Regulation of type I interferon gene expression by interferon regulatory factor-3.

The genes of the family of interferon (IFN) regulatory factors (IRF) encode DNA binding transcriptional factors that are involved in modulation of transcription of IFN and interferon-induced genes (ISG). The presence of IRF binding sites in the promoter region of IFNA and IFNB genes indicates that IRF factors recognizing these sites play an important role in the virus-mediated induction of these genes. We have described a novel human gene of this family, IRF-3, that is constitutively expressed in a variety of cell types. IRF-3 binds to the interferon-sensitive response element (ISRE) present in the ISG15 gene promoter and activates its transcriptional activity. In the present study, we examined whether IRF-3 can modulate transcriptional activity of IFNA and IFNB promoter regions. Our results demonstrate that IRF-3 can bind to the IRF-like binding sites present in the virus-inducible region of the IFNA4 promoter and to the PRDIII region of the IFNB promoter but cannot alone stimulate their transcriptional activity in the human cell line, 293. However, the fusion protein generated from the IRF-3 binding domain and the RelA(p65) activation domain effectively activates both IFNA4 and IFNB promoters. Cotransfection of IRF-3 and RelA(p65) expression plasmids activates the IFNB gene promoter but not the promoter of IFNA4 gene that does not contain the NF-kB binding site. Surprisingly, activation of the IFNA4 gene promoter by virus and IRF-1 in these cells was inhibited by IRF-3. These data indicate that in 293 cells IRF-3 does not stimulate expression of IFN genes but can cooperate with RelA(p65) to stimulate the IFNB promoter.

The genes of the family of interferon (IFN) regulatory factors (IRF) encode DNA binding transcriptional factors that are involved in modulation of transcription of IFN and interferon-induced genes (ISG). The presence of IRF binding sites in the promoter region of IFNA and IFNB genes indicates that IRF factors recognizing these sites play an important role in the virus-mediated induction of these genes. We have described a novel human gene of this family, IRF-3, that is constitutively expressed in a variety of cell types. IRF-3 binds to the interferon-sensitive response element (ISRE) present in the ISG15 gene promoter and activates its transcriptional activity. In the present study, we examined whether IRF-3 can modulate transcriptional activity of IFNA and IFNB promoter regions. Our results demonstrate that IRF-3 can bind to the IRF-like binding sites present in the virus-inducible region of the IFNA4 promoter and to the PRDIII region of the IFNB promoter but cannot alone stimulate their transcriptional activity in the human cell line, 293. However, the fusion protein generated from the IRF-3 binding domain and the RelA(p65) activation domain effectively activates both IFNA4 and IFNB promoters. Cotransfection of IRF-3 and RelA(p65) expression plasmids activates the IFNB gene promoter but not the promoter of IFNA4 gene that does not contain the NF-kB binding site. Surprisingly, activation of the IFNA4 gene promoter by virus and IRF-1 in these cells was inhibited by IRF-3. These data indicate that in 293 cells IRF-3 does not stimulate expression of IFN genes but can cooperate with RelA(p65) to stimulate the IFNB promoter.
Viral infection leads to the transient expression of early inflammatory genes. The proteins encoded by these genes enhance recognition of the infected cells by the host immune system. A group of proteins, called interferons (IFNs), 1 can directly inhibit viral replication. Type I IFNs are encoded by a family of closely related ␣ genes and a single IFNB gene, which are all localized on chromosome 9 (1,2). The sequences that regulate inducible transcription of these genes are localized within the 100-nucleotide 5Ј-end of the transcriptional start of IFNA and IFNB genes (3,4). These regions contain a number of a short overlapping GAAAGT-rich sequences that serve as a binding sites for multiple transcriptional factors. Several elements that function as positive regulatory domains (PRD) in virus-infected cells were identified in the promoter region of the IFNB gene (5). It was further shown that the region designated as PRDII contains an NF-kB binding site (6 -8) to which the NF-kB heterodimers induced upon viral infection bind as well as the HMG protein (9). The PRDIV domain was shown to bind the ATF-2 factor and octamer binding protein (10), and the PRDI and PRDIII regions serve as high affinity sites for the binding of the interferon regulatory factors (IRF-1 and IRF-2) (6,11,12). The transcriptional activation of this promoter, in virus-infected cells, is a result of the interaction among these multiple transcriptional factors (13) where the virus-induced binding of p50⅐p65 heterodimer plays a crucial role (14,15). In contrast, the IFNA gene promoter region does not contain an NF-B binding site. However, both the murine and human IFNA virus-inducible regions (VRE) contain multiple repeats to the GAAAGT and AAGTGA elements that could serve as binding sites for the IRF-1 and IRF-2 factors (16 -18). In addition, the deletion and mutation analysis of the murine IFNA4 promoter region identified an essential element, named ␣F1 (19), which serves as a recognition site for DNA binding proteins p68 and p96 (20). In the human IFNA1 promoter region, another essential binding site interacting with the TG protein was identified (21). Recent studies have identified a distinct multisubunit complex in virus-induced cells (22).
