A Distal Region in the Interferon- (cid:1) Gene Is a Site of Epigenetic Remodeling and Transcriptional Regulation by Interleukin-2*

Interferon- (cid:1) (IFN- (cid:1) ) is a multifunctional cytokine that defines the development of Th1 cells and is critical for host defense against intracellular pathogens. IL-2 is another key immunoregulatory cytokine that is involved in T helper differentiation and is known to induce IFN- (cid:1) expression in natural killer (NK) and T cells. Despite concerted efforts to identify the one or more transcriptional control mechanisms by which IL-2 induces IFN- (cid:1) mRNA expression, no such genomic regulatory regions have been described. We have


Interferon-␥ (IFN-␥) is a multifunctional cytokine that defines the development of Th1 cells and is critical for host defense against intracellular pathogens. IL-2 is another key immunoregulatory cytokine that is involved in T helper differentiation and is known to induce IFN-␥ expression in natural killer (NK) and T cells.
Despite concerted efforts to identify the one or more transcriptional control mechanisms by which IL-2 induces IFN-␥ mRNA expression, no such genomic regulatory regions have been described. We have identified a DNase I hypersensitivity site ϳ3.5-4.0 kb upstream of the transcriptional start site. Using chromatin immunoprecipitation assays we found constitutive histone H3 acetylation in this distal region in primary human NK cells, which is enhanced by IL-2 treatment. This distal region is also preferentially acetylated on histones H3 and H4 in primary Th1 cells as compared with Th2 cells. Within this distal region we found a Stat5-like motif, and in vitro DNA binding assays as well as in vivo chromosomal immunoprecipitation assays showed IL-2-induced binding of both Stat5a and Stat5b to this distal element in the IFNG gene. We examined the function of this Stat5-binding motif by transfecting human peripheral blood mononuclear cells with ؊3.6 kb of IFNG-luciferase constructs and found that phorbol 12-myristate 13-acetate/ionomycin-induced transcription was augmented by IL-2 treatment. The effect of IL-2 was lost when the Stat5 motif was disrupted. These data led us to conclude that this distal region serves as both a target of chromatin remodeling in the IFNG locus as well as an IL-2induced transcriptional enhancer that binds Stat5 proteins.
IFN-␥ 1 is a cytokine expressed primarily by, but not limited to, lymphocytes of T and NK cell lineages. NK cells are a major early source of IFN-␥ in vivo, and T cell expression of IFN-␥ is generally restricted as T helper 1 (Th1) cells (1). These cell types and their predominate cytokine IFN-␥ are intimately involved in protection against intracellular pathogens and the development of tumors (2). This is evident in humans with gene mutations that interfere with IFN-␥ signaling, including the IFN-␥ receptor (3) and Stat1 (4) as well as in IFN-␥-deficient mice (5), which are highly susceptible to infectious agents and prone to certain cancers. In contrast, circumstances in which IFN-␥ is overproduced or its effects fail to induce negative feedback loops lead to widespread pathology (6). Not surprisingly, IFN-␥ expression patterns are tightly regulated in a tissue-specific manner by precise induction signals.
Much effort has focused on characterizing the proximal IFNG promoter and intervening sequences in an effort to identify potential regulatory regions that are responsible for signaland cell-specific control of gene expression. This approach has been useful in defining putative regulatory regions near the core IFNG promoter. It is notable, however, that regions proximal to the IFNG gene fail to explain tissue-specific patterns of IFN-␥ expression as evidenced by transgenic mice expressing only the proximal IFNG promoter linked to a reporter gene (7,8). In addition, transgenic mice that contain the full-length human IFNG gene and ϳ2.7 kb of upstream sequence have similar signal-specific regulatory patterns to the endogenous gene, yet the responses are relatively weak (8). Therefore, it is likely that regions more distal from the core IFNG promoter are involved in determining tissue-specific and optimal signalspecific expression of IFN-␥ (9).
In naïve T cells, induction signals for activation and IFN-␥ expression are provided optimally through T cell receptor (TCR) and co-stimulatory receptor (CD28) engagement (10). Following primary activation, differentiation of effector CD4 ϩ T helper (Th) cell populations occurs over several days and is directed in large part by cytokines such as IL-2, IL-4, and IL-12 (11). Not only do IL-2 and IL-12 induce Th1 cell development, they also directly induce expression of IFN-␥. This is most apparent in NK cells, because no preceding activation/differentiation signals are required for cytokine responsiveness and transactivation of the IFNG gene (12). Therefore, NK cell models offer an advantage in studying regulation of IFN-␥ expression by isolating the effects of autonomous cytokine receptor signaling events.
IL-12 acts via the transcription factor signal transducer and activator of transcription 4 (Stat4) that is thought to mediate trans-activation of the IFNG gene by binding to sequence motifs therein or enhancing the binding of AP-1 to the core IFNG promoter (13). The requirement of Stat4 in regulating IFN-␥ is underscored in Stat4 Ϫ/Ϫ T cells and NK cells that are refractory to IL-12-induced IFN-␥ expression. In both mice (14) and humans (15), 2 Stat4 has been shown to interact with the IFNG gene in the proximal promoter and/or first intron (16,17). These elements are based on non-consensus STAT motifs, and their functional role is still not firmly established. Additionally, the human intronic STAT binding region is not present in murine genomic DNA, suggesting its importance in regulating IFNG gene expression may be a recent evolutionary event. Despite concentrated efforts and that fact that IFN-␥ is highly regulated by cytokines, the identification of functional STAT binding elements in the IFNG gene has proved to be challenging.
