EZH2 and Histone 3 Trimethyl Lysine 27 Associated with Il4 and Il13 Gene Silencing in T H 1 Cells* □ S

Differentiation of naı¨ve CD4 T cells toward the T helper 1 (T H 1) and T helper 2 (T H 2) fates involves the transcriptional repression and enhancement, respectively, of Il4 and Il13 , adjacent chromosome 11 genes encoding the canonical T H 2 cytokines interleukin-4 and interleukin-13. Proper execution of this developmental fate choice during immune responses is critical to host defense and, when misregulated, leads to susceptibility to infectious microbes and to allergic and autoimmune diseases. Here, using chromatin immunoprecipitation and real time reverse transcription PCR we identify the Polycomb family histone methyltransferase EZH2 as the enzyme responsible for methylating lysine 27 of histone H3 at the Il4 - Il13 locus of T H 1 but not T H 2 cells, impli- cating EZH2 in the mechanism of Il4 and Il13 transcriptional silencing. Acting on instructions issued by cells of the innate immune system, naı¨ve CD4 T cells commit to effector cell fates specifically tailored to address the confronting class of pathogens and the constraints of infected host tissue environments. IP PCR products were subjected to 3-fold serial dilutions over a 27-fold range prior to gel electrophoretic separation. The relative intensities of the gel-resolved major centromeric repeat products (showing a typical laddering pattern) reveal a 3–9-fold enrichment of H3K9me3 in the IP versus input samples, as described previously (14).

Acting on instructions issued by cells of the innate immune system, naïve CD4 T cells commit to effector cell fates specifically tailored to address the confronting class of pathogens and the constraints of infected host tissue environments. The best characterized of these fates are called T helper 1 (T H 1) 1 and T helper 2 (T H 2), distinguished primarily by their non-overlapping patterns of cytokine gene expression potential. Il4 and Il13, adjacent genes on chromosome 11, encode the canonical T H 2 cytokines responsible both for promoting immune responses against extracellular pathogens and, when misregulated, causing allergic and autoimmune diseases (1). Heritable states of transcriptional repression and enhancement of these genes during T H 1 and T H 2 development, respectively, are associated with developmentally programmed chromatin structural changes at the Il4-Il13 locus. The molecular nature of these changes, the cis-acting regulatory elements that coordinate them, and the relationship of these changes to Il4 and Il13 transcriptional potential are not well understood.
The flexible amino termini of nucleosomal histones harbor multiple residues that can undergo a variety of post-translational modifications, including acetylation, methylation, and phosphorylation. Combined with the octameric structure of the nucleosomal core, this flexible modification system harbors tremendous combinatorial diversity and information-coding potential (the socalled histone code) that can endow discrete genetic intervals with specific functional properties (2). Along a given interval, factors recruited to specific regulatory elements can create heritable patterns of histone modification that, in turn, can influence the transcriptional behavior of associated genes by creating or destroying binding platforms for transcriptional activators and repressors, nucleosome remodeling, and chromatin-packaging machinery (3). Histone H3, for example, can undergo a variety of post-translational modifications that can be classified as transcription-associated (acetylated lysine 9/lysine 14 (H3K9/14ac), dimethylated lysine 4 (H3K4me2), and phosphorylated serine 10) and silence-associated (dimethylated and trimethylated lysine 9 (H3K9me2 and H3K9me3) and trimethylated lysine 27 (H3K27me3)) (for review see Ref. 4).
Thirteen clusters of DNase I hypersensitive sites (HSs) have been mapped at the Il4-Il13 locus of T H 2 cells. Based upon lineage specificity and activation dependence, these can be classified as T H 2-specific/constitutive (HS V , HS III , HS II , HS I , HS 0 , HS S1 , HS S2 , HS 2 , and HS 1 ), T H 2-specific/activation-dependent (HS Va ), and naïve/T H 1/T H 2-shared/constitutive (HS IV and HS S3 ) (see Fig. 2, top). Although useful as a roadmap of potential regulatory elements at the Il4-Il13 locus, DNase I hypersensitivity alone is insufficient to reveal functional regulatory elements and their modes of action at each stage of effector T cell development. For example, the transcriptional states of Il4 and Il13 in naïve CD4 T cells and T H 1 cells are permissive and silent, respectively, despite sharing identical patterns of DNase I hypersensitivity at the Il4-Il13 locus (5,6).