The presence of IRF-like binding sites in the promoter region of the IFNA and IFNB genes indicated that the IRF factors recognizing these sites play an essential role in the induction of IFN genes. The original results of Harada et al. (11) suggested that the up-regulation of these genes is mediated by IRF-1, while the closely related IRF-2 suppresses the expression of these genes. However, the essential role of IRF-1 and IRF-2 in the regulation of IFNA and IFNB gene expression in infected cells was disputed by the finding that mice containing the homozygous deletion of IRF-1 or IRF-2, or fibroblast derived from these mice, were able to induce IFNA and IFNB gene expression upon infection with Newcastle Disease virus (NDV) to the same level as the wild-type mice or cells (23,24). IRF-1 was shown to play an important role in the antiviral effect of IFNs (25). IRF-1 binds the ISRE element present in many IFN-inducible gene promoters and activates expression of some of these genes (26,27). However, activation of ISG genes by IFNA and IFNB was shown to be mediated generally by the multiprotein ISGF3 complex (28). The binding of this complex to DNA is mediated by another member of the IRF family, p48, which, in IFN-treated cells, interacts with phosphorylated STAT1 and STAT2 transcriptional factors forming the heterodimer complex, ISGF3 (29 -32). The homozygous deletion of p48 in mice abolishes the sensitivity of these mice to the antiviral effect of IFNs (25), and the infected macrophages from p48 Ϫ/Ϫ mice show an impaired induction of IFNA and IFNB genes (33). However, induction of IFN genes in virus-infected p48 Ϫ/Ϫ splenocytes is not affected.
Several other members of the IRF family have been identified. The ICSBP gene is expressed exclusively in the cells of the immune system (34,35), and its expression can be enhanced by IFN␥. ICSBP was shown to form a complex with IRF-1 and inhibit the transactivating potential of IRF-1 (36,37). The homozygous deletion of ICSBP in mice leads to the alteration in the development of the cell of macrophage lineage (38). Another lymphoid cell-specific IRF, Pip/LSIRF, was identified (39,40), which can interact with phosphorylated PU.1. It was shown that the Pip⅐PU.1 heterodimer can bind to the immunoglobulin light chain enhancer and function as a B cell-specific transcriptional activator. Expression of Pip/LSIRF can be induced by antigenic stimulation, but not by IFN, and it was recently shown that the Pip/LSIRF Ϫ/Ϫ mice failed to developed mature T and B cells (41). Another novel member of the IRF family, IRF-7, was recently identified by its ability to bind to an ISRElike element in the promoter region of the Qp gene of EBV (42,43). Furthermore, the genome of human herpesvirus 8 contains four open reading frames, which show homology to the cellular IRF family of genes (44). These data indicate that transcriptional factors of the IRF family may modulate not only the expression of cellular genes but viral genes as well.
We previously identified and described another member of the human IRF family, IRF-3 (45). The IRF-3 gene encodes a 55-kDa protein and is expressed constitutively in all tissues. Recombinant IRF-3 binds to the ISRE element of the IFNinduced gene, ISG15, and overexpression of IRF-3 activates transcription of this promoter in the transient expression assay. Viral infection or IFN treatment does not activate the expression of the IRF-3 gene.
