Synergy between interferon-gamma and tumor necrosis factor-alpha in transcriptional activation is mediated by cooperation between signal transducer and activator of transcription 1 and nuclear factor kappaB.

Interferon-gamma (IFNgamma) and tumor necrosis factor-alpha (TNFalpha) cooperate to induce the expression of many gene products during inflammation. The present report demonstrates that a portion of this cooperativity is mediated by synergism between two distinct transcription factors: signal transducer and activator of transcription 1 (STAT1) and nuclear factor kappaB (NF-kappaB). IFNgamma and TNFalpha synergistically induce expression of mRNAs encoding interferon regulatory factor-1 (IRF-1), intercellular adhesion molecule-1, Mig (monokine induced by gamma-interferon), and RANTES (regulated on activation normal T cell expressed and secreted) in normal but not STAT1-deficient mouse fibroblasts, indicating a requirement for STAT1. Transient transfection assays in fibroblasts using site-directed mutants of a 1.3-kilobase pair sequence of the IRF-1 gene promoter revealed that the synergy was dependent upon two sequence elements; a STAT binding element and a kappaB motif. Artificial constructs containing a single copy of both a STAT binding element and a kappaB motif linked to the herpes virus thymidine kinase promoter were able to mediate synergistic response to IFNgamma and TNFalpha; such response varied with both the relative spacing and the specific sequence of the regions between these two sites. Cooperatively responsive sequence constructs bound both STAT1alpha and NF-kappaB in nuclear extracts prepared from IFNgamma- and/or TNFalpha-stimulated fibroblasts, although binding of individual factors was not cooperative. Thus, the frequently observed synergy between IFNgamma and TNFalpha in promoting inflammatory response depends in part upon cooperation between STAT1alpha and NF-kappaB, which is most likely mediated by their independent interaction with one or more components of the basal transcription complex.

Intercellular communication by cytokines during an inflammatory reaction is integral to the subsequent orchestration and resolution of the response. IFN␥ 1 and TNF␣ are pleiotrophic cytokines that play often critical roles in this process (1,2). Although both cytokines independently exert a number of biological activities in a cell type-specific fashion, they have been shown in many circumstances to function cooperatively or antagonistically in controlling expression of a variety of cytokines and cell surface molecules (3)(4)(5)(6)(7).
Much recent work on cytokine-mediated intracellular signaling pathways has provided a general paradigm for the molecular mechanisms by which extracellular signals induce transcription of target genes (8 -11). A variety of cytokines, growth factors, and hormones trigger phosphorylation of latent cytoplasmic transcription factors termed signal transducers and activators of transcription (STATs) via one or more members of the Janus (Jak) family of protein tyrosine kinases. Tyrosinephosphorylated STATs assemble in dimeric or oligomeric form, translocate to the nucleus, and bind to specific DNA sequence motifs or STAT binding elements (SBEs) (12). IFN␥ has been shown to induce tyrosine phosphorylation of STAT1␣, and a homodimeric form of STAT1␣ binds to the IFN␥-activation sequence (13), an SBE that has been identified as a critical sequence motif involved in the transcriptional activation of many IFN-inducible genes including the IRF-1 and ICAM-1 genes (14 -17).
The B sequence motif has been shown to be an essential cis-acting regulatory element for mediating the TNF-, interleukin-1-, and lipopolysaccharide-induced transcriptional activation of multiple cytokines and cell surface molecules (18 -20). Although this sequence motif is recognized by members of the Rel homology family, including NF-B1 (p50/p105), NF-B2 (p52/p100), RelA, c-Rel, and RelB, various forms of the B sequence motif have been shown to exhibit differential affinity for and functional response to different dimeric combinations of Rel family proteins. Cell type-specific expression of the Rel family members also mediates specificity for B-dependent gene expression. Furthermore, members of the Rel family have been shown to physically and functionally interact with members of other transcription factor families (21)(22)(23). The combination of these variables generates high potential for diversity in the control of gene expression during inflammation.
Components of the JAK-STAT and the B signaling pathways appear to be indispensable for stimulus-dependent, transcriptional activation of many inflammatory genes. Furthermore, SBE and B motifs are found in the promoter regions of many inflammatory genes. Many studies have reported functional synergy between TNF␣ and IFN␥ in promoting inflammatory function and gene expression, some of which could involve an interplay between STAT1 and B binding factors (3)(4)(5)(6). The present study was undertaken to determine whether IFN␥-activated STAT1 can cooperate with TNF␣-induced NF-B to promote enhanced transcription. The results show that IFN␥ and TNF␣ synergize to induce expression of several genes that contain both SBE and B motifs. The findings indicate that both the SBE and B motifs are required for cooperativity and that the synergistic function of STAT1␣ and NF-B appear to result from independent activation and recognition of cognate nucleotide sequence motifs.

