ERG-associated protein with SET domain (ESET)-Oct4 interaction regulates pluripotency and represses the trophectoderm lineage

Background Pluripotency, the capacity for indefinite self-renewal and differentiation into diverse cell types is a unique state exhibited by embryonic stem (ES) cells. Transcriptional regulators, such as Oct4, are critical for pluripotency, but the role of epigenetic modifiers remains to be fully elucidated. Results Here, we show that ERG-associated protein with SET domain (ESET), a histone methyltransferase enzyme, maintains pluripotency through repression of Cdx2, a key trophectoderm determinant, by histone H3 lysine 9 trimethylation (H3K9me3) of the promoter region. Notably, this repression is mediated through the synergistic function of small ubiquitin-related modifier (SUMO)ylated ESET and Oct4. ESET localises to the promyelocytic leukaemia (PML) nuclear bodies and is SUMOylated in ES cells. Interaction of ESET with Oct4 depends on a SUMO-interacting motif (SIM) in Oct4, which is critical for the repression of Cdx2. Conclusion Loss of ESET or Oct4 results in strikingly similar phenotypes both in ES cells with their differentiation into trophectoderm cells, and in early embryos where there is a failure of development of the pluripotent inner cell mass (ICM) of blastocysts. We propose that SUMOylated ESET-Oct4 complex is critical for both the initiation and maintenance of pluripotency through repression of differentiation, particularly of the trophectoderm lineage by epigenetic silencing of Cdx2.


Background
ERG-associated protein with SET domain (ESET), also known as SET domain bifurcated 1 (SETDB1), is a histone methyltransferase that catalyses a repressive mark on euchromatin by mediating histone 3 lysine 9 trimethylation (H3K9me3) [1,2]. The ESET protein contains a Tudor domain, a methyl-CpG binding domain and a bifurcated SET domain that is responsible for its catalytic activity [3]. Proteins that associate with ESET are mainly corepressors [1, [4][5][6][7], in agreement with its repressive role. ESET is critical for very early development since the Eset-null embryos die at the peri-implantation stage with defective development of the inner cell mass (ICM), from which no embryonic stem (ES) cells could be derived [8]. The Eset-null phenotype is similar to that of Oct4-null embryos [9] but the basis for this is unknown. We decided to address this question by investigating the role of ESET in the ICMderived pluripotent ES cells.

