Abstract

INTRODUCTION

In order to mount effective immunity towards diverse microbial pathogens, naive CD4⁺ T helper cells adapt distinct fates, becoming T helper 1 (Th1), Th2 or Th17 cells, and thereby orchestrating effector responses (Abbas et al., 1996; Murphy and Reiner, 2002; Weaver et al., 2006). These T helper effector subtypes are defined by their signature cytokines, namely interferon-γ (IFN-γ) for Th1 cells, interleukin-4 (IL-4) and IL-13 for Th2 cells and IL-17A for Th17 cells (Zhou et al., 2009). CD4⁺ T cells can also differentiate to become regulatory T cells, which serve to constrain immune-mediated damage (Sakaguchi et al., 2008). Thus, the attainment of these cell fates is critical not only for proper elimination of pathogens, but also for avoidance of damage to the host (Reiner, 2007).

The microenvironment of the naïve CD4⁺ T cell greatly influences its specification, with a key factor being the cytokine milieu. A major means by which cytokines exert their effect is through activation of signal transducer and activator of transcription (STAT) family transcription factors, which translocate to the nucleus, bind to the regulatory region of genes and induce transcription (Levy and Darnell, 2002; Schindler and Darnell, 1995). Among the 7 STAT family members, STAT4, in conjunction with STAT1, plays pivotal role in Th1 cell differentiation by transmitting IL-12 signals to produce IFN-γ and enhancing expression of T-box 21 (T-bet) (Thieu et al., 2008). STAT6, activated by IL-4, is the key player in Th2 cell differentiation and serves to enhance expression of IL-4, IL-13 and GATA binding protein 3 (Gata3) (Zhu and Paul, 2008). Among the STATs, STAT4 and STAT6 stand out as having highly specific functions, because the phenotype of gene-targeted mice is largely confined to Th1 or Th2 cell defects respectively and the mice develop otherwise normally with normal numbers of T cells (Wurster et al., 2000).

Transcriptional profiling has helped to clarify genes whose expression is controlled by STAT4 and STAT6, and these data argue that these transcription factors are clearly important for regulating a wide array of genes (Chen et al., 2003; Hoey et al., 2003; Lund et al., 2004; Lund et al., 2007; Rogge et al., 2000). However, this technology does not permit the discrimination of genes that are directly versus indirectly regulated. Chromatin immunoprecipitation (ChIP) can identify genes bound by transcription factors, but relatively few genes have been confirmed to be occupied by STATs. In the case of STAT4, known targets include Ifng, Il18r1, Il12rb2, Il2ra and Furin (Letimier et al., 2007; O'Sullivan et al., 2004; Pesu et al., 2006; Thieu et al., 2008), whereas STAT6-bound genes include Il4, Gata3, and Ccl17 (Kubo et al., 1997; Wirnsberger et al., 2006). Therefore it was far from evident what proportion of genes are directly occupied by STAT4 or STAT6, and ultimately contribute to Th1 or Th2 cell phenotypes or whether STATs work predominantly in an indirect manner. ChIP coupled with hybridization to arrays (ChIP-on-Chip) or massive parallel sequencing (ChIP-seq) provides the opportunity to identify transcription factor binding sites on a broader scale (Good et al., 2009; Jothi et al., 2008). This has expanded our understanding of STAT-bound genes; however, relating transcription factor binding to functional consequence, i.e. transcriptional control, has been difficult (Beima et al., 2006; Good et al., 2009; Wunderlich and Mirny, 2009).

