One of the best-studied epigenetic mechanisms of heritability is the covalent modification of cytosine to 5mC, a mark deposited by the DNA methyltransferase (DNMT) enzymes. DNA methylation occurs predominantly at cytosine residues that are followed by guanine (CpG) in mammalian genomes, and about 60–80% of CpGs are methylated in somatic tissues (13). The classic model of DNA methylation holds that de novo DNA methylation is deposited in the genome by Dnmt3a and Dnmt3b along with their non-enzymatic co-regulator Dnmt3L (14, 15). Maintenance DNA methylation is carried out by Dnmt1, which associates with the replication fork through PCNA and with hemimethylated CpGs through the E3 ubiquitin ligase Uhrf1 during DNA replication (16–18). However, these distinctions are not absolute as Dnmt1 has been shown to exhibit de novo methyltransferase function, and Dnmt3 can participate in the maintenance of methylation marks (19). Also, as discussed later, the model of DNA methylation was further revised with the discovery of an active enzymatic process of demethylation.

In the 1970s, two laboratories hypothesized that DNA methylation could act as a cellular mechanism of transcriptional memory through cell division due to the symmetrical nature of the CpG dinucleotide (20, 21). Since then, DNA methylation has been shown to be critical for genomic imprinting, X chromosome inactivation, and long-term repression of mobile genetic elements (22). Mechanistically, DNA methylation can lead to gene silencing by inhibiting the binding of factors that activate transcription through the addition of methyl groups in the major groove of the double helix or through the recruitment of repressive complexes (13). For example, the binding of CTCF, an insulator protein involved in the formation of chromosomal domains, is inhibited by DNA methylation, allowing enhancer-mediated activation of the paternal allele at the imprinted H19/Igf2 locus (23). DNA methylation can also mediate gene repression through methyl-CpG-binding domain proteins that bind to 5mC. Some of these proteins such as Mbd2 and Mbd3 have been found to associate with the nucleosome remodeling and deacetylase (NuRD) complex (nucleosome remodeling and histone deacetylation), thus linking DNA methylation with other epigenetic mechanisms such as histone deacetylation and nucleosome positioning (24, 25). Functions beyond gene silencing have also been reported, such as genomic DNA methylation defining introns and exons during splicing (26, 27).

Although DNA methylation is a stable epigenetic mark that can be propagated across cell divisions, DNA demethylation had been observed in different biological contexts (28).

In contrast to passive demethylation, which results in the loss of 5mC during successive rounds of replication in the absence of DNA methylation maintenance machinery, replication-independent demethylation processes had also been observed invoking the possibility of active enzymatic removal of 5mC marks (29). The discovery of the ten eleven translocation (TET) family of enzymes that can modify 5mC through iterative oxidation converting 5mC to 5-hydroxymethyl cytosine (5hmC), and subsequently to 5-formylcytosine and 5-carboxylcytosine (5caC), and the detection of these intermediates in vivo in mammalian DNA, were critical in elucidating the mechanism of active DNA demethylation (30–32). There are three TET family members in mammals (Tet1/2/3), each possessing a core catalytic domain at the carboxyl terminus, a double-stranded β-helix fold containing crucial metal-binding residues, and a CpG DNA-binding CXXC domain toward the amino terminus of the protein (33). Although Tet1 and Tet3 contain an internal CXXC domain, Tet2 partners with IDAX, an independent CXXC-containing protein (34). Following iterative oxidation of 5mC to 5caC by the TET enzymes, unmodified cytosines can be generated via passive dilution of the oxidized base or enzymatic removal via thymine DNA glycosylase and the base excision repair pathway (35–38).

