Conserved Intergenic Elements and DNA Methylation Cooperate to Regulate Transcription at the il17 Locus*

Background: IL-17 is an inflammatory cytokine that mediates immunopathology in autoimmune disease. Results: DNA methylation at the il17 locus is lineage-restricted and blocks STAT3 binding. Conclusion: Expression of the il17 genes is regulated by promoter methylation and a novel intergenic enhancer. Significance: Understanding the mechanisms regulating IL-17 production will facilitate therapeutic control of inflammatory immune responses. Naive CD4+ T cells can differentiate into distinct lineages with unique immune functions. The cytokines TGFβ and IL-6 promote the development of Th17 cells that produce IL-17, an inflammatory cytokine not expressed by other T helper lineages. To further understand how IL-17 production is controlled, we studied an ∼120-kb genomic region containing the murine il17a and il17f genes and seven evolutionarily conserved, intergenic noncoding sequences. We show that the +28-kb noncoding sequence cooperates with STAT3, RORγt, and Runx1 to enhance transcription from both il17a and il17f promoters. This enhancer and both promoters exhibited Th17 lineage-specific DNA demethylation, accompanied by demethylation of lysine 27 of histone H3 (H3K27) and increased H3K4 methylation. Loss of DNA methylation tended to occur at STAT3 consensus elements, and we show that methylation of one of these elements in the il17a promoter directly inhibits STAT3 binding and transcriptional activity. These results demonstrate that TGFβ and IL-6 synergize to epigenetically poise the il17 loci for expression in Th17 cells, and suggest a general mechanism by which active STAT3 may be epigenetically excluded from STAT3-responsive genes in non-Th17 lineages.


Naive CD4 ؉ T cells can differentiate into distinct lineages with unique immune functions. The cytokines TGF␤ and IL-6 promote the development of Th17 cells that produce IL-17, an inflammatory cytokine not expressed by other T helper lineages.
To further understand how IL-17 production is controlled, we studied an ϳ120-kb genomic region containing the murine il17a and il17f genes and seven evolutionarily conserved, intergenic noncoding sequences. We show that the ؉28-kb noncoding sequence cooperates with STAT3, ROR␥t, and Runx1 to enhance transcription from both il17a and il17f promoters. This enhancer and both promoters exhibited Th17 lineage-specific DNA demethylation, accompanied by demethylation of lysine 27 of histone H3 (H3K27) and increased H3K4 methylation. Loss of DNA methylation tended to occur at STAT3 consensus elements, and we show that methylation of one of these elements in the il17a promoter directly inhibits STAT3 binding and transcriptional activity. These results demonstrate that TGF␤ and IL-6 synergize to epigenetically poise the il17 loci for expression in Th17 cells, and suggest a general mechanism by which active STAT3 may be epigenetically excluded from STAT3-responsive genes in non-Th17 lineages.
Naive CD4 ϩ T cells activated in the presence of TGF␤ and IL-6, a physiologic stimulus provided by antigen-presenting cells that phagocytose apoptotic debris from pathogen-infected cells (1), gain the capacity to secrete inflammatory cytokines IL-17A and IL-17F. IL-17F is involved in host defense at mucosal surfaces, whereas IL17A mediates immunity to fungi and bacteria and is involved in autoimmune disease processes such as multiple sclerosis and inflammatory bowel disease (reviewed in Ref. 2). IL-17-producing Th17 2 cells behave as a distinct lineage from cells of the Th1, Th2, and regulatory T cell lineages, which do not produce IL-17, suggesting that distinct transcriptional mechanisms may control il17 gene expression. Genetic studies have established that a combination of STAT3, Runx1, Batf, and IRF4 and the nuclear receptors ROR␥t, RORa, and Ahr is involved in the induction of IL-17 during Th17 differentiation (reviewed in Ref. 3), but why so many factors are required and how they interact with the genomic architecture of the closely linked il17a and il17f genes is not known.
We show that the il17a promoter is primarily responsive to STAT3 or Runx1, whereas the il17f promoter is largely ROR␥tresponsive, and neither promoter is capable of supporting significant transcriptional cooperativity among these three factors. Cytokine genes are commonly regulated by long distance cis regulatory elements (4); therefore, we interrogated seven noncoding sequences (CNS) across an ϳ120-kb region of the mouse genome encompassing the il17a and il17f loci and identified two intergenic regions capable of enhancing transcription from the il17a and il17f promoters. One of these elements is located 5 kb upstream of the il17a gene and corresponds to a previously identified enhancer (5). We now show that this enhancer is particularly responsive to ROR␥t and does not synergize with STAT3 or Runx1. The second element is located between the two il17 genes, 28 kb downstream of il17a and 24 kb downstream of il17f. This is a previously unidentified enhancer, and we show that it synergizes with STAT3, ROR␥t, and Runx1 to drive transcription from both il17a and il17f promoters.