Although transcription factors like STATs play a critical role in regulating differentiation, gene expression is also dependent on the status of chromatin structure, which is regulated by DNA methlylation, ATP-dependent remodeling of nucleosomes, incorporation of histone variants, and posttranslational modification of histone tails (Bernstein et al., 2007). As STATs play a critical role in initiating a defined gene transcription program, it is of great interest to determine how the distribution of histone epigenetic marks and recruitment of STAT transcription factors coordinate to orchestrate gene regulation in differentiating T cells (Wei et al., 2009; Wilson et al., 2009). Whether histone modifications are prerequisite events for STAT recruitment or vice versa has not been resolved. In the case of the Ifng and Il4, Il13 and Il5 loci, there is evidence that epigenetic modifications and higher order structural remodeling are STAT-dependent (Bernstein et al., 2007; Lee et al., 2006; Wilson et al., 2009; Zhang and Boothby, 2006). However, other than the signature cytokine genes, there is a paucity of information on the epigenetic control of genes that participate in T helper cell differentiation and the involvement of STATs in the process (Thieu et al., 2008; Yu et al., 2007). Whether signature cytokine genes (Ifng and Il4) represent a very limited subset of STAT-dependent T helper cell genes with respect to epigenetic control or whether they are prototypical of a large number of similarly regulated genes has not been determined; only very limited segments of the genome have previously been interrogated. By contrast, recent genome-wide mapping of STAT1 binding sites in HeLa cells indicated that epigenetic modifications preceded IFNγ-triggered STAT1 binding (Robertson et al., 2008; Robertson et al., 2007). Thus, the exact interplay between transcription factors and epigenetic modifications is still an open question.

We set out to address these questions by simultaneously measuring genome-wide STAT recruitment, histone modifications and mRNA expression during differentiation of Th1 and Th2 cells in wild-type and STAT-deficient cells. By mapping the distribution of H3K4me3, H3K27me3 and H3K36me3 along with STAT4 and STAT6 binding, we determined the entire spectrum of specific STAT4- and STAT6-occupied genes and linked STAT binding to corresponding gene transcription and epigenetic modifications. For the majority of STAT-bound genes, the absence of the cognate STAT had little effect on epigenetic modifications and gene transcription. However, we identified a core subset of STAT-occupied genes whose gene expression and chromatin configuration was highly STAT-dependent. Furthermore, we found that the global effects of STAT4 and STAT6 are distinct in that STAT4 primarily promoted accessible epigenetic marks whereas STAT6 had a prominent role in antagonizing repressive marks. This analysis also allowed the identification of subset of genes for which STAT4 and STAT6 appeared to serve as repressors. Therefore, our global analysis identified wide regulatory roles for STAT4 and STAT6 in regulating epigenetic state and transcription in their cognate T helper lineages.

RESULTS

Genomic distribution of STAT binding sites and identification of consensus binding motifs

To identify genes directly bound by STAT4 or STAT6, we polarized naïve CD4⁺ T cells under Th1 or Th2 cell conditions, and re-activated the cells with anti-CD3 and anti-CD28 antibodies and the respective cytokines for 2 hours prior to ChIP-seq analysis. The appropriate polarization of Th1 and Th2 cells was confirmed by intracellular staining (data not shown) and mRNA expression, as determined by microarray analysis. A total of 7 to 8 million non-redundant reads were uniquely aligned onto mouse genome (mm9 Build 37 assembly by NCBI). We used CisGenome to identify peaks (Ji et al., 2008), and preimmune serum as a negative control to define nonspecific binding. Genome-wide, we identified 10,831 STAT4-occupied peaks and 8,434 STAT6-bound peaks in wild-type cells cultured under Th1 and Th2 cell conditions, respectively (Table S1).

To examine the comparative distribution of STAT4 and STAT6 binding sites, we divided the mouse genome into four regions: promoter (Transcriptional start site (TSS) to upstream 10 kb), exonic, intronic, and intergenic regions. As shown in Figures 1A, 31% of STAT4-occupied sites and 36% of STAT6-occupied sites localized to promoters. This represents a statistically significant (p<10⁻⁶) enrichment, as promoter regions constitute only 2% of the total mouse genome (Figure S1). In contrast, only 5% of STAT4 and STAT6 binding sites localized to introns and less than 1% of the binding sites localized to exons, (Figures 1A), with intronic and exonic sequences accounting for 37% and 0.2% of the total mouse genome (Figure S1) respectively. Notably, 63% of STAT4 binding sites and 59% of STAT6 binding sites localized to intergenic regions, where they could potentially function as distal enhancers or promoters of unannotated transcripts (Figures 1A).