In the present study, we address the question whether IRF-3 can modulate the expression of IFNA and IFNB genes. The IRF-1 site(s) plays an important role in the transcriptional activation of these genes; therefore, it is important to determine which member of the IRF family plays a role in the activation of IFNA and IFNB genes in virus-infected cells. Our results demonstrate that: 1) recombinant IRF-3 can bind to the IRF-like binding sites present in the virus-inducible region of the IFNA and IFNB promoters; 2) in 293 cells, overexpression of IRF-3 neither activates transcription of the IFNA or IFNB promoters in a transient expression assay nor induces expression of the endogenous IFN genes; 3) fusion of the IRF-3 DNA binding domain with the RelA(p65) transactivation domain generated a fusion protein that effectively activated both the IFNA and IFNB promoters; and 4) activation of the IFNA4 promoter region by IRF-1 in these cells is inhibited in the presence of IRF-3.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfections-293 cells were grown in Dulbecco's modified Eagle's medium. In the transfection assays, subconfluent 293 cells (2.5 ϫ 106 cells/plate) were cotransfected with 2.5 g of reporter plasmid chloramphenicol acetyltransferase (CAT) and the indicated amounts of the IRF or p65 expression plasmids, using the calcium phosphate coprecipitation method (46). When indicated, treatment with IFN␣ (500 units/ml) or infection with NDV (multiplicity of infection ϭ 5) was done 24 h after transfection for 16 h. Protein extracts, prepared by the freeze-thaw method, and CAT assays were done as described previously (19). To compensate for the possible differences in transfection efficiency, each sample was cotransfected with ␤-galactosidase expressing plasmid (2.5 g), and CAT activity was normalized to the constant levels of ␤-galactosidase (47). The total transfected DNA was kept constant in each experiment. Each transfection with the CAT reporter plasmids was repeated at least three times, and the data presented represents averages of these experiments.
Expression of GST Fusion Proteins-The GST-IRF-1 and IRF-3 fusion proteins were purified from bacterial lysates by affinity chromatography on a glutathione-agarose column (Sigma).

Modulation of IFNA4 Promoter Activity by IRF-3, IRF-1, and
Mutants of IRF-3-The amino-terminal region of IRF-3 shows a high degree of homology to IRF-1 and IRF-2 (Fig. 1A). Au et al. (49) showed that IRF-1 is an effective activator of the IFNA4 promoter in transient cotransfection assays, whereas IRF-2 inhibits the IRF-1-mediated activation (14). We have analyzed whether IRF-3 plays any role in the regulation of the IFNA4 promoter. The IFNA4 promoter contains an IRF-1 binding sites in its inducible region (IE) (Fig. 1B) and additional IRF-1-like sites can be located in the upstream region of this promoter. Human 293 cells were cotransfected with murine IFNA(Ϫ464)-CAT reporter plasmid and IRF-1 or IRF-3 (Fig. 2). Results of the transient transfection showed that IRF-1 is an effective transactivator as shown previously (20), whereas IRF-3 does not transactivate the IFNA promoter. Next, we wanted to determine whether IRF-3 modulates the IRF-1-mediated transactivation of this promoter. Cotransfection of IRF-1 and IRF-3 showed that IRF-3 inhibited IRF-1-mediated activation of IFNA promoter in a concentration-dependent manner. At the ratio of IRF-3 to IRF-1 of 2:1, the activation was reduced by 50%. The next question was if IRF-3-mediated inhibition of IFNA was through competition for the same binding site or due to protein-protein interaction as shown with ICSBP and IRF-1 (34). We, therefore, examined whether variants of IRF-3 with carboxyl-terminal deletions would also be inhibitory. Since proline-rich regions are often involved in protein-protein interaction, we also cotransfected IRF-1 with IRF-3 that had its proline-rich region deleted (Fig. 3). However, the proline-deleted form of IRF-3 could still inhibit IRF-1 activation of IFNA promoter, indicating that the proline-rich region of IRF-3 did not play any role in the observed inhibition. Plasmids encoding the truncated forms of the IRF-3 protein (327 and 240 aa) were also potent inhibitors of IRF-1 activity. These data suggest that IRF-3 and IRF-1 compete for the same or overlapping binding site on the IFNA promoter and that the truncated proteins may bind more strongly than the full-length IRF-3.
