Helper T Cell Identity and Evolution of Differential Transcriptomes and Epigenomes

Sometimes you do not have to make up your mind

Early work in vitro argued that polarized Th1 and Th2 cells were phenotypically fixed states; that is, once a CD4⁺ T cell had chosen its fate, it would not easily switch to another fate, even if exposed to the cytokines that drove differentiation to the opposing subset (14, 39, 40). One way this is accomplished is the downregulation of cytokine receptors required to sense the environmental cues that drive an opposing fate (41). However, there are many lines of evidence that challenge a strict view of helper T cells as distinct lineages. For instance it is not infrequent for CD4⁺ T cells to co-express more than one ‘signature’ cytokine, particularly in vivo. Th17 cells, for example, readily become IFN-γ producers (42-46). In fact, the conversion of IL-17 producers to IFN-γ producers is an important aspect of immunopathogenesis disease models and likely autoimmune disease in humans (44).

Because Tfh cells do not produce a unique ‘signature’ cytokine and can also express cytokines produced by other fates, it should come as no surprise that Tfh cells have been argued to exhibit considerable flexibility. In vitro generated IL-21-producing Tfh-like cells can easily be re-differentiated to make IFN-γ, IL-4, or IL-17 (47). Conversely, early in their differentiation, Th1 cells exhibit features of Tfh cells (26). Later though, Tfh features are repressed and the ‘Th1’ aspects dominate. In vivo, during the peak response of a helminth infection, all of the IL-4-producing cells found in the lymph node can be viewed as Tfh cells, based on their location in B-cell follicles (23, 48, 49). Similarly, during bacterial infection, expression of IFN-γ is readily apparent in Tfh cells (23). For this reason, viewing Tfh cells as a fate similar to Th1 or Th2 cells has been questioned. One can envision a model in which the IFN-γ or IL-4-producing cells retained in the lymph node can be considered Tfh cells, while those that exit this milieu represent bona fide polarized Th1 or Th2 cells. Exposure of CD4⁺ T cells to signals provided by B cells likely transition the cells retained in the lymph node to a stable Tfh phenotype (22, 50, 51). The precise sequence of signals necessary for a cell to acquire both Tfh and Th1 or Th2 characteristics are still unclear, and thus the plasticity of Tfh cells is an area of active research.

Another dramatic example of flexible cytokine expression is the finding that committed Th2 cells can be reprogrammed to become IL-4⁺IFN-γ⁺ double producing cells in the context of adoptive transfer and viral infection (52). These so-called ‘Th2+1’ cells exhibit aspects of both Th1 and Th2 programs and challenged the original model of the stable Th1/Th2 paradigm. Thus, all CD4⁺ T cells may in fact be quite flexible rather than fixed and stable fates; though they are among the most plastic, flexibility is not limited to Th17 and Tfh cells.

The flexibility or stability of Tregs is perhaps one of the most controversial and studied areas of helper T-cell plasticity. Aside from the interesting basic implications, transfer of Tregs has been proposed as a therapeutic approach for autoimmune diseases and cases where limiting inflammation may be beneficial. If Treg cells are highly unstable, this approach will have little likelihood of success (53). Several studies have reported that a fraction of Treg cells lose their stability, particularly during inflammatory environments where cues can potentially influence a Treg to convert to an effector cell (54-58). However, other groups have argued that these unstable cells are not bona fide Tregs or that conversion to effector fates is an experimental artifact (59, 60). Finally, whether these cells are thymically derived nTregs or peripherally derived iTregs may complicate our ability to assess the stability or flexibility of the Treg cell fate. While this area of research is still under investigation and unresolved, there are clear examples where cells with suppressive capacity can in the right circumstances take on a more effector cell fate.