In the case of IL-2, the exact mechanisms by which this cytokine regulates IFN-␥ have been surprisingly neglected, despite reports dating to 1983 indicating that IL-2 induces IFN-␥ expression (18 -20). IL-2 is an important factor in the differentiation of both Th1 and Th2 cells (11). Accordingly, this cytokine is routinely added to in vitro culture systems to enhance Th cell development and is thought to act via Stat5 (21). In IFN-␥-secreting cells, induction of IFN-␥ mRNA by IL-2 is rapid and occurs at the level of transcription, as demonstrated by nuclear run-on analysis. The induction occurs in the presence of cycloheximide, indicating that protein synthesis is not necessary for the induction of transcription (22). Although IL-2 has long been associated with IFN-␥ expression, the molecular mechanisms involved are still unknown.
In this report, we identify and characterize a distal region of the IFNG gene that is a target for cytokine-driven epigenetic modifications, which correlate with competency of tissue-specific IFN-␥ expression profiles. Contained within this same region is a distal cis-acting element that binds Stat5 proteins and enhances transcription of the IFNG promoter in an IL-2dependent fashion. These data indicate that this distal 5Јflanking region of the IFNG gene is a critical part of the functional IFNG gene and provides the first indication of the molecular mechanisms involved in IL-2 regulation of IFN-␥ expression.

EXPERIMENTAL PROCEDURES
Antibodies and Cytokines-Antibodies directed against Stat5a, Stat5b, and Stat3 were purchased from R&D Systems and were used in supershift analyses as well as chromatin immunoprecipitation assays. Recombinant human IL-2 was obtained from Hoffmann-La Roche. Recombinant mouse IL-12 was generously provided by Genetics Institute (Cambridge, MA). Recombinant human IL-4 was purchased from R&D Systems. Anti-CD3⑀, anti-CD28, anti-IL-4, anti-IL-12, and anti-IFN-␥ antibodies were purchased from BD Pharmingen. Anti-acetylated H3, anti-acetylated H4, and normal rabbit IgG were from Upstate Biotechnology (Lake Placid, NY).
Human PBMC were isolated from healthy volunteers over Ficoll-Hypaque, and for T cell differentiation studies, naïve CD4 ϩ T cells were purified with the human CD4 ϩ /45RO Ϫ naïve T cell subset column kit or human CD4 ϩ T cell subset column kit (both from R&D Systems, Minneapolis, MN) followed by positive selection using CD45RA beads (Mylteni Biotec Inc., Auburn, CA). Purity was typically Ͼ90%. Cells were stimulated with plate-bound anti-CD3 (10 g/ml) plus anti-CD28 (5 g/ml) Abs and IL-2 (50 units/ml) for 3 days, and expanded in a new flask for a further 3 days with exogenous IL-2. To generate Th1 cells, IL-12 (10 ng/ml) was added throughout the culture period and anti-IL-4 Ab (10 g/ml) was added for the first 3 days. For Th2 cells, IL-4 (25 ng/ml) was added throughout the culture period, and anti-IL-12 Ab (5 g/ml) was added for the first 3 days.
Promoter Cloning and Reporter Constructs-Human genomic DNA (200 ng) was used as a template for all PCR reactions in 10-to 50-l volumes. The full Ϫ3.662 kb to ϩ38 bp IFNG 5Ј-flanking region was amplified in 10-l PCR reactions containing 10 mM Tris-HCl (pH 8. For amplification of the Ϫ3.6-kb fragment, the conditions were as follows: 95°C for 5 min, then 30 cycles of denaturation at 95°C for 30 s, annealing at 60°C for 60 s, and extension at 68°C for 5 min. These PCR reactions were performed using a Hybaid Touchdown thermal cycler (Hybaid Limited, UK).
The Ϫ3.6-kb IFNG upstream region was amplified from genomic DNA with MluI linkers and ligated into a TA vector (Invitrogen). Following MluI and PvuI double digestion, the Ϫ3.6-kb insert was extracted from the 0.8% agarose gel, purified, and ligated into the PGL3 Basic Luciferase vector (Promega, Madison, WI). Restriction digests revealed clones with a 5Ј to 3Ј orientation, and sequencing confirmed the STAT element. Sitedirected mutagenesis was performed on the Ϫ3.6-kb IFNG-Luc construct to disrupt the Stat5 motif from TTCAAGGAA to TTCAACCTT. DNA sequencing confirmed the mutagenesis. The mutations introduced in the Stat5 element are the same as those used in EMSA experiments, which failed to bind IL-2-induced Stat5 proteins and do not complete for Stat5 binding by excess cold competition.