EXPERIMENTAL PROCEDURES
Reagents-Dulbecco's modified Eagle's medium, minimum essential medium nonessential amino acid solution, sodium pyruvate, and antibiotic were obtained from Life Technologies, Inc. Fetal bovine serum was purchased from Bio Whittaker (Walkersville, MA). DEAE-dextran and polydeoxyinosinic-deoxycytidylic acid (poly(dI-dC)) were purchased from Pharmacia LKB Ltd. (Uppsala, Sweden Cell Culture-Fibroblasts from STAT1-deficient and wild type mice were prepared as described previously (24). These cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 10 mM nonessential amino acid solution, 10 mM sodium pyruvate, 20 mM of L-glutamine. NIH3T3 fibroblasts were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, glutamine, penicillin, and streptomycin (complete medium) and subcultured twice weekly. Prior to use in experiments, the cells were grown to confluence in 100-or 150-mm diameter culture dishes and then transferred to medium containing 0.2% fetal bovine serum for 24 h in order to deprive growth factor.
Preparation of Plasmid DNA-A cDNA encoding mouse IRF-1 (25) was cloned from a mouse macrophage cDNA library (26) using a reverse transcriptase-PCR fragment as a probe as described previously. 2 The plasmid encoding the cDNA for mouse ICAM-1 was obtained from the American Type Culture Collection (Rockville, MD) (27). cDNA fragments for mouse Mig and RANTES were prepared by reverse transcriptase-PCR using a set of primers corresponding to the mouse Mig and RANTES cDNA sequences obtained from the GenBank TM data base (28 -30) and subcloned into pBluescript (Stratagene, La Jolla, CA). The nucleotide sequences were independently confirmed. The plasmid encoding GAPDH was obtained from Dr. David Stern (Columbia University, New York, NY). Methods for plasmid DNA preparations were as described in Sambrook et al. (31). One g of plasmid DNA or 100 ng of PCR products were radiolabeled by random priming with [␣-32 P]dCTP. The resultant specific activity was approximately 10 8 cpm/g, which was used at 10 7 cpm/blot.
Preparation of RNA and Northern Hybridization Analysis-Each assay utilized confluent monolayer of fibroblasts cultured in 100-mm diameter plastic Petri dishes for preparation of total RNA. After treatment of the cells with the indicated stimuli, total cellular RNA was extracted by the guanidine isothiocyanate-cesium chloride method (32). Samples of total RNA (5 g) were separated on a 1% agarose, 2.2 M formaldehyde gel and subsequently blotted onto MAGNA nylon membrane with 20 ϫ SSC by capillary transfer according to previously published methods (31). The RNA was cross-linked to the membrane with a UV cross-linker (Stratagene). The blots were prehybridized for 8 -12 h at 42°C in 50% formamide, 1% SDS, 5 ϫ SSC, 1 ϫ Denhardt's solution (0.02% Ficoll, 0.02% bovine serum albumin, 0.02% polyvinylpyrrolidone), 0.25 mg/ml denatured salmon sperm DNA, and 50 mM sodium phosphate (pH 6.5) and then hybridized with 1 ϫ 10 6 cpm/ml of radiolabeled cDNA plasmid probe at 42°C for 16 -24 h. After hybridization, blots were washed with 0.1% SDS, 2 ϫ SSC for 30 min at room temperature followed by two washes at 55°C. The blots were then exposed using XAR-5 x-ray film with intensifying screens at Ϫ70°C.
Preparation of Reporter Gene Plasmid DNA-The luciferase reporter constructs containing the 1.3-kb IRF-1 promoter was kindly provided by Dr. Bryan Williams (Department of Cancer Biology, Cleveland Clinic Foundation). The SBE site at positions Ϫ123 to Ϫ113 and the B site at positions Ϫ49 to Ϫ40 of the IRF-1 promoter (14) were respectively mutated in the 1.3-kb 5Ј-flanking sequence of the IRF-1 gene by oligonucleotide-directed, site-specific mutagenesis as described previously (33). The mutant sequence utilized for the SBE and the B were TTC-CCtcc and GtGGAATCaC, respectively. Lowercase letters represent the mutant nucleotides.
One or two copies of the IRF-1 SBE or the IP-10 B2 were placed in front of the Ϫ105 or the Ϫ81 base pair herpes simplex virus-thymidine kinase (TK) promoter (34) linked to the chloramphenicol acetyl transferase (CAT) gene (pTK-105 CAT) (35) or the luciferase gene (pTK-105 Luc or pTK-81 Luc) (36). These constructs were prepared by ligating the synthetic oligonucleotides (see below) into restriction enzyme sites of the reporter plasmids. To generate constructs containing different nucleotide spacing between the IRF-1 SBE and the B2 sites, one or more SalI linkers (GGTCGACC) were placed between the two sites. The sequences of these reporter gene constructs were confirmed.
