Promoter Organization of the Interferon-A Genes Differentially Affects Virus-induced Expression and Responsiveness to TBK1 and IKKϵ*

Virus-induced expression of interferon (IFN)-A genes is regulated by two members of the IFN regulatory factor (IRF) family, IRF-3 and IRF-7, which are activated by phosphorylation during viral infection by the IKK-related serine/threonine kinases TBK1 and IκB kinase ϵ (IKKϵ). In this study, we demonstrate that three IRF-binding sites located in the virus-responsive element mediate the transcriptional activation of the IFN-A4 promoter by IRF-3. The precise arrangement of these IRF elements is required for synergistic activation of the IFN-A4 promoter following Newcastle disease virus infection or activation by TBK1 or IKKϵ. The ordered assembly of IRF-3 multimers on the promoter also determines cooperative recruitment of IRF-3 and CREB-binding protein and differential virus-induced expression of IFN-A4 gene promoter compared with IFN-A11. Naturally occurring nucleotide substitutions disrupt two of the IRF elements in the IFN-A11 gene promoter, leading to a dramatic decrease in IRF-3 and CREB-binding protein recruitment and in IRF-3-dependent transcription. Transcription of the IFN-A4 promoter by IRF-7 is mediated by two IRF elements; promoter mutants that carry a reversed IRF element retain the ability to respond to IKKϵ or TBK1 expression in the presence of IRF-7 but lose the capacity to respond to virus or kinase-induced IRF-3. Interestingly, IKKϵ or TBK1 stimulates the IRF-7-mediated transcription of IFN-A11, although at a lesser extent compared with IFN-A4. Our data indicate that virus-induced expression of IFN-A genes is dictated by the organization of IRF elements within the IFN-A promoters and that the differential IFN-A gene expression, based on the IRF-3 responsiveness, is partially compensated in the presence of IRF-7 when both factors are activated by IKKϵ or TBK1.


Virus-induced expression of interferon (IFN)-A genes is regulated by two members of the IFN regulatory factor (IRF) family, IRF-3 and IRF-7, which are activated by phosphorylation during viral infection by the IKK-related serine/threonine kinases TBK1 and IB kinase ⑀ (IKK⑀). In this study, we demonstrate that three IRF-binding sites located in the virus-responsive element mediate the transcriptional activation of the IFN-A4 promoter by IRF-3. The precise arrangement of these IRF elements is required for synergistic activation of the IFN-A4 promoter following Newcastle disease virus infection or activation by TBK1 or IKK⑀. The ordered assembly of IRF-3 multimers on the promoter also determines cooperative recruitment of IRF-3 and CREB-binding protein and differential virus-induced expression of IFN-A4 gene promoter compared with IFN-A11. Naturally occurring nucleotide substitutions disrupt two of the IRF elements in the IFN-A11 gene promoter, leading to a dramatic decrease in IRF-3 and CREB-binding protein recruitment and in IRF-3-dependent transcription. Transcription of the IFN-A4
promoter by IRF-7 is mediated by two IRF elements; promoter mutants that carry a reversed IRF element retain the ability to respond to IKK⑀ or TBK1 expression in the presence of IRF-7 but lose the capacity to respond to virus or kinase-induced IRF-3. Interestingly, IKK⑀ or TBK1 stimulates the IRF-7-mediated transcription of IFN-A11, although at a lesser extent compared with IFN-A4.

Our data indicate that virus-induced expression of IFN-A genes is dictated by the organization of IRF elements within the IFN-A promoters and that the differential IFN-A gene expression, based on the IRF-3 responsiveness, is partially compensated in the presence of IRF-7 when both factors are activated by IKK⑀ or TBK1.
The immediate cellular response to virus infection is characterized by the transcriptional activation of type I IFN 4 genes involved in the host antiviral defense program (1). Multiple IFN-A gene family members are differentially expressed in cultured cells or embryonic fibroblasts infected by virus (2)(3)(4), and in vivo studies indicate that the virus species as well as the natural route of infection determines the magnitude of type I IFN production and the nature of the producer cell. For example, plasmacytoid dendritic cells are identified as the major IFN␣/␤-producing cells that rapidly express high levels of IFN-A during systemic infection and acute viremia, whereas local viral infections stimulate delayed and lower IFN-A expression, requiring a positive feedback amplification mechanism dependent on IFN-␤ production in cell types other than IFN␣/␤-producing cells (5)(6)(7)(8). These studies also indicated that differential IFN-A gene regulation is critical for the development, magnitude, and duration of innate and adaptive antiviral responses in the course of viral infection.
