Introduction

Embryonic stem cells (ESC) originate from the blastocyst stage of the developing embryo. They possess self-renewal ability and pluripotency, which make them invaluable in vitro models for studies focusing on early embryonic development. While self-renewal enables culturing ESCs almost indefinitely, through their pluripotent nature, ESCs can be directed to differentiate towards any lineage in vitro. Mimicking embryonic development, ESC commitment towards mesoderm, endoderm and neuroectoderm can be achieved with the carefully timed use of a combination of cytokines (1,2).

ESC stage is established and maintained by pluripotency transcription factors (TF), such as OCT4, SOX2, NANOG and KLF4 (3). During ESC differentiation, the pluripotency network achieved by these TFs need to be shut off while a lineage-specific gene expression program is established. Activation of lineage-specific genes poses a challenge, since their chromatin environment is not permissive of transcription at the ESC stage. Therefore, TFs that initiate differentiation need to gain access to their gene regulatory sequences at the promoter and enhancers and increase the accessibility of these regions to downstream transcription related machinery. The lineage-specific TFs that start this cascade of events are called pioneer TFs (4). Endoderm lineage originates from a mesodermal embryonic structure called the primitive streak and thus share an early common differentiation path (5,6). Neuroectoderm, on the other hand, is distinct from meso/endodermal lineages (7). BRACHYURY (Bry) was identified as a pioneer TF for mesodermal lineage and is used as a marker gene for meso/endodermal commitment. Similarly, FOXA2 is a pioneer TF for definitive endodermal lineage.

TFs instigate ESC differentiation by alteration of the target gene chromatin environment (8). Eukaryotic chromatin is made up of repeating units of conserved proteins called histones (9). Two copies of histones 2A, 2B, 3 and 4 form nucleosomes that wrap the genomic DNA. Modifications of histones at particular modifiable amino-acids, the number or the positioning of nucleosomes across the gene region, the use of histone variants in the nucleosome structure or methylation of DNA can change the chromatin environment and alter the frequency or efficiency of transcription.

Histones can be methylated at lysine or arginines, acetylated at lysines and phosphorylated at serine or threonines (9). Modification itself, as well as the exact residue that gets modified, can have profound effects on transcription. For example, histone acetylation leads to weakening of interaction between nucleosomes and DNA. High levels of histone acetylation, especially around promoter regions, help create an accessible environment for recruitment of factors critical for transcription initiation. Thus, histone acetylation generally correlates with active transcription, although acetylation in specific lysines can have unique outcomes. The effect of histone methylation depends on the level of methylation (mono, di or trimethylation) and which residue is methylated. While H3K4me3 is generally observed over transcriptionally active gene regions, H3K27me3 plays a suppressive role.

During differentiation, lineage-specific genes become more accessible and start harboring histone modifications that correlate with active transcription such as high level of histone acetylation including H3K27Ac and H3K4me3. Pluripotency genes are instead suppressed with repressive histone modifications, such as H3K27me3 or DNA methylation (10).

Our previous study focused on chromatin factors that regulate mouse embryonic stem cells (mESC) differentiation (11). We identified ARID4B as an important factor that is required for both mesoderm and endoderm commitment. In its absence, mESCs fail to express critical pioneer TFs for these lineages and cannot instigate differentiation. The failure of meso/endodermal differentiation was not due to improper pluripotency network shut-off. Instead, ARID4B loss resulted in accumulation of repressive H3K27me3 histone markers on lineage-specific genes.

As part of an RPD3 histone deacetylase complex, ARID4B regulates the overall histone acetylation landscape, primarily around promoter and enhancer regions (12). Histone acetylation levels around regulatory regions can impact transcriptional output by altering their accessibility to binding of other critical factors. ARID4B lacks any enzymatic domain but has a DNA and various protein-protein interaction domains. Particularly, the presence of TUDOR, PWWP and chromobarrel domains indicate interactions with modified histones and chromatin in general. DNA binding is mediated through its AT-rich interaction domain (ARID) in a sequence-independent manner. Recruitment of ARID4B along with the rest of the RPD3 complex generally leads to transcriptional suppression of target genes through histone deacetylation. In order to further delineate the mechanism of ARID4B function in mESC differentiation, we therefore decided to focus on pathways that are activated in the absence of ARID4B.