Transient Transfection-Transfection of the luciferase and the CAT reporter genes into fibroblasts or NIH3T3 cells were as described previously (5). Briefly, the cells were seeded at a density of 3 ϫ 10 6 cells/150-mm diameter dish 24 h prior to transfection. 30 g of reporter luciferase construct plasmid DNA were transfected by the DEAE-dextran method (300 g/ml DEAE-dextran) for 20 min at room temperature. After the incubation, the cells were subjected to dimethyl sulfoxide shock for 1 min (10% dimethyl sulfoxide in phosphate-buffered saline), washed with phosphate-buffered saline, replenished with fresh culture medium, and cultured for 2 h. To standardize transfection efficiencies, the transfected cells were then harvested in trypsin-EDTA solution, pooled, and seeded in four 100-mm diameter Petri dishes. The cells were cultured in medium containing 0.2% fetal bovine serum for 24 h to deprive growth factors and then stimulated with IFN␥ and/or TNF␣ for 16 h for the CAT reporter gene and for 8 h for the luciferase reporter gene, respectively. After stimulation, the cells were washed and extracted in lysis buffer (Promega), and luciferase activity was assayed using reagents provided by Promega according to the manufacturer's instructions. Twenty g of extract protein were utilized in each assay. CAT activity was assessed by determination of the conversion of [ 14 C]chloramphenicol into acetylated forms detected by thin layer chromatography as described previously (35). The acetylated products were quantified using a phosphorescence detection system (Molecular Dynamics, Sunnyvale, CA).
Preparation of Oligonucleotides and PCR-amplified DNA-The following oligonucleotides were used in this study.
The nucleotide sequences of IRF-1 SBE and B were taken from Sims et al. (14). The IP-10 B2 sequence was taken from Ohmori and Hamilton (33,37). Lowercase letters represent the bases included for creat-2 Y. Ohmori and T. A. Hamilton, submitted for publication.
ing restriction sites. Underlined sequences represent the consensus sequences for the SBE and B elements, respectively. Boldface type indicates the substituted bases for mutation. Oligonucleotides were synthesized using an Applied Biosystem DNA synthesizer (model 381A) or obtained from Ransom Hill Bioscience Inc. (Ramona, CA). Doublestranded oligonucleotides were prepared by annealing the complementary single strands. A DNA fragment corresponding to the region between Ϫ129 and Ϫ37 of the IRF-1 promoter (14) was generated by PCR using a sense oligonucleotide of the IRF-1 SBE and an antisense oligonucleotide of the IRF-1 B as primers, and the luciferase reporter plasmid containing the 1.3-kb IRF-1 promoter was used as a template.
A mutant fragment was also generated by using a sense oligonucleotide of mut1 SBE and antisense oligonucleotide mutIRF-1B as described above. Double-stranded oligonucleotides were radiolabeled with the Klenow fragment of DNA polymerase I and [␣-32 P]dCTP in a fill-in reaction for 5Ј protruding ends. PCR-amplified DNA fragments were radiolabeled with T4 kinase and [␥-32 P]ATP. Preparation of Nuclear Extracts-Nuclear extracts were prepared using a modification of the method of Dignam et al. (38) as described previously (5,37). After stimulation, the cells were washed with ice-cold phosphate-buffered saline three times, harvested, and resuspended in 300 l of hypotonic buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml of leupeptin, antipain, aprotinin, and pepstatin) for 10 min on ice. The cells were then lysed in 0.6% Nonidet P-40 by vortexing for 10 s. Nuclei were separated from cytosol by centrifugation at 12,000 ϫ g for 30 s, washed with 300 l of buffer A, and resuspended in buffer C (20 mM HEPES, pH 7.9, 25% glycerol, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml of leupeptin, antipain, aprotinin, and pepstatin) and briefly sonicated on ice. Nuclear extracts were obtained by centrifugation at 12,000 ϫ g for 10 min. Protein concentration was measured by the method of Bradford (39) by using the protein dye reagent (Bio-Rad).
Electrophoretic Mobility Shift Assay (EMSA)-For binding reactions, nuclear extracts (5 g of protein) were incubated in 12.5 l of total reaction volume containing 20 mM HEPES (pH 7.9), 50 mM KCl, 0.1 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 200 g/ml bovine serum albumin, and 1.25 g of poly(dI-dC) for 15 min at 4°C. The 32 P-labeled oligonucleotide (ϳ5 ϫ 10 5 cpm) was then added to the reaction mixture and incubated for 20 min at room temperature. In some experiments, antibodies against NF-B1, RelA, c-Rel, STAT1, STAT3, and Sp1 were included in the reaction mixture. The reaction products were analyzed by electrophoresis in a 4% polyacrylamide gel with 0.25 ϫ TBE buffer (22.3 mM Tris, 22.2 mM borate, 0.5 mM EDTA). The gels were dried and analyzed by autoradiography.

