Essential contribution of IRF3 to intestinal homeostasis and microbiota-mediated Tslp gene induction

Abstract

The large intestinal epithelial cells and immune cells are exposed to a variety of molecules derived from commensal microbiota that can activate innate receptors, such as Toll-like receptors (TLRs) and retinoic acid-inducible gene-I-like receptors (RLRs). Although the activation of these receptors is known to be critical for homeostasis of the large intestine, the underlying gene regulatory mechanisms are not well understood. Here, we show that IFN regulatory factor (IRF)3 is critical for the suppression of dextran sulfate sodium-induced colitis. IRF3-deficient mice exhibited lethal defects in the inflammatory and recovery phases of the colitis, accompanied by marked defects in the gene induction for thymic stromal lymphopoietin (TSLP), a cytokine known to be essential for protection of the large intestine. We further provide evidence that DNA and RNA of the large intestinal contents are critical for Tslp gene induction via IRF3 activation by cytosolic nucleic acid receptors. We also demonstrate that IRF3 indeed activates the gene promoter of Tslp via IRF-binding sequences. This newly identified intestinal gene regulatory mechanism, wherein IRF3 activated by microbiota-derived nucleic acids plays a critical role in intestinal homeostasis, may have clinical implication in colonic inflammatory disorders.

The intestinal microbiota contains a variety of commensal bacteria that can greatly affect mucosal immune response in the large intestine (1–4). Commensal bacteria have the potential to activate large intestinal epithelial cells and immune cells; this constitutive activation of these cells is critical to maintain intestinal homeostasis (1–4). Indeed, germ-free mice or mice treated with antibiotics are sensitive to develop colonic inflammation (5, 6). Homeostasis is mediated by the operation of a complex network of signaling pathways and transcription factors that are triggered by molecules derived from commensal bacteria via signal-transducing innate receptors, such as Toll-like receptors (TLRs), retinoic acid-inducible gene-I–like receptors (RLRs), Nucleotide-binding oligomerization domain-containing protein (NOD)-like receptors, and C-type lectin receptors (1–4, 7–9). Consistent with their involvement in the gut, the integrity of TLR and RLR signaling pathways were shown to be critical in the control and development of colitis in animal models (2, 3, 7, 8). Mice lacking either of these receptors or their associated adaptor molecules manifest severe sensitivities to dextran sulfate sodium (DSS)-induced colitis (3, 7, 8). However, the underlying mechanisms how these innate signaling pathways suppress the colitis still remain to be clarified.

In general, the evocation of innate immune responses by innate receptors involves the activation of several families of transcription factors that ultimately induce expression of their target genes. Members of the IFN regulatory factor (IRF) family of transcription factors function distinctly from one another depending on the nature of the signal emanating from a given innate receptor; these IRFs often cooperate with other transcription factors, such as NF-κB (10). Several IRFs are known to be activated by the TLR and RLR pathways. Briefly, IRF3 is robustly activated by RLR signaling and plays a critical role in the activation of type I IFN genes in cooperation with IRF7, whereas IRF5 is activated by TLR signaling and is involved in the gene activation of proinflammatory cytokines (10, 11). To date, however, little is known regarding the involvement of IRFs in the regulation of intestinal homeostasis and associated diseases.

In this study, we first focused on IRF3 and IRF5, which mainly function downstream of the RLR signaling and TLR signaling pathways, respectively, in the context of DSS-induced colitis, a widely used, chemically induced model of inflammatory bowel disease (12). We found the development of severe and lethal DSS-induced colitis in IRF3-deficent, but not IRF5-deficient, mice. The IRF3-deficient mice also showed impaired recovery from DSS-induced colitis. This was accompanied by a lack of gene induction for thymic stromal lymphopoietin (TSLP) and IL-33, cytokines crucial for the recovery stage of colitis (13, 14). Interestingly, we also observed that the basal expression level of colonic Tslp and Il33 genes in IRF3-deficient mice was markedly lower than that of WT mice before DSS treatment. Antibiotic treatment of WT mice also resulted in dramatic reduction of basal gene expression levels. We further demonstrate that nucleic acids derived from large intestinal contents are responsible for the induction of these genes via IRF3 activation by cytosolic nucleic acid sensing receptors. We discuss the link between IRF3 and enteral nucleic acids for the expression of colonic Tslp and Il33 genes in the development of colonic inflammatory diseases.