ESET-depleted ES cells differentiate towards the trophectoderm lineage
To investigate the role of ESET in ES cells, we induced knockdown of Eset using a short hairpin RNA (shRNA)expressing vector that contains an enhanced green fluorescent protein (EGFP) reporter (Figure 1a). An empty vector was used as a control. At 7 days following transfection of Eset shRNA, we observed a dramatic loss of pluripotent ES cells as judged by a marked reduction in the number of alkaline phosphatase positive cells (Figure 1b, Additional file 1). To visualise the morphology of the knockdown cells, EGFP-positive cells were purified by fluorescent activated cell sorting (FACS), 24 h after transfection and cultured in ES medium for 2 to 4 days (Figure 1c, Additional file 1). Notably, ES cells transfected with Eset shRNA appeared flat and differentiated, which was clearly distinct from control ES cells, which formed compact colonies ( Figure 1c). This experiment demonstrates that ESET is important for the maintenance of undifferentiated ES phenotype.
Next, 5 days after transfection of Eset shRNA, we carried out reverse transcription quantitative polymerase chain reaction (Q-PCR) analysis to determine the precise consequences of Eset knockdown. Concomitant with the change in the phenotype in ESET-depleted ES cells, we found downregulation of pluripotency-specific genes, including Oct4, Nanog and Sox2. At the same time, we detected upregulation of genes associated with differentiation, including, Cdx2, Hand1, Dlx3, Ets2, Fgf5 and Gata6 ( Figure 1d). Notably, this response is similar to that triggered by the knockdown of Oct4 in ES cells, where genes associated with the trophectoderm lineage, such as Cdx2 and the primitive endoderm lineage, such as Gata6 are upregulated [10]. Thus, apart from the failure of ICM development and pluripotent cells in early embryos following the loss of either Eset [8] or Oct4 [9], the response of ES cells to the loss of these two genes is also comparable.
The upregulation of trophectoderm-specific genes is of particular interest as mouse ES cells have a low propensity to differentiate into trophectoderm cells [11]. This is, at least in part, because Oct4 and Cdx2 induce reciprocal inhibition to achieve a mutually exclusive expression in the pluripotent and the trophectoderm lineages, respectively [12]. Since Eset knockdown evidently exhibits a similar outcome as Oct4 knockdown in ES cells, we investigated whether ESET, like Oct4, might also play a role in repressing trophectoderm differentiation. We tested this possibility by investigating whether ESETdepleted ES cells could give rise to trophectoderm cells. To do this, we cultured ESET-depleted ES cells in a medium that is conducive for development of trophectoderm cells (henceforth called TS medium) [13]. Notably, Cdx2-positive cells were observed in about 40% of Eset knockdown cells after 4 days of culture in TS medium (Figure 1e). By contrast, the majority of EGFP-negative cells cultured in the TS medium, which represents a concurrent control, were negative for Cdx2, except for a few cells that are occasionally positive (less than 1% of total cells), which arise spontaneously when ES cells are exposed to TS medium. The heterogeneity of expression of Cdx2 in ESET-depleted ES cells suggests that, apart from the trophectoderm cells, these ESET-depleted cells may differentiate into other lineages. Nevertheless, ESET-depleted ES cells clearly have a propensity to differentiate towards the trophectoderm lineage, a response that is uncharacteristic of normally differentiating ES cells. To gain mechanistic insight on the role of Eset in regulating pluripotency of ES cells, we decided to focus our analysis on Cdx2, since this is an early and a key determinant of the trophectoderm lineage [14].

ESET-mediated H3K9me3 represses Cdx2 in ES cells
Since the loss of Eset from ES cells results in the upregulation of Cdx2 at the mRNA and protein levels, we reasoned that Cdx2 might be repressed by ESET-mediated epigenetic modification at its promoter region via H3K9me3.
To test for this possibility, we first performed chromatin immunoprecipitation (ChIP) for H3K9me3 at the Cdx2 promoter region in wild-type ES cells. We detected H3K9me3 marks that were spread throughout the 5 kb region surrounding the transcription start site of Cdx2 gene (Figure 2a, b, Additional file 2). In contrast, H3K9me3 levels were not enriched at the promoter region of Oct4 gene, which unlike Cdx2 is transcriptionally active in ES cells (Figure 2a, b).
To investigate whether ESET is the histone methyltransferase responsible for H3K9me3 modification at the Cdx2 promoter region, we performed ChIP analysis on ES cells. H3K9me3 was detected at the Cdx2 promoter region, but this signal was downregulated in ESET-depleted cells by an average of 40% to 50% compared to control cells (Figure 2c, Additional file 2). This observation was confirmed by Q-PCR analysis using primers C1 and C2 ( Figure 2d). As an additional control, we also analysed the H3K9me3 at the major satellite region that is known to be mediated by Suv39h1/2, a histone methyltransferase that governs H3K9me3 at the pericentric heterochromatin region [15,16]. In contrast to the observations on the Cdx2 promoter region, H3K9me3 levels at the major satellite region were unaffected in ESET-depleted ES cells ( Figure  2c, Additional file 2). These results suggest that the H3K9me3 marks at the Cdx2 promoter region were most likely mediated by ESET.
To further confirm that ESET mediates H3K9me3 at the Cdx2 promoter region in ES cells, we performed ChIP using an anti-ESET antibody. We detected enrichment of ESET on Cdx2 promoter with a peak at the region immediately upstream of the transcription start site, which was amplified by primer C7 (Figure 2e). Collectively, these results demonstrate that ESET-mediated H3K9me3 represses Cdx2 in ES cells.

