Genomic views of STAT function in CD4⁺ T helper cell differentiation: new technology brings new insights and new questions

Abstract

Signal transducer and activator of transcription (STAT) proteins are well known for their essential roles in transmitting cytokine-mediated signals and specifying T helper (T_H) cell differentiation; however, recent technological advances have led to a new level of understanding of transcription factor action. This work has revealed that STAT proteins have broad and complex roles in gene regulation and epigenetic control, including important roles as functional repressors. However, the challenge remains how to link signal transduction, nucleosome biology and gene regulation. The relevance of tackling this problem is highlighted by genome-wide association studies that link cytokine signalling and STATs to a variety of autoimmune or immune deficiency disorders. Defining exactly how extrinsic signals control the specification and plasticity of T_H cells will provide important insights and perhaps therapeutic opportunities in these diseases.

CD4⁺ T cells are essential for host defence, as exemplified by the effects of depletion of CD4⁺ T cells associated with HIV/AIDS. This loss of T cells leads to profound impairment of the immune response and a range of opportunistic infections. Conversely, CD4⁺ T cells are also fundamental drivers of autoimmunity when a loss of tolerance occurs. CD4⁺ T cells mainly direct immune responses through the cytokines they produce, and our understanding of the range of cytokines produced by CD4⁺ T cells has become considerably expanded than before¹. In addition to T helper 1 (T_H1) and T_H2 cells, which produce interferon-γ (IFNγ) and interleukin-4 (IL-4), respectively, new subsets of T cells continue to be recognized. These include regulatory T (T_Reg) cells², which comprise both natural and induced T_Reg cells, IL-17-producing T_H17 cells^2–4, IL-9-producing T_H9 cells^5–7 and a subset of IL-22-producing T_H cells^{7, 8}. IL-21-producing follicular helper T (T_FH) cells provide help to B cells, but their identity as a distinct lineage and relationship to other CD4⁺ T cell subsets remain a source of some controversy^{9, 10}.

What is clear is that the cytokine milieu is crucial for CD4⁺ T cell differentiation. Signal transducer and activator of transcription (STAT) family proteins have essential roles in transmitting many cytokine-mediated signals and thereby have similarly crucial roles in T_H cell differentiation^{1, 11}. The first STAT proteins (STAT1 and STAT2) were discovered as a mediators of interferon (IFN) action to induce de novo gene transcription^{12, 13}; the essential, non-redundant functions of the seven members of the STAT family have been extensively defined by generating individual gene knockout mice and by careful analysis of the effects on gene expression^14–18. One of the challenges to interpreting such gene expression data is to distinguish direct actions of STATs on individual genes from secondary, indirect effects of STAT deficiency. Recent technologies have enabled investigators to construct a genome-wide view of transcription factor binding to distinguish direct from indirect effects.

In this Review, we discuss the impact of next generation sequencing^{19, 20}, and illustrate how this technology has allowed us to begin to construct a quantitative map of not only of genome-wide transcription factor binding, but also of their effect on genome-wide epigenetic changes. Specifically, we review the genome-wide STAT binding studies that have been reported so far. We discuss the relationship between STAT binding and local epigenetic patterns and how STAT proteins can integrate extrinsic signals to influence epigenetic changes associated with T cell lineage commitment. Finally, we review emerging new information regarding mutations and polymorphisms of STAT proteins that are associated with human immune disorders.

Genome-wide views offered by next generation sequencing

STATs are DNA-binding transcription factors that induce the transcription of their target genes by recognizing specific DNA consensus sequences. Analysis of direct STAT binding to DNA was initially analyzed by electrophoretic mobility shift assay (EMSA). Later, analysis of binding to specific genes was carried out by chromatin immunoprecipitation followed by the detection of precipitated DNA by PCR using primers against pre-selected regions. This type of targeted analysis inevitably limited its application to a small subset of genes and regions. However with the arrival of next generation sequencing methods, an unbiased genome-wide view of protein–DNA association has become a reality (Box 1 and Fig. 1), allowing us to catalogue the entire range of STAT target genes on a genome-wide scale. Equally important has been the capability to map histone epigenetic marks throughout the entire genome to gain insight into how chromatin accessibility relates to STAT binding and ultimately to transcriptional regulation.

Text box 1. Genome-wide chromatin immunoprecipitation (ChIP) to study protein–DNA interactions.

Chromatin immunoprecipitation (ChIP) has been used to profile DNA–protein interactions. By choosing appropriate antibodies specific for the protein or epigenetic modification of interest, both transcription factor binding and histone epigenetic marks can now be profiled on a genome-wide scale. After protein-associating DNA fragments are enriched and purified through immunoprecipitation, the DNA fragments can be measured and mapped to reference genomes by either hybridization (ChIP -on chip) or high-throughput next generation sequencing (ChIP-seq).

ChIP -on chip, which is based on microarray hybridization technology^{133, 134}, has the intrinsic limitation that only pre-selected regions of the genome are included in the arrays, such as proximal gene promoter regions. Also, array-based methods are restricted by the variation and limitation implicit in nucleotide hybridization. By contrast, chromatin immunoprecipitation coupled with next generation sequencing methods (ChIP-seq) covers the entire genome without any preconceived bias^{19, 20, 135}. Because the DNA fragments of interest are sequenced directly instead of being hybridized to microarray chips, ChIP-seq provides higher resolution, greater genomic coverage, fewer artifacts and a larger dynamic range of signal strength than ChIP-on chip. Although the relatively short reads (35–75 bp) generated by various next generation sequencing platforms could pose technical difficulties for certain other applications such as RNA-seq, the technology is well-suited for a ChIP-seq approach ¹³⁶.

