What Happens in the Thymus Does Not Stay in the Thymus: How T Cells Recycle the CD4⁺–CD8⁺ Lineage Commitment Transcriptional Circuitry To Control Their Function

T lymphocytes constitute a critical arm of the immune system and serve multiple functions in responses against both external and internal offenses. Conventional T cells recognize MHC–peptide complexes (pMHC) through a heterodimeric TCR comprising an α-chain and a β-chain (1, 2). Such αβ T cells are distributed into two subsets based on their expression of CD4 and CD8 surface molecules (hereafter referred to as coreceptors). CD4⁺ T cells, which recognize peptides bound to class II MHC (MHC-II), are traditionally referred to as “helper” cells (3, 4). Upon antigenic stimulation, they can adopt any of multiple specialized Th fates defined by unique cytokine- and transcription factor–expression patterns. CD8⁺ T cells, which express CD8α and CD8β molecules as CD8αβ dimers, recognize peptides bound to class I MHC (MHC-I). Contrasting with the polymorphism of helper cell differentiation, CD8⁺ T cells are heavily skewed toward cytotoxic effector differentiation and are responsible for eliminating infected or transformed cells.

CD4⁺ and CD8⁺ T cells develop from a common precursor through a differentiation process that has long served as a model for binary lineage decisions and is of interest from an immunological and a developmental standpoint. This common precursor, which expresses CD4 and CD8 and is thus called “double positive” (DP) originates from hematopoietic progenitors that have entered the thymus and initiated their development into T cells as CD4⁻CD8⁻ (“double-negative”) thymocytes (5–8). The developmental sequence that leads these progenitors to become DP thymocytes includes multiple differentiation and proliferation events that we will not discuss in this article. Critical for the CD4⁺-CD8⁺–differentiation decision is the rearrangement of the genes encoding TCRβ and TCRα. This allows the surface expression of TCRαβ complexes whose reactivity against pMHC expressed by the thymic stroma determines the death or survival of thymocytes (6, 9). Because of the broad allelic polymorphism among MHC molecules at the species level, most DP cells have little or no affinity for self-MHC ligands at the individual level; such cells die in the thymic cortex in a few days through death by neglect. At the opposite end, those thymocytes with high affinity for self-MHC, with the potential for causing autoimmune disease, are thought to be eliminated through active cell death (a process called negative selection); however, recent studies emphasize that a fraction of these cells is redirected toward regulatory or alternative functional fates (10, 11). As a result, only thymocytes with an intermediate affinity for self-pMHC survive, a process called positive selection, and become mature T cells.

In addition, the pMHC reactivity of positively selected thymocytes determines their choice of CD4⁺ versus CD8⁺ lineage, so that MHC-I–restricted DP cells become CD4⁻CD8⁺ “single-positive” (SP) thymocytes, whereas MHC-II–restricted DP cells become CD4⁺CD8⁻ SP thymocytes (4, 8). Such matching is important because CD4 and CD8 coreceptors facilitate TCR recognition of the appropriate class of MHC molecules and subsequent initiation of intracellular signaling (12). Furthermore, there is evidence that the thymic choice of CD4 or CD8 coreceptor expression is accompanied by “preprogramming” for helper or cytotoxic functions, respectively (13–15). From this final differentiation stage, SP thymocytes egress to the peripheral immune system as naive CD4⁺ or CD8⁺ T cells, prepared to respond to initial encounter with Ag. Although they can embrace multiple functional fates characterized by distinct gene-expression patterns, MHC-I– and MHC-II–restricted T cells retain the coreceptor they committed to in the thymus. This review discusses the mechanisms enforcing this “lineage stability” that are emerging as essential for proper T cell function. Before addressing these issues, a brief introduction to the transcription factors that promote the emergence of CD4⁺ and CD8⁺ lineages in the thymus is in order.

Work from several laboratories led to a model of CD4⁺–CD8⁺ lineage differentiation in which commitment to either lineage, biologically defined as the loss of the alternative developmental fate, is enforced by two transcription factors, Runx3 and Thpok, with mutually exclusive expression and opposite activities in thymocytes (16–18) (discussed below). However, other factors, which we refer to as specification factors, are involved in initiating expression of Thpok and Runx3, as well as additional lineage-specific genes. The CD4⁺ lineage specification factors, to which we return later in this review, include E-box binding proteins (E-proteins) E2A and HEB, Gata3, and three HMG proteins (Tox, Tcf1, and Lef1), with the last two being highly related and carrying out partially overlapping functions in developing T cells (19–24). Gata3 and Tcf1 bind the Thpok gene, suggesting a direct effect on its transcription (24–26). In addition to their role in CD4⁺ lineage specification, E-proteins, Gata3, and Tox are important for positive selection, whether MHC-I or MHC-II induced (21, 24, 26, 27), and Gata3 represses Runx3 in thymocytes (28). The impact of Tcf1 on the differentiation of DP thymocytes into SP thymocytes is even broader, because it contributes to Cd4 silencing in CD8⁺ lineage cells (24). The combined activity of these transcription factors lays the groundwork for Thpok expression and CD4⁺ lineage commitment, and an equivalent circuitry is thought to promote Runx3 expression and CD8⁺ lineage commitment (17, 29, 30).

