The Loss of TET2 Promotes CD8⁺ T Cell Memory Differentiation

This article is featured in In This Issue, p.1

In response to infection, naive CD8⁺ T cells proliferate and differentiate into a heterogeneous pool of Ag-specific cells having divergent cell fates. Following pathogen clearance, most Ag-specific CD8⁺ T cells die, but a subset persists to become long-lived memory cells, which are able to rapidly respond to rechallenge.

Several cell surface proteins can be used to identify cells with differing memory potential. Ag-specific CD8⁺ T cells that are CD127^hi and KLRG1^lo preferentially differentiate into long-lived memory cells, whereas CD127^lo and KLRG1^hi cells are largely short-lived terminally differentiated effector cells (1–3). This program of CD8⁺ T cell differentiation is regulated through the integration of signals from the TCR, costimulatory/coinhibitory receptors, and inflammatory cytokines, which direct transcriptional changes that control cell fate. Although it is evident that particular transcription factors, such as T-bet, eomesodermin (Eomes), Blimp-1, Bcl-6, IRF4, and Runx3, are important in determining the fate of activated cells (4–11), it is also becoming clear that epigenetic programming plays a crucial role in T cell fate determination.

DNA methylation is one epigenetic mechanism by which T cell differentiation is regulated, and recent genome-wide studies have identified coordinated epigenetic changes associated with transcriptional programs during CD4⁺ or CD8⁺ T cell differentiation (12–19). It is now appreciated that DNA undergoes regulated demethylation. Recently, the ten-eleven translocation (TET) family of methylcytosine dioxygenases was shown to mediate this process by catalyzing the conversion of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) and, subsequently, to 5-formylcytosine and 5-carboxylcytosine, critical enzymatic steps necessary for generating unmodified cytosines (20–22). The TET family member TET2 is widely expressed in the hematopoietic system, and murine models reveal that TET2 loss leads to expansion of hematopoietic stem cells (HSCs) and myeloid compartments (23–26). In T cells, TET2 contributes to CD4⁺ Th cell differentiation (27) and cooperates with TET3 to stabilize Foxp3 expression in regulatory T cells (28).

However, the function of TET2 in CD8⁺ T cell differentiation is unknown. In this study, we investigated TET2’s role in directing CD8⁺ T cell fate following acute lymphocytic choriomeningitis virus (LCMV) infection. We found that TCR signaling rapidly and dynamically regulates TET2 expression and TET activity. Although mice with selective loss of TET2 in T cells have no overt thymic or peripheral T cell phenotypes at steady-state, following acute viral infection, CD8⁺ T cells preferentially adopt a memory phenotype in a cell-intrinsic manner and demonstrate superior pathogen control upon rechallenge. Methylation analysis of LCMV-specific CD8⁺ T cells identified genomic loci that gained 5mC/5hmC in TET2-deficient cells, including several transcriptional regulators known to direct CD8⁺ T cell effector versus memory differentiation. Together, these data demonstrate a novel role for TET2 in directing CD8⁺ T cell fates.

B6;129S-Tet2^tm1.1Iaai/J (TET2^fl/fl) mice, C57BL/6J mice, and CD4Cre⁺ mice were obtained from The Jackson Laboratory and bred at the University of Pennsylvania or the University of Michigan. B6.SJL-Ptprc^a (CD45.1⁺) mice were from Taconic. For the LCMV memory phenotyping experiments, TET2^fl/flCD4Cre⁺ mice were backcrossed twice to C57BL/6J mice. For all other experiments, TET2^fl/flCD4Cre⁺ mice were backcrossed 5–10 times to C57BL/6J mice. Control mice for experiments included age-matched TET2^fl/fl, TET2^fl/+CD4Cre⁻, TET2^+/+CD4Cre⁺, or C57BL/6J animals. P14 mice (29) were bred to TET2^fl/flCD4Cre⁺ mice to generate TET2^fl/flCD4Cre⁺P14⁺ animals. All experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (numbers 803976 and 803071) and the University of Michigan (PRO00007214).

Mice were infected with 2 × 10⁵ PFU LCMV-Armstrong i.p. or 0.5–2 × 10⁵ CFU Listeria monocytogenes that expresses the LCMV gp33 epitope (LM-gp33), as indicated. LM-gp33 was grown and used as previously described (30). LCMV viral titers and Listeria bacterial loads were measured as described (30, 31).