In addition to mapping trascription factor binding and histone epigenetic marks, Chip-seq has also been applied to map the binding of CCCTC-binding factor (CTCF), which regulates chromatin architecture, and p300, which marks enhancer elements²⁷. This next generation sequencing platform is also used to define nucleosome positioning and accessibility by coupling with micorococcal nuclease digestion (MNase-seq) ¹³⁷ or with the detection of DNAse hypersensitivity sites (DNase-seq)^{138, 139}. Next generation sequencing can also be used to generate comprehensive methylome mapping ¹⁴⁰, which will provide yet more insights on stable and heritable aspect of epigenome. Next generation sequencing has also been used to profile various types of RNA, such as microRNAs (miRNAs)^{141, 142}, long noncoding intervening RNAs¹⁴³ and enhancer RNAs¹⁴⁴.

Figure 1. Experimental flow of ChIP-seq analysis.

A technique to combine chromatin immunoprecipitation with next generation sequencing to map DNA–protein interactions across the whole genome is shown. Chemically cross-linked DNA-protein complexes are immunoprecipitated and the protein-bound DNA fragments are isolated. The crosslinks are reversed, and the purified DNA was used to generate a library for sequencing. Automated reactions yield 36 nt long sequence reads of over 20 million per sample (Illumina GA platform). The sequence reads are aligned onto the reference genome and the distribution of protein-DNA interaction sites are visualized as “peaks” on the genome browser.

Evolving views of epigenome

It has become clear since the original discovery of the STATs that in addition to transcription factor binding, a crucial part of gene regulation is epigenetic regulation and that the modifications comprised by this term are highly dynamic²¹. It is beyond the scope of the present Review to comprehensively discuss this incredibly dynamic field; but briefly, factors that influence the accessibility of chromatin for active transcription include DNA methylation, ATP-dependent nucleosome remodeling and a large number of post translational histone modifications. Together, these modifications, in addition to classic transcription factors, have major effects on gene expression ^22–24. Early comprehensive genome-wide histone-modification maps generated by ChIP-seq suggested novel functions for histone modifications and showed the importance of combinatorial patterns of modifications ^{25, 26}. Although acetylation is always associated with active chromatin regions, the functional significance of histone methylation is more complex with respect to gene expression. For example, histone 3 lysine 4 (H3K4), H3K36 and H3K79 trimethylation are associated with active genes (‘permissive’ marks), whereas H3K27, H3K9 and H4K20 di/trimethylation are linked to gene silencing (‘repressive’ marks). Importantly, in contrast to classic views of the epigenome, it is now clear that some of these modifications can occur rapidly in response to exogenous signals^27–29. Therefore, the nucleosome is increasingly viewed as a nuclear sensor that responds to various signals from the cellular environment.

Cytokine signaling and the T cell epigenome

The ability to measure genome-wide changes in histone modifications by ChIP-seq provided an opportunity to ask a simple but crucial question about T cell biology, namely whether the observed epigenetic modifications in T cells are more consistent with a model of stable terminal differentiation of CD4⁺ T cells or intrinsic flexibility in T cell responses ^{6, 7}. The stability of various T cell subsets continues to be intensively debated, with remarkable examples of T cell plasticity appearing in the literature^{30, 31 Rubtsov, #320}, measuring the epigenetic landscape of T_H cells has proven to be illuminating.

Genome-wide H3K4 (permissive) and H3K27 (repressive) trimethylation maps in naïve CD4⁺ T cells and fully polarized T_H1, T_H2, T_H17, induced T_Reg and natural T_Reg cells have now been obtained ³². The data show that the histone modifications of genes encoding T_H cell subset signature cytokines are consistent with the signature of terminal commitment, such that permissive marks on a particular cytokine gene are selectively present in the relevant lineage that expresses that cytokine and repressive marks are present in other CD4⁺ T cell lineages that do not express the cytokine. However, genes encoding “master regulator” transcription factors, such as Tbx21 for T_H1 cells and Gata3 for T_H2 cells, were found to have “bivalent poised domains”, meaning that both permissive and repressive histone marks are present on these genes in the lineages of alternative fates³². Originally identified in stem cells, bivalent domains seem to allow for flexibility in expression once the cells receive signals for a differentiation ^{33, 34}. Bivalent domains were also present on genes encoding other key transcription factors including Runx3, Bcl6 and Blimp1 ³²,. Thus, the answer to the question initially posed is that epigenetic analysis provided evidences for both terminal commitment (e.g. cytokine genes) and flexible plasticity (e.g. master regulators genes) of T helper cells depending which genes are examined. As such, the extent to which T cells subsets really behave as “lineages” or behave as flexible populations will continue be the focus of ongoing research and controversy.

Although many interesting observations have arisen from this genome-wide epigenetic profiling, many other questions remain. First, the cell preparation used for these studies was generated in vitro and characterized at a single time point. The dynamic nature of chromatin remodeling/modification over the course of T cell differentiation is yet to be fully elucidated by genome-wide assays. It will be interesting to know whether the bivalent marks noted on the master transcriptional regulators are already in place in the very early phases of T_H cell differentiation to guide the transcriptional programme or if these marks gradually evolve over time. Equally, it will be crucial to determine how the recruitment of STAT proteins affects the deposition/removal of epigenetic marks, and how all the aspects of nucleosome remodeling are acquired over time. We also do not yet know the degree of similarity between in vitro-generated cells and bona fide T_H cells that arise in vivo during the course of infection or autoimmunity in mice and in humans.