Different members of the interferon regulatory factor (IRF) family participate in the IFN-A gene regulation, either positively or negatively, although virus-induced differential IFN-A gene expression is essentially mediated by IRF-3 and IRF-7 (9 -12). These factors possess overall structural and biochemical similarities, including their virus-induced activation by C-terminal serine phosphorylation, dimerization, nuclear transport, and DNA binding to different IRF elements in the promoter of target genes (13)(14)(15)(16)(17)(18). IRF-3 is constitutively expressed in all cell types, whereas the expression of IRF-7, constitutive in lymphoid cells and certain subsets of dendritic cells, is induced by IFN-␣/␤ produced during viral infection in other cell types (19,20). The ubiquitous constitutive expression of IRF-3 implies that this factor is required for the primary activation of IFN-A, IFN-B, and a subset of IFN-stimulated genes, yet generation of mice deficient for IRF-3 or IRF-7 indicated that IRF-7 functions even in the absence of IRF-3 and attributed a critical role for both factors in the activation of the initial phase of IFN-A/B gene induction (21). Once the initial activation of genes is achieved, according to a multistep IFN-A/B gene expression model established for mouse embryonic fibroblasts, IRF-7 cooperates with IRF-3 to amplify IFN-A4 and IFN-B gene transcription and induce a set of delayed mouse IFN-A genes, including IFN-A2, -A5, -A6, and -A8 (18,22,23). Among the different members of mouse IFN-A multigenic family, the IFN-A4 gene is the only immediate early gene expressed following virus infection in different cell lines and exhibits higher expression levels compared with other IFN-A genes (2,3,22,24,25). In this regulatory network, IRF-3 and IRF-7 are potent activators of virus-induced IFN-A gene transcription, although the mechanisms of their cooperative interaction with target sequences remain to be elucidated.
IRF-3 phosphorylation is initiated by binding of double-stranded RNA, a product of viral replication considered as an early inducer of type I IFN synthesis, on membrane-associated Toll-like receptor-3 that leads to the recruitment of the specific TIR domain-containing adaptor TRIF and the activation of two serine/threonine kinases, TBK1 (TANKbinding kinase) and IKK⑀, that phosphorylate IRF-3 (26,27). IRF-3 activation also occurs upon recognition of double-stranded RNA by RIG-I, a retinoic acid-inducible RNA helicase involved in cytoplasmic immune surveillance (28). TBK1 or IKK⑀ is required for IRF-3 activation by RIG-I, the mitochondrial antiviral signaling protein MAVS functioning as an adaptor between RIG-I and both kinases (29 -33). IRF-3 phosphorylation by TLR4, a receptor for LPS from Gram-negative bacteria, also requires the recruitment of TBK1 or IKK⑀ via the adaptor proteins TRIF and TRAM (34). Ligand binding on TLR-7 and -8 that deliver signals following their interaction with viral single-stranded RNA or TLR-9 that recognizes unmethylated CpG-containing DNA and senses infection by cytomegalovirus or herpes simplex virus type-1 leads directly to IRF-7 activation (35)(36)(37). TLR7-and TLR-9-dependent pathways require the general TLR adaptor, MyD88 (myeloid differentiation factor) associating with TRAF6 and the serine-threonine kinase IRAK1 that induces IRF-7 phosphorylation (38 -40). Whether IRF-7 mediates differential activation of IFN-A genes when TLR7 or TLR-9-dependent signaling is triggered in plasmacytoid dendritic cells infected by virus is unknown. TBK1 and IKK⑀ are not involved in these signalings, and the relative transcriptional capacity of IRF-7 activated by IRAK1 or these kinases has to be determined. Studies with TBK1-and IKK⑀-deficient mice have confirmed the critical role of TBK1 for IFN-A/B or IFNstimulated gene expression in mouse embryonic fibroblasts following virus infection (41)(42)(43)(44), although TBK1 is not pivotal for IRF-3 or IRF-7-mediated gene transcription in bone marrow-derived macrophages or thymocytes where high levels of constitutive IKK⑀ are detected. These studies have clearly demonstrated the involvement of TBK1 and IKK⑀ in virus-induced type I IFN expression but did not determine the individual role of these kinases in IFN-A gene regulation.
To determine the individual effects of IKK⑀ and TBK1 on the differential expression of IFN-A genes and elucidate the mechanism of the cooperative transcription of IFN-A gene promoters by IRF-3 or IRF-7, the effect of each kinase on IRF3-or IRF7-mediated transcription of mouse IFN-A4 and IFN-A11 gene promoters was evaluated. We demonstrate that the precise arrangement of multiple IRF elements within the IFN-A4 gene promoter determines in vivo recruitment of IRF-3 and CBP in correlation with virus inducibility. This organization is also critical for the differential stimulation of IFN-A4 gene promoter activity by IKK⑀ and TBK1. Our data indicate that naturally occurring nucleotide substitutions that alter the IRF-3 response of IRF elements on different IFN-A gene promoters dictate the differential expression of IFN-A genes.

EXPERIMENTAL PROCEDURES
Cell Culture and Virus Infection-Human embryonic kidney 293 (ATCC CRL 1573), mouse L929 (ATCC CCL 1), and human HeLa S3 cells (ATCC CCL 2.2) were grown in ␣-minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum, glutamine, and antibiotics. Infection of L929 and HeLa cells with Newcastle disease virus (NDV; La Jolla strain) and HEK293T cells with Sendai virus were performed at a multiplicity of infection of 1-5. After a 1-h incubation in serum-free medium, cells were washed and placed in growth medium containing 5% fetal bovine serum for 7 h.
