Chromatin remodeler ARID1A binds IRF3 to selectively induce antiviral interferon production in macrophages

Transcription factor IRF3 is critical for the induction of antiviral type I interferon (IFN-I). The epigenetic regulation of IFN-I production in antiviral innate immunity needs to be further identified. Here, we reported that epigenetic remodeler ARID1A, a critical component of the mSWI/SNF complex, could bind IRF3 and then was recruited to the Ifn-I promoter by IRF3, thus selectively promoting IFN-I but not TNF-α, IL-6 production in macrophages upon viral infection. Myeloid cell-specific deficiency of Arid1a rendered mice more susceptible to viral infection, accompanied with less IFN-I production. Mechanistically, ARID1A facilitates chromatin accessibility of IRF3 at the Ifn-I promoters by interacting with histone methyltransferase NSD2, which methylates H3K4 and H3K36 of the promoter regions. Our findings demonstrated the new roles of ARID1A and NSD2 in innate immunity, providing insight into the crosstalks of chromatin remodeling, histone modification, and transcription factors in the epigenetic regulation of antiviral innate immunity.


INTRODUCTION
Chromatin remodeling plays a critical role in regulating gene transcription [1,2]. Epigenetic factors remodel chromatin architecture to facilitate or inhibit the access of the transcription factors or regulatory proteins to the promoter region of genes, including those involved in the innate immune response and inflammation [3,4]. As the most studied epigenetic regulators, histone modifications such as methylation, acetylation, and ubiquitination, are precisely regulated to control the transcription of immune-associated genes [5]. Among those most important cytokines, type I interferons (IFN-I) including IFN-α and IFN-β have strong antiviral activities and also play important roles in the pathogenesis of inflammatory diseases [4,6,7]. The methylation status (mono-, di-or tri-methylation) of specific lysine (K) residues in the histone N terminal are associated with gene activation or repression. For example, methylation of K4 or K36 of histone 3 (H3K4, or H3K36) promotes gene transcription, whereas methylation of H3K9 or H3K27 inhibits gene expression [8,9]. Although di-/tri-methylation of K27 of histone (H3K27me2/3) was recently reported to promote IFN-β production in macrophages [10], the potential roles of histone modification in IFN-I signaling regulation are worth further investigation. Also, the crosstalk among multiple regulatory proteins and their synergistic effects in regulating transcription of Ifn-I during innate immune response remains to be fully understood.
BAF complex (Brg/Brahma-associated factors), known in mammals as the mSWI/SNF (SWItch Sucrose non-fermentable) complex, is an evolutionarily conserved ATP-dependent chromatin remodeler [11]. The mSWI/SNF complex has been intensively investigated in the context of stem cell development, the pathogenesis of the metabolic disease, and cancer. The mechanism for the action of the SWI/SNF complex was previously thought to regulate nucleosomes and facilitate chromatin accessibility after recruitment by transcription factors to the DNA promoter locus [12]. The mSWI/SNF complex has been reported to be essential for T cell development [13,14]. Besides, a previous study showed that ATPase BRG1, the core component of the mSWI/SNF, could bind with NF-κB and activate Ifnb transcription upon viral infection [15]. As another key component of the mSWI/SNF complex, ARID1A (AT-rich interacting domain-containing protein 1 A, also known as BAF250a) binds DNA in a non-sequencespecific manner by its ARID domain to guide the location of the mSWI/SNF complex and is also involved in the protein-protein interactions via its C-terminal region [16]. However, the role of the ARID1A subunit in innate immunity remains poorly understood and is worth further investigation.
In our study, we investigated the role of ARID1A in the antiviral innate response and found that ARID1A can selectively promote IFN-I production in macrophages. We revealed that, in response to viral infection, ARID1A can be recruited by IRF3 at the Ifn-I promoters and interact with the protein methyltransferase NSD2 which methylates H3K4 (H3K4Me3) and H3K36 (H3K36Me2) to facilitate chromatin accessibility at the Ifn-I promoters, resulting in the increased transcription of Ifn-I. Therefore, we demonstrated the interplay among the nuclear proteins including chromotin remodeler ARID1A, transcriptional factor IRF3, and histone modification protein NSD2 at the Ifn-I promoter regions which jointly regulate the production of IFN-I in innate immune cells upon virus infection. Our findings about the promotion of Ifn-I transcription by ARID1A during innate response provide new insight to the epigenetic regulation of innate immunity.

