Histone H3 Lysine 36 Trimethylation Is Established over the Xist Promoter by Antisense Tsix Transcription and Contributes to Repressing Xist Expression

Abstract

One of the two X chromosomes in female mammals is inactivated by the noncoding Xist RNA. In mice, X chromosome inactivation (XCI) is regulated by the antisense RNA Tsix, which represses Xist on the active X chromosome. In the absence of Tsix, PRC2-mediated histone H3 lysine 27 trimethylation (H3K27me3) is established over the Xist promoter. Simultaneous disruption of Tsix and PRC2 leads to derepression of Xist and in turn silencing of the single X chromosome in male embryonic stem cells. Here, we identified histone H3 lysine 36 trimethylation (H3K36me3) as a modification that is recruited by Tsix cotranscriptionally and extends over the Xist promoter. Reduction of H3K36me3 by expression of a mutated histone H3.3 with a substitution of methionine for lysine at position 36 causes a significant derepression of Xist. Moreover, depletion of the H3K36 methylase Setd2 leads to upregulation of Xist, suggesting H3K36me3 as a modification that contributes to the mechanism of Tsix function in regulating XCI. Furthermore, we found that reduction of H3K36me3 does not facilitate an increase in H3K27me3 over the Xist promoter, indicating that additional mechanisms exist by which Tsix blocks PRC2 recruitment to the Xist promoter.

INTRODUCTION

In mammals, X chromosome inactivation (XCI) provides dosage compensation between the sexes for X-linked genes (1). The noncoding RNA (ncRNA) Xist initiates chromosome-wide inactivation of one of the two X chromosomes of female cells. On the active X chromosome in males and females, Xist is repressed by several mechanisms. In mice, the Tsix ncRNA is transcribed over the Xist locus in the antisense orientation and functions as a repressor of Xist on the chromosome from which it is transcribed (2). The function of Tsix has been extensively studied in mouse embryonic stem (ES) cells, which constitute a model for studying the initiation of random XCI (1, 3–5). Disruption of Tsix leads to derepression of Xist whose extent varies with experimental details in a number of different studies (6–9). In mouse preimplantation development, imprinted XCI leads to inactivation of the paternally inherited X chromosome in female embryos. Overexpression of Tsix from the paternal X chromosome prevents XCI and causes lethality (10). Conversely, disruption of Tsix on the maternally inherited X chromosome in males and females causes lethality due to misregulation of imprinted XCI in the extraembryonic lineages (11, 12). However, in the embryonic lineages, the Tsix disruption-bearing X chromosome is fated to become the inactive X chromosome (Xi) (6, 12).

Mutation of Tsix causes death of male embryos due to initiation of X inactivation in extraembryonic tissues. This lethality can be prevented by complementing the extraembryonic lineages, suggesting that Tsix-independent mechanisms can act to repress Xist in the embryonic lineages (13). Tsix-independent mechanisms can also be inferred from other mammals, including humans, which lack a functionally conserved Tsix homologue (14). Our previous work linked Tsix-independent Xist repression to Polycomb repressive complex 2 (PRC2) (15). PRC2 contains the Polycomb genes Eed and Suz12 and the SET domain histone H3 methyltransferase gene Ezh2. Eed is required for PRC2-mediated trimethylation of histone H3 lysine 27 (H3K27me3) (16). Combined mutations in Tsix and Eed lead to deregulation of Xist in male ES cells, leading to activation of Xist in a majority of the cells (15). Although it appears that Tsix and PRC2 act in parallel to repress Xist, the precise function of Polycomb complexes in repressing Xist remains to be established. Notably, transient enrichment of H3K27me3 on the Xist promoter has also been proposed as one of the sequential events for Xist activation (17). However, PRC2 is generally correlated with repression of genes, and no molecular mechanism for an activating function has been identified yet. Additional indirect effects of PRC2 disruption also cannot be ruled out.

Several regulators of Xist have been identified, including the X-linked Rnf12, Ftx, and Jpx genes. Rnf12 inhibits Xist repression in part through targeting Rex1 protein for degradation (4, 18). Several transcription factors associated with ES cell pluripotency, including Oct4, Sox2, Nanog, and Rex1, have been proposed to be implicated in the repression of Xist in ES cells (3, 19, 20), but their precise function in the embryos remains to be resolved (21, 22). Recently, the activation of Xist during the progression from naive to primed pluripotency of mouse ES cells was examined in detail in chemically defined medium (5). Ftx and Jpx are ncRNA genes which are located upstream of Xist and positively regulate Xist. Jpx may function through evicting Ctcf and changing chromatin conformation (23, 24). Mutation of Ftx leads to decreased Xist expression in ES cells (25), but Ftx is dispensable for imprinted XCI in embryos (26). Furthermore, a number of studies have suggested that changes in chromatin organization and pairing of the X chromosomes along the X chromosome inactivation center (Xic) regions contributes to the regulation of XCI (27–29). Taken together, these studies illustrate that multiple factors interact in the regulation of Xist.

Here, we investigated repressive mechanisms of Xist in male ES cells, which possess a single X chromosome, and thus, trans interactions and pairing are not expected to be relevant. We show that genetic disruption of Eed and Tsix leads to loss of Xist repression despite the presence of other regulators of Xist, including Rnf12, Nanog, and Oct4. Moreover, DNA methylation and PRC2 recruitment are not essential for Xist repression as long as Tsix transcription is unperturbed. We show that Tsix transcription induces trimethylation of histone H3 lysine 36 (H3K36me3) at the Xist promoter, which contributes to the repression of Xist expression, among other mechanisms.