Activation of IFNA4 Promoters by IRF-3, IRF-1, and RelA (p65) Chimeric Fusion Proteins-Since IRF-3 could not transactivate the IFNA4-CAT promoter in 293 cells, we wanted to determine if this was a result of weak binding or the inability to transactivate this promoter in 293 cells. Therefore, a plasmid encoding a chimeric protein containing the amino-terminal binding domain of IRF-3 (1-133 aa) and the carboxyl-terminal transactivating domain of RelA(p65) (397-550 aa) was generated. 293 cells were cotransfected with IFNA4-CAT and the chimeric fusion protein IRF-3/p65 or with the previously described chimeric plasmid IRF-1/p65 used as a positive control (48). The levels of transactivation show that IRF-1/p65 and IRF-3/p65 chimeric fusion proteins were equally effective transactivators of the full-length IFNA(Ϫ464) promoter (Fig.  3), indicating that IRF-3 does bind but does not transactivate the IFNA4 promoter. In contrast to the IRF-3/p65 plasmids, cotransfection of IRF-3 and p65 plasmids in trans could not transactivate the IFNA promoter (data not shown). Since there is no obvious NF-kB binding site in IFNA promoters, the inability of p65 to cooperate with IRF-3 in trans further supports the idea that NF-kB binding does not play a role in the activa-tion of the IFNA4 promoter. Transfection of the chimeric fusion plasmids with a series of IFNA promoter deletions resulted in a gradual reduction in reporter gene transactivation, suggesting the presence of multiple IRF-1 and IRF-3 binding sites in the upstream (Ϫ464) region of the IFNA promoter, which may act in a cooperative manner.
Virus Induction of IFNA4-CAT-We previously found that in L929 cells transfection of IRF-3 alone did not activate the IFNA promoter, but it enhanced NDV-induced activation (45). To determine whether IRF-3 also cooperates with NDV in 293 cells, the cells were cotransfected with IFNA4-CAT, IRF-3, or IRF-1 and infected with NDV. As shown in Fig. 4, NDV induces the IFNA promoter effectively, but the induction is inhibited in the presence of IRF-3. In contrast, combination of IRF-1 with the viral infection shows an additive effect on the activity of the IFNA4 promoter as previously shown by Au et al. (49). Effect of IRF-3 on the Activation of the IFNB Promoter-The VRE of IFNB promoter contains two IRF binding sites (PRDI and PRDIII) and an NF-kB binding site (PRDII). To examine the effect of IRF-3 on the activity of IFNB promoter, we used the IFNB-CAT plasmid, which contains the IFNB promoter (Ϫ281 to ϩ19) inserted in front of the CAT gene (Fig. 5A). In a transient transfection assay, overexpression of IRF-1 and IRF-3 did not transactivate the IFNB-CAT promoter in 293 cells, while a slight activation of this promoter was seen upon cotransfection with RelA(p65) expression plasmid. When IRF-1 and IRF-3 expressing plasmids were cotransfected together with the RelA(p65) expression plasmid (in trans), transactivation of the IFNB promoter was observed (13-and 8-fold, respectively), demonstrating that IRF-3 and IRF-1 can synergize with p65 and activate transcriptional activity (2-fold) of the IFNB promoter. The chimeric fusion protein IRF-1/p65 did not significantly transactivate the IFNB promoter. However, the IRF-3/p65 chimeric fusion protein was a strong transactivator (16fold), suggesting that IRF-3/p65 is able to bypass the requirement for cooperation between NF-kB and IRF-3 binding for the activation of the IFNB promoter (14,15). These results also suggest that IRF-3 can strongly interact with one or more PRD domains of the IFNB promoter.
To further characterize the IFNB transactivation mediated by IRF-3, we used the TH-CAT reporter, which contains a tetrahexamer (AAGTGA) element that functions as a strong binding site for IRF-1 (Fig. 5B). As expected, both IRF-1 and IRF-1/p65 chimeric fusion proteins strongly transactivated reporter gene activity. However, transactivation of TH-CAT by IRF-3/p65 was only marginal, suggesting that IRF-3 does not bind effectively to this PRDI-like element. These data suggested that IRF-3 binding to the IFNB promoter and IRF-3/p65 transactivation may be mediated by an IRF site present in the IFNB promoter that is distinct from the PRDI element.
