A Distinct Region of the Murine IFN-γ Promoter Is Hypomethylated from Early T Cell Development through Mature Naive and Th1 Cell Differentiation, but Is Hypermethylated in Th2 Cells

Cytokine gene methylation has been increasingly studied and recognized as an important mechanism of transcriptional repression in the immune system (1, 2, 3, 4, 5, 6, 7, 8). Methylation of the IFN-γ promoter was shown to correlate with transcriptional inactivity in mouse and human (9, 10, 11, 12, 13, 14). Methylation of a cytosine in a CpG motif at the −53 position of the IFN-γ promoter was originally observed in several murine Th2 cell clones that produced no measurable IFN-γ (11). In contrast, this site was hypomethylated in several Th1 cell clones that made IFN-γ upon in vitro stimulation (11). Subsequent studies in populations of murine CD8⁺ T cells suggested that the IFN-γ promoter is methylated in unstimulated cells, and that following stimulation with anti-CD3, anti-CD8 and anti-LFA-1 Abs, the promoter progressively demethylates until a stable, totally demethylated state is reached in the population of IFN-γ-producing cells (9, 10). The level of demethylation of the promoter in the cell population correlated with the amount of IFN-γ produced. This led to a model in which differentiating to cytokine production was associated with DNA demethylation of the cytokine gene. More recent experiments, however, have contradicted this model. In primary murine CD4⁺ T cells, it was demonstrated that, regardless of Th1 or Th2 culture conditions, all the cells were hypomethylated on the −53 site, as well as a second position closer to the transcription start site (15).

Many of the CpG sites, including the −53 site, are conserved in the human and mouse genomes. The first published study on human neonatal CD4⁺ and CD8⁺ T cells, as well as thymocytes, found that the promoter of the IFN-γ locus was methylated (16). A subsequent study compared the methylation status of IFN-γ in human adult CD3⁺CD45RO⁻ PBMC with that of neonatal cord blood T cells. The authors found the promoter hypermethylated in CD4⁺, but not CD8⁺ neonatal subpopulations, and comparatively hypomethylated in both subpopulations of the adult (13). Another study examining the in vitro differentiation of human peripheral blood CD4⁺ T cells into Th1 or Th2 subtypes, found the promoter hypomethylated in undifferentiated and Th1 cells, but methylated in Th2 cells (14).

These somewhat conflicting reports prompted us to re-examine the methylation status of the IFN-γ promoter during CD4⁺ and CD8⁺ T cell development in the mouse. We show that the murine IFN-γ promoter is hypomethylated in a well-defined region of the gene in naive populations of both CD4⁺ and CD8⁺ T cells as well as in NK cells. This hypomethylated status was detectable very early in T cell development, but not in cells that do not make IFN-γ. We also found increased methylation of the promoter after naive CD4⁺ T cells were stimulated in vitro under Th2-polarizing conditions. These results suggest a new model for chromatin structural changes in the mouse IFN-γ gene during T cell differentiation in which the locus opens up very early in development and does not close again except in certain end-stage effector cells, which may undergo remethylation.

The 5C.C7 mouse is a B10.A TCR-5C.C7 transgenic (Tg)², Rag-2^−/− with T cells that specifically recognize pigeon cytochrome c (PCC) peptide 81–104 bound to I-E^a and I-E^k (17). The OT-1 mouse is a C57BL/6 TCR-OT-1 Tg, Rag-1^−/− with T cells specific for OVA peptide 257–264 bound to H-2K^b (18). CD3ε^−/− mice, Rag-2-deficient mice and common γ-chain (γ_c)/Rag-2 double knockout mice (γ_c^−/−Rag-2^−/−), all on a C57BL/10 or B10.A background, were obtained from the National Institute of Allergy and Infectious Diseases (NIAID) breeding colony at Taconic Farms (Germantown, NY). The mice were housed in our American Association for the Accreditation of Laboratory Animal Care-approved, specific pathogen-free colony and protocols were approved by the NIAID animal care and use committee. The NIH 3T3 transformed fibroblast cell line (National Institutes of Health/Swiss strain, CRL-1658) was obtained from the American Type Culture Collection (Manassas, VA).