MATERIALS AND METHODS

Cell culture and generation of ES cell lines.

Details of plasmid construction and Setd2 knockdown are provided in the supplemental material. ES cells were cultured as previously described (30, 31). The Dnmt triple-knockout (Dnmt TKO) ES cells were a gift from Masaki Okano (Kumamoto University, Japan) (32). For generating R^−/− ΔTsix cells, the conditions for electroporation, antibiotic selection, Southern hybridization, and removing the selection cassette by transient expression of Cre recombinase were described previously (33). The expression vector pPyCAG-EGFP-IZ (34) was a gift from Hitoshi Niwa (Kumamoto University, Japan) and provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. For generating enhanced green fluorescent protein (EGFP), H3.3wt, and K36M transgenic lines, 10 μg of each expression vectors pPyCAG-EGFP-IZ, pPyCAG-H3.3wt-FH-IZ, and pPyCAG-K36M-FH-IZ was linearized by SalI, purified, and transfected into 2 million J1 ES cells with a Neon electroporator (Invitrogen, CA). The settings were as follows: 1,400 V; pulse width, 10 ms; 3 times. The cells were selected with 300 μg/ml zeocin, and a pool of approximately 100 clones of transgene-expressing cells was used for the experiments.

ChIP.

Three different chromatin immunoprecipitation (ChIP) methods were used: the H2AK119ub1-formulated method (see Fig. 4E) and the standard method without nuclear isolation (see Fig. 4D and 6C) and with nuclear isolation especially for improving the immunoprecipitation efficiency for nuclear transcription factors (see Fig. 2B and C, 3B and C, 5B, 6B, and 7C and D). ChIP for H2AK119ub1 was performed as previously described (35) with some modifications. The details of the procedures are provided in the supplemental materials and methods. The following antibodies were used: anti-H2AK119ub1 (05-678; Upstate Biotechnology, NY) and normal mouse IgM (M5909; Sigma-Aldrich, MO) as a mock-immunoprecipitated control (refered to as “mock”). A standard ChIP method without nuclear isolation was performed as previously described (36) except that a Covaris S1 system (Covaris Inc., MA) was used for DNA fragmentation, using the following settings: 20% duty, intensity of 10.0, 500 cycles/burst, 360-s duration, and 1 cycle. The following antibodies were used: anti-H3K27me3 (07-449; Upstate), anti-H3K36me3, (ab9050; Abcam, Cambridge, United Kingdom), and normal rabbit IgG (I5006; Sigma-Aldrich) as a mock control. The ChIP method with nuclear isolation was performed as previously described (37). The following antibodies were used: anti-Nanog (8822; CST, MA), anti-Oct4 (5677; CST), anti-H3K4me2 (CMA303; a gift from Hiroshi Kimura) (38), anti-H3K27me3 (CMA323; a gift from Hiroshi Kimura) (38), anti-H3K36me3 (ab9050; Abcam), and normal rabbit IgG (I5006; Sigma-Aldrich) as a mock control.

FIG 4.

Xist expression is largely repressed in R^−/− ΔTsix ES cells. (A) Xist RNA-FISH (red) in E^−/− ΔTsix and R^−/− ΔTsix cells differentiated for 4 days. Nuclei were counterstained with DAPI (blue). Bar, 10 μm. (B) Number of Xist-positive nuclei measured by Xist RNA-FISH (n > 150). (C) Xist expression (spliced product; exons 1 and 2) by qRT-PCR analysis is shown relative to undifferentiated J1:rtTA ES cells and normalized to Gapdh. (D and E) ChIP analysis of H3K27me3 (D) and H2AK119ub1 (E). Hoxa7 and Oct4 promoters were used as positive- and negative-control loci, respectively. Values are means and standard deviations from three independent experiments. Cells were cultured in ES medium with (undifferentiated) or without (differentiated) LIF. *, P < 0.05 (Student's t test).

FIG 6.

H3K36me3 is accompanied by Tsix transcription. (A) ChIP analysis of H3K36me3 in J1 cells covering the entire transcription unit of Tsix (39). (B and C) ChIP analysis of H3K36me3 in Dnmt TKO cells and its parental line, J1 (B), and J1:rtTA, Eed^−/−, E^−/− ΔTsix, Ring1b^−/−, ΔTsix, and R^−/− ΔTsix cells (C). The Sox2 gene body and its promoter were used as positive- and negative-control loci, respectively. Data are means and standard deviations from three independent experiments.

FIG 2.

Tsix prevents H3K27me3 invasion from the hot spot in naive ES cells. (A) Map of Xic. The H3K27me2/3 hot spot (43), in the 340-kb region 5′ of Xist, is shown. (B and C) ChIP analysis of H3K27me3 (B) and H3K4me2 (C) across the Xic locus. Hoxa7 and Oct4 promoters for H3K27me3 and Gapdh and H1foo promoters for H3K4me2 were used as positive- and negative-control loci, respectively. Data are means and standard deviations from three independent experiments. All cells were cultured with 2i medium. *, P < 0.05; **, P < 0.01 (Student's t test).