Electrophoretic Mobility Shift Assay-Complementary singlestranded oligonucleotides were commercially synthesized (Invitrogen) to span ϳ10 bp on either side of the Stat element (in bold). Another oligonucleotide with a 4-bp mutation in the Stat site (underlined) was made, and the sequences are as follows: 5Ј-AGT TAT TAG AAA TTT CAA GGA AGT GAC AAC AGA G-3Ј (Stat WT) and 5Ј-AGT TAT TAG AAA TTT CAA CCT TGT GAC AAC AGA G-3Ј (Stat Mutant).
Complementary strands were annealed by combining 2 g of each oligonucleotide and 6 l of 10ϫ annealing buffer (500 mM Tris, 100 mM MgCl 2 , and 50 mM dithiothreitol) in a 60-l reaction, denatured in a boiling water bath for 5 min, and then allowed to cool to room temperature. Double-stranded DNA overhangs were labeled with Klenow enzyme by a fill-in reaction, [␣-32 P]dCTP and free dNTPs (Amersham Biosciences). The DNA-protein binding reaction was conducted in a 20-l reaction mixture consisting of 5-10 g of nuclear protein extract, 1 g of poly(dI-dC) (Sigma), 4 l of 5ϫ binding buffer (60 mM Hepes, 7.5 mM MgCl 2 , 300 mM KCl, 1 mM EDTA, 2.5 mM dithiothreitol, 50% glycerol, and 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride) and 1.5 ϫ 10 4 cpm of 32 P-labeled oligonucleotide probe. For supershift analysis to identify Stat proteins interacting with the IFNG promoter, 1 l of antibody was added to the reaction prior to the addition of the probes and incubated on ice for 20 min. In cold oligonucleotide competition experiments, 10-to 100-fold excess of unlabeled oligonucleotide probe was added 10 min prior to adding the radiolabeled oligonucleotide probe.
Cytoplasmic and Nuclear Extraction-Nuclear extracts were prepared from PHA-stimulated human peripheral blood T cells or NK cells, as described previously (23). Briefly, following no stimulation (NS) or treatment with cytokines for 30 -60 min, cell pellets were resuspended in lysis buffer (50 mM KCl, 25 mM Hepes, pH 7.8, 0.5% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 20 g/ml aprotinin, 100 M dithiothreitol) and subsequently incubated on ice for 5 min. Cellular suspensions were collected by centrifugation at 2,000 rpm, and the supernatant was harvested as the cytoplasmic protein fraction. Nuclei were washed in wash buffer (lysis buffer without Nonidet P-40) and harvested by centrifugation at 2,000 rpm. Nuclear pellets were resuspended in extraction buffer (500 mM KCl, 25 mM Hepes, pH 7.8, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 20 g/ml aprotinin, and 100 M dithiothreitol), frozen in dry ice, thawed slowly on ice, and finally centrifuged at 14,000 rpm for 10 min. The supernatant was harvested, and nuclear proteins were quantified with the bicinchoninic acid protein assay reagent (Pierce).
Transfection Assays-PBMCs were transfected following overnight culture in RPMI 1640 medium containing 10% FCS with or without 10 -100 units/ml rhIL-2 (Chiron Therapeutics, Emeryville, CA). Cells were then washed and resuspended in 250 l of fresh medium at 2 ϫ 10 7 cells/ml and electroporated in the presence of 50 g of reporter construct (600 V, for 9 pulses of 500 s, with 100 s between pulses) using 4-mm (gap width) cuvettes in a BTX Electro Square Porator ECM 830 (Genetronics, Inc., San Diego, CA). A control plasmid containing the ␤-actin promoter driving a Renilla luciferase (provided by Dr. Christopher Wilson, University of Washington) was co-transfected as an internal standard, and values were normalized to correct for transfection efficiency (generally 30 -50%). After electroporation the cells were diluted in fresh medium, allowed to rest for 1 h prior to plating, and then stimulated with 20 ng/ml PMA plus 400 nM calcium ionophore for 4 h. Luminescence was measured using a Promega luciferase assay kit and counted on a sixdetector PerkinElmer Life Sciences 1450 Microbeta liquid scintillation counter with coincidence counting deactivated.
Nuclei Isolation and DNase I Hypersensitivity Analysis-NK92 cells were rested in RPMI medium plus 10% FCS without IL-2 and IL-15 for 12 h. Approximately 100 ϫ 10 6 cells were used for each stimulation. The cells were stimulated with IL-2 (100 units/ml) and IL-12 (10 units/ml) alone and in combination for 1 h. For nuclei isolation NK92 cells were spun at 600 ϫ g, washed once in 1ϫ phosphate-buffered saline, and spun again. The cell pellet was resuspended in 3 ml of cold lysis buffer consisting of 0.35 M sucrose, 60 mM KCl, 15 mM NaCl, 2 mM EDTA, 0.5 mM EGTA, 15 mM Hepes, pH 7.4, 0.6% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 0.15 mM spermine, 0.5 mM spermidine and incubated on ice for 5 min. The lysate was spun at 800 ϫ g for 5 min at 4°C. The nuclear pellet was washed twice in 2 ml of nuclei wash buffer (lysis buffer without Nonidet P-40) and spun at 800 ϫ g for 5 min at 4°C. The nuclei were resuspended in 500 l of nuclei storage buffer consisting of 60 mM KCl, 15 mM NaCl, 0.1 mM EDTA, 0.1 mM EGTA, 75 mM Hepes, pH 7.5, glycerol (40% by vol.), 0.1 mM phenylmethylsulfonyl fluoride, 0.15 mM spermidine, 0.5 mM dithiothreitol, and stored at Ϫ70°C until needed.