Open in a separate window

Figure 1

Genomic distribution of STAT binding sites and identification of consensus binding motifs

(A, B) ChIP-seq analysis was used to map STAT4 and STAT6 binding from polarized Th1 and Th2 cells. The distribution of (A) STAT4 and STAT6 binding sites was analyzed based on location: promoter (within 10 kb upstream from the transcription start site), exon, intron, and intergenic regions. (B) Venn diagram showing the number of genes uniquely bound by STAT4, STAT6 or genes bound by both STATs. STAT-bound genes were identified if at least one peak of binding was present within the gene region defined as 10 kb upstream of TSS to the transcription end site. (C) A parallel version of MEME (Bailey and Elkan, 1994) was used to perform a de novo search of consensus binding motifs for STAT4 and STAT6 and resulted in distinctive GAA palindromes for STAT4 with 3 bp spacer and STAT6 with 4 bp spacer.

To enumerate STAT-bound genes, we defined a gene locus as a region encompassing potential promoter elements (10 kb upstream from TSS), as well as the exonic and intronic regions of the gene itself (TSS to transcription termination site). Using this definition, we identified 2,334 genes uniquely bound by STAT4 in Th1 cells and 1,984 genes selectively bound by STAT6 in Th2 cells. Interestingly, 2,152 genes were bound by both factors (Figure 1B).

Previous in vitro studies have indicated that the STAT family proteins bind to a GAA palindromic motif (Decker et al., 1997). To determine if such an element was identifiable in our genome-wide analysis, we carried out de novo searches of consensus motifs of STAT4- and STAT6-bound regions using MEME (Bailey and Elkan, 1994). We identified the consensus motif for STAT4 as a GAA palindrome with a 3-nucleotide spacer. In contrast, STAT6 preferentially bound to a palindrome with a 4 base spacer (Figures 1C). Of the total of 4486 STAT4-bound genes, 100% exhibited at least one copy of the motif that was depicted in Figure 1C. However, only 30% of 4136 STAT6-bound genes contained at least one copy of the 4-space motif that was identified in Figure 1C. Among 2152 genes that were bound by both STAT4 and STAT6, 800 genes (37%) contained both 3-space and 4-space consensus motifs.

Colocalization of STAT binding with histone modifications

Other genome-wide surveys of transcription factor binding have suggested that in many cases, binding does not necessarily correlate with regulation of gene expression, particularly in eukaryotes (Wunderlich and Mirny, 2009). The interplay between transcription factor binding and epigenetic events has been equally unclear (Robertson et al., 2008; Zheng et al., 2007). To begin to relate epigenetic changes and occupancy of STAT transcription factors, we also mapped the distribution of H3K4me3, H3K27me3 and H3K36me3 using ChIP-seq (Table S1). We found ~60% of STAT4 and STAT6 binding sites colocalized with H3K4me3 islands (Figures 2A and 2B). Only 0.3% of STAT4 binding sites colocalized with H3K27me3 islands, whereas 5% of STAT6 binding sites colocalized with this modification. Of note, under the culture conditions used, roughly 1.7 fold more total H3K27me3 islands were present in Th2 cells than in Th1 cells, whereas the numbers of total H3K4me3 islands were similar (Figures 2A and 2B). We next compiled the localization of H3K4me3, H3K27me3 and H3K36me3 marks for all genes, relating this to their transcriptional start site (TSS) and transcriptional end site (TES) to visualize tag density profile for each mark (Figures 2C and 2D). As previously reported, H3K4me3 (in red) was highly enriched around the TSS (Roh et al., 2006). STAT4 and STAT6 binding sites (in blue) were also enriched around TSS; however, the precise peak of STAT binding coincided with a valley of H3K4me3 binding (Figures 2C and 2D lower panels). This suggests that maximal STAT binding likely occurs in a relatively nucleosome-poor region in the midst of abundant H3K4me3 modification encompassing several nucleosomes. Of note, H3K36me3 modifications (in purple) increased along gene bodies and peaked at the transcription end sites (Figures 2C and 2D upper panels), whereas H3K27me3 marks (in green) were relatively depleted in regions in which STATs were bound (Figures 2C and 2D lower panels).