FIG 3.

Nanog and Oct4 enrichment is not changed in Eed^−/− ΔTsix cells. (A) Transcription factors associated with pluripotency, including Nanog, Oct4, and Sox2, are proposed as Xist repressors through a binding site within Xist intron 1. (B and C) Chromatin recruitment of the putative Xist repressors Nanog (B) and Oct4 (C) at its binding site at Xist intron 1, measured by ChIP. Klf2 promoter and Xist exon 1 for Nanog and Nanog promoter and Xist exon 1 for Oct4 were used as positive- and negative-control loci, respectively. For Nanog, data are means and standard deviations from three independent experiments. For Oct4, data from two independent experiments are shown. All cells were cultured with 2i medium.

FIG 5.

Xist expression is largely repressed in Dnmt TKO ES cells. (A) Cytosine DNA methylation was measured by bisulfite sequencing of the Xist promoter (filled circles, methylated; open circles, unmethylated). The percent methylation is given below the graphs. (B) ChIP analysis of H3K27me3 (n = 3). Hoxa7 and Oct4 promoters were used as positive- and negative-control loci, respectively. (C) qRT-PCR analysis for Xist and Tsix expression in Dnmt TKO and parental J1 ES cells. Expression relative to undifferentiated J1 ES cells was normalized to Gapdh (n = 2). (D) Xist (red) and Tsix (green) RNA-FISH in parental J1 (right) and Dnmt TKO (left) undifferentiated ES cells. Nuclei were counterstained by DAPI (blue). Bar, 10 μm. (E) Percent Tsix- or Xist-positive nuclei, revealed by RNA-FISH (n > 100).

FIG 7.

H3K36me3 is functionally involved in Xist repression. (A) Schematic representation of expression constructs. (B) Western blot analysis of histone modifications H3K36me3 and H3K27me3 in J1 cells expressing the indicated transgenes. H3 was used as a loading control, and Flag was used to confirm expression of a Flag-tagged transgene. The relative amount of histone modification, with the amount of H3.3wt set to 1 and normalized to histone H3, is given beneath the lanes in the top two panels. (C and D) ChIP analysis of H3K36me3 (C) and H3K27me3 (D) in J1 cells expressing H3.3wt and K36M transgenes. The Sox2 gene body and its promoter for H3K36me3 and the Hoxa7 and Oct4 promoters for H3K27me3 were used as positive- and negative-control loci, respectively. Data are means and standard deviations from three independent experiments. (E) qRT-PCR analysis for Xist and Tsix expression in H3.3wt-expressing (H3.3wt) and K36M-expressing (K36M) J1 ES cells. Expression relative to H3.3wt-expressing J1 ES cells normalized to Gapdh (n = 3) is shown. *, P < 0.05; **, P < 0.01 (Student's t test).

Quantitative real-time PCR.

Total RNA was purified with RNeasy minikit (Qiagen, Hilden, Germany) and cDNA was generated by reverse transcriptase SuperScript II (Invitrogen, MA). Quantitative real-time PCR for gene expression and ChIP was performed using iQ SYBR green Supermix (Bio-Rad, CA) with a single-color detection MyIQ i cycler (Bio-Rad), SYBR green PCR master mix (Life Technologies, California, USA), or Thunderbird SYBR qPCR mix (Toyoko, Osaka, Japan) with a StepOnePlus real-time PCR system (Life Technologies). Primer sequences covering the entire Xist/Tsix transcription unit (see Fig. 6A) were described previously (39), and those used for quantitative real-time PCR for both RNA expression and ChIP analyses are listed in Table S1 in the supplemental material.

RNA-FISH.

RNA fluorescence in situ hybridization (RNA-FISH) was performed according to the protocols described previously (10). The strand-specific RNA-FISH probe was generated as previously described (40).

Immunostaining and Western blotting.

Methods for immunostaining and Western blotting were performed as described previously (33). The antibody for immunostaining was Ring1b (gift from Haruhiko Koseki) (41). The antibodies for Western blotting were H3K36me3 (ab9050; Abcam), Flag (A8692; Sigma-Aldrich), H2AK119ub1 (05-678; Upstate), H3 (see Fig. S1G in the supplemental material) (ab1791; Abcam), H3 (see Fig. 4D) (39763; Active Motif), H3K27me3 (see Fig. S1G in the supplemental material) (07-449; Upstate), H3K27me3 (see Fig. 4D) (gift from Hiroshi Kimura; CMA323) (38).

Bisulfite sequencing.

DNA methylation analysis by bisulfite sequencing was performed following the manufacturer's protocol (Imprint DNA modification kit MOD50; Sigma-Aldrich) as described previously (10).

RESULTS

Eed is required for Tsix-independent Xist repression in naive ES cells.