Plasmids-Reporter plasmids pIF4T and pIF11T carrying the Ϫ470 to ϩ19 fragment of the mouse IFN-A4 gene promoter and the Ϫ457 to ϩ19 fragment of the mouse IFN-A11 gene promoter, respectively, and the pIF4J and pIF11J constructs containing the Ϫ120 to ϩ19 fragment of both promoters, upstream of the chloramphenicol acetyltransferase (CAT) or luciferase reporter gene have been previously described (45).
Transfections and Reporter Assays-Human HEK 293T cells were seeded at 7 ϫ 10 4 cells in 12-well plates in minimal essential medium supplemented with 5% fetal calf serum. After 18 -24 h, 0.5 g of reporter plasmid and 62.5 ng of reference plasmid pIEF-LacZ in the presence of various amounts of IRF-or kinase-expressing plasmids (see the legends to Figs. 7 and 9) were transfected by the calcium phosphate method (48). 48 h after transfection, cells were harvested to monitor the enzymatic activities. The CAT or luciferase values, normalized according to the ␤-galactosidase levels, were obtained in at least three independent transfection experiments, each performed in triplicate with two or three different clones. Transcription levels were expressed as relative transcription (RT) and inducibility values correspond to either the ratio between RT values obtained in overexpression conditions (with IRFand/or kinase-expressing plasmids) to values obtained by cotransfection of the empty vector pcDNA3 or to the ratio between RT values obtained in virus-induced and in mock-induced conditions.
Electrophoretic Mobility Shift Assays-The electrophoretic mobility shift assay was performed with probes corresponding to the 4J, 4J [Brev] and 4J[Crev] fragments generated by PCR of the corresponding plasmid using a [␥-32 P]ATP (Amersham Biosciences; 3000 Ci/mmol) 5Ј-labeled upstream primer corresponding to the Ϫ119 to Ϫ103 region of IFN-A4 promoter and an unlabeled downstream primer corresponding to the Ϫ53 to Ϫ31 region of the IFN-A4 promoter. The recombinant glutathione S-transferase proteins fused with the DNA-binding domain (DBD) of IRF-3 (residues 1-133) and IRF-7 (residues 1-150) were previously described (47). Binding reactions performed in the presence of 20 g/ml IRF-3(DBD), 400 g/ml IRF-7(DBD), or both recombinant proteins together for 30 min at room temperature and electrophoresis were described previously (46). The DNA-protein complexes were resolved on a 7% polyacrylamide (30:1) gel in 25 mM TEB, pH 8.3. The bound DNA was quantitated using a Storm PhosphorImager (Amersham Biosciences).

Preferential Binding Sites for IRF-3 and IRF-7 Define the Virus-responsive Element of the Mouse IFN-A4 Gene Promoter-The IFN-A4
gene is the only immediate early IFN gene expressed following virus infection in different murine cell lines (25). Maximal virus-induced transcription of the mouse IFN-A4 gene requires the presence of the virus-responsive element, VRE-A4, delimited to the Ϫ120 to Ϫ40 region of IFN-A4 gene promoter (45,52). IRF-3 and IRF-7 are potent activators of virus-induced IFN-A gene transcription, although their mechanism of interaction with target IFN-A sequences remains to be elucidated. To define the specificity of IRF-3 and IRF-7 for the three IRF-elements (B, C, and D) located in the VRE-A4, mutated IFN-A4 promoters were generated by substituting these motifs with the GAAATA motif ( Fig. 1), which was previously shown to be unresponsive to both IRF-3(5D) and IRF-7A ectopic expression (46). The capacity of these mutated VRE-A4 constructs to respond to virus or to activation by IRF-3(5D) or IRF-7A was then examined (Table 1).
Transcription mediated by IRF-3 was mainly affected by Ϫ81,Ϫ80TG/TA pair substitution (7-fold decrease of transcription with MC1 construct compared with pIF4T), and to a lesser extent by Ϫ94,Ϫ93GT/TA, Ϫ75,Ϫ74GC/TA, or Ϫ46,Ϫ45CT/TA substitutions (4 -5-fold decrease observed in MB1, MC2, and MD2 constructs). Among the doubly mutated versions of the IFN-A4 promoter, IRF-3dependent transcription was impaired more than 20-fold in the MC1-2 construct carrying the Ϫ81,Ϫ80TG/TA and Ϫ75,Ϫ74GC/TA substitutions. Inhibition was due to the Ϫ81,Ϫ80TA pair, since IRF-3-mediated transcription was not significantly affected in the MC2D1 mutant also carrying the Ϫ75,Ϫ74GC/TA pair substitution. The 7-fold decrease observed with MB2C1 was also due to the Ϫ81,Ϫ80TA pair, since the same -fold inhibition was obtained with MC1. Thus, the TG residues in the GAATTGGAAAGC sequence (C element) were pivotal for the transcription of the IFN-A4 promoter by IRF-3. Comparison of the data obtained with MB1, MB2, and MB1-2 further indicated that the 5-fold decrease in activation observed with the MB1-2 construct was due to Ϫ94,Ϫ93GT/TA pair substitution and indicated that the sequence GAAAGTGAAAAG (B element) in VRE-A4 defines another binding site for IRF-3. In contrast, the GG/TA and CT/TA substitutions in the GAAAGGAGAAACT sequence (D element) only slightly affected the IRF-3 response, suggesting that IRF-3 weakly recognized the D element.