Open in a separate window

Figure 2

Global STAT binding and epigenetic modifications

(A, B) ChIP-seq was performed to map histone epigenetic modification in Th1 cells and Th2 cells. The concordance between H3K4me3, H3K27me3 modifications and (A) STAT4 binding sites in Th1 cells and (B) STAT6 binding sites in Th2 cells is shown. (C, D) Compiled tag density profiles for H3K4me3, H3K27me3, H3K36me3 and (C) STAT4 in Th1 cells and (D) STAT6 in Th2 cells are shown. The diagrams represent all genes that showed positive signals for the respective marks. Upper panel: tag density profile across gene body ± 5 kb flanking regions with 200 bp resolution. Lower panel: tag density profile across promoter ± 5 kb flanking regions with 25 bp resolution.

ChIP-seq provides a comprehensive view of STAT-occupied genes that can participate in helper cell function

The ChIP-seq data set generated allowed us to assess global quantitative views of transcription factor binding and STAT-dependent epigenetic modifications, but its precision and resolution also provided a refined view at the gene level. Therefore, we next evaluated localized STAT-dependent epigenetic modification of STAT-occupied genes that contribute to the Th phenotype. Figure 3 depicts a compilation of browser views of selected genes showing relevant marks (STAT binding, H3K4me3 and H3K36me3) in wild type and STAT-deficient cells. The complete list of genes occupied by STAT4 and STAT6 are provided (Table S2) and genome-wide STAT binding and STAT-dependent epigenetic modifications can be visualized via the genome browser; http://www.niams.nih.gov/Research/Ongoing_Research/Branch_Lab/Molecular_Immunology_and_Inflammation/Supplemental/OShea/KannoImmunity10/

Open in a separate window

Figure 3

STAT-bound genes represent multiple aspects of Th differentiation

ChIP-seq signal profile maps are shown as a genome browser view. (A) Th1 cell genes, Ifng (chr10:117,875,074-117,885,977), Tbx21 (chr11:96,947,000-96,996,774), Lag3 (chr6:124,848,219-124,863,009), Zbtb32 (chr7:31,374,500-31,385,000), Il21 (chr3:37,119,708-37,136,092), (B) Th2 cell genes, Il4 (chr11:53,418,425-53,444,775), Gata3 (chr2:9,770,098-9,802,379), Il24 (chr1:132,778,117-132,786,123), Plcd1 (chr9:118,980,999-119,011,401) and Hipk2 (chr6:38,641,845-38,850,017). STAT4 and STAT6 binding in wild-type is shown in red upward peaks. Epigenetic marks (H3K4me3, H3K36me3) in wild-type cells are depicted as red upward peaks whereas the accompanying blue colored downward peaks depict the binding in STAT4-deficient cells. Where no signal was detected, the corresponding columns are blank.

Among selected Th1 cell genes shown in Figure 3A, the Ifng is well-known to be regulated by STAT4 (Xu et al., 1996). In addition to STAT4 binding in the promoter, we identified multiple peaks across the extended Ifng locus (Figure S2). Consistent with previous work, (Wilson et al., 2009), permissive H3K4me3 and H3K36me3 modifications were identified throughout the entire Ifng locus and nearly all permissive marks were STAT4-dependent (Figure S2). The Tbx21 gene encodes the key transcription factor T-bet, which is regulated by both IFN-γ-STAT1 and IL-12-STAT4 (Afkarian et al., 2002; Lighvani et al., 2001; Thieu et al., 2008). STAT4 binding localized to an upstream region, which included a previously described enhancer at −13 kb (Yang et al., 2007). Of note, H3K4me3 and H3K36me3 marks throughout the Tbx21 gene were highly STAT4-dependent. We found that STAT4 bound many other Th1 cell-expressed genes and influence their permissive epigenetic marks (see below).