Profiling of genome-wide STAT binding

Initial work using Chip-seq to map STAT1 binding sites in the genome revealed more than 11,000 sites in unstimulated HeLa cells and 40,000 sites after IFN-γ stimulation ³⁵. However, it was not clear from these data whether STAT1 is an important initiator of gene regulation in all cases of binding or whether STAT1 has a major role in creating the local epigenetic patterns around these binding sites. Subsequent work showed that for most genes, deposition of the local histone modification has preceded the recruitment of ligand-induced STAT1 binding³⁶. While this study was an important breakthrough, it will be important to link STAT1 action in various primary cells and to link transcriptional and epigenetic changes using STAT1-deficient cells. (This point is discussed further with additional ChIP-seq data for STAT binding below). Since these initial reports, all STATs, with the exception of STAT2, have been profiled by ChIP-seq and the original datasets are publically available through Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo/) as shown in Table 1. These datasets will serve as enormous resources to promote further genomic research in the scientific community, and deposition of original datasets to publicly accessible domains such as GEO will be key. On the other hand, it will be important to keep in mind the degree of compatibility among different datasets, which have been generated by different sequencing platforms and by different investigators under different experimental conditions could impose certain limitation over comparable analysis. As the field matures, we await a better ways to “quality control” these sequencing dataset to allow broader across-the-board analyses including the utilization of appropriate reference controls to score ChIP-seq peaks³⁷. Nevertheless, from the STAT binding data listed in Table 1, we can start to address the unique as well as shared functions of individual STAT proteins in directing epigenetic modifications and gene expression in T cells. For the sake of briefness, we will only discuss genome wide STAT data in Table 1 that are derived from T cells with genome scale analysis provided.

Table 1.

Summary of ChIP-seq data reported for STAT family proteins

	Stimuli	Cell-type	Species	GEO Accession number	Reference
STAT1	IFNg	HeLa	human	GSE15353	35
	IFNg	HeLa	human	GSE12783	37, 145
STAT2				N/A
STAT3	IL-6	TH17 polarized T cells	mouse	GSE21670	76
	IL-21	CD4+ T cells	mouse	GSE19198	89
	LIF?	ESC cells	mouse	GSE11431	146
STAT4	IL-12	TH1 polarized T cell	mouse	GSE22105	42
STAT5(a,b)	IL-2?	CD4+ T cell	mouse	GSE12346	68
STAT6	IL-4	TH2 polarized T cell	mouse	GSE22105	42
	IL-4	TH2 polarized T cell	human	GSE18017	59

IFN, interferon; IL, interleukin; T_H, T helper

All of the original data listed in the table are accessible through Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo/) by the corresponding GEO accession number.

STAT4 in T_H1 cells

Landscape of STAT4 targets

Unlike other STATs, the expression of which is found in a wide range of cell types, the expression of STAT4 is relatively restricted to immune cells and testis. Accordingly its nonredundant functions are manifested mainly in immune cells, resulting in a very discrete phenotype of STAT4 deficiency involving decreased IFNγ production³⁸. STAT4 is activated mainly by IL-12, IL-23 and type I IFNs, and it functions predominantly in promoting T_H1 cell differentiation. STAT4 is also the major regulator of Ifng expression in innate immune cells such as NK cells^{39, 40}.

Before the advent of genomic approaches, only a limited number of direct STAT4 target genes had been identified (such as Ifng, Il18r1, Hlx, Map3k8 and Furin). The first effort to increase our knowledge of STAT4 targets was through the use of ChIP-on-chip technology, which showed that STAT4 bound the promoters of many previously unidentified targets such as Gadd45g, Lcp2 and Myd88 ⁴¹. The expanded list of STAT4 target genes also showed that all genomic STAT4-binding events are not equal and that in some cases the binding of STAT4 to a target gene was not translated into a change in gene expression as a result of IL-12 stimulation. Binding per se was not the only determinant of STAT4-dependent gene programming during T_H1 cell differentiation.

Whereas the analysis provided by ChIP-on-chip technology is limited to predefined regions of the genome, ChIP-seq data generate an unbiased genome-wide map of where STATs bind. Using ChIP-seq ⁴², STAT4 was found to have 10,000 binding sites in in vitro-differentiated murine T_H1 cells, 40% of which were localized to the promoters or gene bodies of approximately 4,000 annotated genes. 60% of the STAT4-binding sites occurred in intergenic regions, where some of the distal enhancer elements are thought to reside away from annotated genes. In sharp contrast to the implications of the STAT1 data described above, comparative epigenomic analysis of wild-type versus STAT4-deficient T_H1 cells provided evidence that of the ~4,000 genes bound by STAT4, nearly 1,000 of these genes showed STAT4-dependent alterations in epigenetic modifications. Of these 1,000 genes, 200 genes showed highly STAT4-dependent gene expression as determined by microarray analysis of wild-type versus STAT4-deficient cells. These genes therefore represent a core subset of direct STAT4 targets, which are highly dependent on STAT4 for promoting gene expression and the local epigenetic signature. Importantly their dependence on STAT4 cannot be compensated for by other STAT proteins or transcription factors. The subset not only included signature T_H1 cell genes such as Ifng and Tbox21 but also others including Il18rb, Icos, Lilrb4 and Nkg7. This implies a potential role for these genes in maintaining the phenotype of fully polarized T_H1 cells, which may be of interest to examine in the future.