Like Th17 differentiation from naive CD4 ϩ precursors, astrocyte differentiation from neuroepithelial precursors in the developing brain is driven by the combination of a TGF␤ family member bone morphogenic protein 2 (BMP2) and an IL-6 family member leukemia inhibitory factor (LIF) (6). LIF/BMP2 cooperativity in this system results in demethylation of STAT3 cis elements in astrocyte-specific gene promoters, an epigenetic process required for the astrocyte cell fate decision. We therefore hypothesized that lineage-specific induction of the il17 gene by TGF␤ and IL-6 in Th17 cells may be regulated by an analogous mechanism. We demonstrate that CpG dinucleotides in the il17a promoter, the ϩ28-kb enhancer, and the il17f promoter are preferentially demethylated during Th17 lineage commitment and that DNA methylation directly blocks STAT3 binding to a CpG dinucleotide sequence required for full il17a promoter activity. These studies demonstrate that the il17a and il17f genes are regulated transcriptionally by at least two intergenic enhancers, and epigenetically by DNA methylation, which acts as a biochemical switch for STAT3 binding.

EXPERIMENTAL PROCEDURES
Mice-C57BL/6 mice, 4 -6 weeks old, were obtained from The Jackson Laboratory and maintained at the laboratory animal facility of The Children's Hospital of Philadelphia. All animal experiments were conducted according to approved institutional protocols and guidelines.
Vectors-Vectors pGL4-il17a containing the 0.6-kb promoter (5), pRc/CMV-STAT3C (7), pTATA-TK-Luc containing four STAT3 binding elements, and retroviral MIGR-ROR␥t were obtained from Addgene. MIGR1-Runx1 was a obtained from S. Sakaguchi (Kyoto University, Japan). Control Renillaluciferase vector was obtained from Promega. Mutations were introduced in the STAT3 binding site in the il17a 0.6-kb promoter vector using a mutagenesis kit (Stratagene). Conserved noncoding sequence elements surrounding the il17a-il17f loci were cloned into the il17a promoter vector at the XhoI site located at the 5Ј end of the promoter (supplemental Table 1). The pGL4-il17f 0.6-kb promoter construct was generated by cloning the region 0.6 kb upstream of the il17f coding region into the XhoI and HindIII sites. CNS elements were subcloned into the XhoI site of the pGL4-il17f promoter construct.
DNA Methylation-DNA methylation at the il17a-il17f multilocus region was assessed by sodium bisulfite conversion and cloning as described previously (8). Genomic DNA isolated from T helper cells was treated with sodium bisulfite to convert cytosines to uracil. The cis regulatory elements described were PCR-amplified (supplemental Table 1) and cloned into PGEM-T easy vector (Promega), and plasmid DNA isolated from 25-30 clones was sequenced.
Chromatin Immunoprecipitation (ChIP) Analysis-T helper cultures were formaldehyde-fixed, and chromatin was sheared and solubilized by sonication. Chromatin containing histone 3 lysine 4 trimethylation (H3K4me3) and histone 3 lysine 27 trimethylation (H3K27me3) was enriched using specific antibodies (Millipore, Cell Signaling Technologies) and a ChIP assay kit (Millipore), and enriched DNA was purified. cis regulatory elements at the il17 and ifng loci were detected in pre-ChIP versus post-ChIP fractions and quantified by quantitative PCR (supplemental Table 1) as described (9).
Electrophoretic Mobility Shift Assay (EMSA)-EMSA of STAT3 binding to [␥-32 P]ATP-labeled double-stranded oligonucleotides representing sequences in the il17a promoter containing a STAT3 consensus element was determined using nuclear extracts prepared from Th17 cells. To check the methylation effect on STAT3 binding, a methylated probe was synthesized by incorporating the cytosine methylation at a CpG site located within the STAT3 binding element. EMSA reactions were performed as described previously (10).
Oligonucleotide Pulldown Assay-Oligonucleotide probes used for EMSA analysis were 5Ј end-labeled with biotin and incubated (0.5 M) with total cell lysate (400 g) prepared from stimulated Th17 cells under the same ionic conditions used in EMSA reactions. Magnetic streptavidin-conjugated agarose beads (MagSelect, Sigma-Aldrich) were used to collect protein complexes bound to the DNA probes. Beads were washed following the protocol provided by the vendor, bound protein complexes were eluted with Laemmli buffer, and STAT3 was visualized by immunoblot using antibody (C-20) purchased from Santa Cruz Biotechnologies.