In eukaryotes, DNA is wrapped around an octamer of histones composed of two copies of H2A, H2B, H3, and H4. The linker histone H1 is present where the DNA enters and exits the core nucleosome, and this nucleosome filament can be further compacted, facilitating the packing of large eukaryotic genomes within the limited space of the nucleus (39). Covalent changes in histones through posttranslational modifications (PTMs) are critical to epigenetic processes as they can alter chromatin structure or allow recruitment of activators or repressors of transcription (40). Examples of PTMs include acetylation, methylation, phosphorylation, ribosylation, and ubiquitination although it is unclear whether all of these are self-propagating through cell division, a typical criterion of epigenetic processes. One well-studied modification is histone acetylation at lysine residues mediated by histone acetyltransferases (HATs). Acetylation is typically associated with gene activation, whereas histone deacetylase enzymes (HDACs) that catalyze the hydrolysis of acetyl-lysine residues generally result in silencing and chromatin compaction (41–43). Histone methylation can also occur on lysine and arginine residues, and the methylation of distinct lysines on the histone H3 tail is associated with either gene silencing or activation. H3K9me2/3 is a conserved hallmark of heterochromatin among eukaryotes and serves as a binding site for heterochromatin protein 1 (HP1) family members to facilitate heterochromatin integrity (44). Furthermore, H3K9me3 also interacts with the DNA methylation machinery, and cooperation between these silencing marks may be vital for heritable silencing. For instance, Uhrf1 was found to bind H3K9me3 marks and facilitates H3 ubiquitination via its RING domain for recruitment of Dnmt1 at the replication fork (45–47). Deacetylation of histones can also provide H3 lysine substrates for the histone methyltransferase G9a, which also facilitates de novo DNA methylation at gene promoters through the recruitment of Dnmt3a and Dnmt3b in ES cells (48, 49). Interestingly, the histone methyltransferase activity of G9a is dispensable for de novo methylation at G9a-target gene promoters in ES cells (50). A more direct connection between DNA methylation and H3K9 methylation exists in the fungus Neurospora crassa in which the H3 histone methyltransferase dim-5 is required for DNA methylation through trimethylation of H3K9 (51). Other repressive histone marks include H3K27me2/3, mediated by the polycomb group 2 complex (PRC2), modifications associated with the silencing of genes during developmental decisions (52, 53). In contrast to K9/K27 methylation, H3K4 trimethylation (H3K4me3) is associated with active promoters, whereas H3K4 monomethylation (H3K4me1) is often a mark of enhancer function (54). H3K4me3 was also suggested to inhibit DNA methylation, as the de novo DNA methylation co-factor Dnmt3L was blocked from binding methylated H3K4 (55). Several histone modifications have also been found to be reversible by enzymes such as the histone demethylase Lsd1 that removes monoethyl or dimethyl marks from H3K4 (H3K4me1/2) (44).

Non-covalent histone modifications can also impact the state of chromatin through changes in nucleosome positioning or altered exposure of DNA along the nucleosome through ATP-dependent nucleosome remodelers. Indeed, nucleosome remodelers are critical for chromatin assembly and enable dynamic changes in genomic architecture in response to signaling or differentiation (56). Many of these complexes can both activate and repress transcription, possibly through combinatorial assembly of different subunits (57). Their specificity is thought to stem from recruitment by histone modifications, transcription factors, or DNA methylation, and therefore they have an intimate connection to epigenetic processes. Thus, both covalent and non-covalent histone modifications extend the capacity of chromatin to influence gene expression.

Variants of the core histones H2A, H2B, and H3 and the linker histone H1 exist and differ from their canonical histone counterparts in primary sequence and expression pattern. In contrast to S phase-coupled expression of the canonical histones, histone variants are expressed throughout the cell cycle (39). Substitution of canonical histones with these variants is thought to modulate gene expression by altering the physical properties of nucleosomes through changes in PTMs, chromatin structure, or recruitment of additional co-activators or repressors (58). Some examples of histone variants that have a putative role in transcription include the H2A variant H2A.Z, which is enriched at the transcription start site (TSS) of active genes, and the H3.1/H3.2 variant H3.3, which has been suggested to have roles in both gene activation and heterochromatin integrity (59, 60).