RESULTS
ARID1A selectively promotes IFN-I production in antiviral immune response Arid1a expression was quite abundant in immune-related cells and tissues according to the BioGPS website (BioGPS.org). The higher expression level of Arid1a was verified in the spleen and lymph nodes of mice compared to other organs (Fig. S1A). Immunofluorescence assay and Western blot confirmed that ARID1A protein was located in the nucleus of macrophages (Fig. S1B, C). However, we found no significant change of ARID1A protein expression in peritoneal macrophages (PMs), bone marrow-derived macrophages (BMDMs), and RAW264.7 cells before and after infection with RNA virus such as vesicular stomatitis virus (VSV) and Sendai virus (SeV) or DNA virus herpes simplex virus 1 (HSV-1) (Fig. S1D-F). Similarly, ARID1A protein expression remained constant in splenocytes of mice after in vivo infection with VSV (Fig. S1G).
Despite of the unchanged expression during the acute virus infection, we are curious about the function of ARID1A in antiviral immune response, considering its high abundance in immunerelated cells. Therefore, we firstly interfered with Arid1a expression in PMs and examined the production of cytokines including IFN-α, IFN-β, TNF-α, and IL-6 in response to VSV infection. The expression of IFN-I (IFN-α and IFN-β) was significantly decreased in Arid1asilenced PMs after VSV infection (Fig. 1A, B; Fig. S2A), whereas the expression of TNF-α and IL-6 remained unchanged (Fig. S2B, C). Then we generated Arid1a knock-out RAW264.7 cells using the CRISPR-Cas9 system (Fig. S2D). Compared with control cells, Arid1a-deficient RAW264.7 cells showed a substantial reduction of IFN-I but not TNF-α or IL-6 after VSV infection (Fig. 1C, D; Fig. S2E, F). Accordingly, the expression of interferon-stimulated genes (ISGs) such as Ifit1, Cxcl10, Ccl5, and Mx1 also reduced in the Arid1a knock-out RAW264.7 cells infected with VSV (Fig. S2G). On the contrary, Arid1a-overexpressed RAW264.7 cells exhibited substantially increased IFN-I production in response to VSV infection (Fig.  S2H, I and J). These data indicated that ARID1A selectively promotes IFN-I production in macrophages upon viral infection.
Next, we established myeloid cell-specific Arid1a conditional knock-out mice by crossing Arid1a flox/flox mice [17] with Lyz2-cre mice and confirmed the efficiency of Arid1a deletion in BMDMs (Fig. S3A, B). Previous studies have reported that deficiency of Arid1a does not affect the formation and stability of other subunits in the mSWI/SNF complex, but may influence the assembly and the recruitment of complex [18,19]. Therefore, we examined the expression of other important subunits in the mSWI/SNF complex and the interaction between different subunits in both Arid1a -deficient (Arid1a f/f Lyz-Cre + ) mice and control (Arid1a f/f Lyz-Cre − ) mice. Indeed, no difference was observed between the two groups with or without VSV infection (Fig. S3C), which suggested Arid1a deletion in BMDMs has no impact on the integrity or stability of the mSWI/SNF complex.
Consistently, we discovered the expression of IFN-I and ISGs were substantially decreased in Arid1a-deficient BMDMs infected with VSV, SeV, or HSV-1 respectively (Fig. 1E-J; Fig. S3D). However, the expression of TNF-α and IL-6 in Arid1a-deficient BMDMs was comparable with that in control BMDMs upon virus infection ( Fig.  S3E-J). By RNA-seq analysis on the Arid1a-deficient BMDMs infected with VSV, we further confirmed the selective decrease of Ifn-I expression accompanied with Arid1a deletion. Meanwhile, the differentially expressed genes included multiple genes involved in virus-associated signaling pathways ( Fig. S4A-D). Taken together, our data demonstrated that ARID1A selectively promotes IFN-I production in antiviral innate response.
ARID1A protects mice from virus infection by promoting IFN-I production FACS analysis showed that the ratios of splenic CD4 + T cells, CD8 + T cells, B cells, macrophages, dendritic cells, NK cells, neutrophils, and eosinophils were all comparable between Arid1a f/f Lyz-Cre + mice and littermate controls ( Fig. S5A and B), indicating deletion of Arid1a in myeloid cells has no effect on the development of immune cells. We also confirmed that the generation of BMDMs and PMs were not affected in Arid1a-deficient mice (Fig. S5C).
After challenging Arid1a-deficient mice and control mice by intravenously injecting VSV, we found that the production of IFN-I in the sera of Arid1a-deficient mice was much lower than those in control mice ( Fig. 2A), whereas the production of TNF-α and IL-6 were comparable (Fig. 2B). We also detected the significantly decreased Ifnb mRNA expression as well as increased VSV titer in the spleen, liver, lung, and PMs of Arid1a-deficient mice compared to their littermate controls (Fig. 2C, D). Consequently, the obviously increased infiltration of inflammatory cells into the lung of Arid1a-deficient mice was observed (Fig. 2E). Moreover, the lethality of VSV infection in Arid1a-deficient mice was higher than that in control mice (Fig. 2F). Thus, our results demonstrated ARID1A protects mice from virus infection by selectively promoting IFN-I production.