A key feature of lineage commitment is the repression of alternative fates in concert with promotion of the desired lineage. Zbtb32, which encodes repressor of GATA3, is a factor that attenuates Th2 cell responses (Hirahara et al., 2008; Miaw et al., 2000). It was also a STAT4-occupied gene that exhibited STAT4-dependent epigenetic modifications. Thus, STAT4 not only directly regulates Ifng itself, but also regulates transcription factors that promote Th1 cell differentiation and repressors that inhibit Th2 cell differentiation (Figure 3A). Specific STAT4 binding of selected genes was confirmed by ChIP-qPCR (Figure S3, Table S3). By contrast, a number of other genes preferentially expressed in Th1 cells were not directly bound by STAT4 (e.g. Cdh3, Cxcr3, Etv5, and Cebpb) (Rogge et al., 2000).

IL-21-producing follicular helper T cells have recently been recognized as a new subset of T cells; however, their origins are complex (Fazilleau et al., 2009; King et al., 2008). They can arise separately as a lineage, but can also be generated from other polarized subsets. Recently, it has been shown that IL-12 can promote the generation of human Tfh cell (Schmitt et al., 2009). We found that the Il21 gene was a very prominent STAT4-bound gene whose epigenetic regulation was strongly STAT4-dependent (Figure 3A). These data argue that STAT4 is a factor that can regulate Th1 cell differentiation but can also contribute to Tfh cell differentiation.

Previous studies have argued that STAT6 directly contributes to the regulation of the Il4 and Gata3 genes; although limited regions of the genes were analyzed (Ansel et al., 2006; Pai et al., 2003; Zheng and Flavell, 1997). As shown in Figure 3B, STAT6 bound to both the Il4 and Gata3 loci, which were associated with strong permissive epigenetic modifications. Other STAT6-bound genes whose permissive epigenetic modifications showed dependency on STAT6 included: Il24 (Schaefer et al., 2001; Wang and Liang, 2005), Plcd1 and Hipk2 (Chen et al., 2003) (Figure 3B).

Differential patterns of epigenetic regulation by STAT4 and STAT6

Having established that the epigenetic modifications of many Th cell-defining genes are highly STAT-dependent, we sought to evaluate the impact of the cognate STAT on the global epigenetic patterns of occupied genes. To this end, we calculated the total tag counts of epigenetic modifications (K4me3, K27me3 and K36me3) encompassing each occupied gene. Using the K-means algorithm, we clustered STAT-bound genes based on the ratio of tag counts in wild-type and STAT-deficient cells for each of three epigenetic marks (Figure 4, Table 4). Based on this analysis, six epigenetic patterns for STAT4-occupied genes emerged (Figure 4A). For the largest cluster (75% of the total STAT4 bound genes), the absence of STAT4 had minimal effects on epigenetic modifications, suggesting that other factors were the major regulators (denoted as “indeterminate pattern”). Five other clusters of genes were also apparent. We identified clusters with high amounts of H3K4me3 (denoted “K4 high”, 5.7%) and H3K36me3 (“K36 high”, 4.3%) that were clearly STAT4-dependent (Figure 4A). In addition, the third cluster showed STAT-dependent inhibition of H3K27me3 marks (“K27 low”, 7.0%). Analysis of STAT6-occupied genes revealed similar but not identical clusters (Figure 4B). Like STAT4, roughly 25% of STAT-occupied genes showed strong dependency on STAT6 for their distinct epigenetic patterns. However, a discrete cluster of genes in which STAT6 induced high levels of H3K4me3 was not apparent. Although analysis of selected genes showed that STAT6 can positively regulate H3K4me3 (Figure 3B), these genes did not cluster as a discrete group. Rather, the proportion of genes for which STAT6 reduced H3K27me3 was greatly expanded compared to the corresponding cluster of STAT4-occupied genes (11.8% vs. 7.0%).

