Interferons Inhibit Tumor Necrosis Factor-α-mediated Matrix Metalloproteinase-9 Activation via Interferon Regulatory Factor-1 Binding Competition with NF-κB*

Enhanced expression of matrix metalloproteinase-9 (MMP-9) correlates with invasion during tumor progression. Interferons (IFNs) inhibit MMP-9 activation in response to tumor necrosis factor-α (TNF-α), and the latter activates theMMP-9 gene through NF-κB. Understanding the molecular basis for MMP-9 inhibition may provide tools to control cell invasion. The data reported here show the critical role of interferon regulatory factor-1 (IRF1) in the inhibition of MMP-9. (i) IFN treatment suppresses TNF-α-induced MMP-9 reporter activity in STAT1(+/+) cells but not in STAT1(−/−) cells. (ii) IRF1 transfection blocks TNF-α-mediated MMP-9 activation. (iii) IFNs phosphorylate STAT1 and induce IRF1 but do not affect Iκ-B degradation nor NF-κB nuclear translocation. (iv) Nuclear NF-κB (p50/p65) and IRF1, but not STAT1, bind to the MMP-9 promoter region containing an IFN-responsive-like element overlapping the NF-κB-binding site. (v) Recombinant IRF1, although unable to bind to an NF-κB consensus sequence, competes with NF-κB proteins for binding to theMMP-9 promoter. (vi) Conversely recombinant p50/p65 proteins reduce IRF1-DNA binding. (vii) In cells cotransfected with IRF1 and/or p65 expression vectors, an excess of IRF1 reduces MMP-9 reporter activity, whereas an excess of p65 blocks the inhibitory effect of IFN-γ. Thus, in contrast to the known synergism between IRF1 and NF-κB, our data identify a novel role for IRF1 as a competitive inhibitor of NF-κB binding to the particularMMP-9 promoter context.

In the current work, we chose to determine the transcriptional mechanism by which IFN represses MMP-9 expression. We utilized EW-7 human cells derived from Ewing's sarcoma previously shown by us to be sensitive to the antiproliferative effects of IFNs (35), IRF1-transfected EW-7 cells (35), as well as the deficient STAT1 (Ϫ/Ϫ) human fibrosarcoma U3A cells and their parental STAT1 (ϩ/ϩ) 2Ftgh cells (36). We report that IFN-induced IRF1 inhibits NF-B-dependent activation of MMP-9 by binding to the promoter region overlapping the NF-B-binding site.
Transient Transfections-Luciferase constructs containing elements of the MMP-9 promoter (Ϫ634 bp to ϩ30 bp, Ϫ603 to ϩ30, Ϫ531 to ϩ30, Ϫ144 to ϩ30, and Ϫ73 to ϩ30) have been described by Gum et al. (14). Plasmids DNA cloned in DH 5 ␣ competent cells (Invitrogen) were prepared with the Quantum Preparation Kit (Bio-Rad). 20 ϫ 10 6 cells were mixed gently with 10 g of dried supercoiled MMP-9 promoter plasmid and were electroporated as described previously (37). Cells were stimulated 16 h after electroporation with IFNs (1000 units/ml) or with TNF-␣ (10 units/ml), alone or in combination, in fresh medium containing 10% FCS. After 24 h of stimulation, cells were analyzed for luciferase activity according to the manufacturer's recommendations (Luciferase Assay System, Promega) or for chloramphenicol acetyltransferase activity as described previously (37).
Cotransfection Experiments-EW-7 cells (20 ϫ 10 6 in 200 l of Dulbecco's modified Eagle's medium) were mixed with 15 g of supercoiled plasmid DNA, p65/pc-DNA-3, or IRF1/pc-DNA-3 and were electroporated as described previously (37). Basic pc-DNA3 was used as control. After 6 h, the medium was changed, and cells were distributed in a 12-well tissue culture plate, and the culture was continued for 36 h. Then supercoiled MMP-9 promoter DNA (Ϫ634) was transfected using LipofectAMINE TM 2000 (Invitrogen) according to the manufacturer's recommendations (1 g of DNA/1 ϫ 10 5 cells/well). After 6 h, cells were stimulated in the same medium for 24 h with IFN-␥ (1000 units/ml) or with TNF-␣ (10 units/ml) or both stimuli. Cells were further analyzed for luciferase activity according to the manufacturer's recommendations (Luciferase Assay System, Promega).