Runx proteins are characterized by an N-terminal region of homology, the Runt domain, which binds specific DNA sequences (31). Runx1 and Runx3 are expressed in thymocytes and mature T cells and act as obligate heterodimers with the structurally unrelated CBFβ protein (31). Runx1 is expressed throughout thymocyte development and in CD4⁺ T cells, whereas Runx3 expression starts in differentiating CD8⁺ SP cells, after positive selection, and is maintained in mature CD8⁺ T cells (32–34). In the absence of Runx function (which, because of the functional redundancy between the two proteins, requires inactivation of both Runx1 and Runx3 or of CBFβ), MHC-I–restricted thymocytes fail to become CD8⁺ T cells and are redirected into the CD4⁺ lineage, severing the link between MHC restriction and coreceptor expression (34, 35). Runx3 directly contributes to CD8⁺ lineage differentiation by direct binding to Cd4 and Cd8 loci, which it represses and activates, respectively (32, 36). Runx1 and Runx3, together with Mazr, a zinc finger transcription factor related to Thpok, contribute to inhibit expression of Thpok, by binding a silencer located upstream of the Thpok promoter (35, 37–39).

The opposing transcription factor, Thpok, belongs to a large family characterized by a zinc finger–based DNA-binding domain and an N-terminal BTB motif that promotes dimerization and recruitment of additional transcriptional regulators (40, 41). Although not expressed in DP cells, Thpok is upregulated in MHC-II–selected thymocytes, and its expression persists in CD4⁺ SP thymocytes and mature CD4⁺ T cells (42, 43). Thpok is required for MHC-II–restricted thymocytes to become CD4⁺ T cells, although not for their positive selection and maturation. Disruption of Thpok, or a mutation in one of its zinc finger domains, causes the redirection of MHC-II–restricted thymocytes into the CD8⁺ lineage (26, 37, 42, 44, 45). Conversely, enforced expression of Thpok in DP thymocytes inhibits CD8⁺ T cell differentiation and redirects MHC-I–restricted cells toward the CD4⁺ lineage (42, 43). Thpok represses expression of Runx3 and Cd8 (37, 45–47) and binds to enhancers within the Cd8 genes (48, 49). In addition, Thpok antagonizes the activity of Runx proteins (37, 50), so that Cd4 and Thpok are subject to Runx-mediated transcriptional repression in thymocytes lacking Thpok molecules (37, 50). Thpok molecules bind the Cd4 and Thpok silencers, suggesting a direct antagonism of DNA-bound Runx (37), including of Runx1 in CD4⁺ lineage cells, although there is also evidence that Thpok promotes expression of Cd4 and CD4⁺ lineage genes through indirect mechanisms (50, 51).

Although it is not fully understood how they are matched to MHC restriction (52, 53), the mutually exclusive and opposite activities of Thpok and Runx3 promote CD4⁺ and CD8⁺ commitment in thymocytes. Because both factors remain expressed in postthymic T cells (16), the question arises whether they contribute to maintain the postthymic stability of CD4⁺ and CD8⁺ lineages.

Conceptually, two nonmutually exclusive processes can be envisioned to maintain lineage stability: an active transcriptional circuitry, possibly involving Thpok and Runx3, or epigenetic DNA or chromatin modifications, established during CD4⁺–CD8⁺ lineage differentiation in the thymus and subsequently inherited through cell division. In fact, as we will see below, T cells have borrowed from these two mechanisms to maintain CD4⁺ and CD8⁺ lineages in a way that fits requirements for stable, MHC-matched coreceptor expression, as well as for functional plasticity, which tailors cell responses to specific stimuli and environments.