The analysis of STAT4 target genes also showed that some cytokine genes previously considered to define other T_H cell lineages were STAT4 targets in T_H1 cells. Initially denoted as a T_H2 cell gene in mice, subsequent work has shown that IL-10 is expressed by multiple T cell subsets⁴³. It was interesting to note that the Il10 gene was bound and positively regulated by STAT4 in T_H1 cells, but also by STAT6 in T_H2 cells. Initially noted to be a product of T cells following TCR stimulation, IL-21 was later reported to be produced by T_H17 cells in a STAT3-dependent manner ^44–47. More recently IL-21 has also been reported as a lineage-defining cytokine for T_FH cells ¹⁰. However, Chip-seq data indicated that STAT4 can bind and regulate the Il21 gene, which is consistent with the recent finding that IL-12 (acting via STAT4) can induce the expression of IL-21 in human T cells ⁴⁸. Thus, IL-10 and IL-21 are two examples of genes that can be regulated by multiple STATs.

STAT4 as a transcriptional repressor

Generally, transcription factors that drive lineage commitment positively regulate the expression of phenotype-defining genes, but they can also repress the expression of genes associated with alternative fates. Although STATs were originally discovered as activators of gene transcription, a point that has been well confirmed by genome-wide analysis, there have been indications that STATs can also function as functional repressors. Microarray data provided evidence of genes whose expression was increased when a given STAT was deleted ^{49, 50}, but there were few examples of genes for which STATs seemed to function as direct transcriptional repressors ^{51, 52}. In this regard, several possible ways for a STAT protein to cause gene silencing were reported that include recruitment of DNMT1 and HDAC1⁵³ or direct interaction with HP1 for heterochromatin formation⁵⁴. In T cells, STAT4-dependent repressive histone marks were identified on several T_H2 cell-expressed genes, including STAT6 target genes, which are actively repressed by STAT4 in T_H1 cells. Although the total number of such genes is small (around 40 genes), the data clearly point to a role for STAT4 as a transcriptional repressor as well as its more widely recognized role as a transcriptional activator (Figure 2). It is not yet clear how a transcription factor can drive expression of one gene and repress the expression of another gene in the same cell, but it will be informative to analyze the associating factors/proteins that are locally recruited to genes bound and repressed by STAT proteins. The successful identification of specific chromatin modifications associated with STAT binding represents different way of utilizing the genomic approach, aside from obtaining a list of target genes. The genomic approach that integrates different types of readouts enables us to “examine the forest” rather than just “finding the trees” on the genome. Not to say that the trees are not interesting, but the big picture is important as well.

Figure 2. Distinctive epigenetic patterns are formed by STAT proteins in differentiated T helper effector cells.

A key role of signal transducer and activator of transcription (STAT) proteins includes shaping epigenetic patterns on target gene loci to maintain cell lineage specificity. Five distinct epigenetic patterns were found to be STAT4 dependent in T helper 1 (T_H1) cells that included both permissive chromatin signatures (high histone 3 lysine 4 trimethylation (H3K4me3) marks, high H3K36me3 marks or low H3K27me3 marks) and repressive chromatin signatures (high H3K27me3 or low H3K36me3). Permissive chromatin signatures are found on T_H1 cell-expressed genes, whereas repressed chromatin signatures are found on T_H2 cell-expressed genes in T_H1 cells. The figure was reproduced from online version of Wei et al, Immunity 2010⁴².

STAT6 in T_H2 cells

STAT6 as a driver of T_H2 cell differentiation

T_H2 cell differentiation is induced by IL-4, and the importance of STAT6 for this process has been well established in mice ^55–58. The actions of STAT6 and its downstream targets in pathological T_H2 cell responses such as asthma and allergy are also of great interest, given the public health impact of these diseases. Consequently, both mouse ^{16, 18} and human ⁵⁹ systems have been studied in terms of STAT6 functions. In human cells, STAT6 mediates the expression of more than 80% of IL-4-regulated genes, a higher proportion than was reported in previous studies using mouse cells ¹⁶. The function and cellular distribution of identified STAT6 targets are broad, reflecting the fundamental role of STAT6 in regulating multiple aspects of activities in cells.

Genome-wide kinetic profiling of STAT6-dependent gene expression and analysis of the STAT6-dependent gene network in humans⁵⁹ confirm that STAT6 is a major, direct contributor to the transcriptional profile associated with the T_H2 cell phenotype. The findings also show that IL-4-induced regulation of gene transcription in human cells is highly dynamic; only a subset of the genes that were differentially regulated within the first few hours after IL-4 treatment remained differentially expressed at later time points up to 72 hrs. These findings indicate that in addition to genes that provide the appropriate ‘switch signal’ for T_H2 cell differentiation at early times, factors other than STAT6 are required for the transition along the developmental pathway and maintenance of the acquired phenotype. Through genome-wide differential gene expression analysis using small interfering RNA (siRNA), 453 genes were identified as being regulated by STAT6 in human cells⁵⁹. Of these, only 6% of genes had been previously identified as STAT6 targets, including genes such as GATA3, SOCS1 and IL24. The new target genes suggest new functions and processes that are mediated by STAT6 signalling. In general, the findings underscore the importance of using genome-wide approaches to further explore whether there are species-specific roles of STAT proteins in humans and mice.