Murine lymphocytes were isolated from spleen and lymph nodes, and T cells were purified by negative selection and magnetic separation (Pan T Cell Isolation Kit; Miltenyi Biotec). T cells were cultured in T cell media (10% FCS, 50 μM 2-ME, 2 mM l-glutamine/penicillin/streptomycin in IMDM) and activated with plate-bound anti-CD3 (1 μg/ml; 2C11) and 5 μg/ml anti-CD28 (37.51; both from eBioscience) for the indicated times. Pharmacologic activation of T cells was performed using PMA at 10, 25, or 50 ng/ml and ionomycin at 100, 250, or 500 ng/ml.

Lymphoid and nonlymphoid organs were processed, and single-cell suspensions were obtained. Peripheral blood was collected into 4% sodium citrate, purified with a Ficoll gradient (Ficoll-Paque Plus; GE Healthcare), and stained for flow cytometric analysis. CD8⁺ T cells (purified as described above) from “memory” mice were injected into congenic hosts so that 5000 or 7500 CD8⁺gp33⁺ cells were transferred. For P14 adoptive-transfer experiments, 2000 CD8⁺gp33⁺Vα2⁺ cells isolated from the peripheral blood were transferred into congenic hosts.

Cells were isolated, washed, and stained with the indicated Abs. The following Abs were used (all from BD Biosciences unless otherwise noted): CD8α-Pacific Blue or Alexa Fluor (AF)700 (53-6.7; BioLegend); CD4 FITC (GK1.5; eBioscience), PE-Cy7 (RM4-5; BioLegend), or PE–Texas Red (RM4-5; Invitrogen); TCRβ allophycocyanin-e780 (H57-597; eBioscience); CD62L PE–Texas Red (MEL-14; Invitrogen) or Brilliant Violet e605NC (MEL-14; eBioscience); KLRG1 PE-Cy7, FITC or PerCP-e710 (2F1; eBioscience), CD127 PE-Cy7 (A7R34; BioLegend) or Pacific Blue (A7R34; eBioscience); CD27 PE (LG.7F9; eBioscience); CXCR3 PerCP-Cy5.5 (CXCR3-173; BioLegend), 2B4 FITC (eBio244F4; eBioscience; 2B4; BD), CD160 PE (7H1; BioLegend), CD45.1 PerCP-Cy5.5, PE-Cy5, PE-Cy7 (A20; eBioscience) or AF700 (A20; BioLegend); CD45.2 AF700, allophycocyanin-e780 (104; eBioscience) or Pacific Blue (104; BioLegend); CD44 AF700 (IM7; BioLegend); IFN-γ–PerCP–Cy5.5 (XMG1.2; BioLegend); TNFα–Pacific Blue (MP6-XT22; eBioscience); IL-2 allophycocyanin (JES6-5H4); granzyme B PE-Cy7 (NGZB; eBioscience); CD107a FITC or PE (1D4B); human Ki67-FITC (B56); and Eomes AF647 (Dan11mag; eBioscience). Biotinylated monomers specific for H2-D^b–restricted gp33–41 of LCMV were obtained from the National Institutes of Health Tetramer Core Facility and tetramerized using their published protocol. Intracellular staining was performed using a Cytofix/Cytoperm kit (BD Biosciences) or a Foxp3/Transcription Factor Staining Buffer Set (eBioscience), according to the manufacturer’s instructions. Discrimination of live cell populations was performed using LIVE/DEAD Aqua stain (Invitrogen), according to the manufacturer’s instructions.

For experiments involving measurement of intracellular 5hmC, T cells were surface stained prior to fixation/permeabilization with a Cytofix/Cytoperm kit (BD Biosciences), treated with DNase I (300 μg/ml in PBS) at 37°C for 1 h, and intracellularly stained with isotype or anti–5hmC (catalog number 39791; 1 μg/ml; Active Motif) Ab for 30 min, followed by fluorochrome-conjugated goat anti-rabbit secondary Ab (Invitrogen).