Initial studies of Cd4 expression highlighted the importance of epigenetic control. At least four cis-regulatory elements contribute to control Cd4 expression in T cells (54–57): the promoter, near the transcriptional start site (TSS); the “proximal” enhancer, E4p, located 13 kb upstream of the TSS; the silencer, S4, in the first Cd4 intron; and a recently identified presumptive enhancer located near the silencer (Fig. 1). Genetic analyses by homologous recombination showed that Cd4 expression requires E4p in DP thymocytes but not in positively selected thymocytes and resting CD4⁺ T cells (58). However, neither E4p nor the promoter is involved in determining the CD4⁺ lineage specificity of Cd4 expression. Rather, this task falls to the silencer (55, 56), whose germline deletion causes ectopic expression of Cd4 in MHC-I–restricted CD8⁺ T cells (59, 60). Silencing of Cd4 in the thymus requires binding of Runx molecules and of additional silencing factors (including Tcf1 and Lef1) (24, 32, 61).

In contrast with its function in thymocytes, and remarkably, the silencer is dispensable to maintain silencing in mature CD8⁺ T cells, even through cell division (59). This indicates that the absence of Cd4 expression in CD8⁺ T cells reflects truly epigenetic silencing. This spectacular finding stimulated a long search for the silencing mechanisms. Epigenetic gene silencing largely relies on two processes: methylation of DNA itself (at position 5 of cytosine, typically within palindromic CpG dinucleotide pairs) (62, 63) and posttranscriptional modifications of histones, the core proteins permitting assemblage of eukaryotic chromatin into nucleosomes (64). Although there is no evidence that histone modifications contribute to Cd4 silencing, DNA methylation was recently shown to be a key player (65). CpG methylation, initiated by DNA methyl transferases Dnmt3a and Dnmt3b and transmitted during cell division by Dnmt1, is thought to promote gene silencing either by direct inhibition of transcription factor binding or by recruiting methyl CpG-binding proteins associated with transcription-repression complexes (66). Importantly, the symmetry of CpG motifs allows their methylation to be “copied” by Dnmt1 to newly synthesized strands during DNA replication, thereby ensuring truly epigenetic inheritance of this mark. Indeed, the Cd4 locus is hypermethylated in CD8⁺ T cells and hypomethylated in CD4⁺ T cells (65). Hypermethylation of the Cd4 locus is established early at the DN3 stage and maintained through the DP stage to CD8 SP, in a silencer-dependent manner. However, in mature CD8⁺ T cells, hypermethylation is silencer independent and remains stable throughout cell division. Furthermore, Dnmt1 (redundantly with Dnmt3a and Dnmt3b) is necessary to maintain Cd4 silencing during cell division (65). Thus, DNA methylation is critical to Cd4 silencing in CD8⁺ T cells.

Reciprocally, hypomethylation of the Cd4 locus is important for its proper expression in CD4⁺ T cells (65). Because the locus is methylated in DP thymocytes and because DP thymocyte are nondividing cells, hypomethylation involves removing methyl residues from CpG islands. Although there is no known cytosine demethylase that would perform such a function, a mechanism involving the Tet family of DNA dioxygenases has emerged as a functional equivalent (62). These enzymes convert 5-methyl cytosine to 5-hydroxymethylcytosine, leading to its excision and replacement with unmethylated cytosine. Although future genetic studies will be needed to confirm the involvement of Tet enzymes in Cd4 demethylation, analyses of CpG hydroxymethyl modifications in thymocytes are consistent with this possibility (65).

The mechanisms that install and maintain hypomethylation are coming to light. Although E4p is not needed for Cd4 expression except in DP thymocytes, its deletion in thymocytes, but not in postthymic cells, destabilizes Cd4 expression in proliferating postthymic CD4⁺ T cells (58). This suggests that E4p activity in the thymus facilitates demethylation, potentially by Tet-dependent mechanisms, installing an “open” epigenetic context that subsequently contributes to stable Cd4 expression. Of interest in this context is that E4p recruits the E-protein HEB (67), which, redundantly with the related factor E2A, promotes Cd4 expression in DP thymocytes; both factors also support CD4⁺ lineage specification (22, 23). This suggests that E-proteins could promote stable epigenetic opening of the Cd4 locus and affect postthymic Cd4 expression through their thymic functions. Additional experiments are needed to evaluate this hypothesis and determine the mechanisms at play. In agreement with a role for the specification circuitry, the demethylation of Cd4 in postselection class II–restricted thymocytes appears to be initiated independently of Thpok expression (65), and MHC-II–restricted Thpok-deficient T cells, which are CD4⁻CD8⁺ when resting but retain marks of CD4⁺ lineage specification, re-express CD4 upon activation (13)

Despite the importance of such Cd4 epigenetic opening, recent studies (discussed in the next section) indicate that the transcriptional circuitry is needed to maintain Cd4 expression in mature T cells (68). Thus, epigenetic mechanisms do not seem to be as important to maintain Cd4 expression in CD4⁺ T cells as they are to enforce its silencing in CD8⁺ T cells.

