoter consist of the eight-nucleotide palindrome sequence, 5' TGAGCTCA 3'. A four-nucleotide deletion (AGCT) within the -108 to -101 region of the promoter markedly reduced basal activity and abolished the response to TNF, demonstrating that this sequence is critical for both of these activities.
The palindromic sequence is similar to the consensus motifs '' V. Baichwal, unpublished data.
for three previously characterized families of DNA binding proteins, c-jun, CREB, and ATF. AP-1 is a heterodimeric complex comprised of c-jun-and c-fos-related factors (35-37). AP-1 regulates gene transcription by binding to the DNA sequence 5' TGACTCA 3', which is known as the 12-0tetradecanoylphorbol-13-acetate-responsive element (TRE) since AP-1 is activated by phorbol esters (35). The CREB and the ATF families both bind to the sequence, 5' TGACGTCA 3 ' , and are activated by cyclic AMP (38) and E l a from adenovirus (39), respectively. Of these three classes, only AP-1 has been reported to be activated by T N F (22). Therefore, a possible role for AP-1 in the TNF induction of the TNF promoter was examined.
Three copies of the human metallothionein AP-1 element (TRE) were cloned upstream of the -32 to +45 thymidine kinase promoter, which does not contain any upstream enhancer elements and is inactive and not inducible by T N F in K652 and U937 cells. If AP-1 mediates the effects of TNF, then authentic AP-1 binding sites should confer T N F responsiveness to a heterologous promoter. For comparison we also transferred the -125 to -82 fragment of the TNF promoter and the somatostatin CREB binding element upstream of the thymidine kinase promoter. As shown in Fig. 2 three copies of the -125 to -82 T N F fragment conferred a substantial increase in basal activity, as well as a 7or 11-fold stimulation by TNF. The -125 to -82 T N F fragment containing an internal deletion of four nucleotides (AGCT) within the palindrome was unable to confer either basal or T N F stimulation. The addition of three metallothionein TRE sites upstream of the thymidine kinase promoter markedly increased basal activity. However, this construction responded to T N F with a minor decrease in CAT activity in both K652 and U937 cells. This result may indicate that TNF regulation of AP-1 is more complex than previously appreciated (22), possibly depending on either tissue specificity and/or promoter context. CRE binding sites did not enhance basal activity or confer T N F responsiveness to the thymidine kinase promoter. These results demonstrate that only the native -125 to -82 region containing the intact palindromic sequence 5' TGAGCTCA 3' is capable of conferring T N F responsiveness to the heterologous thymidine kinase promoter even though the TRE and CRE sequences are almost identical to the T N F promoter palindromic sequence. These observations also suggest that AP-1 and CREB do not mediate the effects of T N F in these cells.
We next used a band shift assay to detect proteins in U937 or K652 cells that bind to the -125 to -82 region of the TNF  Fig. 3 were cloned upstream of -32 to +45 TKCAT. The plasmids contained three copies of TNF-RE, mutated TNF-RE, or a n oligonucleotide containing the human metallothionein T R E or two copies of an oligonucleotide containing the somatostatin CRE upstream of TKCAT. K652 or U937 cells were transfected by electroporation using 15-pg TKCAT plasmids. Triplicate cell cultures were treated without or with 5 ng/ml T N F for 18 h. Cellular lysates were assayed for CAT activity and P-galactosidase actvity. TNF promoter activity is expressed as CAT activity divided by @-galactosidase activity. T K , thymidine kinase.
promoter. Several shifted bands were observed, one of which was specific since this complex (Fig. 3A, lane 2, solid arrow) was inhibited by the addition of excess unlabeled -125 to -82 T N F fragment (lane 3 ) . Treatment of U937 cells with T N F for 3 or 18 h did not alter the pattern observed (data not shown). These results indicate that the factor that binds to the TNF promoter fragment is expressed constitutively and is not inducible by TNF. The mutated -125 to -82 TNF fragment (lane 4 ) did not compete for binding, indicating that the deleted four nucleotides are critical for binding of the factor to the promoter. Oligonucleotides containing the TRE (lane 5 ) or CRE (lune 6) competed for binding minimally, indicating that the binding site for the U937 cell factor differs from AP-1 or CREB. Similar results were obtained when the band shift assays were performed with nuclear extracts prepared from K652 cells (data not shown). DNase I footprinting of the TNF promoter using affinity-purified c-junA protein revealed only a weak footprint over the palindromic region (data not shown), confirming that the proximal promoter does not contain a high affinity AP-1 binding site.
Although the functional and competition binding studies rule out that the factor from U937 cells is AP-1, a possible relationship of the factor to AP-1 was further examined by comparing their electrophoretic mobilities and anti-c-jun and anti-c-fos antiserum to determine if c-jun or c-fos are components of the factor that binds to the TNF palindrome. AP-1 was generated by binding purified bacterially expressed cjun to in vitro-synthesized c-fos. Fig. 3B shows that c-jun.cfos (lane 2, open arrow) binds very weakly to the -125 to -82 TNF fragment and the band occurs at a lower position in the gel compared with the factor derived from U937 cells (lane 1, solid arrow). The difference in migration indicates that the U937 factor cannot be identical to the c-jun.c-fos heterodimer. Binding of c-jun.c-fos (Fig. 3C, lane 4 ) to an oligonucleotide containing the TRE site was much stronger than to the T N F promoter (Fig. 3B, lane 2). As shown in Fig. 3C, anti-cfos antiserum did not block binding or supershift the factor bound to the -125 to -82 T N F fragment (lane 2), indicating that c-fos is not a component of the complex. Anti-c-jun antiserum did inhibit binding of the U937 cell factor (lane 3 ) . Therefore, the U937 cell factor either contains c-junA or an antigenically related protein such as a CREB/ATF protein (39, 40). In a control experiment, anti-c-jun antiserum abolished binding of the c-jun. c-fos complex to the TRE (lune 6), whereas anti-c-fos antiserum produced a supershift of the complex (lune 5 ) , demonstrating that the antibodies are capable of recognizing c-jun and c-fos.
In this study, we demonstrated that the -125 to -82 T N F promoter region containing the 5' TGAGCTCA 3' sequence functions as a constitutive enhancer and a TNF-responsive element in K652 and U937 cells. This sequence is nearly identical to the consensus AP-1 site, which was previously shown to mediate the activation of the collagenase promoter by T N F (22). However, the results of the present study indicate that AP-1 is not the factor from U937 cells which binds to the TNF palindrome and mediates the TNF induction of the TNF promoter. Interestingly, we did find that antiserum to c-junA blocked binding of the factor to the TNF palindrome, suggesting that c-junA or another member of the c-jun family or CREB/ATF families may be a component of the factor complex. If c-jun is a component of the complex it apparently dimerizes with a factor other than c-fos, possibly a tissue-specific factor which alters the specificity for c-jun binding to DNA. A model for this idea comes from the observation that the optimal binding of the c-jun.c-fos heterodimer is 5' TGACTCA 3', whereas when c-jun is com-

CRE G G G G G C G C C T C C T T G G C~G A C G T C A G A G A G A G A G
plexed with CREB-binding protein 1 the preferred binding site is CRE sequence, 5' TGACGTCA 3' (41). The purification and cloning of a cDNA for the factor will be necessary to establish its composition, its relationship to other factors, and the mechanism of regulation by TNF. The elucidation of the transcription factors that are regulated by TNF should provide a greater understanding of how TNF exerts its pleiotropic effects on physiological and pathological processes.