For experiments involving ex vivo stimulation, single-cell suspensions were stimulated with 200 ng/ml gp33, gp276, or NP396 peptides in the presence of 1 mg/ml brefeldin A for 5 h and then analyzed for intracellular cytokine staining. Data were acquired using a FACS LSR II (BD Biosciences) and analyzed with FlowJo software (TreeStar).

For experiments involving cell sorting, T cells were isolated and sorted on a FACSAria II (BD Biosciences). For isolation of naive CD8⁺ T cells, CD8⁺ T cells from the spleen and lymph nodes were purified by negative selection and magnetic separation (CD8a⁺ T Cell Isolation Kit II; Miltenyi Biotec) and then sorted for naive CD8⁺ T cells (TCRβ⁺CD8⁺CD4⁻CD44⁻CD62L⁺). For sorting of gp33⁺CD8⁺ T cells, CD8⁺ T cells from the spleens of control and TET2^fl/flCD4Cre⁺ mice were negatively isolated as above and sorted for gp33⁺CD8⁺CD4⁻CD44⁺ on a FACSAria II using appropriate biohazardous precautions.

Genomic DNA was bisulfite converted using an EZ DNA Methylation Kit (Zymo Research). Base-pair resolution DNA methylation was performed on control and TET2-deficient gp33⁺CD8⁺ T cells (n = 4 per genotype), as described (32). Differentially methylated cytosines (DMCs) were identified using a 25% methylation difference threshold and q-value < 0.01 using methylKit (33); differentially methylated regions (DMRs) were determined using the bioinformatic algorithm eDMR (34).

Data were analyzed through the use of Ingenuity Pathways Analysis (IPA; Ingenuity Systems, Redwood City, CA; http://www.ingenuity.com) specified for “Mouse.” The differentially methylated genes containing DMRs that corresponded to at least one pathway annotation in the Ingenuity Knowledge Base were eligible for analysis. The p values associated with pathways were calculated using the right-tailed Fisher exact test.

Enhanced reduced representation bisulfite sequencing (ERRBS) data have been deposited to the National Center for Biotechnology Information Gene Expression Omnibus under accession number GSE105176 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE105176).

Statistical significance was calculated as noted in the figure legends. Prism (GraphPad) was used for statistical analysis.

If TET2 plays an important role in T cell differentiation, we reasoned that its expression and/or function might be altered in response to TCR ligation. We found that TET2 mRNA expression was rapidly induced in murine T cells after TCR plus CD28 stimulation, with maximal expression at 2 h (Fig. 1A). This upregulation was dependent on Ca²⁺ signaling, because TET2 mRNA was induced by the Ca²⁺ ionophore ionomycin but not PMA alone, suggesting that the protein kinase C pathway is not sufficient for TET2 induction (Fig. 1B). In addition to regulation of TET2 mRNA levels, TET2 protein is negatively regulated by the CXXC-type zinc finger domain containing proteins IDAX and CXXC5 (35). We evaluated their expression in primary murine T cells as a possible additional means by which TET2 could be regulated. Although IDAX was not detected in T cells, CXXC5 was expressed in resting cells and was downregulated by TCR stimulation (Supplemental Fig. 1A); thus, CXXC5 could serve to further regulate TET2 expression. We also assessed whether TCR stimulation directs TET activity, specifically the oxidation of 5mC to 5hmC. Using flow cytometry, we found that 5hmC levels were rapidly induced in murine T cells after TCR stimulation (Fig. 1C). Together, these data indicate that TET2 gene expression and TET function are regulated by TCR activation.

TET2 deficiency in the hematopoietic compartment results in expansion of HSCs with skewing toward the myeloid lineage (23–26, 36). To specifically study TET2 in T cells, we generated TET2^fl/flCD4Cre⁺ mice that lack TET2 in all mature αβ T cells (TET2 conditional knockout [TET2cKO] mice). These mice had similar frequencies and numbers of thymic and peripheral T cell populations compared with age-matched control mice, without evidence of substantially altered homeostasis, as indicated by similar absolute numbers of naive (CD44⁻CD62L⁺) and CD44⁺ T cell populations (Supplemental Fig. 1B–D). Moreover, we did not see changes in early activation markers (CD25, CD69, Bcl-x_L) or proliferation following CD3/CD28 stimulation (Supplemental Fig. 1E, 1F). Together, these data suggest that TET2 does not substantially alter T cell homeostasis or regulate short-term TCR-induced responses in vitro.