The early signalling network: connection to different T_H cell fates

By gene network analysis, the IL-4- and STAT6-regulated transcription factors were found to form a compact core interaction network of signalling (Figure 3). These data underscore the importance of combinatorial signalling pathways that function together to determine T_H cell commitment and fate. Of the newly identified direct STAT6 targets, three transcription factors that form hubs in the regulatory network — RUNX1, EPAS1 and BATF — are of particular interest. RUNX proteins have a central role in regulating T_H cell differentiation in general ⁶⁰, but RUNX1 preferentially inhibits T_H2 cell differentiation by downregulating GATA3 expression⁶¹ and binds to the IL4 silencer region⁶². In addition, RUNX1 can form a complex with FOXP3 or RORC, and is necessary for T_Reg and T_H17 cell function, respectively ^{63, 64}. Interestingly, EPAS1 binds to the RUNX1 promoter, potentially amplifying the effect of STAT6 on RUNX1 expression ⁶⁵. BATF, which is also directly regulated by STAT6, regulates both T_H17 and T_H2 cell differentiation ^{66, 67}. The connection between Th2 differentiation program and programs for other Th cell subsets can be further examined through the network of key transcription factors. As shown in Fig. 3, it is notable that within STAT6-mediated Th2 differentiation program, a close connection between STAT6 and other STAT family proteins is evident through only a few intermediate molecules. This in turn underscores the importance of understanding cooperative and counter-operative interactions between the STATs as well as between downstream transcription factors in directing T_H cell fate. For example, a comparison of ChIP-seq data for STAT6⁵⁹ and STAT5A ⁶⁸ showed that they have overlapping targets and indicated that these two STATs might have cooperative roles ⁵⁹. This is consistent with the known contribution of STAT5 to STAT6-independent T_H2 cell commitment^68–70.

Figure 3. The STAT6 signalling network identified during the initial T_H2 cell differentiation stage.

Interleukin-4 (IL-4)- and signal transducer and activator of transcription 6 (STAT6)-regulated transcription factors form a core network of interacting nodes. Genes shown in red color boxes are upregulated and those in green color boxes are downregulated by STAT6 in transcriptomics studies. STAT6-mediated regulation of genes detected by ChIP-seq is marked with red arrows (solid line for direct regulation, dashed line for indirect regulation). Furthermore, known direct interactions between the putative downstream transcriptional regulators of STAT6 in humans were added to the figure. Blue lines correspond to protein-protein interactions, and black lines correspond to other types of interaction or regulation. The networks were generated through the use of Ingenuity Pathways

Analysis (Ingenuity Systems, www.ingenuity.com) with some modifications based on published reports. The figure was modified from Elo et al. Immunity 2010⁵⁹ with the permission from Immunity.

A stabilizer of the T_H2 cell phenotype

The crucial role of STAT6 in the initiation stage of T_H2 cell differentiation is evident, but STAT6 also contributes to the maintenance of the T_H2 cell phenotype in differentiated cells, as supported by combinatorial genome-wide analysis of STAT6 binding, STAT6-induced epigenetic patterns and gene expression ⁴². Similar to STAT4, STAT6 is responsible for the maintenance of distinct epigenetic patterns on selected target genes. STAT6 predominantly functions as a transcriptional activator, but as is the case with STAT4, it also functions as a functional repressor for a subset of genes. In terms of activating target genes, STAT6 more frequently opposes the deposition of repressive epigenetic marks than it promotes permissive epigenetic marks; in this regard, STAT6 seems to be subtly different from STAT4.

Of particular interest are a subset of genes that are bound by STAT4 in T_H1 cells and by STAT6 in T_H2 cells, for which the respective STAT has opposing effects on local epigenetic patterns. A notable example is the Il18r1-Il18rap locus⁴². That is, whereas one STAT promotes active marks (in this case STAT4), the other STAT (STAT6) promotes repressive marks on the same locus. This divergent action of STAT4 and STAT6 on the same genes provides a back-up for inducing gene expression in one lineage and repressing gene expression in the other lineage (Figure 2).

STAT3 and T_H17 cell differentiation

T_H cells that selectively produce IL-17 (known as T_H17 cells) are one of the newest T cell subsets to be recognized. They have crucial roles in host defence against extracellular bacteria and fungi, but also in the pathogenesis of various autoimmune diseases ^{2, 3}. Cytokines that promote IL-17 production include IL-1, transforming growth factor-β1 (TGFβ1), IL-6, IL-21 and IL-23. The latter three cytokines activate STAT3. Although STAT3 is activated by a large number of cytokines and has crucial functions in various tissues ^{71, 72}, T cell-specific deletion of STAT3 mainly affects the expression of IL-17 and IL-21 ^73–75, and consequently results in decreased severity of several autoimmune disease models ^{73, 76–79}. Conversely, the deletion of Socs3, which increases STAT3 activation, results in increased numbers of T_H17 cells ⁷⁴.

Landscape of STAT3 targets

ChIP-seq analysis of STAT3 binding in T cells, coupled with gene expression analysis, has confirmed that the Il17 and Il21 genes are direct targets of STAT3 ^{46, 74, 76}. Of note, STAT3 binds to multiple sites in the Il17 locus⁸⁰, the most prominent of which are intergenic regions that coincide with conserved non-coding sequences (CNS) ⁸¹. These sites also bind p300 in a STAT3-dependent manner and so are probably enhancer elements (Golnaz Vahedi, Yuka Kanno, John O’Shea, unpublished observation). Furthermore, analogous to the role of STAT6, STAT3 directly binds to genes encoding multiple transcription factors that are crucial for programming T_H17 cells. These include Rorc ⁸², Rora⁸³, Ahr⁸⁴, Batf ⁶⁶, Irf4 ⁸⁵ and c-Maf ⁸⁶. Other important direct targets of STAT3 that define the T_H17 cell phenotype include Il23r and Il6ra ^{45, 76}. Notably, the ability of STAT3 to positively regulate expression of these genes was associated with permissive H3K4me3 marks. The prominent role of STAT3 in this process led to a re-evaluation of the factors involved in specification of T_H17 cells. While TGF-β signaling constitutes an important aspect of Th17 differentiation, an alternative mode of generating Th17 cells in the absence of TGF-β has also been recognized⁸⁷. It was found that activation of STAT3 in conjunction with IL-1 were sufficient to promote expression of the IL-23 receptor⁸⁷. Acquisition of the receptor allowed responsiveness to IL-23, which has a major role in driving pathogenic IL-17-dependent responses⁸⁸. As a result, pathogenic T_H17 cells were generated in the absence of TGF-β signalling⁸⁷ via activation of STAT3 and other cooperating factors. In this regard, it is interesting that genome wide STAT3 binding sites and that of IRF4 site overlap significantly following IL21 stimulation⁸⁹