Analysis of locus-wide patterns of histone modification provides an avenue to discern the developmental stages at which specific HS-containing regions act to modulate Il4 and Il13 transcriptional permissiveness. A recent survey of transcription-associated H3 modifications revealed the occurrence of H3K4me2 near the T H 2-specific/constitutive HS V in naïve CD4 T cells but not in T H 1 cells. Thus, the HS V -containing region is implicated in the maintenance of Il4 and Il13 transcriptional permissiveness at the naïve CD4 T cell stage, even though hypersensitivity at HS V is not yet detectable in these cells (5,6). Missing still from this emerging picture of the sequence of chromatin structural events at the T H 2 cytokine locus in developing T H 1 and T H 2 cells is the nature and dynamic distribution of transcriptional silenceassociated histone modifications.
EZH2 and SUV39 are histone methyltransferases (HMTs) with in vivo specificity for H3 lysine 27 and H3 lysine 9, respectively (7). SUV39, together with heterochromatin protein 1 with binding specificity for methylated H3 lysine 9, has been shown to be critical in the formation of heterochromatin (7). EZH2, a member of the Polycomb group, occurs in a number of multisubunit complexes with other Polycomb group members including EED (embryonic ectoderm development). A Polycomb group complex containing the eponymous Polycomb has binding affinity for methylated H3 lysine 27 and appears to be involved in the repression of euchromatic genes (4). EZH2 is expressed in lymphocytes and has been shown to play a critical role in B lymphopoiesis (8). Its role in T cell biology is not known.
Here, we report the analysis of repressive H3 modifications at the Il4-Il13 locus during effector T cell development. Our data demonstrate that, in naïve CD4 T cells and T H 1-primed but not T H 2-primed cells, lysine 27-methylated H3 occurs selectively at HS IV and HS S3 , the only two lineage-nonspecific hypersensitive sites at the Il4-Il13 locus. T H 1-specific silencing of Il4 and Il13 correlates with elevated levels of H3K27me3 at HS IV and the subsequent spreading of H3K27me3 into flanking regions. Furthermore, we show that EZH2 associates with HS IV and HS S3 in naïve, T H 1, and, unexpectedly, T H 2 cells. Together, our results suggest that developmentally regulated H3 lysine 27 methyltransferase activity of EZH2 constitutively bound at HS IV and HS S3 is responsible for specifying/maintaining Il4 and Il13 transcriptional silence in T H 1 cells.
Naïve T Cell Purification-Combined lymph node and spleen cell suspensions generated from 4 -6 week BALB/c mice (Jackson Laboratories, Bar Harbor, ME; housed under specific pathogen-free conditions at the University of Washington, Seattle, WA) were subjected to two sequential AutoMACS purification steps. First, CD4 T cells were obtained by negative selection using a CD4 T cell isolation kit (Miltenyi Biotec, catalog number 130-090-860). Next, CD4 ϩ CD62L Hi cells were isolated by positive selection from CD4 T cells stained with anti-CD62L MicroBeads (Miltenyi Biotec, catalog number130-049-701). The purity of CD4 ϩ CD62L Hi T cells, as determined by fluorescence-activated cell sorter analysis, was Ն 97%.