Nuclei were spun at 800 ϫ g and resuspended in a nuclease digest buffer consisting of 10 mM Tris-HCl, pH 7.5, 10 mM NaCl, 5 mM MgCl 2 , and 0.1 mM CaCl 2 . The nuclei were digested with increasing concentrations of DNase I (Roche Applied Science) that ranged from 0 to 0.4 unit per reaction for 10 min at 37°C. The DNase digestion was stopped by the addition of an equal volume of stop solution consisting of 20 mM Tris-HCl, pH 7.5, 10 mM EDTA, 0.6 M NaCl, 1% SDS, and 400 g/ml proteinase K, and the digests were incubated at 55°C for overnight. The genomic DNA was purified from each reaction by phenol/chloroform and subsequent ethanol precipitation. The precipitated DNA was resuspended in 10 mM Tris, pH 7.4, and quantitated by optical density. Approximately 20 g of DNase-treated DNA was digested with HindIII and separated by size on a 0.8% agarose gel using 1 ϫ TBE (Tris/borate/ EDTA) buffer. The DNA was transferred to a nylon membrane, and Southern analysis was performed using a 32 P-labeled human IFN-␥specific XcmI-HindIII DNA fragment as the probe. The DNase I hypersensitivity sites were visualized by exposing the blot to x-ray film for 24 -36 h.
Chromatin Immunoprecipitation Assays-ChIP assays were performed according to the manufacturer's instructions (Upstate Biotechnology, Lake Placid, NY). Briefly, 2 ϫ 10 6 cells were fixed with 1% formaldehyde, washed with cold phosphate-buffered saline, and lysed in buffer containing 10 g/ml aprotinin (ICN Biomedicals, Aurora, OH), 10 g/ml leupeptin (Bachem, Torrance, CA), and 2.5 M 4-nitrophenyl 4-guanidinobenzoate hydrochloride (Sigma). Nuclei were sonicated for a total of 60 s to shear DNA (Heat Systems, Farmingdale, NY), the lysates were pelleted, and supernatants were diluted. Diluted lysates were pre-cleared with salmon sperm DNA/protein A-agarose for 30 min prior to immunoprecipitation with specified antibodies. A proportion (2%) of the diluted supernatant was kept as "input" to quantify genomic DNA. After immunoprecipitation, the protein-DNA complexes were incubated with Protein A beads, and the protein-DNA complexes were eluted in 1% SDS/0.1 M NaHCO 3 , and cross-links were reversed at 65°C. DNA was recovered by phenol-chloroform extraction and ethanol precipitation then subjected to PCR and/or real-time PCR analysis. PCR was carried out with AmpliTaq Gold (Applied Biosystems) for 35 cycles (40 s at 95°C, 40 s at 59°C, and 40 s at 72°C), and the products were visualized by ethidium bromide staining. The immunoprecipitated DNA samples were normalized to "input" DNA samples. PCR primers for the distal IFNG promoter were as follows: 5Ј-GTTTCACTGC-CAGTCTGG-3Ј and 5Ј-ACTACACCCTATCAGGCTTTA-3Ј; primers for the proximal IFNG promoter region were 5Ј-GTTTCACTGC-CAGTCTGG-3Ј and 5Ј-ACTACACCCTATCAGGCTTTA-3Ј.
Ribonuclease Protection Assay-RNase protection assays (RPA) were performed according to the manufacturer's specifications (Ambion, Austin, TX) with ϳ5 g of RNA per reaction. RNA hybridization products were separated by size on a 6% denaturing polyacrylamide gel.

IL-2 Induces the Expression of IFN-␥ mRNA-
The process of Th cell differentiation is complicated and involves diverse extracellular signals through multiple classes of receptors, including the T cell receptor (TCR), co-stimulatory molecules, and cytokine receptors (1,2,11). This process typically: 1) occurs over an extended period of time (5-10 days), 2) seems to occur with cell division (24), and 3) includes IL-2 in the cultures, either endogenously produced by T cells and/or by exogenous addition. We wanted to isolate the effects of IL-2 on IFNG gene regulation per se from the process of Th cell differentiation. Thus, we have primarily used fresh NK cells and an NK cell line (NK92) as a model to study cytokine-induced epigenetic and transcriptional modifications to the IFNG gene.
