Generation of ES cells lacking Eed

To directly investigate the function of the PRC2 complex in the recruitment of PRC1 at the initiation of X inactivation, we disrupted the Eed gene by targeting in clone 36 and ΔSX ES cells. The targeting vector replaced sequences encoding the first and second WD40 domains of the Eed protein with a stop cassette, which terminates transcription resulting in a null allele (Supplementary Figure 1A). After removal of the selection cassette from targeted ES clones by Cre-recombinase mediated deletion, the second allele of Eed was targeted using the same strategy. This yielded the cell lines 36^Eed−/− (clone 1 and 2) and ΔSX^Eed−/−, derivatives of clone 36 and ΔSX ES cells, respectively. Northern analysis confirmed the absence of wild-type Eed transcripts in these cells (Figure 2A). Two truncated Eed RNA species were observed in Eed^−/− ES cells consistent with the termination of transcription at the introduced stop cassette. Western analysis revealed that Eed protein was absent in the Eed^−/− cell lines (Figure 2B), while in control clone 36 and ΔSX ES cell lines the four Eed isoforms were resolved. We further reconstituted Eed expression in 36^Eed−/− (clone 2) ES cells by introducing a transgene expressing an amino terminal fusion of the enhanced green fluorescent protein (EGFP) with the short Eed isoform. In these 36^EedTG ES cells, we observed one protein migrating with the expected molecular weight of the EGFP-Eed fusion protein and a faster migrating product likely due to proteolysis (Figure 2B). In Eed deficient ES cells, Suz12 RNA levels were reduced whereas steady-state levels of the Ezh2 transcripts remained unchanged compared to control cell lines (Figure 2A). Western analysis revealed that Ezh2 is drastically reduced below detection limit and Suz12 was found in reduced amount in Eed^−/− cells (Figure 2B). In 36^EedTG cells, Ezh2 and Suz12 RNA and protein levels were rescued confirming that the effect was specific and caused by the lack of Eed (Figure 2B, and data not shown).

Open in a separate window

Figure 2

Generation of ES cells lacking Eed. (A) Northern analysis of Eed, Suz12 and Ezh2 in undifferentiated control clone 36 and Eed deficient 36Eed^-/- ES cells after Xist was induced for 3 days (+) or not (−); Gapdh as loading control. (B) Western analysis of Eed, Ezh2 and Suz12 in nuclear extracts from uninduced ES cells (−) or induced for 3 days (+). hnRNP A as loading control, asterisk indicates a nonspecific band. (C–E) Indirect immunofluorescence (green) of Eed (C), Ezh2 (D) or Suz12 (E) and subsequent Xist RNA FISH (red) of representative nuclei of undifferentiated 36^Eed−/− and control clone 36 ES cells after 3 days of Xist induction. DAPI (blue) stains DNA. Statistics of the number of nuclei showing colocalisation of Suz12 staining with Xist in 36 and 36^Eed−/− ES cells. Error bars indicate standard deviation (n>600).

Eed deficient 36^Eed−/− and ΔSX^Eed−/− ES cells showed a reduced ability to form colonies compared to control 36 and ΔSX ES cells but proliferation and self-renewal of ES cells was largely independent of Eed (Supplementary Figure 1B and C). Furthermore, the plating efficiency is rescued in 36^EedTG ES cells showing that the defect is specific and due to lack of Eed. Eed^−/− ES cells could be induced to differentiate with retinoic acid, but showed a reduced developmental potential indicated by the formation of irregular shaped embroid bodies and the absence of contractile structures indicative of cardiomyocytes in embroid body outgrowths (Supplementary Figure 1D and E, and data not shown).