Real Time Reverse Transcription-PCR (RT-PCR)-RNA was isolated free of contaminating genomic DNA using the RNeasy mini kit (Qiagen) Random hexamer-primed cDNA was generated using the Superscript II RNase H Ϫ reverse Transcriptase kit (Invitrogen; catalog number 18064-014). Real time PCR reactions were performed and analyzed on a Stratagene MX4000. The PCR reaction buffer contained 10 mM Tris (pH8.3), 50 mM KCl, 4.5 mM MgCl 2 , 0.01% Tween 20, 0.3% Me 2 SO, 0.0025% SYBR Green I solution (Molecular Probes; catalog number S-7563), 50 nM each primer (Supplemental Fig. 1 in the on-line version of this article), and 2 units/reaction Hot Start Taq polymerase (Qiagen) per 25-l reaction. Cycling conditions were 94°C for 15 min followed by 40 cycles of 94°C for 20 s, 61°C for 1 min, and 72°C for 40 s. C T values for no reverse transcriptase controls were at least four times higher than experimental samples, corresponding to Ͻ6% background. More often, no reverse transcriptase background was Ͻ0.8%. For a given cDNA, relative abundance of each target was normalized to Hprt according to the formula 2 Ϫ⌬CT , where ⌬C T ϭ C T TARGET Ϫ C T HPRT . Hprtnormalized target signals were expressed relative to expression in NIH3T3 according to the formula 2 Ϫ⌬⌬CT , where ⌬⌬C T ϭ ⌬C T SAMPLE Ϫ ⌬C T NIH3T3 . Thus one arbitrary unit corresponds to the relative level of a given target in NIH3T3.
Chromatin Immunoprecipitation (ChIP) Assays-DNA recovered from an aliquot of sheared chromatin was used as the "input" sample. The remaining chromatin was pre-cleared with protein A-and protein Gagarose (Upstate Biotechnology; catalog numbers 16-156 and 16-266) and then incubated with antibody overnight at 4°C (anti-H3K9/14ac, catalog number 06-599 and anti-H3K4me2, catalog number 07-030 from Upstate Biotechnology), both used at 2 g/ml; anti-H3K9me3 (sera 4861) and anti-H3K27me3 (sera 6523, number 3200; Ref. 15), both used at 3 g/ml; and anti-EZH2 (9), used at 1:500. Input DNA and DNA recovered after immunoprecipitation (IP) were quantified using PicoGreen fluorescence (Molecular Probes, Eugene, OR). Equivalent masses of IP and input DNA were compared by real time PCR as described above for RT-PCR with the following modifications. Taq polymerase was from Qiagen (Hot Start), and cycling conditions were 94°C for 15 min followed by 45 cycles of 94°C for 20 s, 61°C for 1 min, and 72°C for 40 s. Data are presented as the ratio of IP to input C T values. For analysis of centromeric repeats, end point PCR analysis was performed to compare equivalent masses of input and IP DNA. PCR products were subjected to 3-or 4-fold serial dilution and separated by gel electrophoresis.

Transcriptional Response of T H 1-and T H 2-primed Naïve
CD4 T Cells-Using real time RT-PCR, we characterized the 4-hour phorbol 12-myristate 13-acetate/ionomycin-induced transcriptional response of the clustered Kif3a, Il4, Il13, and Rad50 genes in one-round and two-round T H 1-and T H 2-primed (T H 1R1, T H 1R2, T H 2R1, and T H 2R2) CD4 T cells (Fig. 1). For developing T H 2 cells, transcription of Il4 and Il13 increased whereas that of Ifn␥ decreased with each successive round of priming. The reciprocal pattern was observed for developing T H 1 cells. Interestingly, Il13 expression consistently decreased in T H 1R1 cells and then reappeared at low levels in T H 1R2 cells. Though still, to date, unexplained, this Il13 effect was highly reproducible and consistent with observations made by others. 2 Previously, we demonstrated that Il4 transcriptional permissiveness was maintained at naïve CD4 T cell levels 48 h after the onset of T H 1 priming (6). Together with our current results demonstrating transcriptional silence of Il4 in oneround T H 1-primed cells (Fig. 1), we conclude that transcriptional repression of Il4 requires between 2 and 4 days of T H 1priming. Consistent with a previous report (5), expression of Kif3a and Rad50 was ϳ2-3-fold higher in T H 2-primed versus T H 1-primed cells (Fig. 1). However, unlike Il4 and Il13, Kif3a and Rad50 expression did not increase further in T H 2R2 cells, demonstrating, as shown previously (6), that the genes flanking Il4 and Il13 are insulated from the regulatory events governing the Il4-Il13 locus.
