Loss of methylation at the IFNG promoter and CNS-1 is associated with the development of functional IFN-γ memory in human CD4+ T lymphocytes



This article is corrected by:

  1. Errata: Correction Volume 44, Issue 10, 3144, Article first published online: 13 October 2014

Full correspondence: Dr. Jun Dong, Cell Biology Group, German Rheumatism Research Centre Berlin 1, 10117, Berlin, Germany

Fax: +49-30-28460603

e-mail: dong@drfz.de

Additional correspondence: Prof. Dr. Andreas Thiel, Regenerative Immunology and Aging, Berlin-Brandenburg Center for Regenerative Therapies, Charité, Campus Virchow-Klinikum Augustenburger Platz 1, 13353, Berlin, Germany

Fax: +49-30-450539955

e-mail: andreas.thiel@charite.de


Cytokine memory for IFN-γ production by effector/memory Th1 cells plays a key role in both protective and pathological immune responses. To understand the epigenetic mechanism determining the ontogeny of effector/memory Th1 cells characterized by stable effector functions, we identified a T-cell-specific methylation pattern at the IFNG promoter and CNS-1 in ex vivo effector/memory Th1 cells, and investigated methylation dynamics of these regions during the development of effector/memory Th1 cells. During Th1 differentiation, demethylation occurred at both the promoter and CNS-1 regions of IFNG as early as 16 h, and this process was independent of cell proliferation and DNA synthesis. Using an IFN-γ capture assay, we found early IFN-γ-producing cells from 2-day differentiating cultures acquired “permissive” levels of demethylation and developed into effector/memory Th1 cells undergoing progressive demethylation at the IFNG promoter and CNS-1 when induced by IL-12. Methylation levels of these regions in effector/memory Th1 cells of peripheral blood from rheumatoid arthritis patients correlated inversely with reduced frequencies of IFN-γ-producers, coincident with recruitment of effector/memory Th1 cells to the site of inflammation. Thus, after termination of TCR stimulation, IL-12 signaling potentiates the stable functional IFN-γ memory in effector/memory Th1 cells characterized by hypomethylation at the IFNG promoter and CNS-1.


A unique feature of adaptive immunity is the generation of effector/memory T cells after primary activation that control pathogens and mediate effective protection during secondary challenges [1]. Depending on the stimuli and cytokine environment experienced during activation, naive T cells make decisions to proliferate and differentiate into Th1, Th2, or Th17 lineages [2]. When challenged by intracellular pathogens, naïve CD4+ T cells can adopt a proinflammatory cell fate and differentiate into effector and memory Th1 cells, which are characterized by the production of the signature cytokine IFN-γ. Th1 cells play a key role in intracellular pathogen killing and exert proinflammatory effects in organ-specific autoimmune diseases [3]. During primary activation, the expression of the Ifng/IFNG gene in naïve CD4+ T cells is initiated by TCR signaling, in conjunction with instructive IL-12 signaling through lineage-specific transcription factors (such as T-bet and the IL-12-responsive transcription factor STAT4, respectively) [4-6]. After termination of antigen stimulation, IL-12 also controls Ifng reactivation in effector/memory Th1 cells through STAT4 and T-bet [7]. During secondary activation, TCR signaling in the absence of the original instructive signal is sufficient to trigger rapid Ifng/IFNG reactivation in effector/memory Th1 cells. This functional cytokine memory for IFN-γ is achieved through a transcription factor network that engages the “poised” epigenetically imprinted regulatory regions in the Ifng/IFNG gene established during primary activation [8].

The accessibility of DNA binding sites to transcription factors is regulated by epigenetic modifications, including DNA methylation and covalent histone tail modifications [9]. Several studies have demonstrated an essential role of DNA methylation in determining the capacity of T cells to express Ifng/IFNG. For example, DNA methylation of a few specific CpGs in the IFNG/Ifng promoter is inversely correlated with IFNG/Ifng expression [10, 11]. We have also previously shown that treatment with DNA methylation inhibitors augments IFN-γ production, even in purified cells that do not originally produce IFN-γ [11]. In addition, de novo methylation of the IFNG promoter leads to reduced IFN-γ production in T cells infected with human immunodeficiency virus type I [12].

Analyses of regulatory elements in the Ifng/IFNG gene have provided considerable insight into the control of selective Th1 expression. For example, a conserved element termed CNS-1 (conserved noncoding sequence 1) located 5 kb upstream of the murine Ifng gene and 4.2 kb upstream of the human IFNG gene has been identified by sequence homology analysis between mice and humans. CNS-1 displays Th1-specific DNase I hypersensitivity and enhancer activity. The transcription factors NFAT and T-bet bind to this region in stimulated Th1-cell lines and augment its enhancer activity [13]. A minimal 500 bp stretch of the proximal IFNG promoter has been demonstrated to confer Th1/Tc1 selective expression [14, 15]. However, it remains to be elucidated how changes in the methylation status of these T-cell-specific regulatory elements affect Th1 lineage commitment in general and, in particular, during early Th1-cell differentiation, and whether IL-12 signaling regulates IFN-γ cytokine memory epigenetically. Such knowledge will provide insight into modulating Th1 responses in vaccine design, autoimmune diseases, and organ transplantation.