In this context, CD4⁺ T-cell ‘identity crisis’ – how helper T cells choose an effector function while maintaining the ability to be flexible – is the major focus of much current research of CD4⁺ T-cell differentiation. Understanding the molecular basis of helper T-cell specification and plasticity is a critical question both in terms of the basic science and also clinical/translational implications. As we discuss in detail, there are multiple mechanisms to repress alternative fates when a given fate is adopted. If so, why is it that CD4⁺ T-cell options are carefully regulated, yet provide for the possibility of flexible gene expression? An obvious possibility is that flexible helper T cells may be beneficial to the host for dealing with more than one pathogenic micro-organism: bacterial super-infections are common sequelae of viral infections and poly-microbial sepsis occurs in seriously ill patients. With age and declining T-cell repertoires, flexibility in gene expression may also be advantageous. T-cell plasticity may represent the tuning of the effector response as environmental cues change. Alternatively, the apparent flexibility might also represent specialization within the populations. This cellular heterogeneity could allow for adaptation to responses as environmental cues vary. This hypothesis would fit nicely into a dissection of responses in the peripheral lymph node organs versus the tissues, such as the gut or skin environment. Regardless of the reason, the many examples of T-cell plasticity suggest that there are multiple mechanisms that allow for this flexibility in effector function. As discussed in this review, these mechanisms include the regulation of the transcriptome through carefully controlled expression of transcription factors and by modulating the epigenomic landscape, both of which fundamentally influence the ‘identity’ of the CD4⁺ T cell.

Sensing the environment

There are a number of factors that seem to contribute to CD4⁺ T-cell specification. One well recognized factor is the ‘strength’ of T-cell receptor (TCR) signaling (61-63). Several studies have demonstrated that strong TCR signaling promotes Th1 differentiation, whereas weak TCR signaling favors the Th2 cell fate during in vitro T-cell skewing assays. In addition, strong T-cell receptor signaling is reported to induce Th17 differentiation (64). Strong TCR signals can also reportedly promote Tfh cell fate commitment (65). nTreg differentiation in the thymus is also thought to be the result of antigens that provide signals stronger than those that drive generation of conventional T cells but weaker than antigens that drive negative selection (66). Generation of iTregs on the other hand has been associated with weaker TCR signals (34, 67-70). How ‘strong’ signals can drive the differentiation of diverse subsets is difficult to understand, particularly in vivo, and the precise mechanistic underpinnings of these phenomena remain obscure.

In addition to TCR signaling, ambient environmental cues to which an activated T cell is exposed are key factors in helper cell specification. The secreted cytokines produced by both innate and adaptive immune cells are among the most important and relevant specifying factors. A large proportion of these cytokines, though not all, signal via receptors that belong to the Type I/II cytokine receptor superfamily (71). This family of receptors utilizes the Janus kinase (JAK)-signal transducer and activator of transcription (STAT) signaling pathway signaling pathway to translate the extracellular cytokine signals into meaningful gene transcription programs, which promote differentiation of a given T-cell subset (72). Within the STAT family, there are seven DNA binding proteins (STAT1-5a, 5b, and 6) that function as transcription factors (73). The conventional view focuses on the distinct role of STATs as factors that drive T-cell differentiation in the presence of a given cytokine (72). However, the reality is more subtle and complex. Differentiating helper T cells are exposed to more than one cytokine during immune and inflammatory responses. In addition, most cytokines do not just activate a single STAT; most activate multiple STATs. Recently, the complex functions of STATs have become better appreciated; they do much more than just bind promoter elements and induce transcription. Indeed, the negative regulatory role of STATs in helper cell differentiation has become increasingly apparent.

Classically, we link IL-12-dependent activation of STAT4 with generation of Th1 cells, but STAT1 activated in response to IFN-γ is also a relevant player (74). Additional studies have also implicated IL-2 and STAT5 as contributors to Th1 cell differentiation (75). For Th2 cells, IL-4 and STAT6 promote differentiation, and the absence of STAT6 abrogates in vitro Th2 cell differentiation (74). However, Th2 cells are generated in vivo without IL-4 or STAT6 (76). Accordingly, STAT5 is also a contributor to Th2 differentiation. IL-2-dependent activation of STAT5a and STAT5b causes these factors to bind and transactivate the Il4 and Il4ra genes thereby facilitating Th2 differentiation (77-81). IL-6 and STAT3 can also promote Th2 differentiation (82-84).