CD4-TriColor, CD8-TriColor, and CD4-FITC Ab conjugates were purchased from Caltag Laboratories (Burlingame, CA). Thy1.2-FITC, DX-5-FITC, B220-FITC, TCR-Vβ3-PE, TCR-Vα11-FITC, CD44-PE, CD62L-allophycocyanin, and CD4-PE Ab conjugates were purchased from BD Pharmingen (San Diego, CA). Flow cytometry was performed on a BD Biosciences FACSort or FACSCalibur flow cytometer (San Jose, CA), using CellQuest software for analysis.

Spleens were taken from 5C.C7 mice, crushed on nylon screens and the cell suspensions activated with 10 μM PCC for 4 days in complete medium containing 50% RPMI 1640, 40% Eagle-Hank’s amino acid, 10% FCS, 4 mM glutamine, antibiotic/antimycotic solution (100 U of penicillin, 100 μg/ml streptomycin, 2.5 μg/ml fungizone), plus an additional 100 μg/ml penicillin, 100 μg/ml streptomycin, and 50 μM 2-ME. The cells were then rested for 7–10 days in medium containing 10 U/ml rmIL-2 (R&D Systems, Minneapolis, MN). After resting, the cells were purified on a Ficoll gradient. They were >90% Vβ3 TCR⁺ CD4⁺ by flow cytometry.

Naive CD4⁺ and CD8⁺ T cells were taken from cervical, axillary, inguinal, and mesenteric lymph nodes of the 5C.C7 and OT-1 mice, respectively. The lymph nodes were pooled and crushed on nylon screens. These cell suspensions contained ∼90% CD4⁺ or CD8⁺ T cells, respectively, as measured by flow cytometry. Bone marrow was prepared by clipping the ends of the femur and tibia, followed by flushing with PBS through a 20-gauge needle and filtering on a Falcon 70-μm nylon cell strainer (BD Biosciences). Thymuses were removed from mice and crushed on nylon screens. The thymocyte cell suspensions from all mice were ∼95% Thy1.2⁺. Livers and kidneys were removed from the mice, and cell suspensions were prepared as previously described for T cells, before DNA extraction.

Lymph nodes were taken from B10.A mice, pooled and crushed. The CD4⁺ and CD8⁺ T cells were purified by magnetic bead separation using CD4⁺ and CD8⁺ T cell purification kits (Miltenyi Biotec, Auburn, CA). The final CD4⁺ and CD8⁺ T cell purity was >95% after separation. For a purified B cell population, spleens were taken from CD3ε^−/− mice, crushed on nylon screens and lysed twice with ACK RBC lysis buffer (BioSource International, Rockville, MD). After washing in PBS and resuspending in complete medium, the cells were mixed with cell suspensions taken from the mesenteric lymph nodes of these mice, prepared as described. The pooled cells were purified by negative selection using anti-CD43, anti-CD11c, anti-CD11b, and anti-DX-5 microbeads (Miltenyi Biotec), according to the manufacturer’s protocol. Purity was verified by flow cytometry to be 96% B220⁺. An enriched population of NK cells was purified for analysis from Rag-2^−/− spleens. The spleens were crushed, lysed in ACK lysis buffer, stained with anti-DX-5 FITC (BD Pharmingen), and sorted in the NIAID sorting facility to a purity of 99%.

Spleens and lymph nodes from 5C.C7 mice were taken and crushed on nylon screens as previously described. Spleen and lymph node cells were mixed in a 3:1 ratio and then stimulated at a concentration of 5 × 10⁵ cells/ml with 10-μM PCC in the presence of 10 μg/ml anti-mouse IL-4 Ab 11B11 (BD Pharmingen), 10 ng/ml rmIL-12 (PeproTech, Rocky Hill, NJ) and 400 ng/ml rmIFN-γ (PeproTech) for Th1 expansion, or 2 ng/ml anti-mouse IL-12 neutralizing Ab (R&D Systems), 2 μg/ml anti-mouse IFN-γ Ab (R&D Systems) and 1000 U/ml rmIL-4 (PeproTech) for Th2 expansion. The cells were incubated for 4 days in complete medium and rested 3 days in the presence of 10 U/ml rmIL-2 (R&D Systems). After resting, the cells were purified on a Ficoll gradient before analysis.