In this study, we identified a T-cell-specific methylation pattern at the promoter and CNS-1 of IFNG in ex vivo effector/memory Th1 cells. We further analyzed the methylation changes of these regions in CD4+ T cells during their differentiation from naive T cells into IFN-γ-producing effector/memory Th1 cells. We show that the T-cell-specific hypomethylation at the IFNG promoter and CNS-1 could be verified by both in vitro differentiated Th1 cells and Ag-specific Th1 clones. In particular, the initiation of demethylation of the IFNG gene locus occurred before any cell cycle activity during Th1 differentiation. Progressive demethylation along the IFNG promoter and CNS-1 took place during the differentiation of naive T cells into early IFN-γ-producing Th1 cells and then to effector/memory Th1 cells. After termination of TCR stimulation, IL-12 signaling was required for early IFN-γ-producing Th1 cells to reinforce demethylation at the IFNG promoter and CNS-1. This led to the increased levels of hypomethylation at these two regions that resembled those in ex vivo IFN-γ effector/memory CD4+ T lymphocytes, and the development of effector/memory Th1 cells. Therefore, our study provides an explanation of how epigenetic regulation establishes stable effector/memory during Th1-cell differentiation. Finally, the identification of demethylation of the IFNG promoter and CNS-1 shows clinical relevance in Th1-involved autoimmune diseases such as rheumatoid arthritis (RA).


T-cell-specific hypomethylation of the IFNG locus in IFN-γ effector/memory CD4+ T lymphocytes

To identify the T-cell-specific demethylation pattern of the IFNG gene locus, we first isolated ex vivo IFN-γ-producing and -nonproducing effector/memory Th-cell subsets from human peripheral blood using a cytokine capture assay (Fig. 1A). We then performed bisulfate PCR sequencing of seven CpG sites at CNS-1 (4.2 kb upstream from the transcriptional start site (TSS) of IFNG) and eight CpG sites at the promoter (Fig. 1B). At all CpGs of the promoter and CNS-1 regions, the IFN-γ+ Th-cell subset showed hypomethylation, whereas the IFN-γ Th-cell subset showed hypermethylation (Fig. 1C). Because IFN-γ is also produced by CD8+ T cells and NK cells, we further studied methylation status in different CD8+ T and NK cells. Similarly to the CD4+ Th-cell subset, naïve CD8+ cells were hypermethylated, whereas effector and memory CD8+ cells were hypomethylated (Supporting Information Fig. 1A). In contrast, NK cells displayed a differential methylation pattern with only a few CpGs being hypomethylated at the promoter (Supporting Information Fig. 1B). Notably, no differential methylations among all tested cell types were observed at a control region (2.4 kb upstream from the TSS of IFNG and adjacent to a DNase I hypersensitivity site) (Supporting Information Fig. 2A). Therefore, the distinct methylation patterns of CD4+ and CD8+ T cells and NK cells indicate that both the promoter and CNS-1 regions are important in the epigenetic regulation of selective T-cell differentiation. We next determined whether the T-cell-specific methylation pattern is heritable in Ag-specific Th clones and found that, in both of the promoter and CNS-1 regions, Th1 clones displayed a heritable hypomethylation pattern more similar to that of ex vivo IFN-γ+ Th cells than to Th2 clones (Fig. 1D). We next assayed for the methylation status of naturally occurring effector/memory CD45RO+ Th cells and CXCR3+ Th1 cells in relation to their capacities to produce IFN-γ and found that, in both of the promoter and CNS-1 regions, the methylation status of these two memory cell types inversely correlated with frequencies of IFN-γ-producing cells (Fig. 1E). Finally, using an IFN-γ capture assay, we confirmed the T-cell-specific methylation pattern in ex vivo CMV-specific IFN-γ+ Th cells (Fig. 1F). Taken together, we identified a T-cell-specific hypomethylation pattern of the IFNG gene locus in human effector/memory Th1 cells.

Figure 1.

Inverse correlation between IFNG locus methylation and its expression in CD4+ T cells. (A) Induction and isolation of ex vivo IFN-γ and IFN-γ+ Th-cell subsets. CD4+ T cells were stimulated with PMA and ionomycin, followed by an IFN-γ capture assay. The following Th cell subsets were identified: IFN-γ and IFN-γ+. The data are representative of five separate experiments. (B) Alignment of 10 kb of the human and mouse IFNG/Ifng loci; DNA sequence identity >50% over at least 100 bp is shown in the histogram plot. Arrow, transcription direction; dark blue areas represent conserved exons; light blue areas represent conserved untranslated regions; reddish areas represent conserved noncoding sequences; individual CpGs at CNS-1 and the promoter are depicted in relation to the TSS. (C, D) The DNA methylation status of (C) three pooled DNA samples from each IFN-γ and IFN-γ+ Th-cell subsets and (D) four Th1 and Th2 clones are shown in rows. (E) Inverse correlation between methylation levels of the promoter and CNS-1 with frequencies of IFN-γ-producing cells among memory and Th1 memory CD4+ T cells. (F) T cell-specific hypomethylation at the IFNG promoter and CNS-1 in ex vivo CMV-specific CD4+IFN-γ-producing effector/memory cells. DNA methylation levels were determined as previously described [11] (C and D) or average methylation levels of the promoter and CNS-1 by bisulfate pyrosequencing (E and F) are depicted with box plots with 5th and 95th percentile. ***p < 0.0001. Data shown are representative of two or three independent experiments.