In addition to the role of STAT3 in regulating the expression of T_H17 cell-related cytokines and transcription factors, Chip-seq analysis also pointed to a role for STAT3 in regulating T cell proliferation and survival. This phenotype was not evident in the initial description of deletion of STAT3 in T cells⁹⁰; however, Stat3^−/− T cells have delayed proliferation and poor clonal expansion, particularly in the setting of inflammation⁷⁶. In this regard, newly identified STAT3 target genes in T cells include the anti-apoptotic genes Bcl2⁹⁰ and ler3, and the proto-oncogenic transcription factors Fos, Jun, and Fosl2.

Complex roles of STAT3 in T_Reg cells

IL-6 also inhibits FOXP3 expression, an effect that depends on STAT3 ⁹¹. Accordingly, deletion of STAT3 in T cells resulted in the clonal expansion of induced T_Reg cells in the setting of colitis, but not in the normal gut ⁷⁶, consistent with the relief of IL-6-mediated inhibition of T_Reg cells. Curiously, when STAT3 was deleted only in the T_Reg cell population, the ability of T_Reg cells to constrain a pathogenic T_H17 cell response was selectively impaired, whereas suppression of T_H1 or T_H2 cell responses remained intact ⁹². These data were interpreted to indicate that intrinsic activation of STAT3 in T_Reg cells endows these cells with the ability to specifically suppress T_H17 cell responses. Gene expression analysis of STAT3-deficient T_Reg cells showed impaired expression of genes potentially contributing to the suppressor function of T_Reg cells (such as Prf1, Gzmb, Klrg1, Ccr6, Il1r1 or Il6ra). As Chip-seq datasets are generated, it will be of considerable interest to dissect how STAT3 controls the suppressor function of T_Reg cells that directed towards a specific T_H cell subset.

STAT5 and T_Reg cell differentiation

Essential regulators of lymphoid development and peripheral tolerance

Stat5a and Stat5b genes are adjacent on the same chromosome in both mice and humans, have 96% similarity in sequence and have overlapping functions in diverse tissues^{93, 94}. Similar to STAT3, germline deletion of Stat5a and Stat5b (collectively referred to as Stat5) is embryonic lethal ^{95, 96}. The few mice that survive are extremely runted and anemic. Deletion of Stat5 has marked effects on all lymphoid lineages including T cells (thymic and peripheral), B cells and NK cells, pointing to a crucial role for STAT5 in lymphoid development. It should also be noted that the targeted disruption of the Stat5a or Stat5b genes individually yielded distinctive phenotypes, suggesting signaling mechanisms unique to each⁹⁴.

Elucidation of STAT5 target genes in T cells⁶⁸ and other cells has been accomplished, and these data helped to explain the role of STAT5 in T_H2 cell differentiation by up-regulating expression of the IL-4 receptor⁶⁸. However, there has been remarkably little analysis so far of the non-redundant roles of STAT5 in regulating the development and survival ofn T cells. Given the profound effects of STAT5 deficiency in T cells, more extensive analysis of the STAT5 ChIP-seq data sets is warranted.

The few peripheral T cells that develop in STAT5-deficient mice have an activated phenotype leading to the development of autoimmunity ^97–99. One major factor underlying this autoimmune phenotype is the impaired T_Reg cell development in both thymus and periphery that results from deletion of Stat5 in CD4⁺ T cells. Indeed, it was recognized that STAT5 binds to the promoter and first intron of the Foxp3 gene, to activate transcription of this T_Reg cell master regulator^{95, 96,100}. In addition, STAT5 influences the survival of T_Reg cells by regulating expression of the IL-2 receptor α-chain (Cd25) and the anti-apoptotic gene Bcl2. Although it is probable that STAT5 regulates many aspects of lymphoid survival, a direct comparison of STAT5 targets in T_Reg cells and conventional CD4⁺ T cells has not been carried out. Equally, a direct comparison of STAT5 and STAT3 targets in T_Reg cells would be particularly interesting.

Whereas IL-2-mediated activation of STAT5 is indispensable for the maintenance of T_Reg cells through upregulation of FOXP3 expression, other cytokines negatively influence FOXP3 expression. For example, IL-4-activated STAT6, IL-12-activated STAT4 and IL-6-activated STAT3 all decrease the expression of FOXP3 and affect chromatin modification at this locus ^{73, 95, 101, 102}. However, the exact mechanisms by which STAT4 and STAT6 function to negatively regulate T_Reg cells have not been elucidated on a genome-wide scale.