Recently, the use of chemically defined culture conditions for investigating XCI has been explored (5). Naive pluripotent mouse ES cells (31) cultured in the presence of MAP kinase inhibitors and GSK3 kinase inhibitors (2i) have reduced promoter-specific Polycomb complex-associated histone modifications (37) and DNA methylation, which is likely to resemble the actual situation in the developing epiblast (42). We reasoned that under 2i culture conditions, the effects of chromatin on Xist repression could be discerned from indirect effects on the pluripotent state more readily. We analyzed Xist expression in naive male ES cells that harbor disruptions of Eed (Eed^−/−) and Tsix (ΔTsix) (see Fig. S1E in the supplemental material) (15) by RNA-FISH (Fig. 1A and B) and qRT-PCR (Fig. 1C). J1:rtTA ES cells were used as a parental control cell line (15). A significant derepression of Xist was observed in ΔTsix and Eed^−/− cells by semiquantitative PCR (Fig. 1C, Xist [ΔTsix, 3.47 ± 0.46; Eed^−/−, 2.93 ± 0.39; P < 0.01]). However, the number of Xist clusters observed with Xist RNA-FISH was not significantly increased in ΔTsix and Eed^−/− cells over the control cell line J1:rtTA (Fig. 1B [J1:rtTA, 3.2%; ΔTsix, 1.9%; Eed^−/−, 3.9%]). In contrast, combined disruption of both Eed and Tsix resulted in the appearance of Xist clusters in 46.2% of E^−/− ΔTsix ES cells (Fig. 1B) and high levels of Xist expression (Fig. 1C, Xist [E^−/− ΔTsix, 81.73 ± 15.33; P < 0.001]). Furthermore, the active histone mark H3K4me2 was strongly increased over the Xist promoter in Eed^−/− ΔTsix cells (Fig. 2C [E^−/− ΔTsix, 13.16 ± 0.51; P < 0.01]), in contrast to cells with mutation of either Tsix or Eed (Fig. 2C [ΔTsix, 1.40 ± 0.40; Eed^−/−, 1.26 ± 0.33; P = 0.78 and P = 0.76, respectively]). We conclude that PRC2 function is required for Tsix-independent Xist repression in naive pluripotent cells, a finding that is consistent with earlier observations in serum-LIF-cultured ES cells (15). We did not observe an increase of H3K4me2 on the Xist promoter, when Tsix was truncated in either in naive ES cells (Fig. 2C) or ES cells that were cultured in serum-LIF-based medium (15), contrasting a previous report on serum- and LIF-cultured ES cells (39). We attribute this apparent difference to the use of different ES cell lines and different strategies for mutating Tsix, where we have chosen an insertion of a gene trap cassette that truncates Tsix transcripts before the Xist gene locus.

FIG 1.

H3K27me3 is required for Tsix-independent Xist repression in naive ES cells. (A) Xist RNA-FISH (red) in J1:rtTA control, ΔTsix, Eed^−/−, and E^−/− ΔTsix ES cells (the boxed area is magnified in the rightmost panel). Nuclei were counterstained with DAPI (blue). Bar, 10 μm. (B) Percent Xist-positive cells (n > 200). (C) Map of the X chromosome inactivation center (Xic) and its linked genes Rnf12, Xpct, Cnbp2, Ftx, Jpx, Xist, and Tsix. Rnf12, Ftx, and Jpx are reported as Xist activators. Their expression was measured by qRT-PCR. Expression is shown relative to undifferentiated J1:rtTA ES cells and normalized to Gapdh (n = 3). All cells were cultured with 2i medium. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (Student's t test).

The Xist regulators Ftx and Jpx but not Rnf12 are elevated in Eed-deficient ES cells.

To investigate the mechanisms of Xist activation in Eed^−/− ΔTsix cells further, we analyzed the expression of Rnf12, Ftx, and Jpx and two additional genes within the X chromosome inactivation center (Xic), Xpct and Cnbp2. Increased Rnf12 expression is correlated with activation of Xist (18, 21) but was decreased in Eed^−/− ΔTsix cells (Fig. 1C, Rnf12 [E^−/− ΔTsix, 0.60 ± 0.05; P < 0.05]). This observation excludes upregulation of Rnf12 as a cause of Xist activation in E^−/− ΔTsix cells. The observed repression of Rnf12 could likely be a consequence of the activation of Xist in Eed^−/− ΔTsix cells. Ftx and Jpx are ncRNAs that are located upstream of Xist and have also been proposed to positively regulate Xist RNA expression (23–25). We observed increased Ftx and Jpx expression in Eed-deficient ES cells (Fig. 1C, Ftx [Eed^−/−, 1.50 ± 0.18; E^−/− ΔTsix, 1.34 ± 0.12; P < 0.05] and Jpx [Eed^−/−, 4.47 ± 0.63; Eed^−/− ΔTsix, 4.86 ± 0.17; P < 0.001]). Similarly, Xpct and Cnbp2 expression was also increased, showing that lack of Eed caused derepression of multiple genes within the Xic region (Fig. 1C, Xpct [Eed^−/−, 31.59 ± 4.75; E^−/− ΔTsix, 14.24 ± 0.97; P < 0.001] and Cnbp2 [Eed^−/−, 2.70 ± 0.04 {P < 0.001}; E^−/− ΔTsix, 1.91 ± 0.44 {P < 0.05}).