STAT3 is particularly important for the differentiation of Th17 cells, and without STAT3, Th17 differentiation is severely impaired (85-89). The failure to generate Th17 cells is an important feature of hyperimmunoglobulin E syndrome (HIES) or Job’s syndrome, a disease that is due to dominant negative STAT3 mutations (90, 91). IL-6 and IL-23 both activate STAT3, and along with IL-1 and TGF-β, these cytokines drive IL-17 production (92).

IL-6 and IL-21, again acting via STAT3, can promote the generation of Tfh-like cells (22, 93, 94). However, in vivo the absence of IL-6, IL-21, or STAT3 does not ablate Tfh cell development (93, 95-100). In this regard, it is notable that IL-12 and STAT4 promote Tfh cell formation both in vitro and in vivo, although sustained activation represses Tfh features (26, 100-102).

Finally, STAT5, activated by IL-2, is critical for Treg cell differentiation. In contrast to other STATs and their associated subsets, STAT5A/B are entirely non-redundant for Treg cells. In the absence of STAT5A/B, the differentiation of this subset is completely abrogated.

Although STATs were originally discovered as positive regulators of IFN-induced genes, their ability to inhibit gene expression is increasingly apparent. STAT4 can inhibit Th2 cell formation while promoting Th1 cell differentiation (74). Conversely, STAT6 inhibits Th1 cells while driving the Th2 cell fate (74). Although STAT5 promotes Treg, Th1, and Th2 cell differentiation, it is a negative regulator of Th17 and Tfh cells (103-106), again highlighting the dual role of STATs as both activators and inhibitors. In contrast, STAT3 activation inhibits iTreg differentiation (107, 108). Thus, for the Treg and Th17 fates, STAT5 and STAT3 have opposing roles. STAT1 is also important negative regulator of Th17 differentiation (109, 110). This is vividly demonstrated in the primary immunodeficiency disorder chronic mucocutaneous candidiasis. This disorder is characterized by recurrent fungal infections and impaired Th17 cell generation and is due to gain-of-function STAT1 mutations (111). Thanks to new technologies, discussed in detail below, we can map precisely how STATs inhibit gene expression and antagonize the function of other STATs (112, 113). In summary, the complex nature of environmental cytokine cues to which a differentiating CD4⁺ T cell is exposed greatly influences the activation of different STAT family members, and they in turn are major contributors to the outcome of CD4⁺ T-cell fate determination.

Acquiring master regulators

In addition to the STAT family of transcription factors, each helper cell fate expresses a lineage-defining ‘master regulator’ transcription factor. Classically, these factors are defined as being both necessary and sufficient for characteristic features of the various subsets. In the conventional view, a single master regulator is expressed in a given helper T-cell subset (Fig. 1). Typically, T-bet (encoded by Tbx21) (114, 115), GATA-binding protein 3 (GATA3) (116), retinoic acid receptor-related orphan receptor-γt (Rorγt) (117), SFFV proviral integration 1 (PU.10) (118), B-cell lymphoma 6 (Bcl6) (119, 120), and forkhead box P3 (FoxP3) (121, 122) are referred to as the master regulators of Th1, Th2, Th17, Th9, Tfh, and Treg cells, respectively. However, recent emerging data have demonstrated that, like the signature cytokines, the expression of T-cell master regulators is more complicated than originally appreciated. Indeed, the extent to which the nomenclature master regulator is apt is being revisited.

Unlike STAT family proteins, which are expressed in all subsets, master regulators are either not expressed in naive T cells or expressed at low levels. In many cases, their expression is driven by TCR signals and STATs. Under the influence of cytokines and other environmental cues, the dominant expression of one master transcription factor over the others can promote a specific T-cell fate, at least this was the classical model of T-cell differentiation. What complicates these initial views of master regulators is that their patterns of expression are substantially more complicated than the simple model equating a given subset with a single lineage-defining master regulator. Many examples of stable co-expression in fully differentiated cells now challenge this conventional model. For example, T-bet and GATA3 are transiently co-expressed in recently activated CD4⁺ T cells and can functionally interact, limiting the action of Gata3 (123). It is important though to distinguish this temporary and transitional type of co-expression from stable co-expression in a differentiated population of cells.