Because DNA methylation regulates the cell fate of T cells and other cell types (37, 38), and TET2 is known to control HSC differentiation, we hypothesized that TET2 might influence CD8⁺ T cell differentiation. To test this hypothesis, we infected control and TET2cKO mice with LCMV-Armstrong and followed virus-specific effector and memory CD8⁺ T cell differentiation using tetramers specific for the gp33–41 (gp33) epitope of LCMV. At the peak of the CD8⁺ T cell response (day 8), we found similar frequencies and absolute numbers of gp33⁺CD8⁺ T cells in the spleens, as well as similar absolute numbers (although lower frequencies) of gp33⁺CD8⁺ T cells among PBMCs of TET2cKO mice compared with control mice (Fig. 2A). Additionally, a similar frequency of gp33⁺CD8⁺ T cells expressed the proliferation marker Ki67 on day 8 postinfection (p.i.) in control and TET2cKO mice (data not shown). Together, these data suggest that TET2 is not required for Ag-specific CD8⁺ T cell proliferation and expansion following acute viral infection.

Next, we examined the functional ability of TET2cKO CD8⁺ T cells to respond to antigenic stimulation. Based on findings in CD4⁺ T cells, in which TET2 deficiency led to reduced cytokine production following T_H1 and T_H17 in vitro skewing (27), we predicted that TET2cKO CD8⁺ T cells would produce less cytokine. However, in contrast to our expectation, TET2cKO LCMV-specific CD8⁺ T cells had enhanced IFN-γ expression and higher expression of CD107a, a surrogate marker for degranulation, in response to gp33 and gp276 peptides, with percentages trending higher in response to NP396 peptide stimulation (Fig. 2B, 2C). There was no difference in “double producers” (i.e., IFN-γ⁺TNF-α⁺) or IL-2–producing CD8⁺ T cells following LCMV–peptide stimulation, and control and TET2cKO mice were able to clear virus by day 8 p.i. (data not shown). Together, these data demonstrate that TET2 loss enhances CD8⁺ T cell effector function at the peak of infection for certain TCR specificities.

After T cell expansion in response to infection, the majority of Ag-specific CD8⁺ T cells undergo cell death, whereas a small subset persists and differentiates into long-lived memory cells. Memory CD8⁺ T cells have several stem cell–like properties, including long-term survival, self-renewal, and multipotent potential (39), as well as a shared transcriptional program (40). Given these characteristics and the fact that TET2 loss in HSCs leads to an expanded stem/progenitor pool (23–26), we reasoned that TET2 deficiency might skew CD8⁺ T cells toward a memory phenotype at the expense of short-lived effector cells. We monitored cohorts of mice for persistence of gp33⁺ CD8⁺ T cells in the peripheral blood of LCMV-infected animals. The frequency of control and TET2-deficient gp33⁺CD8⁺ T cells in the blood had a similar pattern of expansion, contraction, and persistence over time for >80 d p.i. (Supplemental Fig. 2A). However, in the spleens on day 8, among TET2cKO gp33⁺CD8⁺ T cells, the memory precursor effector cell (MPEC) population, identified by its KLRG1⁻CD127⁺ phenotype (1–3), was significantly augmented with regard to frequency and absolute number, with a concomitant decrease in the KLRG1⁺CD127⁻ short-lived effector cell (SLEC) population (Fig. 3A). The phenotype was also evident among gp33⁺CD8⁺ T cells in the peripheral blood (Supplemental Fig. 2B). Additionally, during longitudinal monitoring of peripheral blood following LCMV infection, compared with controls, more TET2cKO gp33⁺CD8⁺ T cells expressed the markers CD27, CD62L, and CXCR3 (Supplemental Fig. 2C, 2D), which are highly expressed on central memory CD8⁺ T cells (T_CM) (41–43). Interestingly, the rate of increase over time was similar in the MPEC population among control and TET2-deficient gp33⁺CD8⁺ T cells (Supplemental Fig. 2E). Together, these data suggest that TET2 loss directs an early decision toward a T_CM fate.