ARID1A facilitates Ifn-I promoters accessibility upon viral infection Given the known function of ARID1A in the mSWI/SNF complex, we performed a chromatin immunoprecipitation (ChIP) assay to examine the association between ARID1A and the promoter region of Ifnb and Ifnα. We found the recruitment of ARID1A to Ifn-I promoter regions gradually increased in BMDMs upon VSV infection (Fig. 3A), suggesting that ARID1A was involved in chromatin remodeling of Ifn-I promoter regions thereby affecting their transcriptional activity.
Next, we performed a luciferase reporter assay in HEK293T cells with co-transfection of Arid1a and Tbk1 plasmid. We discovered ARID1A significantly promoted the transcription of Ifnb reporter but not Tnfα or Il6 reporter (Fig. 3B, Fig. S6A). Moreover, DNA accessibility of Ifn-I promoters was significantly decreased in Arid1a-deficient BMDMs compared to control cells (Fig. 3C), but Arid1a deletion had no effect on the accessibility of Il6 or Tnfα promoter regions (Fig. S6B). Therefore, our results demonstrated that ARID1A promotes the production of IFN-I by facilitating Ifn-I promoter accessibility.
Viral infection induces IRF3 to bind and recruit ARID1A to Ifn-I promoters Since ARID1A is a sequence non-specific DNA binding protein, its selectivity is probably dependent on the associated protein(s) of ARID1A which could recruit ARID1A to the promoter regions of the given genes [16,20]. We performed Mass Spectrometry (MS) analysis of VSV infected BMDM lysates subjected to IP with ARID1A antibody and identified IRF3 as a candidate of ARID1A-associated proteins ( Fig. 3D; Fig. S6C). Then we confirmed the interaction between endogenous phosphorylated IRF3 with ARID1A in the nucleus of BMDMs upon infection with VSV, but no interaction was observed without VSV infection (Fig. 3E).
Next, immunoprecipitation from HEK293T cells co-transfected with Arid1a and truncated Irf3 plasmids revealed that deletion of the DNA binding domain (DBD; 1-140aa) or the IRF association domain (IAD; 198-374 aa) of IRF3 abrogated its interaction with ARID1A (Fig. 3F). We then performed a ChIP assay with ARID1A antibody in BMDMs derived from Irf3-deficient mice and control mice. As expected, the recruitment of ARID1A to the promoter regions of Ifn-I was significantly decreased in Irf3 -/-BMDMs in response to VSV infection (Fig. 3G), while absence of IRF3 had no effect on ARID1A recruitment to the promoter regions of Il6 or Tnfα (Fig. S6D). Therefore, our data suggested that ARID1A is selectively recruited by IRF3 to the Ifn-I promoters to facilitate chromatin accessibility.
ARID1A/IRF3 interact with and recruit NSD2 to promote H3K4Me3 and H3K36Me2 of Ifn-I promoters Histone modification plays a critical role in regulating DNA accessibility [11,21], which implies ARID1A might promote DNA accessibility of Ifn-I promoter regions through certain histone modification enzymes. By MS analysis, histone methyltransferase NSD2 (Nuclear receptor binding SET domain protein 2) was identified as another ARID1A-associated protein ( Fig. 3D; Fig. S6C). The interaction and co-localization between ARID1A and NSD2 were confirmed by both co-IP and immunofluorescence assay (Fig.  4A, B). We further determined that NSD2 interacts with ARID1A by its PWW1 and HMG domains (Fig. 4C). Meanwhile, We found IRF3 could interact with the C-terminal (629-1367aa) of NSD2, indicating that the SET domain-containing region of NSD2 mediates its interaction with IRF3 (Fig. 4D). Therefore, our data suggested ARID1A, IRF3, and NSD2 together form the complex to promote IFN-I production. NSD2 is widely known as a histone methyltransferase [22], which plays critical roles in regulating DNA accessibility [19]. Therefore, we analyzed histone methylation in the promoter regions of cytokines in response to VSV infection by ChIP assays. We revealed that both H3K4Me3 and H3K36Me2, two types of histone methylation-mediated activators of transcription, at Ifn-I promoters were substantially decreased in Arid1a-deficient macrophages (Fig. 4E, F). Still, Arid1a-deficiency did not make any difference in either H3K4Me3 or H3K36Me2 of Il6 and Tnfα promoters (Fig. 4G, H). Conclusively, our data demonstrated that ARID1A selectively facilitates chromatin accessibility of Ifn-I promoters by recruiting NSD2 to promote H3K4Me3 and H3K36Me2 of Ifn-I promoters.
In addition, transfection of full-length Arid1a or Nsd2 could rescue IFN-I production of Arid1a-deficient BMDMs (Fig. 6A-D; Fig.  S8A, B) but had no impact on IL-6 or TNF-α production (Fig. S8C,  D). Moreover, H3K4Me3 and H3K36Me2 levels of Ifn-I promoter regions were recovered after re-expression of wild type ARID1A or NSD2 (Fig. 6E, F). While transfection of either Arid1a mutant lacking ARID domain or Nsd2 mutant missing SET domain failed to restore IFN-I expression to the normal level. Taken together, our results demonstrated that ARID1A recruits NSD2 to mediate H3K4Me3 and H3K36Me2 in the promoter regions of Ifn-I thus promotes the production of IFN-I, indicating ARID1A enhances IFN-I expression dependent on its interaction with NSD2.