Nuclear Extracts and Electrophoretic Mobility Shift Assay (EMSA)-Nuclear extracts (from 40 ϫ 10 6 cells) were prepared, and EMSA experiments were performed as described by Sancéau et al. (37). Nuclear proteins (10 g) were incubated for 30 min at 4°C with radiolabeled probe (30,000 cpm). After resolution of the nucleoprotein complexes by a non-denaturing electrophoresis, the gel was dried and exposed overnight to a PhosphorImager screen (Amersham Biosciences). For competition experiments, appropriate antibodies (1 g) were mixed directly with nuclear extracts and binding buffer for 30 min at 4°C before adding the radiolabeled probe. For direct DNA occupancy, recombinant proteins (IRF1, p50, and p65 proteins, or recombinant Sp-1 protein, increasing volumes 0.5-2 l) were mixed with the radiolabeled probe (30,000 cpm) in EMSA buffer conditions. After 30 min of incubation at 4°C, the reactions were resolved by non-denaturing electrophoresis, like the usual EMSA. A 44-bp DNA fragment that spans the Ϫ619 to Ϫ575 region of the human MMP-9 promoter and encompasses the NF-B-binding site was generated by annealing two complementary oligonucleotides (Genset, Paris, France), 5Ј-AGA CAG GGG TTG CCC CAG TGG AAT TCC CCA GCC TTG CCT-3Ј and 3Ј-GTC CCC AAC GGG GTC ACC TTA AGG GGT CGG AAC GGA TCG-5Ј (the NF-B-binding site of MMP-9 is underlined). The synthetic oligonucleotide 5Ј-CTA GAC AGA GGG GAT TTC CGA GAG GTT-3Ј covering the consensus NF-B-binding site from the human immunoglobulin light chain enhancer or the synthetic oligonucleotide 5Ј-CCT TGG GTT TTC CCA TGA GTT C-3Ј covering the NF-B-binding site of the IL-6 promoter as described previously (37) were used. The 32 P-labeling probes were generated by the Klenow fill-in reaction (Stratagene) of the annealing fragment.
Western Blot Analysis-Following cell (10 ϫ 10 6 cells/ml) stimulation with IFNs (1000 units/ml) or with TNF-␣ (10 units/ml), alone or in association, for the indicated times, cells were washed twice with phosphatebuffered saline, and nuclear or cytosolic extracts were prepared as described previously (37). 20 mM anti-phosphatase mixture (sodium orthovanadate, NaF, ␤-glycerophosphate), 0.2 mM phenylmethylsulfonyl fluoride, and 10 g/ml protease inhibitors (pepstatin, leupeptin, aprotinin) were added to each buffer used. Protein content in the supernatants was determined using Bradford's method. Equivalent amounts of protein (60 g) were separated on a 10% SDS-PAGE and blotted to nitrocellulose membranes (Schleicher & Schuell). Membranes were hybridized and processed as described previously (35) using enhanced chemiluminescence protein detection (Renaissance Enhanced Luminol Reagent, PerkinElmer Life Sciences). Immunoreactive protein bands were detected by autoradiography on Hyperfilms (Amersham Biosciences).
Reverse Transcriptase-PCR-RNA extraction (by SV Total RNA Isolation System, Promega) from Ewing cells and subsequent cDNA synthesis were conducted as described previously (35). MMP-9 cDNA (296 bp) was amplified using the sense primer 5Ј-GGA GAC CTG AGA ACC AAT CTC-3Ј and the antisense primer 5Ј-TCC AAT AGG TGA TGT TGT CGT-3Ј according to published sequences (38). To normalize the PCR products, the ␤ 2 -microglobulin gene was amplified as described by Bauvois et al. (6). The PCR products were visualized by electrophoresis in 1.2% agarose gel containing 0.2 g/ml ethidium bromide. The NIH Image 1.44 b11 software was used for the analysis.
Enzyme-linked Immunoabsorbent Assays-The conditioned media from Ewing cells were harvested under sterile conditions and frozen before MMP-9 or IL-6 contents were determined using a commercial ELISA kits from R & D Systems (Abingdon, UK). Controls included FCS-supplemented RPMI 1640 medium alone incubated under the same conditions.
Gelatinolytic Zymography-Analysis of MMP-9 activity was carried out in 7.5% (w/v) SDS-PAGE containing 0.1% gelatin (w/v) as described elsewhere (39). Equal amounts of culture media (10 -20 l) were applied to the gel in Laemmli sample buffer lacking ␤-mercaptoethanol. The recombinant pro-MMP-9 with a molecular mass of 92 kDa (R & D Systems) was used as positive control. Samples were preincubated for 60 min with 0.5 mM aminophenylmercuric acid (APMA, Sigma) which activates the pro-form to the activated form. Gelatinolytic activities of both pro-MMP-9 and active MMP-9 were detected as transparent bands on the background of Coomassie Blue-stained gelatin. The NIH Image 1.44 ␤11 software was used for the analysis of the bands, after acquisition in an Appligene densitometer (Oncor).