Oct4 and ESET synergistically repress Cdx2
Next, we considered whether Oct4 and ESET act synergistically to repress Cdx2 since loss of function of Eset and Oct4 generate similar phenotypes, both during early embryonic development [8,9], and in ES cells, as described above. First, we confirmed that Oct4 is enriched on Cdx2 promoter in wild-type ES cells (Figure 3a) as reported previously [17]. Notably, Oct4 is enriched at the same region that is bound by ESET (Figure 2e), suggesting that both factors might coregulate Cdx2. To address this possibility, we used ZHBTc4 ES cells that contain a tetracycline regulatable Oct4 allele [18]. Depletion of Oct4 from these ES cells following addition of tetracycline to the culture medium results in the induction of Cdx2 after 2 days (Figure 3b). To analyse the effect of Oct4 depletion on the binding of ESET to Cdx2 promoter, we performed ChIP analysis on ZHBTc4 ES cells at day 1 of tetracycline treatment when ESET expression was not yet affected by the depletion of Oct4 ( Figure 3b). As expected, Oct4 binding on the Cdx2 promoter was downregulated by 90% in Oct4-depleted ES cells ( Figure 3c). Notably, this was accompanied by downregulation of ESET binding on the Cdx2 promoter by 30% (Figure 3c). Furthermore, downregulation of ESET binding was coupled with the decline of H3K9me3 levels at the Cdx2 promoter region ( Figure  3d). We further verified that the binding of ESET to Cdx2 promoter is dependent on Oct4 by performing ChIP experiments in ZHBTc4 ES cells overexpressing haemagglutinin (HA)-tagged ESET (Additional file 3). In this experiment, tetracycline was first added to ZHBTc4 cells to initiate loss of Oct4, followed 1 day later by transfection of HA-tagged ESET, followed by a ChIP experiment 1 day further on using an anti-HA antibody. The results show that ESET binding to Cdx2 promoter decreased by 40% confirming that the loss of Oct4 affects the binding of ESET to Cdx2 promoter. Taken together, these results suggest that ESET and Oct4 likely cooperate to repress Cdx2, which is consistent with the observation that depletion of either ESET or Oct4 leads to upregulation of Cdx2.
Since Cdx2 is a known target of Oct4 [12,17,19], it is possible that Oct4 may interact with ESET to repress Cdx2. To test for this possibility, we coexpressed Oct4 and ESET in 293T cells and found that ESET coimmunoprecipitated with Oct4 (Additional file 3). To further investigate which domain of ESET interacts with Oct4, we generated two deletion constructs; HA-ESET-ΔSET and HA-ESET-ΔTudor (Additional file 3). Interestingly, when we coexpressed either HA-ESET-ΔSET or HA-ESET-ΔTudor with Oct4, both mutants were able to coimmunoprecipitate with Oct4 (Additional file 3), suggesting that ESET and Oct4 form a complex through bridging proteins, or that ESET associates with Oct4 through at least two different sites. We also performed immunoprecipitation of endogenous ESET in ES cells using an anti-ESET antibody and found that Oct4 was coimmunoprecipitated independently of DNA, albeit at a low level and only when immunoprecipitation was carried out under mild conditions ( Figure 3e and Additional file 4). However, in the reciprocal immunoprecipitation experiment using an anti-Oct4 antibody, we noticed that immunoprecipitation of endogenous Oct4 not only precipitated the expected ESET protein of 180 kDa, but also ESET proteins of higher molecular weight, which may be post-translationally modified ESET proteins (Figure 4e, lane 7, top panel and Additional file 4). We therefore sought to investigate these modified ESET proteins.

