Cutting Edge: Innate Production of IFN-γ by NK Cells Is Independent of Epigenetic Modification of the IFN-γ Promoter¹

The production of IFN-γ is essential for the resistance to many viral, bacterial, and parasitic infections (1). Although T cells are a major source of IFN-γ, their ability to respond to infection is part of the adaptive response and takes days to develop a prominent IFN-γ response. These kinetics are in concord with studies which established that the ability of T cells to produce IFN-γ correlates with proliferation-dependent remodeling of the repressed IFN-γ locus (2, 3). In naive CD4⁺ T cells, the IFN-γ locus is generally regarded as being in a repressed state that is associated with methylation of CpG motifs and resistance to DNase I. However, after TCR stimulation and concurrent costimulatory signals, a proliferation-dependent remodeling of the IFN-γ locus occurs that leads to a heritable “open” configuration (4, 5). These changes are associated with demethylation of CpG motifs and hypersensitivity to DNase I (6, 7, 8).

NK cells are a major innate source of IFN-γ produced rapidly during many infections before the development of an adaptive response (9). Thus, in contrast to the delay in T cell activation and effector function, NK cells are able to respond quickly to different infectious agents and contribute to the early control of these infections until an adaptive immune response can develop. Many of the same cytokines (IL-12, IL-18) and transcription factors (T-bet, STAT4, NF-κB) implicated in the regulation of T cell production of IFN-γ are also involved in the ability of NK cells to produce IFN-γ (10, 11, 12). Although NK and T cells use similar signaling pathways, and although there is evidence that “memory” phenotype CD8⁺ T cells may act as an innate source of IFN-γ independent of antigenic stimulation (13), there are important differences between these cells types. In particular, cytokines alone fail to stimulate naive T cells to produce IFN-γ and the interaction of the TCR with its specific Ag is a prerequisite to initiate the production of IFN-γ (14). In contrast, NK cells can be stimulated directly with cytokines to produce high levels of IFN-γ, yet the molecular basis for the different requirements for NK and T cells to produce IFN-γ is not well understood.

Given the role of chromatin remodeling in the regulation of IFN-γ production by T cells, studies were performed to determine whether proliferation-mediated remodeling of the IFN-γ locus is involved in the ability of NK cells to produce IFN-γ. The studies presented here reveal that resting NK cells can make IFN-γ within hours of cytokine stimulation, but these events are independent of proliferation. Moreover, the IFN-γ locus in resting NK cells, in contrast to naive T cells, appears to exist in a more open configuration that correlates with the ability of these cells to make IFN-γ rapidly. Together, these findings reveal that differences in the structure of the IFN-γ gene provide a molecular basis for the different kinetics of innate and adaptive IFN-γ responses.

NK cells isolated from the spleens of either female C57BL/6 or C57BL/6 RAG^−/− donors in complete RPMI 1640 (Life Technologies, Gaithersburg, MD) supplemented as previously described (15) were used as the source of a CD25^lowCD28⁻ resting NK cell population. To assess T and NK cell responses in whole splenic cultures, cells from C57BL/6 mice were stimulated with recombinant human IL-2 at 2000 U/ml (Chiron, Emeryville, CA), IL-12 at 10 ng/ml (Genzyme, Cambridge, MA), anti-CD3 mAb at 10 μg/ml (145-2C11) + anti-CD28 mAb together or each individually for 4 or 24 h at 37°C in 5% CO₂.

Cells were stimulated as above, harvested, and surface stained with fluorochrome-conjugated Abs (BD Pharmingen, San Diego, CA) against CD3, CD8α, NK1.1, and CD4, before intracellular staining for IFN-γ, as previously described (15). To detect IFN-γ production in combination with proliferative capacity, RAG^−/− splenocytes were labeled with CFSE, then stimulated with recombinant human IL-2 or IL-12 alone or in combination in the presence or absence of the cell cycle inhibitor, l-mimosine at 100 μM (Sigma-Aldrich, St. Louis, MO). Four hours before harvesting and staining, cultures were treated with 50 ng/ml PMA + 500 ng/ml ionomycin in the presence of brefeldin A (10 μg/ml). Cell collection was performed using a FACSCalibur flow cytometer (BD Biosciences, Mountain View, CA) and data analyzed using FlowJo software (Tree Star, San Carlos, CA). Supernatants were collected after 24 h of stimulation and ELISAs were performed as previously described (11).