Repressive H3 Modifications in T H 1-and T H 2-primed Cells-One previous study has described the occurrence of H3K9me at the Il4-Il13 locus of T H 1 cells, suggesting that this modification might be involved in the transcriptional silencing of Il4 and Il13 (10). To test this possibility, we purified naïve CD4 T cells from young BALB/c mice and cultured them for two rounds in T H 1-priming conditions. Primed cells were harvested and pro-2 M. Kubo and A. Rao, personal communications. cessed for ChIP using an antibody highly specific for H3 trimethylated on lysine 9. Using real time PCR, we measured the relative enrichment of specific target sequences in equivalent amounts of DNA obtained from unprecipitated and precipitated chromatin. PCR primer pairs targeted 17 locations at the Il4-Il13 locus, sampling elements known to be functionally (11,12), structurally (5, 13), or computationally (11,12) implicated in cytokine regulation (Fig. 2, top, and described in Ref. 6). Although in T H 1R2 cells control major centromeric repeat sequences displayed enrichment for H3K9me3 ( Fig. 2B; see Ref. 14), at the Il4-Il13 locus no H3K9me3 enrichment was detected ( Fig. 2A). Centromeric repeat sequences were not enriched in chromatin precipitated with antibodies to H3K9/14ac (data not shown). Similar results were obtained with antibodies to H3K9me2 (data not shown). These results suggest that the transcriptional silence of Il4 and Il13 in T H 1-primed cells does not involve H3 lysine 9 methylation.
As the anti-H3K9me antibody used in the previous study may have cross-reacted with H3K27me (10, 15), we decided to perform ChIP analysis of T H 1-primed cells with an antibody highly specific for H3K27me3 (15). T H 1R1 cells displayed two prominent peaks of H3K27me3 enrichment centered on HS IV and HS S3 , respectively (Fig. 2C). T H 1R2 cells showed, in addition to these peaks, increased H3K27me3 levels across the entire Il4-Il13 locus, extending at least from HS V downstream of Il4 to HS 2 upstream of Il13 (Fig. 2D). This increase was specific insofar as H3K27me3 was at or below background levels at the G6pd locus in both T H 1R1 and T H 1R2 cells (data not shown). These results suggest that the initial report of H3K9me at the Il4-Il13 locus was likely due to cross-reactivity of antibodies that failed to distinguish between H3 methylation on lysine 9 and lysine 27, residues with very similar sequence contexts (16). We conclude that, during the development of T H 1 cells, H3K27me3 occurs initially at HS IV and HS S3 and subsequently spreads to surrounding regions of the locus.
To determine whether any aspect of the H3K27me3 pattern was T H 1-specific, as would be predicted for a histone modification involved in Il4 and Il13 transcriptional silence, we used the H3K27me3-specific antibody in ChIP analysis of developing T H 2 cells. In contrast to T H 1-primed cells, T H 2R1 and T H 2R2 cells displayed significantly lower and almost undetectable H3K27me3 levels, respectively (Fig. 3, A and B), despite exhibiting strong signals at the control MyoD locus (Fig. 3C). The low level of H3K27me3 that was detectable in one-round T H 2-primed cells localized predominantly to HS S3 and could not be found in surrounding regions (Fig. 3A). A faint signal was observed at HS IV in T H 2R1 cells but was no longer detectable in T H 2R2 cells. As was the case in T H 2R2 cells, no H3K27me3 enrichment was detected across the entire Il4-Il13 locus of the T H 2 clone D10.G4 despite strong signals at the control MyoD locus (data not shown).