The 3-kb IFN-γ promoter-driven luciferase pGL3 plasmid (a gift from Dr. H. Young, National Cancer Institute, Frederick, MD) was methylated overnight at 37°C with SssI (CpG) methylase (New England Biolabs, Beverly, MA). Failure of HpaII to digest the plasmid verified successful methylation. Transient transfections of the IFN-γ luciferase plasmids were conducted on preactivated CD4⁺ T cells (previously described) using a microparticle-mediated gene transfer apparatus (Bio-Rad, Richmond, CA) as previously described (19). The pRL-CMV vector Renilla luciferase plasmid was cotransfected as an internal control to normalize transfection efficiencies. After transfection, the cells were stimulated overnight with 10 μg/ml plate-bound anti-CD3 and soluble anti-CD28. Luciferase activity was reported as a percentage of the unmethylated plasmid transfection, after normalization. The results are the average of two separate experiments.

DNA was extracted with the Easy DNA kit (Invitrogen Life Technologies, Carlsbad, CA) and DNA Extraction kit (Serologicals, Norcross, GA), followed by 5 rounds of phenol/chloroform extraction. The DNA was modified with the bisulfite reaction as previously described (19). Briefly, DNA was bisulfite transformed, followed by PCR using PCR primers MIFNGBIS1, 2, 4, and 6 described by Fitzpatrick et al. (10). The PCR products were subcloned, minipreps were made from 15 to 16 clones, and their DNAs were sequenced. Complete conversion of C to U was confirmed by the absence of C in non-CpG sequences.

All enzymes were purchased from New England Biolabs. A total of 10 μg of DNA per reaction was simultaneously digested overnight at 37°C by 30 U of EcoRI alone or EcoRI plus 10 U of SnaB1, 20 U of Aci1, 20 U of BsmB1, or 2 U of BceA1. Reactions containing BsmB1 were incubated at 55°C for 2 h following overnight digestion. The digestions were run on 0.6% agarose and the gels were blotted onto Nytran⁺ membranes using the Turboblotter system (Schleicher & Schuell, Keene, NH). The IFN-γ probe was made by 40-cycle PCR amplification from a Puc8 13-kb IFN-γ plasmid obtained as a gift from Dr. H. Young, using the following primers: 5′-GCACGTTGACCCTGAGTGATTTGTAGTAGGTA-3′ and 5′-CATGCCACAAAACCATAGCTGTAATGC-3′. The PCR conditions were 94°C for 1 min, (94°C for 15 s, 51.3°C for 15 s, and 65°C for 1 min) × 40 followed by 65°C for 10 min. The PCR product was purified using a Geneclean II kit (BIO 101, Carlsbad, CA) and sequenced to confirm it was from the IFN-γ promoter. Radiolabeled probes were made using RediPrime II kits (Amersham Biosciences, Piscataway, NJ). Membranes were hybridized with ExpressHyb solution (BD Clontech, Palo Alto, CA) according to the manufacturers protocol. Radioactivity was detected using a STORM, a THYPOON (Amersham Biosciences) or the FLA-5000 phosphor imager (FujiFilm, Stamford, CT). All densitometry analyses were performed using Gel-Pro Analyzer 3.1 software (Media Cybernetics, Silver Spring, MD) and are expressed as percentage of the demethylated fragment.

To determine the sensitivity of this Southern blotting technique for detecting hypomethylated DNA, we digested thymocyte genomic DNA, which is hypomethylated at this locus, either with EcoR1 alone (yielding a 1.5-kb fragment), or by EcoR1 plus Aci1 (yielding a 1-kb fragment) simultaneously, as described. Then we mixed the EcoR1 and the EcoR1 plus Aci1 digested DNA at a 5:95, 10:90, 25:75, 50:50, 75:25, 90:10, and 95:5 ratio. After blotting and hybridization, we were able to detect either DNA fragment that was added at as little as 10%, but not at 5%. This delineates our sensitivity for detecting demethylation at around 10%. To determine whether we could detect complete methylation of the IFN-γ promoter with each of our methylation-sensitive enzyme digestion conditions, thymocyte genomic DNA was methylated in vitro by SssI, as described earlier for the pGL3 plasmid, followed by digestion. Southern blotting revealed only the 1.5-kb EcoR1 fragment (data not shown). We concluded that this technique was sensitive and accurate enough to allow us to differentiate between hypomethylated and hypermethylated DNA and to measure most changes in the methylation status of the IFN-γ promoter.