We first confirmed that IL-2 treatment of NK92 cells leads to IFN-␥ mRNA expression (Fig. 1A). Strong IFN-␥ mRNA expression in NK92 was evident at 3 h and was slightly diminished after 6 h of IL-2 treatment. Our group and others have also reported that co-treatment of activated T cells or NK cells with IL-2 and IL-12 leads to a highly synergistic effect on IFN-␥ expression (25,26). A time course experiment was performed on NK92 cells under IL-2, IL-12, and IL-2 plus IL-12 conditions to confirm these data (Fig. 1B). The results confirm that IL-2 and IL-12 induce IFN-␥ mRNA expression rapidly and when added together, synergistically induce IFN-␥ transcripts. These data are similar to what was observed in activated T cells (data not shown). Fig. 1C shows by real-time PCR the response to individual cytokine treatment and the synergistic response to IL-2 and IL-12 stimulation in NK92 cells. Thus, we determined that the NK92 cell line was an appropriate model to study IL-2-regulation of IFN-␥ expression.
Identification of a DNase I Hypersensitivity Site ϳ3.5-4 kb Upstream of the IFNG Gene-DNase I HS mapping strategies have been used to successfully identify genomic regions that are functionally linked to gene expression. This approach allows for the investigation of the chromatin structure of the native gene in intact nuclei. It has been shown previously that DNase I HSs in the IFNG gene are found only in cells capable of transcribing the gene (27). Furthermore, the appearance or predominance of these sites in T cells is stimulation-sensitive, indicating a dynamic relationship between cell-specific competency for expression and transcription that is activation-dependent (28). Because the proximal 5Ј-flanking region of the IFNG gene fails to account for normal IFN-␥ regulation (8), we sought to determine if more distal regions of the IFNG gene become remodeled upon cytokine stimulation.
To identify potential regulatory regions in the IFNG gene we performed DNase I HS analysis using NK92 cells (Fig. 2). The results show the appearance of two bands by Southern blot analysis. The band of ϳ12 kb represents the HindIII-HindIII restriction fragment, visible only after phenol-chloroform extraction and complete HindIII digestion. The 3Ј HindIII site in the first intron is located at position ϩ669 and the 5Ј HindIII site is at position Ϫ11,780 relative to the transcriptional start site according to our nomenclature (29). The lower band represents a single DNase I HS that was estimated to be ϳ4.5 kb upstream of the 3Ј HindIII restriction site in the first intron IL-2 Regulation of Chromatin and Transcription of IFN-␥ (Fig. 2, see arrows) and ϳ3.5-4 kb 5Ј to the transcriptional start site. The 4-kb HS was observed with the highest DNase I concentrations in unstimulated cells (lane 4), but when NK92 cells were treated for 6 h with IL-2 (lanes 6 -8), IL-12 (lanes 10 -12), or in combination (lanes 14 -16); however, the accessibility of the HS for cleavage was augmented. This suggests that IL-2 and/or IL-12 treatment may enhance the availability of this region to transcription factors in NK cells. We sought to confirm that this distal region was a site for epigenetic modification by using an alternative approach.

IL-2 Induces Histone Acetylation of the Distal IFNG Site in Primary NK Cells and Is Preferentially Acetylated in Th1
Cells-Alterations in histone acetylation are known to be associated with chromatin remodeling and the ability of a gene to be transcribed. To corroborate our identification of a distal cytokine-regulated site of chromatin remodeling, we assessed histone acetylation of the distal IFNG promoter in primary NK cells using chromatin immunoprecipitation analysis (Fig. 3A). We found that the distal IFNG gene was constitutively acetylated on histone H3 in non-stimulated (NS), freshly isolated NK cells at least to some extent (lane 3). However, when primary NK cells were stimulated with IL-2 alone (lane 6) or in combination with IL-12 (lane 12) we observed a consistent augmentation of histone acetylation. IL-12 treatment alone did not alter the background levels of H3 acetylation at the distal site (lane 9). We confirmed this observation in freshly isolated human NK cells by quantitative ChIP analysis using real-time PCR in which the distal IFN-␥ site was amplified (Fig. 3B). Indeed, IL-2 treatment led to quantitative enhancement of histone H3 acetylation of this upstream region. Similar results

FIG. 2. A cytokine-inducible DNase
I hypersensitivity site is present in the distal IFNG gene. NK92 cells were not stimulated (NS) or treated with IL-2, IL-12 or IL-2 plus IL-12, and the nuclei were isolated. The genomic DNA was subjected to DNase I digestion, phenol/chloroform extraction, and digestion with HindIII. The DNA was transferred to a nylon membrane, and Southern analysis was performed using a 32 P-labeled human IFN-␥-specific XcmI-HindIII DNA fragment as the probe. A DNase I HS site (large arrows) was identified ϳ3.5-4 kb 5Ј to the IFNG transcriptional start site that was responsive to IL-2 and IL-12 stimulation.
were obtained using the NK92 cell line (data not shown).
In T cells, the proximal IFNG promoter has been shown to be hyperacetylated in Th1 cells in comparison to Th2 cells (30). 2 This region contains a constitutive HS that is not predictive of the selective expression patterns of the native IFNG gene (8). Thus, we analyzed both the distal and proximal regions of the 5Ј-flanking region of the IFNG gene for histone modifications in fully differentiated Th1 and Th2 cells (Fig. 3C). At the distal site, we found that this region of the IFNG promoter was preferentially acetylated on histones H3 and H4 in Th1 versus Th2 cells. As a control we also analyzed the previously determined site in the proximal IFNG promoter that is a target for histone acetylation. Our results confirm that the proximal IFNG promoter is preferentially acetylated in Th1 cells, however, we found acetylation of the distal IFNG region to be more predictive of IFN-␥ secreting Th1 cells (30).