STAT1
Is Essential for the Synergistic Induction of IFN␥ and TNF␣-mediated Gene Expression-IFN␥ and TNF␣ have been shown to cooperatively regulate transcription of many inflammatory genes (3)(4)(5)(6)40). Previous studies have demonstrated that the IFN␥-induced transcriptional activation of the IRF-1, ICAM-1, and Mig genes depends upon IFN␥ response elements or SBEs in the promoter region of the genes (Fig. 1), which are recognized by STAT1 or STAT1-containing factor(s) (14 -17, 41-43). The promoter regions of these IFN␥-inducible genes also contain one or more B sequence motifs, and their transcriptional activation in response to TNF␣ is dependent upon activation of B binding activities (44,45). On the basis of these observations, we postulated that IFN␥-induced STAT1 and TNF␣-induced NF-B cooperatively regulate transcription of genes containing both SBE and B motifs. Analysis of RANTES gene expression was also included, since IFN␥ and TNF␣ can cooperatively induce expression of this chemokine gene, although no SBE has been identified in the promoter (6, 30, 46 -48). Levels of endogenous mRNA expression were determined in fibroblasts from wild-type mice or from mice in which the STAT1 gene has been deleted through homologous recombination (24). Serum-starved cultures were stimulated with IFN␥ and TNF␣ either alone or in combination for 3 h prior to isolation of RNA and Northern analysis. While the sensitivity of normal cells to IFN␥ or TNF␣ alone varied with each gene, all four genes were strongly expressed in cells stimulated with both agents (Fig. 2A). IFN␥ and TNF␣ cooperativity was markedly reduced (Ͼ90%) in STAT1-deficient fibroblasts without affecting the sensitivity to TNF␣ alone (Fig. 2B). These four responses were not mechanistically identical; the synergistic enhancement of RANTES mRNA expression was blocked in cells co-treated with cycloheximide (CHX), while expression of IRF-1, ICAM-1, and Mig was unaltered (Fig. 2C). Thus, the synergy between IFN␥ and TNF␣ may involve protein synthesis- Potential cis-regulatory elements and critical defined sequences are schematically shown based on gene sequences obtained from the GenBank TM data base and the following references; human IRF-1 (14,15,49); human ICAM-1 (16,17,44,45,68); mouse Mig (41,42); mouse RANTES (30,46). The numbers above the promotor regions refer to the nucleotide position relative to the transcription start site.
FIG. 2. STAT1 is essential for optimal synergy between IFN␥ and TNF␣. Serum-starved confluent monolayers of fibroblasts from wild type mice (A) or STAT1-deficient mice (B) were either untreated (UT) or stimulated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) for 3 h in the presence or absence of cycloheximide (C, CHX; 5 g/ml) prior to preparation of total RNA and analysis of specific mRNA levels by Northern hybridization as described under "Experimental Procedures." Five g of total RNA were analyzed in each lane. Blots were hybridized with the indicated radiolabeled cDNA probes. Similar results were obtained in three independent experiments. dependent and -independent mechanisms, both of which require IFN␥-induced STAT1.

IFN␥/TNF␣-induced Cooperative Transcription of the IRF-1 Gene Depends upon both the SBE and B Sites-IFN␥-induced
transcription of the IRF-1 gene has been shown to depend upon the SBE located in positions Ϫ123 to Ϫ113 (14,15). A highly conserved B motif has also been identified, although its functional significance remains equivocal ( Fig. 1) (14,49,50). To determine if cooperation between IFN␥ and TNF␣ for induction of the IRF-1 gene depends upon one or both of these sites, luciferase reporter gene constructs of the IRF-1 promoter (1.3 kb) were prepared in which the SBE and/or the proximal B site were individually mutated. These constructs were transiently transfected into wild type or STAT1-deficient fibroblasts, and their activity was assessed following treatment with IFN␥ and/or TNF␣ (Fig. 3). As reported previously, IFN␥ markedly induced luciferase activity in normal fibroblasts transfected with the wild type 1.3-kb IRF-1 promoter construct (14,15). TNF␣ also modestly induced luciferase activity (5-7fold induction). When cells were simultaneously stimulated with IFN␥ and TNF␣, the promoter activity was synergistically enhanced. Mutation of the SBE site nearly abolished the IFN␥ sensitivity and markedly reduced the cooperative response to the combination of IFN␥ and TNF␣ (85% reduction in magnitude) without affecting the TNF␣-induced luciferase activity. Similarly, the mutation of the B site abolished the response to TNF␣ alone and reduced the cooperativity for IFN␥ and TNF␣ (77% reduction in magnitude). Mutation of both the SBE and the B site in the same construct nearly eliminated sensitivity to either stimulus alone, and the cooperative response, although evident, was only 3% of that seen in cells transfected with the intact promoter. Furthermore, in STAT1-deficient fibroblasts, response to IFN␥ either alone or in combination with TNF␣ was less than 5% of the response seen in wild type cells. Taken together, these results indicate that both the SBE and the B sequence motifs are required for optimal cooperativity between IFN␥ and TNF␣ and that activation of STAT1, at least, is also necessary.
While the magnitude of cooperative response is markedly reduced in wild type cells transfected with mutations in individual motifs (either SBE or B) and in STAT1-deficient cells, there is residual cooperativity evident in both circumstances. This apparent leakiness may derive from multiple sources. For example, the mutated motifs may retain some low affinity interaction with individual factors. Alternatively, there may be other sites that are able to participate, providing the lower magnitude cooperativity. Indeed, low but detectable cooperativity is evident using promoter constructs in which both the SBE and the proximal B sites are mutated. The very low but reproducible response to IFN␥ seen in STAT1-deficient cells (both in Fig. 1 and Fig. 3) may reflect minor compensatory action of IFN␥ functioning through STAT1-independent systems. Indeed, in EMSA experiments using nuclear extracts from IFN␥-treated STAT1-deficient fibroblasts, we detected low but significant levels of STAT3 that were not seen in wild type cells (data not shown). Since STAT3 can act to modulate transcription through IFN␥ activation sequence motifs, this could account for the leaky response to IFN␥. It should be emphasized that these low level responses may detectable in our experimental system due to the very high sensitivity of the luciferase reporter gene.