Although epigenetic marks are required for proper regulation of gene expression, an area of active interest is the contribution of small and long non-coding RNAs (ncRNAs) in directing chromatin modifications. In yeast, plants, and Drosophila, a role for small ncRNAs has been implicated in heterochromatin formation, which is disrupted upon interference with the RNAi machinery (61–63). Whether a similar pathway operates in mammals is still unclear as studies using genetic knockout of the endoribonuclease Dicer in ES cells, which is required for the biogenesis of siRNAs and microRNAs, resulted in different conclusions, with one study showing no effect on histone modification or DNA methylation status (64, 65). The PIWI-interacting RNA (piRNA) pathway has also been demonstrated to regulate heterochromatin in Drosophila by directing H3K9 methylation to transcriptionally silence transposons and interacting with HP1 (66, 67). In mice, the piRNA pathway seems to be required for sequence specific de novo methylation of the imprinted Rasgrf1 locus in the male germline (68). The epigenetic landscape can also be shaped via site-specific chromatin modifications mediated by long ncRNAs (lncRNAs). To date, X chromosome inactivation is the canonical model for the epigenetic regulation by lncRNAs. The X-inactive-specific transcript (Xist) is transcribed from the inactive X chromosome in female mammalian cells and has been shown to mediate silencing of the X chromosome in cis through the direct recruitment of PRC2 and YY1 (69, 70). Since the discovery of Xist, additional lncRNAs have been implicated in the modulation of epigenetic processes at other loci through interaction with chromatin modifiers (71). However, the in vivo functions of many lncRNAs have not yet been evaluated.

Taken together, epigenetic changes are dictated by numerous mechanisms that work in a combinatorial manner (Figure 1). An understanding of how these processes work in combination and how they are deployed for heritable transmission of gene expression has proven challenging. As we review below, studies of Cd4 and Cd8 regulation during lineage commitment may help to shape our basic understanding of these processes in a system that is highly tractable.

The E4_P element was mapped to a 339 bp region of DNA that directed reporter activity in T cell lines and in vivo in T cells of transgenic reporter mice (76, 77). Germline targeting of E4_P prevented CD4 upregulation during the DN–DP transition in the thymus, demonstrating its requirement for CD4 expression in preselection DP cells (73). After positive selection, CD4 was reexpressed on MHC class II selected cells, although at reduced levels compared to wild-type cells. Furthermore, CD4 expression was unstable during cell division, as a large proportion of CD4⁺ T cells in the periphery lost CD4 expression after activation. In these cells, H3 acetylation and H3K4Me3 within the Cd4 locus were also reduced. The loss of H3Ac and H3K4Me3 was observed in CD4⁻ DP stage thymocytes, consistent with the loss of CD4 expression. To investigate whether E4_P was required for regulating CD4 expression beyond the DP stage, Cre-mediated excision of a conditional allele was performed. In contrast to germline deletion, conditional deletion of E4_P in mature CD4⁺ T cells did not impair CD4 expression during activation, and the levels of H3Ac and H3K4Me3 were similar to wild-type cells. This suggested that, during thymic development, E4_P modulates a stable epigenetic state of the Cd4 locus that can be maintained in its absence thereafter.

Sequence analysis and nuclear extract binding studies suggested that the HMG-box transcription factors T cell factor 1 (Tcf1)/Lef1 and the basic helix-loop-helix (bHLH) proteins E2A/HEB bind to motifs within E4_P (76, 78). Accordingly, gene targeted inactivation of the genes encoding E2A and HEB impaired CD4 expression during the DN–DP transition (79, 80). Similarly, mutation of Tcf7 (which encodes Tcf1) also reduced CD4 expression in DP thymocytes (81). ChIP analyses confirmed that these factors bound to E4_P, suggesting that they directly promote expression of CD4 in DP thymocytes (81, 82). However none of these factors is individually required for CD4 expression in DP thymocytes, as single deletions do not phenocopy E4_P^−/− mice. Interestingly, the NuRD complex, typically associated with gene repression, was also implicated in CD4 activation (83). Deficiency in mi-2β, an ATPase chromatin remodeling subunit of the NuRD complex, led to impaired CD4 upregulation during the DN–DP transition (83). mi-2β also bound E4_P by ChIP assays and co-immunoprecipitated with HEB and the HAT p300, linking histone acetylation with factors recruited to E4_P. The cooperative mechanisms by which these transcription factors drive transcription of Cd4 and their contribution to epigenetic programs are yet to be assessed.