Nucleotide substitutions in any IRF element had minor effects on IRF-7-driven transcription of the IFN-A4 promoter (Table 1); IRF-7mediated activation was affected by the Ϫ94,Ϫ93GT/TA pair substitution in the B element that decreased activation of MB1 and MB1-2 mutants more than 3-fold compared with pIF4T. As observed with IRF-3, the GAAAAG motif in the B element appeared dispensable for IRF-7-mediated transcription (1.5-2-fold decrease with MB2 and MB2C1 compared with pIF4T). IRF-7-mediated transcription was only slightly affected by substitutions in the C element. The 3.3-fold inhibition observed with MD2 compared with the 2.6-and 2-fold inhibition observed with MD1 and MD1-2, respectively, indicated that Ϫ46,Ϫ45CT/TA substitutions in the D element affected the IRF-7-mediated transcription.
Interestingly, virus inducibility was reduced by 20-fold in the doubly mutated promoter mutant MB1-2, which was deficient in its response both to IRF-3 and IRF-7; the MC1-2 mutant was defective only in its response to IRF-3 (25-fold inhibition) ( Table 1). The importance of the D element in the virus responsiveness of this promoter was highlighted by the 12-fold decrease in inducibility observed in the MD1-2 mutant. Taken together, these results demonstrate that disruption of any IRF element, B, C, or D, severely impaired the virus inducibility of the IFN-A4 gene promoter. Although three IRF elements that form VRE-A4 participate to different extents in the cooperative transcription mediated by IRF-3, the C element is crucial; similarly, the B element is pivotal for IRF-7 transactivation.

Reversal of IRF Element B or C Abolishes IRF-3-mediated Cooperative Transcription and Virus Inducibility without Affecting IRF-7-dependent
Transactivation of IFN-A4 Gene Promoter-Virus-induced expression of the IFN-B gene requires correct combination of different factors (ATF-2/c-Jun, IRF-3, IRF-7, and NF-B) with the architectural factor HMGI(Y) and their alignment on the same face of the DNA helix. The assembly of an enhanceosome on the IFN-B gene promoter has been well characterized (53)(54)(55), whereas formation of IFN-A enhancesosomes involving IRF-3 or IRF-7 has not been documented. To determine whether the orientation of an IRF element in the IFN-A promoter is critical for transcription by IRF-3 or IRF-7, IFN-A4 promoter mutants pIF4J[Brev] and pIF4J[Crev] carrying the B or C element in reverse orientation were generated ( Fig. 2A); these mutants were tested for responsiveness to IRF3(5D), IRF7A, virus infection, or IKK⑀/TBK1 expression. Reversal of the B element in pIF4J[Brev] abolished IRF3(5D)-mediated transcription without affecting the transcription by IRF-7A (Fig. 2B). The increase in inducibility (260-fold compared with 80-fold) was due to the 3-4-fold decrease in the constitutive expression of the reporter construct. The electrophoretic mobility shift assay performed with 4J and 4J [Brev] probes indicated that reversal of the B element did not affect the c1 complex resulting from the binding of IRF-3(DBD) homodimers to the B, C, or D element of the IFN-A4 promoter (Fig. 2D, lanes 1 and 2). Reversal of the B element affected only the formation of high stoichiometry electrophoretic mobility shift assay   Differential IFN-A Gene Regulation by IKK⑀ and TBK1 FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 complexes c2 and c3, hallmarks of cooperative binding of IRF-3(DBD), representing IRF3(DBD) homodimers that interact with two and three IRF elements, respectively. Similarly, when both IRF-3 and IRF-7 DBDs were used, only the high stoichiometry complexes c6 and c7 containing multimers of IRF-3 and IRF7 were affected (Fig. 2D, lanes 7 and 8). As expected, IRF-7 binding was not altered by reversal of the B element.
Reversal of the C element in pIF4J[Crev] had the same differential effect on IRF-3 and IRF-7 transcriptional and binding activities (Fig. 2, B  and D, lanes 3, 6, and 9), indicating again the critical role of the orientation of IRF elements in cooperative transcription mediated by IRF-3. Most importantly, this particular arrangement was also crucial for virus-induced transcription, since reversal of the B or C element totally abolished the NDV responsiveness of IFN-A4 promoter, as indicated by the 108-fold inducibility of pIF4J compared with the 5-8-fold inducibility values of pIF4J [Brev] or pIF4J [Crev] in L929 cells (Fig. 2C). These results clearly indicated that cooperative transcription mediated by IRF-3 dimers bound to three IRF-elements on the IFN-A4 promoter correlates with virus-induced transcription and requires the ordered assembly of protein-DNA complexes in the proximal region of the IFN-A4 promoter. In contrast, transactivation by IRF-7 homodimers, which does not necessarily correlate with the virus responsiveness of the promoter, may require a distinct organization.