IFNs Inhibit TNF-␣-dependent Activation of MMP-9 in EW-7
Cells-As assessed by enzyme-linked immunoabsorbent assay (ELISA), day 1-cultured EW-7 cells in the absence or presence of IFNs spontaneously released no detectable amounts of MMP-9 into the culture conditioned medium (Fig. 1A). TNF-␣ up-regulates the expression of MMP-9 in various cell types (4,5,13). Similarly, treatment of EW-7 cells with TNF-␣ (10 units/ml) resulted in marked stimulation of MMP-9 protein levels ( Fig. 1A). MMP-9 stimulation in response to TNF-␣ was dose-dependent, with maximal effects being obtained for doses of 1000 units/ml. Although without any effect on unstimulated EW-7 cells, IFN-␤ and IFN-␥ reduced production of MMP-9 of TNF-␣-treated cells (Fig. 1A). Under our conditions, TNF-␣/ IFN combination (TNF 10 units/ml, IFN 1000 units/ml) did not induce any death signaling in EW-7 cells. Zymography analysis showed the presence of a gelatinase activity at 92 kDa in the conditioned media of TNF-␣-stimulated cells, consistent with the pattern of recombinant pro-MMP-9 ( Fig. 1B, top panel). Preincubation of TNF-␣-stimulated cell supernatant with APMA resulted in conversion of pro-MMP-9 to an active form of 82 kDa (Fig. 1B, top panel). Gelatinolytic activity of MMP-9 was decreased in the conditioned media of TNF-␣-stimulated cells treated with IFN-␤ or IFN-␥ for 24 h (Fig. 1B, top panel), and such inhibition of MMP-9 production in response to IFN was dose-and time-dependent; maximal inhibitory effects were obtained for doses of 5000 units/ml IFN (Fig. 1B, middle panel), and TNF-␣-mediated accumulation of MMP-9 in the culture medium was blocked 12 h after addition of IFN (Fig. 1B, bottom  panel). Moreover, the protein levels were related to the levels of expression of MMP-9 transcript (296 bp) detected by reverse transcriptase-PCR (Fig. 1C) suggesting that inhibition of MMP-9 production by IFNs reflected decreased MMP-9 mRNA amounts in TNF-␣-activated cells.
TNF-␣-dependent Activation of the MMP-9 Promoter Requires a Region Containing One Potential IFN-responsive Element Overlapping the NF-B-binding Site-To define the promoter region of MMP-9 necessary for both transcriptional MMP-9 activation by TNF-␣ and subsequent inhibition by IFNs in EW-7 cells, functional analysis of the 5Ј-flanking region of the human MMP-9 gene was carried out using the Ϫ670 to ϩ30 fragment (700 bp) as well as a series of 5Ј-deletions of the MMP-9 promoter, linked to a luciferase reporter plasmid ( Fig. 2A). We transiently transfected EW-7 cells with these plasmids. As shown in Fig. 2B, the Ϫ670 construct of the MMP-9 5Ј-flanking region yielded a 9-fold increase in luciferase activity in response to TNF-␣ treatment. No significant response to IFNs alone was observed (Fig. 2B). Combined treatment with TNF-␣ and IFN resulted in significant reduction of luciferase activity (30% inhibition with IFN-␣, 45% inhibition with IFN-␤, and up to 70% inhibition with IFN-␥). Similar results were observed after transfection of the Ϫ634 and Ϫ603 fragments (Fig. 2B). However, cells transfected with shorter fragments expressed a low basal luciferase activity that was not enhanced by TNF-␣ (Fig. 2B). Interestingly, this promoter fragment between Ϫ603 and Ϫ531 regions contained one nonconsensus NF-B-binding site (between Ϫ600 and Ϫ590) and an Sp-1-binding site (between Ϫ562 and Ϫ555) (13,14). Equally important, this region also contains eight nucleotides matching an IFN enhancer core sequence (5Ј-GGAATTCC-3Ј) previously identified in IFN-response genes (40, 41) ( Fig. 2A). Thus, it is possible that this putative IFN-responsive element (IRE), which overlaps with the NF-B-binding site ( Fig. 2A), participates in the negative regulation of MMP-9 gene transcription.
The transcription factors STAT1 and IRF1 are intracellular signal-transducing molecules in IFN signaling (28,29,(31)(32)(33)(34). STAT1 activated by both classes of IFNs is able to bind to ISRE (28,31,32). IRF1 binds ISRE sequences that may overlap with IRE sequences (33,34,42). Moreover, human IL-6, whose mechanism of transcriptional activation involves the STAT3 signaling pathway (43), did not affect TNF-␣-activated MMP-9 gene transcription (Fig. 2B) nor the MMP-9 activity induced by TNF-␣ in EW-7 cells (data not shown). Based on these observations, we next investigated the involvement of STAT1/IRF1-  20). B, analysis of gelatinolytic activity in the culture media of EW-7 cells cultured for 24 h in the absence or presence of TNF-␣ and/or IFNs. The enzymatic activity of MMP-9 was analyzed using zymography performed with equal amounts of protein loaded. Gelatinolytic activities are detected as clear bands in the gel. Activation of pro-MMP-9 by APMA (0.5 mM) results in an 82-kDa active form (bottom panel). 10 units/ml TNF-␣-induced gelatinase activity was inhibited in a dose-dependent manner with increasing concentrations of IFN-␥ (10, 100, 1000, and 5000 units/ml, middle panel) and a time-dependent manner (10 units/ml TNF-␣ alone or in the presence of 10 3 units/ml IFN-␥ for 3, 6, 12, 18, 24, 36, or 50 h, two lower panels). C, the cDNAs from EW-7 cells cultured for 24 h in the absence or presence of TNF-␣ and/or IFNs were used as templates for PCRs using specific primers for MMP-9 or ␤ 2 -microglobulin as described under "Experimental Procedures." PCR products were run on 1.2% agarose gels.