To examine whether TET2 loss promotes “bona fide” memory formation, we examined control and TET2cKO mice on day 45 p.i. In the spleens of TET2cKO mice, there was an increase in the frequency and absolute number of T_CM gp33⁺CD8⁺ T cells (defined by CD62L and CD44 expression; Fig. 3B). Additionally, we saw a significant increase in the expression of other T_CM markers, including CD27 and CXCR3 (Fig. 3C). Consistent with increased T_CM formation, we found an increase in the absolute number of gp33⁺CD8⁺ T cells in the lymph nodes of TET2cKO mice compared with control mice (Fig. 3D). To further characterize the TET2-deficient memory CD8⁺ T cells, we examined expression of transcription factors associated with CD8⁺ T cell differentiation, including Eomes and T-bet. TET2-deficient gp33⁺CD8⁺ T cells had higher expression of Eomes compared with control gp33⁺CD8⁺ T cells (Fig. 3E), but no significant differences were noted in T-bet expression (data not shown). Together, these data show that TET2 loss promotes CD8⁺ T cell memory formation.

TET2cKO mice lack TET2 in mature CD4⁺ and CD8⁺ T cells; therefore, we sought to determine whether the preferential CD8⁺ memory differentiation in TET2cKO mice was intrinsic to TET2 deficiency in the CD8⁺ T cell population or dependent on an altered environment due to TET2 loss in all T cells. To address this issue, we used two complementary approaches. In one approach, we transferred CD8⁺ T cells isolated from congenic (CD45.1⁺CD45.2⁺) mice expressing the P14 TCR transgene, which is specific for the LCMV gp33 epitope presented by H-2D^b, into control and TET2cKO mice and subsequently infected these mice with LCMV (Fig. 4A). As expected, endogenous host (CD45.2⁺) TET2cKO gp33⁺CD8⁺ T cells rapidly differentiated into MPECs at the expense of SLEC formation (Fig. 4B). In contrast, similar frequencies of SLEC and MPEC populations derived from P14 cells were present in control and TET2cKO mice (Fig. 4B; CD45.1⁺CD45.2⁺ gate). Additionally, the transferred P14 cells expressed CD62L at similar frequencies in control and TET2cKO mice (data not shown). As a second approach, we transferred equal numbers of wild-type (WT) or TET2-deficient P14 CD8⁺ T cells into congenic (CD45.1⁺) hosts and infected them with LCMV-Armstrong (Fig. 4C). Consistent with prior results, a higher frequency of TET2-deficient P14 cells differentiated into MPECs compared with WT P14 cells, which predominantly differentiated into SLECs (Fig. 4D). Together, these data demonstrate that TET2 regulates CD8⁺ T cell memory differentiation in a cell-intrinsic manner.

Thus far, we have shown that TET2 loss promotes a central memory phenotype. To determine whether TET2cKO cells displayed enhanced functional memory properties, we evaluated the ability of TET2cKO memory CD8⁺ T cells to mount an effective memory response. Purified CD8⁺ T cells from control and TET2cKO mice previously infected with LCMV (>100 d p.i.) were transferred into WT congenic hosts, such that equal numbers of gp33⁺CD8⁺ T cells were injected. Hosts were subsequently infected with LM-gp33. Five days following LM-gp33 infection, more gp33⁺ TET2cKO cells than control cells were present in the peripheral blood of infected hosts (Fig. 5A). Comparable gp33⁺CD8⁺ T cell numbers were found in the spleen, and frequencies within both organs were similar (Fig. 5A, Supplemental Fig. 3A). To assess the function of these memory populations, bacterial burden was measured. Mice receiving TET2cKO memory CD8⁺ T cells had lower Listeria burden in their livers at 5 d p.i. (Fig. 5B), indicating that TET2-deficient memory cells exhibit greater pathogen control. This enhanced bacterial control was not accompanied by increased production of IFN-γ, TNF-α, or CD107a (Supplemental Fig. 3B) but rather was associated with higher pretransfer expression of CD62L and CXCR3 (Supplemental Fig. 3C), two T_CM receptors that have been associated with more efficient recall responses (42–44). Together, these data demonstrate that loss of TET2 favors a more effective memory immune response.