Xist recruits Suz12 independent of functional PRC2

Western analysis of Eed deficient ES cells revealed reduced Suz12 protein levels compared to control 36 ES cells and a loss of Ezh2 protein (Figure 2B). This was verified by combined immunofluorescence Xist RNA fluorescence in situ hybridisation (FISH) analysis on ES cells, after 3 days of Xist induction with doxycycline. In control clone 36 ES cells, 89, 79 and 88% of cells showed colocalisation of Xist RNA with Eed, Ezh2 and Suz12, respectively (Figure 2C–E; Table I). In Eed deficient 36^Eed−/− ES cells Xist RNA showed normal localisation and no signal for Eed and Ezh2 was detected consistent with the loss of these proteins (Figure 2C and D). The Suz12 signal was markedly decreased in Eed deficient cells. However, we observed colocalisation of Suz12 with Xist RNA in 13 and 7% of 36^Eed−/− clone 1 and clone 2 ES cells, respectively (Figure 2E). This demonstrates that recruitment of Suz12 by Xist RNA can occur, at least in part, independent of Ezh2 and Eed suggesting a role for Suz12 in PRC2 recruitment in X inactivation.

Xist recruits PRC1 independent of Eed and H3K27me3 in ES cells

To study the chromosomal marks at the initiation of X inactivation in Eed deficient ES cells, we performed combined Xist RNA FISH immunofluorescence analysis on 36^Eed−/− and control ES cells (Figure 3 and Table I). After Xist induction for 3 days, we observed a strong focal H3K27me3 staining colocalising with Xist RNA in clone 36 ES cells. However, in 36^Eed−/− ES cells di- and tri-methylation of H3K27 were drastically reduced and no colocalisation with Xist was observed consistent with a loss of PRC2 function in these cells (Figure 3B and C). A faint H3K27me3 signal was still observed at pericentric heterochromatin possibly due to weak cross-reactivity of the antibody with H4K20me3 (Peters et al, 2003). We detected a robust H3K27me1 signal at pericentric heterochromatin in 36^Eed−/− cells comparable to controls (Figure 3A). In 36^EedTG ES cells transgenic expression of EGFP-Eed rescued H3K27me3 (Supplementary Figure 1F).

Open in a separate window

Figure 3

Histone modifications in Eed deficient ES cells. (A–E) Combined Xist RNA FISH (red) indirect immunofluorescence of indicated histone modifications (green) analysis on undifferentiated 36^Eed−/− and control 36 ES cells after 3 days Xist expression. Representative images are shown, statistics see Table I. (F) Western analysis of mono-, di- and tri-methylation of H3K27 in 36^Eed−/− clones 1 and 2 and control 36 ES cells after Xist induction for 3 days (+) or not (−); loading control hnRNP A. (G) Mass-spectrometric analysis of histone H3 lysine 27 methylation in clone 36, 36^Eed−/− and 36^EedTG ES cells. The percentage of the indicated modification state is given for three independent experiments; error bars indicate standard deviation.

H2AK119ub1 and H4K20me1 are two marks associated with the initiation of X inactivation. H2AK119ub1 colocalised with Xist RNA in 97 and 98% of clone 36 and 36^Eed−/− ES cells, respectively (Figure 3E). A robust H4K20me1 signal colocalising with Xist RNA was detectable in 82% control clone 36 ES cells. In 36^Eed−/− ES cells, the H4K20me1 signal appeared less intense and was detected in 50% (clone 1) and 36% (clone 2) cells (Figure 3D). We conclude that ubiquitination of H2A on lysine 119 is independent of Eed, but PRC2 function supports the establishment of H4K20me1 by Xist (Table I).