Results

Exacerbation of DSS-Induced Colitis in IRF3-Deficient Mice.

Given that the TLR and RLR pathways contribute to the suppression of colitis (2, 3, 7, 8), we asked whether IRF3 and IRF5, which are known to be activated by either or both of these pathways, play a role in DSS-induced acute colitis that is triggered independently of adaptive immune cells (15, 16). First, WT and IRF3- and IRF5-deficient mice treated with 2% (wt/wt) DSS were evaluated for survival and colon lengths (Fig. 1 A–C). Surprisingly, all IRF3-deficient mice died within 11 d, whereas WT and IRF5-deficient mice survived. Colons from IRF3-deficient mice, measured 8 d after the start of DSS treatment, were markedly shorter than those of WT or IRF5-deficient mice, a phenotype understood to reflect excessive tissue damage (Fig. 1 B and C). Consistent with these observations, histological analysis showed severe cell invasion and destruction of tissue structure in the colon from IRF3-deficient mice (Fig. 1D). These results indicate a protective role of IRF3, but not IRF5, in DSS-induced colitis.

Fig. 1.

DSS-induced colitis in IRF3-deficient mice. A total of 2% DSS is administered for 6 d and replaced with water on day 7. (A) Survival of WT (n = 6), IRF3-deficient (n = 5), or IRF5-deficient mice (n = 5) monitored every 24 h. (B) Colons from WT, IRF3-deficient, and IRF5-deficient mice on day 8. (C) Quantification of the colon length described in B. *P < 0.05 compared with colon from WT mice. (D) Histological analysis of H&E-stained colon sections from WT and IRF3-deficient mice treated with 2% DSS for 8 d. (Original magnification: 200×.) (E) Body weight of WT (n = 6) or IRF3-deficient mice (n = 5) monitored every 24 h. **P < 0.01 and *P < 0.05 compared with WT mice. (F) qRT-PCR analysis of Tslp and Il33 mRNA in colons from WT or IRF3-deficient mice (n = 5) on day 8. **P < 0.01 and *P < 0.05, compared with WT mice. Statistical data are presented as mean ± SD.

To further examine the role of IRF3 in the protection of DSS-induced colitis, we next examined changes in body weight during DSS treatment (Fig. 1E). IRF3-deficient mice showed rapid weight loss compared with WT mice. The weight continued to decrease even after DSS was replaced with water (Fig. 1E). These results indicate that IRF3-deficient mice have defects in recovery from colitis. Although type I IFNs are known to be important for the suppression of colonic inflammation, type I IFN does not contribute to the recovery from DSS-induced colitis; there is no lethality in mice defective in IFN signaling at the same dose (2%) of DSS (17). Hence, these observations for IRF3-deficient mice are not likely the result of a defect in IRF3-driven type I IFN responses. We also examined the mRNA expression of proinflammatory cytokines but no abnormalities were found in their expression from total colon of IRF3-deficient mice compared with WT colon (Fig. S1).

Gene Expression for Protective Cytokines in IRF3-Deficient Mice.

The observations presented here prompted us to search for target genes of IRF3, whose expression would be required for the recovery from colitis. We focused on the genes for TSLP and IL-33, both of which are known to have a critical, T-cell–independent role in the recovery from DSS-induced colitis (13, 14). It has also been reported that Il33 mRNA induction by TLR or RLR signaling is impaired in IRF3-deficient macrophages in vitro (18).