Material and Methods

mESC culture and differentiation

The study was done using already established mESC lines. Therefore, ethics committee approval form or institutional review board forms were not required. ARID4B knockout mESCs were prepared previously (11). Wild-type and arid4b∆ mESCs were cultured using standard serum containing medium on irradiated MEFs (13). mESC differentiation towards mesoderm and endoderm was performed using previously optimized protocols (11,14).

Transcriptomic analysis and RT-qPCR

RNA extraction, cDNA synthesis and quantitative polymerase chain reaction (qPCR) were done using previously published protocols (11). rRNA minus RNA-sequencing was previously performed (11). The transcriptomic data is available at NCBI GEO (GSE153633). Primers used for real-time quantitative polymerase chain reaction (RT-qPCR) are listed in Table 1.

Western blot

Cellular extract preparation and Western blot was done using previously published protocols (11). Antibodies used were: ARID4B (Bethyl Laboratories, A302-233A), ACTIN (Millipore, MAB1501), STAT1 (CST, 9172), pSTAT1 (CST, 9167), STAT3 (CST, 9139), pSTAT3 (CST, 9145).

Chromatin immunoprecipitation-sequencing and analysis

H3K4me3, H3K27Ac and H3K27me3 chromatin immunoprecipitation (ChIP)-seq experiments and subsequent data analysis were previously performed in wild-type and arid4b∆ cells on day 5 of endoderm differentiation (11,15). Sequencing data is available at NCBI GEO (GSE153634).

Statistical analysis

Experiments were conducted with three independent biological replicates. RNA-seq and ChIP-seq experiments, the statistical analysis was done within the analysis package (11). Algorithms used within the study, such as Ingenuity Pathway Analysis (IPA), Gene Set Enrichment Analysis (GSEA) and Genomic Regions Enrichment of Annotations Tool (GREAT) have their own statistical analysis embedded within and were used as default. RT-qPCR results were graphed using GraphPad Prism Software and statistically analyzed using Student’s t-test (^*: p≤0.05, ^**: p≤0.01, ^***: p≤0.001).

Results

Meso/endodermal differentiation defect observed in arid4b∆ mESCs might stem either from pathways that cannot be turned on or from those that are aberrantly upregulated. Our previous studies had focused on which pathways and functions were lost in arid4b∆ mESCs as they differentiate towards meso/endoderm lineages. In this paper, we decided to investigate abnormally upregulated functions.

We had previously optimized the embryoid body-based mesoderm and endoderm differentiation protocol and could observe upregulation of pioneer TFs for either lineage by day 5 of differentiation (14). arid4b∆ mESCs could not induce the expression of these TFs, leading to meso/endodermal commitment defect (11). RNA-seq analysis of day 5 samples from wild-type or arid4b∆ cells showed no apparent increase in meso/endoderm related signaling and gene expression pathways. Inappropriate pathway induction in arid4b∆ cells might hinder successful differentiation. We, therefore, looked for pathways that are enriched in the differentially upregulated genes in arid4b∆ cells over wild-type cells on day 5 of endoderm differentiation. IPA (Qiagen) revealed interferon signaling, activation of cytosolic pattern recognition receptors and the role of pattern recognition receptors in recognition of bacteria and viruses as the most enriched pathways (Figure 1a). Similarly, GSEA (the Broad Institute) showed cellular defense response and interferon α/β signaling as aberrantly upregulated in arid4b∆ cells (Figure 1b, c). Additionally, the IRF TF motif was found to be enriched in genes that were more expressed in arid4b∆ cells over wild-type (Figure 1d). The overrepresentation of innate immune response-specific genes (red dots) in arid4b∆ cells was visualized using the volcano-plot (Figure 1e, Table 2) and Integrative Genomics Viewer [(IGV), the Broad Institute] (Figure 1f-k).