Open in a separate window

Figure 4

STAT-bound genes form clusters that share common epigenetic signatures

(A, B) The total tag count of each epigenetic modification (H3K4me3, H3K27me3, and H3K36me3) was computed across the body of each STAT-occupied gene in wild-type versus STAT-deficient cells. The ratio of tag counts for three epigenetic modifications was used to cluster STAT-occupied genes by applying the K-means clustering method with squared Euclidean distance with 1000 iterations. (A) STAT4-bound genes in Th1 cells clustered in 6 patterns based on the ratio of wild type vs. STAT4-deficient cells: H3K4me3(K4)-high, H3K36me3 (K36)-high, H3K27me3 (K27)-low, H3K27me3 (K27)-high, H3K36me3 (K36)-low, and an indeterminate pattern. (B) STAT6-bound genes in Th2 cells showed 5 distinct epigenetic patterns similar to A without the K4-high cluster. The accompanying tables list the number and percentage of genes in each cluster.

To confirm the clustering analysis, we plotted the global distribution of epigenetic marks on STAT-occupied genes in wild type and STAT-deficient cells. In the absence of STAT4, H3K4me3 modifications on STAT-bound genes were markedly reduced (Figure 5A upper panels), but as expected from the previous analysis, the absence of STAT6 did not affect the global pattern of H3K4me3 marks on the bound genes. The global distribution of H3K36me3 modifications (Figures 5A and B middle panels) showed minimum changes in the absence of STAT4 or STAT6. In contrast, STAT4 and STAT6 showed notably different influences on H3K27me3 marks. Although deficiency of STAT4 had little effect on the global pattern of H3K27me3 marks, the absence of STAT6 resulted in a significant, global increase of H3K27me3 marks (p= 8E-42, Student T-test) (Figure 5B lower panels). Taken together, our results suggested that STAT4 and STAT6 differentially affected distinct epigenetic modifications of their cognate genes. Specifically, STAT4 primarily promoted H3K4me3 marks whereas STAT6 predominantly acted to antagonize H3K27me3 modifications of its bound genes.

Open in a separate window

Figure 5

Differential regulation of epigenetic marks by STAT4 and STAT6 on STAT-bound genes

(A, B) Compiled tag density profiles for histone modification marks (H3K4me3, H3K36me3, H3K27me3) were calculated from all STAT4-bound genes in Th1 cells (A) and STAT6-bound genes in Th2 cells (B) in wild type cells (red) and in STAT-deficient cells (blue). The tag density profiles across promoter ± 5 kb flanking regions are shown for H3K4me3 (top panels) and H3K27me3 (bottom panels). The tag density profiles across gene body ± 5 kb flanking regions for H3K36me3 are shown in middle panels.

Correlations between STAT-dependent epigenetic regulation and gene expression

Having identified distinct clusters of STAT-dependent epigenetic modifications, we next assessed how these patterns correlated with regulation of gene expression. Transcriptional profiling was performed in Th1 and Th2 cells (Table S5) to identify STAT-dependent genes (≥2-fold difference in expression between wild type versus STAT-deficient cells). Among them, clusters of genes that exhibited strong STAT-dependent H3K4me3 or K36me3 marks were overwhelmingly positively regulated by STAT in terms of mRNA expression (Figure 6). Genes assigned to these clusters for STAT4 include Ifng, Tbx21, and Il21. Interestingly for STAT6, the gene cluster for which this factor inhibited H3K27me3 marks included the major lineage-specifying genes such as Il4, Gata3, Il24, and Il4ra. The entire lists of genes included in each gene cluster are presented in Table S4 (also see Figure S4 for summary). For both STAT4 and STAT6, STAT-dependent positive epigenetic changes (K4 high, K36 high, K27 low) correlated well with the STAT-dependent positive gene expression (Figure 6).

Open in a separate window

Figure 6

STAT-dependent epigenetic signature correlates with gene expression

(A, B) STAT-dependent gene expression change was evaluated by microarray for each epigenetic cluster as described in Figure 4. Two-fold change was used as a cut off to sort genes into positively regulated (red bars) and those negatively regulated (blue bars) by STAT4 (A) and by STAT6 (B). Genes whose expression was not affected by STAT4 or STAT6 were not depicted in the graphs.