The observation that CD4 was re-expressed following positive selection in E4_P^−/− thymocytes suggested the existence of another regulatory element, now referred to as the “maturation enhancer,” E4_M. The existence of such an element was also supported by a previous study employing a Cd4 transgene containing the Cd4 first intron, promoter, and E4_P element, whereby transgenic reporter activity was lost in activated mature T cells without changes in endogenous Cd4 expression (84). Indeed, our recent ATAC-seq analysis of the Cd4 locus in CD4SP thymocytes revealed a previously uncharacterized chromatin-accessible region situated in the first intron of the gene, 3′ to the silencer (Priya D. A. Issuree and Dan R. Littman, unpublished results). The potential role of this region was highlighted by Egawa and colleagues, who showed that CD4 expression was unstable in T cells from mice with targeted deletion of 1.5 kb encompassing the silencer and the region 3′ to it, but was stable when only the core silencer was deleted (75). Our studies using mice lacking E4_M revealed that this element controls Cd4 expression in postselected CD4SP thymocytes, and in the absence of both E4_M and E4_P, there is a complete loss of CD4 in T cells (Priya D. A. Issuree and Dan R. Littman, unpublished results). As observed in the absence of E4_P, deletion of E4_M resulted in the gradual loss of CD4 expression by activated CD4⁺ T cells following multiple rounds of cell division [(75) Priya D. A. Issuree and Dan R. Littman, unpublished results]. Analogous to E4_P, retroviral Cre-mediated excision of this region in mature T cells did not diminish CD4 expression, suggesting that E4_M may be required for establishing an epigenetically active state during development. However, it remains possible that both enhancers function cooperatively in the establishment of stable CD4 expression and are insufficient to do so individually. Importantly, these studies demonstrate that the main enhancers dictating CD4 expression in αβ T cells are E4_P and E4_M, and no additional elements are capable of compensating for their loss. The minimal region encompassing E4_M activity and the transacting factors recruited there to drive expression are currently under investigation. It also remains to be determined whether the lack of E4_M without a larger intronic deletion results in redirection of MHC class II selected cells into the CD8⁺ lineage, as was observed in mice with the 1.5-kb intronic deletion (75).

The silencer (S4) is a 434 bp core element in the first intron of Cd4 that was identified by its ability to inhibit expression from reporter constructs in cell lines and in both DN thymocytes and CD8⁺ T cells of transgenic mice (77, 85). Deletion of this element in the germline of mice caused CD4 derepression in DN thymocytes and CD8⁺ cytotoxic T cells, confirming its physiological function (86, 87). Importantly, Cre-mediated excision of a conditional LoxP-flanked S4 allele in mature CD8⁺ T cells did not reverse CD4 repression (87). The results demonstrated that S4 activity is reversible during the DN–DP transition, but becomes irreversible following positive selection, in mature cytotoxic T cells. Therefore, CD4 silencing in peripheral cytotoxic T cells is independent of S4 and mediated by a heritable silencing mechanism after CD8⁺ lineage commitment. To determine when S4-independent repression occurred, stage-specific S4 deletion with different Cre transgenic mice was carried out. By using CD4-Cre, which drives Cre expression in the late DN and early DP stages of development, DN thymocyte repression was intact but mature CD8⁺ T cells derepressed CD4 similarly to the germline deletion (87). S4 deletion with CD8-E8_I Cre, which is active only in MHC class I-selected T cells, also caused CD4 derepression, suggesting that heritable Cd4 silencing occurs after positive selection. Deletion of S4 at additional stages after MHC class I-positive selection may provide more insight into the initiation of heritable CD4 silencing.