Transient Promoter-Reporter Assays-293T or EL4 cells were transfected with wild-type or mutant 0.6-kb il17a promoterluciferase vectors for 48 h and stimulated with PMA (30 ng/ml) and ionomycin (1 M) for 5 h before harvesting. Renilla luciferase driven by a constitutive vector was used to normalize the transfection. Transfection experiments were performed with cells co-transfected with STAT3C (constitutively active form), Runx1, or ROR␥t. STAT3-negative and STAT3-positive reporter luciferase vectors were used as positive and negative controls for STAT3 transfections. Bioluminescence produced by firefly versus Renilla luciferase was estimated using cell lysate prepared from the transfected cells by a Dual-Luciferase assay kit purchased from Promega.
The cytokine IL-23 has been reported to enhance Th17 differentiation and is required for pathogenicity in vivo (11,12). The addition of IL-23 to Th17 cultures did not influence IL-17A production, but augmented the secretion of IL-17F (Fig. 1, J-L).
To identify potential cis regulatory elements for the il17a and il17f loci, we measured methylation of lysine 4 of histone H3 (H3K4me3), a chromatin mark associated with gene activity or expression potential, and methylation of lysine 27 of histone H3 (H3K27me3), a chromatin mark associated with active gene silencing (reviewed in Ref. 13), at several defined (14,15) evolutionarily conserved CNS surrounding the il17a and il17f genes ( Fig. 2A). Naive CD4 ϩ T cells exhibited significant H3K27 trimethylation at the il17a and il17f promoter regions and at CNS regions Ϫ60, Ϫ5, Ϫ2.5, ϩ10, ϩ23, ϩ28, and ϩ36 kb from the il17a TSS (Fig. 2B, left panel). The levels of H3K27me3 at the il17 loci were as high or higher than that observed at the ifn␥ locus in naive cells (Fig. 2B, right panel). Although a small amount of H3K4 trimethylation could be detected near the ifn␥ gene, the il17 multilocus regions was completely devoid of this mark. The chromatin at the il17a and il17f loci was maintained in this same configuration in differentiated Th1 cells (Fig. 2C), whereas the ifn␥ locus exhibited the expected increase in H3K4me3 and a loss of H3K27me3. Th2 cells strongly silenced both of these genomic regions (Fig. 2D). Similar to previous studies (14,16), we observed strong trimethylation of H3K4 at the il17a and il17f promoters, as well as at the Ϫ5, ϩ10, ϩ23, and ϩ28 CNS, but not at the Ϫ60 CNS (Fig. 2E). The increase in H3K4 methylation was accompanied by a significant loss of the opposing H3K27me3 mark, except at Ϫ60 and Ϫ2.5 CNS, which maintained naive levels of the silencing H3K27me3 mark. Conversely, the ifn␥ locus in these cells did not accumulate H3K4me3, but instead was subjected to increased methylation at lysine 27. Th17 cells cultured with IL-23, which led to increased IL-17F secretion (Fig. 1L), retained strong H3K4me3 levels at the il17f promoter, but actually reduced H3K4 trimethylation at the il17a locus (Fig. 2F). These results highlight intergenic regions that may be involved in the transcriptional control of the il17a and il17f loci and indicate that these genes are subject to epigenetic regulation at the level of chromatin remodeling.
The ϩ28 Intergenic CNS Is a STAT3-, ROR␥t-, and Runx1responsive Enhancer for il17 Transcription-To determine whether the CNS regions analyzed in Fig. 2 are directly involved in the regulation of il17 gene transcription, we cloned each CNS (Fig. 3A) in front of the il17a promoter or the il17f promoter and tested the capacity of these DNA elements to drive luciferase activity upon transient transfection in 293T cells. As reported previously (5,17), the 0.6-kb il17a promoter was moderately active by itself in unmanipulated cells (Fig. 3D, yellow bars), and none of the CNS regions were able to augment transcription under these conditions (Fig. 3D, green to blue bars). IL-6 induces il17a gene expression through the activation of STAT3 (18,19); therefore, we tested whether the CNS regions surrounding the il17a and il17f genes were responsive to the activity of this transcription factor. Co-transfection with a constitutively active form of STAT3 (STAT3C (Ref. 7)) resulted in increased promoter activity (Fig. 3D, yellow bars), and the ϩ28 CNS was the only one of the seven conserved regions able to further enhance il17a promoter-driven transcription under these circumstances (Fig. 3D, blue bars). Consistent with this, in silico analysis of the ϩ28 CNS sequence identified