Although the Cd8 locus is methylated (69, 70) and subject to histone modifications (71), and although there is evidence for epigenetic control of its expression (71, 72), whether there is a similar “imprint” of thymic events on postthymic Cd8 gene expression remains to be determined. In fact, the current evidence points to a greater control by the transcriptional circuitry: Runx3 is important to sustain expression of Cd8 in CD8⁺ T cell effectors (72), whereas (as discussed below) Thpok is critical for its repression in CD4⁺ T cells (47, 68, 73).

Does the tentative dichotomy in the control of coreceptor expression, predominantly epigenetic for Cd4 and dynamic for Cd8, also apply to the associated programming for helper versus cytotoxic functions? Relevant to this question is the diversity of CD4⁺ T cell functional responses, corresponding to the variety of pathogens that these cells fight (74–76). Th1 cells produce IFN-γ in response to intracellular pathogens; Th2 cells fight extracellular parasites through production of IL-4, IL-5, and IL-13; and Th17 cells make IL-17 against extracellular bacteria and fungi. In addition, CD4⁺ T cells can postthymically upregulate Foxp3, adopting a regulatory fate and suppressing immune responses (77). Differentiation toward each of these fates is controlled by specific transcription factors, including T-bet, Gata3, and RORγt, in Th1, Th2, and Th17 cells, respectively, in conjunction with cytokine-specific signal transducers of the Stat family. The current perspective is that each naive CD4⁺ T cell emerging from the thymus, irrespective of its antigenic specificity, can potentially adopt any of these fates and preserves such potential by repressing expression of fate-specific transcription factors until Ag encounter (78). At that point, the cytokines generated by Ag-presenting and innate immune cells initiate transcriptional programs specific to each effector fate and, thereby, match T cell functional responses to pathogenic stimuli.

Of particular interest to the present discussion is that differentiation into Th1 (CD4⁺) and cytotoxic (CD8⁺) T cells, which target intracellular pathogens and produce IFN-γ, relies on similar key factors in both cell types (76, 79, 80): Stat4 and Stat5, activated by IL-12 and IL-2, respectively; a T-box–binding factor, typically T-bet in Th1 effectors and Eomes or T-bet in cytotoxic cells; and Runx3. Although Runx3 is CD8⁺ lineage specific in the thymus, it is expressed at similar levels in Th1 CD4⁺ effectors and cytotoxic CD8⁺ T cells, which raises puzzling issues. First, it implies that, although Thpok represses Runx3 in the thymus, the circuitry in Th1 effectors must be plastic enough to accommodate Thpok and Runx3 coexpression. Although Runx3 expression presumably involves cytokine signals that could be relayed by Stat5, as was reported in thymocytes (29), how it is enabled, despite concurrent Thpok expression, has not been determined and will require a better understanding of the mechanisms of Thpok-mediated Runx3 repression. Second, Th1 and Th2 differentiation are mutually antagonistic; notably, Runx3 represses IL-4 expression and Th2 differentiation (79, 81). Thus, the question arises whether preserving the Th2 potential of CD4⁺ T cells requires Thpok restraint of Runx3 expression. Last, because Th1 effectors express Cd4 and Thpok (47, 79, 80), they must counteract the repressive function of Runx3 on these genes, a task that was suggested to fall to Thpok molecules (37, 50).

A recent study shed light on the latter two points by evaluating Thpok functions through conditional deletion in naive postthymic CD4⁺ T cells, after their Thpok-dependent differentiation into CD4⁺ T cells in the thymus (68). These analyses showed that Thpok enforces Runx3 repression in naive CD4⁺ T cells and, thereby, a strict CD4⁺CD8⁻ coreceptor expression pattern. However, even after Thpok disruption and loss of Thpok protein, most MHC-II–restricted naive cells remain CD4⁺CD8⁻ and express little or no Runx3, suggesting that additional mechanisms, possibly epigenetic, contribute to control Cd4, Cd8, and Runx3 expression. Of note, the total number of MHC-II–restricted cells was not affected by postthymic Thpok disruption; thus, consistent with its dispensability for positive selection in the thymus, this factor is not essential for CD4⁺ T cell survival (13, 44, 47).