We posited that TET2 loss promotes CD8⁺ T cell memory differentiation through altered epigenetic regulation of gene expression. To examine cytosine methylation differences in control and TET2cKO T cells, we performed ERRBS, a method that allows for quantitative genome-wide single–base pair resolution analysis of cytosine methylation, with coverage within and outside of CpG islands (32, 45). Similar to bisulfite sequencing, ERRBS does not distinguish between 5mC and 5hmC, reading both as a methylated cytosine residue and, thus, referred to as 5mC/5hmC to indicate detection of either.

We isolated gp33⁺CD8⁺ T cells from control and TET2cKO mice on day 8 p.i. with LCMV. Genomic DNA was digested with a methylation-insensitive restriction enzyme and subjected to standard next-generation sequencing library preparation. Unmethylated cytosines underwent bisulfite conversion prior to final amplification and sequencing. The ERRBS analysis covered an average of 1.54 × 10⁶ individual CpGs with minimum 10× coverage of the mouse genome. Analysis of DMCs (>25% methylation difference) demonstrated that the majority gained 5mC/5hmC in TET2cKO LCMV-specific CD8⁺ T cells compared with control (Fig. 6A). These DMCs occurred primarily in intergenic (35%) and intronic (47%) regions (data not shown). Next, we grouped DMCs into DMRs, in which multiple adjacent CpG sites have differential methylation (≥25% difference in all control versus TET2cKO samples), using the bioinformatic algorithm eDMR (34). Three hundred and fifty-five DMRs mapped to 184 unique ReqSeq genes (mm 9; Supplemental Table I). The identified DMRs were mapped to different genomic regions, with the majority located in introns and coding sequences (Fig. 6B), and most commonly occurred in genes involved in cellular growth and proliferation, cellular development, cell death and survival, and gene expression, according to IPA (Fig. 6C).

Several proteins that were differentially expressed on TET2cKO versus control gp33⁺CD8⁺ T cells, including KLRG1 (15), CD127 (46), CD27 (47), CXCR3 (15, 48) and CD62L (15), are known to be regulated by DNA methylation. Thus, mechanistically, we reasoned that TET2 could individually regulate the methylation status of the loci of differentially expressed memory markers. Alternatively, TET2 could control the expression of key transcription factors and, thus, promote a CD8⁺ T cell memory transcriptional program. Some differentially expressed molecules, including CD127 and CD62L, did not have 10× ERRBS coverage in all eight samples (our cutoff for assessing DMCs) and, thus, could not be evaluated. However, gene loci encoding KLRG1, CD27, and CXCR3 were appropriately represented. Interestingly, these loci did not contain DMRs, suggesting that TET2 does not regulate expression of these genes through direct demethylation. Therefore, we evaluated DMRs associated with transcription factors known to direct CD8⁺ T cell effector versus memory differentiation, including Tbx21 (encoding T-bet), Eomes, PRDM1 (encoding Blimp-1), Bcl6, Runx3, and IRF4 (4, 6–10, 49). Bcl6 and Eomes did not contain DMRs, whereas Tbx21, PRDM1, IRF4, and Runx3 had associated DMRs that gained 5mC/5hmC in TET2cKO gp33⁺CD8⁺ T cells (Fig. 6D). Taken together, these data suggest that acquisition of a memory fate is not due to TET2-mediated loss of 5mC/5hmC at each locus encoding effector/memory markers (i.e., KLRG1, CD27, CXCR3) but rather TET2 may regulate transcriptional drivers of effector or memory cell fate.

Because we observed early differences in CD8⁺ T cell differentiation in TET2cKO CD8⁺ T cells, we questioned whether methylation differences seen at this time point were a cause or consequence of altered differentiation. Therefore, to evaluate early gene expression changes, we isolated control and TET2cKO naive CD8⁺ T cells, activated them in vitro with anti-CD3/CD28 and IL-2 for 3 d, and examined gene expression using RT-PCR and flow cytometry. Activated TET2cKO CD8⁺ T cells had diminished Blimp-1 mRNA, although this did not reach statistical significance. Because Blimp-1 and Bcl-6 negatively regulate one another and have opposing effects on CD8⁺ T cell effector versus memory development (6, 7, 9), we also assessed Bcl-6 mRNA. We found that TET2 loss promotes Bcl-6 expression and increased the Bcl-6/Blimp-1 ratio (Fig. 6E). Additionally, we noted a modest increase in Runx3 mRNA and IRF4 protein (Fig. 6E, 6F) but no significant difference in expression of T-bet or Eomes by flow cytometry at this early time point (data not shown).