Demethylation dynamics of the IFNG gene locus during Th1 cell differentiation

The development of effector/memory status of CD4+ T cells is reflected by the differentiation process from naïve Th cells to effector/memory IFN-γ-producing Th1 cells. With respect to this differentiation process, we next analyzed the methylation changes of the IFNG gene promoter and CNS-1 in naïve Th cells and in vitro-differentiated 1-week old Th1 cells separated according to the secretion of IFN-γ. In both the promoter and CNS-1 regions, naïve Th cells were hypermethylated, whereas IFN-γ+ and IFN-γ Th1-cell subsets isolated from Th1 cultures activated with either plate-bound anti-CD3 and anti-CD28 or with alloDCs were hypomethylated and hypermethylated, respectively (Fig. 2). These results suggest that the T-cell-specific demethylation pattern of the IFNG promoter and CNS-1 is required for the induction of effector and/or memory Th1 cells during Th1-cell differentiation.

Figure 2.

Th1 differentiation is accompanied by demethylation of CNS-1 and promoter of the IFNG. The methylation status of pooled (10 donors) naive Th cells and IFN-γ and IFN-γ+ Th-cell subsets separated from 1 week Th1 cultures with anti-CD3/28 or alloDC stimulation from two independent experiments are shown in rows. DNA methylation levels were determined as described in Fig. 1C and D.

To explore the dynamic changes of demethylation of the IFNG gene triggered by early Th1 differentiation, we performed a 3-day time course analysis of the methylation changes in naïve Th cells activated by plate-bound anti-CD3 and anti-CD28 under Th1 conditions using bisulfate pyrosequencing technology. As shown in Figure 3A, in ex vivo-isolated naïve Th cells (0 h), CpGs methylation at CNS-1 was between 80 and 100%, and CpG methylation at the promoter was between 68 and 91%. Of note, 16 h after activation, the methylation level of CpG-4227 at CNS-1 was reduced from 80 to 70%. Similarly, after 19 h of activation, a 13% reduction in methylation (from 76 to 63%) was detected at CpG-186 at the promoter. In addition, between the 16 and 19 h time points, a 15% decrease in the methylation level of CpG-53 and a 35% decrease in the methylation level of CpG-186 (at the promoter) was also observed (Fig. 3A). However, methylation levels of the 2.4 kb upstream control region remained unchanged at these analyzed time points (Supporting Information Fig. 2B). Together, these results suggest that the dynamic changes in demethylation of the T-cell-specific epigenetic regulatory regions of the IFNG locus may be important in initiating the early Th1 differentiation.

Figure 3.

Dynamic demethylation at the IFNG locus during Th1 differentiation. (A) Time course analysis of the methylation status of individual CpGs in the CNS-1 and promoter regions during Th1 differentiation using bisulfate pyrosequencing. The data shown are from one experiment representative of two performed. Time course analysis of (B) DNA synthesis and (C) cell proliferation, and (D) dynamic methylation at the IFNG gene locus and (E) cytokine productions according to cell divisions during Th1-cell differentiation of the cells described in (A). The data shown are representative of three individual experiments.

To determine whether and at which time point the activated cells synthesize DNA, we combined time-course analysis with a BrdU incorporation assay. No BrdU+ cells were detected at early time points (16, 19, or 22 h), while in contrast, increased numbers of BrdU+ cells were observed later between 36 and 70 h (Fig. 3B), therefore suggesting that the initiation of demethylation of the IFNG gene implicates an active regulatory mechanism, independent of DNA replication.

We next investigated whether the methylation changes involve a cell division-dependent regulatory mechanism as measured by CFSE dilution of CFSE labeled naïve Th cells. Over time, primed cells expressed dynamic levels of IFNG and the Th1-lineage transcription factors TBX21 and RNX3 (unpublished observations), in agreement of recent results in mouse Th1 cells [7]. No cell divisions were observed 22 h after activation under Th1 conditions, and while some proliferation was detectable at 46 h, most cells required 70 h of activation to divide at least once (Fig. 3C). At these time points, drastic reduction (about 50%) in methylation was detectable at CpGs-4227 and -186 at CNS-1 and promoter, respectively (Fig. 3A), suggesting a potential contribution of proliferation to demethylation. To test this possibility, we analyzed methylation changes in cells sorted by cell division (Fig. 3D). In primed nondividing cells, the methylation levels of CpGs-4227, -4276, -4291, and -4323 at CNS-1 was reduced between 12 and 21% in comparison with ex vivo-isolated naive Th cells. However, when cells had undergone one or three cell divisions, the methylation levels of these CpGs were markedly reduced to half or even less than half of those of their previous cell division. Interestingly, a dynamic methylation pattern of CNS-1 was observed in subsequent cell divisions, similarly to that of the promoter and CNS-1 in bulk differentiating cultures. In addition, IFN-γ production according to distinct cell division was comparable with results shown in mouse Th1-cell cultures [16, 17] (Fig. 3E). Together, these results suggest an unrevealed complexity of demethylation of the IFNG locus during early Th1 differentiation, and this process is both proliferation-independent and -dependent.