We first examined the expression of Tslp and Il33 mRNA in the whole colon of the WT and IRF3-deficient mice by quantitative (q) RT-PCR. As shown in Fig. 1F, basal expression of these mRNAs, observed even before DSS treatment, was further augmented after DSS treatment in the colon of WT mice (Fig. S2). However, expression levels were dramatically reduced in the colon of IRF3-deficient mice before and after DSS treatment (Fig. 1F). These results, therefore, suggest that Tslp and Il33 genes are both regulated by IRF3 in the colon and that IRF3 is activated even before DSS treatment. Expectedly, the expression of the mRNA for these genes was normal in the colon of IRF5-deficient mice (Fig. S3). Of further note, Tslp and Il33 mRNA expression remained unaffected in the colon of mice lacking the IFNAR1 chain of the type I IFN receptor, indicating that IRF3 likely directly regulates these genes (Fig. S4).

To investigate the cell type in which IRF3 regulates Tslp and Il33 gene expressions, we isolated cells from colonic epithelia and lamina propria from WT and IRF3-deficient mouse colons and then examined mRNA expression of these genes. Interestingly, expression of these mRNA observed in both colonic epithelial cells and lamina propria cells of WT mice is significantly diminished in cells from IRF3-deficient mice (Fig. S5), suggesting that IRF3-mediated gene expression occurs in both epithelial cells and immune cells.

IRF3-Dependent Tslp and Il33 Gene Induction by Microbiota.

Gut microbiota contain a variety of commensal bacteria that have the potential to activate gene expression through the release of innate receptor ligands (1–3, 8, 9). To further examine the mechanism(s) underlying IRF3-dependent inductions of Tslp and Il33 genes in the colon, we next investigated the role of commensal bacteria in this process. Mice treated with a mixture of antibiotics to eliminate most of the commensal bacteria (5, 19) were analyzed 7 wk later for colonic expression of Tslp and Il33 mRNA (Fig. 2A). Interestingly, antibiotic treatment resulted in a dramatic decrease of mRNA levels in the colon of WT mice, indicating that commensal bacteria contribute to their expression.

Fig. 2.

IRF3-dependent Tslp and Il33 gene induction by treatment with large intestinal contents. (A) qRT-PCR analysis of Tslp and Il33 mRNA in colons from WT mice treated (n = 5) or untreated (n = 3) with antibiotics. **P < 0.01 compared with untreated mice. Data are presented as mean ± SD (B) qRT-PCR analysis of Tslp and Il33 mRNA in MEFs (Left), peritoneal macrophage (PEC; Center), or bone marrow–derived dendritic cells (BMDC; Right) stimulated with large intestinal contents (enteral contents) or feces for 3 h. Data are presented as mean ± SD of triplicate determinations. (C) qRT-PCR analysis of Tslp and Il33 mRNA in WT or IRF3-deficient MEFs stimulated with feces for 3 h. **P < 0.01 compared with WT cells. Data are presented as mean ± SD of triplicate determinations.

These results prompted us to examine the potential of large intestinal contents for IRF3-mediated gene induction. When mouse embryonic fibroblasts (MEFs), peritoneal macrophages or bone marrow-derived dendritic cells from WT mice were treated with a suspension of WT mice-derived large intestinal contents or feces, a significant induction of mRNA was observed for Tslp and Il33 genes in WT cells (Fig. 2B), but not in cells derived from IRF3-deficient mice that were severely impaired for gene induction (Fig. 2C). Feces from IRF3-deficient mice also induced these gene expressions (Fig. S6). Thus, there is no alteration of the microbiota in terms of its composition that affects these gene expressions in the absence of IRF3. Further, these results indicate that large intestinal contents activate IRF3-dependent Tslp and Il33 gene inductions in the colon.

Cytokine Gene Induction by Nucleic Acid in Large Intestinal Contents.

We next asked which molecule(s) within the large intestinal contents mediates the Tslp and Il33 gene induction. To address this, we first investigated the signaling pathway that activates IRF3-dependent gene induction by colon contents. Given that IRF3 is activated downstream of the TLR-TRIF (TIR-domain-containing adapter-inducing interferon-β) and cytosolic pathways mediated by mitochondrial antiviral signaling protein (MAVS) (also referred to as IFN-β promoter stimulator-1/virus-induced signaling adaptor) or stimulator of IFN genes (STING) (10, 11), we knocked-down the expression of these molecules by siRNA in MEFs and then examined feces-mediated Tslp and Il33 gene induction. As shown in Fig. 3A, Tslp gene induction was markedly decreased in Mavs or Sting knock-down cells, whereas Il33 gene induction was decreased only in the Sting knock-down cells (Fig. 3A). Similar results were obtained when siRNAs targeting different sequence within Mavs and Sting mRNA were used (Fig. S7A).