Repressive H3 Modifications in Naïve CD4 T Cells-The occurrence of H3K27me3 at HS IV and HS S3 (HS sites that are shared among T H 1, T H 2, and naïve CD4 T cells) in T H 1R1 and T H 2R1 cells prompted us to investigate whether H3K27me3 might pre-exist in naïve CD4 T cells. To test this possibility, we isolated naïve CD4 T cells from spleen and lymph nodes of young BALB/c mice and performed quantitative ChIP analysis for H3K27me3. Indeed, H3K27me3 was detected in naïve CD4 T cells in a pattern similar to that of T H 1R1 cells, focused in two peaks at HS IV and HS S3 but absent from surrounding regions of the locus (Fig. 4). Interestingly, the HS S3 peak was significantly higher in naïve CD4 T cells than in T H 1-primed cells, suggesting greater involvement of this putative regulatory element early in the development of CD4 effector cells. This result demonstrates that low level Il4 and Il13 transcriptional permissiveness in naïve CD4 T cells is compatible with the occurrence of H3K27me3 at HS S3 (and, perhaps, to a much lower level at HS IV ).

EZH2 Binds to HS IV and HS S3 in Naïve T H 1-and T H 2primed Cells-
The only known HMT capable in vivo of trimethylating H3 lysine 27 is the Polycomb family protein EZH2 (17). Apart from the inactive X chromosome in XX female cells, physiological targets of EZH2 binding have only recently been described and have not yet been verified functionally (18 -20). To test whether H3 lysine 27 trimethylation at the Il4-Il13 locus might be catalyzed by EZH2, we asked whether it binds in vivo to the Il4-Il13 locus. We used an EZH2-specific antibody to perform quantitative ChIP analysis of naïve CD4 T cells and T H 1R2 and T H 2R2 cells. In T H 1R2 cells we detected peaks of EZH2 binding that coincided precisely with the locations of maximal H3K27me3 enrichment at HS IV and HS S3 (Fig. 5B). No EZH2 binding was detected in the regions surrounding HS IV and HS S3 (Fig. 5B), where H3K27me3 was detected at lower levels in T H 1R2 cells (Fig. 2D). A similar binding pattern was also detected in naïve CD4 T cells, correlating quantitatively with the increased magnitude of H3K27me3 enrichment at HS S3 (Figs. 4 and 5A). Surprisingly, in T H 2R2 cells, where H3K27me3 was no longer detectable at the Il4-Il13 locus, we still detected strong EZH2 binding at HS IV and HS S3 (Figs. 3B and 5C). Together, these results demonstrate that EZH2 binds constitutively to HS IV and HS S3 in naïve CD4 T cells and in T H 1R2 and T H 2R2 cells. In naïve CD4 T cells and T H 1-primed cells, quantitative and spatial correlations of EZH2 binding and H3K27me3 abundance are consistent with EZH2 being the HMT responsible for H3 lysine 27 trimethylation. The persist-

ence of HS IV -and HS S3 -bound EZH2 in T H 2R2 cells where
H3K27me3 is no longer detectable is consistent with developmental regulation of EZH2 activity at a post-chromatin binding step or with the involvement of a novel H3 lysine 27-specific HMT. Finally, the T H 1-specific appearance of H3K27me3 in the regions surrounding HS IV and HS S3 may indicate that EZH2 can interact transiently with these regions, perhaps processively by tracking or by looping from points of nucleation at HS IV and HS S3 (21).