Differential Transcription of Mouse IFN-A4 and IFN-A11 Gene Promoters-We previously showed that the weak virus-induced expression of the IFN-A11 gene compared with IFN-A4 was due to the presence of two nucleotide substitutions in the IFN-A11 gene promoter (45). These substitutions, Ϫ78A/G and Ϫ57G/C, are located in the second GAAA core motif of the C element (GAATTGGAAAGC sequence) and in the first half of the D element (GAAAGGAGAAACT), respectively (Fig. 3). IRF-3-or IRF-7-mediated transcription and the synergism between these factors were dramatically affected in the case of IFN-A11 (Fig. 4A). Interestingly, when IKK⑀ or TBK1 were expressed, IKK⑀ exerted a strong stimulatory activity on IFN-A4 promoter transcription (57-fold), compared with the 3-fold enhancement displayed by TBK1 (Fig. 4B); kinase-dead versions of IKK⑀ or TBK1 did not stimulate transcription. When each kinase was expressed together with IRF-3, IFN-A4 gene promoter transcription was stimulated more efficiently ( Fig. 4C). No stimulatory effect on IFN-A11 promoter transcription was observed when IKK⑀ or TBK1 was expressed alone or in the presence of IRF-3, although a 4 -5-fold increase was obtained in the latter case (Fig.  4, B and C), confirming that the IFN-A11 gene promoter was unresponsiveness to IRF-3. Moreover, only the IFN-A4 promoter transcription was induced in HEK293T cells infected by Sendai virus or in L929 cells infected by NDV (Fig. 4D). These results indicated that the differential virus-induced transcription of IFN-A4 and IFN-A11 promoters correlated with their differential response to IRF-3 stimulated by IKK⑀ or TBK1. Interestingly, differential activation was also observed when RIG-I, an intracellular receptor for viral RNA or the mitochondrial antiviral signaling protein MAVS that functions downstream of RIG-I (28,30) was expressed in 293T cells (Fig. 4D). The IFN-A11 gene promoter did not respond to IRF-3 activation even when cells expressed a constitutively active N-terminal form of RIG-I (⌬RIG-I), whereas the IFN-A4 promoter was stimulated 15-fold in similar conditions.
Chromatin immunoprecipitation experiments were performed to assess whether the differential transactivation observed with IKK⑀ or TBK1 expression may be due to differential recruitment of IRF-3. Anti-IRF3 and anti-CBP antibodies were used to immunoprecipitate nuclear proteins cross-linked to chromatin in IRF-3-and kinase-expressing cells. Specific amplification of the IFN-A4 promoter was easily observed when IRF-3 was coexpressed with IKK⑀, compared with TBK1 (Fig. 5,  lanes 4 and 5). Weak amplification with IFN-A11 (lanes 9 and 10) further indicated that IKK⑀ expression resulted in differential in vivo recruitment of IRF-3 and CBP on IFN-A4 and IFN-A11 gene promoters. Interestingly, reversal of the B or C element in the IFN-A4 promoter did not affect the IRF-3 recruitment but completely abolished the recruitment of CBP (lanes 11 and 13). These results, together with the data obtained with DNA-binding assays (presented in Fig. 2D), demonstrated that CBP recruitment to the IFN-A4 promoter requires the ordered assembly of IRF-3 dimers bound to three IRF elements on the proximal promoter region.
We have then examined whether the Ϫ78A/G and Ϫ57G/C substitutions that naturally occur in the C and D elements of IFN-A11 also affected IFN-A4 transcription by introducing the Ϫ78A/G and Ϫ57G/C substitutions into the IFN-A4 promoter (Fig. 6A). The . Sequence comparison of the ؊119 to ؊26 region of the IFN-A4 and IFN-A11 promoters. Sequences were numbered relative to the transcription start site. The virus-responsive element VRE-A4 was located in the proximal Ϫ119 to Ϫ26 region of the IFN-A4 promoter. In the IFN-A11 promoter, Ϫ78A/G and Ϫ57G/C substitutions, shown to be responsible for the differential expression of IFN-A4 and IFN-A11 gene promoters following virus infection, disrupt the C and D elements, respectively.
Ϫ57G/C substitution that disrupted the D element of IFN-A4 did not affect the response to IRF-3(5D) or virus inducibility in Sendai virusinfected HEK293 cells or NDV-infected L929 cells (Fig. 6, B and D). In contrast, disruption of the C element dramatically reduced the IRF-3(5D) responsiveness of the promoter (from 186-to 19-fold), impaired its stimulation by IKK⑀ or TBK1, and reduced its virus inducibility by more than 2-fold (Fig. 6, B-D). The M78 construct exhibited similar IRF-7-mediated transcription levels to that observed with M57. Interestingly, the mutant promoter carrying both the Ϫ57G/C and Ϫ78A/G substitutions was not responsive to IRF-3(5D), IRF-7A, IKK⑀, TBK1, or virus infection (Fig. 6, A-D). Moreover, in vivo IRF-3 recruitment to the IFN-A4 promoter carrying both substitutions was totally abrogated in virus-infected cells, as shown by chromatin immunoprecipitation assays using wild-type and mutated IFN-A4 promoters (Fig. 6E, lanes 3  and 6). The weak signal obtained in uninfected cells (lane 1) suggested a constitutive recruitment of IRF-3 to the IFN-A4 promoter, consistent with the detection of a small fraction of nuclear IRF-3 in unstimulated cells (15,56); no amplification of the IFN-A4 promoter was observed in the absence of immunoprecipitating antibodies (lane 2). Furthermore, the Ϫ78G/C substitution strongly reduced IRF-3 recruitment, compared with Ϫ57A/G substitution (lanes 4 and 5). These results illustrated the critical role of the C element in recruitment of IRF-3 to the IFN-A4 gene promoter and indicated that efficient IRF-3-mediated cooperative transcription was due to the number, precise arrangement, and specificity of the IRF elements.