We next investigated the distribution of H3K27me3 across the Xic region (Fig. 2A). In J1:rtTA and ΔTsix cells, H3K27me3 could be detected at the Xpct, Ftx, and Jpx promoters, which are located within a 340-kb region 5′ of Xist that has been previously characterized as a hot spot region harboring high levels of H3K27me3 and H3K27me2 (Fig. 2B, Xpct, Ftx, and Jpx promoters) (43). An H3K27me3 hot spot was prominent in ES cells cultured in 2i and serum-based medium and encompassed the Xpct, Ftx, and Jpx promoters. As expected, H3K27me3 was lost in Eed^−/− and E^−/− ΔTsix cells (Fig. 2B, Xpct, Ftx, and Jpx promoters). This finding suggests that the expression of Xic-linked genes Xpct, Cnbp2, Ftx, and Jpx is repressed (Fig. 1C) by PRC2 and H3K27me3 spreading from the hot spot (Fig. 2B). Interestingly, the expression of the hot spot-linked genes Xpct and Cnbp2 was decreased in Eed^−/− ΔTsix cells compared with Eed^−/− cells (Fig. 1C), suggesting that these genes are potentially also partially repressed by ectopically expressed Xist in cells lacking Tsix and Eed. However, this repressive effect of Xist activation is small compared with the derepression observed after loss of Eed, such that a net activation is observed between wild-type control and Eed^−/− ΔTsix cells.

We next investigated whether Tsix prevented H3K27me3 spreading over the Xist promoter from the hot spot in J1:rtTA-naive ES cells. H3K27me3 was enriched at the Xpct, Ftx, and Jpx promoters (Fig. 2B) (Xpct, 1.99% ± 1.30% of input; Ftx, 2.49% ± 0.45% of input; Jpx, 1.51% ± 0.37% of input) but not at the Xist promoter (Fig. 2B) (0.41% ± 0.32% of input), suggesting a boundary of the H3K27me3 hot spot between the Jpx and Xist promoters. In ΔTsix ES cells, H3K27me3 extended over the Xist promoter and gene body (Fig. 2B [Xist promoter, 5.44- ± 1.77-fold change {P < 0.05}; Xist exon 1, 3.49- ± 1.29-fold change {P < 0.05}; Xist intron 1, 4.14- ± 0.81-fold change {P < 0.01}] [enrichment is measured relative to J1:rtTA]). This observation showed that Tsix prevented the spreading of H3K27me3 from the hot spot into the Xist locus, which is consistent with earlier results in serum and LIF cultures (44).

Nanog and Oct4 binding does not prevent Xist activation in Eed- and Tsix-deficient cells.

A number of recent studies have implicated the binding of transcription factors, including Nanog and Oct4, in Xist repression through a prominent binding site within Xist intron 1 (Fig. 3A) (19). To assess whether these pluripotency factors remain bound when Xist is activated in Eed^−/− ΔTsix cells under 2i conditions, we performed ChIP analysis. We observed that despite activation of Xist in Eed^−/− ΔTsix cells, Nanog and Oct4 binding to Xist intron 1 remained unchanged from the pattern in control cells (Fig. 3B, P = 0.54, n = 3; Fig. 3C, P = 0.19 and 0.50, n = 2). This observation that Oct4 and Nanog binding is not sufficient for repressing Xist in 2i, where Rex1 and Nanog are homogenously expressed, is further consistent with previous reports that the binding site for Oct4 and Nanog in Xist intron 1 is dispensable for Xist regulation in embryos (22). Although a deletion of the binding site slightly skews the X chromosome for inactivation in female cells (21), we conclude that loss of pluripotency factor binding to Xist intron 1 does not explain the activation of Xist in Eed^−/− ΔTsix cells.

Ring1b is dispensable for Tsix-independent Xist repression.

PRC2-catalyzed H3K27me3 can act as a signal for recruitment of PRC1, consistent with observations of corecruitment of PRC1 and PRC2 to a large number of genes (45). PRC1 contains the RING finger domain proteins Ring1a and Ring1b and catalyzes monoubiquitination of histone H2A lysine 119 (H2AK119ub1) (46), which has been proposed to repress transcription by restraining poised RNA polymerase II (35). To investigate the potential function of PRC1 in Xist repression, we disrupted Tsix in ES cells that lack Ring1b (R^−/− ΔTsix) (see Fig. S1 in the supplemental material). In contrast to Eed^−/− ΔTsix ES cells, which showed activation of Xist (Fig. 4A to C), Xist remained repressed in R^−/− ΔTsix ES cells (Fig. 4A to C). Although we could observe a few Xist clouds in R^−/− ΔTsix ES cells (Fig. 4A and B), there was no significant difference between R^−/− ΔTsix and ΔTsix ES cells (Fig. 4B). Quantitative PCR confirmed that Xist remained repressed in R^−/− ΔTsix ES cells (Fig. 4C). This observation demonstrated that Ring1b is not critical for Tsix-independent Xist repression.