A Novel Stat5 Binding Motif in the Distal 5Ј-Flanking Region of the IFNG Gene Binds IL-2-induced Stat5 Proteins in Vitro and in Vivo-Stat5 is a key IL-2-activated transcription factor that is critically involved in the regulation of IL-2-inducible genes (31). IL-2 receptor ␣ (CD25) is one of the few IL-2inducible genes whose promoter has been extensively characterized. This gene contains functional IL-2 response elements and Stat5 binding motifs (31,32). Upon sequence comparison, however, no similar motifs were identified in the Ϫ2.7-kb IFNG promoter, and previous attempts to identify IL-2-responsive elements in the Ϫ2.7-kb IFNG promoter were unsuccessful. 3 Thus, we began to search for IL-2-responsive elements further upstream of the IFNG gene, particularly in regions that are targets for epigenetic modifications. A computer search identified several putative regulatory motifs distal to the IFNG promoter. Of particular interest was a potential STAT binding motif located at position Ϫ3.607 (Fig. 4A). The core sequence of the palindrome and the spacing of the internal nucleotides suggested it may be capable of binding Stat5 proteins (31). Interestingly, this motif was also identified at a similar position (Ϫ3.554 kb) upstream of the murine IFNG gene. This motif is within the previously identified HS region (Fig. 2) and is contained in the distal region that is acetylated in NK cells and Th1 cells (Fig. 3).
To assess if this motif was capable of binding Stat5 proteins, we made a double-stranded oligonucleotide spanning the site for use in EMSA assays. When activated human peripheral T FIG. 3. IL-2 induces histone modifications in the distal IFNG gene. A, freshly isolated, negatively selected, peripheral human NK cells were cultured in media alone (NS) or with the indicated cytokines for 18 h, after which histone H3 acetylation of the distal IFNG gene was assessed by ChIP. One of three separate experiments is shown. B, quantitative ChIP analysis of histone H3 acetylation in fresh human NK cells stimulated by the indicated cytokines via real-time PCR. C, naïve CD4 ϩ T cells were isolated and differentiated under Th1 or Th2 conditions, and ChIP analysis was performed on the proximal and distal IFNG gene. Histone H3 and H4 acetylation was assessed in Th1 and Th2 cells by PCR in the distal (top) and proximal (bottom) IFNG promoter regions. One of three independent experiments is shown. cells or NK92 cells (data not shown) were stimulated with IL-2, a complex was formed that was not present in non-stimulated cells (Fig. 4B, lane 2). Interestingly, when NK92 cells were treated with IL-15, an inducible complex migrated to the same location to that of the IL-2-induced band, and both Stat5a and Stat5b antibodies shifted the complex (data not shown). Of note, the IL-2-activated complex evident in lane 2 was not inducible by treating cells with IL-12 (lane 3). The specificity of the complex was documented by cold competition experiments where a 100-fold molar excess of the unlabeled wild type probe completely inhibited formation of the IL-2-induced complex (lane 4). Meanwhile, addition of a cold SP-1-specific probe failed to compete for binding of the IL-2-induced band (lane 6). Additional cold competition experiments were used with IL-2treated extracts to determine if the STAT sequence was involved in complex formation. An identical probe was generated with the exception that the four 3Ј bases in the STAT element were mutated. Addition of this mutant probe to gel shift reactions did not compete with the IL-2-induced complex indicating an intact STAT site was required for binding (lane 5). To identify the proteins in the IL-2-induced complex, antibodies raised against Stat5a (lane 7) or Stat5b (lane 8) were used in supershift assays. Both Stat5a-and Stat5b-specific antibodies were able to supershift the complex (lanes 7-9) arguing that IL-2 (as well as IL-15) treatment leads to Stat5a/Stat5b heterodimer formation on the distal IFNG Stat element in NK cells and activated T cells.
The preceding data indicated that Stat5 isoforms have the ability to bind this distal site in the IFNG gene. However, to establish that IL-2-activated Stat5a and Stat5b interact with the endogenous IFNG promoter in vivo, we again utilized ChIP analysis. We used NK92 cells to test for IL-2-induced Stat5 proteins interacting with the distal IFNG region in vivo. As shown in Fig. 4C, cells that were not stimulated (NS) (lanes 3 and 4) or treated with IL-12 alone (lanes 11 and 12) did not recruit Stat5a or Stat5b to the distal IFNG gene. When cells were stimulated in the presence of IL-2 alone (lanes 7 and 8) or IL-2 plus IL-12 (lanes 15 and 16), however, both Stat5a and Stat5b were immunoprecipitated, bound to the distal IFNG region. To control for nonspecific antibody interactions, normal rabbit IgG was used as an immunoprecipitation control but no evidence of nonspecific binding was observed (lanes 2, 6, 10,

FIG. 4. Identification of a Stat motif within the distal IFNG gene that binds IL-2-induced Stat5 proteins in vitro and in vivo.
