Indeed, there has already been significant data in the literature to support the mechanism(s) of epigenetic regulation in autoimmune diseases. In particular, DNA from SLE T cells was found to be less methylated than control DNA from normal T cells by measuring the cellular deoxymethylcytosine content (23). Interestingly, non-T cells from lupus patients displayed normal DNA methylation levels (24). Decreased DNA methyltransferase (DNMT) activity in lupus T lymphocyte nuclear proteins was considered to be responsible for the observed DNA hypomethylation in lupus T cells. Patients with lupus had significantly lower levels of DNMT1 mRNA, but not DNMT3A or DNMT3B, as compared with healthy controls (25). A preliminary analysis of microarray data from immature monocyte-derived dendritic cells (MDDCs) revealed that they express abundant amounts of DNMT1, which is downregulated after LPS stimulation. The methylation status of DNA from SLE PDCs and the levels of DNMT1 expression in this important IFN-α producing cell type are not currently known.

In general, hypermethylation in the promoter of a gene is associated with gene suppression, while hypomethylation is linked to gene expression; methylation within the gene body is also associated with gene expression. Two next-generation sequencing technologies have recently been developed for the analysis of gene methylation – methylated DNA immunoprecipitation sequencing (MeDIP)-seq, to detect methylated CpGs (26, 27), and MRE-seq, to detect unmethylated CpGs (28). Integrative methodologies that combine both MeDIP-seq and MRE-seq can differentiate hypermethylation, intermediate, and hypomethylation regions of DNA. An integrative analysis of KLHL9 indicates that the CpG islands of the KLHL9 promoter are highly hypomethylated (Figure 1). These islands are highly conserved since they were found to be present in virtually all cell types queried. Combining these data with ChIP-seq histone modification data in the same tissues, we found that hypomethylated CpGs of KLHL9 are occupied by significant levels of trimethylated lysine 4 on histone H3 (H3K4me3) (Figure 1). Two other hypomethylated regions in the type I IFN cluster, located in the genomic region between IFNA2 and IFNA8, have relative low levels of enrichment for H3K4me3 peaks (Figure 1). H3K4me3 is a histone modification that accumulates at the transcription-start site (TSS) of active genes and is believed to be important for transcription activation. Loss of H3K4me3 occurs at TSSs and leads to gene transcriptional inactivation as a result of promoter hypermethylation. The occupancy of H3K4me3 in the promoter of KLHL9 may ensure the protection of CpG islands from methylation. In contrast, the other two hypomethylated sites that are located quite far from the TSSs of IFNA2 and IFNA8, may not be functional for transcription. DNA methylation by RRBS from various cell types, including B cells, failed to reveal strong methylation signals in the IFNA gene cluster. One exception to this is that MeDIP-seq defined methylation peaks were found to be distributed between IFNA genes from brain tissue.

Thus far, data do not support that methylation is the likely major mechanism by which IFNA gene expression is suppressed in most non-IFN-producing cells. Further experimental studies will be necessary to determine whether constructive hypomethylation, as well as H3K4me3 occupancy, is important for regulating IFNA gene transcription in IFN-producing cells such as monocytes, and PDCs.

There are multiple IFNA and IFNB genomic regions that have open chromatin structure in an evolutionally conserved pattern across species and most human cell types. Since DNase I hypersensitive sites (DHSs) reflect the local openness and accessibility of chromatin, chromatin structure or accessibility of IFNA clustering may be similar among different cells. In general, hypersensitive sites are found only in the chromatin of cells in which the associated gene is being expressed, and do not occur when the gene is inactive. Therefore, mapping DHSs within nuclear chromatin is a powerful method of identifying genetic regulatory elements (29). However, the distribution of DHSs in promoters and other gene regions of similarly expressed genes differs among different chromosomes. Furthermore, silenced genes have a more open chromatin structure than previously thought and DHSs in 3′-untranslated regions (3′-UTRs) have been shown to negatively correlate with gene expression levels (30), thus going against the standard dogma. Bioinformatics analysis of DHSs in the IFN gene cluster between different cell types revealed a highly conserved pattern (Figure 2); however, we found additional DHSs in CD14⁺ monocytes that can produce type I IFNs. We also found that CD34⁺ stem cells have more DHSs close to promoters within the IFN gene cluster (Figure 2). These data support the presence of unique cell-specific chromatin structures which may play important regulatory roles in the control of type I IFN expression.

