To gain more insight into how the innate and induced antiviral activities are controlled on gene expression levels, we studied the molecular mechanisms that control the expression of the TLR3 gene that is induced by viral infection. Both the basal level expression and induction of TLR3 require the chromatin remodeling activity of the SWI/SNF-like BAF complexes, since the knockdown of an essential subunit of the complex, BAF47, severely inhibited its expression [12-15]. To confirm that the TLR3 promoter is a direct target site of the BAF complex, the promoter region [16] was cloned into pGL3, which does not form regular chromatin structure, or pREP4 reporter vector, which replicates and forms regular chromatin structure when transiently transfected into cells [17]. Co-transfection with BRG1 expression construct into SW-13 cells, a BRG1-deficient cell line, significantly activated the promoter with pREP4 vector but not with the pGL3 vector (Figure 1A), suggesting that the BAF complex regulates the TLR3 promoter in a chromatin-dependent manner. Chromatin immunoprecipitation (ChIP) assays revealed that anti-BRG1 antibody pulled down the endogenous TLR3 promoter without IFN-α treatment (Figure 1B), indicating the BAF complex is constitutively associated with the promoter. IFN-α stimulation enhanced BRG1 binding to the promoter as shown by quantitative-PCR (q-PCR) analysis (Figure 1B). These results indicate that the TLR3 promoter is a direct target site of the BAF complex.
To identify the DNA elements that mediate the BAF complex activity, 5′-deletion analysis of the TLR3 promoter was performed. The 509 bp DNA fragment of promoter was responsive to IFN-α stimulation and to the BRG1 expression (Figure 1C). Both together can activated the promoter further. Deletion to −251 reduced the activity of the promoter. However, the responsiveness to IFN-α and BRG1 remained, even though an apparent ISRE was deleted. Deletion to −11, which removed an IRF-E sequence, completely abolished the activity of the promoter. These data suggest that the Sp1 binding site contributes to and the IRF-E plays an essential role in mediating the activity of the BAF complex. Consistent with this, the point mutation of the Sp1 binding site decreased the promoter activity, and point mutation of the IRF-E site completely abolished the activity of the promoter (Figure 1D). These results are consistent with the observation that Sp1 stabilizes the BAF complex binding to target promoters [18].
IRF-E sequence is the binding site for IRFs [19]. ChIP assays showed that both the IRF1 and IRF2 antibodies enriched the TLR3 promoter sequence relative to the control in non-stimulated cells (Figure 2A and B). Interestingly, IFN-α treatment strongly enhanced IRF1 binding but inhibited IRF2 binding (Figure 2A and B). No binding was detected for other IRFs (data not shown). These data suggest that IRF1 and IRF2 may regulate both the basal and induced expression of the TLR3 gene. To confirm this, both genes were knocked down by small interference RNA (siRNA) constructs (Figure 2C). When IRF1 was knocked down, the induction of the TLR3 gene (Figure 2D) and its promoter activity (Figure 2E) were significantly reduced. The basal level expression was only slightly decreased by siIRF1. Surprisingly, knocking down IRF2 resulted in a dramatic reduction of both the basal and induced levels of the TLR3 gene expression and also its promoter activity (Figure 2D and E). These data reveal that while IRF1 is critical for the induction of the TLR3 gene, IRF2 binding to the TLR3 promoter plays an essential role in both the basal and induced expression of the TLR3 gene in response to interferon stimulation.
Why do the cells take the trouble to make two proteins binding to the same site? We hypothesize that the TLR3 gene needs to be expressed at low level in the absence of viral infection and needs to be rapidly induced to high level in the presence of viral infection. Furthermore, the chromatin structure at the promoter should be prepared for rapid activation in response to viral infection. Based on the in vivo binding and knock-down results (Figure 2), we hypothesize that IRF2 may be the factor to prepare the chromatin structure and direct low level expression, and IRF1 may be the factor to direct highly induced expression.
IRF2 has been considered as a transcriptional repressor or activator [5]. To test our hypothesis above, first, we decided to distinguish if IRF2 is a repressor or weak activator. When fused to the GAL4 DNA binding domain, IRF1 efficiently activated a promoter containing five GAL4 DNA binding sites, whereas IRF2 showed only low activity (Figure 3A). These results indicate that IRF1 is a strong activator and IRF2 is a very weak activator. Second, we addressed how the IRF1 and IRF2 binding to the promoter were controlled. Western blotting indicated that IRF1 protein was highly induced by IFN-α treatment, with the highest protein level detected after 2 hours of stimulation (Figure 3B). The IRF1 level decreased significantly after 8 hours and went back to basal level at 24 hours post stimulation. In contrast, the IRF2 protein level was relatively stable (Figure 3B). Remarkably, IRF1 binding to the TLR3 promoter paralleled with its protein level in the cells. The highest binding level was detected after 2 hours of IFN-α treatment and the binding level decreased to almost basal level after 8 hours of stimulation (Figure 3C). These data argue that the binding levels of IRF1 and IRF2 at the TLR3 promoter are controlled by their relative protein levels in the cells. If this true, artificially increasing IRF1 level in the cells would be able to compete with IRF2 and activate the TLR3 promoter. Indeed, over-expression of IRF1 efficiently activated the TLR3 promoter (Figure 3D), consistent with its ability to compete with IRF2 to bind to the promoter and its activity as a strong activator.