Requirement of IL-12 signaling for the development of functional IFN-γ memory

To further decipher the impact of initial demethylation of the IFNG promoter and CNS-1 on the formation of IFN-γ memory, we investigated the mechanism by which naïve human CD4+ T cells develop cytokine memory for IFN-γ after primary activation. Although it has been shown that IFNG mRNA is detectable as early as 2 h after activation [18], the first IFN-γ-producing effector cells were detectable only after 24 h (<0.5%), and their numbers increased over time (d2, 8 to 19%; d3, 30 to 40%) upon Th1 polarization by anti-CD3 and anti-CD28 stimulation (Fig. 4 and unpublished observations). To mimic Ag-specific TCR stimulation, we primed naive T cells for 2 days and separated early IFN-γ+ and IFN-γ cells using the cytokine capture assay. IL-12, the major instructive signal in Th1 differentiation, is postulated to play a role in maintaining Th1-cell immunity in vivo in several experimental models [4, 19-21]. In addition, IL-12 has recently been shown to control Ifng reactivation after termination of antigen stimulation [7]. Therefore, the sorted d2-IFN-γ+ and -IFN-γ cells were cultured in the presence or absence of IL-12 for 3 days and then restimulated with PMA and ionomycin (d7). In response to the initial IL-12 stimulation (1° + rIL-12), d2-IFN-γ+ cells were able to reactivate IFNG expression by 86%; d2-IFN-γ cells were induced to activate IFNG expression by 45%. In contrast, in the absence of IL-12 (1° – rIL-12), the majority of d2-IFN-γ+ cells lost their ability to reproduce IFN-γ, whereas d2-IFN-γ cells remained negative for IFN-γ production. To determine whether the initial exposure to IL-12 is sufficient for d2-IFN-γ+ cells to maintain long-term memory and, for d2-IFN-γ cells to maintain induced IFNG activation, cells at d7 were further cultured with or without IL-12 (2° ± rIL-12) and restimulated as described earlier. The d2-IFN-γ+ cells that had received the initial IL-12 signal were able to reactivate IFNG expression by 90 and 89%, respectively, in response to further treatment with and without IL-12. Notably, IL-12-treated d2-IFN-γ cells behaved similarly in response to the second round of treatments (2° ± rIL-12) as did d2-IFN-γ+ cells in response to the first round of treatments (1° ± rIL-12) (Fig. 4). Taken together, these data provide strong evidence that, after primary activation and termination of TCR stimulation, the original instructive IL-12 signal is required by early IFN-γ-producing Th1 cells to develop into stable effector/memory Th1 cells.

Figure 4.

Th1 instruction signals and subsequent exposure to IL-12 are necessary and sufficient to establish long-term IFN-γ memory. Naïve T cells were primed in a Th1-cell culture with anti-CD3/28, followed by the IFN-γ capture assay. Presort and postsort histograms are shown. Isolated cells were cultured under neutral conditions with or without rIL-12 (1° ± rIL12) at day 0 and day 2. At day 7, IFN-γ reactivation was assessed by PMA and ionomycin stimulation. Cells receiving the initial IL-12 treatment were washed and further cultured under neutral conditions with or without rIL-12 (2° ± rIL12) and cytokine production was assessed as described above. The numbers indicate the percentage of IFN-γ-producing cells after PMA and ionomycin stimulation. The data shown are representative of at least three independent experiments.

Epigenetic regulation of the IFNG for IFN-γ memory by IL-12 signaling in early IFN-γ+ cells

Consistent with previous observations [6, 22], the responses of d2-IFN-γ+ and -IFN-γ cells to IL-12 involved the tyrosine phosphorylation of STAT4 (unpublished observation). Both d2-IFN-γ+ and -IFN-γ cells responded to 1° IL-12 treatment, whereas only d2-IFN-γ+ cells promptly adopted a stable effector/memory fate. This observation led us to examine the demethylation status of the IFNG promoter and CNS-1 that was induced in early effector cells that received additional IL-12 stimulation. Therefore, we performed methylation analysis in d2-IFN-γ+ and -IFN-γ cells, 1° IL-12-treated d2-IFN-γ cell-converted IFN-γ+ and -remained IFN-γ cells, and 1° IL-12-treated d2-IFN-γ+ cells that developed memory for IFNG reactivation.