IKK⑀ and TBK1 Differentially Affect the IRF-3-mediated Transcription of IFN-A4 Gene Promoter-The robust response of the IFN-A4 gene promoter to IKK⑀ expression, compared with a weak stimulation by TBK1 (Fig. 4, B and C, and Fig. 6C) was an intriguing and unexpected observation. Actually, transfection of 12.5 ng of the IKK⑀-expressing plasmid in HEK 293T cells was sufficient to obtain threshold levels of transcriptional activity (up to 5-6 relative units), whereas 25 times higher amounts of pcDNA3-TBK1 were required to observe a similar effect ( Fig. 7A and Table 2). The 20-fold induction of the IFN-A4 gene promoter observed by expressing 75 ng of pcDNA3-IKK⑀ was only slightly augmented by further increasing the amount of kinase up to 300 ng. In this case, transcriptional activity was 10-fold higher than observed in the presence of 300 ng of TBK1-expressing plasmid. A 4 -8-fold difference between IKK⑀ and TBK1 was also observed when each kinase was expressed in the presence of exogenous IRF3. In this case, threshold promoter activity required 3 ng of pcDNA3-IKK⑀ and 10 times higher amounts of pcDNA3-TBK1; a further increase in the amount of exogenous kinase resulted in RT values 3-6-fold and 5-8-fold higher with IKK⑀ and TBK1, respectively (Fig. 7B, Table 2). The differential effect of IKK⑀ or TBK1 on IRF-3-mediated transcription was less pronounced (2-3-fold) when each kinase was expressed in the presence of high concentrations of exogenously added substrate IRF-3 ( Fig. 7C and Table  2). Immunoblot analysis with anti-FLAG antibodies monitoring the expression of IKK⑀ and TBK1 indicated surprisingly that, although IKK⑀ displayed higher stimulatory activity on IRF-3-mediated transcription, this kinase was barely detected in cytoplasmic extracts compared with TBK1 (Fig. 7D). Furthermore, in nuclear and cytoplasmic cell extracts, IKK⑀ was only detected in the nuclear extracts of 293T cells when expressed alone, whereas co-expression of exogenous IRF-3 resulted in IKK⑀ cytoplasmic localization (Fig. 8, lanes 2 and 7 and lanes 3 and 8). In contain the Ϫ120 to ϩ19 fragment of the IFN-A4 or IFN-A11 promoter, respectively, upstream of the CAT reporter gene. The relative transcription levels were determined in the presence of 1 g of plasmid expressing IRF-3(5D), IRF-7A, or both (A) and when 75 ng of pcDNA3-IKK⑀, TBK1, or the kinase-inactive mutant IKK⑀-KD or TBK1-KD was transfected either alone (B) or together with 25 ng of plasmid expressing IRF-3wt (C) in human 293T cells. Virus inducibility of both promoters was determined in HEK293T cells infected with Sendai virus or in L929 cells infected with NDV (D). Relative transcription levels of plasmids pIF4T-L and pIF11T-L containing the Ϫ470 to ϩ19 and Ϫ457 to ϩ19 fragments of the IFN-A4 or IFN-A11 promoter, respectively, upstream of the luciferase reporter gene were determined in the presence of 0.5 g of plasmid expressing MAVS, RIG-I, or ⌬RIG-I (E).  1-5), pIF11T (lanes 6 -10), pIF4J[Brev] (lanes 11 and 12), and pIF4J[Crev] (lanes 13 and 14) together with pcDNA3- IKK⑀ (lanes 2, 4, 7, 9, 11, and 13) or pcDNA3-TBK1 (lanes 3, 5, 8, 10, 12, and 14) in the absence or presence of IRF-3wt (lanes 4, 5, and 9 -14). Cross-linked chromatin extracted from these cells was sonicated, purified, and immunoprecipitated with polyclonal IRF-3 or CBP antibodies. Decross-linked chromatin was subjected to a PCR using primers described under "Experimental Procedures." The PCR products were analyzed on a 1% agarose gel. FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 contrast, TBK1 was detected in the cytoplasmic extracts of cells expressing exogenous IRF-3 but not in nuclear extracts (Fig. 8, lanes 4  and 5 and lanes 9 and 10). However, expression of IKK⑀ or TBK-1 together with IRF-3 resulted in IRF-3 phosphorylation, detected by the phosphoserine 396 phosphospecific antibodies (Fig. 8, lanes 3, 5, 8, and  10) and correlated with IFN-A4 promoter activity. Detection of poly-  2-6) and fixed by formaldehyde 5 h later. DNA was sonicated following extraction and immunoprecipitated with polyclonal IRF-3 antibodies (lanes 1 and 3-6) or treated with protein A-Sepharose in the absence of antibodies (lane 2). Purified DNA was then subjected to a PCR using specific primers for pIF4J, M57, M78, or M57/78. The PCR products were analyzed on a 1% agarose gel.  Table 2. Expression levels of IKK⑀ and TBK1 proteins was monitored in cytoplasmic extracts from 293T cells transfected with increasing amounts of kinase-expressing plasmid by Western blot experiments performed with monoclonal anti-FLAG antibodies (D).