In contrast to uniform CD4 derepression in S4^−/− mice, variegated CD4 derepression occurred with smaller deletions in S4 (88). This phenotype was reminiscent of position-effect variegation (PEV) studied extensively in Drosophila. PEV occurs when a gene is stochastically silenced in a population of cells due to spreading of heterochromatin from an adjacent locus (89). In agreement with the spreading of heterochromatin, CD4 repression was stable in cells that achieved silencing and the frequency of repression could be increased by overexpressing HP1β in vivo (88). There was no variegation observed in DN cells in which CD4 derepression occurred following mutations in S4. Instead, there was a partial but uniform derepression, consistent with active repression and the contribution of multiple transcription factor binding sites toward this process (88). Motif analysis revealed two sites in S4 that contained consensus binding motifs for members of the evolutionarily conserved family of Runx transcription factors (see Runx) (90). Deletion of both sites caused CD4 derepression in CD8⁺ T cells, and interaction between S4 and the Runt domain of Runx1 was found in a yeast-one-hybrid screen (91). Runx1 is expressed at the highest level in DN thymocytes, whereas Runx3 is predominantly expressed in CD8SP cells. Consistent with their expression pattern, loss of function studies demonstrated that Runx1 is indispensable for active repression of Cd4 in DN thymocytes, whereas Runx3 is responsible for establishing Cd4 silencing in CD8⁺ T cells (91–94). Interestingly, deficiency in the nucleosome remodeling Brg/Brahma-associated factor (BAF) complex was also reported to cause CD4 derepression in DN thymocytes (82). Combining BAF deficiency with point mutants in S4 significantly enhanced CD4 derepression, and BAF subunits were found to bind S4 (82). The BAF complex was subsequently implicated in regulating S4 accessibility for Runx1 binding in DN thymocytes (95). Finally, the zinc finger transcription factor Ikaros, which associates with BAF and other chromatin remodeling complexes, was also found to be required for CD4 repression in DN thymocytes (96). Active repression of CD4 hence appears to require remodeling of the locus and binding of Runx1, which is discussed further below (Figure 3).

S4 deletion increased recruitment of HAT p300 at E4_P and increased H3K9Ac at the promoter in DN thymocytes (93). Although E4_P is 13 kb upstream of the Cd4 promoter, these elements have been suggested to physically communicate through looping (97). Chromatin conformation capture assays suggested that in DN thymocytes, Runx1 mediates a loop between S4 and E4_P, which sequesters the positive transcription elongation factor (P-TEFb) from activating RNAPol-II at the promoter (97). Another factor that cooperates with Runx1 in Cd4 silencing at the DN stage is the bHLH transcription factor adaptor-related protein complex 4 (AP4), which binds to E4_P (98). This raises the possibility that a loop forming between E4_P and S4 via Runx1 and AP4 prevents transcription in DN thymocytes and, upon Runx1 downregulation during the DN–DP transition, P-TEFb associates with RNAPol-II at the promoter to drive expression. However, Runx1 has also been reported to bind to S4 in DP thymocytes, suggesting that its repressive function is overridden by an unknown mechanism at this stage. Another study using three-dimensional fluorescent in situ hybridization showed that the Cd4 and Cd8 loci are in close proximity when CD8 is expressed (in DP thymocytes and CD8⁺ T cells), but in the absence of S4, this association was lost, suggesting that S4 has a role in locus architecture (99). Strikingly, association of the co-receptor loci was conserved in human T cells even though Cd4 and Cd8 are on different chromosomes, suggesting that their proximity may be functionally important. Thus, S4 together with E4_P and E4_M are able to direct helper lineage-specific expression of CD4 in vivo.