We observed normal H2A ubiquitination upon Xist expression in Eed deficient cells, which in ES cells is thought to be mediated by Ring1b, a core component of PRC1 (Figure 3E). To assess if PRC1 was indeed recruited by Xist independent of PRC2, we performed immunofluorescence analysis using antisera specific for the PRC1 core components Ring1b, Mph1 and Mph2. Colocalisation of Ring1b with Xist RNA was observed in ES cells independent of Eed (Figure 4A). The Mph1 signal colocalised with Xist in 48% of control 36 ES cells, but no colocalisation was observed in 36^Eed−/− ES cells (Figure 4B). Colocalisation of Mph2 with Xist RNA was observed only in differentiated cells (Figure 4C), and was detected in 33% of clone 36 but not in Eed deficient 36^Eed−/− ES cells on day 8 of differentiation. We conclude that recruitment of Mph1 and Mph2 by Xist is dependent on PRC2 function, but Ring1b is recruited independently of PRC2, Mph1 and Mph2. Despite the lack of detectable Mph1 and Mph2 recruitment, the Ring1b protein is enzymatically active as shown by ubiquitination of H2A.

Open in a separate window

Figure 4

Recruitment of PRC1 components in the absence of Eed. (A, B) Indirect immunofluorescence (IF) of Ring1b (A), Mph1 (B) and subsequent Xist RNA FISH (red) analysis on undifferentiated 36^Eed−/− and control clone 36 ES cells after Xist expression for 3 days. (C) Analysis for Mph2 in ES cells differentiated for 8 days in the presence of doxycycline. The percentage of nuclei showing focal IF staining colocalising with Xist RNA is given for undifferentiated (ES), day 3 (DD3) and day 8 (DD8) of differentiation. Error bars represent standard deviation (n>350).

PRC2 is critical for H3K27me2 and H3K27me3 in ES cells

To assess if disruption of Eed in 36^Eed−/− and ΔSX^Eed−/− ES cells indeed caused a loss of PRC2 function, we performed an analysis of histone modifications. By Western analysis H3K27me2 and H3K27me3 were lost in 36^Eed−/− and ΔSX^Eed−/− ES cells, but we found mono-methylation of H3K27 only slightly reduced consistent with our immunofluorescence data (Figure 3F). The mono-, di- and tri-methylation states of histone H3 lysine 9 or of H4 lysine 20, and ubiquitination of histone H2A lysine 119 were not altered in Eed deficient ES cells (data not shown).

To further quantify the histone methylation marks, we performed a mass spectrometric analysis of nuclear extracts prepared from undifferentiated 36^Eed−/− and ΔSX^Eed−/− and control ES cells. In control 36 ES cells, 17% of bulk histone H3 were mono-methylated, 58% di-methylated and 14% tri-methylated on lysine 27 (Figure 3G) consistent with previous reports (Peters et al, 2003). In Eed deficient ES cells, H3K27me3 and H3K27me2 were dramatically reduced compared to controls, but only a moderate reduction in the H3K27me1 signal was observed (Figure 3G). The loss of H3K27 di- and tri-methylation in Eed deficient ES cells resulted in a concomitant increase in unmodified but not mono-methylated H3K27. Di- and tri-methylation of H3K27 was restored in 36^EedTG ES cells to 42 and 7%, corresponding to 72 and 50% of wild-type levels, respectively (Figure 3G). The methylation levels of H3K9 or H4K20 were unchanged by the absence of Eed (Supplementary Figures 2B and 3). However, H3K36me2 levels were significantly reduced from 50% in control clone 36 ES cells to 34% in 36^Eed−/− ES cells, and 31% in ΔSX^Eed−/− ES cells (Supplementary Figures 2A and 5). Restoration of H3K36me2 levels in 36^EedTG to 43% demonstrated that the PRC2 complex regulates global H3K36me2 marks.

We conclude that in ES cells PRC2 is crucial for H3K27 di- and tri-methylation, but has no detectable contribution to H3K9 methylation. The H3K27me1 mark on pericentric heterochromatin was unaffected in Eed deficient ES cells (Figure 3A).

Initiation of silencing by Xist is independent of PRC2

In clone 36 ES cells inducible Xist expression causes reversible silencing of a puromycin resistance gene, which was cointegrated with the Xist cDNA transgene on chromosome 11 (Wutz and Jaenisch, 2000). To establish whether Eed is required for initiation of silencing, we induced Xist expression in 36^Eed−/− ES cells for 3 days and analysed puromycin resistance gene expression by Northern (Figure 5A). Silencing was equally efficient in control 36, 36^EedTG and Eed deficient ES cells, demonstrating that Eed and H3K27me3 are dispensable for initiation of silencing by Xist.