Differential IFN-A Gene Regulation by IKK⑀ and TBK1
(ADP-ribose) polymerase, Bcl-xL, and ␣-tubulin used as markers for nuclear, mitochondrial, and cytoplasmic proteins, respectively, showed that the subcellular distribution of IKK⑀ and TBK1 was not due to contamination of nuclear extracts. Taken together, these results indicate that IRF-3-mediated transcription is activated at low concentrations of IKK⑀ due to the predominant nuclear localization of this kinase, whereas higher levels of cytoplasmic TBK1 are required.
IKK⑀ and TBK1 Stimulate the IRF-7-mediated Transcription of IFN-A11 Gene Promoter-Surprisingly, IKK⑀ or TBK1 strongly stimulated IRF-7-mediated transcription of IFN-A11, 133-and 36-fold, respectively, compared with 408-and 56-fold stimulation of IFN-A4 inducibility (Fig. 9A), indicating that the differential response of these two promoters to IRF-7 can be partially compensated upon stimulation with IKK⑀ or TBK1. Interestingly, the compensatory effect of IKK⑀ or TBK1 on IRF-7-mediated differential transactivation of IFN-A4 and IFN-A11 promoters was not observed when the MyD88-dependent pathway stimulated the IRF-7 transcriptional activity (Fig. 9B). When IRF-7A was expressed together with increasing amounts of each kinase, IFN-A4 promoter transcription was up-regulated more strongly by IKK⑀ than TBK1 (Fig. 9C). Since these stimulatory effects could be overestimated by a potential cooperative effect of endogenous IRF-3 activated by IKK⑀ or TBK1, the IFN-A4 promoter mutants pIF4J[Brev] and pIF4J [Crev], which are responsive to IRF-7 but not to IRF-3, were used to determine the effect of TBK1 or IKK⑀ on IRF-7A-mediated activation. As illustrated in Fig. 9C, IRF-7-mediated transcription of these constructs was stimulated more efficiently by IKK⑀ than by TBK1. Using a newly developed IRF-7 phosphoserine 477/479-phosphospecific antibody, it was possible to identify C-terminal phosphorylation of IRF-7 by TBK1 or IKK⑀ (Fig. 9D, lanes 3 and 5), indicating that IRF-7 activation correlated with transcriptional activity of the IFN-A4 promoter. These data confirmed that the differential effect of IKK⑀ and TBK1 on IRF-7-mediated transcription was due to their distinct subcellular localizations (Fig. 9D,  lanes 8 and 10).

DISCUSSION
The present study demonstrates that three IRF elements (designated as B, C, and D) constitute binding sites for IRF-3 and IRF-7 and define the virus-responsive element (VRE-A4) of the murine IFN-A4 gene promoter. Transactivation of the IFN-A4 gene promoter by IRF-3 or IRF-7 requires these distinct binding sites that contribute to different extents to cooperative transcription. The first IRF element corresponds to the Ϫ98 to Ϫ87 GAAAGTGAAAAG sequence (B element), where the GT residues determine the specificity for IRF-7 and IRF-3. The Ϫ86 to Ϫ75 GAATTGGAAAGC sequence (C element) corresponds to the second IRF element and is a selective target for IRF-3, whereas the Ϫ57 to Ϫ45 GAAAGGAGAAACT sequence (D element) is weakly recognized both by IRF-3 and IRF-7. Our data suggest that when both factors are simultaneously activated during viral infection, synergistic activation of the IFN-A4 gene promoter is achieved by preferential recognition of B and C elements by IRF-7 and IRF-3 homodimers, respectively, whereas the D site may be occupied by either IRF-7 or IRF-3. We also demonstrate that the in vivo recruitment of IRF-3 and CBP to the IFN-A4 gene promoter following virus-or kinase-dependent activation is dictated by the precise arrangement of the IRF elements in the promoter. These results are in agreement with previous in vitro binding studies demonstrating that disruption of the C or D elements decreased IRF-3 affinity by 6-and 3-fold, respectively, whereas disruption of both elements almost completely abolished the IRF-3 binding (46). DNA distortions in the GAAA core sequences of cognate binding sites have been proposed as a determinant for cooperativity in IRF binding (57,58). Accordingly, transcriptional cooperativity might be favored by IRF-3 binding to the C element by similar distortions, thus preparing the IFN-A4 template for the cooperative IRF-3 interaction with the B and D modules. Virus-induced transcription of IFN-A gene promoters also depends on the ability of IRF-3 or IRF-7 homodimers to interact with CBP/p300 coactivators (18,59,60). Our results showed that reversal of the B or C module abolishes IRF-3-mediated transcription of IFN-A4 promoter in correlation with a dramatic decrease of in vivo CBP recruitment to the promoter, indicating that ordered assembly is required for IRF-3-mediated cooperative transcription. Furthermore, the absence of a single IRF element significantly disturbs the interaction of IRF-3 with the promoter and destabilizes the assembly required for    1 and 6), pcDNA3-FLAGIKK⑀ (lanes 2 and 7), or pcDNA3-FLAGTBK1 (lanes 4 and 9) alone or together with pcDNA3-IRF-7A (lanes 3, 5, 8, and 10). The phosphorylated form of IRF-7 is indicated by an asterisk.