By comparing STAT4 and STAT6 regulated genes, we were particularly interested in a group of genes bound by both factors and how STATs regulated their function in their cognate lineages. For 5% of the genes bound by both STATs, both factors served to positively regulate these genes in the respective subsets. Examples of these genes include: Il10 (Saraiva et al., 2009), Il7r (Kallies, 2008), Socs3 (Dimitriou et al., 2008), and Id2 (Kusunoki et al., 2003). As shown in Figure 7A, permissive epigenetic marks on those genes, as well as gene expression, were dependent upon the cognate STAT.

Open in a separate window

Figure 7

STAT4 and STAT6 work in concert and in opposition to influence gene expression

(A) ChIP-seq profiles for binding of STAT4 and STAT6, as well as epigenetic modifications in Th1 and Th2 cells. Illustrative genes include: Il10 (chr1:132,911,300-132,926,671), Il7r (chr15:9,430,427-9,470,839), Socs3 (chr11:117,821,153-117,836,725) and Id2 (chr12:25,769,886-25,787,125). (B, C) ChIP-seq data illustrating STAT binding and epigenetic modifications of (B) Ccr8 (chr 9:119,988,491-120,013,456) and (C) Il18r1-Il18rap (chr1:40,521,609-40,620,416).

DISCUSSION

In this study, we examined the landscape of transcription factor binding of STAT4 and STAT6 through genome-wide approaches and identified the genes uniquely bound by each STAT, as well as those occupied by both factors. Our analysis revealed a core subset of STAT-occupied genes, whose expression and local epigenetic fingerprint are clearly dependent upon the presence of the given STAT. Using this strategy, a large number of genes were identified, many of which were not previously known to be directly regulated by STATs. Consequently, the direct functional roles of STAT4 and STAT6 are considerably broader than previously appreciated. Furthermore, we found that the two STATs differ in their ability to influence global epigenetic profiles. On the level of each bound gene, STATs induce a variety of localized positive or negative epigenetic patterns. However genome-wide, promoting H3K4me3 modifications is a more prominent global feature of STAT4, whereas antagonizing repressive marks (H3K27me3) is a major property of STAT6. Our analysis also uncovered a number of genes for which STAT4 and STAT6 appear to serve as repressors. Among them, a considerable proportion were bound by both factors, and regulated in an opposing manner.

With the ability to measure transcription factor binding on a genome-wide scale, a fundamental issue is the functional relevance of factor binding and how to evaluate it. In this study, we identified over 4000 genes bound by the cognate STAT in Th1 and Th2 conditions. Of these, 15% (684 genes for STAT4) to 29% (1200 genes for STAT6) showed STAT-induced expression changes. An emerging concept coming from genome-scale studies is that only a fraction of factor-bound genes showed clear functional dependence as evaluated by gene expression changes. For example, a recent study enumerating GATA-1 binding showed that ~40% of genes (790 genes out of 1800 bound genes) showed GATA-1 induced expression changes (Yu et al., 2009). Our study produced similar absolute numbers of functionally relevant transcription factor-bound genes when evaluated by microarray-based expression changes.

By adding analysis of STAT-induced epigenetic changes to gene expression changes, we were able to further separate highly relevant STAT binding that substantially impacted both epigenetic signature and gene transcription, from binding events that had minimal functional outcomes. Our analysis revealed subsets of gene clusters whose epigenetic signature was highly STAT dependent. These genes comprise roughly 1000 genes of total 4000 STAT-occupied genes. Our data firmly establish that there is a substantial subset of genes for which the recruitment of STAT during the course of T helper cell differentiation serves to maintain the distinctive epigenetic pattern of the genomic region. For these STAT-regulated genes, it will be of considerable interest to carefully dissect the kinetics of epigenetic pattern formation during the process of lineage specification.

Also of interest was the identity of the STAT-bound genes whose transcriptional and epigenetic regulation was highly dependent on the factor. The genes identified include phenotype-defining cytokines (Ifng, Il4 and Il24), receptors (Il18rap, Il18r1, Lag3, Il4ra), transcription factors (Gata3, Tbx21) and transcriptional repressor (Id2 and Zbtb32). In addition to these recognized regulators of T helper cell differentiation, we identified a number of previously unrecognized STAT-regulated genes with preferential expression in Th1 and Th2 cells. These include genes that encode transcriptional repressors (Ikzf3 [Aiolos] and Nfil3), and diverse signaling molecules that included kinases, phosphatases, G-proteins and adaptors (Hipk2, Plcd1, Gbp2 and Skap2). Clearly, it will be of interest to define the role of these molecules in T cells and assess how their actions relate to the specification of cognate lineages.