Binding of IFN␥-induced STAT1 and TNF␣-induced NF-B Is Not Cooperative-IFN␥ is well documented to stimulate the phosphorylation and nuclear localization of STAT1␣ homodimers (8 -11). Similarly, TNF␣ is a potent stimulus of the nuclear translocation of various members of the Rel homology family (18 -20). The functional cooperativity between IFN␥ and TNF␣ might result from cooperative effects on DNA binding activities of the respective transcription factors. Thus, we next compared the binding activities of STAT1 and NF-B to their respective sequence motifs using nuclear extracts prepared from cells stimulated with IFN␥ and/or TNF␣ for 30 min by EMSA (Fig. 4). Although nuclear extracts from fibroblasts stimulated with IFN␥ showed little or no inducible B binding activity, cells stimulated with TNF␣ exhibited two inducible complexes designated as C1 and C2 in Fig. 4A. When cultures were co-stimulated with IFN␥ and TNF␣, the magnitude of the binding activity and pattern of complex formation were essentially the same as seen in cells stimulated with TNF␣ alone. As shown in Fig. 4B, the C1 complex was fully reactive with antiserum specific for NF-B1, while the C2 complex showed partial reactivity with anti-NF-B1 and full reactivity with antiserum specific for RelA. Antibodies specific for c-Rel and STAT1 did not recognize any of the IFN␥and/or TNF␣-induced B binding activities. These results suggest that the binding activity induced by TNF␣ that recognizes the IRF-1 B site is composed of NF-B1/RelA heterodimers and RelA homodimers. Specificity for the Rel protein binding was further assessed by oligonucleotide competition assays (Fig. 4C). Oligonucleotides containing the wild type IRF-1 B motif competed effectively for the binding of these Rel proteins, while a mutant oligonucleotide was inactive (lanes 2 and 3). Interestingly, oligonucleotides containing a wild type IRF-1 SBE or a mutant SBE in which two adenine residues in the 3Ј half of the inverted repeat were changed (m1 SBE) also partially competed for the binding of NF-B1 and RelA (lanes 4 and 5). Another mutant SBE (m2 SBE), in which the intervening sequence between the inverted repeats was also altered, could not compete for binding to the B motif. Since the adenine residues in the inverted repeat have been previously shown to be critical for recognition by STAT1 (14), sequence preferences for STAT1 and NF-B appear to be distinct. This ability of SBE to compete for NF-B recognition may result from the B-like site in the 5Ј portion of the IRF-1 SBE motif. In addition it may reflect the low affinity recognition of SBEs by NF-B as previously reported (51).
Consistent with previous reports, nuclear extracts from IFN␥-treated fibroblasts contained a prominent stimulus-dependent DNA binding activity specific for the IRF-1 SBE (Fig.  5A), and this complex is fully reactive with antibody to STAT1 (data not shown). Interestingly, TNF␣ also induced a DNA binding activity that recognized the IRF-1 SBE forming a complex that migrated at a slightly different mobility. This complex was reactive with antibodies specific for NF-B1 and RelA (Fig.  5B, lanes 3 and 4). These findings are also consistent with the results in Fig. 4C showing competition between the SBE and NF-B on the IRF-1 B site. When nuclear extracts from cells stimulated with both IFN␥ and TNF␣ were analyzed, a single broad band was observed, consistent with the presence of both the STAT1 and NF-B complexes seen with IFN␥ or TNF␣ stimulation alone. The most prominent component in this complex was STAT1, as indicated by immunoreactivity with anti-STAT1 (Fig. 5B, lane 6). The more slowly migrating complex, which was not reactive with anti-STAT1, was reactive with anti-NF-B1 and anti-RelA (lanes 8 and 9). Competition assays showed that an oligonucleotide containing a wild type SBE effectively competed for the binding of all complexes (Fig. 5C,  lane 4), while the wild type B motif either did not compete or did so poorly (lane 2). The m1 SBE did not compete, indicating that most of the binding activity present was STAT1, since this mutation appears to affect primarily the formation of STAT1 complexes and not NF-B. A large DNA fragment containing both the SBE and the B sites was also able to compete complex formation in response to treatment with IFN␥ and TNF␣. When this larger fragment (spanning positions Ϫ129 and Ϫ37 of the IRF-1 promoter) was used as a probe in EMSA, each complex was formed independently, and no evidence was obtained for cooperativity in binding between factors activated

FIG. 4. NF-B1 (p50) and RelA (p65) bind to the IRF-1 proximal B site in TNF␣-induced fibroblasts.