Treg cells are good example of the latter. Simultaneous expression of FoxP3 with T-bet, GATA3, or Bcl6 has documented. This has been argued to be functionally relevant for control of Th1, Th2, and Tfh cell responses, respectively (57, 124-129). As STATs are important inducers of master regulators, it comes as no surprise that STAT1 is important for Treg cells to express T-bet via sensing the cytokines IL-12 or IL-27 (129, 130). Similarly, although STAT3 is not a master regulator, its expression in FoxP3⁺ Treg cells is important for suppression of Th17 responses (131). Together, these findings suggest that co-expression of master regulators may create populations of cells that are sub-specialized for specific functions of CD4⁺ T-cell subsets (Fig. 2A).

Somewhat counter-intuitively, T-bet and GATA3 can be stably co-expressed in previously committed Th2 cells to create a population termed Th2+1 that has features of both Th1 and Th2 cells (52). This example of co-expression suggests a way for CD4⁺ T cells to ensure some flexibility that may be crucial to host defense against diverse microbial pathogens, rather than a sub-specialized population that has unique functional properties (Fig. 2B).

T-bet and Bcl6 can also be simultaneously expressed (26, 132-134). These studies suggested that low levels of T-bet can be co-expressed with low levels of Bcl6 in CD4⁺ T cells. However, at higher concentrations these factors are mutually antagonistic. Thus, the balance between T-bet and Bcl-6 expression could be important for the decision between a Th1 and Tfh cell (Fig. 2C).

These examples of co-expression make it appropriate to re-visit and re-define our views of master regulator. Functionally, helper cells have the ability to co-express more than one master regulator to provide something necessary for the immune response. But the molecular mechanisms of how this co-expression can exist in certain settings, and how it shapes the identity of the cell both from a transcriptomic and epigenomic view likely underlies the basis for helper T-cell plasticity.

Towards a transcription factor network for helper T cells

While STATs and master regulators are critical for helper cell differentiation, it is overly simplistic to try to explain the diverse functionalities of helper T cells based on these two classes of factors. In reality, specification requires a cohort of critical transcription factors working in concert. During T-cell development, CD4⁺ T cells express an array of transcription factors that dynamically change over the course of commitment in the thymus, allowing them to diverge from CD8⁺ T cells (135). Factors such as Thpok, Runx3, Runx1, Ets1, Tox, and the E proteins E2A and HEB are all induced at discrete steps to drive commitment to either the CD4⁺ or CD8⁺ lineage. Importantly, GATA3, the so-called master regulator of Th2 cells, has critical roles in thymic development as well (136). Thus, any given transcription factor may have stage-specific functions.

In addition to the factors required for development, several other transcription factors have been described that are critical for CD4⁺ T cells but are not specifically required for only one subset. Interferon regulatory factor 4 (IRF4) is important for the differentiation of Th2, Treg, Th17, Th9, and Tfh cells (137-141). To function, IRF4 complexes with members of the AP1 family, making family members like BATF and c-Maf necessary for several CD4⁺ T-cell subsets (142-146). Additional transcription factors that are important for Th1 cells include Hlx, Runx3, and the Ets family members (147-149). In addition to c-Maf, the AP1 family member JunB and the transcription factor Gfi-1 are required for Th2 cells (150, 151). Recently, HIF1, Runx1, Aiolos, and Fosl2 have all been demonstrated to be important for Th17 cells (144, 152-154).

The recognition that helper T cells express many key transcription factors has been aided by improved technologies for measuring their simultaneous expression. This allows us to view helper T cells with a ‘wide-angle lens’ rather than taking narrowly focused snapshots. The complexity of the transcription factor networks is daunting to consider; however, at this point it is rather naive to simply measure one transcription factor, master regulator or otherwise, and try to make strong functional conclusions. A much more reasonable view is to simply accept the intricacy of transcription factor action, knowing that these factors work in multimolecular complexes. Thus, we need to appreciate that the master regulators and STATs exert their effect in the context of other transcription factors needed for CD4⁺ T-cell identity and try to be comprehensive in assessing the expression of all these factors.