ChIP analysis showed that H3K27me3 was enriched on the Xist promoter R^−/− ΔTsix cells to a level comparable to that of ΔTsix cells (Fig. 4D, Xist promoter [ΔTsix, 5.27- ± 1.59-fold enrichment compared to the mock control {P < 0.05}; R^−/− ΔTsix, 3.34- ± 0.81-fold enrichment {P < 0.05}). H2AK119ub1 was weakly enriched on the Xist promoter in ΔTsix cells and furthermore was reduced in Ring1b^−/− ES cells (Fig. 4E) (1.95- ± 0.31-fold change [P < 0.05]) and R^−/− ΔTsix cells (Fig. 4E) (2.89- ± 0.46-fold change [P < 0.05]). In Ring1b^−/− ES cells, a very small amount of H2AK119ub1 could still be observed (see Fig. S1G in the supplemental material) that was not significantly increased upon deletion of Tsix in R^−/− ΔTsix cells (Fig. 4E, Xist promoter [2.48- ± 1.65-fold change {P = 0.20} relative to the mock control]). We cannot fully rule out the possibility that Ring1a catalyzes low levels of H2AK119ub1 in the absence of Ring1b in R^−/− ΔTsix cells. However, in R^−/− ΔTsix cells, H2AK119ub1 enrichment on the Xist promoter was similar to the Oct4 promoter, which served as a negative control (Fig. 4E, compare R^−/− ΔTsix data obtained with the Xist promoter to those obtained with the Oct4 promoter; 1.19- ± 0.79-fold change [P = 0.74]). Therefore, we conclude that establishment of H3K27me3 on the Xist promoter does not lead to efficient H2AK119ub1 recruitment, consistent with earlier results of low enrichment of H2AK119ub1 at the Xist promoter in serum-cultured ES cells (15). In addition, Ring1b and H2AK119ub1 appeared to be largely dispensable for Xist repression when Tsix was disrupted. Since H3K27me3 was still present at the Xist promoter, these observations pointed toward a PRC1-independent function of PRC2 in repressing the Xist promoter. This finding is surprising but consistent with observations of PRC1-independent PRC2 recruitment at a subset of genes in ES cells (45) as well as the idea of an evolutionarily older origin of PRC2, which is present in plants where PRC1 is not conserved (46).

DNA methylation is dispensable for Xist repression by Tsix.

Taken together, our results suggested that a repression mechanism based on PRC2-mediated H3K27me3 acts on genes within the Xic region. However, this mechanism is apparently disrupted at the Xist promoter by Tsix, which prevents the spreading of H3K27me3 from the hot spot over the promoter and gene body of Xist. Since previous studies have shown that Tsix is important for DNA methylation at the Xist promoter (10, 17, 36, 39, 47, 48) and DNA methylation is inversely correlated with Polycomb recruitment (49), we investigated whether DNA methylation might explain the mechanism Tsix uses for blocking H3K27me3 spreading. DNA methylation was present on the Xist promoter in the naive ES cells and decreased after Tsix disruption in ΔTsix cells (Fig. 5A [J1, 84.6%; ΔTsix, 77.5%]). This observation confirmed that Tsix contributed to DNA methylation on the Xist promoter in naive male ES cells, similar to earlier findings in serum-cultured ES cells. To analyze the function of DNA methylation, we used ES cells harboring combined mutations of all three DNA methyltransferases Dnmt1, Dnmt3a, and Dnmt3b (32). In these Dnmt triple-knockout (TKO) cells, DNA methylation at the Xist promoter was essentially absent (0.3%) (Fig. 5A). Despite the loss of DNA methylation, we observed that H3K27me3 remained unchanged and was excluded from the Xist promoter and gene body (Fig. 5B [Dnmt TKO], Xist promoter [0.47 ± 0.05], Xist exon 1 [0.55 ± 0.04], and Xist intron 1 [0.14 ± 0.05; values are relative to input values], and (Fig. 2B [ΔTsix], Xist promoter [2.23 ± 0.73], Xist exon 1 [6.01 ± 2.23], and Xist intron 1 [1.32 ± 0.26]). We further observed that Tsix transcription was slightly increased in Dnmt TKO cells (Fig. 5C, Tsix) (1.87- ± 0.66-fold change). In Dnmt TKO cells, DNA methylation was lost (Fig. 5A) and spreading of H3K27me3 from the hot spot to the Xist promoter was blocked by Tsix (Fig. 5B), leaving the Xist promoter unaffected by these two repressive modifications. This prompted us to investigate whether, in this situation, reactivation of Xist could be observed, similar to that in Eed^−/− ΔTsix cells, in which DNA methylation at the Xist promoter was also significantly decreased (Fig. 5A, E^−/− ΔTsix [55.8%]). We observed only a moderate upregulation of Xist in Dnmt TKO cells (Fig. 5C, Xist) (3.01 ± 0.33), comparable to the increased Xist expression in single Eed^−/− and ΔTsix mutant ES cells (Fig. 1C). Furthermore, no Xist clusters were observed in Dnmt TKO cells by Xist RNA-FISH (Fig. 5D and E), demonstrating that Tsix represses Xist through additional mechanisms, when neither DNA methylation nor H3K27me3 is present at the Xist promoter.

Tsix recruits H3K36me3 to the Xist promoter.