Inhibition of TNF-␣-dependent Activation of MMP-9 by IFNs Is Accompanied by STAT1 Activation and IRF1 Induction-We next performed Western blot analyses to determine whether STAT1 was effectively activated in IFN-stimulated cells. Binding of IFNs to their receptors resulted in oligomerization of receptor subunits and subsequent activation of JAK1, TYK2 (for IFN-␣ and IFN-␤), JAK1, and JAK2 (for IFN-␥) which then activate STAT1 through tyrosine phosphorylation resulting in dimerization, nuclear translocation, and DNA binding of STAT1 (28,31,32). Serine (Ser-727) phosphorylation was also required for maximal transactivating capacity of STAT1 (29,31). Nuclear extracts were prepared from untreated, IFN-, and/or TNF-␣-treated EW-7 cells for 30 min or 18 h. As shown in Fig. 4A, induced levels of tyrosine-and serine-phosphorylated STAT1 were already observed at 30 min of stimulation with IFN-␤, and IFN-␥ STAT1 activation by IFNs was rather persistent because amounts of tyrosine-and serine-phosphorylated STAT1 were still augmented in the nuclei of EW-7 cells treated with IFNs for 18 h (Fig. 4B). In parallel, STAT1 protein levels were not significantly affected with time following IFN treatment (Fig. 4, A and B). TNF-␣ neither influenced STAT1 protein nor its phosphorylation (Fig. 4, A and B). 18 h of stimulation by IFNs also resulted in tyrosine and serine phosphorylation of STAT1 in 2Ftgh STAT-1 (ϩ/ϩ) cells (Fig. 4C). As expected, no STAT1 was detected in nuclei extracts obtained from U3A STAT-1 (Ϫ/Ϫ) cells (Fig. 4D).
Unlike STAT1, which is activated within minutes of ligand binding to the receptor, the IRF family members including IRF1 constitutes a secondary wave of response to the IFN signal (33,34). We therefore tested whether IFNs could induce IRF1 in EW-7, 2Ftgh, and U3A cells. As shown in Fig. 4, B and C, stimulation of EW-7 or 2Ftgh cells for 18 h with IFN-␤ or IFN-␥ induced IRF1 protein. In agreement with previous observations (45), IFN-␥ was the best inducer of IRF1. The TNF-␣/IFN combination resulted in a more pronounced induction of IRF1 protein in these two cell lines when compared with stimulation with cytokines alone (Fig. 4, B and C). As expected, IRF1 was not detected in IFN-unresponsive U3A cells untreated or treated with IFNs and/or TNF-␣ (Fig. 4D). These data indicate that IRF1, as a downstream mediator of STAT1, is efficiently induced in STAT1-positive cells. Finally, in contrast to TNF-␣, IFN-␥ neither blocked the I-B degradation nor the TNF-␣-induced translocation of p65 from cytosol to nucleus (Fig. 4E).