Fig. 3.

IRF3-dependent Tslp and Il33 gene inductions by fecal nucleic acids. (A) qRT-PCR analysis of Tslp and Il33 mRNA in Trif, Mavs, or Sting knocked-down MEFs stimulated with feces for 3 h. **P < 0.01 and *P < 0.05 compared with control siRNA-treated MEFs. (B) qRT-PCR analysis of Tslp and Il33 mRNA in WT MEFs stimulated with nontreated or nuclease-treated feces for 3 h. **P < 0.01 and *P < 0.05 compared with MEFs treated with feces. (C) qRT-PCR analysis of Tslp and Il33 mRNA in WT MEFs stimulated with fecal nucleic acids (2.5, 5.0, or 10.0 μg/mL) for 6 h. (D) qRT-PCR analysis of Tslp and Il33 mRNA in WT and IRF3-deficient MEFs stimulated with B-DNA (10 μg/mL) or poly(I:C) (10 μg/mL) for indicated periods. **P < 0.01, compared with WT MEFs. (E) qRT-PCR analysis of Tslp and Il33 mRNAs in WT or IRF3-deficient MEFs transduced with retrovirus expressing WT or mutant IRF3. Cells were stimulated with B-DNA for 6 h. **P < 0.01 compared with WT MEFs. All data are presented as mean ± SD of triplicate determinations.

Because MAVS and STING are critical for cytosolic RNA- and DNA-mediated activation of innate immune responses, respectively, these results suggest there is a contribution of nucleic acids to feces-mediated gene induction. It is interesting to note that gene induction is observed without the entrapment of nucleic acids within liposomes, which is usually required for nucleic acids to cross into the cytoplasm for the activation of the cytosolic pathways. To further examine the importance of nucleic acid present in the feces, we treated WT mice-derived feces with various nucleases and then examined cytokine gene induction in MEFs. Interestingly, DNase I treatment of feces resulted in a notable loss (ca. 45%) of its activity to induce expression of the Il33 gene, but no effect on Tslp gene expression, whereas treatment of RNase A did not show any effect on the induction of either gene. Furthermore, the induction of these genes was not abolished by the treatment of feces suspension with Benzonase, which degrades all forms of DNA and RNA (Fig. 3B).

Given the involvement of the MAVS or STING pathways and that gene induction by feces occurs without nucleic acid entrapment (Fig. 3A), we surmised that the nucleic acids are in a complex with other molecules that helps render them nuclease-resistant. We therefore next extracted nucleic acids from WT mice-derived feces and then examined their potential to induce these genes. As shown in Fig. 3C, the Tslp and Il33 genes were induced by fecal nucleic acids, but only when they were mixed with lipofectamine, further supporting the notion that the immunostimulatory nucleic acids in feces are bound with other molecules to activate the cytosolic pathways. Taken together, these results suggest there is a critical role for nucleic acids in the feces-mediated induction of Tslp and Il33 genes, wherein cytosolic DNA- and RNA-induced signaling pathways may differentially contribute to the induction of these IRF3 target genes.

To address this issue further, MEFs were stimulated by synthetic poly(dA-dT)⋅poly(dT-dA) DNA (termed B-DNA) or double-stranded RNA (poly(I:C)). As shown in Fig. 3D, IRF3-dependent Tslp gene expression in MEFs was observed upon stimulation by either poly(I:C) or B-DNA, whereas Il33 gene induction was induced by B-DNA much more strongly than poly(I:C) at all doses examined (Fig. S7C). Although IRF3 usually exerts its function via its direct binding to IRF-binding sequences [interferon-stimulated response elements (ISREs)] of the target genes, IRF3 is also known to function as a cofactor to activate transcription of a set of NF-κB–dependent genes without directly binding to DNA (20). To clarify this issue further, we introduced WT IRF3, an activation-defective mutant or a DNA-binding-defective mutant of IRF3 into IRF3-deficient MEFs (also null for Bcl2L12) (21) by retrovirus-mediated gene transfer and then examined the cells for the induction of Tslp and Il33 by B-DNA stimulation (Fig. 3E). We found that WT IRF3, but none of these mutants, fully restored gene induction, indicating that IRF3 indeed directly activates these genes via its binding to ISREs.