a strong STAT3 consensus element (Fig. 3C), and a previous genomewide ChIP analysis indicates that STAT3 is bound to this region in Th17 cells (20). The transcription factor ROR␥t is specifically induced during Th17 differentiation and is required for il17 gene expression (21); therefore, we tested whether the CNS regions surrounding the il17a and il17f genes were responsive to the activity of this transcription factor. The Ϫ5 CNS was recently shown to contain an ROR consensus element (Fig. 3F) and to enhance il17a promoter activity in response to ROR␥t (5). Consistent with this study, we found that co-transfection with ROR␥t did not affect minimal promoter activity (Fig. 3D, yellow bars), but the promoter became ROR␥t-responsive when linked to Ϫ5 CNS (Fig. 3D, green bars). However, in addition to the Ϫ5 CNS, we find that the ϩ28 CNS is also able to cooperate with ROR␥t to drive transcription from the il17a promoter (Fig. 3D, blue bars). Consistent with the fact that the ϩ28 CNS, but not the Ϫ5 CNS, contains consensus elements for both STAT3 and ROR (Fig. 3, B and C), the ϩ28 CNS was the only region capable of enhancing transcription over that of the minimal promoter when co-transfected with both STAT3C and ROR␥t (Fig. 3E, blue bars). The ϩ28 enhancer also contains multiple putative binding sites for Runx1/AML1 (Fig. 3C), another factor known to transactivate the il17a promoter (5); therefore, we tested whether this enhancer is responsive to Runx1. The promoter itself was moderately responsive to Runx1 when expressed alone (Fig. 3E, yellow bars in empty vector versus Runx1), but neither the Ϫ5 CNS nor the ϩ28 CNS was able to significantly enhance transcription under these conditions (Fig. 3E, green and blue bars). Likewise, when Runx1 was co-expressed with ROR␥t, it did not drive transcription from any of the promoter-enhancer constructs significantly more than ROR␥t did by itself (Fig. 3E, ROR␥t versus Runx1ϩROR␥t). However, Runx1 was able to cooperate strongly with STAT3 at both the promoter (Fig. 3E, yellow bars in Runx1 ϩSTAT3) and the ϩ28 enhancer (Fig. 3E, blue bars in Runx1ϩSTAT3) to drive il17a promoter activity to levels 4 -5fold higher than in the presence of either Runx1 or STAT3 alone.
Like the il17a promoter, the il17f promoter was also moderately responsive to STAT3 (Fig. 3F, yellow bars). However, unlike il17a, none of the intergenic CNS elements, including the ϩ28 element, was able to further enhance il17f promoter activity under these conditions (Fig. 3F, STAT3C). The il17f promoter was likewise unresponsive to Runx1 (Fig. 3G, yellow bars in Runx1), and neither the Ϫ5 nor the ϩ28 CNS elements were able to enhance il17f transcription in the presence of this transcription factor (Fig. 3G, green and blue bars in Runx1). The il17f promoter was responsive to RORgt (Fig. 3, F and G, yellow bars in ROR␥t), and both the Ϫ5 and the ϩ28 CNS were able to further enhance il17f promoter activity in the presence of this transcription factor (Fig. 3, F and G, green and blue bars in ROR␥t). Unlike at the il17a promoter, ROR␥t was unable to synergize with STAT3 or Runx1 to drive transcription from the il17f promoter (Fig. 3G). These data indicate that STAT3, ROR␥t, and Runx1 synergize to transactivate the il17a/Ϫ5/ ϩ28 promoter-enhancer unit, whereas the il17f/Ϫ5/ϩ28 promoter-enhancer unit is largely responsive to ROR␥t. These data establish the ϩ28 CNS as an intergenic cis regulatory element that undergoes lineage-specific chromatin remodeling in Th17 cells and is capable of coordinating the activities of STAT3, ROR␥t, and Runx1 to enhance transcription from the il17a and il17f promoters.
Lineage-specific Demethylation of DNA at the ϩ28 Enhancer and the il17a and il17f Promoters in Th17 Cells-Another important mode of epigenetic regulation occurs at the level of DNA methylation at CpG dinucleotides. CpG islands, defined as a Ͼ200-bp region with Ͼ55% GC content, are often found at cis regulatory regions and are used by cells to nucleate DNA methylation machinery at silenced genes (22). The il17 multilocus region is devoid of CpG islands; however, the il17a promoter, the ϩ28 enhancer, and the il17f promoter regions each contain approximately a dozen CpG dinucleotides (Fig. 4). Therefore, we measured DNA methylation at these elements to

Regulatory Architecture of the il17 Locus
determine whether this epigenetic mark is regulated in a lineage-specific manner in T helper cells.