Modifying the chromatin template at a particular gene locus can also serve as an important mechanism of gene transcriptional activation that exhibits cell-type specific expression patterns. The functional importance of histone acetylation in type I IFN production has been supported by studies that show increased IFN-β expression in cells treated with histone deacetylase inhibitors, such as Trichostatin A (TSA) (31), and decreased IFN-β expression in murine macrophages where the binding of bromodomain-containing BET (bromodomain and extraterminal) transcriptional regulators to acetylated histones was inhibited (32). Di- or tri-methylation of H3K9 is capable of suppressing gene expression not only passively, by inhibiting acetylation, but also actively, by recruiting transcriptional repressors of the heterochromatin protein 1 (HP1) family. We found that H3K9me2 occupancy at IFN and ISG promoters is inversely correlated with gene expression. Furthermore, human MDDCs that are capable of producing type I IFNs, as compared with human lung fibroblasts that do not, show decreased H3K9me2 occupancy at the IFNB promoter. In the absence of G9A, a methyltransferase for H3K9me2, non-professional IFN-producing cells were shown to be converted into potent IFN-β producers (33). Together, these data support the importance of histone modifications in the regulation of type I IFN expression.

H3K27me3, on the other hand, are found to be associated with the repression of gene transcription in a cell-type specific manner. Polycomb Repressive Complex 2 (PRC2) is a histone methyltransferase that catalyzes tri-methylation of Histone 3 at Lysine 27 (H3K27me3) (34). A detailed profile of H3K27me3 peaks reveal that broad peaks at TSS are associated with transcriptional suppression while skewed peaks up-stream of the TSS may not be suppressive (35). Indeed, we found that IFNA regions in B cells, which are incapable of producing IFN-α, are widely occupied with H3K27me3, as shown by the substantive peaks found along the gene cluster (Figure 3). In contrast, ChIP-seq data from monocytes demonstrate that H3K27me3 peaks occupy some IFNA genes, such as IFNA2, IFNA14, and intergenic regions between IFNA2 and IFNA8, while the remaining IFNA genes were not suppressed by H3K27me3.

As mentioned above, current dogma holds that H3K4me3 represents a chromatin landmark that is present at the TSS for genes that are either actively transcribed or permissive for gene transcription. However, H3K4me3 are not sufficient to license cells to produce IFN-α. For example, multiple H3K4me3 occupancy peaks can be identified in the IFN regulatory regions in B cells that do not express IFN-α. A good comparison would be with PDCs, yet the histone codes are not yet available for this cell type. In PDCs, TLR-7 signaling quickly turns on transcription of IFNB, IFNA2, IFNA8, and IFNA14 genes at 30 min post-stimulation with peak levels being achieved at this time point. In comparison, peak levels of IFNA5, IFNA6, IFNA10, IFNA13, and IFNA21 were observed around 4 h post-stimulation (36). Based on our bioinformatics analysis, we reason that transcriptional suppression by H3K27me3, if it exists in PDCs in a pattern similar to that found in CD14⁺ monocytes, may not be functional in PDCs or can quickly be replaced by H3K4me3 after TLR-7 activation. Alternatively, the IFN gene cluster in PDCs may not have H3K27me3 markers. It is not known whether chromatin change is necessary for IFNA transcriptional activation or whether chromatin status is responsible for differentially transcribed type I IFN genes. Further studies in human PDCs will be required to address this.

The transcriptional repressor CTCF (11-zinc finger protein) or CCCTC-binding factor is thought to regulate the 3-dimensional (3D) structure of chromatin by binding strands of DNA together and forming DNA loops (37). CTCF represses gene expression by blocking the interaction between enhancers and promoters (38). This phenomenon may serve as a chromatin barrier to block the spread of heterochromatin structures and set boundaries between active chromatin regions marked by histone H2A acetyl Lys5 (H2AK5ac) and repressive regions marked by H3K27me3 (39). The cohesin complex, consisting of cohesion proteins SMC1, SMC3, SCC3, and the α-kleisin SCC1, may contribute to CTCF-mediated repression. Many CTCF/cohesin binding sites are located at promoter regions suggesting a joint regulatory role for these factors (40). Although most cohesin sites overlap with CTCF, a significant proportion of each factor’s sites are independent of the other, implying CTCF-independent functions of cohesin as well as cohesin-independent CTCF functions. Bioinformatics analysis of CTCF ChIP-seq data from ENCODE cell lines identified several CTCF insulators that are basally located in the promoters and intergenic regions of IFNA5, A1, A2, and A8 (Figure 4A). In gum fibroblast cells (AG09319), we found five CTCF binding sites in the region covering IFNA14, A17, A16, A10, A7, and A4. There is only one CTCF/SMC3 binding site in the IFNB gene. The regulatory region of the IFNA2 gene contains two CTCF binding sites. The second CTCF site yields co-binding with SMC3, suggesting the cohesin complex may function in this IFN genomic region. Based on these data, we speculate that CTCF may indeed function as an IFNA suppressor and block promoter activation.