Activation of the Tslp Promoter by IRF3.

The lethal deficiency observed for IRF3-deficient mice from DSS-induced colitis is more reminiscent of the phenotype of the TSLP-deficient mice than IL-33–deficient mice in that the former mice fail to recover from DSS-induced colitis, whereas the latter mice can recover from the colitis (13, 14). Although the Il33 gene is known to be regulated by IRF3 and cAMP response element-binding protein (CREB) upon activation of TLRs or RLRs (18), little has been reported regarding the transcriptional regulation of the Tslp gene. We therefore next focused on the IRF3-dependent gene induction of the Tslp gene.

We first analyzed the promoter region of Tslp gene and found six putative IRF-binding sequences (ISREs), termed ISRE-1 through ISRE-6, within 4 kb upstream of the transcriptional start site (TSS); two of which are located near NF-κB binding sites (Fig. 4A) (22). We then asked whether IRF3 has the potential to activate the promoter activity of Tslp gene in cooperating with NF-κB. First, a reporter gene construct pTSLP-4k-luc, which contains the 4-kb Tslp promoter fragment, was examined for transcriptional reporter activity upon coexpression of an active form of IRF3 (IRF3-5D) (23) with or without the p65 subunit of NF-κB, which widely activates NF-κB–binding elements (20). As shown in Fig. 4B, this reporter gene was activated by IRF3-5D, albeit weakly, and synergistically increased by p65 coexpression.

Fig. 4.

IRF3 activates Tslp promoter in cooperation with p65. (A) Schematic view of murine Tslp promoter and reporter genes (Upper). There are six putative IRF binding sites (ISREs) and four NF-κB sites in the Tslp promoter. The previously reported NF-κB site is κB3 (22); the arrow indicates the transcription start site. For simplicity, binding sites for other transcription factors are not denoted (32). Reporter plasmids containing ISRE sequences (Lower). (B) Reporter assay in HEK293T cells transiently cotransfected with a Tslp reporter plasmid (pTSLP-4k–Luc) and combinations of expression plasmid for IRF3-5D (100 ng) and/or p65 (1 ng); results are presented in relative light units (RLU) relative to Renilla luciferase activity. (C) Reporter assay in HEK293T cells with pTSLP-ISRE reporter plasmids performed as described in B. (D) Schematic view of pTSLP-ISRE1mt reporter genes (Upper). Reporter assay in HEK293T cells with pTSLP-ISRE1-luc and pTSLP-ISRE1mt-luc reporter plasmids and combinations of expression plasmid for IRF3-5D (100 ng) and p65 (1 ng) (Lower); results are presented as in B. **P < 0.01 and *P < 0.05 compared with RLU by pTSLP-ISRE1 reporter gene. All data are presented as mean ± SD of triplicate determinations.

We next examined the activity of each ISRE located within the 4-kb region upstream of the TSS for its IRF3-dependent gene transcription by fusing each ISRE-containing segment with a minimal promoter; here, a TATA box served as an acting cis-element (Fig. 4A). The reporter construct that was most notably activated, albeit not strongly, by IRF3-5D is pTSLP-ISRE1-luc (Fig. 4C), which contains two partially overlapping ISREs in proximity to the TSS (Fig. 4A). Interestingly, this activity was dramatically enhanced by coexpression of p65, suggesting that effective activation of ISRE-1 by IRF3 requires cooperation with NF-κB (Fig. 4C). Although relatively weak in its activity, pTSLP-ISRE6-luc was also activated by the coexpression of IRF3 and p65 (Fig. 4C). Expectedly, both ISRE segments, but not other ISRE fragments, contain an NF-κB–binding site in proximity to the ISREs (Fig. 4A).