CpG dinucleotides within the ϳ650-bp region upstream of the il17a TSS (Fig. 4A), an ϳ300-bp region including the ϩ28 enhancer (Fig. 4B), and the region ϳ200 -500 bp upstream of the il17f TSS (Fig. 4C) were all highly methylated (at frequencies of 80 -100%) in naive CD4 ϩ T cells (Fig. 4, first row). During Th1 and Th2 differentiation, CpG at the il17a and il17f promoters remained highly methylated, with the exception of the Ϫ33 dinucleotide adjacent to the TATA box (Fig. 4A, second  and third rows) and the Ϫ245 dinucleotide in the il17f promoter (Fig. 4C, second and third rows), which were 60 -70% methylated in these lineages. Th1 cells exhibited significant loss of methylation at the ϩ28 enhancer (Fig. 4B, second row), whereas methylation was largely maintained in Th2 cells (Fig. 4B, third  row). Conversely, differentiating Th17 cells exhibited a 80 -90% loss of DNA methylation across an ϳ175-bp region of the il17a promoter (Fig. 4A, fourth row) and across an ϳ250-bp region upstream of the il17f gene (Fig. 4C, fourth row). The ϩ28 enhancer also exhibited a marked loss of DNA methylation at CpG dinucleotides in Th17 cells, to a much larger extent than that observed in Th1 or Th2 cells (Fig. 4B, fourth row). The Mammalian Conservation

CNS-5 CNS+28
175 125 75 25 0 P -6 0 -5 -2 .5 + 1 0 + 2 3 + 2 8 + 3 6 P -6 0 -5 -2 .5 + 1 0 + 2 3 + 2 8 + 3 6 P -6 0 -5 -2  addition of IL-23, which selectively enhanced the secretion of IL-17F by Th17 cells (Fig. 1), caused preferential demethylation of the il17f promoter, particularly at the Ϫ326 CpG (Fig. 4C, fifth row). These data suggest that DNA demethylation at the il17 loci facilitates il17 gene expression. In this in vitro culture model, Th17 differentiation is driven by the combination of TGF␤ and IL-6 (in the presence of neutralizing Ab to IFN␥ and IL-4). To determine the contribution of TGF␤ receptor-coupled signaling versus IL-6 receptor-coupled signaling to the demethylation of DNA observed at the il17 loci in the Th17 lineage, we stimulated naive CD4 ϩ T cells under the same Th17 culture conditions described above, except that either IL-6 was omitted (Fig. 4, sixth row) or TGF␤ was omitted (Fig. 4, seventh row). Neither signals from IL-6 in the absence of TGF␤ nor signals from TGF␤ in the absence of IL-6 were able to induce full DNA demethylation at the il17a or il17f promoters (Fig. 4, A and C). However, TGF␤, which in the absence of IL-6 promotes regulatory T cell differentiation, was able to mediate partial demethylation of the il17a Ϫ144 CpG (Fig. 4A, sixth row). Furthermore, demethylation at the ϩ28 CNS occurred in cells cultured in either TGF␤ alone (Fig. 4B) or cells cultured with IL-6 alone (Fig. 4B). These data suggest that TGF␤ by itself may be capable of partial epigenetic priming of the il17 loci, but that full demethylation of the il17a-f multilocus region and commitment to the Th17 lineage require the action of both cytokines.
STAT3 Binding to the il17a Promoter Is Lineage-restricted and Is Inhibited by DNA Methylation-Interestingly, the most drastic drop in DNA methylation in the Th17 lineage tended to occur at CpG dinucleotides within consensus elements for STAT3, e.g. Ϫ144 in the il17a promoter and Ϫ309 and Ϫ326 in the il17f promoter (Fig. 4, A and C, red asterisks). This led us to hypothesize that DNA methylation at STAT3 cis elements in non-Th17 cells inhibits STAT3 binding at the il17 genes. The minimal il17a promoter contains a putative STAT3 binding site at Ϫ144 (CCGTCA), and our results in Fig. 3 show that this promoter is STAT3-responsive. To determine whether this element directly contributes to the transcriptional activity of the il17a promoter, we mutated the second cytosine of this element to thymidine (CTGTCA), a substitution (indicated by bold) that would be expected to reduce the affinity of this sequence for STAT3 (23). When transiently transfected into EL-4 cells, the wild-type promoter was moderately active and was responsive to stimulation with PMA, ionomycin, and IL-6 ( Fig. 5A, dark gray bars). The C 3 T mutation in the Ϫ144 element resulted in a 50 -70% reduction in promoter activity under both resting and stimulated conditions (Fig. 5A, light gray bars). To specifically determine the effect of this mutation on STAT3 activity, we utilized constitutively active STAT3C. The wildtype il17a promoter showed moderate activity in 293T cells and was increased 2-fold when co-transfected with STAT3C (Fig.  5B, dark gray bars). However, when co-transfected with a promoter containing the mutated Ϫ144 element, the stimulatory activity of STAT3 was abrogated (Fig. 5B, light gray bar). These data demonstrate that the STAT3 consensus element at Ϫ144 is required for full il17a promoter activity; therefore, we tested whether DNA methylation of this element influences STAT3 binding.