In addition to these effects in naive T cells, postthymic Thpok is needed for the proper effector differentiation of CD4⁺ T cells and specifically to restrain the expression of cytotoxic genes, including those encoding Runx3, Eomes, granzymes A and B, perforin, and CD8 (47, 68). In the absence of Thpok, this cytotoxic program takes over Th2 differentiation and strongly impairs IL-4 production, both in vitro and in vivo (68). This impact of Thpok disruption is consistent with the role of Runx3 in IFN-γ production during Th1 differentiation; however, and unexpectedly, Thpok is also needed for Th1 differentiation (68). Although production of IFN-γ per se does not depend on Thpok, this factor is necessary to prevent the cytotoxic “diversion” of Th1 effectors. Transcriptome analyses show that Thpok disruption causes an almost-complete Th1 to cytotoxic conversion of the effector response, accompanied by the acquisition of actual cytolytic functions. In addition to restraining the expression of cytotoxic genes, Thpok is essential for sustained expression of helper genes, first and foremost Cd40lg, which is critical for CD4⁺ help to B cells and dendritic cells (82).

Given the role of Runx3 in inhibiting Il4 expression, is Runx3 repression underpinning the protective role of Thpok in Th2 differentiation? That is indeed the case, because disruption of both Thpok and Cbfβ fully restores Th2 differentiation and prevents cytotoxic gene expression (68). In contrast, because Th1 effectors normally express Runx3, Thpok-mediated inhibition of Runx3 transcription could not account for its role in Th1 differentiation. Nonetheless, Cbfβ disruption showed that the expression of cytotoxic genes by Thpok-deficient “Th1” effectors, and, to a lesser extent, their repression of Cd40lg, are dependent on Runx3 (68). Thus, Thpok serves two Runx-related functions by which it preserves the potential diversity of CD4⁺ T cell responses (Fig. 2). The first is to restrain Runx3 expression and, thereby, enable cytokine activation of Th2 responses. The second, which operates in Th1 effector cells that normally express Runx3, constrains Runx3 functions and prevents its activation of the cytotoxic program (68).

Unexpectedly, and in contrast with its role in the thymus, the postthymic disruption of Thpok spares two genes characteristic of the CD4⁺ lineage: Cd4 itself and Thpok (assessed by a GFP-based reporter transgene) (47, 68). Functional redundancy with LRF, a zinc finger transcription factor closely related to Thpok that also inhibits Runx-mediated Cd4 repression (13, 68), underpins Thpok-independent Cd4 expression in CD4⁺ T cells. Indeed, the postthymic disruption of Thpok and LRF causes CD4⁺ T cells to fully convert into CD4⁻CD8⁺ cells (68). Because Cd4 hypomethylation correlates with stable expression in CD4⁺ T cells (65), it will be interesting to determine whether Thpok and LRF disruption results in remethylation of the Cd4 locus or allows its repression by Runx molecules while remaining unmethylated. In any case, these findings emphasize that active epigenetic marking of Cd4 is not sufficient for its sustained expression in CD4⁺ T cells, for which the overlapping functions of Thpok or LRF are also needed.

Contrary to Cd4, sustained expression of the Thpok gene in MHC-II–restricted T cells requires neither Thpok nor LRF molecules, despite the inhibitory potential of Runx3 (68). Whether additional transcription factors, or lack of transcriptional repressors, support Thpok expression or whether it is maintained epigenetically remains to be determined.

With the exception of the epigenetic maintenance of Cd4 silencing, the mechanisms that postthymically maintain CD8⁺ lineage differentiation are less well understood. Additional evidence for epigenetic repression of the helper program in CD8⁺ T cells comes from ectopic (retroviral) Thpok expression in postthymic CD8⁺ T cells (46). Although Thpok impairs the expression of CD8, cytotoxic, and IFN-γ genes, mirroring the effects of Thpok disruption in CD4⁺ T cells (47, 68), it only modestly upregulates markers of helper function (including Gata3 and IL-4), with little or no effect on Cd40lg and none on Cd4. Although they do not exclude that transcription factors critical for helper gene expression are missing in CD8⁺ T cells, these observations support the idea that epigenetic mechanisms similar to those at work at the Cd4 locus impose broad restraints on CD4⁺ helper differentiation in CD8⁺ T cells. Such epigenetic control could involve establishment of silencing marks (e.g., DNA methylation), or removal of active marks, during CD8⁺ T cell differentiation in the thymus. Of note, the repression of Thpok itself in CD8⁺ T cells is not as stringent as that of Cd4, because it is expressed upon Ag-induced activation (25, 83). Thpok amounts in activated CD8⁺ T cells are much lower than in CD4⁺ T cells and were reported to promote cytotoxic responses and memory CD8⁺ T cell differentiation (83) through mechanisms that remain to be defined.