IRF1 Induced in IFN-␥-activated EW-7 Cells Binds to the MMP-9
Promoter-To determine the contribution of STAT1 and/or IRF1 in the MMP-9 promoter activity, we performed EMSA. Nuclear extracts from EW-7 cells stimulated for 3 or 16 h with TNF-␣ (10 units/ml) and/or IFN-␥ (1000 units/ml) were isolated, and equal amounts of proteins were incubated with the radiolabeled MMP-9 probe covering the region between Ϫ619 and Ϫ575 (see Fig. 2A). As shown in Fig. 5A, the patterns of the protein-DNA complexes were different for EW-7 cells stimulated for 3 or 16 h. At 3 h, two inducible complexes (C1 and C2) were detectable after either TNF-␣ or TNF-␣/ IFN-␥ stimulation (Fig. 5A). Addition of antibodies against NF-B p50 subunit supershifted the C2 complex, whereas antibodies against NF-B p65 subunit supershifted both C1 and C2 complexes (Fig. 5B). Addition of IRF1 or STAT1 antibodies did not affect the formation of C1 and C2 complexes bound to the MMP-9 probe, indicating that such complexes contained neither IRF1 nor STAT1 proteins (Fig. 5B). Moreover, no changes were detected with the addition of respective matched isotype antibodies (data not shown). This demonstrated that the C1 and C2 complexes inducible by TNF-␣ after 3 h were, respectively, the dimers of NF-B p65/p65 and p50/p65. After 16 h induction, TNF-␣ only induced the binding of the C2 complex (Fig. 5A), whereas IFN-␥ induced the binding of a third lower DNA-protein complex C3 (Fig. 5A). Both C2 and C3 complexes were detected in response to TNF-␣ and IFN-␥ (Fig. 5A). Again, the identity of C2 was confirmed by supershift with antibodies against p50 and p65, whereas anti-IRF1 and anti-STAT1 had no effect on the mobility of this complex bound to the MMP-9 probe (Fig. 5C, panels TNF-␣ and TNF-␣ ϩ IFN-␥). The mobility and abundance of the C3 complex were not affected by antibodies against p50, p65, and STAT1 (Fig. 5C, panels IFN-␥ and TNF-␣  ϩ IFN-␥), whereas a decrease in C3 formation associated with a complex supershift was seen with IRF1 antibodies (Fig. 5C, panels IFN-␥ and TNF-␣ ϩ IFN-␥). In STAT1 (Ϫ/Ϫ) U3A cells, 16 h of TNF-␣ treatment induced the formation of the C2 complex (p50/p65), whereas IFN-␥ treatment did not lead to C3 complex formation (data not shown). Moreover, the comparison of levels of the IRF1-DNA complex (C3) for each IFN (compare Fig. 5, A and D) correlated with the difference in magnitude of repression of each IFN toward MMP-9 as shown above by zymography (Fig.  1B, zymography (Fig. 1B, IFN-␥ Ͼ IFN-␤ Ͼ IFN-␣). Again, both C2 and C3 complexes were formed in the presence of TNF-␣ ϩ IFN-␣/-␤ (Fig. 5D).
Moreover, when EW-7 nuclear extracts were analyzed for binding to the immunoglobulin NF-B consensus motif (37), we observed only the appearance of C1 and C2 complexes with TNF-␣, composed of p50 and p65 NF-B subunits as assessed by the addition of respective antibodies (Fig. 5E). Control isotype antibodies were ineffective (data not shown). No C3 complex was detectable in cells stimulated with IFN-␥ (Fig. 5E). 5. Binding of NF-B (p50 and p65) and IRF1 to the MMP-9 promoter. A, nuclear extracts were prepared from EW-7 cells after a 3or 16-h treatment with TNF-␣ (10 units/ml) or IFN-␥ (1000 units/ml) or both, and then were incubated with the 32 P-labeled MMP-9 probe (Ϫ619 to Ϫ575), and the protein DNA binding was analyzed by EMSA. B, EW-7 nuclear extracts after a 3-h IFN-␥/TNF-␣ treatment were assayed for protein DNA binding activity in the presence of p50 or p65 or IRF1 or STAT1 antibodies. C, EW-7 nuclear extracts after a 16-h treatment with TNF-␣, or IFN-␥, or both were assayed for protein DNA binding activity as in B. D, EW-7 nuclear extracts after a 16-h IFN-␣ -␤/TNF-␣ treatment were assayed for protein DNA binding activity. E, EW-7 nuclear extracts after a 3-or 16-h treatment with TNF-␣, or IFN-␥, or both were incubated with the 32 P-labeled immunoglobulin NF-B consensus probe and assayed for protein DNA binding activity as in B. NS, nonspecific band.
To ensure further that IRF1 binds to the MMP-9 promoter, we performed direct DNA/protein binding experiments using human recombinant IRF1, p50, and p65 proteins (0.5, 1, and 2 l), and the radiolabeled MMP-9 promoter fragment covering nucleotides between Ϫ619 and Ϫ575, in the absence of endogenous nuclear proteins (Fig. 6A). As controls, in vitro translated proteins derived from the empty vector or the Sp-1 coding sequence were used. No DNA-protein complexes were evident with in vitro translated proteins derived from the empty vector or the Sp-1 (Fig. 6A). As expected, recombinant NF-B p50 and p65 proteins bind to their respective sites of the MMP-9 promoter (Fig. 6A). The increased formation of DNA-protein complexes correlated with the increasing quantities of recombinant p50 and p65 proteins added to each reaction (Fig. 6A). When   FIG. 6. Competitive binding of IRF1 to the NF-B-binding site of the MMP-9 promoter. A, for direct DNA occupancy, recombinant proteins IRF1, p50, and p65 NF-B proteins (increasing volumes 0.5-2 l) were mixed with the radiolabeled probe in EMSA buffer conditions. The reaction products were then resolved by non-denaturing electrophoresis like the usual EMSA as detailed under "Experimental Procedures." The empty vector and the purified Sp1 protein were used as controls. B, nuclear extracts of EW-7 cells after a 16-h IFN-␥ or IFN-␥/TNF-␣ treatment were assayed for protein DNA binding activity in the presence of increasing quantities of recombinant proteins (p50, p65, or IRF1) as described under "Experimental Procedures." C, EW-7 cells were transiently cotransfected with the expression vectors p65 or IRF1 or the empty vector as control and the MMP-9 promoter region Ϫ634 construct as described under "Experimental Procedures." Luciferase activity was determined from cell-stimulated lysates. Values represent the mean Ϯ S.D. (n ϭ 3). D, cotransfections of the expression vectors p65 and IRF1 using the indicated amount (1 or 5 g) in EW-7 cells. Luciferase activity was determined 24 h after the MMP-9 promoter region Ϫ634 construct transfection, without cells stimulation. Values represent the mean Ϯ S.D. (n ϭ 3). the radiolabeled MMP-9 probe was incubated with recombinant IRF1, a major DNA-IRF1 complex formed (Fig. 6A). These results indicate that recombinant p50 and p65 proteins (activated by TNF-␣) as well as recombinant IRF1 (induced by IFN-␥) specifically bind to the MMP-9 promoter.