As shown in Figure 5, at CNS-1, d2-IFN-γ cells were hypermethylated at all CpG sites and d2-IFN-γ+ cells were generally hypomethylated. In particular, compared with naive cells (Fig. 3A), d2-IFN-γ+ cells acquired 79% demethylation at CpG-4323 and 52, 40, 19, and 18% demethylation at CpG-4276, -4227, -4358, and -4291, respectively. Strikingly, in memory cells (IL-12-treated d2-IFN-γ+) consisting of 86% IFN-γ-producers, the demethylation levels of these CpGs reached their maximal values (76 to 86%) compared with naive cells (Fig. 4). In addition, considerable levels of demethylation (62 and 21%) were also detected at another two CpGs (-4397 and -4375) in memory cells compared with the levels in naïve cells (Fig. 3A). Interestingly, 1° IL-12-treated d2-IFN-γ cell-converted IFN-γ+ cells displayed a demethylation pattern similar to that of d2-IFN-γ+ cells, whereas 1° IL-12-treated d2-IFN-γ cell-maintained IFN-γ cells were hypermethylated similar to d2-IFN-γ cells. Moreover, memory cells were able to maintain their demethylation status without further IL-12 signaling, whereas d2-IFN-γ cells required repetitive IL-12 signals to induce the demethylation of CNS-1 (unpublished observations). Moreover, the imprinting patterns of the promoter among analyzed cell types mirrored those of CNS-1. In particular, compared with naive cells, 60 and 61% demethylation was observed in 1° IL-12-treated d2-IFN-γ+ cells at CpG-186 and -53, respectively (Fig. 5).

Figure 5.

Distinct DNA methylation profiles of d2-IFN-γ+ cells and -IFN-γ cells, 1° IL-12 treated d2-IFN-γ cell-converted IFN-γ+ cells and -maintained IFN-γ cells, 1° IL-12-treated d2-IFN-γ+ cells (as described in Fig. 4) at the IFNG locus. The data shown are representative of two independent experiments analyzed by bisulfate clone sequencing and pyrosequencing (not shown).

Thus, these results indicate that differential levels of demethylation of the IFNG promoter and CNS-1 in early IFN-γ+ and IFN-γ cells underlie their differential fates. In particular, after termination of TCR stimulation, IL-12 signaling leads to reinforced levels of demethylation of the IFNG locus and ultimately the development of effector/memory Th1 cells from early effectors.

Methylation status of the IFNG promoter and CNS-1 in periphery memory T cells of patients with RA

Our results described earlier have clearly shown the association of demethylation of the IFNG promoter and CNS-1 and the development of IFN-γ memory Th1 cells. This promoted us to further study their clinical relevance in Th1-involved autoimmune diseases such as RA [23, 24]. Here, we analyzed the peripheral distribution of IFN-γ effector/memory Th1 cells by delineating the DNA methylation status of the IFNG promoter and CNS-1, among naïve and memory CD4+ Th cell subsets between RA and healthy subjects. We also analyzed memory CD4+ Th cells directly isolated from synovial fluid (SF) of inflamed joints of RA patients. Memory CD4+ T cells consist of central memory and effector memory cell subsets, but IFN-γ are mainly produced by TEM cells [25]. Because naturally occurring CD45RO+ and CXCR3+ effector memory Th1 cells showed a similar pattern of epigenetic regulation which were inversely correlated with their frequencies of IFN-γ production, respectively (Fig. 1E), therefore in the following analyses we used total memory CD4+ cells instead of further purifying memory cell subsets. In the naïve T-cell compartment, lower than 5% demethylation was observed in both the promoter and CNS-1, and there were no differences between RA patients and healthy controls. In contrast, in the memory CD4+ Th-cell compartment of RA patients, 38% of such cells were demethylated, whereas the same cell type in healthy controls were 55% demethylated in the promoter. Similar to the promoter, in CNS-1, memory cells from RA patients were only 22% demethylated, whereas those from healthy controls were 39% demethylated. As expected, memory cells from SF were 85 and 67.5% demethylated in the promoter and CNS-1, respectively. Of note, the differences in both the promoter and CNS-1 between RA and healthy groups were significant (Fig. 6A). However, no differences in methylation levels were detected between purified IFN-γ-producers isolated from RA and healthy subjects (unpublished observations). In addition, the frequencies of IFN-γ-producing cells among effector/memory T cells of RA patients were significantly lower than those of healthy donors. The levels of DNA methylation correlated inversely with frequencies of IFN-γ-producing cells among memory T cells in both subject groups (Fig. 6B). Taken together, these results suggest that in RA patients reduced Th1 memory cells reside in the periphery blood, coincident with a selective recruitment of effector/memory Th1 cells to the site of inflammation.

Figure 6.

Coincidence of higher methylation levels of the IFNG promoter and CNS-1 in memory CD4+ Th cells in the periphery and lower cytokine production in RA patients. (A) Methylation levels of the IFNG promoter and CNS-1 in peripheral memory CD4+ Th cells of RA patients and healthy donors (HD) are shown. Ex vivo naive and/or memory CD4+ T cells from RA patients (blood, n = 8; SF, n = 2) and HD (n = 8) were analyzed for methylation levels using bisulfate PCR sequencing and pyrosequencing approaches. (B) IFN-γ production in peripheral memory CD4+ Th cells of RA patients and in HD are shown. Intracellular IFN-γ production was assessed after PMA and ionomycin stimulation. Data shown are from two independent cohorts of studies. Average methylation levels of the promoter and CNS-1 are depicted with box plots with 5th and 95th percentile. **p < 0.01; ***p < 0.0001.