To identify such additional mechanisms, we considered histone marks associated with coding and noncoding transcription. H3K36me3 is correlated with transcriptional elongation and repression of inappropriate initiation of transcription within the gene body (50–54). We performed ChIP to investigate whether H3K36me3 was established by Tsix transcription over the Xist gene body and promoter (39). We detected a strong enrichment of H3K36me3 over the entire Xist locus that, importantly, also included the Xist promoter (Fig. 6A). The enrichment of H3K36me3 over Xist is also observed in genome-wide data sets of serum-cultured ES cells (see Fig. S2 in the supplemental material). We observed H3K36me3 in both J1 and Dnmt TKO cells at the Xist promoter (Fig. 6B, Xist promoter [J1, 10.35% ± 0.15% of input; Dnmt TKO, 7.17% ± 0.09%]) and exon 1 (Fig. 6B [J1, 10.52% ± 2.34%; Dnmt TKO, 7.57% ± 0.69%]). H3K36me3 was not detected upstream (Fig. 6B, Jpx promoter [J1, 1.98% ± 0.24% of input; Dnmt TKO, 1.81% ± 0.16%]) or far downstream of Xist (Fig. 6B, Xist intron 1 [J1, 1.98% ± 0.24% of input; Dnmt TKO, 1.00% ± 0.06%]), where we observed levels of H3K36me3 that are comparable to the negative-control Sox2 promoter (Fig. 6B [J1, 1.27% ± 0.10% of input; Dnmt TKO, 1.24% ± 0.11%]). An enrichment of H3K36me3 on the Xist promoter was also observed in J1:rtTA, Eed^−/−, and Ring1b^−/− ES cells (Fig. 6C, Xist promoter [J1:rtTA, 8.55% ± 1.89% of input; Eed^−/−, 9.20% ± 0.59%; Ring1b^−/−, 11.09% ± 2.53%]), demonstrating that H3K36me3 is enriched at the Xist promoter and does not require PRC function. Importantly, H3K36me3 was strongly reduced to near background levels in ES cells that lack Tsix (Fig. 6C, Xist promoter [ΔTsix, 1.69% ± 0.73%; E^−/− ΔTsix, 0.99% ± 0.67%; R^−/− ΔTsix, 1.63% ± 0.31%]). Taken together, these findings show that Tsix transcription is required for establishing H3K36me3 at the Xist promoter.

H3K36me3 contributes to repression of Xist transcription.

H3K36me3 recruitment on the Xist promoter via Tsix transcription prompted us to investigate whether it could exert a repressive function on the Xist promoter in cis. In mammals, to date, at least eight histone H3 lysine K36 methyltransferases (H3K36 HKMTs) have been reported, including those encoded by the NSD1, NSD2, NSD3, SETD2, SETD3, ASH1L, SETMAR, and SMYD2 genes (55). The number of H3K36 HKMTs and their potential redundancy make it difficult to eliminate H3K36 methylation. To address this issue comprehensively and also to consider potential cell viability issues, we selected two complementary strategies. First, we aimed to disrupt H3K36 methylation through expression of a mutant histone protein, H3.3-K36M (K36M), that has a substitution of methionine for lysine at position 36. This mutant protein has been previously shown to reduce endogenous K36 methylation through binding and sequestering H3K36 HKMTs (56). We established stable transgenic J1 ES cell pools that express wild-type and mutated histone H3.3 proteins (Fig. 7A). Anti-Flag immunoblotting (Fig. 7B, Flag) and long exposure of anti-H3 (see Fig. S3 in the supplemental material) confirmed the expression of H3.3wt and H3.3-K36M. The total amount of endogenous H3K36me3 was only slightly increased by transgenic expression of H3.3wt (Fig. 7B, H3K36me3) (1.22-fold change compared to H3.3wt with EGFP cells, normalized to H3). In contrast, in K36M-expressing cells, H3K36me3 was decreased to approximately half the level of control cells (Fig. 7B, H3K36me3) (0.40-fold change for K36M compared with H3.3wt, normalized to H3). Importantly, H3K36me3 was also reduced over the Xist promoter (Fig. 7C) (0.48- ± 0.07-fold change for the K36M mutant compared with H3.3wt [P < 0.05]) and Xist exon 1 (Fig. 7C) (0.31- ± 0.07-fold change for the K36M mutant compared with H3.3wt [P < 0.01]). To test for specificity, we also measured H3K27me3, which remained unchanged globally (Fig. 7B, H3K27me3) (1.08-fold change for the K36M mutant compared with H3.3wt, normalized to H3) and locally on the Xist promoter (Fig. 7D) (1.04- ± 0.18-fold change for the K36M mutant compared with H3.3wt [P = 0.82]) and Xist exon 1 (Fig. 7D) (1.35- ± 0.56-fold change for the K36M mutant compared with H3.3wt [P = 0.34]). We did not measure an increase in the number of Xist clusters by RNA-FISH (see Fig. S4 in the supplemental material), but we did observe a significant derepression of Xist in K36M-overexpressing cells by qRT-PCR (Fig. 7E, Xist) (3.95- ± 1.49-fold change [P < 0.05]), whereas the expression of Tsix remained unchanged (Fig. 7E, Tsix) (1.15- ± 0.04-fold change [P = 0.39]). To obtain independent evidence by a second method, we aimed to reduce H3K36me3 using an RNA interference-mediated depletion of Setd2. The product of Setd2 is considered a major H3K36 methylase responsible for H3K36me3 (55). We observed a significant derepression of Xist in J1 ES cells after RNA interference-mediated depletion of Setd2 (see Fig. S5 in the supplemental material). Importantly, under our experimental conditions, the expression of Tsix remained unaffected. Taken together, our data suggest that H3K36me3 is established in a Tsix-dependent manner over the Xist promoter and contributes to a repressive effect on the Xist promoter.