To examine further the contribution of the ISRE and NF-κB sequences within the ISRE1 segment, we next introduced mutations in its sequence (Fig. 4D Upper) and examined its effect on reporter gene activation by the IRF3 and p65 coexpression. As shown in Fig. 4D Lower, reporter gene activation levels were significantly decreased when each of the ISRE sequences were mutated (pTSLP-ISRE1mt1-luc and pTSLP-ISRE1mt2-luc), particularly the upstream ISRE sequence, indicating the essential role of the upstream ISRE and ancillary role of the downstream ISRE. Because the very low level of the reporter gene activation observed with pTSLP-ISRE1mt1-luc remains the same with pTSLP-ISRE1mt1/2-luc, carrying mutations in both ISREs, we surmise that this residual activation is mediated by p65. Indeed, mutations within the NF-κB site essentially abolished reporter gene activation (Fig. 4D). Thus, these results underscore the critical cooperation of IRF3 and NF-κB in the activation of the Tslp promoter through direct binding to these ISREs and NF-κB sites.

Discussion

In the present study, we demonstrate an essential role for IRF3 to suppress DSS-induced colitis. In particular, we show that IRF3-deficient mice exhibit lethality in contrast to control mice, with severe defects in both the inflammatory and recovery phases of colitis (Fig. 1 A–E). We also found that colonic expression of Tslp and Il33 genes, which are critical to recover from colitis, was significantly decreased in the colon of the IRF3-deficient mice before and after DSS treatment (Fig. 1F). These results therefore indicate that the deficiency for the expression of these genes in the absence of IRF3 accounts for this pathogenesis, although in a strict sense it remains to be clarified whether IRF3 may also additionally regulate other genes involved in the suppression of colitis.

It is interesting that the IRF3-mediated induction of these genes occurs in both colonic epithelial cells and lamina propria cells, indicating that they are activated in many gut-associated cell types; these observations are consistent with previous report showing that Tslp and Il33 are expressed in many kinds of cells (24, 25). Although further work will be required to more rigorously analyze which particular cell types are principally responsible for the gene induction of these cytokines for the protection from colitis, we infer from our data that both epithelial cells and immune cells cooperate for the gut homeostasis in this disease model. In view of the well-known function of TSLP and IL-33 in Th2-type T-cell polarization, our current findings may also add an explanation for previous report showing that viral infections results in an enhancement of Th2-type response via the activation of the RLR-IRF3 pathway (26).

We provided evidence that the Tslp and Il33 gene inductions are mediated by DNA and RNA derived from the large intestinal contents via the STING and MAVS pathways (Fig. 3 A and C). The exact source and nature of these nucleic acids remains to be clarified. In view of our observation that antibiotic treatment of the mice results in a significant reduction of the mRNA expression for these genes (Fig. 2A), it is likely that the nucleic acids from commensal bacteria of the gut microbiota are the major source, the levels of which may become elevated by the destruction of colonic epithelia and mucosal layer upon DSS treatment. In this regard, it has been reported that bacteria-derived genomic DNA and cyclic di-GMP can activate the STING pathway (10, 27). It has also been reported that bacteria-derived double-stranded RNA has the potential to activate the RLR pathway in the intestine, although in this case RNA was entrapped in liposomes (8). In addition, we cannot rule out the interesting possibility that the augmentation of mRNA expression levels for these cytokines is caused by nucleic acids that are released by DSS-induced necrotic epithelial cells. Taken together, we surmise that the large intestine contains several types of immunogenic nucleic acids derived from bacteria and possibly necrotic epithelial cells, which have the potential to active Tslp and Il33 gene expression via stimulation of the cytosolic DNA and RNA pathways.