A, sequence comparison of a potential STAT element upstream of the human IFNG gene, a consensus Stat5 element, and a motif upstream of the murine IFNG gene (arrow indicates the numbered base). B, EMSA analysis with nuclear extracts from primary human T cells. Stimulation of T cells with IL-2 results in the formation of a unique complex that is not induced with IL-12 treatment (arrow) and is not present in non-stimulated (NS) cells. Cold-competition experiments using excess unlabeled probe reveal the specificity of the complex and requirement for an intact Stat motif for binding (lanes 4 -6). Supershift analysis with antibodies against Stat5a or Stat5b displaces the IL-2-induced complex (lanes 7-9). C, ChIP analysis in NK92 cells of the distal IFNG gene using Stat5a or Stat5b antibodies to immunoprecipitate following cytokine stimulation. IL-2 treatment results in the binding of both Stat5a and Stat5b to the upstream IFNG gene (lanes 7-8 and 15-16). One of three representative experiments is shown. . Because IL-2 can activate both Stat3 and Stat5 (33), we also performed ChIP assays with antisera to Stat3. However, we did not find Stat3 interacting with the distal Stat5 region suggesting that this motif may preferentially bind Stat5 proteins (data not shown). Therefore, based on EMSA and ChIP data we conclude that, upon IL-2 stimulation, Stat5a and Stat5b bind to a Stat5 motif in the distal IFNG gene. We next questioned whether this Stat5 binding site has a functional role in IL-2-induced IFN-␥ transcription.

IL-2 Regulation of Chromatin and Transcription of
The Stat5 Element Confers IL-2 Inducibility of the IFNG Promoter in Reporter Gene Assays-To further define the functional role of IL-2 treatment in modulating the levels of expression of the IFN-␥, a Ϫ3.6-kb reporter construct was generated. The involvement of the Stat5-binding element was assessed using a corresponding construct in which the Stat5 motif is disrupted (Fig. 5A). This mutation was shown to no longer bind Stat5 proteins by EMSA (Fig. 4B, lane 5). The constructs were transfected into human PBMC, because neither primary NK cells nor the NK92 cell line are easily transfectable. Because unactivated T cells are refractory to cytokine stimulation and require activation signal(s) for optimal cytokine responsiveness (34), PBMC were stimulated with PMA/ionomycin (P/I), and then the ability of IL-2 to enhance transcription was assessed. As expected, expression of the Ϫ3.6-kb IFNG construct was barely detectable in unstimulated PBMC treated with or without IL-2 alone (Fig. 5B). However, following P/I treatment, transactivation of the Ϫ3.6-kb IFNG construct was readily detected. When P/I stimulation cells were co-cultured with exogenously added IL-2, transcription was enhanced by ϳ100% compared with P/I stimulation alone (despite that T cell activation results in some endogenous IL-2 production (11)). Furthermore, mutation of the Stat5 binding region within the Ϫ3.6-kb region abolished IL-2-mediated enhancer activity. These results suggest that IL-2-mediated activation of the Ϫ3.6-kb Stat5 binding region may be important in IFN-␥ regulation.

DISCUSSION
The biochemical and molecular events that control expression of IFN-␥ have been intensively studied, yet our understanding of the genomic regions that are required for tissuespecific expression and cytokine-induced transactivation of the IFNG gene is still very superficial. In this study we have identified a distal, 5Ј-flanking region of the IFNG gene that is a site for cell-specific epigenetic modifications. This region is ϳ3.5-4 kb upstream of the IFNG gene and contains a cytokineresponsive DNase I HS and is a target of IL-2-activated histone acetylation. This same region includes a Stat5 binding element that is critical to IL-2-induced transcription of the IFNG promoter. Our data show that, in human NK cells, the IFNG locus is constitutively in an "open" conformation yet is responsive to cytokine-induced augmentation of a distal HS site (both IL-2 and IL-12) and to acetylation of the distal promoter within the HS site (IL-2 only). Given that NK cells are capable of expressing IFN-␥ in a matter of minutes to hours following stimulation or local infection without prior activation (12,35), it is logical that the IFNG locus may be regulated differently in NK cells compared with T cells (36). It is interesting that the observed HS region is responsive to both IL-2 and IL-12 treatment, whereas histone acetylation around the Stat5 site responds only to IL-2 stimulation. Because the location of the DNase I HS site is an estimate, and the length of DNA it spans is undetermined, it is possible that other, more IL-12-specific sites may be nearby, possibly further upstream.
Recent studies in T cell subsets have described a CD4 ϩ T cell-specific role for the transcription factor T-bet (37) and a CD8 ϩ T cell-specific role for the related Eomesodermin (38). It is still not clear exactly how these factors exert their influence in the IFNG gene to affect T cell differentiation. In terms of general activation, however, multiple transcription factors appear to play a role in regulating T cell-specific IFN-␥ expression. Upon T cell receptor activation signals, AP-1, NF-AT, and CREB proteins bind to regions close to the transcription start site and act to initiate transcription (39 -43). The proximal promoter region, bound by these nuclear proteins, is also a site of DNA methylation, and expression of IFN-␥ correlates with the methylation status of this site (44 -46). Interestingly, this methylation site is evolutionarily conserved, because it is present in other vertebrates with the exception of avians (47). Regardless, the proximal promoter regions important for TCRinduced activation are not sufficient for normal tissue-specific expression or cytokine-induced IFNG gene induction (8,9).