Taken together, the study of cis elements controlling Cd4 expression during T cell development revealed a key role for these elements in the establishment of key epigenetic marks that can be transmitted thereafter in a heritable manner. Thus, a thorough analysis of the stage-specific activity of these elements provides a unique opportunity to delve into molecular requirements for heritable transmission.

The Cd8 locus is composed of two linked genes, Cd8a and Cd8b1. In mice, they are found on chromosome 6, separated by 36 kb and aligned in the same transcriptional orientation. DP thymocytes and most cytotoxic TCRαβ T cells express CD8 as a CD8αβ heterodimer, whereas on intraepithelial lymphocytes (IELs) and a subset of dendritic cells, CD8 can be expressed as a homodimer of CD8αα. This suggests that both genes can be coordinately and independently regulated in different cell types. An 80-kb genomic fragment encompassing the Cd8 locus was found to drive developmental stage and lineage-specific expression of a reporter transgene (103, 104). Four DHS clusters within this fragment (CI-IV) contain at least six enhancers. E8_I enhancer (CIII-1,2) was found to drive CD8 expression in CD8SP thymocytes, mature CD8⁺ T cells, and IELs (104–106). Interestingly, the onset of E8_I activity was observed in positively selected thymocytes, suggesting a link between this element and cytotoxic T cell commitment. E8_II (CIV-4,5) directed expression in both DP thymocytes and CD8⁺ T cells, while E8_III (CIV-3) drove CD8 expression in immature DP thymocytes. The E8_IV (CIV-1,2) element was active in DP and CD8⁺ T cells. In transgenic studies, E8_V (CII) exhibited no enhancer function by itself, but a combination of E8_V and E8_I directed expression in both CD8⁺ T cells and DP thymocytes, unlike E8_I alone, suggesting that stage-specific activity could arise through combinations of CD8 enhancers (107). Recently, another cis element termed E8_VI (CIV-6) was described, which directed transgenic reporter expression in mature CD8⁺ T cells, memory-phenotype-like CD44^hiCD62L⁺CD8⁺ T cells, and CD8αα⁺ dendritic cells (108). In summary, at least six enhancers direct CD8 expression in cytotoxic T cells (Figure 5).

Studies examining the regulation of the endogenous Cd8 locus through genetic deletion revealed both specific and cooperative modes of enhancer regulation. In E8_I^−/− mice, both TCRγδ and TCRαβ IELs showed a significant decrease in CD8α, while CD8 expression in DP thymocytes, CD8SP thymocytes, and peripheral CD8⁺ T cells was largely intact (106, 107). E8_I was also found to be crucial for the maintenance of CD8α expression in activated CD8⁺ T cells (109). In some cases, combinatorial enhancer activity occurred as shown by combined deletion of E8_I/E8_II, resulting in variegated expression in DP thymocytes that did not occur with single deletion of either enhancer (110). In contrast, combined deletion of E8_I/E8_II had no additional effect on the maintenance of CD8 expression in activated CD8⁺ T cells than deletion of E8_I alone (109). These results suggested that E8_I might have a non-redundant role in maintaining transcription of CD8α during activation, as IELs are also thought to exist in a partially activated state (111). Similar to combined E8_I/E8_II deletion, variegated CD8 expression was observed during the DN–DP transition with deletion of E8_V or combined deletion of E8_II/E8_III (112, 113). Taken together, these data indicate that in vivo the CD8 enhancers have both non-redundant and stage-specific functions.