Open in a separate window

Figure 5

Initiation and maintenance of silencing independent of Eed. (A) Northern analysis of PGKpuromycin, (puro) silencing in 36^Eed−/− and control clone 36 cells after Xist induction for 24, 48 and 72 h; Gapdh as loading control. (B) Maintenance of puro silencing in cells differentiated in the presence (+; lanes 2, 5 and 8) or absence (−; 1, 4 and 9) doxycycline, or differentiated for 4 days in the presence followed by 4 days in the absence of doxycycline (lanes 3, 6 and 7). (C) Northern analysis of puro expression in embryoid bodies outgrowths established in the presence of doxycycline (+) or without (−) after 4 weeks. (D) Quantitative expression analysis of Cct4, Npm1, Igf2b and Tk1 on chromosome 11 in control 36 and 36Eed^-/− ES cells at day 8 of differentiation in the absence (red bars), continuous presence (blue bars) of doxycycline, or presence of doxycycline for the first 4 days (green bars). Means of three independent measurements normalised to Gapdh are shown, error bars represent standard deviation. Scheme on the left shows the genes relative to the Xist transgene.

To investigate the role of Eed for the maintenance of silencing, we induced Xist in differentiating 36^Eed−/− and control 36 ES cells. In retinoic acid differentiated cells in the presence of doxycycline for 8 days or for 4 days followed by 4 days without Xist induction, we observed efficient maintenance of silencing of the puromycin gene compared to cultures, in which Xist had not been induced (Figure 5B). Notably, there was no difference between Eed deficient 36^Eed−/− and control 36 ES cells demonstrating that the shift from reversible to irreversible gene silencing had occurred in the absence of PRC2 function. To test the function of Eed for the maintenance of silencing in a more physiological differentiation model, we established embryoid body outgrowth cultures from Eed deficient 36^Eed−/− and control 36 ES cells in the presence or absence of doxycycline and measured expression of the puromycine marker gene after 4 weeks of differentiation. Northern analysis revealed that silencing was maintained in the absence of Eed (Figure 5C). Finally to establish that long-range silencing was maintained as opposed to merely silencing of the marker gene in proximity of the Xist transgene, we established differentiated cultures by induction with retinoic acid and measured expression of genes on chromosome 11 by quantitative PCR analysis. The Npm1 gene, which is located 20 Mb from the transgene integration site, was repressed by Xist on day 8 of differentiation in 36^Eed−/− as well as in control cells to approximately 50%. Npm1 repression was also maintained in cells that were differentiated in the presence of doxycycline for 4 days followed by 4 days without (Figure 5D). The repression of three other genes Cct4, Igf2b1 and Tk1 in differentiated cells showed the same trend but was more variable, probably because of heterogeneous regulation in the differentiated cultures. We conclude that Eed is not required for the initiation of Xist mediated silencing, and that PRC2 function and H3K27me3 are dispensable for the maintenance of long-range silencing.

H2AK119ub1 activity of PRC1 does not require Mph1 or Mph2

Using Eed-deficient ES cells we show that the recruitment of the PRC1 core proteins Mph1 and Mph2 by Xist is dependent on Eed. This observation is consistent with the idea that PRC2 has a recruitment function for PRC1 components. A fly PRC1 core complex has been reconstituted containing the four components Psc, Pc, Ph and dRing (Francis et al, 2001). A similar composition has been proposed for mammalian PRC1 like complexes based on purification and reconstitution experiments (Lavigne et al, 2004; Wang et al, 2004). However, we observe in Eed deficient ES cells that Ring1b is not only recruited by Xist in the absence of the PRC1 core components Mph1 and Mph2 but also appears to be functional as demonstrated by the concomitant ubiquitination of lysine 119 on histone H2A. Thus, in X inactivation Ring1b is either functional alone and can be recruited independently of other PRC1 members by Xist, or is part of a distinct complex that is yet to be identified (Ogawa et al, 2002; Dou et al, 2005; Isono et al, 2005b).