We have annotated CpG sites in the IFN-γ promoter (+1 to +8) from the transcription start site to −380 bp upstream. Two CpG sites inside the coding region are numbered 0 and −1, moving downstream. A diagram of the IFN-γ promoter using these annotations is shown in Fig. 1,A. To determine whether methylation of the IFN-γ promoter correlates with a loss of transcription, we transfected previously activated murine CD4⁺ T cells, which produce IFN-γ upon stimulation, with a CpG methylated or unmethylated pGL3 plasmid containing luciferase driven by a 3-kb IFN-γ promoter. The plasmid containing the IFN-γ promoter was methylated in vitro with SssI. HpaII digestion verified successful methylation (data not shown). Preactivated CD4⁺ T cells transfected with the methylated plasmid followed by stimulation with plate-bound anti-CD3 and soluble anti-CD28 showed a 99% reduction in luciferase activity, when compared with those transfected with the unmethylated plasmid (Fig. 1 B). We have previously established that pGL3 vector methylation has no effect on transcription of luciferase, suggesting that the transcriptional inhibition we observed was due to promoter methylation only (19). Therefore, we conclude that methylation of the IFN-γ promoter can inhibit transcription.

Because methylation of the IFN-γ promoter correlates with reduced transcription, we examined the methylation status of the promoter in IFN-γ-producing and nonproducing cell types. Bisulfite sequencing analysis identified a hypomethylated region in the IFN-γ promoter in preactivated CD4⁺ T cells that was hypermethylated in the NIH 3T3 fibroblast cell line (Table I). By this method, CpG sites 1–6 (see diagram, Fig. 1,A) upstream of the transcription start site were hypomethylated in T cells (Table I). The sites 0 and −1 inside the coding sequence were consistently hypermethylated in the T cells as were sites 0 to +5 in the fibroblast cell line. Sites −1 and +6 were not tested in the fibroblasts. To confirm this result, Southern blotting was performed on DNA from the preactivated T cells. DNA from the T cells was digested by EcoRI, yielding a 1.5-kb fragment from +238 to −1292 containing CpG sites −1 through +8. This fragment was visualized by hybridization to a 462 bp probe spanning from −130 to −592 (Fig. 1,A), which was synthesized by PCR. Digestion of the EcoRI fragment by CpG methylation-sensitive restriction enzymes revealed hypomethylation of specific sites in this region. In DNA from 5C.C7 preactivated CD4⁺ T cells, the EcoRI fragment was cut by AvaI at site 2, SnaBI at site 3, AciI at site 6, and BsmBI at site 7 (Fig. 1,C and data not shown). BceAI was unable to cut at site 8. These data show that sites 2, 3, 6, and 7 were >90% hypomethylated, whereas site 8 was >90% methylated. In DNAs from non-IFN-γ-producing cell types such as the liver and kidney, all of the above enzymes were unable to cut the 1.5-kb EcoRI fragment (Fig. 1 C and data not shown). Therefore, these sites were fully methylated in these latter cell types. All these results are consistent with the bisulfite data. Thus, we conclude that the IFN-γ promoter has a well-circumscribed region of around 300 bp that can be demethylated in T cells.