It is unclear whether there is a silencer in the Cd8 locus, analogous to that in Cd4, that restricts its expression to the cytotoxic lineage. Experiments in a hybridoma cell line suggested the possibility that DNA elements in cis to the Cd8 locus were able to silence transcription (114). Referred to as L2a, this region within E8_V was identified as a matrix-associated region (MAR), an AT-rich sequence that mediates attachment to the nuclear matrix (115). In transgenic experiments, L2a was found to have negative transcriptional activity using an E8_I/E8_V hCD2 reporter construct that was reversed by the MAR binding protein, Satb1 (special AT-rich sequence binding protein 1) (116). Deletion at the endogenous locus as well as assays for position and orientation independence will help clarify whether this is a bona fide silencer. Thus, in contrast to the Cd4 locus, there is currently no known negative element that confers helper versus cytotoxic lineage specificity of CD8 expression. One ontological explanation is that there may be a need for greater plasticity in CD8 expression compared to CD4. For example, positive selection of DP thymocytes induces the downregulation of CD8, which is reversed during cytotoxic lineage commitment (7). The overexpression of Thpok in CD8⁺ cytotoxic T cells silences CD8 but does not induce CD4 expression (8, 117). In vivo, some CD4⁺ helper T cells in the intestine undergo transcriptional reprogramming toward a cytotoxic phenotype including upregulation of CD8αα following the downregulation of Thpok (118, 119). Thus, dynamic and flexible CD8 expression in T cells may require regulation mediated directly through a transcriptional circuit.

The variegated CD8 expression in DP thymocytes due to deletion of E8_I/_II or E8_V suggested that these enhancers were required to protect the Cd8 locus from repressive chromatin modifications. Interestingly, mutations in the SWI/SNF-like nucleosome remodeling BAF complex caused similar CD8 variegation in DP thymocytes as the enhancer-deficient mice (82, 110, 112). BAF likely regulates Cd8 directly, as it was found to bind to the locus by ChIP analysis (82). The DNA-binding transcriptional regulator Ikaros, which was reported to associate with the BAF complex, was also required for efficient CD8 upregulation during the DN–DP transition when on an Aiolos-deficient background (another Ikaros family member) (120, 121). In T cells, Ikaros was also found to associate with the repressive chromatin remodeling complex NuRD (122). The NuRD ATPase chromatin remodeling subunit, mi-2β, was found to bind the Cd8 locus until pre-TCR signaling, upon which its eviction coincided with increased DNAse I hypersensitivity and expression of CD8 (123). Thus, different chromatin remodeling complexes, presumably recruited by transcription factors such as Ikaros to the CD8 enhancers, can activate or repress CD8 expression at different stages of development.

Double-positive thymocytes that failed to upregulate CD8 in E8_I/II-deficient mice showed decreased histone acetylation at the Cd8 locus concomitantly with a decrease in H3K4Me3, both consistent with the loss of expression (124). H3Ac was also decreased along with an increase in H3K27Me3 in activated cytotoxic T cells from E8_I^−/− mice that had lost CD8 expression (109). Treatment of E8_I^−/− CD8⁺ T cells with the HDAC inhibitor TSA partially rescued CD8 expression during activation although there was no restoration of CD8 in T cells that had already lost expression (109). Interestingly, the unstable CD8 expression during proliferation was reminiscent of CD4 expression in the absence of either E4_M or E4_P (73, 75).

Histone acetylation appears to play a particularly prominent role in CD8 regulation, as the deletion of HDAC1/2 with CD4-Cre caused CD8 derepression in CD4⁺ T cells (125). In contrast, Dnmt1 deletion did not affect CD8 expression in CD4⁺ T cells but caused CD8 derepression in TCRγδ cells, suggesting differential dependence on DNA methylation and histone modifications for CD8 silencing in different cell types (126). DNA methylation was also analyzed in thymocytes that had failed to upregulate CD8 during the DN–DP transition in E8_I/_II-deficient mice (124). Previous studies showed that DNA methylation patterns correlate with CD8 expression during thymocyte development and in peripheral T cells (126–129). However, in the absence of E8_I/_II, there was a gain in methylation of several CpGs at the Cd8 locus at E8_V, and the expression of CD8 could be partially rescued by crossing the mice onto a Dnmt1-deficient background (124). These results suggest that factors binding to E8_I/_II promote Cd8 expression through inhibition of the DNA methylation machinery.