To directly test whether methylation of the il17a promoter opposes STAT3 DNA binding, we probed nuclear extracts of primary Th17 cells with a radiolabeled dsDNA oligonucleotide representing the putative Ϫ144 STAT3 element (CCpGTCA) from the il17a promoter using an electrophoretic mobility shift approach. This sequence, which simulates the unmethylated Ϫ144 site as it would exist in Th17 cells (i.e. Fig. 4A), formed two protein-DNA complexes (Fig. 5D, lane 1, I and II). Formation of both complexes was inhibited by the addition of an anti-STAT3 antibody (Fig. 5D, lane 2), but not by an irrelevant control antibody (Fig. 5D, lane 3), and a defined STAT3 element from the c-fos promoter could effectively compete with the Ϫ144 element for binding to both complexes (Fig. 5D, lanes 4  and 5). These data demonstrate a sequence-specific, STAT3dependent binding activity toward the Ϫ144 il17a promoter element in Th17 cells. To simulate the Ϫ144 STAT3 site as it would exist in non-Th17 cells (i.e. Fig. 4A), we used an oligonucleotide that was synthetically methylated at the CpG dinucleotide (CC(CH3)pGTCA) to probe Th17 nuclear extracts. This methylated STAT3 oligonucleotide completely failed to form complex II and exhibited a reduced capacity to form complex I (Fig. 5D, lanes 6 -8). Indeed, although the unmethylated Ϫ144 element competed very effectively with its own binding (Fig.  5D, lanes 11 and 12), the methylated form was a much less efficient competitor of the unmethylated element binding activity (Fig. 5D, lane 9), comparable with a control NFAT/ AP-1 site in the ARRE-2 element from the il2 promoter (Fig.  5D, lane 13), and only showed partial inhibition of Ϫ144 binding at greater excess (Fig. 5D, lane 10). These EMSA data show that methylation of the Ϫ144 STAT3 element in the il17a promoter greatly reduces its binding affinity.
In a separate approach, we coupled the unmethylated CCpGTCA or methylated CC(CH3)pGTCA oligonucleotides to streptavidin beads. These beads were then incubated with extracts from Th17 cells, washed, and subjected to SDS-PAGE followed by immunoblot analysis with an antibody specific for STAT3. These experiments showed that native STAT3 from Th17 cells binds to the unmethylated Ϫ144 il17a promoter element (Fig. 5E, lane 3) with an affinity comparable with the binding of STAT3 to a defined STAT3 element from the c-fos promoter (Fig. 5E, lane 2). This binding was sequence-specific as a control NFAT/AP-1 site in the ARRE-2 element from the il2 promoter was unable to precipitate STAT3 (Fig. 5E, lane 5). However, STAT3 exhibited greatly reduced binding to the methylated Ϫ144 il17a promoter element (Fig. 5E, lane 4). Furthermore, although the c-fos STAT3 element and the unmethylated Ϫ144 element efficiently competed with Ϫ144 for binding to STAT3 (Fig. 5E, lanes 6 and 8), the methylated Ϫ144 was a less efficient competitor of STAT3 binding to the unmethy- . Lineage-specific, methylation-sensitive binding of STAT3 contributes to transcriptional activity of the il17a promoter. A, EL4 cells were transfected with a wild-type Ϫ0.6-kb il17a promoter-luciferase construct (dark gray bars) or a mutant il17a promoter construct with a C 3 T substitution in the Ϫ144 STAT3 element (light gray bars). Cells were cultured with or without PMA, ionomycin, and IL-6, and luciferase activity was normalized to Renilla for each sample. B, 293T cells were co-transfected with wild-type (dark gray bars) or mutant (mut) (light gray bar) il17a promoter-luciferase and either control empty vector (lane 1) or STAT3C (lanes 2 and 3). Luciferase activity was normalized to Renilla as in A. C, primary naive, Th1, Th2, and Th17 lated Ϫ144 element (Fig. 5E, lane 7). These data demonstrate that the Ϫ144 CpG, which is demethylated specifically in differentiating Th17 cells, resides in a bona fide STAT3 binding element that is required for full il17a promoter activity. Binding of STAT3 to this element is inhibited by DNA methylation, establishing this process as an epigenetic switch that regulates il17 gene expression during T helper differentiation.