Analogous to the relationship between STAT4 and STAT6 in determining T_H1 versus T_H2 cell differentiation, the relationship and balance between STAT5 and STAT3 seem to dictate the dichotomy of T_Reg cells and T_H17 cells⁸⁰. In addition to its role in positively regulating T_Reg cell function, STAT5 inhibits T_H17 cell differentiation¹⁰³. To address the potential mechanisms underlying this action, mapping of STAT5 targets in IL-2-activated T_H17 cells was carried out by ChIP-seq. One important finding was the extensive overlap between STAT3 and STAT5 binding sites in the Il17 gene ⁸⁰. It was found that STAT5 competes with STAT3 for binding to Il17 and inhibits the function of STAT3 in activating Il17 transcription. The opposing effects of STAT3 and STAT5 on Il17 transcription explain why IL-2 inhibits IL-17 production, although effects of signaling molecules other than STATs, which activated by these distinct cytokines, may also be contributors. In many other cases, STAT3 and STAT5 will work together to enhance gene expression, but given the example of Il17, the precedent is clear that these two highly related transcription factors can act in opposition. Exactly how these factors can accomplish global versus gene-specific effects warrants further investigation.

STATs and human disease

In addition to abundant data pointing to critical functions of STAT proteins in animal models, evidence of their importance in humans is quickly emerging from studies in primary immunodeficiency and in autoimmune diseases.

New insights into primary immunodeficiency disorders

Previous work has established that STAT5A/B mutations in humans are associated with impaired T_Reg cell function¹⁰⁴ and that STAT1 mutations are associated with susceptibility to viral and mycobacterial infections^{105, 106}. Furthermore, recent work has established that another classic primary immunodeficiency, Hyper-IgE or Job’s syndrome (HIES) is a result of dominant-negative mutations of STAT3 ^{107, 108}. This finding was interesting because germline deletion of Stat3 in mice is embryonic lethal. As a result, the restricted pathology seen in humans with STAT3 mutations was not anticipated; presumably, this is because the mutant allele interferes with, but does not totally abrogate, STAT3 function. An important aspect of this immunodeficiency disorder is absence of T_H17 cells ^109–112. A classic feature of HIES is infection without the typical signs of inflammation (i.e. redness and warmth) that results in “cold abscesses”. It is tempting to speculate that what might underlie this unique feature of HIES is the defect of T_H17 cells to produce IL-17, which in turn results in the failure to recruit neutrophils to sites of infection. It remains to be elucidated how impaired function of STAT3 in other tissues contributes to the pathology seen in HIES in various tissues.

Genetic polymorphisms and human autoimmunity

Although various animal models have implicated STATs and altered cytokine signalling in autoimmunity, the issue always arises as to whether these models really mirror immunopathogenic mechanisms in humans. However, large scale genome-wide association studies (GWAS) have now provided evidence that various genes involved in cytokine signalling, and STATs in particular, are linked to the development of autoimmunity in humans (Table 2). For example, polymorphisms in STAT3 are linked to susceptibility to Crohn’s disease (a form of inflammatory bowel disease) and ankylosing spondylitis ^{113, 114}. Equally compelling is the evidence that polymorphisms of IL23R and JAK2, both of which signal through STAT3, are linked to the same diseases ^115–120, suggesting a profound involvement of the IL-23–STAT3 axis in the genesis of autoimmune diseases.

Table 2.

Genetic analysis of human autoimmune diseases showing linkages to JAK-STAT signalling pathway.

disease	Immunological phenotype	gene	Mutation/linkage	referense
Hyper-IgE syndrome	Skin abscess, cystic lung infection, elevated serum IgE, impaired Th17 generation	STAT3	Missensemutations leading to dominant negative STAT3 protein	107 108
Immune dysfunction/growth hormone insensitivity	Impaired Treg function	STAT5b	Missense mutation -A630P in SH2 domain leading to failure to respond to activation signal	104
Immune dysfunction	Susceptibility to infection	STAT1	Missense mutation-L600P, 1757- 1758delAG, L706S leading to loss of function	106 105
Crohn’s disease (CD)	Overactive mucosal immune reaction of GI tract triggered by commensal intestinal bacteria	STAT3 JAK2 STAT4	11 previously reported loci + 21 additional loci linked to CD by meta-analysis A SNP (rs7574865) linked to early onset and colonic CD	113 122
Reumatoid arthritis Systemic lupus erythematosus	Immune reaction against the lining of small joints Systemic immune reaction against own tissue and organs	STAT4	A SNP(re7574865)linked to both RA and SLE	121
Primary Sjögren’s Syndrome(pSS)	Inflammation of salivary and lacrimal glands leading to a dry mouth and dry eyes	STAT4	A SNP(rs7574865) linked to pSS	123

Components of the JAK–STAT signaling pathway have been identified as causal genes for autoimmune diseases and have also been implicated in genetic linkage studies as having statistically significant differences between patients and controls.

In multiple studies, a variant allele of STAT4 has been found to be associated with increased risk of developing systemic lupus erythematosus (SLE), rheumatoid arthritis, Sjogen’s syndrome and Crohn’s disease ^121–123. The connection between STAT4 and SLE is perhaps unexpected insofar as this disease is not a prototypical T_H1 cell-mediated disease. However, it is worth noting that STAT4 can be activated by type I IFNs¹²⁴, and an important aspect of the pathogenesis of SLE is the “interferon signature” ¹²⁵. Polymorphisms in TYK2, whose gene product is activated by IFNs, IL-12 and other cytokines, have also been reported to be associated with SLE ¹²⁶, providing further evidence in support of the importance of STAT4 in the pathophysiology of this disorder. As the STAT4 polymorphisms do not fall within the coding region of the gene, they presumably influence the level of gene expression, but clearly much more work is required to confirm this hypothesis.