Global transcriptional profiling in T cells

Although CD4⁺ T-cell subsets are typically defined by their expression of cytokines and master regulator transcription factors, the question arises how many other genes are unique to each helper T-cell subset? For more than a decade, microarray based technologies have been the technology of choice for high-throughput transcriptome profiling in CD4⁺ T cells (155). These studies have indicated that hundreds of genes are differentially expressed among various T-helper subsets, many of which can be directly related to their differentiation and function (112, 155, 156). Despite their routine use, expression microarrays have limitations. These include hybridization and cross-hybridization artifacts and design constraints that prevent the detection of RNA splicing and previously unknown transcripts. In contrast, RNA sequencing (RNA-seq) provides detailed information about microRNAs (miRNAs), long non-coding RNAs (lncRNAs), unannotated genes, and splicing isoforms (157). This technology now provides a more complete catalogue of transcription and sequence based profiling of the transcriptome is slowly replacing expression microarrays.

Using RNA-seq technology, the transcriptomes during thymic T-cell development have been obtained (135). Unexpectedly, the largest differential gene expression occurred between double negative 2a (DN2a) and DN2b cells, not between DN1 and DN2, where Notch signaling induces the first signs of T-lineage entry. The transcriptional profiling in Th17 cells also identified more than hundred genes to be either expressed or repressed in these cells compared to the naive state (144). Such RNA-seq datasets not only quantify gene expression levels but also provide detailed information about the choice of exons that is in use in different T-helper subsets (135).

RNA sequencing has allowed the repertoire of non-coding RNAs to be glimpsed for the first time. Among these categories of transcripts, miRNAs are small single-stranded non-coding transcripts that post-transcriptionally regulate their target genes. Drosha and Dicer are two major components that allow for miRNA processing and function (158). Loss of these factors affects antibody diversity in B cells and also leads to instability of helper T cells and consequently autoimmunity (158-160). A comprehensive profiling of miRNAs during lymphopoiesis revealed the underlying transcriptional forces that shape miRNA homeostasis (161). Various miRNAs were reported to play key roles in T-cell specification. For example, miR-125b contributes to enforcing a naive state of Th cells (162), whereas miR-182 promotes clonal expansion (163). By contrast, miR-10a attenuates the phenotypic conversion of iTregs into Tfh cells by targeting Bcl6 and the co-repressor Ncor2 (164).

The importance of a limited number of lncRNAs has long been appreciated (165). Examples include Xist and Air, which are associated with heterochromatin formation and imprinting (166, 167). Overall though, relatively few lncRNAs have been functionally characterized and even fewer have been studied in T cells. One lncRNA, Tmevpg1, located in relative proximity to the Ifng gene, positively controls expression of this key gene (168). Interestingly, STAT4 and T-bet are both regulators of Tmevpg1 expression. ENCODE has identified ten-thousand new lncRNAs (169). Thus, this area of T-cell biology is clearly in its infancy. Undoubtedly, there will be many more non-coding RNAs that will have an impact on helper cell specification, and we are only beginning to understand their roles in helper T-cell function.

Chromatin status of CD4⁺ T cells

As explained in the introduction, the expression of a given transcription factor may not be sufficient for it to exert its effect; its targets need to be accessible. In this way, the chromatin landscape permits transcription factors to act on the genome and thus both factors contribute to selective gene expression and acquisition of cell identity. In fact, it has long been recognized that CD4⁺ T cells respond to changes in their environments by reshaping the chromatin architecture of signature cytokine loci and other key genes expressed by effector CD4⁺ T cells (170-175).