Binding of IRF-3 to the PRD Domains of the IFNB Promoter-Binding of recombinant IRF-3 to the PRD domains of the IFNB promoter was, therefore, analyzed by gel mobility shift analysis. The amino-terminal 133-aa DNA binding fragment of IRF-3, produced as a GST-fusion protein in Escherichia coli, bound with high affinity to PRDIII but only weakly to PRDI; no stable binding with PRDII was observed (Fig. 6). Interestingly, the IRF-3 protein bound efficiently to probes consisting of PRDI-II and PRDIII-I regions. At low concentrations of IRF-3, only a single DNA-protein complex was formed with the PRDIII-I probe; however, as the concentration of IRF-3 was increased, a second complex of slower mobility was formed, indicating a cooperative interaction of IRF-3 with other sites within the PRDIII-I element.
Binding of IRF-3 to the IFNA Promoter-Since IRF-3 displayed strong binding to the PRDIII region of the IFNB promoter, we searched for the sequence homologies to PRDIII (AGGAAAACTG) region in the IFNA promoter. Two sequences similar to PRDIII were found. One sequence TGGAGTAGTG is located at positions Ϫ252 to Ϫ243, and the second, GTGAAAA-GAG, is located in the IE region of the promoter from position Ϫ94 to Ϫ85. This latter sequence overlaps with the IRF-1 binding site. The binding of GST-IRF-3 and truncated GST-IRF-3 (133 aa) to the Ϫ252 to Ϫ243 radiolabeled oligonucleotide probe was analyzed by EMSA. However, neither the fullsize IRF-3 nor its amino-terminal fragment were able to bind this sequence (data not shown). We next analyzed the binding of the GST-IRF-3 proteins to IE probe (Ϫ109 to Ϫ75) and compared it with binding of GST-IRF-1 to this probe. As shown in Fig. 7A, 50 ng of GST-IRF-1 bound the radiolabeled IE probe, resulting in a formation of single complex. As the concentration of protein was increased from 50 to 100 ng, a second complex of slower mobility was formed. Full-length GST-IRF-3 bound very effectively to the IE probe with formation of multiple bands When the concentration of GST-IRF-3(133) was increased from 15 to 50 ng, a second IRF-3/DNA complex of a slower mobility was formed (Fig. 7B). In addition, lower amounts of GST-IRF-3 were required for formation of the slower mobility complex than that of GST-IRF-1, suggesting that IRF-3 binds more tightly to the IE probe than GST-IRF-1. Competition studies showed that formation of the IRF-3(133) homodimers (slower mobility complex) was prevented by 50-fold excess of cold IE probe but to compete formation of the faster mobility complex required 100-fold excess of cold IE probe (Fig. 8C). The formation of GST-IRF-3(133) complexes were also prevented in the presence of oligomers corresponding to ISRE (ISGI5) or by a tetramer of the PRDI site but not by the oligonucleotide corresponding to the ␣F1 probe (49). Since the ␣F1 probe did not bind the GST-IRF-3 protein (data not shown), we concluded that the IRF-3 binding site is located in the 3Ј portion of the IE region but does not extend into the 5Ј portion of the IE, where the IRF-1 and ␣F1 binding sites overlap (Fig. 1B).
To determine a putative binding region for IRF-3, we made point mutations in the IE region of IFNA4 that showed homology to PRDIII and flanking regions. As shown in Fig. 8A, a mutation at nucleotide 92, which is part of the putative IRF-1 binding site, resulted in a decrease in the levels of the slower mobility complex. Point mutations at positions 87 and 103 of IE did not affect the binding; however, both of these nucleotides are external to the region of PRDIII sequence homology. We have previously shown that mutations at positions 103 and 92 decrease virus-mediated activation of the IFNA4 promoter. 2 However, the change in nucleotide 90 (A to G) decreased overall binding of GST-IRF-3(133) as shown in Fig. 8B (a lighter exposure than Fig. 8A). This part of IE has homology to PRDIII. Competition studies using cold IE90 oligonucleotide demonstrated that this oligonucleotide competed less efficiently than IE (data not shown), supporting the conclusion that nucleotide 90 is important for IRF-3 binding. The double mutant, which had alterations at positions 92 and 87, showed only a slight reduction in IRF-3 binding when compared with IE (Fig. 8B) and competed with binding to labeled IE probe as efficiently as wild-type IE (data not shown). Thus, from this gel shift analysis, we conclude that IRF-1 and IRF-3 share overlapping binding sites in the IFNA promoter and that the GAAAA sequence is important for IRF-3 binding.