DISCUSSION
Production of IFN-I is tightly regulated in the antiviral innate response to effectively eliminate the invading virus but avoid the overproduction-induced tissue damage [23,24]. Multiple transcription factors including NF-κB, IRF3, IRF7, and AP1 have been reported to promote IFN-I production upon virus infection. Among those transcriptors, IRF3 plays the most critical role in the initiation of IFN-I production [25]. Here we show that ARID1A is recruited by IRF3 to the promoter regions of Ifn-I, where it binds histone methyltransferase NSD2 to increase chromatin accessibility and promote IFN-I  production in response to virus infection. Therefore, ARID1A binds IRF3 and NSD2 to form a complex in the promoter regions of Ifn-I to selectively promote expression of IFN-I, which establishes the crosstalk between histone modification and chromatin remodeling in transcriptional regulation of antiviral gene expression.
Previous reports showed that the mSWI/SNF complex is recruited to the promoter of Ifnb to force nucleosome displacement from the transcription start site [15,26,27]. However, the sequential recruitment of ATP-dependent remodelers and other modifiers to the promoter regions is still controversial. Also, whether the mSWI/SNF complex or specific components might participate in the regulation of interferon production remains unclear to date. ARID1A, which mediates the location of the mSWI/ SNF complex by its DNA-binding domain, is not essential for the ATP-dependent chromatin remodeling function of the complex [16,28]. Recent studies showed that ARID1A mediates the interaction between mSWI/SNF complex and YAP/TAZ thus involves in carcinogenesis and liver regeneration [29,30]. Our results indicated that ARID1A is not only a subunit of mSWI/SNF complex that helps to target specific regions of the genome, but also selectively and epigenetically promotes antiviral IFN-I production. On the one side, we found ARID1A interacts with IRF3, a crucial transcription factor that regulates the production of IFN-I in response to virus infection [31,32]. IRF3 could recruit ARID1A to the promoter sites of Ifn-I which could explain ARID1A involves specific regulation of IFN-I but not other cytokines production. On the other hand, ARID1A also binds with methyltransferase NSD2 to change the chromatin accessibility of Ifn-I promoter regions by histone modification.
Histone acetylation has a critical role in the regulation of IFN-I production during the innate response against viruses [33]. However, histone methylation has been rarely recognized as a mechanism of transcriptionally regulating IFN-I production [3,34]. As one of the most important methyltransferase groups, the NSD family could monomethylate and dimethylate histone H3 on lysine 36 (H3K36) [35]. Besides, the NSD family also targets the non-histone protein. For example, NSD1 or NSD3 could promote antiviral innate response by mediating p65 methylation at lysine residues K218 and K221 [36], or promoting IRF3 methylation at lysine residue K366 respectively [37]. Also, NSD2 was reported to interact with histone chaperone SPT6 and facilitate interferoninduced transcription [38]. Here, we illustrated NSD2 could interact with ARID1A to promote antiviral immune response by methylating H3K4Me3 and H3K36Me2, thus increasing chromatin accessibility in the promoter regions of Ifn-I. Our results suggested IRF3, ARID1A and NSD2 together compose the complex at the promoters of Ifn-I to cooperate in the regulation of IFN-1 expression. Further work is needed to clarify the exact association inside the IRF3-ARID1A-NSD2 complex and evaluate the potential interactions between the IRF3-ARID1A-NSD2 complex and the mSWI/SNF complex. In addition, the question of whether there would be other protein regulators involved in the abovementioned complex or whether this complex might target to other possible genes is yet to be answered.
Pathogens have evolved a variety of epigenetic strategies for their survival and replication, such as induction of host protein and chromatin modifications to inhibit the expression of activators and enhance the expression of repressors in innate immunity [ 39,40]. The analyses of the GEO database showed the decreased Arid1a mRNA expression in hepatitis C patients (GSE38597), gastric cancer infected with Epstein Barr virus [41], epithelial cell model of kaposarcoma virus infection (GSE1377), and cervical cancer infected with human papillomavirus [42]. These results indicated that in chronic viral infection, especially along with carcinogenesis, Arid1a expression may be downregulated and involved in the immune escape and pathogenesis. In the current study, we utilized virus-infected macrophages as the cell model and found that the protein expression of ARID1A remains unchanged in these cells in response to acute infection with RNA or DNA viruses. The results indicated that ARID1A may have different responses to acute and chronic viral infections. More studies are further needed to explore the expression pattern and roles of ARID1A protein in infectious diseases and virus infection-related cancers.
In conclusion, our study provides a mechanistic insight into the epigenetic regulation of antiviral innate immune response by integrating histone modification with chromatin remodeling. Our study may provide a potential target for the control of virus infectious diseases and other inflammatory diseases.