The impact of histone modifications, nucleosome positioning, or DNA methylation in CD4⁺ T-cell differentiation has been established through the use of genetic deletions or knockdown approaches. Brahma-related gene 1 (BRG1)-containing complexes, which displace nucleosomes, are important for remodeling the Ifng locus and production of this cytokine is important for Th1 differentiation (176). The cohesin protein complex is also important for the maintenance of gene architecture (177). Knocking down just one component of this complex, RAD21, also causes a reduction of IFNG transcript levels (178). Conversely, the Trithorax complex functions as a histone 3 lysine 4 methyltransferase and thereby serves to repress gene expression. Impaired expression of the components of this complex are associated with aberrant Th2 differentiation (179, 180). Similarly, H3K9me3 is another important marker of gene silencing. Deletion of the H3K9 methylase results in inappropriate expression of Th1 genes in Th2 cells (181). Another key factor in silencing gene expression is DNA methylation. Conditional ablation of the DNA methyltransferase, DNMT1, or methyl-CpG binding domain protein 2 results in the inability of Th1 or Th2 cells to properly silence the expression of genes typically expressed in the opposing lineage (91-93).

Not surprisingly, epigenetic modifications extend beyond the control of the cytokine genes, affecting the expression of the key transcription factors associated with helper T-cell differentiation. For example, a conserved CpG rich region resides upstream of the FoxP3 promoter (182). This region is partly methylated in iTreg cells but completely demethylated in nTreg cells. This helps explain the relative stability of expression of Foxp3 in nTreg cells (183-185).

Recent developments in deep sequencing technologies have now replaced the painstaking mapping of chromatin modifications at individual loci with genome-wide catalogues. In particular, chromatin immunoprecipitation followed by sequencing (ChIP-seq) is widely used for profiling histone tail modifications, distribution of chromatin-associated proteins, and binding of transcription factors (186). Genome-wide mapping of the trimethylation status of naive CD4⁺ T cells, and fully polarized Th1, Th2, Th17, iTreg, and nTreg cells has now been accomplished (56). Consistent with the standard ‘lineage commitment’ view of helper cell differentiation, signature cytokine genes exhibit unopposed permissive marks (H3K4me3) in the appropriate subsets (e.g. Ifng in Th1 cells) and repressive marks (H3K27me3) in other subsets that do not express these cytokines. Genes other than cytokine genes also exhibited this pattern of regulation. Surprisingly though, genes encoding master regulators such as Tbx21, Gata3, Bcl6, Runx3, and Prdm1 have more complex marks. In many cases these genes exhibit “bivalent poised domains”, meaning that both accessible and repressive marks are present (56). This helps explain why T-bet can be expressed in Treg, Th17, and even Th2 cells (30, 38, 100). It also explains the flexible expression of Bcl6 and how a novel GATA3- and RORγt-expressing subset of Th2 memory/effector cells can arise (187). Thus, interrogating the epigenetic modifications of genes encoding cytokines and transcription factors provide insights into stability and flexibility of expression.

Action of the environmental sensors

As critical sensors of the cytokine milieu that shape T-cell differentiation, it might be anticipated that various STATs can have a substantial effect on transcriptomic and epigenomic changes. However, it was an open question as to which factor would have a more pronounced effect: STATs or master regulators. Would these factors influence both transcription and epigenetic modifications? The role of STATs on transcriptomic changes has been documented using microarray technology and the major impact of STATs is very clear (65, 102). However, microarray technology does not allow one to distinguish direct versus indirect effects. In addition to measuring genome wide changes in histone modifications, ChIP-seq can be used to measure genome wide transcription factor binding. Coupling this information with transcriptional profiling by microarrays or RNAseq and histone ChIP-seq in wildtype versus transcription factor-deficient cells, one can begin to achieve a more complete view of transcription factor action.

Genome-wide binding studies for STAT3, STAT4, STAT5, and STAT6 proteins identified target genes of these transcription factors in Th1, Th2, and Th17 cells (52, 65, 102, 103). While most of STAT4 binding sites (80%) occurred outside the promoter regions in Th1 cells, 4000 genes were directly bound by STAT4. Similar results were found for STAT6 in Th2 cells (112). Analysis of global STAT3 binding sites in Th17 cells also revealed thousands of genes that are bound by this protein (144, 188). Thus, the expectation would be that STATs have potential for having broad effects on the genome; this turns out to be the case.