Based on our data we therefore propose two mechanisms for recruitment of PRC1 function by Xist in X inactivation (Figure 6b). A PRC2 dependent mode involves the binding of the Polycomb chromodomain to the H3K27me3 mark and operates via Mph1 or Mph2. This is predicted from biochemical evidence that H3K27me3 acts as a affinity signal recognised by the chromodomain of mammalian homologues of Polycomb (Fischle et al, 2003; Min et al, 2003). Our data provide evidence for a second mode of recruitment for PRC1 function. In the absence of PRC2 function, Xist can recruit Ring1b independent of the PRC1 core proteins Mph1 or Mph2. Both recruitment modes for PRC1 by Xist act synergistically to mediate H2AK119ub1 in the initiation of X inactivation.

Ring1b and PRC2 are regulated by a chromosomal memory

Establishment of H2AK119ub1 is restricted to an early time window in ES cell differentiation such that little H2AK119ub1 is imposed if Xist is induced at late time points in ES cell differentiation. Thus, Xist expression during an early window in differentiation establishes a chromosomal memory that in differentiated cells is required for H2AK119ub1. This memory is established at the time when X inactivation becomes irreversible and is stably maintained independent of Xist expression. To date, the molecular nature of this chromosomal memory is unknown. Our data demonstrate that the memory regulating the imposition of H2AK119ub1 and H3K27me3 in differentiated cells is independent of silencing. Recruitment of both PRC1 and PRC2 is also dependent on Xist RNA localisation and both histone marks are lost from the chromosome when Xist is turned off. Hence, H2AK119ub1 and H3K27me3 are reversible modifications and depend on Xist expression. We conclude that PRC1 and PRC2 are not stably maintained on the Xi throughout ES cell differentiation and can be excluded as integral components of the memory. Yet, the recruitment of PcG proteins early in X inactivation is consistent with a role in the establishment of a special chromatin structure that functions as chromosomal memory. Importantly, our data demonstrate that once established this chromatin structure is self-perpetuating and stable in the absence of Xist, PRC1 and PRC2. We show that a chromosomal memory regulating H2AK119ub1 is established independent of PRC2, Mph1 and Mph2.

Immunostaining and RNA FISH

ES cells were attached to poly-l-lysine coated coverslips or cytospun using a Cytospin 3 centrifuge (Thermo Shandon, USA). Differentiated cells were grown on Roboz slides (CellPoint Scientific, USA). Immunostaining was performed as described (Peters et al, 2003; Kohlmaier et al, 2004). Briefly, cells were fixed for 10 min at RT in 4% PFA in PBS, permeabilised for 5 min at RT in 0.1% Na Citrate/0.1% Triton X-100, blocked for 60 min at RT in PBS containing 5% (wt/vol) BSA, 0.1% Tween-20. For H2AK119ub1 immunostaining cells were pre-extracted in 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 10 mM Pipes pH 6.8 and 0.5% Triton for 2 min at RT before fixation.

RNA FISH probes were generated by random priming (Stratagene, USA) using Cy3-dCTP (Amersham). After immunostaining, cells were fixed in 4% PFA in PBS for 10 min at 4°C, de-hydrated, hybridised and washed as described (Wutz and Jaenisch, 2000). Images were obtained using a fluorescence microscope (Zeiss Axioplan) equipped with a CCD camera and the MetaMorph image analysis software (Universal Imaging, USA).