ESET localises to promyelocytic leukaemia (PML) nuclear bodies and is post-translationally modified by small ubiquitin-related modifier (SUMO)
We had noted that ESET has a very striking pattern of subcellular localisation in ES cells, where it is mainly found in punctate foci specifically within the euchromatin region as indicated by the absence of 4',6-diamidino-2phenylindole (DAPI) staining ( Figure 4a). Notably, we found that these ESET punctate foci overlap with PML nuclear bodies ( Figure 4b). Coimmunoprecipitation experiments in 293T (Additional file 5) and ES cells under mild conditions (Figure 4c and Additional file 4) show that ESET interacts with PML. PML nuclear bodies are of interest because these are highly organised structural and functional domains that are composed of the PML protein and various other proteins that are involved in different biological functions [20,21]. Post-translational modification of PML through addition of SUMOs is pivotal for the formation of PML nuclear bodies [22,23], and many proteins found in the PML nuclear bodies are SUMOylated [24,25]. Since ESET colocalises to PML nuclear bodies and interacts with PML in ES cells, we investigated whether ESET undergoes SUMOylation.
Oct4 and ERG-associated protein with SET domain (ESET) synergistically repress Cdx2 To test this hypothesis, HA-ESET and Flag-SUMO-1 were cotransfected in 293T cells and cell lysates were subjected to immunoprecipitation in the presence of N-ethylmaleimide (NEM), an inhibitor of SUMO isopeptidases. When cell lysate was immunoprecipitated with anti-HA antibody (to immunoprecipitate ESET) and immunoblotted with anti-Flag antibody (to visualise SUMOylated ESET), we observed several slower migrating ESET proteins, which may be indicative of different degrees of SUMOylation of ESET protein (Figure 4d). This result suggests that ESET may be SUMOylated in ES cells and that the high molecular weight ESET proteins that interact with Oct4 are probably SUMOylated forms of ESET proteins.
To confirm that the various larger forms of ESET proteins being immunoprecipitated by Oct4 are SUMOylated ESET, we performed immunoprecipitation of ESET from ES cell lysate in the presence of NEM. We found that ESET was More importantly, these bands were less prominent when immunoprecipitation was carried out without the addition of NEM (Figure 4e, lanes 6 to 9). In addition, the higher molecular weight ESET proteins corresponding to SUMOylated ESET were also observed with the immunoprecipitant from anti-PML antibody (Additional files 4 and 5). Collectively, these data confirmed that the higher molecular weight ESET proteins that interact with Oct4 are SUMOylated ESET suggesting that SUMOylated ESET-Oct4 interaction is required for the repression of Cdx2.
Consistent with this, we show by ChIP experiments that SUMO-1 is highly enriched on Cdx2 promoter (Figure 4f), most notably at the region that was amplified by primer C7, which happens to be the region where ESET was also most enriched (Figure 2e). The presence of SUMO-1 and ESET at the same region on Cdx2 promoter suggests that SUMOylated ESET binds to Cdx2, although we cannot rule out a possibility that other SUMOylated proteins might be present in the same region.

Delocalisation of ESET from PML nuclear bodies enhances the interaction of SUMOylated ESET and Oct4
We then considered whether the localisation of ESET to the PML bodies is crucial for its interaction with Oct4. To do this, we performed shRNA knockdown of Pml in ES cells. Notably, we found that ESET punctate foci were lost in PML-depleted ES cell nuclei (Figure 5a, white border), although the ESET protein levels remained unaltered (Figure 5c: compare lanes 1 and 2). Furthermore, confocal images suggested that the ESET protein was now uniformly dispersed throughout the ES nuclei. Notably, this delocalisation of ESET from PML nuclear bodies increased its interaction with Oct4 ( Figure 5b, lane 4). In the reciprocal immunoprecipitation experiments, more SUMOylated ESET was found to associate with Oct4 (Figure 5c, lanes 4 and 6). This observation suggests that the scaffold composed of PML regulates the interaction between SUMOylated ESET and Oct4, perhaps by modulating the levels of SUMOylated ESET. These results are consistent with our observations that Oct4 preferentially interacts with SUMOylated ESET.