Conversely, Runx complexes are not needed to maintain CD8 expression in activated cytotoxic effectors (72), and, similar to the Cd4 silencer itself, Runx3 was reported to be dispensable to maintain postthymic Cd4 silencing (58). However, the postthymic functions of Runx3 in CD8⁺ T cells remain to be clarified by in vivo stage-specific inactivation. In large part, this is because Runx3 is not strictly required for CD8⁺ T cell development (owing to the functional overlap with Runx1) (32–34), so that studies of Runx3 functions in mature CD8⁺ T cells have primarily been conducted using genetic models that delete Runx3 in the thymus. These experiments provided evidence that Runx3 is important for cytotoxic gene expression in CD8⁺ effectors and, particularly, suggested that it promotes the expression of T-bet and Eomes (84). Of note, the disruption of T-bet and Eomes causes the diversion of cytotoxic into IL-17–producing responses (85), and it remains to be determined whether the disruption of Runx activity would have similar consequences.

The preceding discussion highlights that the Thpok-Runx3–based loop that decides CD4⁺–CD8⁺ lineage commitment in the thymus is also involved in maintaining CD4⁺ lineage integrity and helper responses after thymic egress (Fig. 2). The critical importance of these functions is highlighted by the fact that postthymic deletion of Thpok in CD4⁺ T cells impairs long-term control of Toxoplasma gondii, resulting in the death of infected animals (M.S. Vacchio and R. Bosselut, unpublished observations). Whether postthymic functions of Thpok are needed for differentiation into other effector fates (e.g., Th17 or T follicular helper) remains to be determined. Current evidence indicates that Thpok and LRF are not needed for Th17 differentiation or to restrain the expression of IFN-γ by Th17 cells, suggesting that the Th17 circuitry efficiently represses IFN-γ (68). Contrasting with this broad impact of Thpok in CD4⁺ T cells, the available evidence, although limited, suggests that Runx3 contributes to the expression of cytotoxic genes in CD8⁺ T cells but not to the repression of CD4⁺ lineage genes, although it serves such functions in thymocytes or in Thpok-deficient CD4⁺ T cells.

Continued involvement of Thpok and Runx3 beyond the thymus also contrasts with stage-specific functions of some of the CD4⁺ lineage specification factors mentioned earlier, including Tcf1 and Lef1, or Gata3. Postthymic disruption of Tcf1 and Lef1 compromises T follicular helper differentiation but does not affect overall CD4 expression or other helper functions (86). Similarly, disruption of Gata3 in mature CD4⁺ T cells impairs Th2 differentiation but was not reported to affect CD4⁺ lineage integrity (87).

It is also important to emphasize that, as illustrated by a recent study of histone deacetylase (HDAC) functions in T cells (71), epigenetic and active transcriptional control collaborate to maintain appropriate gene expression. Acetylation of histone H3 on lysine residues 9 and 27, typically at gene promoters and enhancers, is associated with increased transcriptional activity (88). HDACs are part of large transcriptional-repression complexes, some of which associate with members of the BTB zinc finger family (41, 89), and their recruitment dampens gene expression. Inactivation of HDAC1 and HDAC2 in the thymus causes the conversion of CD4⁺ effectors into CD4⁺CD8⁺ cells expressing cytotoxic genes and Runx3 and producing IFN-γ (71). Accordingly, activated CD4⁺ T cells from HDAC1 and HDAC2–deficient mice show increased H3K9 acetylation at the promoters of cytotoxic genes and at a Cd8 enhancer active in effector cells. Although these experiments assessed the consequences of thymic HDAC inactivation, treating postthymic CD4⁺ T cells with an inhibitor of HDACs mimicked the effects of HDAC1 and HDAC2 deletion, indicating a persistent need for HDACs to repress CD8⁺ lineage genes (71).

The mechanistic bases for these effects of HDAC1 and HDAC2 remain to be elucidated. The parallel with Thpok functions suggests that Thpok-mediated repression of cytotoxic genes could involve its recruitment of class I HDACs (HDAC1 and HDAC2), similar to its homolog Bcl6 (89). However, it was reported that Thpok binds class II HDACs (HDAC4, HDAC5, and HDAC10), which it recruits to the Cd8 locus, but not HDAC1 and HDAC2 (48); additionally, it is possible that Thpok and HDACs serve independently to repress CD8⁺ lineage expression. Thus, a comprehensive picture of Thpok association with HDACs has not yet emerged. Further highlighting the challenges of integrating such studies into a unified conceptual framework, HDACs act on nonhistone acetyl-lysine substrates. A comparison of mass spectrometry and mRNA expression analyses showed that HDAC1 and HDAC2 inactivation affected almost 200 genes at the protein, but not at the mRNA, level. Thus, it is possible that re-expression of CD8⁺ lineage genes in HDAC-deficient CD4⁺ T cells is mediated by deacetylation of nonhistone proteins. Although such substrates could include Thpok and Runx3, both of which were shown to be acetylated and, thereby, protected from ubiquitin-mediated degradation (90, 91), the impact of such modifications in vivo remains to be assessed.