STAT4 activates genes important for Th1 function such as Tbx21, Ifng, and Il12rb2 (189, 190). Likewise, STAT6 drives Gata3, Il4, Il4ra, and other genes necessary for Th2 cells (191). STAT5 plays a pivotal role in regulating Treg differentiation by directly binding the Foxp3 gene and inducing its expression (192). STAT3 positively regulates many genes important for Th17 cell formation and function, including Rorc, Il17a/f, and Il23r (188, 193). In addition, STAT3 is a strong activator of Il21, expressed by both Th17 and Tfh cells, suggesting why STAT3 has a role in both of these subsets. It should be emphasized that of the STAT-bound genes, only a subset of genes are highly dependent upon the cognate STAT. Overall, less than 25% of STAT-bound genes require STAT for gene regulation (112, 144). This makes sense however, when one considers that STATs are just one of many families of transcription factors in the genome.

In addition to their action in promoting gene expression, STAT deficiency has elucidated their role in inhibiting gene expression (194). Prior to the generation of ChIP-seq datasets however, it was not possible to discern direct versus indirect effects. A surprise from these new data was the extent to which STATs have a direct repressive function. STAT4 and STAT6 can both bind to target genes and reciprocally antagonize each other’s function in Th1 and Th2 cells. For example, Ccr8 is preferentially expressed in Th2 cells in a STAT6-dependent manner, whereas STAT4 inhibits this gene in Th1 cells (112). In Th17 cells, IL-2-activated STAT5 directly competes with IL-6-activated STAT3, binding the same genomic elements and thereby limiting gene expression (113). STAT5 can also bind and repress the Bcl6 locus, using the same manner of competition with STAT3 (134). STAT3 also functions to limit FoxP3 expression (86, 103, 107, 108, 195, 196); however in this case, the direct versus indirect actions of STAT3 have not been comprehensively defined. An example of the indirect action of STAT3 is its role in limiting IL-2 by induction of the Ikaros family protein Aiolos, which acts directly to represses the Il2 gene (154). Thus, STATs have multiple ways of mediating repression of T-cell subsets, either by directly competing with activators or by inducing expression of other transcriptional repressors.

Although STAT binding occurs at promoters, they also bind within genes, in 3′ untranslated regions and at distal sites. This suggested the possibility that in addition to just regulating transcription, STATs might also influence the epigenetic landscape in which genes reside. In fact, this is the case. STATs have major effects on both accessible and repressive marks. Globally, STAT4 has a significant effect on H3K4me3. In contrast, the major genome wide action of STAT6 had a greater effect on the repressive histone mark H3K27me3. In some cases, STAT6 deficiency promoted the appearance of repressive marks for genes expressed in Th2 cells (e.g. Il4, Gata3, and Il4ra), thus a major action of STAT6 is to remove repressive marks (112).

We now know that STATs can impact target genes in three ways: by affecting transcription and epigenetic modifications (~5%), by only affecting transcription (~11%), and by only affecting epigenetic modifications (~20%). For most STAT-bound genes however, there is no impact on either transcription or epigenetics (~64%). These observations suggest that STATs act on genes via multiple mechanisms, presumably in concert with other transcription factors (174).

Action of master regulators

The fact that the master regulators were both necessary and sufficient for differentiation of each corresponding T-cell subset suggested that the action of these transcription factors might encompass many aspects of regulating helper cell identity. In fact, T-bet regulates thousands of genes in Th1 cells (26, 197, 198). Similar to STAT4, T-bet binds to the Ifng locus (199) and promotes its expression, as well as the loci for Il12rb2 and Cxcr3 (115, 148, 200). Global profiling of T-bet binding and its impact on transcription and epigenetics has now been accomplished (201). Roughly 6% of genes bound by T-bet are transcriptionally regulated by this factor, but overall the number of genes positively or negatively regulated by T-bet are comparable. For genes expressed in Th1 cells, T-bet binding correlates with H3K4me1. For genes normally expressed in Th2 cells, T-bet binding is associated with H3K27me3 (201). Unlike T-bet, GATA3 is constantly available at almost all stages of T-cell development and in differentiated T cells. However, the distribution of GATA3 occupancy varies greatly in different stages of T-cell development (135). In Th2 cells, GATA3 is required for the expression of many Th2 associated genes, such as Il4, Il5, Il13, and Batf (202). Globally, 60% of the genes that require GATA3 for transcription also exhibit GATA3 binding, arguing for direct mode action in a relatively large proportion of genes (202). However, many genes were found to be negatively regulated: 30% of the genes repressed by GATA3 were bound by this factor in Th2 cells. These data suggest that like the STATs and T-bet, GATA3 can act as both a transcriptional activator and repressor. Interestingly, GATA3 regulates both H3K4me2 and H3K27me3 at many target genes. In the absence of GATA3, Il10 had more repressive marks, but at the Tbx21 and Ifng loci, reduced H3K27me3 marks were observed. The global effect of GATA3 deletion on H3K4me2 and H3K27me3 has also been shown (202). Thus, T-bet and GATA3 effect gene expression via changes in transcription and epigenetic modification.