IRF1 Binds to a DNA Sequence of the MMP-9 Promoter Fragment Overlapping the NF-B-binding Site-To establish which DNA elements in the MMP-9 promoter are recognized by IRF1, we tested the effects of recombinant proteins (p65, p50, and IRF1, from 0.5 to 2 l) on the formation of C1, C2, and C3 complexes. As shown in Fig. 6B, recombinant p65, p50, and IRF1 proteins, added to nuclear extracts from untreated EW-7 cells, led to the formation of three protein-DNA complexes with different mobilities (Fig. 6B, cont). Two of the complexes corresponded to the C2 and C3 positions, and the third complex migrated slower than the expected C1 complex, suggesting a p65 multimer. Whether these recombinant proteins interfere with the formation of C1, C2, or C3 complexes was investigated next. Increasing amounts of p50, p65, or IRF1 proteins were added to the binding reaction mixtures containing the nuclear extracts and the MMP-9 promoter probe (Ϫ619 to Ϫ575). As increasing amounts of p65 or p50 proteins were added to IFN-␥-and/or TNF-␣-treated nuclear extracts, the abundance of C1 and C2 complexes increased (Fig. 6B), and this was associated with a concomitant decrease in the amounts of the C3 complex (Fig. 6B, IFN-␥ and TNF-␣ ϩ IFN-␥). Increasing amounts of IRF1 led to the appearance of the C3 complex in TNF-␣-treated nuclear extracts (Fig. 6B, TNF-␣), whereas IRF1 diminished the abundance of the C2 complex in TNF-␣and/or IFN-␥-treated nuclear extracts (Fig. 6B, TNF-␣, TNF-␣ ϩ IFN-␥). This result demonstrates that IRF1 is capable of competing with p50 and p65 NF-B subunits at their respective binding sites. Moreover, our competition assays suggest that the preferential binding of p50, p65, or IRF1 to the MMP-9 promoter is related to their relative nuclear abundance in EW-7 cells.
To provide further evidence that competition between IRF1 and NF-B (p65) dictates MMP-9 promoter activity, EW-7 cells were cotransfected with the IRF1 and/or p65 expression vectors or the empty vector and the luciferase reporter driven by the MMP-9 promoter region Ϫ634 construct (Fig. 6, C and D). Following cell stimulation for 18 h with IFN-␥ and/or TNF-␣, we observed that production of p65 blocked the inhibitory effects of IFN-␥ in EW-7 cells treated by TNF-␣ ϩ IFN-␥ (Fig.  6C). Conversely, production of IRF1 blocked MMP-9 promoter activity in TNF-␣-treated cells (Fig. 6C). When EW-7 cells were transiently cotransfected with IRF1 and p65 expression vectors alone or together, and the luciferase reporter driven by the MMP-9 promoter region Ϫ634 construct (without exogenous stimulation), p65 alone increased luciferase activity, whereas IRF1 alone did not markedly modify the basal luciferase activity (Fig. 6D). Moreover, an excess of IRF1 (5 g) relative to p65 (1 g) reduced MMP-9 reporter activity by ϳ30%, whereas an excess of p65 (5 g) relative to IRF1 (1 g) restored the level of luciferase activity observed after transfection of p65 expression vector alone (Fig. 6D).