for IFN-A gene promoters, although IRF-3 and IRF-7 have been suggested to bind to human IFN-A1 promoter as heterodimers (62). In the case of the IFN-A4 promoter, virus-induced transcription is dependent solely on the capacity of IRF-3 or IRF-7 homodimers to bind on the IRF elements of the VRE-A4. The factors facilitating this enhanceosome or "irfosome" formation are unknown, although IRF-3 itself may play this role. Our data suggest that IRF-3 homodimers bound to the B, C, and D elements adopt precisely defined orientations with respect to each other on the DNA helix. Accordingly, nucleotide substitutions disrupting specifically the C element (the M78 construct) or the C and D elements (IFN-A11 gene promoter and the M78/57 construct) led to a proportional decrease in IRF-3 and CBP recruitment, affected synergistic transcription, and altered cooperativity between IRF-7 and IRF-3. This particular organization of the IFN-A4 promoter accounts also for its differential responsiveness to IKK⑀ or TBK1 compared with IFN-A11. IRF-3-mediated induction of IFN-A4 was activated at low concentrations of IKK⑀ but required high levels of TBK1 expression. The distinct stimulatory effect of IKK⑀ and TBK1 on transcription may be related to subcellular localization; IKK⑀ was localized predominantly to the nuclear compartment, whereas TBK1 was only detected in the cytoplasm. Since IRF-3 constitutively shuttles in and out of the nucleus (41,56), IKK⑀ may promote phosphorylation of nuclear pools of this factor, leading to rapid and efficient IRF-3-mediated transcription. In contrast, TBK1 phosphorylates only the cytoplasmic pool of IRF-3, thereby stimulating a relatively delayed transcription that initiates following IRF-3 translocation to the nucleus. The IFN-A11 gene promoter, on the other hand, lacking the irfosome required for IRF-3-mediated transcription, did not respond to IKK⑀ or TBK1 stimulation. These data predict that differential IRF-3-mediated transcription of IFN-A promoters might persist, even when virus infection triggers multiple IRF-3 activation pathways by both IKK⑀ and TBK1. This "gradient" effect may influence the differential expression of IFN-A genes in different cell types, since the promoters of other members of the IFN-A multigenic family differ from the IFN-A4 promoter by nucleotide substitutions affecting the number, IRF specificity, and arrangement of IRF elements. For example, the C module of IFN-A2 is altered or disrupted by Ϫ80G/A and Ϫ78A/G substitutions and Ϫ85G/C and Ϫ82T/G substitutions in IFN-A1, IFN-A5, and IFN-A6 genes; the low virus inducibility of these genes compared with IFN-A4 (3, 63-65) may therefore be due to an impairment of their IRF-3-mediated transcription. Cooperative transcription of IFN-A4 promoter by IRF-7 homodimers bound to B and D elements is distinct from its IRF-3-dependent transcription. Reversal of the B or C element in the IFN-A4 gene promoter did not affect IRF-7-dependent transactivation in contrast to IRF-3-mediated cooperative transcription. Interestingly, a promoter mutant carrying a reversed B or C element retained its ability to respond to IKK⑀ or TBK1 overexpression in the presence of IRF-7 but lost its ability to respond to IRF-3 stimulated by IKK⑀ or TBK1 and, most importantly, to virus induction. Our observation is also based upon the fact that IKK⑀ and TBK1 stimulate the IRF-7-mediated transcription of the IFN-A11 gene promoter, which is unresponsive to IRF-3. These data indicated that cooperative transcription by multiple IRF-7 homodimers is not necessarily related to IRF-3-mediated virus inducibility of IFN-A4. Robust stimulation of IFN-A11 by kinase-activated IRF-7 also suggests that IFN-A promoters lacking some IRF elements may nevertheless respond to virus infection in cells expressing IRF-7. Interestingly, all IFN-␣ subtypes were expressed at the same level in influenza virusinfected monocyte-derived or plasmacytoid dendritic cells (66). The constitutive expression of IRF-7 and IKK⑀ in lymphoid cells or mature dendritic cells may thus reduce the differential expression levels of IFN-A gene family members. Since the expressions of IRF-7 and IKK⑀ are both virus-inducible (11,19,41), the differential expression of IFN-A may be preeminent in the early phase of virus infection but may be balanced in fibroblasts or epithelial cells during the delayed phases. The "gradient" effect of IRF-3-mediated transcription that favors differential IFN-A gene expression might constitute a regulatory mechanism allowing the return to homeostasis in different cells, in comparison with particular subsets of dendritic cells expressing both IRF-7 and IKK⑀ and acting as high IFN-A producers.