We demonstrate here that early Th1-cell differentiation is accompanied by dynamic demethylation of CpGs at both the promoter and CNS-1 regions of the IFNG locus. Early epigenetic modification of the IFNG gene locus appears to be not only essential for early Th1 effector function, but that it is also indispensable for the establishment of stable functional Th1 cytokine memory. IL-12 signaling is required in this process after termination of TCR stimulation by strengthening the level of demethylation. Stable effector/memory Th1 cells are identified by hypomethylation of CpGs at the promoter and CNS-1 regions. In addition, analyzing methylation status of these two regions in memory Th-cell compartment can be used to evaluate the distribution and/or re-distribution of effector/memory Th1 cells. Our data indicate an epigenetic mechanism of CpG demethylation that governs Th1-cell commitment.

Despite the identification of multiple distal regulatory elements regulating transcription of the murine Ifng gene, only the CNS-54, CNS-6 (correspond to human CNS-1), and CNS+18–20 showed murine Th1-specific demethylation [26]. However, a T-cell-specific role for CNS-54 in regulating transcription of Ifng/IFNG has not yet been shown, and regulation by CNS+20 is not specifically required by T cells [27]. Although conflicting data regarding a murine Th1-specific demethylation of the Ifng promoter have been reported [28, 29], the human IFNG proximal promoter confers Th1/Tc1 selective expression [14, 15]. Therefore, the promoter and CNS-1 of the human IFNG are to date the defined T-cell-specific regulatory elements. Indeed, in the present study the inverse correlation between DNA methylation and IFNG expression and the heritability of hypomethylation at the IFNG promoter and CNS-1 clearly demonstrate a T-cell-specific epigenetic regulatory role for these two regions. Interestingly, CpG-186 and -53 in the proximal promoter which are targets of general transcription factors such as CREB/ATF and AP-1, were about 70–90% methylated in naïve T cells, consistent with results obtained from murine naïve T cells [18]. This finding implies a limited degree of chromatin accessibility that allows rapid transcription, yet a low level, in the course of primary activation.

Two mechanisms by which DNA demethylates have been described to date. Passive demethylation occurs during DNA replication because of a failure to methylate CpGs in the newly synthesized strand if levels of DNA methyltransferase I (DnmtI) are insufficient. Active demethylation, on the other hand, takes place in the absence of DNA replication and cell division [30]. Both passive and active mechanisms have been observed during genome-wide demethylation after fertilization in mouse [31]. Active demethylation has been demonstrated in several systems [32-37]. Our observations that the demethylation of selective CpGs at the IFNG gene locus occurs in cell proliferation-independent and -dependent manner strongly suggest that the early demethylation of the IFNG gene during Th1-cell differentiation is regulated by both active and passive mechanisms. Active demethylation has been proposed to be mediated by the demethylases methyl CpG-binding domain (MBD) 2b and/or MBD 4 [38], through a DNA repair-like pathway [39], through activation-induced cytidine deaminase [40, 41], or through Tet enzymes that catalyze cytosine 5-hydroxymethylation [42, 43]. In this study, although the underlying mechanism remains unclear, both active and passive demethylation of the IFNG locus may positively regulate its expression during early Th1 differentiation.

By analyzing a 3-day time course of methylation changes in Th1-primed naive cells, we observed dynamic demethylation at several CpGs, such as CpG-4227, -186, and -53. Interestingly, recent studies also revealed that during the transcription dynamic DNA demethylation occurs in the promoter of genes such as the estrogen receptor α (ERα)-responsive gene pS2 [44]. With respect to cytokine gene regulation, a direct structure-function relationship between the spatial organization of the chromatin around the mouse Ifng gene and its transcriptional potential has been demonstrated [45]. Thus, the dynamic demethylation in early Th1 differentiation could be associated with the transcriptional activation of IFNG, thereby allowing a certain degree of flexibility in the regulation of gene expression in response to distinct stimuli and/or in conjunction with dynamic alterations of chromatin conformation.

We have previously shown that demethylation primarily drives IFNG expression in purified ex vivo CD4+ T cells that do not originally produce IFN-γ [11]. However, the present data do not exclude the possibility that transcription factors (such as NFAT, PolII, T-bet, and STAT4, induced by TCR stimulation and instructive IL-12 signaling) affect IFNG activation by recruiting proteins that also modify the methylation status of IFNG. In this regard, a role for STAT4 in driving murine Th1 differentiation has been shown by STAT4-dependent changes in epigenetic histone modifications [46]. Importantly, the impact of CpG demethylation at the two Th1-specific regulatory regions in the IFNG locus on the development of memory cells was reflected by the differential demethylation status of d2-IFN-γ+ and -IFN-γ cells and their different cell fates. Thus, when continuously receiving an IL-12 signal, d2-IFN-γ+ cells with an initial level of CpG demethylation were able to maintain their effector function accompanied by the reinforced demethylation of these CpGs, whereas primed d2-IFN-γ cells needed repetitive IL-12 signaling to activate IFNG. Notably, the frequency of IFN-γ-producing cells in IL-12-cultured d2-IFN-γ cells remained consistent after further IL-12 treatment, but decreased in the absence of IL-12, suggesting that a fraction of the positive cells resemble d2-IFN-γ+ cells. This is most likely because the demethylation pattern of the IFN-γ+ cells sorted from 1° rIL12-cultured d2-IFN-γ cells resembles that of early d2-IFN-γ+ cells, indicating that under certain circumstances, activated non-IFN-γ-producing Th cells may be geared towards effector-like or memory-like Th1 cells. In this regard, it has been shown that murine TCR-Tg IFN-γ cells can acquire the effector/memory phenotype in vivo through a delayed differentiation program [47].