Cell culture
PMs were obtained from mice 3 days after intraperitoneal injection with thioglycolate, BMDMs were prepared by culturing in recombinant 50 ng/ml mouse M-CSF (Perprotech) for 7 days. HEK293T, RAW264.7, and Vero cells were obtained from American Type Culture Collection. Cells were all cultured in DMEM supplemented with 10% FBS.
Pathogens VSV (Indiana strain) and HSV-1 were amplified from Vero cells and SeV (Sendai virus) was amplified from chicken embryos as described previously [43].

In vivo animal experiments
The mice were infected with VSV by tail vein injection (5 × 10 7 pfu/g). Serum was collected for ELISA detection at indicated times. The RNA of the spleen, liver, lung, and PM cells were extracted by TRIzol (Invitrogen). The titer of VSV in cell lysates (0.1%Triton in PBS) of mouse organs was detected by TCID50. The lung lobes of mice were fixed with 4% paraformaldehyde. After embedding, sections were stained with hematoxylin and eosin, and observed with Pannoramic SCAN. For survival curve analysis, mice were intravenously injected with VSV (1 × 10 8 pfu/g), then generated and analyzed for statistical significance with GraphPad Prism 5.0.

Immunofluorescence
PMs plated on glass coverslips were infected with or without VSV for 8 hours and labeled with antibodies. Antibody to ARID1A (32761) was from Santa Cruz. Antibody to NSD2 (A7938) was from ABclonal. Antibody to Alexa488 (A1108) and Alexa594 (A11005) were from Invitrogen. Slides were subjected to microscopy analysis with Zeiss LSM780 confocal laser microscope.

ELISA
ELISA kits (Biolegend and Ebioscience) were used for examining the concentrations of cytokines in cell supernatants or serum of mice according to the manufacturer's instructions.

RNA-seq detection and analysis
Total RNAs were isolated with TRIzol. Then RNA quantification and qualification were confirmed. Library construction and sequencing were performed on Illumina Novaseq 6000 (Novegene, http://www.novogene. com/). Gene FPKMs were computed by summing the FPKMs of transcripts in each gene group. Transcripts with a P-adjust < 0.05 were assigned as differentially expressed. KOBAS software was used to test the statistical enrichment of differentially expressed genes in KEGG pathways.

ChIP-PCR
The cells were treated as the instructions of the ChIP assay kit (Beyotime P2078). For each ChIP assay, antibodies were added to cell samples and incubated overnight at 4°C. Antibody-chromatin complexes were pulled down by magnetic protein G beads (Invitrogen). DNA was eluted, purified, and subjected to qPCR. Antibodies to ARID1A (ab182560) and H3K4Me3 (ab8580) were from Abcam. Antibody to H3K36Me2 (39255) was from Active Motif. Sequences of the primers for qPCR are the same as the primers in DNase I Sensitivity Assay.