DISCUSSION
In this paper, we have analyzed the role of IRF-3 in the virus-mediated induction of the IFNA and IFNB genes. Our results show that overexpression of IRF-3 in 293 cells does not activate the transcription of IFNA4 or IFNB promoter regions in a transient expression assay while it can bind to the VRE of the IFNA and IFNB genes. Two IRF-1 binding sites, PRDI and PRDIII, were identified in the virus-inducible region of the IFNB promoter. Our binding studies demonstrated that the recombinant IRF-3, or its 3Ј-deleted mutant (133 aa), binds strongly to PRDIII but not to PRDI. When binding of IRF-3 to an oligonucleotide encompassing PRDIII-PRDI was analyzed, only a single protein-DNA complex was detected at low concentration; however, at higher IRF-3 protein concentrations, IRF-3 bound as two molecules either as a dimer or more likely to two distinct sites. This result indicates cooperativity of IRF-3 bind-2 W. C. Au and P. M. Pitha, unpublished data.  (133) were incubated with radiolabeled IE probe and resolved by EMSA. C, 50 ng of GST-IRF-3(133) was preincubated with the 50ϫ and 100ϫ cold IE probe or 50ϫ cold ISRE, PRDI (which is a tetramer of the PRDI site), and ␣F1 probes followed by incubation with radiolabeled IE and resolved by EMSA.
ing to high affinity (PRDIII) and low affinity (PRDI) binding sites.
Although IRF-1 contains an activation domain in its carboxyl-terminal end (48), it cannot by itself activate transcription of the IFNB promoter in a transient transfection assay. In our previous study (45), we had not detected the presence of an activation domain in IRF-3. However, our recent results with the two-hybrid yeast assay 3 as well as the results of others (50) indicate that IRF-3 has transactivation potential when expressed as a Gal4 fusion protein. Similarly to IRF-1, IRF-3 alone does not activate the IFNB promoter in 293 cells. In contrast, cotransfection of IRF-3 with a RelA(p65) expressing plasmid resulted in transcriptional activation of the IFNB promoter. These data indicate that the interaction between NF-kB (p50/p65) and IRF-3 is sufficient for a transcriptional activation of this promoter. Interestingly, the requirement for this interaction can be bypassed by the IRF-3/p65 and IRF-2/p65 (48) chimeric fusion proteins but not by the chimeric fusion protein IRF-1/p65. We assume that this difference indicates that IRF-3 binds more strongly to the IFNB promoter than IRF-1. It has been shown that virus-induced binding of p50/p65 heterodimer to the NF-kB site in the IFNB promoter plays a critical role in the activation of IFNB promoter in infected cells (14,15). Mutations of the PRDI site in the IFNB promoter that abolish binding of IRF-1 were shown to decrease virus-mediated inducibility of this promoter (6,51), indicating that nuclear factor binding to this element is essential for the transcriptional activation of this promoter. However, neither IRF-1 nor IRF-3 can act alone as a transcriptional activator of this promoter in transient cotransfection assays, indicating that the activation may require cooperation between different transcription factors (14,15). The requirement of complex formation between Pip and PU.1 for activation of the immunoglobulin light chain gene was recently demonstrated (39). Our recent data 4 shows complex formation between IRF-3 and the p300/ CBP proteins, which are basal components of the transcriptional complexes (52). This interaction is facilitated by virusmediated phosphorylation of IRF-3. Since the interaction of RelA(p65) with the amino-terminal domain of p300/CBP was recently demonstrated (53), it is likely that the cooperation between IRF-3 and p65 may be facilitated by p300/CBP. Promoters of the various IFNA genes differ from the majority of other cytokine gene promoters by the absence of a consensus NF-kB binding site. Nevertheless, these genes are induced in cells of lymphoid origin by viral infection in a cell type-specific manner (17,54). Au et al. (49) has shown that, in a transient expression assay in L929 cells, overexpression of IRF-1 can induce both the murine IFNA4 promoters and synergize with the NDV-activated stimulation of the IFNA4 