IFNs and TNF-␣ Synergistically Induce IL-6 Expression in EW-7 Cells-We showed previously (37) that synergistic induction of the IL-6 gene by TNF-␣ and IFN-␥ in human monocytic cells involved cooperation between the IRF1 and NF-B p65 homodimers. We thus investigated whether TNF-␣ plus IFN treatment of EW-7 cells could regulate the expression of IL-6. As shown in Fig. 7A, TNF-␣ alone was able to induce IL-6 production, whereas IFNs were without effect. Addition of IFN-␣, IFN-␤, or IFN-␥ to TNF-␣ enhanced such production (Fig. 7A). EW-7 cells, transiently transfected with the plasmid construct linking IL-6 promoter containing the NF-B site to the chloramphenicol acetyltransferase reporter gene, expressed increased activity in the presence of TNF-␣ (Fig. 7B), and the addition of the three IFNs significantly enhanced this activity (Fig. 7B). IFN treatment alone did not induce the IL-6 promoter activity (Fig. 7B). In parallel, EMSA indicated that stimulation with TNF-␣, but not by IFN-␥, resulted in the formation of two DNA-protein complexes to the IL-6-NF-B probe, respectively identified as C1 and C2 complexes of NF-B (Fig. 7C). IFN-␥ did not induce the formation of the IRF1-DNA complex (Fig. 5C). This was confirmed by assessing the abilities of recombinant proteins p50 and p65, but not IRF1, to bind to the IL-6 promoter (Fig. 7D). These data are in agreement with our previous study showing that the synergistic induction of the IL-6 gene by IFN-␥ and TNF-␣ involved cooperation between NF-B and IRF1 which binds to an IFN enhancer core sequence close but distinct to the NF-B-binding site (37). These results show that the combination of TNF-␣ and IFNs in EW-7 cells divergently acts on IL-6 and MMP-9 gene expression.

DISCUSSION
The efficacy of IFNs has been exploited in maintenance therapy for patients with malignancies, including leukemias and various types of carcinoma (46 -49). The antitumor effects of IFNs have been attributed to the inhibition of cell proliferation, induction of cell differentiation, immunomodulation, and alteration of the level of gene expression in target cells (46 -49). Previous in vivo and in vitro studies indicated that MMP-9 production and activity in various normal and tumoral cell types could be inhibited by IFNs (6, 19 -27). In the present study, we show that IFN-mediated inhibition of MMP-9 expression occurs through IRF1 which competes for NF-B binding to the MMP-9 promoter.
Results from luciferase reporter assays in the EW-7 cell line indicated that the MMP-9 promoter region spanning from Ϫ603 bp to Ϫ531 bp, which contains the NF-B and Sp-1binding sites, is required for stimulation by TNF-␣ as well as for subsequent inhibition by IFNs. Studies on MMP-9 activation have identified NF-B as a main effector of MMP-9 gene expression activated by the TNF-␣ signaling pathway (13,14). The EMSAs performed here demonstrating activation of NF-B (p50/p65) DNA binding activity by TNF-␣ in EW-7 cells confirmed this. A recent study (26) reported the existence of two potential ISRE-binding sites within the Ϫ700-bp 5Ј-flanking sequence of the human MMP-9 gene. Moreover, we found an IRE-like sequence (sharing homologies with the motif designated pIRE, for palindromic IFN-responsive element, according to Miyamoto et al. (40)), located in the promoter of MMP-9 within the Ϫ601/Ϫ584 bp overlapping the non-consensus NF-B-binding element of the MMP-9 gene (Fig. 1D). The transcription factors STAT1 and IRF1 activated by IFNs may bind ISRE or IRE sequences (28, 24 -31, 42). Ma et al. (50) recently reported the critical role of STAT1 in IFN-␤ and -␥ suppression of MMP-9 gene expression in using STAT1-deficient murine astrocytes and U3A cells. The present data are in agreement with these observations and reveal the mechanism beyond STAT1. We provide different lines of evidence for the involvement of IRF1 in the IFN-inhibited transcriptional response of the MMP-9 promoter. First, our transfection experiments using STAT1 (Ϫ/Ϫ) U3A cells confirmed the absolute requirement of STAT1 in the inhibition of TNF-␣-mediated MMP-9 activation. Second, ectopic expression of IRF1 in EW-7 cells prevented MMP-9 activation by TNF-␣, indicating that the presence of IRF1 was associated with MMP-9 gene inhibition. Third, IFNs induced IRF1 while not affecting I-B degradation and NF-B activation. Fourth, in contrast to STAT1, IRF1 bound to the MMP-9 promoter, specifically by an IFN-response element that overlapped with the non-consensus NF-B-binding site (Ϫ600 to Ϫ590). Fifth, IRF1 diminished the DNA binding of NF-B p50/p65 subunits, and conversely p50 or p65 proteins competed for IRF1-DNA binding. Finally, in cells cotransfected with IRF1 and p65 expression vectors, an excess of IRF1 protein reduces MMP-9 reporter activity, whereas an excess of p65 protein blocks the inhibitory effect of IFN-␥. Together, these results underline the direct inhibitory action of IRF1 on NF-B-mediated MMP-9 gene activation. Further support was given by mutation studies showing that mutation of the IRF1binding sequence (5Ј-GGAATTCC-3Ј), which blocks IRF1 binding, also blocked the binding of NF-B and consequently prevented the increase in activity of the MMP-9 promoter construct upon TNF-␣ treatment (data not shown). TNF-␣ may activate IRF1 directly through NF-B sequences located in its promoter (51). In the present report, TNF-␣ also increased IRF1 levels in EW-7 cells at 18 h. Thus, inhibition of MMP-9 expression in the long term may possibly be achieved by a complementary mechanism resulting from the combined effects of TNF-␣ and IFN that increase the synthesis of IRF1.