DISCUSSION
In this study, we show that the closely linked il17a and il17f loci exhibit Th17 lineage-specific epigenetic modification of histones and DNA. In particular, marked DNA demethylation occurred at the il17a promoter, at the il17f promoter, and at an intergenic conserved noncoding sequence ϩ28 kb from the il17a gene in CD4 ϩ T cells stimulated in the presence of both TGF␤ and IL-6. One of the CpG dinucleotides preferentially demethylated in Th17 cells is located within a STAT3 consensus element in the il17a promoter. We show that this element is a bona fide STAT3 binding site, that this element is required for full il17a promoter activity, and that DNA methylation at this site abrogates STAT3 binding.
These data suggest that CD4 ϩ T cells use methylation of STAT3 DNA elements in cell type-specific promoters as an epigenetic switch for regulation of cell fate decisions during an immune response. An analogous mechanism has been shown to regulate the astrocyte versus neuronal cell fate decision during brain development (6). Similar to Th17 differentiation from naive CD4 ϩ T cells, astrocyte differentiation from neuroepithelial precursors is driven by the combined action of the IL-6 family member LIF and the TGF␤ family member BMP2. Combined LIF-BMP2 signaling leads to demethylation of a STAT3 element in the promoter of the gene encoding the glial fibrillary acidic protein (GFAP), a STAT3-dependent factor required for astrocyte differentiation. We therefore hypothesized that TGF␤ receptor-coupled morphogenic signals during T cell activation might lead to the initial demethylation of STAT3 elements in the il17a and il17f promoters, whereas IL-6 would induce the STAT3 activity required to transactivate these and other Th17-specific genes. We tested this and found that TGF␤ in the absence of IL-6 was able to mediate full demethylation of the ϩ28-kb enhancer and partial demethylation of the STAT3 element in the il17a promoter (Fig. 4). This partial priming of the il17 locus provides an explanation for previous results in which CD4 ϩ T cells expressing constitutively active STAT3 exhibited IL-17 production when stimulated with TGF␤ in the absence of IL-6 (18,19). However, IL-17 production under these circumstances was still ϳ10-fold lower than that produced by cells stimulated with TGF␤ and IL-6, suggesting that STAT3-independent IL-6 receptor signaling may be involved and/or that the residual DNA methylation at the il17a-f loci under these conditions can oppose STAT3 activity.
The cytokine IL-23 is required for in vivo Th17 function, but IL-6 can substitute for its activity in vitro (19). Consistent with this, we found that IL-23 did not influence IL-17A production in TGF␤ plus IL-6 cultures, but did augment IL-17F secretion in a manner associated with reduced DNA methylation at the Ϫ326 STAT3 consensus element in the il17f promoter. IL-23 has also been shown to promote the appearance of IFN␥-pro-ducing Th17 cells in vivo during the development of experimental autoimmune encephalitis (26). However, we found no effect of IL-23 on either histone or DNA methylation at the ifn␥ locus in our short term in vitro Th17 cultures, 3 suggesting that IL-23 acts indirectly to promote IFN␥ production by Th17 cells in vivo or that factors in addition to IL-23 are required for this phenomenon.
A combination of several transcription factors including STAT3, ROR␥t, and Runx1 is required to induce il17 expression. Deletional analysis of the il17a promoter by Ichiyama et al. (17) indicated that the core promoter is contained within the first ϳ300 bp of the region upstream of the TSS, and Zhang et al. (5) found that inclusion of sequences up to Ϫ2 kb increased transcriptional activity. Whether the Ϫ2-kb region is a bona fide part of the promoter versus a proximal enhancer (e.g. like the il2 promoter/enhancer) was not determined; therefore, we chose to use the Ϫ0.6-kb region as the promoter in our studies. We find that the Ϫ0.6-kb il17a promoter contains a bona fide STAT3 cis element and is highly STAT3-responsive, whereas it is not responsive to ROR␥t and is only mildly responsive to Runx1. The il17f promoter has not been previously characterized to our knowledge. We determined that the region 0.6 kb upstream of the il17f coding region contains a well defined TATA box and multiple ROR, AP-1, NFAT, and Runx1 consensus elements and was demethylated specifically in Th17 cells. Therefore, we used this region as the promoter in our study. Our results demonstrate that although the il17a promoter is responsive to STAT3 and Runx1, these factors exhibit no cooperativity at this promoter. Likewise, we show that the il17f promoter is moderately STAT3-responsive, highly ROR␥t-responsive, but by itself is also incapable of supporting synergy between STAT3, ROR␥t, and Runx1.