RNA and protein analysis

Northern analysis was performed using 20 μg of RNA (Trizol; Invitrogen) as described previously (Wutz and Jaenisch, 2000). Antibodies for histone lysine methylation states and Western analysis were previously described (Peters et al, 2003; Kohlmaier et al, 2004) and the following dilutions were used (immunostaining/Western blot): α-H3K9m1 (#4858, 1:1000/1:500); α-H3K9m2 (#4677, 1:1000/1:1000); α-H3K9m3 (#4861, 1:750/1:1000); α-H3K27m1 (#8835, 1:6,000/1:1000); α-H3K27m2 (#8841, 1:1000/1:2,000); α-H3K27m3 (#6523, 1:1000/1:7,000); α-H4K20m1 (#0077, 1:500/1:3,000); α-H4K20m2 (#0080, 1:1000/1:1000); α-H4K20m3 (#0083, 1:3,000/1:3,000). Additional antibodies were as follows: α-H2AK119ub1 (α-ubiquityl-Histone H2A, clone E6C5; #05-678 Upstate Biotechnology, Lake Placid, New York, USA), 1:50/1:400; α-Suz12 (# 07-379; Upstate), 1:1000/1:1000; α-Eed (rabbit polyclonal antiserum, AKS and AW, unpublished results); 1:1000 α-Ezh2 (rabbit polyclonal antiserum, M Busslinger unpublished results); 1:1000/1:1000); α-Ring1b (Atsuta et al, 2001), 1:100 for IF; α-Mph1 (Isono et al, 2005a), 1:5 for IF; α-Mph2 (Isono et al, 2005a), 1:100 for IF; α-hnRNP A1 (4B10 mouse monoclonal antiserum), 1:1000 for Western. Secondary antibodies: Alexa A-11034 Fluor 488 goat anti-rabbit IgG (H+L) and Alexa A-11034 Fluor 488 goat anti-mouse IgG (H+L) all at 1:500 (Molecular Probes, USA); HRP-conjugated AffiniPure goat anti-rabbit IgG (H+L), 1:10 000 and HRP-conjugated AffiniPure goat anti-mouse IgG (H+L), 1:5000 (Jackson ImmunoResearch Laboratories. Inc., USA).

Quantitative PCR expression analysis

Random primed cDNA was generated from 10 μg total RNA from clone 36 and Eed^−/− ES cells using the Superscript II Reverse transcription kit (Invitrogen). Quantitative PCR using the Taqman method (Applied Biosystems) for Tk1 (primers:

GCAACAGCTTCTCCACACATGA, GCGGAGCATGCAGGCT;

probe: CGGAACACCATGGACGCATTGC),

Npm1 (TGTAGAGGAAGATGCAGAGTCTGAA, CCTCCAGGAGCAGATCGCT;

AGGAGGACGTAAAACTCTTAGGCATGTC),

Igfbp2 (CGGCAAAGGCGGCAA, TGGCACTACCACCTCAGCTG;

ACGGTGAATGAGCTGCAGAACTTGACC),

Cct4 (CCTACCAGGACCGCGACA, GCTTTGGCCGCGGAA;

CCAGCCCAGATCCGCTTCAGCAAT) and

Gapdh (CATGGCCTTCCGTGTTCCTA, TGTCATCATACTTGGCAGGTTTCT;

TCGTGGATCTGACGTGCCGCC)

on a ABI PRISM 7000 detection machine was performed in triplicate as described (Pauler et al, 2005). Quantification was achieved by the standard curve method using serial dilutions of cDNA generated from uninduced ES cells at day 8 of differentiation. Samples were normalised to Gapdh and the expression levels of uninduced clone 36 ES cells at day 8 of differentiation were set to 100 for each gene.

We thank Gideon Dreyfuss, Arie Otte and Meinrad Busslinger for kindly providing antibodies, and Leonie Ringrose for critically reading the manuscript. This research is supported by the IMP through Boehringer Ingelheim and by grants from the GEN-AU initiative of the Austrian Ministry of Education, Science and Culture, and the European Union 6th framework program Epigenome Network of Excellence.