Recent evidence (19) indicates that IFN-␥-mediated inhibition of MMP-13 expression by transformed human epidermal keratinocytes is associated with STAT1 activation. Whether STAT1 is involved in direct suppression of MMP-13 gene transcription, however, remains to be demonstrated (19). In another study, IFN-␥ (through IRF1 induction) sensitized ME-180 cervical cancer cells to TNF-␣-induced apoptosis by inhibiting NF-B activity (51), but IRF1 did not interact with NF-B (52).
Interestingly, our findings have to be reconsidered in the context of the recent discovery of the negative regulation of FIG. 7. Synergistic induction of IL-6 by TNF-␣ and IFNs in EW-7 cells. EW-7 cells (1 ϫ 10 6 cells/ml) were cultured for 24 h in the absence or presence of TNF-␣ (10 units/ml) and/or IFNs (10 3 units/ml). A, IL-6 production (pg/ml) in the culture supernatants of EW-7 cells was determined by ELISA. B, EW-7 cells were transiently transfected with the human IL-6 promoter and then cultured for 24 h in the absence or presence of TNF-␣ or IFNs or TNF-␣ ϩ IFNs. Chloramphenicol acetyltransferase activity was determined from cell lysates as described under "Experimental Procedures." C, nuclear extracts prepared from EW-7 cells treated for 18 h with TNF-␣ (10 units/ml), or IFN-␥ (1000 units/ml), or both were incubated with the 32 P-labeled IL-6-NF-B probe, and the protein DNA binding was analyzed by EMSA. Nuclear extracts were assayed for protein DNA binding activity in the presence of p50, p65, or IRF1 antibodies. NI, non-immune serum; NS, nonspecific band. D, for direct DNA occupancy, recombinant proteins IRF1, p50, and p65 NF-B proteins (0.5 and 1 l) were mixed with the IL-6 probe, and the reaction products were then resolved by EMSA. MMP-9 expression by a metastasis suppressor Kiss-1 (53). Kiss-1 indirectly down-regulated NF-B binding to the MMP-9 promoter by preventing nuclear localization of NF-B (p50/p65) (34). The gene of Kiss-1 maps to chromosome 1q32, and its sequence predicts a secreted protein with a molecular mass of 15.4 kDa (54,55). Thus, IFNs and Kiss-1 may exert possible overlapping antitumoral activities through different modes of action, as seen for the regulation of MMP-9 expression.
By contrast, published data (37,45,(57)(58)(59) reported the existence of a synergistic in vitro and in vivo cross-talk between NF-B and IRF1 for the induction of inflammatory genes. NF-B synergizes with IRF1 on adjacent consensus sequences to induce the transcription of the genes coding for vascular cell adhesion molecule 1 (58), major histocompatibility complex class I (56), inducible nitric-oxide synthetase (59), and IFN-␤ (57). According to our previous results (37), we report here that the 5Ј-flanking region of the human IL-6 gene transfected in EW-7 cells is up-regulated in response to TNF-␣ and that IFNs synergize IL-6 activity through IRF1 and Sp-1 activation. The present data serve to demonstrate an opposite effect of IRF1 in which this molecule counteracts the effects of NF-B in the particular MMP-9 promoter context.
In other cell types, MMP-9 has been shown to influence cell invasion through proteolysis of the extracellular matrix (1,(3)(4)(5), leading to the subsequent release of the molecules sequestered in the extracellular matrix which exert themselves through direct chemotactic and/or proliferative activities on cells (1,(3)(4)(5). IFN-␤ in vitro diminished the capacity of leukocytes to infiltrate an endothelial cell monolayer through a reduction of released MMP-9 (23,27). In line with these observations, we found a good correlation between loss of EW-7 cell adhesion to collagen IV and MMP-9 proteolytic activity. Activation of recombinant MMP-9 with APMA inhibited EW-7 cell adhesion to collagen IV, thus suggesting the capacity of IFNs, through MMP-9 inhibition, control EW cell adhesion.
In summary, our study demonstrates that IRF1 is essential for the repression of MMP-9 expression by blocking NF-B binding to the collagenase promoter. The comprehension of this mechanism may provide new therapeutic strategies for targeting MMP-9 gene expression involved in tumor growth and metastasis.