We found that IL-12-treated d2-IFN-γ+ cells exhibited a “poised” demethylated status of the IFNG gene similar to that of ex vivo memory Th1 cells. This demethylation imprinting of CpG sites at the proximal promoter and CNS-1 may provide direct access for the binding of lineage-transcription factors such as T-bet and STAT4, and for the binding of other general transcription factors, thereby serving as the basis for functional cytokine memory. DNA demethylation as the basis for such functional cytokine memory has been shown for murine Il4 and human IL2 [36, 48]. Moreover, our findings are in agreement with recent data from murine Th1 cells showing that enhanced IL-12 signaling through T-bet in the late phase of Th1 priming strongly correlates with the frequency of IFN-γ-producing cells in a later recall response [7]. Given the role of STAT4 in driving murine Th1 differentiation by STAT4-dependent changes in histone modifications [46, 49], it is attempting to speculate a IL-12-STAT4-dependent changes in DNA methylation of the IFNG gene may occur during the development effector/memory Th1 cells. Indeed, repetitive IL-12 stimulations lead to an inverse correlation between demethylation of the IFNG gene and IFNG expression in primed d2-IFN-γ cells, demonstrating that IL-12 is required not only in the differentiation of Th1 cells, but it is also required at multiple time points of Th1 differentiation to establish stable Th1 effector programs. Other cytokines such as IL-7 has been previously reported to enhance IFNG mRNA expression in activated T cells after 3 to 6 h of activation [50]. However, in our experimental system IL-7 stimulation alone or in combination with IL-15 did not lead to the transition from early effectors to effector/memory cells, therefore confirming that the effect of IL-12 stimulation cannot be replaced by IL-7 and/or IL-15.

Our findings of higher methylation levels of the IFNG promoter and CNS-1 in memory CD4+ Th cells of peripheral blood in RA patients compared with those in SF indicate reduced memory Th1 cells in the periphery, coincident with recruitment of effector/memory Th1 cells to the site of inflammation. In support of this data, several studies have shown Th1-cell activity in the joint and reduced Th1 response in the periphery [23, 24]. Thus, analyzing DNA demethylation status of IFNG may serve as a novel and reliable method for evaluating the distribution and/or redistribution of effector/memory IFN-γ-producing cells in patients with Th1-involved autoimmune diseases.

In conclusion, TCR signaling and instructive Th1 signals (such as IL-12) induce dynamic demethylation of the IFNG locus, and the latter of which stabilizes heritable and stable IFNG hypomethylation and expression during the transition from effector to effector/memory Th1 cells. Our findings have major implications for researchers attempting to modulate Th1 responses in vaccine design, autoimmune diseases, and organ transplantation.

Materials and methods

Media and reagents

Cells were cultured in RPMI 1640 supplemented with 1% glutamax, 100 U/mL penicillin, 100 μg/mL streptomycin (Invitrogen Life Technologies), and 10% human AB serum (PAA). PMA (5 ng/mL; Sigma-Aldrich) and 1 μg/mL ionomycin (Sigma-Aldrich) were used for stimulation. Brefeldin A (5 μg/mL; Sigma-Aldrich) was used to block cytokine secretion. Recombinant (r) IL-7, rIL-15, rIL-12, and rIFN-γ, each at 10 ng/mL (R&D systems), were used to induce cell differentiation as indicated in the text.

Ex vivo cell purification

Buffy coats from anonymous healthy adult donors and patient materials were obtained with local ethical committee approvals and informed consents. PBMCs were isolated by density gradient sedimentation using Ficoll-Hypaque (Sigma-Aldrich). CD45RA+CD31+CD45RO naïve, CXCR3+CD45RO+ and CD45RO+ memory CD4+ Th-cell subsets were purified as described [51]. Monocytes were separated by the MACS positive selection of CD14+ cells using CD14 microbeads (Miltenyi Biotec). The purity of the sorted populations was 95 to 99% as assessed by FACS.

Real-time quantitative PCR analysis

Total RNA from ex vivo or primed naïve CD4+ Th cells at days 1, 2, and 3 was extracted using an RNeasy Mini Kit (Qiagen) and was reverse transcribed using TaqMan Reverse transcription reagents (Roche Applied Biosystems) according to the manufacturer's recommendations. cDNA was analyzed for the expression of IFNG, TBX21, RUNX3 and the ubiquitin gene UBCH5B by real-time quantitative PCR analysis as previously described [11]. Primer sequence information is provided in Supporting Information Table 1.