We wondered whether ARID4B loss would lead to a similar upregulation of interferon signaling in mesoderm differentiation. To investigate this hypothesis, we directed wild-type and arid4b∆ mESCs towards mesodermal lineage and assessed differentiation success using the induction of mesodermal pioneer TF BRACHYURY. arid4b∆ mESCs were unable to express BRACHYURY, as well as its downstream target genes (11). IPA and GSEA similarly identified interferon signaling and activation of IRF by cytosolic pattern recognition receptors among top canonical pathways in arid4b∆ cells (Figure 2a-c). However, volcano plot representation showed that the interferon signaling related genes were not as highly expressed in mesoderm as they were in endoderm (Figure 2d, Table 3). Regardless, they showed apparent upregulation in arid4b∆ cells compared to wild-type (Figure 2e-j). These results collectively point to a role of ARID4B in suppression of interferon response during ESC differentiation towards meso/endodermal lineages.

The next step was to validate transcriptomic data and perform RT-qPCR for a list of highest differentially upregulated genes in arid4b∆ cells. Interferon induced transmembrane protein encoding IFITM3 was the top of this list and showed clear induction upon ARID4B loss on day 5 of endoderm differentiation (Figure 3a). As the major interferon signaling mediator, STAT1 was expressed markedly higher in arid4b∆ endodermal cells (Figure 3b).  Cytosolic dsRNA sensor DDX58 (RIG-I) and related antiviral protein encoding DDX4 were similarly upregulated (Figure 3c, d). The expression of these genes was still significantly higher than wild-type in mesoderm commitment (Figure 3e-h) but their fold induction was more severe during endoderm differentiation. Therefore, we focused on endoderm differentiation hereafter.

To this point, results had been obtained at a specific time point (day 5) during mESC differentiation. In order to gain a more detailed understanding of aberrant induction kinetics of immune-related genes in arid4b∆ cells, RT-qPCR analysis was performed in wild-type and arid4b∆ mESCs through the time course of endoderm differentiation. At the mESC stage (depicted as day 0 of differentiation in graphs), IFITM3 was already highly expressed in arid4b∆ cells, although the fold difference seemed to increase later during endoderm differentiation (after day 3) (Figure 3i). STAT1 showed a highly variable but elevated expression in arid4b∆ mESCs and its expression remained elevated throughout our time-course (Figure 3j). DDX58 (RIG-I) followed a similar pattern to STAT1 (Figure 3k) while DDX4 levels were similar between wild-type and arid4b∆ mESCs and became elevated in arid4b∆ cells only late in endoderm differentiation (Figure 3l). These results show that ARID4B loss leads to higher interferon-related gene expression at the mESC stage and this effect gets stronger through the endoderm differentiation time-course with variable kinetics and extent for each target gene.

Transcriptional upregulation might not necessarily lead to protein level alterations and pathway activation. To understand whether observed increased transcripts activate interferon signaling in arid4b∆ cells, we performed Western blot using total cell lysates of mESCs (day 0 of differentiation) and during the time course of endodermal differentiation. Cytosolic double-stranded viral RNAs are recognized by DDX58 (RIG-I), which in turn activate expression and secretion of interferons including interferon α and interferon β (16). Interferons bind to their cognate receptors on neighboring cells and lead to activation of STAT1 and STAT2 through phosphorylation. Our western blot results showed elevated STAT1 level in arid4b∆ mESCs (day 0 of differentiation) (Figure 4a). Following exit from pluripotency, in both cell type, STAT1 levels declined although they still remained higher in arid4b∆ cells than in wild-type. Since interferon pathway activation correlated with the active phosphorylated STAT1 (pSTAT1), we compared its level in wild-type or arid4b∆ cells. At the mESC stage, there was a dramatic increase in pSTAT1 upon ARID4B loss. Mirroring total STAT1 level, the phosphorylated form was reduced upon endoderm differentiation. However, arid4b∆ cells had markedly more pSTAT1 than wild-type, even during differentiation.

Pluripotency of mESCs is regulated through a related STAT protein, STAT3 (17). We wanted to test whether the effect seen in arid4b∆ mESCs is specific to STAT1 or it also affects the STAT3 pathway. Unlike STAT1, total and phosphorylated forms of STAT3 were similar in wild-type and arid4b∆ cells at the mESC stage and through endoderm commitment (Figure 4b). These results show that the transcriptional upregulation of the pathway, originally observed through our RNA-seq results, is indicative of interferon-related STAT1 pathway activation in arid4b∆ cells.