Spleens were harvested from C57BL6 RAG^−/− mice and surface stained for PE-conjugated anti-DX5 and sorted on a MoFlo cytometer (DakoCytomation, Fort Collins, CO) to purify resting NK cells. Cell purity was 98–99% for each sort. DNase I digestions (0, 0.0008, 0.004, 0.02, and 0.1, 0.2 μg/ml) were performed on isolated nuclei as described previously (16). Analysis of the IFN-γ locus was performed as described (17). Briefly, genomic DNA was digested overnight with BamHI and resolved through 0.8% agarose before transfer to nylon membranes and blots were probed with a 410-bp fragment of IFN-γ exon 4.

Resting NK (purity >98%) and CD4⁺ T (purity, 88.5%) cells were purified by cell sorting as before. Genomic DNA was isolated, sheared, and denatured by incubating for 20 min at 75°C in 0.3 N NaOH. DNA was then incubated in a 4.8 M bisulfite solution (sodium metabisulfite, 5 N NaOH; Sigma-Aldrich) + 10 mM hydroquinone (Sigma-Aldrich) for 5 h at 55°C. DNA was then isolated with a GeneClean II kit (Bio 101, Vista, CA), desulfonated with 5 N NaOH and precipitated using NH₄Ac and ethanol. PCR generated ∼500-bp fragments of the IFN-γ intron I region, which spans HS site I and contains CpG sites at +40, +62, +69, +80, +120, +168, +195, +338, +383, and +462 relative to the first base of intron I. Sense 5′-GGTATAGTTATTGAAAGTTTAGAAAGTTTG-3′ and antisense 5′-CAAAATTACTCCTCAAAATAAAACAACTTC-3′. PCR products were cloned into pGEM T easy plasmid (Germantown, Promega, Madison, WI) and 7–14 individual colonies were isolated per experiment. Plasmid DNA was isolated using the miniprep kit from Qiagen Sciences (Germantown, MD) and sequenced using Applied Biosystems (Foster City, CA).

To directly compare the ability of NK and T cells to produce IFN-γ, splenocytes were stimulated with anti-CD3 + anti-CD28 + IL-2 + IL-12 and intracellular staining was used to assess the presence of cytokine protein in both cell populations. Analysis of the CD4⁺ T cell population revealed that IFN-γ staining was not detected after 4 h of stimulation, but by 24 h 17% of cells were positive for IFN-γ (Fig. 1). In these studies, whole T cell populations were used and the presence of memory cells likely contributes to the small percentage of IFN-γ⁺ cells at 24 h poststimulation. However, exposure of splenocytes to the combination of IL-2 + IL-12 in the absence of TCR stimulation failed to stimulate T cell production of IFN-γ (data not shown). In contrast, analysis of the response of NK cells in the same cultures revealed that 83% of these cells were IFN-γ⁺ after only 4 h of stimulation and these numbers were maintained for 24 h after stimulation (Fig. 1). Although the overall percentage of NK cells expressing IFN-γ has not significantly changed after 24 h of stimulation, the mean fluorescence was increased, which suggests that other regulatory elements (i.e., transcription factors) may be involved with the regulation of this cytokine or more simply that there is an accumulation of cytokine within the cell over time. This NK response was not observed in unstimulated cultures and was not dependent on the presence of either anti-CD3 or anti-CD28. Moreover, a similar response could be observed with IL-2-activated lymphokine-activated killer cells (CD28⁺, CD25⁺) or using enriched splenic NK cells (data not shown). Although stimulation of NK cells with a single cytokine did not result in detectable levels of intracellular IFN-γ at any time point examined, stimulation with other cytokine combinations (IL-12 + IL-18, IL-2 + IL-18) produced cytokine expression profiles similar to those shown in Fig. 1 (data not shown). Thus, the ability of NK cells to produce IFN-γ rapidly is not limited to one stimulatory pathway. In addition, these studies illustrate the different kinetics and requirements for cytokines for the ability of NK and T cells to produce IFN-γ and are consistent with the respective role of these lymphocytes in innate and adaptive responses.