A, fibroblasts from wild type mice were either untreated (UT) or treated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) as indicated for 30 min prior to the preparation of nuclear extracts. Five g of each nuclear extract were analyzed for B binding activity by EMSA using a radiolabeled oligonucleotide containing the IRF-1 B sequence motif. Two major complexes are indicated as C1 and C2. B, nuclear extracts from IFN␥ and TNF␣-stimulated cells were incubated with the indicated antibodies (1 g) before analysis of the B binding activity as described above. C, competition analysis of IRF-1 B binding activity. Specificity of binding was assessed by competition with a 50-fold molar excess of unlabeled wild type or mutant oligonucleotide corresponding to the B or SBE motifs as shown at the bottom. Mutated nucleotides are indicated in italic type. Underlined sequences represent the B and SBE motifs, respectively. The overlined sequence indicates a potential B motif contained within the SBE site. An oligonucleotide fragment corresponding to the region between Ϫ129 and Ϫ37 (SBE-B) or an oligonucleotide fragment in which both the SBE and the B sites have been mutated (mSBE-mB) was also used as a competitor. Nuclear extracts (5 g) from wild type fibroblasts treated with IFN␥ (100 units/ml) and TNF␣ (10 ng/ml) for 30 min were analyzed. Similar results were obtained from three independent experiments. independently by IFN␥ or TNF␣ (data not shown).
SBE and B Sequences Cooperate in an Artificial Promoter-To explore the generality of the B motif and SBE functional cooperativity, we asked whether transcriptional synergy could be reconstituted using isolated sequence elements placed in a heterologous promoter. Initially, one copy of the IRF-1 SBE and/or the B2 motif from the mouse IP-10 gene (33,37) were placed in front of the TK promoter (TK-105) linked to the CAT reporter gene and tested for sensitivity to IFN␥ and/or TNF␣ following transient transfection in NIH3T3 cells (Fig. 6). Although one copy of the B2 motif exhibited little sensitivity to IFN␥ or TNF␣ either alone or in combination with the SBE, one copy of the IRF-1 SBE motif was sensitive to IFN␥ or IFN␥ and TNF␣. When a construct containing one copy each of the B2 and the SBE was analyzed, a strong synergistic response was seen in cells stimulated with the combination of agents. Mutation of the SBE site abolished all stimulus sensitivity of the combination construct and was essentially identical to that of a construct containing only a single B site. Thus cooperativity was not due to creation of fortuitous binding sites in the region where inserted sequences are coupled. Interestingly, cooperativity between IFN␥ and TNF␣ was also seen using the construct containing only the IRF-1 SBE and using the construct containing wild type SBE and mutant B. The synergistic response of such constructs to IFN␥ and TNF␣ appears to depend upon the distal portion of the TK promoter, which contains a GC box and a CCAAT box; no cooperativity was seen in a truncated form of the TK promoter in which the distal GC box and CCAAT box have been deleted (pTK-81, see Fig. 7).
The spacial relationship between the two cooperating sites may be an important determinant of their synergistic interaction. To examine this possibility, reporter constructs in which the sequence motif orientation and the nucleotide spacing be-tween motifs were varied were prepared and examined in transient assays (Fig. 7). For these experiments, a truncated form of the TK-luciferase vector (pTK-81) was utilized in which both a GC box (Sp1 binding site) and a CCAAT box have been deleted. When a single copy of either a B site or the IRF-1 SBE were linked to this reporter plasmid, no cooperative response was obtained. As mentioned above, this result suggests that the cooperative response seen with constructs containing a single SBE site (see Fig. 6) requires one or both of the sites deleted from the TK promoter. When a construct containing a single copy each of the SBE and the B motif was examined, a strong synergistic response was obtained. The synergy was not dependent upon the relative order of sites. Although constructs containing the SBE in either a distal or proximal relationship to the TK promoter exhibited variable response to IFN␥ alone, cooperative responses were comparable (6 -7-fold). The variability in sensitivity to IFN␥ is also observed in Fig. 7 when comparing the response of pTK SBE and pTK mB2 ϩ SBE, where the spacing of the SBE relative to the TK promoter is comparably altered. When the spacing between the two sites was incrementally increased, sensitivity to individual and combination stimulation was reduced. An increase of 5 nucleotides only modestly reduced the cooperativity, indicating that the orientation of bound factors relative to each other and the turn of the helix was not a limiting feature of the response. As the spacing interval was increased, the response was much more dramatically reduced. Interestingly, when the sites were separated by 64 nucleotides, a distance equivalent to that separating the SBE and B sites in the endogenous IRF-1 promoter, sensitivity to stimulation was lost entirely. These results indicate that while spacing may influence the magnitude of cooperativity, other features of the sequence between sites are probably of more critical importance.

FIG. 5. SBE binding activity in fibroblasts treated with IFN␥ and TNF␣.
A, fibroblasts from a wild type mouse were either untreated (UT) or treated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) as indicated for 30 min prior to the preparation of nuclear extracts. Five g of each nuclear extract were analyzed for DNA binding activity by EMSA using radiolabeled oligonucleotides containing the IRF-1 SBE sequence motif. B, nuclear extracts (5 g) from wild type fibroblasts treated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) for 30 min were incubated with the indicated antibodies before analysis of the SBE binding activity. C, competition analysis of IRF-1 SBE binding activity. Specificity of binding was assessed by competition with a 50-fold molar excess of unlabeled wild type or mutant oligonucleotide corresponding to the B or SBE motifs as described under "Experimental Procedures" and as shown in Fig. 4. A double-stranded oligonucleotide corresponding to the region between Ϫ129 and Ϫ37 (SBE-B) or an oligonucleotide in which the SBE and the B sites have been mutated (mSBE-mB) was also used as a competitor. Nuclear extracts (5 g) from wild-type fibroblasts treated with IFN␥ (100 units/ml) and TNF␣ (10 ng/ml) for 30 min were analyzed. Similar results were obtained from three independent experiments.