The TET family of methylcytosine dioxygenases mediates active DNA demethylation. In multiple cell types, including embryonic stem cells (20, 50), HSCs (23–26), and CD4⁺ T cells (27), TET activity has been shown to regulate cellular differentiation. It is known that DNA methylation dynamically changes in CD8⁺ T cells following acute viral infection, with both gain and loss of DNA methylation at individual loci and across the genome (15); however, the role that TET enzymes and active DNA demethylation play in CD8⁺ T cell fate decisions has not been fully explored. In this study, we identify TET2 as a novel regulator of cell fate choice between terminally differentiated effector CD8⁺ T cells and T_CM formation.

TET expression and activity are regulated through multiple mechanisms in nonlymphoid cells, including protein stability via IDAX and CXXC5 (35), ascorbic acid (51–53), and acetylation (54). In this study, we find that TET2 gene expression is rapidly and transiently upregulated in primary murine T cells by TCR signaling. This upregulation occurs in a Ca²⁺-dependent manner and coincides with a decrease in CXXC5 gene expression, a known inhibitor of TET2 protein stability (35). We also found that TCR signals regulate TET enzymatic activity, although the timing of changes in 5hmC levels appeared to occur faster than changes in TET2 mRNA expression. The discordant timing of 5hmC induction versus TET2 gene expression may be reflective of the sensitivity of the respective assays or indicate that TCR signals regulate TET2 transcription and TET enzymatic activity independently of each other. Together, these data suggest that TET2 expression and TET activity are tightly controlled in T cells, likely through multiple mechanisms, following T cell activation.

Studies of DNA methylation in T cells have focused primarily on regulation of cytokine and effector molecule gene loci. Previous genome-wide analyses of DNA methylation during CD8⁺ T cell differentiation revealed decreased methylation and increased gene expression at effector gene loci, including Gzmb (granzyme B) and Ifng (IFN-γ), in day-8 effector CD8⁺ T cells following LCMV infection compared with naive T cells (15). How TET2 might contribute to these changes was not explored. However, based on a study in which diminished cytokine production was noted in TET2-deficient T_H1 and T_H17 cells (27), and our unpublished results (L.B. Banks, S.A. Carty, and M.S. Jordan, unpublished observations) indicating a similar finding in CD4⁺ cells, we predicted that TET2-deficient CD8⁺ T cells would have decreased cytokine production and/or effector molecule expression following viral infection. To our surprise, we found increased cytokine production and effector molecule expression in TET2-deficient CD8⁺ T cells in vivo during acute viral infection. These data indicate that there are alternative mechanisms regulating the removal of repressive DNA methylation at these loci, which could include removal by other TET family members or by DNMT1 inhibition. The lack of DMRs associated with cytokine, effector, and individual memory marker gene loci was contrasted with the several DMRs found in known transcriptional regulators of CD8⁺ T cell differentiation. In this study, we noted that, rather than occurring at individual loci of differentially expressed memory markers (such as KLRG1, CD27, and CXCR3), the DMRs identified occur in known transcriptional regulators of CD8⁺ T cell fate. To evaluate whether these methylation changes altered gene expression, we evaluated gene and protein expression in control and TET2-deficient CD8⁺ T cells activated in vitro. We found that there were modest increases in Runx3 mRNA, as well as IRF4 protein expression, which may contribute to the increase in IFN-γ and cytolytic potential seen early during infection (8, 55). Moreover, we found an alteration of the Bcl-6/PRDM1 ratio in a manner that would be predicted to promote memory CD8⁺ T cell differentiation. Although some of these expression changes were modest, they could be synergistic in promoting memory CD8⁺ T cell formation while maintaining strong effector responses. Taken together, these findings suggest that TET2 may direct CD8⁺ T cell fate via regulation of transcriptional programs.