Although Thpok preserves the functional diversity of helper effector responses by restraining the cytotoxic program, there is evidence that CD4⁺ T cells can acquire cytotoxic functions, especially during chronic infections (reviewed in Refs. 92, 93). These CD4⁺ T cells are distinct from Th1 cells, with upregulated Eomes, as well as cytolytic activity, mediated by perforin and Fas. Such cells could function to eliminate pathogen-infected cells that express MHC-II during infection (94) or tumor cells (95) or conceivably to dampen Ag presentation to avoid excessive immune responses.

It remains to be determined whether acquisition of cytotoxic gene expression implies the cessation of Thpok expression or antagonism of its functions. A potential scenario has emerged from studies of a specific subset of intraepithelial lymphocytes (IEL), defined by the coexpression of CD4 and CD8α but not CD8β (49, 73). Thus, these cells differ from the larger IEL subset expressing CD8α but not CD4 or CD8β (96). Most such CD4⁺CD8α⁺ IEL fail to express Thpok, and adoptive-transfer experiments, together with elegant fate-mapping strategies, suggest that at least some of them derive from bona fide MHC-II–restricted CD4⁺ T cells that have terminated Thpok expression, driven by TGF-β and retinoic acid signaling in the intestinal milieu (73). Although such postthymic Thpok silencing was shown to depend on Mazr and the Thpok silencer, it was only partially dependent on Runx3 expression (73). The latter observation suggests that mechanisms distinct from the Runx-mediated Thpok repression operate during thymic selection. Consistent with this idea, the double disruption of Thpok and Lrf, despite promoting Runx3 expression, does not impair Thpok gene activity in CD4⁺ effectors, unlike its effect on Cd4 (68).

Determining whether postthymic Thpok silencing underpins the differentiation of “conventional” cytotoxic CD4⁺ effectors during chronic infections is an important goal for future research. Such cells do not express CD8α, and their Thpok-expression status has not been reported. Conversely, neither the antigenic specificity of CD4⁺CD8α⁺ IEL nor their function has been definitely identified. Although there is evidence that at least some CD4⁻CD8α⁺ IEL derive from self-reactive thymocytes escaping deletion by negative selection in the thymus (10, 11), it remains to be determined whether the same is true of CD4⁺CD8α⁺ IEL. Functionally, they could serve as a first line of defense against viruses trophic for class II–expressing cells in the gut or carry regulatory functions against “pathogenic” inflammatory Th17 or Th1 cells (49, 73).

Additional studies in nonhuman primates (including African green and patas monkeys) reported a shift in class II–restricted T cells from their normal CD4⁺CD8⁻ phenotype to a CD4⁻CD8^lo phenotype (97, 98). Such conversion is only observed in memory cells (naive cells remain CD4⁺CD8⁻) and is accompanied by impaired expression of helper-specific genes at the single-cell level. However, at the organism level, the cessation of CD4 expression is thought to prevent viral dissemination after SIV infection and, therefore, to be advantageous by preserving the MHC-II–restricted repertoire. The transcriptional control of this conversion and whether it involves Thpok or LRF remain to be determined.

Despite their potential importance, such changes in coreceptor expression are the exception rather than the rule, as most T cells strictly maintain coreceptor expression matched to their MHC specificity. This is highlighted by the strong epigenetic Cd4 silencing in CD8⁺ T cells. However, there is evidence that the epigenetic silencing of Cd4 is not as stringent in human as in mouse CD8⁺ T cells, despite conservation of the Cd4 silencer in the human genome (99). Indeed, naive human CD8⁺ cells from adults or neonates can reinduce CD4 expression upon long-term activation (100, 101), and this was proposed to make them susceptible to HIV infection. However, such “leakiness” of CD4 expression was limited, because most CD8⁺ T cells remained CD4⁻ in such experiments, and it is not known whether “leaky” cells acquired helper functions together with CD4 expression.