DISCUSSION

In our study, we addressed the function of chromatin modifications regulating the Xist promoter. Our results suggest two independent repressive mechanisms: one is mediated by Tsix transcription, and a second and independent mechanism is mediated by PRC2. Although this is a relatively simple scenario, which accurately predicts the derepression of Xist when both repressive mechanisms are impaired in Tsix- and Eed-deficient cells, further insights into the molecular interactions are needed. Here, we dissected the functions of several known and one novel chromatin modification in this system. Combined disruption of Tsix and Eed caused activation of Xist in naive ES cells. Notably, Xist was activated, although the binding of Nanog and Oct4 in Xist intron 1 was preserved, demonstrating that Nanog and Oct4 binding is not sufficient for Xist repression. This observation is consistent with a recent report showing that the intron 1 binding site is not essential for Xist repression (21, 22). However, our data do not rule out the possibility that Oct4 and Nanog repress Xist through recruitment of PRC2. Indeed, a previous study reported a synergistic effect between the loss of the Oct4 binding site in Xist intron 1 and Tsix on Xist activation (57). Furthermore, Oct4 could possess independent modulatory functions in the repression of Xist that are not critical in male cells.

In our system, repression of Xist by Tsix required neither DNA methylation nor Polycomb recruitment, suggesting that additional mechanisms exist by which Tsix exerts its repressive function. We find that H3K36me3 is recruited to the Xist promoter in naive as well as serum-cultured ES cells. We show that H3K36me3 recruitment at the Xist promoter and gene body depends on Tsix transcription but requires neither the PRC complex proteins Ring1b and Eed nor DNA methylation. H3K36me3 is associated with transcriptional elongation and a function in repression of inappropriate transcription (50–54). To address the function of H3K36me3 in repressing Xist, we performed two experiments aiming at reducing H3K36me3 globally. The first approach is based on a mutated form of histone H3.3 that is able to sequester histone methylases specifically (56). Expression of a histone H3.3 carrying a substitution of methionine for lysine at position 36 led to a reduction of endogenous H3K36me3 to half the amount observed in control cells. Reduced H3K36me3 caused a marked derepression of Xist but had little effect on Tsix and a control gene. Our second and independent attempt at reducing H3K36me3 was based on RNA interference-mediated depletion of Setd2. Setd2 is considered the major histone methyltransferase catalyzing H3K36me3. This approach was less efficient in reducing H3K36me3 but also resulted in Xist upregulation. Taken together, our experiments support the view that H3K36me3 recruitment contributes to Xist repression. However, we were not able to eliminate H3K36me3 entirely and thus may have observed a partial effect on Xist expression. In addition, depletion of H3K36me3 might also have indirect effects through perturbation of transcription units of other genes. We did not observe Xist clusters after depletion of H3K36me3 by RNA-FISH, despite the fact that H3K27me3 and H3K36me3 were reduced at the Xist promoter, suggesting that additional mechanisms exist through which Tsix represses Xist.

In the absence of Tsix, H3K27me3 is recruited to the Xist promoter through PRC2. H3K27me3 apparently spreads from a hot spot that is located 5′ to Xist. When Eed is mutated, H3K27me3 is lost and Xist is activated. We find that loss of H3K27me3 enrichment also leads to derepression of the Ftx, Jpx, Cnbp2, and Xpct genes, suggesting that multiple genes within the Xic are repressed by PRC2. The expression of Ftx (25) and Jpx (23) is elevated in differentiating female ES cells and correlates with Xist activation. Currently, the function of Cnbp2 in XCI is unknown, and overexpression of Xpct does not lead to Xist expression (58).

Notably, H3K27me3 at the Xist promoter does not efficiently recruit H2AK119ub1. H2AK119ub1 is reduced in Ring1b-deficient cells to near background levels, but simultaneous loss of Ring1b and Tsix did not result in activation of Xist. This observation could suggest that at the Xist promoter H3K27me3 exerts a repressive function that is largely independent of H2AK119ub1. A PRC1-independent function of PRC2 is presently not regarded as a prominent silencing mechanism. However, genes have been identified in ES cells that are targets of PRC2 but not PRC1, suggesting that this mode of regulation is not entirely specific to the Xist gene promoter (45).

In cells carrying an intact Tsix gene, H3K27me3 appears to spread from a hot spot upstream of Xist over several genes within the Xic region but is prevented from entering the Xist locus. We aimed to clarify whether known chromatin modifications can explain how Tsix prevents the spreading of H3K27me3. Previous reports have shown that DNA methylation is recruited by Tsix to the Xist promoter (10, 17, 36, 39, 47, 48). Notably, DNA methylation and H3K27me3 are inversely correlated in undifferentiated ES cells (49). Using ES cells that lack DNA methylation, we demonstrated that DNA methylation is not required for restricting the spreading of H3K27me3. H3K27me3 also remained excluded from the Xist promoter after depletion of H3K36me3. This observation suggests that H3K36me3 is not critical for restricting spreading of H3K27me3 over the Xist promoter. However, we cannot rule out the possibility that we have not been able to reduce H3K36me3 to low enough levels. Blocking of H3K27me3 spreading by Tsix might involve functional redundancies between H3K36me3 and DNA methylation. Alternatively, additional, as-yet-unidentified mechanisms might exist.

Our data implicate H3K36me3 as a chromatin modification that is established cotranscriptionally by Tsix over the Xist promoter and contributes to Xist repression. This modification can now be considered in future analyses of Xist regulation and will facilitate progress in understanding the chromatin-based mechanisms that contribute to the initiation of XCI. It will also be interesting to see if H3K36me3 is also relevant for gene regulation by noncoding transcription at other gene loci.