Following LCMV infection, Ag-specific TET2-deficient CD8⁺ T cells rapidly acquired surface markers associated with memory potential in a cell-intrinsic manner. These differences occurred early and correlated with more TET2-deficient T_CM. Interestingly, a recent report identified DNMT3A, another enzyme involved in DNA methylation, as a repressor of CD8⁺ T cell effector differentiation, demonstrating that mice with a T cell–specific deletion of DNMT3A have enhanced T_CM differentiation following acute infection (56). The similarity of these phenotypes was somewhat surprising at first, because DNMT3A catalyzes the de novo methylation of cytosines (57), and TET2 oxidizes 5mC to 5hmC. However, loss of both genes affects HSC differentiation and self-renewal in a similar (although not identical) manner (23–26, 58). Moreover, mutations in both TET2 and DNMT3A occur in an overlapping set of human myeloid and T cell neoplasms (59, 60), as well as age-related clonal hematopoiesis (61–63). Recent work with TET2 and DNMT3A single- and double-deficient HSCs identified methylated regions that were independently and interdependently regulated (64). These observations suggest that loss of DNMT3A or TET2 may deregulate similar pathways or an overlapping set of targets.

We also found that TET2-deficient memory cells have intact expansion and effector responses and provide enhanced pathogen clearance upon rechallenge. The increase in the T_CM population in TET2cKO mice could potentially explain the improved pathogen control upon antigenic rechallenge. At day 45 following LCMV-Armstrong infection, there were more central memory gp33⁺CD8⁺ T cells with high CD62L, CD27, and CXCR3 expression in TET2cKO mice compared with WT mice. Prior studies have demonstrated that the T_CM subset (defined by CD62L⁺) has enhanced pathogen clearance compared with the effector memory subset (42). Additionally, TET2-deficient memory CD8⁺ T cells expressed higher levels of CXCR3, a chemokine receptor that is important for T cell migration within the lymph node and required for rapid control of pathogens by memory CD8⁺ T cells (44). The role of these receptors in mediating the enhanced pathogen clearance mediated by TET2-deficient memory CD8⁺ T cells is being explored.

We and other investigators, have seen genome-wide 5mC and/or 5hmC changes in cells with altered TET activity (reviewed in Ref. 65). However, understanding the functional significance of these changes and linking them mechanistically to cellular outcomes remain a challenge for the field. The majority of the DMRs identified in this study occurred in intragenic regions, with a relatively low frequency (∼8%) occurring in promoter regions. These data are consistent with prior studies of DMRs in TET2-deficient embryonic stem cells (66), which suggest that TET2 activity may preferentially occur at intragenic sites. Although DNA methylation at promoters is understood to repress gene transcription, the functional significance of DNA methylation/hydroxymethylation at other genomic sites is less well understood. Two recent studies have demonstrated a role for TET2 in maintaining hypomethylation at enhancer regions (67, 68), suggesting that its action at enhancer regions can modulate gene expression. However, in a separate study examining TET2-knockdown cells, a similar number of genes were found to be upregulated and downregulated (66). Taken together, these studies highlight the complexity of TET2-regulated gene expression. Moreover, as is the case in any study of genetic deficiency, there remains a question of whether the DNA methylation changes noted are directly or indirectly mediated by TET2, a question further compounded by technical limitations of determining chromatin occupancy of TET2 with currently available reagents (reviewed in Ref. 65). Additionally, there remains the possibility that TET2 may partially (or wholly) regulate CD8⁺ T cell memory differentiation through a noncatalytic function because other epigenetic modifiers have noncatalytic functions (69, 70). In myeloid cells, TET2 has been shown to regulate cytokine expression and proliferation via mechanisms independent of DNA hydroxymethylation (71, 72); additionally, other TET family members have been shown to function via recruitment of chromatin modifiers (73).

Collectively, our data identify TET2 as a novel epigenetic regulator of CD8⁺ T cell effector versus memory cell fate decisions following acute viral infection and highlight the importance of epigenetic modifiers in shaping T cell immunity. Further studies of TET2-mediated function, including the significance of 5mC and 5hmC modifications, will advance our understanding of epigenetic regulation of T cell fates and may provide novel methods to modulate immune responses.

We thank the Cornell Epigenomics Core and Children’s Hospital of Pennsylvania NAPCore facility for data analysis. We thank Edward Behrens for assistance with analysis of publically available microarray data and members of the Jordan, Koretzky, Behrens, Kambayashi, and Oliver laboratories for helpful discussions. We thank Sydney Drury for technical assistance.

The authors have no financial conflicts of interest.