To examine the methylation status of T cells in vivo, Southern blotting was performed on DNA directly from the lymph nodes of mice. A mixture of B10.A lymph nodes was examined containing ∼30% CD8⁺ T cells, 30% CD4⁺ T cells and 40% other cell types, such as B cells and NK cells, as measured by flow cytometry (data not shown). Fig. 2,A shows that DNA from these cells was partially digested by SnaBI (50%), AciI (45%), and BsmBI (35%), indicating substantial demethylation at sites 3, 6, and 7, but full methylation at site 8. We postulated that this intermediate pattern resulted from a mixed population of non-T cells and T cells in various states of activation. To test this, we purified CD4⁺ and CD8⁺ T cells from B10.A lymph nodes by negative selection with mAbs and magnetic beads. The T cell preparations were >95% pure. As shown in Fig. 2 B, digestion with SnaB1 revealed >90% demethylation at site 3 in the DNA from CD8⁺ T cells. In contrast, DNA from CD4⁺ T cells was only 75% demethylated at site 3. In one experiment enough cells were obtained to examine the other sites in the CD4⁺ T cell DNA. In this case, the Aci1 site 6 was 82% demethylated and the BsmB1 site 7 was 84% demethylated, whereas the control BceA1 site 8 had no detectable demethylation. These observations suggest that the CD4⁺ T cell population contains a subset of cells in which the IFN-γ promoter is hypermethylated.

One such subset could be within the preactivated T cell pool, of which our CD4⁺ T cell populations contained ∼10–15% as measured by CD44 and CD62L staining (data not shown). To examine a fully naive T cell compartment, DNA was extracted from cell suspensions of lymph nodes from either 5C.C7 or OT-1 TCR Tg Rag^−/− mice. Flow cytometry analysis demonstrated that the CD4⁺ T cells from the 5C.C7 lymph nodes and the CD8⁺ T cells from the OT-1 lymph nodes comprised ∼90% of the total cells in each preparation, and they contained only 1.2 or 0.4% CD44^high, CD62L^low T cells, respectively (data not shown). Therefore, the T cell preparations from these Tg mice were mostly naive. In Fig. 2,C, digestion of the DNA from these cells by SnaBI, AciI, and BsmBI demonstrated that the IFN-γ promoter is >90% hypomethylated at sites 3, 6, and 7 in naive CD4⁺ T cells and mostly unmethylated at these same sites in CD8⁺ naive T cells. The inability of BceAI to cut the DNA fragment in either cell population indicates that site 8 is hypermethylated in both cell types (Fig. 2 C). These observations on DNA from TCR Tg cells demonstrate that truly naive T cells from peripheral lymphoid tissues are hypomethylated in a discrete region of their IFN-γ promoter. The nature of the CD4 subset(s) in normal lymph nodes that contains methylated IFN-γ promoter DNA will be explored later.

To examine the IFN-γ promoter methylation status of non-T cells in lymphoid tissues, we examined B cells and NK cells. B cells were purified from CD3ε^−/− mice by magnetic bead selection as described in Materials and Methods. The separated cells were 96% B220⁺ (data not shown). Southern blotting of DNA from these cells indicated that the IFN-γ promoter was hypermethylated at sites 3, 6, 7, and 8 (Fig. 2,D). NK cells were purified by cell sorting for DX-5⁺ cells from the spleens of Rag-2^−/− mice. The purity achieved was 99% DX-5⁺ (data not shown). Because the number of cells recovered was small (∼5 million), we were only able to examine one site in the promoter. In Fig. 2 D, the digestion of NK cell DNA by SnaB1 indicates that the promoter was >90% hypomethylated at site 3. We conclude that the IFN-γ promoter is hypomethylated in NK cells, but hypermethylated in B cells.

The surprising observation that DNA from naive T cells was hypomethylated in a discrete area of the IFN-γ promoter, whereas DNA from B cells or liver cells was not, led us to ask when this chromatin structural change first took place in T cells. To determine whether hypomethylation of the locus was already evident during maturation of T lymphocytes in the thymus, DNA was extracted from the thymuses of B10.A, 5C.C7, and OT-1 mice. The cell suspensions prepared from these thymuses were ∼95% Thy1.2⁺ and contained mostly CD4/CD8 double positive and single positive thymocytes with very few double negative cells (∼3%) (data not shown) (20). In Fig. 3, Southern blots of the thymuses from B10.A, OT-1, and 5C.C7 mice showed hypomethylation at sites 3, 6, and 7, and hypermethylation at site 8. The hypomethylation was to the same degree as that observed in the naive CD4⁺ and CD8⁺ Tg T cells from lymph nodes.