Within the IFNA gene cluster, we have yet to identify any other TF in the ENCODE datasets that basally occupies the promoter regions between TSSs and the proximal CTCF sites. In contrast, multiple TFs, such as NF-κB and PU.1 (in B cells), do constitutively occupy regions up-stream of CTCF sites that control individual IFNA genes. CTCF binding sites are not conserved but cell-type specific. While the majority of cells show CTCF occupancy up-stream of the IFNA2 gene, binding is absent in fibroblast cells. Similarly, at the IFNB promoter, CTCF binding was identified in some B cells lines, HeLa cells, MCF-7, and osteoblast cells, but not in any fibroblast cell lines or A549 lung carcinoma cells. Lack of binding of this insulator may render fibroblast cells to produce type I IFNs upon the appropriate stimulation, such as viral infection, thus supporting that CTCF binding to the IFNA gene may be regulated. In this regard, dexamethasone treatment in A549 cell lines induces CTCF to bind to the IFNA8 promoter. Finally, the discrepancy of CTCF binding patterns in Epstein–Barr virus (EBV)-transformed B cell lines suggests that viral infection may interfere with CTCF function. It is known that CTCF/cohesin occupancy is essential for IFN-gamma (IFNg) gene transcription (41). Thus, this complex may have a similar function and be important for regulating IFNB gene transcription via maintaining the 3D chromatin structure of the IFNB locus in fibroblast cells (Figure 4B). Based on these data, we propose that the DNA regions in the IFN gene cluster that contain CTCF occupancy may be subject to control by this factor to ensue IFNA transcription during viral or viral-like challenges in IFN-producing cells. This region may also be used as a landmark to demarcate the promoter region that spans from a TSS to the CTCF binding sites and enhancer regions located up-stream of CTCF binding site.

Interferon regulatory factors, as their name suggests, have been long known to regulate type I IFN gene expression (42). Of the nine mammalian IRF family members currently identified to date, IRF7 has garnered the most attention for its role in regulating IFNA gene expression (3). IRF7 is highly expressed in human PDCs and allows bypass of the classic autocrine feedback loop that is regulated by IFN-β (43). IRF7 was also shown to be required for murine PDCs to produce an antiviral IFN immune response (44). Similarly, IRF5 has been implicated in the regulation of type I IFN gene expression (45). Early data in human cell lines revealed the regulation of type I IFN genes and IFN-stimulated genes (ISGs) by IRF5 in response to virus (46). Later data in mice supported these findings. For example, splenic PDCs from mice lacking Irf5 were shown to produce less type I IFNs in response to virus infection (47). IRF5 has also been recently reported to regulate IFN-β production in myeloid dendritic cells downstream of the mitochondrial antiviral-signaling protein (MAVS) (48). Furthermore, recent studies demonstrate that IRF5 and NF-κB p50 are key co-regulators of IFN-β and IL-6 expression in TLR9-mediated activation of human PDCs (49). Although both of these IRF family members have been implicated as key regulators of IFN-α production, no ChIP-seq data is available to support these findings. Interestingly, data from the aforementioned STAT4/IRF5 ChIP-seq datasets in PBMCs did not support the direct regulation of type I IFN expression by IRF5 since no peaks were detected in the IFN gene cluster after immunoprecipitation with anti-IRF5 antibodies (22). In this case, PBMCs were stimulated with either IFNa2 or SLE immune complexes before immunoprecipitation with anti-IRF5 or anti-STAT4 antibodies. In the case of IRF7, a cursory review of the literature and publicly available datasets indicate that no ChIP-seq data is currently available for this TF. We have recently performed IRF5 and IRF7 ChIP-seq in human PDCs stimulated with virus. Our unpublished data indicate that these two TFs bind to different regions in the IFNA gene cluster (Figure 5). These data support the distinct and differential roles for IRF5 and IRF7 in type I IFN gene regulation (45, 49). With regard to autoimmune diseases such as SLE and SS that display a pathogenic type I IFN gene signature, determination of the mechanisms by which these two IRF family members cooperatively and distinctly regulate IFNA subtype expression in the critical IFN-α producing cell types will be important for the design of new therapeutic strategies targeting these two factors.