SUMO-interacting motif (SIM) of Oct4 is crucial for the interaction with ESET and the repression of Cdx2
To further confirm the interaction of Oct4 and SUMOylated ESET, we attempted to map the sumoylation sites of ESET and the SIM in Oct4. The consensus sumoylation site sequence is ¬-Lys-X-Glu/Asp (¬KxE/D), where ¬ is a large hydrophobic amino acid and X is any amino acid [26]. Based on the sumoylation prediction software, SUMOplot (Abgent, San Diego, California, USA), many putative sumoylation sites were found in the ERG-associated protein with SET domain (ESET) localises to promyelocytic leukaemia (PML) nuclear bodies and is post-trans-lationally modified by small ubiquitin-related modifier (SUMO) Figure 4 (see previous page)   ESET protein (Additional file 6). This is consistent with our observation that SUMOylated ESET exhibited multiple bands when immunoprecipitated with Oct4. Furthermore, this would also explain why both the Tudor domain-containing and SET domain-containing fragments of ESET were found to interact with Oct4 (Additional file 3). Nevertheless, the presence of a large number of putative sumoylation sites in ESET complicates their functional analysis through mutation of these sites. However, only one putative SIM was found in Oct4 ( Figure  6a). We thus focussed our studies on investigating the Oct4 SIM.

ERG-associated protein with SET domain (ESET) localises to promyelocytic leukaemia (PML) nuclear bodies and is post-translationally modified by small ubiquitin-related modifier (SUMO). (a) ESET (red) localises to dis-
Proteins that contain the SIM with the consensus Val/Ile-X-Val/Ile-Val/Ile (V/I-X-V/I-V/I) in which X is any amino acid, have a high propensity to interact non-covalently with SUMOylated proteins [27][28][29][30]. An inverse SIM (amino acids 266-VVRV-269) was found in Oct4 ( Figure  6a). Importantly, when this SIM was mutated in Oct4 (Oct4-ΔSIM), the interaction with ESET was abrogated ( Figure 6b, lane 2). This result shows that the SIM of Oct4 is crucial for SUMOylated ESET-Oct4 interaction. The precise nature of this interaction remains to be elucidated.
We then considered if the SIM of Oct4 is functional in the repression of Cdx2. We investigated this possibility by substituting the endogenous Oct4 in ZHBTc4 ES cells with exogenous Oct4-ΔSIM. Since the SIM of Oct4 is critical for the interaction with ESET, we anticipated that Cdx2 repression would be less effective in cells that express Oct4-ΔSIM compared to cells that express wild-type Oct4. Indeed, this was the case. In control ZHBTc4 ES cells treated with tetracycline, Cdx2 was upregulated concomitant with the depletion of Oct4 (Figure 6c, lane 2). When these cells were supplied with exogenous Oct4 (Figure 6c . This suggests that the difference in the outcome of response to Oct4 or Oct4-ΔSIM on the repression of Cdx2 in the ES cells, might be due to the nature of the repressor complexes that it recruits through its SIM, as Oct4-ΔSIM is able to bind DNA (Additional file 7). It is possible that Oct4 may interact with SUMOylated ESET through the subunits of the transcriptional repression complexes such as HDAC2 [31,32]. We note that Gata6, a primitive endoderm-specific gene, as well as Hand1 and Dlx3 that were upregulated in Eset knockdown ES cells (Figure 1d), were similarly repressed by Oct4 and resistant to repression by Oct4-ΔSIM (Additional file 7). These data suggest that SUMOylated ESET-Oct4 complex may act to repress genes other than Cdx2. It is equally likely that ESET might interact with other transcription factors to repress a different subset of genes in ES cells. Further work is also required to delineate how Oct4 functions in ES cells to discriminate between its role both as an activator, and as a repressor of genes through ESET-Oct4 complex.

Conclusion
We have uncovered an important epigenetic mechanism that maintains pluripotency by preventing differentiation of ES cells, notably into trophectoderm cells. The Oct4-ESET mediated H3K9me3 epigenetic modification involved in the repression of Cdx2 may affect other genes, including Gata6, to underpin pluripotency. The synergistic action of Oct4 and ESET could also explain the previously described observation that Oct4 regulates expression of Cdx2 [12]. The SIM of Oct4 is apparently critical for the interaction of ESET and Oct4, which highlights the importance of this motif in forming an effective SUMOylated ESET-Oct4 repressive complex. Since ESET, like Oct4 is a maternally inherited protein in the oocyte [8], they may also be critical for the establishment of