We then tested whether the hypomethylated promoter existed at a very early stage of T cell development. DNA was extracted from the thymuses of Rag-2^−/− mice, which are arrested at the double negative stage 3. This cell population contained >95% double negative thymocytes (data not shown) (21). Southern blotting of DNA from these cells again showed that the IFN-γ locus was hypomethylated at sites 3, 6, and 7 and hypermethylated at site 8 (Fig. 3). Therefore, hypomethylation of the IFN-γ promoter appears at or before the double negative stage 3. To attempt to find an earlier hemopoietic developmental stage at which the IFN-γ promoter was (de)methylated, we analyzed DNA from the bone marrow. We used bone marrow from γ_c^−/−Rag-2^−/− mice, which contains no mature T, NK, or B cells (22) to rule out possible contamination by these cell types circulating through this tissue. Fig. 3 shows a small amount of demethylation in DNA from these cells. Therefore, there must be some cell type in the bone marrow that is hypomethylated at the IFN-γ promoter, but it is not a T cell, B cell, or NK cell. It may be from the myeloid lineage, but we have not made any attempt to characterize it further.

After discovering the remarkable stability of the hypomethylated IFN-γ promoter in T cells throughout their development, we wanted to know whether this state changes after T cell activation in the periphery, to explain the 15–25% methylation pattern seen in DNA from CD4⁺ T cells in normal lymph nodes. To investigate this question, naive 5C.C7 CD4⁺ T cells were stimulated with splenic APC and Ag for 4 days and then rested for several days. As shown in Fig. 1,C, the hypomethylation pattern seen at the IFN-γ promoter was the same as that seen with DNA from naive T cells (compare with Fig. 2,C). Other laboratories, however, have reported that Th2-differentiated T cells are methylated at SnaB1 site 3 of the promoter (11, 14). To determine whether the conditions for activation were a critical variable, naive 5C.C7 CD4⁺ T cells were stimulated with splenic APC and Ag in the presence of Th1 or Th2 polarizing conditions, and the DNA extracted after 7 days. Aliquots of the Th1 and Th2 cell preparations were also stimulated for 48 h by anti-CD3 plus anti-CD28, and IFN-γ and IL-4 measured in the supernatants by ELISA. In Fig. 4,A, the Th1 cells produced 551 ng/ml IFN-γ, whereas the Th2 cells produced a 50-fold lower amount at 11 ng/ml. The Th1 cells produced no measurable IL-4, whereas the Th2 cells produced 67 ng/ml IL-4. Similar results were obtained in four separate experiments. The geometric mean of the fold difference in IFN-γ production between Th1 and Th2 cells was 77 ÷/× 1.43. Thus, the Th1 and Th2 cells expressed the appropriate cytokine profiles. Southern blotting revealed a pronounced increase in methylation of the IFN-γ promoter in Th2-polarized cells (Fig. 4,B). All three sites (3, 6, and 7) that were consistently >90% hypomethylated in naive CD4⁺ T cells were now partially methylated. Densitometry calculations on Southern blots from four separate experiments showed an average of 56% methylation at site 3 (Table II). The other two sites were around 30% methylated. In contrast, Southern blotting of DNA from CD4⁺ T cells stimulated under Th1 conditions showed far less and more variable methylation at these sites (<20%) (Fig. 4,B and Table II). Overall, Th2 cells were between 2- and 3-fold more methylated than the Th1 cells at sites 3, 6, and 7 (p < 0.05). Therefore, the IFN-γ promoter appears to greatly change its methylation status in CD4⁺ T cells after stimulation under Th2 polarizing conditions. Such cells could at least in part account for the intermediate methylation pattern seen in the IFN-γ promoter of CD4⁺ lymph node T cells (Fig. 2 B). This does not, however, preclude other subsets (such as T regulatory cells) from also playing a part.