We next wondered what the mechanism behind ARID4B-dependent regulation of interferon signaling was. To investigate this, changes in the chromatin environment of target genes were compared in endoderm committed cells (day 5). ChIP experiments were done for H3K4me3, H3K27Ac and H3K27me3 marks in wild-type or arid4b∆ endoderm cells (11). H3K4me3 and H3K27Ac generally correlates with active transcription while H3K27me3 plays a suppressive role (9). Our results suggest unique chromatin-based mechanisms lead to upregulation of target genes. For example, OASL2 did not show a change in H3K4me3 or H3K27me3 level but had higher H3K27Ac in arid4b∆ cells, presumably leading to higher transcript levels (Figure 5a). The STAT1 region had elevated H3K27Ac markers concomitant with decreased H3K27me3 (Figure 5b).  In comparison, DDX58 and IRGM2 only had marked H3K4me3 increase in arid4b∆ cells (Figure 5c, d), while in OAS1A and IFITM3, H4K4me3 increase was accompanied with a similar increase in H3K27Ac (Figure 5e, f).

In order to understand which of these histone marks is more specifically associated with interferon pathway activation in arid4b∆ cells, we performed GREAT analysis (Stanford University) for ChIP-seq DNA sequences. This tool associates ChIP peaks with nearby genes and surveys a possible pathway enrichment. We performed GREAT for H3K4me3, H3K27me3 and H3K27Ac ChIP results in comparison with arid4b∆ over wild-type. We found that relevant pathways, such as defense response to virus, immune effector process and innate immune response, were only enriched in regions with higher H3K4me3 level in arid4b∆ cells over wild-type (Figure 5g). Although increase in H3K27Ac or loss of H3K27me3 markers were seen in individual target genes (Figure 5a, b, e, f), our results indicate a more generalized effect on H3K4me3 leads to interferon pathway activation upon ARID4B loss.

Discussion

Our previous results showed a clear mESC differentiation defect upon ARID4B loss. Derepression of lineage-specific meso/endoderm TF was not observed in arid4b∆ cells, despite cytokine-directed differentiation conditions.

We had previously investigated TF binding motifs enriched in our ChIP results (15). No particular TF motif was found enriched in H3K4me3 or H3K27me3 ChIP peaks. Interestingly, STAT1::STAT2 binding motif was enriched in regions with higher H3K27Ac level in arid4b∆ cells. When the ChIP peaks were separated based on their distance to transcription start site, it became clear that the H3K27Ac increases in arid4b∆ cells coincided with STAT1::STAT2 binding sites in promoter-distal regions that are presumably enhancers.

Viral RNA and DNA sensors in the cytoplasm initiate the antiviral response that leads to type I interferon (interferon α, β) expression (16). Secreted interferons bind to IFNAR 1-2 receptors through autocrine and paracrine mechanisms and activate JAK1-TYK2. This leads to phosphorylation of STAT1 and its nuclear translocation to transcribe interferon-stimulated genes.

mESCs naturally have limited antiviral response, at least partially due to the low levels of viral RNA sensors and inactive nuclear factor kappa B (18,19,20). Type 1 interferon pathway becomes mature coincident with mESC differentiation. Although intrinsic low expression of interferon-stimulated genes independent of viral infection was recently reported in mESCs, this was shown to be dependent on the TF, IRF9 (21). We did not observe a change in IRF9 transcript level in the absence of ARID4B, suggesting other mechanisms that include ARID4B in ISG regulation might be at play.

Our RNAseq results show no significant change in the level of sensor, interferon β or receptors. Instead, we observed a dramatic increase in STAT1 and interferon-stimulated gene transcript levels. This argues against an active viral response in arid4b∆ mESCs but suggests an aberrant transcriptional response due to high STAT1 expression. Interestingly, in addition to STAT1 mRNA levels, its protein, as well as active (phosphorylated) levels, are also elevated. Since JAK1 phosphorylates STAT1, this presumably points to the presence of other interferons that are secreted and can activate IFNAR-JAK receptor-kinase axis. Consistent with this, the less well studied interferon ϵ, interferon κ and interferon ω can also bind to IFNAR1/2 (22).