The Ϫ5-kb CNS and the ϩ28-kb CNS exhibit permissive epigenetic marks and DNase hypersensitivity in activated Th17 cells (our studies and Ref. 14) and were able to enhance transcription from both promoters in transient reporter assays, indicating that these enhancers are part of the transcriptional architecture of the il17a and il17f multilocus region. The Ϫ5-kb CNS was previously shown to enhance transcription from a Ϫ2-kb il17a upstream construct in response to ROR␥t and Runx1 (5). Our studies confirm that the Ϫ5-kb enhancer is highly ROR␥t-responsive; however, this CNS was unable to enhance transcription from the minimal Ϫ0.6-kb promoter in response to either STAT3 or Runx1. The Ϫ5-kb CNS was also able to enhance transcription from the already highly ROR␥tresponsive il17f promoter, but did not respond to Runx1 or STAT3, individually or in combination. Therefore, we conclude that the primary role for the Ϫ5-kb enhancer is to coordinate ROR␥t activity and that another element must provide cooperativity to the il17a-il17f transcriptional unit.
It is possible that the action of multiple cis regulatory elements, although these elements are not capable of organizing cooperativity between STAT3, Runx1, and ROR␥t as a single element, might synergize to drive transcription when brought together in space. However, the combinatorial possibilities for such cooperativity in cis within the ϳ120-kb il17a-il17f multilocus region appears to be limited as our studies indicate that the Ϫ5-kb CNS and the promoters do not synergize under conditions of forced proximity and that the other intergenic CNS fail to exhibit enhancer activity in the context of either promoter. This could be explained by the fact that the Ϫ2.5-kb CNS retains a significant level of H3K27me3 mark in the Th17 lineage and contains no STAT3 or ROR consensus elements, and the ϩ36-kb CNS exhibits low levels of H3K4me3 and DNase hypersensitivity in Th17 cells (our studies and Ref. 14), contains only a weak match for a STAT3 consensus element, and does not contain an ROR site. The ϩ23 CNS is primed in Th17 cells (i.e. accumulates high levels of the H3K4me3 mark and loses H3K27me3), but contains no STAT3 nor ROR consensus elements. The Ϫ60 CNS contains a single, weak match for the STAT3 consensus and a single weak match for the ROR consensus, but this region remains epigenetically silent (i.e. high H3K27me3) in the Th17 lineage. The ϩ10 CNS exhibits a strong change in H3K4 and H3K27 trimethylation and DNase hypersensitivity in Th17 cells (our studies and Ref. 14), contains a strong STAT consensus element, and binds STAT3 in vivo in activated Th17 cells (20). Why this element does not enhance in our system is not clear, but as with all these CNS, it is possible that additional factors not expressed in 293T cells may be required.
However, an excellent candidate for an element that per se supports cooperativity among the various factors required for IL-17 production is the ϩ28-kb CNS, which we show is a strong enhancer of transcription from both the il17a and the il17f promoters. Although many of the CNS examined here exhibited DNase hypersensitivity in activated Th17 cells in the study by Mukasa et al. (14), the ϩ28-kb CNS was the only CNS that showed constitutive hypersensitivity in resting Th17 cells, indicating the presence of a unique chromatin structure in this region. This enhancer shows strong STAT3 binding in vivo (20) and is highly responsive to STAT3 in our transient reporter assays. This enhancer is also responsive to ROR␥t and Runx1, but importantly, this enhancer is able to support transcriptional cooperativity between all three factors when expressed together. This CNS may also cooperate with additional DNAbinding proteins as this region also contains putative binding sites for SP1 and NFAT. Interestingly, one of these NFAT sites contains a CpG dinucleotide that we show is methylated in non-Th17 cells and unmethylated in Th17 cells. Methylation of an NFAT element has been shown to block NFAT binding and transactivation of the human il2 promoter (27), suggesting that DNA methylation could act in an analogous manner at the il17 loci to oppose NFAT-mediated transactivation. The ϩ28-kb enhancer also contains multiple AP-1 consensus elements, suggesting that Fos/Jun heterodimers may act at this region. Alternatively, these sites may recruit Batf, an alternative Jun partner that was recently shown to be required for il17a gene expression (28). Indeed, this previous study demonstrated Batf binding to the endogenous ϩ28 CNS region in Th17 cells, further indicating that this factor transactivates via interacting with the ϩ28-kb enhancer. Therefore, we propose that the ϩ28-kb enhancer is an integral component of the il17a-il17f transcriptional architecture that organizes the activity of multiple tran-scription factors around the il17a and il17f promoters. Although this work does not strictly establish whether the ϩ28-kb CNS is required for Th17 development, these studies do establish the element as a classically defined enhancer that can be targeted by homologous recombination in future studies.
In summary, we show that a conserved intergenic element can coordinate the activities of STAT3, ROR␥t, and Runx1 to enhance transcription from the il17a and il17f promoters and that Th17 cells utilize an evolutionarily conserved mechanism involving methylation at DNA binding elements to restrict STAT3 activity at the il17a gene in non-Th17 lineages.