Why would CD4 re-expression be so detrimental that CD8⁺ T cells have developed redundant mechanisms to avoid it? It can be envisioned to have distinct consequences based on the cell’s reactivity to MHC-II. For most CD8⁺ T cells carrying TCR unable to cross-react with MHC II–peptide complexes, the main impact of CD4 re-expression is on TCR signaling. Both CD4 and CD8 bind the tyrosine kinase Lck, which initiates intracellular TCR signaling upon pMHC engagement (2, 102). However, Lck binds with greater affinity to CD4 than to CD8; as a result, in cells expressing both CD4 and CD8, CD4 molecules trap a large fraction of Lck away from CD8 and, therefore, diminish the contribution of CD8 to MHC-I–induced TCR signaling (103). Additional and deeper consequences are predicted in CD8⁺ T cells carrying TCR cross-reacting with MHC II–peptide complexes, in which Cd4 re-expression could increase MHC-II–induced TCR signaling. Recent studies suggest that a substantial fraction of thymocytes carry TCR that can recognize both MHC I– and MHC II–peptide complexes and, therefore, possess the potential for cross-reactivity (11, 104). Although most such cross-reactive cells die by negative selection in the thymus, some escape deletion, as shown by the existence of MHC-II–selected CD4⁺ T cells that can bind MHC I–peptide complexes (104). Acquisition of MHC-II reactivity by CD8⁺ T cells has the potential to allow their inappropriate activation by MHC-II–expressing APC and lead to autoimmunity.

Several conclusions arise from the studies summarized in this review. First, the current evidence favors asymmetric control of Cd4 and Cd8 expression in postthymic cells. The transcriptional circuitry remains a key controller of Cd8 expression, in contrast with Cd4, for which silencing in CD8⁺ T cells and stable expression in CD4⁺ T cells have a strong epigenetic component, involving cytosine methylation.

Second, the core components of the thymic commitment circuitry, Thpok and Runx proteins, remain involved in maintaining postthymic T cell lineage integrity (Fig. 2). The current evidence suggests that Runx3 in CD8⁺ T cells promotes expression of the cytotoxic program but not repression of helper or CD4⁺ lineage genes. This contrasts with the role of Thpok (in part redundantly with LRF) as both a repressor of the cytotoxic program and a “protector” of Th1 and Th2 helper differentiation. Although it remains to be determined whether Thpok is also needed for additional helper fates (including Th17 and T follicular helper), the emerging picture is that postthymic Thpok maintains the ability of naive CD4⁺ T cells to adopt any of the helper fates upon activation. The latter appears to be a key requirement for a functional adaptive immune system, because there is no a priori link between antigenic specificity and functional responses (determined by the type of pathogen involved). In contrast with these functions of Thpok, the persistent expression of Runx3 in CD8⁺ T cells strongly biases their responses toward cytotoxicity, suggesting that only strong cytokine-induced signals can overcome such a bias and direct CD8⁺ T cells toward alternative fates, similar to those adopted by effector CD4⁺ T cells, including IL-17 production (reviewed in Ref. 105). Whether Runx3 similarly inhibits the adoption of such alternative functional fates remains to be assessed.

Recent studies highlighted the similarities between T cell effector programs and those elicited in innate lymphoid cells (ILC) (76, 106). A major difference between these two cell types is that the function of ILC does not require Ag recognition by a clonotypic receptor. Accordingly, ILC acquire effector-specific transcriptional programs (e.g., promoting IFN-γ production for ILC1 or cytotoxic gene expression for NK cells) during their development (106, 107), presumably because no benefit would be expected from maintaining these cells in a naive status with multifunctional potential. From this perspective, functions similar to those of Thpok in CD4⁺ T cells would not be needed once ILC have completed their development, an hypothesis that remains to be experimentally validated.

The critical role of Thpok in CD4⁺ T cell effector differentiation raises the question of whether it could be targeted to inhibit or redirect effector differentiation for therapeutic purposes. A correlate of this question is whether Thpok also controls CD4⁺ lineage integrity in human cells. Thpok and LRF amino acid sequences, as well as those of Runx proteins, are highly conserved among mammalian species, and there is evidence for preferential CD4⁺ lineage Thpok expression in human thymocytes and T cells, although possibly not as pronounced as in mice (108, 109). However, there are also differences in the permissivity of Cd4 silencing and expression of CD40L (110) in human CD8⁺ T cells, as well as in the lineage plasticity in nonhuman primate T cells described above. Determining the role of Thpok in human T cell responses should be facilitated by the emergence of tools for genome editing and would be a first step toward exploring the potential of Thpok as a target for therapeutic manipulation of immune responses for human disease.

We thank P. Love and T. Ciucci for useful discussions and J.D. Ashwell, P. Love, and T. Ciucci for reading the manuscript. We apologize to colleagues whose work could not be discussed because of space limitations. Research work in the authors’ laboratory is supported by the Intramural Research Program of the National Cancer Institute, Center for Cancer Research.

The authors have no financial conflicts of interest.