Repressive H3 Modifications in a Fibroblast Cell Line-The fibroblast cell line NIH3T3 does not express Il4 or Il13 and lacks all known DNase I hypersensitive sites at the Il4-Il13 locus, including HS IV and HS S3 (5). At the Il4-Il13 locus the transcription-associated H3 modification phosphoserine 10, found broadly distributed in T H 2-primed cells and the T H 2 clone D10.G4, does not occur in NIH3T3 (6). Similarly, despite strong signals for each at the control G6pd locus, no H3K9/14ac and H3K4me2 occurred across the Il4-Il13 locus of NIH3T3, (Fig. 6, A-C). To assess whether transcriptional silence of Il4 and Il13 in a non-T cell lineage would be associated with the same pattern of repressive chromatin modifications detected in T H 1-primed CD4 T cells, we analyzed NIH3T3 by quantitative ChIP for the occurrence of H3K9me3. Despite control signals at major centromeric repeat sequences, we detected no H3K9me3 across the Il4-Il13 locus (Fig. 6, D and E). Similar results were obtained for H3K27me2 (data not shown). These results were confirmed in primary mouse embryonic fibroblasts (data not shown). Thus, the constitutive transcriptional silence of Il4 and Il13 in fibroblast lineage cells appears to involve neither H3K27me3 nor H3K9me3, suggesting lineage-specific epigenetic mechanisms for the maintenance of Il4 and Il13 transcriptional silence. DISCUSSION Knowing the chemical nature of post-translational histone modifications associated with Il4 and Il13 gene silencing as well as the responsible catalytic enzymes is important for understanding the mechanisms by which heritable transcriptional states are developmentally specified at the Il4-Il13 locus. In this report we exploited antibodies with proven specificity for distinct transcriptional silence-associated H3 modifications in ChIP analyses to demonstrate that the principal H3 modification associated with T H 1-dependent silencing is the trimethylation of lysine 27 and not of lysine 9. The striking localization of H3K27me3 to HS IV and HS S3 suggests that elements containing these sites function either together or separately as transcriptional silencers in T H 1 cells. Indeed, 2) at the Il4-Il13 locus in naïve CD4 T cells following T H 1-priming for one round (panel C) or two rounds (panels A and D). Shown in panel B is the relative abundance of major centromeric repeat sequences in two-round T H 1-primed naïve CD4 T cell DNA before (Input) and after (IP) ChIP using an antibody directed against H3K9me3. Input and IP PCR products were subjected to 4-fold serial dilutions over a 64-fold range prior to gel electrophoretic separation. The relative intensities of the gel-resolved major centromeric repeat products (showing a typical laddering pattern) reveal a 4 -16-fold enrichment of H3K9me3 in the IP versus input samples, as described previously (14). deletion of one of them (HS IV ) prevents efficient silencing of Il4 in T H 1-primed cells (23). We also demonstrate that the Polycomb family H3 lysine 27-specific HMT EZH2 binds to HS IV and HS S3 , suggesting it is likely to be responsible for trimethylating H3 lysine 27 at the TH2 cytokine locus.
Located in phylogenetically conserved regions of the T H 2 cytokine locus, HS IV and HS S3 are unique in being the only two DNase I hypersensitive sites that occur in T H 1, T H 2, and naïve CD4 T cells. Our data demonstrate that they are also the only sites at the T H 2 cytokine locus to display EZH2 binding in all EZH2 and Silencing of Il4 and Il13 31474 three lineages, suggesting that this binding is the basis for their shared DNase I hypersensitivity. The level of H3K27me3 enrichment at HS S3 and HS IV was much higher in naïve CD4 T cells than in developing T H 2 cells, suggesting a progressive loss of H3K27me3 that was essentially complete after two rounds of priming. This loss may be due to the activity of an as yet unidentified H3 lysine 27 demethylase or perhaps progressive dilution of the H3K27me3 mark through successive cell cycle rounds in the absence of maintenance H3 lysine 27 methylation. The pattern in developing T H 1 cells was more complex, comprising an increase and decrease at HS IV and HS S3 , respectively, such that in T H 1R1 cells comparable H3K27me3 levels occur at both sites. In T H 1R2 cells H3K27me3 at HS IV and HS S3 remains and appears to have spread into flanking regions. However, interactions between EZH2 and flanking regions must be transient, as EZH2 binding was not detectable. Alternatively, H3K27me3 in flanking regions is dependent upon an unidentified H3 lysine 27-specific HMT.
The feature that correlates best with the T H 1-dependent silencing of Il4 and Il13 is the developmental elevation of H3K27me3 levels initially at HS IV and subsequently in flanking regions, both of which are observed only in T H 1-primed cells. Contrary to previous reports (10), we did not detect H3 lysine 9 methylation at the T H 2 cytokine locus of T H 1 cells. Because both H3 lysine 9 and H3 lysine 27 occur in an identical sequence motif (ARKS), a likely explanation for the discrepancy is that the antibody used in the earlier study cross-reacted with the H3 methylated on lysine 27, which had spread from HS IV and HS S3 to the Il4 promoter region. Alternatively, H3K9me may occur only in the mature stage of T H 1 development represented by the T H 1 clone analyzed by Grogan et al. (10).