Our time-course RT-qPCR analysis of ARID4BΔ cells showed interferon-related genes have different expression kinetics through endoderm differentiation and in comparison to wild-type. Out of four genes tested, STAT1 and DDX58 had highest, though variable, arid4b∆/wild-type expression difference in mESC stage. IFITM3 and DDX4, on the other hand, reached highest arid4b∆/wild-type expression difference later in endoderm differentiation. These differences might reflect the roles of the genes in early versus late interferon pathway activation. As one of the main TF of the interferon pathway, STAT1 is higher in the pathway hierarchy compared to the rest of the tested genes (16). Consistent with our data, it was reported that some interferon-related genes including IFITM3 were expressed in mESCs without prior viral infection (19). Nevertheless, the expression of many interferon-related genes were altered during ESC differentiation (18,19). At the exit from pluripotency, IRF1 was shown to activate some interferon pathway genes including DDX58 (21).

Alternatively, the differences might be due to the chromatin environment of these genes. The effect of ARID4B loss in the chromatin environment might be more or less permissive for efficient transcription. Consistently, we observed a decrease in H3K27me3 level that accompanied H3K4me3 increase in arid4b∆ cells in STAT1 and DDX58. On the other hand, there was some residual, if not elevated, level of H3K27me3 in arid4b∆ cells in IFITM3, presumably dampening the transcriptional upregulation.

We found that H3K4me3 increase was associated strongly with interferon response and immune response pathway activations in arid4b∆ cells while H3K27me3 decrease or H3K27Ac increase did not lead to any related pathway enrichments. This result argues against a direct link between ARID4B and the altered chromatin environment of these pathway genes, since ARID4B is part of the RPD3L histone deacetylase complex. It is also possible that the heterogenous chromatin landscape of interferon-related genes masks direct effect of ARID4B. In fact, we found such differences in the IGV results. ARID4B-dependent histone acetylation changes might lead to a transcriptionally permissive environment that is conducive to H3K4me3 accumulation before or coincident with transcription activation. Since both histone acetylation and H3K4me3 is correlative with transcription (9), it is possible our results reflect a downstream rather than an immediate direct effect of ARID4B at these gene loci.

We observed higher antiviral response in endoderm committed arid4b∆ cells compared to mesoderm committed arid4b∆ cells. Although mesoderm and endoderm lineages originate from a common embryonic structure and thus share early progenitors, endoderm emergence occurs later during embryonic development. It is conceivable that the higher level of interferon-related gene expression in arid4b∆ cells in endoderm differentiation reflects the further maturation of an antiviral response through the differentiation timeline.

Previous reports identified the effect of histone deacetylation and STAT1 expression and transcription activity (23,24,25). Since ARID4B is part of the RPD3 histone deacetylase complex, our results are consistent with these findings and underline a possible link between STAT1 activity and mESC differentiation.

Histone acetyltransferases and deacetylases can modify proteins other than histones (26). Another connection between acetylation and an interferon pathway comes from STAT1 post-translational modifications. Along with phosphorylation, STAT1 activity can be modulated by acetylation at its DNA-binding domain (27,28,29). Acetylation follows STAT1 phosphorylation and dampens its prolonged transcriptional activity. Our data is consistent with a model where the loss of ARID4B and its deacetylase complex might extend the duration of active STAT1 bound to its target genes, resulting in high level interferon-stimulated gene expression.

Study limitations

Our results show high expression of interferon related genes in arid4b∆ mESCs and differentiating cells. We found STAT1 protein level and its phosphorylated active form to be elevated in arid4b∆ cells. To gain a more comprehensive perspective on pathway activation, other proteins involved in STAT1 phosphorylation and downstream pathway activation would need to be verified through Western blot analysis.

Conclusion

Overall, we present evidence of transcript, protein and pathway level activation of interferon response, specifically through STAT1, in ARID4B-deficient mESCs through endoderm differentiation. Furthermore, our results show that the chromatin environment of interferon related genes is altered through changes, predominantly in H3K4me3 and H3K27Ac and to a lesser extent H3K27me3. Collectively, our data points to a role of ARID4B in the suppression of innate immune response via STAT1 through H3K4me3 and H3K27Ac regulation in mESCs and during endoderm differentiation.

Ethics Committee Approval: Ethics committee approval form or institutional review board forms were not required.

Informed Consent: Informed consent approval form were not required.

Peer-review: Externally peer-reviewed.

Financial Disclosure: The author declared that this study received no financial support.