Previous studies have demonstrated that the ability of naive T cells to produce IFN-γ is dependent on their ability to progress through the cell cycle (4, 18). Since the cytokines IL-2 and IL-12 are potent inducers of NK cell proliferation, studies were performed to assess whether the ability of resting splenic NK cells to produce IFN-γ rapidly was dependent on the ability of these cells to proliferate. Therefore, splenic NK cells were labeled with CFSE, stimulated in vitro with cytokines for up to 4 days, and the kinetics of proliferation and IFN-γ production was assayed. This analysis revealed that after 24 h of stimulation these NK cells remained CFSE^high, but contained a significant population of IFN-γ producers (Fig. 2,A). Despite increasing numbers of IFN-γ⁺ cells over the 4 days of stimulation, NK cells did not begin to proliferate until days 3 and 4, with almost all of the cells dividing at least once by day 4 (Fig. 2,A and data not shown). Although these findings suggest that the ability of NK cells to produce IFN-γ is independent of proliferation, it was possible that entry into the cell cycle was a prerequisite for remodeling of the IFN-γ locus and optimal cytokine production. However, when naive NK cells were stimulated with cytokine in the presence of l-mimosine, a G₁ cell cycle inhibitor, IFN-γ production was not affected (Fig. 2 B). Furthermore, cell cycle analysis was performed to assess DNA content of NK cells over the 4-day stimulation period. Cell cycle profiles of stimulated NK cells revealed that the majority of cells remain in G₀-G₁ for the first 2 days after stimulation and only progressed through the cell cycle after 3 days (data not shown). These data correlate with a lack of significant proliferation in NK cell cultures 3 days after stimulation, in which only one or two additional peaks of CFSE were observed (data not shown).

Since the ability of naive CD4⁺ T cells to produce IFN-γ is dependent on progression through the cell cycle and associated with derepression of the IFN-γ locus, the finding that a significant percentage of NK cells can do so within 4 h of stimulation independently of division and cell cycle progression suggested that the IFN-γ locus in NK cells exists in a constitutively open configuration. Such an “open” configuration is characterized by areas of demethylated chromatin that can be cleaved with the enzyme DNase I to yield a laddering pattern of DNA. In contrast, if chromatin is heavily methylated, then these sites are typically resistant to DNase I. Previous studies have identified intron I of the IFN-γ locus as a site regulated by epigenetic changes that is resistant to DNase I in naive T cells, but sensitive to this enzyme in differentiated Th1 cells (17, 19). In these studies, the acquisition of DNase sensitivity was observed in Th cells cultured under Th1-polarizing conditions for 4 days (Fig. 3,A). Similarly, analysis of FACS-sorted splenic NK cells (>99% purity) revealed the presence of DNase hypersensitivity sites within the IFN-γ locus (Fig. 3 B). Moreover, when DNase hypersensitivity was used to assess the accessibility of the IL-4 locus in resting NK cells, only a nonspecific hypersensitivity site was detected (data not shown), consistent with the inability of NK cells to make IL-4 (20). These data indicate that the sensitivity of the IFN-γ locus in NK cells to DNase I is not a characteristic of cytokine genes in NK cells and correlates with the functional capacity of these cells to make IFN-γ. Thus, these studies reveal that NK cells have chromatin changes that are consistent with previously published reports which show that T cells that have acquired transcriptional competence for IFN-γ also have this DNase hypersensitivity site.

To further assess the status of the IFN-γ locus in NK cells, a bisulfite conversion assay was used to compare the methylation status of a series of CpG motifs within the intronic region of the IFN-γ locus in naive and Th1 T cells as well as NK cells. In naive T cells, almost all of these motifs were methylated, but after polarization under Th1 conditions for 4 days these levels were decreased (Fig. 3, C and D). These findings are consistent with previous studies which associate these changes with the acquisition of defined epigenetic characteristics that contribute to the maintenance of heritable cytokine expression patterns (2). Analysis of purified splenic NK cells displayed levels of methylation that were comparable to the 4-day polarized Th1 cells and indicate that this region of the IFN-γ locus in naive NK cells was demethylated at many of the CpG sites (Fig. 3, C and D). Furthermore, stimulation of NK cells in the presence of methyltransferase or histone deacetylase inhibitors (5-azacytidine and trichostatin A) did not enhance the production of IFN-γ (data not shown). Together, these data suggest that in resting NK cells the IFN-γ locus is not in a repressed state and that the IFN-γ locus exists in a constitutively modified form that is accessible to transcription factors which up-regulate the production of IFN-γ.