Figure 5 (see previous page) Delocalisation of ERG-associated protein with SET domain (ESET) from promyelocytic leukaemia (PML) nuclear bodies promotes the interaction of small ubiquitin-related modifier (SUMO)ylated ESET and Oct4. (a)
In pluripotent cells in the ICM, at least in part through repression of Cdx2 (Figure 7). Notably, the loss of ESET or Oct4 results in the loss of the pluripotent ICM [8,9]. Furthermore, disruption of the SUMO pathway by the inactivation of the SUMO E2-conjugating enzyme Ubc9 is detrimental to the development of the ICM [33]. The use of a conditional allele of Eset will address the precise developmental role of this gene in the future. SUMOylated ESET-Oct4 interaction is probably pivotal for both the establishment of pluripotency in the ICM and its maintenance in ES cells.

RNAi constructs and transfection
shRNA oligonucleotides were cloned into the BglII and HindIII sites of the pSuper.puro vector (Oligoengine, Seattle, WA, USA). Sequences for Eset shRNA which has been described previously [5] are 5'-GATCCCCGATGTGAGTGGATATATCGTTCAAGAGAC-Small ubiquitin-related modifier (SUMO)-interacting motif (SIM) of Oct4 is crucial for the interaction with ERG-associated pro-tein with SET domain (ESET) and the repression of Cdx2
For construction of shRNA-poliovirus internal ribosomal entry site (pIRES)-EGFP, pSuper.puro with or without shRNA insert that has been digested with NotI and HincII were ligated to pIRES-EGFP (Clontech, Mountain View, CA, USA) that has been digested with NruI and NotI. ZHBTc4 ES cell cultures were as described previously [18]. Oct4 expression was regulated by the addition of 1 μg/ml doxycycline (Sigma, St Louis, MO, USA) to the culture medium. TS cultures were as described previously [13].

RNA extraction and real-time Q-PCR
Total RNA was prepared using RNeasy Mini Kit (Qiagen, Hilden, Germany) and cDNA was synthesised from 1 μg of RNA using SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, USA). Endogenous mRNA levels were measured by Q-PCR based on SYBR Green detection with the ABI Prism 7000 PCR machine (Applied Biosystems, Foster City, CA, USA). Each reaction in a total volume of 20 μl contained 1 μl of 10 × diluted cDNA, 1 μM of forward and reverse primer and 1 × QuantiTect SYBR Green Master Mix reagent (Qiagen, Hilden, Germany). Standard curves for each primer were performed in the same sample plate to determine the relative quantification of the transcript. Q-PCR was performed in triplicates and normalised with Gapdh, a housekeeping gene. The data were then normalised against vector control or untreated ES cells which were defined as 100% or 1.0. Each experiment was performed independently on at least two occasions. Sequences of primers used for Q-PCR are available in Additional file 8.

Cotransfection and immunoprecipitation
Full details on vector constructions are available in Additional file 9. For cotransfection experiments, 3.5 × 10 6 293T cells plated on a 10 cm dish and cultured overnight were transfected with 18 μg of DNA comprising of 9 μg of two different constructs with Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA). Cells were harvested for immunoprecipitation 48 h after transfection. Cells were washed twice in cold phosphate-buffered saline (PBS) and scraped into 1 ml of PBS. Cell pellet were resuspended in 200 μl of immunoprecipitation buffer (50 mM Tris pH 7.5, 150 mM NaCl, 5 mM ethylenediaminetetraacetic acid (EDTA), 15 mM MgCl2, 0.75% sodium deoxycholate and 1% NP40 added with protease inhibitors from Roche, Basel, Switzerland), incubated on ice for 30 min and centrifuged at 13,000 rpm for 30 min at 4°C. A total of 100 μl of the supernatant was then diluted with 900 μl dilu-tion buffer (50 mM Tris pH 7.5 and 150 mM NaCl added with protease inhibitor from Roche, Basel, Switzerland) so that the final concentration of NP40 in the immunoprecipitation reaction was 0.1%. The primary antibody was incubated with the protein lysate at 4°C overnight. Precipitation was performed by adding 50 μl of Dynabeads Protein G (Invitrogen, Carlsbad, CA, USA) to the reaction for 1 h followed by five washes in buffer containing 50 mM Tris pH 8.0, 150 mM NaCl and 0.1% NP40. Beads were boiled for 5 min with 50 μl 2 × Laemmli sample buffer (Bio-Rad, Hercules, CA, USA) and 20 μl of the supernatant were subjected to western blot. Where indicated, lysis buffer was also added with 20 mM NEM (Sigma, St Louis, MO, USA) and 1:100 diluted phosphatase inhibitors I and II (Sigma, St Louis, MO, USA).