Concluding remarks and future directions

The powerful genome-wide approaches now available to researchers have enabled a comprehensive evaluation of the role of individual STAT proteins in specifying T_H cell lineages and they provide a quantitative determination of the target genes that are mobilized during the process of T_H cell differentiation. These findings have established that STATs have multiple roles during the initiation stage as well as the maintenance stage of a T_H cell fate decision. For the latter case, a key role of STATs involves the induction and/or maintenance of epigenetic patterns on target loci. STAT proteins induce both permissive and repressive epigenetic patterns. Although a particular STAT can be assigned to each T_H cell lineage as a dominant regulatory factor, it is clear that this is an overly simplistic view of defining T_H cell lineages. Emerging evidence points to a functional network, with STATs working cooperatively and in opposition with other transcription factors to ensure the desired balance between different T cell fates, and in certain cases even to promote phenotypical plasticity.

Fortunately, we now have genome-wide approaches to define the breadth of transcription factor action. In addition, we also have the ability to carry out many chromatin-related assays on a genome-wide scale to examine the activity of genomic regions ¹²⁷ through common chromatin signatures ¹²⁸, and to determine the state of dynamic genomic organization (Figure 4). In particular, the extensive coverage of the genome afforded by next generation sequencing offers the possibility of exploring so called “gene desert” or intergenic regions for distal enhancers and other types of regulatory element^{21, 24, 129}. This is an exciting opportunity to analyse previously unexplored regions of the genome, and in fact, recent reports have shown that the patterns of distal enhancers are quite unique and different between different cell types ¹³⁰. The challenge now of course is to understand how “master regulators” of cell fate and other transcription factors, such as STATs, contribute to the activity of distal enhancers in a manner that creates cell identity^{21, 130}. It is quite possible that some of the polymorphisms that have been linked to autoimmune diseases reside in enhancer regions of the genome ¹³¹ that are crucial for regulating tissue-specific patterns of gene expression from a distance.

Figure 4. Markers of genomic organization to define activities of chromosome regions.

Genomic organization encompassing the interferon-g (Ifng) locus in T helper (T_H) cells. In T_H1 cells, in which the Ifng locus is actively transcribed in a signal transducer and activator of transcription 4 (STAT4)-dependent manner, the promoter is marked by permissive histone 3 lysine 4 trimethylation (H3K4me3) and STAT4 binding, and the gene body is marked by permissive H3K4me3 and H3K36me3 marks. One of the distal enhancer elements (shaded in gray) is marked by H3K4me1 and STAT4 binding in T_H1 cells and by repressive H3K27me3 in T_H2 cells. Further 5′ upstream of the Ifng locus, an insulator site marked by CTCF binding is located and all permissive histone marks and DNase hypersensitivity sites are restricted beyond that point.

Components of the JAK–STAT signaling pathway have been identified as causal genes for autoimmune diseases and have also been implicated in genetic linkage studies as having statistically significant differences between patients and controls.

Equally, the entire notion of the epigenome is in the midst of a not so quiet revolution^{27, 132}. Increasingly, “epigenetics” is being viewed as an extension of signal transduction. Nonetheless, it is certainly not clear how all of the components of epigenetic information are linked to each other, to signalling and to transcription factor binding. Indeed, we are really in our infancy of understanding how signal transduction and nucleosome biology relate. Because the STAT pathway provides a rapid means of transmitting signals from the extracellular environment to the nucleus, investigating how cytokines drive T_H cell specification will provide remarkable opportunities to link signal transduction, transcription factor binding, nucleosome biology and gene expression. Hopefully, we will witness valuable new information about gene regulation coming from systematic database analysis of STAT-mediated cytokine signalling in differentiating T cells.

Glossary

Next generation sequencing

High throughput sequencing methods that produce rapid, inexpensive, accurate sequencing data that can cover entire genomes. Based on different chemistries, several different platforms are available including: Illumina/GA, Roche/454, ABI/SOLID and Helicos/HeliScope

Chromatin immunoprecipitation

A technique for the detection of proteins bound to specific regions of chromatin. These assays involve chemically crosslinking proteins bound to the DNA sequences, followed by immunoprecipitation with an antibody that is specific for the crosslinked protein

Epigenome and Epigenetic regulation

These terms refer to the heritable, but potentially reversible, states of gene activity that are imposed by the structure of chromatin such as covalent modifications of DNA or of nucleosomal histones. The epigenome pertains to the aspects of heritable cellular phenotype that is not explained by DNA sequence

ChIP-seq

A technique in which chromatin immunoprecipitation is followed by high throughput sequencing to generate maps the genome-wide distribution of protein-DNA interaction. This technique can be used to measure transcription factor binding or histone modifications

Nucleosome

A nucleosome consists of histone protein core and a segment of DNA wrapped around it. It is the minimum unit to make up chromosome

ChIP-on chip

The term refers to a technique that combines chromatin immunoprecipitation (“ChIP”) with microarray technology (“chip”) to investigate DNA-protein interaction in vivo on a genome-wide basis

Enhancer element

A control element in DNA to which regulatory proteins bind and influence the rate of gene transcription of the associated gene(s). Enhancers function in an orientation- and position-independent manner so that they can function either from upstream or downstream of the associated gene, or in an intron

Genome-wide association study (GWAS)

GWAS refers to a study in which genomewide genetic variation is linked to a particular phenotype, most often a clinical disorder by applying high-throughput genotyping techniques to profile single-nucleotide polymorphisms (SNPs) of controls vs. patient