Activation of the TLR3 gene by viral infection or polyI/polyC leads to the induction of the IFN-β gene, which is mediated by NF-κB and TLR3 [10]. Our data that IRF2 is required for the basal and induced levels of TLR3 expression suggest that knocking down IRF2 may cripple the cellular response to polyI/polyC. To confirm this, HeLa cells, which were either transfected with a control vector or siIRF2, were treated with polyI/polyC. The induced expression of the IFN-β gene was abolished by knocking down IRF2 (Figure 3E), suggesting that IRF2 is a key molecule that controls the cellular antiviral activities.
To investigate how IRF2 regulates the induction of TLR3, we examined the accessibility of the TLR3 promoter to restriction enzyme Hind III, which has a recognition site −39 bp upstream of its TSS. We found that interferon-α treatment increased the accessibility of the Hind III site, while no changes were detected at the Nhe I site which is located at +52, downstream of the TSS (Figure 4A, lanes 1 and 2). knocking down of BAF47, a key subunit of the BAF chromatin remodeling complex, resulted in decreased accessibility of the Hind III site as compared to the control Nhe I site (Figure 4A, lanes 3 and 4), consistent with our previous observation that the BAF complex is involved in the induction of IFN-inducible genes [12,20]. Interestingly, knocking down of IRF2 also compromised the accessibility of the Hind III site (Figure 4A, lanes 5 and 6). These results indicate a critical role of IRF2 in maintaining an open chromatin structure at the TLR3 promoter in both non-stimulated basal and stimulated states.
Next we investigated whether IRF2 binding impacts histone modifications. Histone H3 is subject to extensive modifications including acetylation and methylation, which is correlation with transcriptional activation or repression [21]. To test whether IRF2 facilitates histone modifications at the TLR3 promoter, we measured H3K4me3 and H3K9/K14ac using ChIP assays. Our data revealed that these modifications at the promoter region were dramatically decreased by knocking down IRF2, while no significant changes were detected at the exon II region of the TLR3 gene (Figure 4B, C). These data confirmed our hypothesis that IRF2 is essential for keeping an open chromatin structure at the TLR3 promoter.
The data that knocking-down IRF2 reduced the chromatin accessibility of the TLR3 promoter to a similar extent as knocking-down BAF47 suggest that IRF2 may be responsible for the recruitment of the BAF complex to the promoter in non-stimulated cells. Thus, we performed co-immunoprecipitation assays to test the direct interaction between IRF2 and the BAF complex. Our data revealed that BRG1 was co-immunoprecipitated from the nuclear extracts using both anti-IRF1 and anti-IRF2 antibodies (Figure 5A). Addition of ethidium bromide did not inhibit the co-immunoprecipitation, indicating that the interaction was not mediated by DNA. These data suggest that both IRF1 and IRF2 are capable of recruiting the BAF complex to the TLR3 promoter to maintain an open chromatin structure through protein-protein interactions.
IRF1/2 are known to recognize ISRE [22], which mediates the induction of IFN-α target genes by the ISGF3 complex. Since the BAF complex regulates most of the IFN-α target genes [12], IRF2 may serve as a general recruiter of the BAF complex to the IFN-α target genes. To test this idea, we analyzed IRF1 and IRF2 binding at several IFN-α inducible genes by ChIP assays. As shown in Figure 5B and C, both IRF1 and IRF2 bound to the promoters of TLR3, IFITM1, IFITM3, and STAT2. Interestingly, IRF2 disappeared from all of these promoters after IFN-α treatment. We assume that IRF2 was replaced by the ISGF3 complex at the IFITM1, IFITM3, and STAT2 promoters as suggested by the ChIP data summarized in Figure 5D. Similarly to the induction of TLR3, knocking down of IRF2 also significantly inhibited the expression of IFITM1, IFITM3, and STAT2 (Figure 5E). Therefore, these data define a general role of IRF2 to potentiate the induction of the IFN-α target genes in response to viral infection.
Recent studies suggested that epigenetic mechanisms play roles in controlling the function of IRF1 and IRF2 [23]. In particular, histone acetyltransferases can directly enhance their transcriptional activity by modifying nucleosome or IRFs themselves [24,25]. Our previous studies suggested that the ATP-remodeling BAF complexes are required for maintaining the basal and induced expression of IFN-inducible genes [12]. However, it is not clear what are the roles of the constitutively expressed IRFs in this process. Although IRF2 has been considered a transcriptional repressor and an antagonist of IRF1 [5,26], it has been difficult to explain certain phenotypes in mice associated with disruption of IRF1 and IRF2 [27,28]. In this study, our data argue against the notion that IRF2 is a repressor and thus antagonizes the activity of IRF1. Instead, we demonstrate that IRF1 is a strong transcription activator and IRF2 is a weak activator, which act at different stages of TLR3 activation. IRF2 serves two roles at the TLR3 promoter: 1) to recruits the BAF complex to prepare an open chromatin structure for rapid activation upon viral infection; 2) to maintain a basal level expression of the gene for the innate antiviral activity. IRF1 serves to rapidly activate the promoter by replacing IRF2 upon induction. However, if the chromatin structure at the promoter was not prepared, induction completely failed. The functional difference between IRF1 and IRF2 may arise from two mechanisms. One is the control of their expression: IRF2 is constitutively expressed while IRF1 is inducible by viral infection or interferon treatment, which results in different binding patterns during different stages of cellular antiviral activity. The other may reflect the differences of their activation domains: although IRF1’s activation domain can potently activate transcription, our data argue that only the activation domain of IRF2 can act as a pioneering factor to prepare chromatin for rapid transcription induction in response to stimulation. Therefore, our data in this report demonstrate that different members in a transcription factor family are made to meet different requirements for elaborate transcriptional regulation. The vertebrate organisms gain an extra level of transcriptional control by duplication of gene families.