Tfh cells represent a recently recognized subset of helper T cells and we were intrigued to see that a major gene expressed by Tfh cells, Il21, was also a STAT4-bound and positively regulated gene. Although previous data has argued that IL-6 acting via STAT3 is the main driver of Tfh cell differentiation (Fazilleau et al., 2009; King et al., 2008), recent data indicates that in human T cells, IL-12 can also promote Tfh cell specification (Saraiva et al., 2009). Our data are entirely consistent with this result and suggest that this mechanism is not unique to human T cells. Interestingly, we also found that STAT4 bound the Bcl6 gene, but unlike IL-21, the expression of the former was modest in Th1 cell polarizing conditions. This would argue that IL-12-STAT4 signals alone are not sufficient to drive Bcl6 expression and presumably other signals are required. The data are also consistent with the evolving notions that Tfh cells represent a flexible subset. In related studies, we found that the Bcl6 and Il21 genes are the targets of STAT3 as well (Durant et al., 2010) and it will be important to differentiate the roles of different STATs in the transcriptional and epigenetic regulation of these important genes.

An unanticipated finding in our studies became apparent when we compared and contrasted the effects of STAT4 and STAT6 on epigenetic modification of their cognate genes. We found that STAT4 primarily promoted accessible epigenetic marks whereas STAT6 had a more prominent role in antagonizing a repressive mark on its cognate genes. One possibility was that the expression of chromatin modifying enzymes differed in Th1 and Th2 cells such that the balance is shifted toward promoting H3K27me3 modification in the absence of STAT6 in Th2 cell condition. The enhancement of H3K27me3 modifications might be explained by over-expression of the H3K27 methyltransferase, Ezh2, and down-regulation of the H3K27 demethylase, Jmjd3. However, this was not the case. We noted decreased expression of Ezh2 associated with a reciprocal increase of Jmjd3 in STAT6-deficient Th2 cells. Therefore, an alteration in the balance between the “writer” (Ezh2) and “eraser” (Jmjd3) (Wang et al., 2007) was evidently not the primary cause of increased H3K27me3 mark associated with the absence of STAT6. On the contrary, these enzymes appear to be compensating for STAT6-dependent alteration of H3K27me3 in Th2 cells. At present, we do not have an explanation for these findings, but clearly additional studies are warranted to dissect the molecular mechanisms underlying the differential regulation of H3K27me3 modifications in Th1 and Th2 cells.

The present study also revealed a large number of genes that are bound and negatively regulated by STAT4 and STAT6. Although the ability to concomitantly promote the expression of some genes and repress others is a common feature of transcription factors associated with lineage commitment (Cobaleda et al., 2007), there is a paucity of circumstances in which STATs have been documented to behave as functional repressors. This was a more prominent feature of STAT4 than STAT6. Clearly, it will be of considerable interest to carefully dissect the molecular basis of the repression of these genes. Particularly interesting were genes bound by both STAT4 and STAT6, for which these factors had opposing functions. In this case, the two STATs appear to function in concert to ensure differential expression of lineage-specifying genes (e.g. Il18r1-l18rap). These will be an interesting subset of genes to analyze in further detail.

In summary, our data provide a platform for understanding how STATs function to modulate epigenetic events to shape gene expression profile unique to T helper subtype. The emerging picture is that STATs bind to a broad, yet clearly defined set of genes and contribute in a very substantial manner to helper T cell differentiation. They do so in both a positive and negative manner. Investigating the function of newly identified genes as they relate to helper cell function and elucidating the mechanisms by which STATs regulate the expression, both positively and negatively, will clearly be important.

METHODS

Supplementary Material

ACKNOWLEDGEMENTS

Footnotes

References