Once a naive T cell is presented with its Ag, it will proliferate and produce cytokines. Epigenetic changes occur at the cytokine loci that are necessary for the T cell to optimally transcribe them. It has been shown that the promoters of IL-2 and IL-4 are CpG-methylated in murine naive T cells (19, 23). These genes become increasingly hypomethylated as their products accumulate, either by an active demethylation, or passive, cell division-dependent demethylation. In contrast, we have demonstrated the distinctiveness of the mouse IFN-γ gene, for which the promoter already exists in an unmethylated state in naive CD4⁺ and CD8⁺ T cells, as well as throughout T cell development. In our study, the methylation of multiple sites in the IFN-γ promoter was analyzed in T cells directly ex vivo. This is the first study to demonstrate a discrete hypomethylated region in the mouse IFN-γ promoter flanked by methylated CpGs in naive CD4⁺ and CD8⁺ T cells. The region we identified has been shown to be sufficient for murine IFN-γ production, and the binding sites for NF-κB, NF-AT, AP-1, and other transcription factors have been mapped here (24, 25, 26). In contrast, previous studies by Fitzpatrick et al. (9, 10) in murine CD8⁺ T cells implied that the IFN-γ promoter is methylated in naive cells. The data they presented, however, were from T cells that had undergone several rounds of in vitro stimulation, which can cause physiological (Fig. 4) and/or nonphysiological changes in methylation status (27). To determine whether in vitro stimulation could have such a methylation effect for CD8⁺ T cells in our hands, we cultured naive CD8⁺ T cells from OT-1 mice, as described by Reis e Sousa et al. (28). We found that the preactivated T cells, recovered in high yields, remained hypomethylated at the SnaB1 site 3 (data not shown). This discrepancy with the observations of Fitzpatrick et al. (9, 10), however, could be due to differences in the stimulation conditions used because we showed for CD4⁺ T cells that only Th2 polarizing conditions yielded substantial methylation at the IFN-γ promoter (Fig. 4 B). Further studies will be necessary to clarify this situation.

The IFN-γ promoter was methylated in B cells and in several nonhemopoietic cell types we tested. This is one likely mechanism for silencing transcription of the gene in these cells. The B cells in our system are from a CD3ε^−/− mouse, genetically devoid of T cells. Therefore, many of them (B2 cells) are likely to be naive. Interestingly, there have been reports of human tonsillar B cells producing IFN-γ in vitro (29, 30). Also, some human B cell lines are able to produce IFN-γ in response to IL-2 or PMA, and they were also shown to be hypomethylated at the IFN-γ promoter (31). Thus, it is possible that human B cells are able to acquire a competent IFN-γ promoter after several rounds of stimulation. Studies on the methylation status of IFN-γ in human T cells are also interesting. Several reports suggest that the human IFN-γ promoter is methylated in naive CD4⁺ T cells and thymocytes (13, 14, 16). This is an apparent distinction from our studies in the mouse. Perhaps DNA methylation plays a different role in regulating IFN-γ expression in human than in mouse.

The hypomethylated region of the IFN-γ promoter was seen as early as the double negative stage 3 of thymocyte development (Fig. 3). This hypomethylated promoter may be an epigenetic “tag,” acquired very early on in T cell development, at or before migration of prothymocytes to the thymus, allowing the cell’s progeny to acquire a competent IFN-γ promoter. Unfortunately, the limits of our assays were such that it was not possible to extract enough cells from an earlier thymocyte developmental stage to pinpoint exactly when the locus is first demethylated. Nonetheless, we are suggesting that IFN-γ is unique in its methylation pattern among other murine T cell cytokines, such as IL-2 and IL-4. One possible reason for the early demethylation of the IFN-γ promoter might be because of the close evolutionary relationship between T cells and NK cells. As shown in Fig. 2 D and in a recent publication by Tato et al. (32), the IFN-γ promoter and intron 1 are hypomethylated in NK cells. The innate immune system needs to respond quickly to pathogens with effector cytokine production, thus requiring an open, hypomethylated IFN-γ gene. If T cells evolved from NK cells, they may have preserved this aspect of IFN-γ gene regulation in the mouse because IFN-γ has little or no effect on T cell development and is needed quickly by naive CD8⁺ T cells during their initial response to pathogens. B cells might have branched off from this T cell-NK cell lineage before a demethylation mechanism was established, or, perhaps, subsequently remethylated their promoter. Human CD4⁺ T cells might also have remethylated the IFN-γ promoter as discussed.