In T H 2R2 cells, HS IV and HS S3 contain bound EZH2 but no H3 lysine 27 trimethylation. This result is consistent with either EZH2 not being responsible for HS IV and HS S3 H3 lysine 27 trimethylation or with the HMT activity of EZH2 being regulated independently of chromatin binding. We support the latter interpretation. First, in naïve CD4 T cells and in T H 1 cells EZH2 is observed to bind specifically at HS IV and HS S3 , precisely the locations where peak H3K27me3 enrichment occurs. Second, in naïve CD4 T cells and T H 1 cells there is a strong correlation between the degree of EZH2 binding and H3K27me3 enrichment. Finally, following polyclonal stimulation by phorbol 12-myristate 13-acetate plus ionophore, CD4 T lymphocytes harboring a homozygous targeted deletion inactivating the Ezh2 gene rapidly lose all detectable nuclear H3K27me3, suggesting that EZH2 is responsible for most if not all H3 lysine 27 tri-methylation. 3 One possible mechanism for regulating the activity of EZH2 independently of its chromatin binding is through developmental regulation of cofactors capable of influencing the HMT activity of EZH2. Consistent with this possibility, EZH2 is known to occur in at least two distinct multisubunit complexes (PRC2 and PRC3) along with another Polycomb family member, EED (9). Alternative translation initiation generates four distinct EED isoforms that feature differentially truncated amino termini (9). Depending on the isoform of its EED partner, EZH2 has been shown in vitro to switch its substrate specificity from H3 lysine 27 to histone 1 (H1) lysine 26 or to become unable to mediate methylation of either substrate (9). Thus, EZH2 that remains bound to HS IV and HS S3 in T H 2primed cells may have acquired a new EED partner that prevents H3 lysine 27 methylation, thereby preventing transcriptional silencing. Alternatively, the T H 2-specific absence of H3K27me3 may arise from H1 recruitment shown in in vitro biochemical studies to prevent the H3 lysine 27 methyltransferase activity of EZH2 (9).
In somatic cell lineages, distinct biochemical pathways may execute facultative and constitutive gene silencing. The textbook example of facultative heterochromatin is the inactive X chromosome in female somatic cells. Recent work has demonstrated that the process of X inactivation involves H3 lysine 27 methylation (24). By contrast, constitutive heterochromatin at telomeres and centromeres involves principally H3 lysine 9 methylation (25). Transcriptional silencing of Il4 and Il13 is facultative in CD4 T lymphocytes and constitutive in fibroblasts. From this perspective it is interesting to note that transcriptional silence of Il4 and Il13 was accompanied by H3 lysine 27 methylation in CD4 T lymphocytes but not in fibroblasts ( Fig. 6 and data not shown). This difference may be indicative of a more general phenomenon in which, for a given cell lineage, constitutive silencing and facultative silencing are functionally segregated to different biochemical pathways, with the establishment of only the latter (facultative silencing) involving H3 lysine 27 methylation. Consistent with this hypothesis is the absence of H3K27me3 from the constitutively silenced muscle-specific gene MyoD in fibroblast lineage cells (data not shown). The occurrence of H3K27me3 at the MyoD locus of CD4 T cells (Fig. 3E) is not necessarily inconsistent with this hypothesis, as it has been reported that hematopoietic cells can trans-differentiate into muscle (22).
In summary, we have shown that the principle H3 modification associated with T H 1-dependent silencing is trimethylation of lysine 27 initially at HS IV and subsequently in regions flanking HS IV and HS S3 , consistent with the known role of HS IV as a transcriptional silencer in T H 1 cells. We also demonstrate that the Polycomb family H3 lysine 27-specific HMT binds to HS IV and HS S3 , suggesting that it is likely to be responsible for trimethylating H3 lysine 27 at the T H 2 cytokine locus and, hence, is an important mediator of Il4 and Il13 gene silencing.