promoter. Furthermore, in these cells, overexpression of IRF-3 significantly enhanced virus-mediated activation of this promoter (45). In contrast, in 293 cells, IRF-3 does not active the IFNA4 promoter region but inhibits the IRF-1-mediated activation. The IFNA4 promoter contains in its inducible element (IE) an IRF binding site (55); the right half of the IE element of the IFNA4 promoter contains four adenosine residues (GAAAAG), thus resembling the half of the PRDIII site (GAAAAC) that strongly binds IRF-3. This is in contrast to the PRDI binding site that contains only three adenosine nucleotides (GAAAG) and is a weak binding site for IRF-3 but a strong site for IRF-1. In this study, we have shown that IRF-3 binds effectively to IFNA4 IE and that mutations in positions 92 and 90 affect IRF-3 binding, while the replacement of the 3Ј G by A in position 87 is without any effect. By mutation analysis, we have previously shown (49) that the IE site is important for the virus-inducible transcriptional activation of this promoter and that a single mutation at nucleotide 92 completely abolished virus-mediated induction of the IFNA4 promoter in L929 cells. 2 Two other IRF family members have been identified that inhibit IRF-1 mediated transactivation. IRF-2 shows the same DNA binding specificity as IRF-1 and inhibits IRF-1-stimulated transcription of both the IE and ISRE elements by competing for the binding site (48,56). In contrast, ICSBP, which binds only weakly to the ISRE element and not to PRDI, inhibits the IRF-1 transactivation by direct interaction with the 3Ј-end of the IRF-1 protein, thus interfering with its transactivation capability (34,37). Our data indicate that the observed IRF-3-mediated inhibition of IRF-1 in 293 cells occurs through competition for the binding site(s) since the 5Ј part of IRF-3(133) protein, containing only the DNA binding domain, is also an effective competitor of IRF-1. Furthermore, we have been unable to demonstrate a direct interaction between immobilized GST-IRF-3 and IRF-1 (data not shown). These data indicate that several members of the IRF family may function as conditional repressors of IRF-1-targeted genes. Since IRF-1 is effectively induced both by virus and IFNs, the expression of IRF-1-regulated genes may reflect a balance between levels of IRF-1 and the negative regulators of the IRF family. However, experiments with mice in which various IRF genes were deleted indicate that the role of the transcriptional factors of the IRF family extends beyond the response to viral infection. Most of these factors were found to be critical for the proper development and/or function of immune lineage cells. In this respect, it is interesting to note that IRF-4/LSIRF/Pip can be induced only by antigenic stimulation of lymphocytes (41) and IRF-3 can be induced in peripheral blood mononuclear cell and macrophages by phytohemagglutinin of phorbol 12-myristate 13-acetate, respectively. 5 In conclusion, our data demonstrate that IRF-3 modulates the expression of type I IFN genes in 293 cells. It cooperates with RelA(p65) to stimulate the transcriptional activity of the IFNB gene promoter, while it inhibits IRF-1 and virus-mediated transactivation of the IFNA4 gene promoter. Studies done since this manuscript has been reviewed indicate that in mouse cells, containing homozygous deletions of IRF-1 and IRF-2 genes, and in embryo fibroblasts, high levels of overexpression of IRF-3 alone can stimulate transcriptional activity of IFNA4 and IFNB promoters and greatly enhance the virus-mediated induction of these promoters. Interestingly, our preliminary results also indicate that overexpression of E1A can inhibit the IRF-3-mediated transcriptional modulation of the IFNA4 and IFNB promoters. The fact that we failed to detect activation of IFNA4 and IFNB promoters by IRF-3 alone in 293 cells may be due to the expression of E1A in these cells. 6 Although the exact role of IRF-3 in the virus-mediated pathway is presently unclear, the fact that IRF-3 is phosphorylated in infected cells but not in IFN-treated cells 4 suggests that IRF-3 may be an important component of virus-induced signaling.