Ideally, one would like to compare the global epigenomic impact of T-bet and GATA3 with STAT4 and STAT6 (112, 201, 202). It is clear that deletion of these factors has an impact on the epigenome. However, usage of different epigenetic marks as well different computational approaches makes a direct comparison difficult. No doubt this information will be available in the not too distant future.

Unlike T-bet and GATA3, Rorγt and Foxp3 have little effect on the epigenetic landscapes of their respective subsets (144, 203). Instead, their binding is highly enriched at promoters where they behave predominantly as direct activators or repressors. The designation as master regulator might imply broad acts on numerous genes. However, the new data argue that Rorγt has a relatively focused mode of action serving as modulator rather than a master transcription factor. In fact, Rorγt binding is associated with modest changes in gene expression in Th17 cells relative to Th0 cells (144). It positively regulates the expression of 68 genes but only strongly regulates a small number of core genes associated with the Th17 fate (e.g. Il17a, Il17f, and Il23r) (144). In addition, Rorγt acts as a repressor, limiting expression of genes such as Il10, Hif1a, Foxo1, and Il7r, as well as genes associated with alternate fates, such as Il4ra, and Il12rb2 (144). Thus, in contrast to other factors including STAT3, IRF4, or BATF, Rorγt’s direct functions are rather limited.

The absence of FoxP3 in Treg cells diminishes the expression of many Treg signature genes (204). However, gene expression is not abrogated, suggesting that additional factors contribute to regulation of expression. In fact, many partners of Foxp3 protein have been identified that may have a role in regulating Treg-specific gene expression (205). Like other factors, Foxp3 can either promote or repress gene expression both in Tregs and non-Tregs (192, 203, 206). Overall, similar numbers of genes are regulated by FoxP3, T-bet, or GATA3 (~200 genes) (192, 201-203). Although binding of FoxP3 is correlated with a variety of active and repressive histone marks, a major surprise is that deletion of Foxp3 does not have substantial impact on global chromatin status (203, 207). Rather, it appears that Foxp3 target genes are ‘prepared’ for Foxp3 arrival by other transcriptional factors, which act as ‘place holders’ (203, 207). The focused effect of FoxP3 is in line with its global enrichment at promoters (203).

Bcl6 is a transcriptional repressor that acts by binding to co-repressors and recruiting histone deacetylase complexes (208). Several co-repressors for Bcl6 have been identified, but the combinatorial effects of each of these co-repressors and the genes regulated by these complexes have not been elucidated (208). A handful of specific target genes have been identified in T cells including Prdm1 (which encodes the transcriptional repressor Blimp1), Tbx21, Gata3, Rorc, and Il5 (22, 209-211). Unfortunately, Chip-seq of Bcl6 in T cells has not been reported, and thus the genome-wide binding targets and effects of Bcl6 are unknown.

While it is clear that the master regulators are important for T-cell subset specific gene expression, it is apparent that they do not have pervasive effects on gene expression and chromatin landscapes; on the contrary, their effects are rather focused. In fact, current data suggest the master regulators control the transcription of a select group of gene modules associated with each subset while repressing a roughly similar number of genes important for other fates. Yet, it appears that the cell may already be ‘primed’ by other factors such as the STATs or additional transcription factors that mediate appropriate epigenetic changes, so that the master regulators can take advantage of this newly accessible landscape.