DISCUSSION
IFN␥ and TNF␣ utilize distinct signaling pathways leading to altered gene transcription (8 -11, 52). When these cytokines have been used in combination, both cooperative and antagonistic effects on gene transcription have been observed (3)(4)(5)(6)(7)40). The present study was undertaken to define the mechanisms involved in such a synergistic response. The results demonstrate that STAT1 activation by IFN␥ and NF-B activation by TNF␣ are the principle events necessary for cooper-ative induction of genes containing appropriate SBE and B sequence motifs. Independent interaction of STAT1 and NF-B with their cognate binding sites is sufficient for mediating the cooperativity. These conclusions are based on the following observations. 1) IFN␥ and TNF␣ synergized strongly to promote expression of multiple genes that contain at least one copy of an SBE and a B site, including the IRF-1, ICAM-1, and Mig genes. 2) This activity was abolished in fibroblasts prepared from mice in which the STAT1 gene has been deleted by ho-FIG. 6. B and SBE motifs confer functional cooperativity on a heterologous promoter in response to IFN␥ and TNF␣. One copy of a oligonucleotide corresponding to the wild type or mutant form of the IRF-1 SBE or the IP-10 B2 site (see "Experimental Procedures") was linked upstream of the TK-105 promoter (pTK-105) containing the CAT gene. The nucleotide length between the SBE and the B sites was 25 bases. The combination of the SBE and the B sequences are schematically indicated. These CAT constructs were transiently transfected into NIH3T3 cells as described under "Experimental Procedures." Twenty-four hours after transfection, the cells were either untreated or stimulated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) for 16 h and assayed for CAT activity. The relative CAT activity is shown as -fold induction of stimulated versus unstimulated samples. Mean percentage of acetylation for different constructs in unstimulated cultures ranged from 0.2 to 0.4%. The maximum CAT activity was 27% acetylation, which was obtained in cultures transfected with pTK B2 ϩ SBE and stimulated with IFN␥ and TNF␣. Each column and bar represents the mean Ϯ S.E. from three independent experiments.

FIG. 7. Effect of motif order and spacing on IFN␥ and TNF␣-induced cooperativity.
One copy of an oligonucleotide corresponding to the IRF-1 SBE or the B2 motif from the IP-10 gene (see "Experimental Procedures") was placed in front of the TK-81 promoter linked to the luciferase gene (pTK-81 Luc) as indicated schematically. Constructs with increasing nucleotide spacing between the SBE and the B site were prepared as described under "Experimental Procedures." These constructs were transiently transfected into NIH3T3 cells, and following a 24-h rest, the cells were either untreated or stimulated with IFN␥ (100 units/ml) and/or TNF␣ (10 ng/ml) for 8 h prior to harvest and determination of luciferase activity. The relative luciferase activity is presented as a percentage of activity obtained in cells transfected with the pTK-81 SP25 B ϩ SBE plasmid stimulated with IFN␥ and TNF␣. The -fold induction of stimulated versus unstimulated samples is also indicated. Each column and bar represents the mean Ϯ S.E. from three independent experiments. mologous recombination. 3) Synergistic transcription induced by IFN␥ and TNF␣ was observed in normal fibroblasts transfected with a reporter gene under control of a 1.3-kb fragment of the IRF-1 gene promoter. 4) The synergistic induction of the IRF-1 promoter activity was nearly abolished in a STAT1deficient cell line. 5) Site-directed mutagenesis of the SBE and the proximal B site in the IRF-1 gene promoter significantly reduced the magnitude of the synergistic response. 6) IFN␥ and TNF␣ independently activated STAT1 and NF-B (NF-B1/ RelA), respectively, as measured by binding to their cognate sequence motifs. 7) No cooperative effects on DNA binding activities were observed. 8) The SBE and B motifs could confer transcriptional synergy in response to IFN␥ and TNF␣ when examined in a heterologous promoter.