Chromatin immunoprecipitation
Chromatin immunoprecipitation was performed according to a published protocol [35] with some modifications. Briefly, cells were crosslinked with 1/10 volume of fresh 11% formaldehyde solution for 15 min and quenched with 1/20 volume of 2.5 M glycine. Cells were sonicated to an average of 500 base pairs (bp) and immunoprecipitated overnight with antibody that was preincubated with 100 μl Dynabeads M-280 Sheep Anti-Rabbit (Invitrogen, Carlsbad, CA, USA) overnight. For isolation of DNA, 100 μl of 10% Chelex (w/v) was added to the washed beads, vortexed and boiled for 10 min [36]. After cooling to room temperature, 100 μg/ml of proteinase K was added and beads were incubated for 30 min at 55°C in a shaking block. Beads were boiled for another 10 min, centrifuged, and the supernatant was collected. The Chelex/bead fraction was vortexed with another 100 μl of water, centrifuged and the supernatant collected was combined with the first supernatant. Immunoprecipitated DNA (3 μl) was used as template for PCR amplification with Red Taq (Sigma, St Louis, MO, USA) or 1 to 2 μl was used for Q-PCR analysis using 1 × QuantiTect SYBR Green Master Mix reagent (Qiagen, Hilden, Germany). Data were normalised to the input or H3 ChIP, control IgG and the least enriched region (as indicated in the figure legends). To compare the enrichment of H3K9me3 between samples, carrier ChIP was performed by adding 3 × 10 7 293T cell to 1 × 10 6 FACS-sorted, EGFP-positive ES cells transfected with either Eset shRNA or empty vector for 3 days (selected with puromycin for 2 days). Carrier ChIP of H3K9me3 was also performed in Figure 3d. Immunoprecipitated DNA was concentrated by ethanol precipitation before PCR analysis. Data were normalised against the input and relative to the controls. Semi quantification of band intensity was performed using Quantity One software (Bio-Rad, Hercules, CA, USA). ChIP experiments were performed independently on at least two occasions. Primers and PCR conditions for ChIP are available in Additional file 10.

Mutagenesis and transfection
Oct4-ΔSIM construct was generated using the Quik-Change XL Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). Details on the construct are available in Additional file 9. For transfection, 0.5 × 10 6 ZHBTc4 ES cells plated on 1 well of a 6-well plate and cultured overnight were transfected with 3 μg of plasmid using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA). Doxycycline 1 μg/ml (Sigma, St Louis, MO, USA) was added to the culture medium after 6 h of transfection to remove the endogenous Oct4. Transfected cells were sorted by FACS after 48 h of transfection.

Immunofluorescence microscopy
Single cells on poly-l-lysine (Sigma, St Louis, MO, USA) treated 12 well microscope slides (Erie Scientific Company, Waltham, MA, USA) were centrifuged at 1,000 rpm for 1 min and fixed with 4% paraformaldehyde (Sigma, St Louis, MO, USA) in PBS for 15 min. This was followed by three washes in PBS and blocking and permeabilisation in 0.1% Triton X (v/v) and 1% bovine serum albumin (BSA) (w/v) in PBS for 30 min. Primary antibodies diluted in blocking buffer were incubated at 4°C overnight. Secondary antibodies conjugated to Alexa Fluor 488 and 568 (Molecular Probes, Carlsbad, CA, USA) and DAPI (1 μg/ ml) were incubated at room temperature, in the dark for 1 h. Cells were mounted on slides with Vectashield (Vector Laboratories, Burlingame, CA, USA) containing DAPI and images were captured with a Bio-Rad Radiance 2000 confocal microscope.