Although innate and adaptive immune cells are thought to be fundamentally disparate, many of the same environmental stimuli that lead to activation of NK cells also regulate T cell functions. Thus, many of the same cytokines and transcription factors that regulate NK cell production of IFN-γ are also involved in the development of Th1-type responses. However, previous studies have shown that the proliferation of naive T cells is required in order for them to overcome a scaffold of epigenetic controls that modulate the ability to produce IFN-γ (4, 5, 6, 17, 21). The results presented here demonstrate that the ability of NK cells to produce IFN-γ is not subject to these same controls. Rather, NK cells can produce IFN-γ in a rapid fashion independently of the need to undergo proliferation-dependent chromatin remodeling associated with the hyperacetylation of histones and demethylation of DNA. In agreement with the data presented here are recent studies that used chromatin immunoprecipitation assays to demonstrate that the IFN-γ locus of NK cells is acetylated (22). Thus, it appears, that in contrast to T cells, the IFN-γ locus in NK cells is already accessible to transcription factors, which in turn allows immediate production of IFN-γ in response to cytokines. Of relevance to these findings are studies which demonstrated that the chromatin of the IL-12 p40 promoter in macrophages is constitutively remodeled (23). Thus, it appears that the innate response to many pathogens that leads to the IL-12-mediated production of IFN-γ by NK cells (9, 24) is a function of an intrinsic genetic pattern in these cells of the innate immune system.

The molecular basis for the ability of NK cells to produce high levels of IFN-γ remains poorly understood. These swift responses may be due to the abilities of cytokines to activate the STAT and NF-κB signaling pathways that directly promote transcription of the IFN-γ gene (11, 15). However, recent studies using transgenic mice in which a bicistronic reporter was knocked into the IFN-γ locus indicate that unstimulated mature NK cells contain a pool of pre-existing IFN-γ mRNA (22). Although these low levels of IFN-γ mRNA may be a consequence of promiscuous basal transcription associated with the open configuration of this locus, it is possible that the stabilization of this pool of RNA may contribute to the rapid production of IFN-γ protein. These two mechanisms are not mutually exclusive and both the stabilization and translation of existing stores of mRNA into protein, followed by the de novo transcription of cytokine, may contribute to the ability of NK cells to produce IFN-γ. Indeed, previous studies have shown that, depending on the stimulus, RNA stability and increased transcription play a prominent role in NK cell production of IFN-γ (25). Nevertheless, regardless of the mechanism, the data presented here provide a genetic basis for the functional difference between NK and T cell production of IFN-γ at the level of DNA structure.

The identification of differences in the chromatin structure of the IFN-γ locus between NK and T cells raises fundamental questions about the environmental cues that regulate these events during NK cell development. Recent studies have shown that ∼50% of NK cell precursors (NK1.1^low, CD122⁺) in the bone marrow appear to express transcripts for this cytokine (22), although other studies which examined the role of GATA3 in NK cell development suggest that immature NK cells do not have the capacity to produce IFN-γ (26). Whether the same factors that are involved in remodeling of the IFN-γ locus in T cells are involved in this developmentally regulated process in NK cells remains unclear. In addition, just as there is in an effector T cell population, there is undoubtedly a phenotypic mixture of NK cells that exist in the periphery that will warrant further analysis. Nevertheless, the identification of developmental differences in the structure of the IFN-γ locus between NK and T cells raises fundamental questions about other genes involved in both innate and adaptive immunity. For example, it is unclear whether the innate ability of basophils, mast cells, and eosinophils, to produce the cytokine IL-4, is subject to the same epigenetic restrictions that govern the production of IL-4 in T cells (27, 28). Additional studies are needed to determine whether differences in chromatin structure provide a general mechanism that governs the basis for the differences between innate and adaptive production of cytokines.

We thank the members of the Hunter Laboratory and Pathobiology Department for their support and helpful discussion.