IFN␥-induced transcriptional synergy appears to be mediated by multiple pathways involving both protein synthesis-dependent and -independent mechanisms (5,53). The results presented in this study indicate that cooperative effects involving IFN␥ and TNF␣ exhibit similar behavior (Fig. 2). An important observation is that both protein synthesis-dependent and -independent cooperativity still depends largely on STAT1, consistent with the recent reports showing STAT1 to be obligatory for IFN-mediated biological activities (24,54). The requirement for protein synthesis during IFN␥/TNF␣-mediated RANTES gene expression suggests that some IFN␥-induced protein(s) (e.g. IRF-1) might be necessary for cooperativity in this circumstance, consistent with such roles for other genes (24,53,55). Inspection of the RANTES promoter sequence suggests the presence of IRF-binding motifs (56). Furthermore, functional B motifs have been identified in the promoter ( Fig.  1) (46) and cooperative regulation of transcription by IRF-1 and NF-B has been previously reported (57,58). In contrast, direct activation of STAT1, which may include the formation of STAT1␣ homodimers, heterodimers, or other oligomeric interactions, appears to be involved in the cooperative induction of IRF-1, ICAM-1, and Mig gene expression. IFN␥-dependent transcription of the IRF-1, ICAM-1, and Mig genes has been shown to depend upon SBE motifs that bind STAT1 in homo-or heterodimeric forms (14 -17, 41-43). Furthermore, synergistic induction of IP-10 gene transcription by IFN␥ and TNF␣ also depends on an IFN␥-inducible factor that contains STAT1 and binds the IFN-stimulated response element found in the IP-10 promoter (5).
TNF␣ is well documented as a potent inducer of NF-B and has been reported elsewhere to cooperate functionally with other transcription factors (5,21,23,57,58). The results in the present study indicate that NF-B (NF-B1/RelA) can cooperate with STAT1 to promote synergistic transcriptional activity. The proximal B site in the IRF-1 promoter is a functional B motif, which is recognized by a combination of NF-B1 and RelA in fibroblasts. TNF␣-mediated ICAM-1 gene transcription has been shown to depend upon a B motif recognized by Rel family members (44,45). Interestingly, despite the fact that the Mig gene is not independently induced by stimuli that activate NF-B (e.g. TNF␣ and lipopolysaccharide) (28), the cooperative induction of this gene by IFN␥ and TNF␣ suggests that the B motifs found in the Mig promoter are functional when STAT1 is also available.
Interestingly, we noted that the IRF-1 SBE appeared able to mediate a synergistic response to stimulation with IFN␥ and TNF␣ independently of the proximal B site (see Figs. 3 and 6). Since the SBE site was also recognized by STAT1 and NF-B, this dual recognition might contribute to the functional synergy. While this possibility cannot be ruled out, several considerations suggest that the cooperativity observed derives from other sources. For example, in Fig. 3, the constructs containing mutations in the SBE, in the B site, and in the double mutant all showed some synergistic response to the stimulus combination. Because this fragment is large (1.3 kb), there are apparently other independent sites that can cooperate with the SBE, the B site, or each other. Second, although the cooperative behavior of the artificial construct utilized in Fig. 6 (pTK SBE) appears to depend solely upon the SBE, data shown in Fig. 7 illustrate that such cooperativity is dependent upon a 25-base pair fragment of the TK promoter between positions Ϫ105 and Ϫ81. When the pTK-81 promoter was used with the isolated SBE, no cooperativity was evident. While we do not understand the mechanism(s) through which cooperativity occurs in this setting, the results suggest that the SBE is not independently capable of mediating cooperative response to IFN␥ and TNF␣.
The molecular mechanisms involved in functional synergy between distinct transcription factors appear to be multifactorial (23, 59 -63). In some cases, direct protein-protein interaction between activator proteins has been observed (23,59). The physical interaction may result in cooperative DNA binding, more stable protein-DNA interactions, and/or increased affinity of one or both activator proteins, ultimately creating a highly stable multiprotein complex that has markedly enhanced functional properties (23,63). In this regard, members of the NF-B and the STAT families have been observed to interact with members of other distinct factor families (21-23, 64, 65) although not with each other. Both NF-B and STAT1 formed complexes on the IRF-1 SBE, but these appeared to be independent interactions between individual factors and DNA, since each complex exhibited a distinct mobility in EMSA. Furthermore, the presence of one factor did not alter the interaction of the other with its cognate site, nor did the presence of both factors promote the formation of any unique complexes not detected in cells treated with either stimulus alone. Nevertheless, we cannot completely rule out the possibility that a weak interaction between STAT1 and NF-B in vivo might produce the observed functional cooperativity, since in vitro study of protein-protein interaction will only detect relatively high affinity interactions. Furthermore, analysis of nucleotide spacing between these sequence motifs indicated that, although spacial distances may quantitatively modify the response, the specific intervening nucleotide sequences were more important. This latter observation may suggest a role for other factors or an influence of flanking sequence on the functional behavior of transacting factors bound to DNA. This possibility is also supported by the finding that a single SBE motif could mediate moderate cooperative response to IFN␥ and TNF␣ when other stimulus-insensitive sites are present.
An alternative mechanism for transcriptional synergy might involve independent interaction of the activation domains of individual factors with components of the general transcription machinery such as the TATA-binding protein, TATA-binding protein-associated factors, TFIIA, and TFIIB (61,62). The same activator domain may interact with more than one component of the RNA polymerase complex. These multiprotein interactions could facilitate assembly of a preinitiation complex, stabilize the complex on promoter DNA, and thus promote the frequency of transcriptional initiation and elongation. Members of the Rel family have been reported to interact directly with TATA-binding protein and TFIIB (66,67), and thus it is conceivable that the activation domains of these factors and of STAT1 may differentially interact with basal transcription components.