Why is the promoter of the murine IFN-γ gene unmethylated in naive T cells whereas those of other cytokine genes are not? Perhaps it is because other transcriptional repression mechanisms, such as histone methylation and acetylation status, are adequate for IFN-γ repression in murine naive T cells. For example, chromatin immunoprecipitation studies in murine CD8⁺ T cells demonstrated that the IFN-γ gene is not associated with acetylated histones in neutrally stimulated or Tc2 polarized cells, both of which, like the naive CD8⁺ T cells, do not produce IFN-γ (33). This suggests the locus is associated with deacetylated histones in these cells, and thus is inaccessible. Furthermore, IFN-γ production is partially dependent on signals from other cytokines such as IL-2 and IL-12; e.g., naive T cells from IL-2 knockout mice don’t produce much IFN-γ in the absence of exogenous IL-2 or IL-12 (34, 35). Thus, the primary level of control may lie with these other cytokine genes.

Our data differ somewhat from a previous publication of Falek et al. (15). In agreement with our findings, they reported that hypomethylation exists at CpG sites 2 and 3 in the thymus and in primary T cells from normal BALB/c mice. However, the Southern blot they used was unable to detect any unmethylated IFN-γ promoters in the bone marrow. We have detected hypomethylation in a small subset of non-T, non-B, non-NK cells in the bone marrow. These cells may be dendritic cells or macrophages, as numerous reports have shown that these cells produce IFN-γ in some situations in both human and mouse (36, 37, 38). Falek et al. (15) also could not detect methylation in Th2-differentiated cells. As with Fitzpatrick et al. (9, 10), this may again be due to differences in the stimulation conditions. Falek et al. (15) activated naive CD4⁺ cells with anti-CD3 and anti-CD28 plus IL-4 in the absence of APCs and without blocking IFN-γ and IL-12. In our case, we used Ag and APC with IL-4, while blocking IL-12 and IFN-γ. Our Th2 data are consistent with other reports (11, 14) and can be explained in several different ways. First, a remethylation event may occur as CD4⁺ T cells differentiate into Th2 cells and then methylated-DNA binding proteins could be recruited to help silence the gene. It has been shown that murine single positive CD4⁺ thymocytes can remethylate their IL-4 promoter as they differentiate into naive CD4⁺ T cells (23). Alternatively, a small subset of CD4⁺ T cells, undetectable by our assays, may be methylated at the IFN-γ promoter in the naive state, and these cells might preferentially proliferate to become Th2 cells in the presence of IL-4 and the absence of IL-12 and IFN-γ. There is also the possibility that the remethylation is an artifact of in vitro stimulation (27). In any case, methylation is unlikely to be the only silencing mechanism in Th2 cells. IFN-γ chromatin structural studies in the mouse have demonstrated that the spatial distribution of the gene itself is more compacted in Th2 than in Th1 cells (39). This could mean that the promoter is spatially inaccessible in Th2 cells. Methylated DNA may contribute to this state, or perhaps methylation is a consequence of the compaction of the gene. Finally, Th2 cells are likely not the only cells that have undergone chromatin structural changes to silence IFN-γ production in the mouse. In whole lymph node CD4⁺ T cell populations, we found that 15–25% of the cells appeared to have a methylated IFN-γ promoter (Fig. 2). One other CD4⁺ cell type that might contribute to this phenotype is T regulatory cells (40).

The current prevailing view is that Th1 cells demethylate their IFN-γ promoter as they differentiate, whereas Th2 cells maintain CpG methylation of the promoter. We have concluded that this is not the case in the mouse. Naive T cells are hypomethylated in a discrete region of the IFN-γ promoter from early in T cell development, and they mostly maintain that status as they differentiate into Th1 cells. We prefer the hypothesis that a remethylation event occurs when a naive T cell differentiates into a Th2 cell, but further study is needed to resolve this issue.

We thank Drs. Mike Lenardo, Nevil Singh, and Howard Young for their critiques of the manuscript, Erika Schmierer for help in the redaction, and all members of the Laboratory of Cellular and Molecular Immunology for their support, friendship, and advice. We acknowledge Elizabeth Majane for maintaining the transgenic mice and Dr. Gregoire Altan-Bonnet for assistance in preactivating CD8⁺ T cells.