Of note, we found that large intestinal contents or feces per se can activate the cytosolic nucleic acid-sensing pathways to induce gene expression in cells including MEFs, and their immunogenic potential is highly resistant to nucleases (Fig. 3B). Thus, we speculate that because they cross into the cytoplasm, these immunogenic nucleic acids formed complexes with other molecules such as phospholipids. It is also interesting that Tslp gene induction occurs by the activation of either the RNA-MAVS or DNA-STING pathways, whereas Il33 gene induction is mainly dependent on the latter (Fig. 3 A and D). Although further work is required to address this point, these observations suggest the possibility that the two pathways activate distinct sets of transcription factors to evoke different patterns of gene expression.

Upon focusing on the transcriptional regulation of the Tslp gene, we found that of the 6 potential IRF-binding sites within the gene promoter, ISRE-1 and -6 are activated by IRF3. Notably, these ISREs are located in proximity to NF-κB sites, which together function to synergize gene activation by IRF3 and NF-κB, respectively (Fig. 4). These results suggest that IRF3 and NF-κB, both of which are activated by the cytosolic receptor signaling pathways, cooperate for the full-blown activation of the Tslp gene. However, we cannot rigorously rule out the possibility that other ISREs, not activated in our reporter assay, may also contribute to the gene induction.

It has been reported previously that TLR- or RLR-activated IRF3 is required for Il33 gene induction (18). This gene has two different TSSs, a so-called “upstream” termed uTSS and a “downstream” called dTSS that are more than 10 kb apart from one another and encode mRNA that differ only in the untranslated region. Transcription of the uTSS and dTSS is thought to depend on the kind of cell stimulation (18). Although the involvement of IRF3 and other transcription factors in the induction of Il33 gene has been argued (18, 28), the molecular regulatory mechanism(s) are still unclear. Here, our ChIP assay indicates that IRF3 binds to several regions within 1 kb upstream of the uTSS in RLR-stimulated macrophage (Fig. S8A). Consistent with these results, we found several potential ISRE sequences in this region (Fig. S8B). Although further investigation is required for more a detailed mechanism, these observations indicate the direct involvement of IRF3 also to Il33 gene transcription.

Our present study reveals the importance of the IRF3-Tslp/Il33 axis activated by cytosolic nucleic acid-sensing receptors, yet certainly other innate pathway(s) also contribute to this process and the protection of the colon from colitis. Indeed, we noted that in MEF cells with a knock-down of the MyD88, an adaptor protein for TLR that is not involved in IRF3 activation, gene induction by feces is also impaired (Fig. S9A). This suggests that an IRF3-dependent cytosolic pathway cooperates with TLR signals for the full-brown induction of these genes. Consistently, simultaneous TLR and RLR stimulation of MEFs resulted in synergistic induction of Tslp and Il33 gene expression (Fig. S9C). Although the nature of the signaling pathway and downstream molecules that cooperate with IRF3 await further investigation, these observations are consistent with previous reports showing the critical contribution of TLRs in the protection from DSS-induced colitis (3, 7, 8).

In conclusion, our present study provides insight into the complex regulatory mechanism of suppression of DSS-induced colitis, wherein the activation of IRF3 by nucleic acid-sensing innate receptors is critical through its induction of protective Tslp and Il33 gene expression. These findings may have critical implication for the suppression of colonic inflammatory diseases.

Materials and Methods

Mice and Reagents.

The generation of Irf3^−/−/Bcl2l12^−/− (21) and Irf5^−/− (29) mice, described previously, were maintained on a C57BL/6 (B6) genetic background. Ifnar1^−/− mice were purchased from B & K Universal Group. Peritoneal macrophages, bone marrow–derived dendritic cells, and MEFs were prepared as previously described (30, 31). Poly(dA-dT)⋅poly(dT-dA) (B-DNA) was purchased from Sigma. Poly(I:C) was purchased from GE Healthcare Biosciences. For large intestinal contents stimulation, colon from B6 mice was cut open and the contents were suspended in 20 mL of DMEM) SIGMA) supplemented with 10% (vol/vol) FCS. For feces stimulation, 25 μg of feces freshly isolated from B6 mice were suspended into 20 mL of DMEM. All animal studies and procedures were approved by the Committee on Animal Experimentation of the University of Tokyo, Tokyo, Japan. Additional information is available in SI Materials and Methods.