In vitro generation of DCs

DCs were generated by culturing purified CD14+ monocytes (1 × 105 cells/mL), in the presence of 50 ng/mL GM-CSF and 50 ng/mL IL-4 (both R&D Systems) for 5 days and 1 μg/mL LPS (Strathmann Biotech AG) for an additional 2 days. CD80, CD83, and CD86 (PE, FITC, and allophycocyanin, respectively; BD Biosciences) expression was used to monitor the maturation status of DCs. All of these markers were upregulated (higher than 80%) in all experiments.

In vitro generation of Th1 cells by alloDC or TCR stimulation

For Th1 differentiation, naive CD4+ Th cells or CFSE-labeled (1 μM) naïve CD4+ Th cells (1–2 × 106/mL), as indicated, were primed with CFSE-labeled (5 μM; Molecular Probes) allogeneic monocyte-derived DCs (10:1) or plate-bound anti-CD3/CD28 (0.5 μg/mL/1 μg/mL; BD Biosciences) in the presence of 10 ng/mL rIFN-γ, rIL-12 (R&D Systems), and anti-IL-4 (10 μg/mL; BD Biosciences) for 2 or 3 days. Primed Th1 cells were analyzed for surface expressions such as CD45RO, IL18Rα (R&D Systems), CXCR3 (eBiosciences), and CD27, and intracellular cytokine productions such as IFN-γ and IL-2 using house conjugates.

IFN-γ capture assay

IFN-γ-producing Th cells from CMV-seropositive donors were isolated as previously described [11]. High frequency (>5%) of IFN-γ-producing Th cells were isolated with some modifications. Briefly, in vitro polarized Th1 cells, highly purified primary CD4+ T lymphocytes or human antigen-specific Th1 clones were stimulated with PMA and ionomycin for 4 h, followed by the detection and isolation of IFN-γ-producing Th cells using a cytometric cytokine secretion assay (Miltenyi Biotec) and FACSDiva™ (BD Biosciences). To prevent nonspecific signaling, 10 or 1 μg/mL anti-IFN-γ was applied to the cell suspension before adding the catch reagent and the detection antibody, respectively. IFN-γ-producing Th cells and their counterparts were also directly isolated after 2 or 3 days Th1 priming with plate-bound anti-CD3/CD28.

Intracellular cytokine staining

For IFN-γ reactivation, naive cells were removed from the stimuli 2 days after priming and washed, or sorted into d2-IFN-γ+ and -IFN-γ fractions. Subsequently, the cells were cultured for 1 to 3 weeks under one of the two following conditions: neutral (rIL-7/IL-15) or neutral plus rIL-12. The cell fractions were stimulated weekly with PMA and ionomycin for 4 h (the last 3 h with Brefeldin A), fixed, permeabilized, stained, and analyzed for intracellular IFN-γ. Alternatively, IFN-γ+ and IFN-γ cells were sorted for methylation analysis.

Bisulfate-specific PCR (BSP) sequencing, clone sequencing, and pyrosequencing

Genomic DNA extraction and the bisulfate conversion were performed as described [11]. Regions of interests in the IFNG locus were BSP-amplified by primers designed by Primer3 software. PCR was performed and PCR product sequenced using the PCR primers and methylation levels calculated as described [11]. Alternatively, PCR product was cloned using a TOPO TA cloning kit (Invitrogen) and sequenced using vector-based primers from both directions (Eurofins MWG Operon or GATC biotech, Germany). In addition, BSP products were pyrosequenced by an ordered service (Varionostic GmbH, Germany). Briefly, 40 ng of bisulfite-converted DNA was used in a PCR using the following program: 95°C for 3 min, 50 cycles of 95°C for 35 s, 55°C for 35 s, and 72°C for 40 s, and a final extension at 72°C for 5 min. Methylation levels were analyzed using Pyro Q-CpG software. Primer sequences for the bisulfate PCR and sequencing are listed in Supporting Information Table 1.

Cell cycle analysis by BrdU incorporation

Cell cycle analysis was performed using an APC BrdU Flow kit (BD pharmingen) according to the manufacturer's instructions. Briefly, naive CD4+ T cells were incubated with 10 μM BrdU under Th1 conditions. As a negative control, incubated cells were cultured in medium without activation. At indicated time points, cells were fixed, permeabilized, and stained with 7-AAD and anti-BrdU antibody. The stained cells were measured and the data were analyzed using FACSCalibur and Flowjo (Tree Star) software, respectively.


Unpaired two-tailed Mann–Whitney tests were used for statistical analysis, with GraphPad Prism software.


We would like to acknowledge the assistances of Flow Cytometry Labs at the Berlin-Brandenburg Center for Regenerative Therapies and Deutsches Rheuma-Forschungszentrum Berlin. Special thanks to Dr. Mairi Mcgrath for critical proofreading of this manuscript. This work was in part supported by the BMBF (S-T-THERA 01GU0802 to A.T. and A.R.), the Berlin-Brandenburg Center for Regenerative Therapies (1. BCRT grant to J.D. and A.T.), and SFB 633 (A.T.).

Conflict of interest

The authors declare no financial or commercial conflict of interest.


bisulfate-specific PCR


healthy donor


rheumatoid arthritis


synovial fluid