In contrast to IFN-␥, regulation of the il4 locus is significantly better understood. Early studies identified sites proximal to or within the il4 gene that are involved in controlling/ enhancing transcriptional events (48 -51). During the process of differentiation, the il4 locus in naïve Th2 cells undergoes dramatic changes in the chromatin structure. No fewer than ten Th2-specific HS sites have been defined in this locus (52,53). Some of these sites provide a physical link between epigenetic modification and transcriptional regulation, including one HS located 3Ј of the il4 gene that corresponds to an enhancer element that binds the Th2-specfic transcription factor GATA3 as well as NF-AT (54). The use of functional genomics strategies and transgenic approaches has elucidated the contribution of multiple, distal regulatory regions in the il4 locus that encompasses intergenic space between several genes and over 100 kb in distance (53,(55)(56)(57)(58).
Like the il4 locus, the chromatin structure around the IFNG gene is regulated in a tissue-specific fashion (56,59). HS sites have been demonstrated in the proximal promoter and introns of the IFNG gene that correlate with a cell-specific ability to express IFN-␥. Nonetheless, Soutto et al. (8), using a transgenic approach, determined that greater than 2.7 kb of IFNG promoter was required to mimic endogenous IFN-␥ expression profiles (8). Thus, in vivo control of IFN-␥ transcription is likely to involve the cooperative effects of regions both proximal and distal to the IFNG gene (8). Accordingly, the 5Ј-flanking region of the IFNG gene has been suspected of harboring crucial regulatory elements (9), and a PHA/PMA-inducible HS site estimated to be ϳ3 kb upstream was found in Jurkat T cells in 1987 (28). A recent paper by Lee and colleagues (60) describes a distal Th1-specific HS site that was estimated to be ϳ5 kb upstream of the murine IFNG gene. This region has nearby binding sites for T-bet, AP-1, and NF-AT (60), transcription factors previously known to regulate IFN-␥ transcription. In addition, this group compared 50 kb of genomic sequence between mouse and human IFNG loci and found a distal region of conservation between species that maps to this HS site. In our study, we have estimated the HS site observed in NK cells to be ϳ3.5-4 kb 5Ј to the human IFNG gene. Within this region, we have characterized a Stat5 binding site identified in the IFNG gene that lies 3.6 kb upstream of the transcriptional start site. Interestingly, by sequence comparison, we also identified this Stat5 motif at position Ϫ3.554 kb 5Ј to the mouse IFNG gene (Fig. 4A). Taken together, this upstream region of the IFNG gene appears to have important functions both in controlling the chromatin structure of the IFNG locus as well as in enhancing transcription.
Originally described as T cell growth factor, IL-2 has been known to up-regulate IFN-␥ transcription in both T cells and NK cells. Although IL-2-induction of IFN-␥ was reported 20 years ago, little data exists regarding the mechanisms involved. IL-2-induced IFN-␥ production has been implicated in several in vivo infectious disease models, including lymphocytic choriomeningitis virus (61) and Toxoplasma gondii (62,63). In fact, by using IL-2 Ϫ/Ϫ mice, Villegas et al. (64), recently showed that maximal CD3-or PMA and ionomycin-induced IFN-␥ ex-pression is dependent on IL-2. We have conducted similar experiments using Stat5a Ϫ/Ϫ Stat5b Ϫ/Ϫ mice (65) and found defects in IL-2-regulation of IFN-␥ production (data not shown) in CD-3-stimulated splenocytes. It should be noted that these mice have pleiotropic defects in lymphoid development and function. Therefore these data need to be interpreted with some caution. Nevertheless, our PBMC transfection data (Fig. 5B) are consistent with these results and suggest that the distal Stat5 binding element is a functional site for IL-2 enhancement of IFN-␥ transcription.
It is important to appreciate, however, that the regulation of IFN-␥ transcription is complex, and no single element is altogether responsible for IFN-␥ activation. As stated earlier, IL-2 also up-regulates NF-B proteins that may have both positive and negative regulatory effects on IFN-␥ transcription. The mechanisms by which these different transcription factors converge to alter chromatin conformation may well be different depending upon the cell type (66) and extracellular signal. Furthermore, the tremendous synergy observed between IL-2 and IL-12 with respect to IFN-␥ expression is likely due, at least in part, to the cooperative activation of multiple transcriptional regulatory regions within the IFNG genomic DNA (such as Stat4 and Stat5 interactions). Although the synergy between IL-2 and IL-12 on IFN-␥ expression is dependent on Stat4 and independent of T-bet expression (25), the one or more precise mechanisms are still not clear. Further studies analyzing in vivo interactions of DNA binding proteins with specific regions of the IFNG gene as well as changes in chromatin conformation following stimulation will be required to more precisely define the epigenetic and transcriptional regulatory mechanisms of this gene. Nevertheless, this study supports the notion that the distal 5Ј-flanking region of the IFNG locus contains specific elements that are involved in tissue-specific and optimal signal-specific expression of IFN-␥.