We and others have previously shown that sense as well as antisense transcription across the Xist locus prior to the onset of X inactivation plays a role in X chromosome inactivation choice 789111213. The mechanism for this is unknown although there is evidence that Tsix has an influence on chromatin structure and DNA methylation of the Xist promoter.

Evidence to date indicates that antisense Tsix transcription has an influence on Xist promoter DNA methylation in differentiating ES cells 6, and in somatic cells 4617, but not in undifferentiated ES cells that are representative of Xist promoter status prior to the onset of X inactivation 4617. To determine whether sense transcription can influence DNA methylation of the Xist promoter prior to X inactivation we analysed two XY ES cell lines carrying mutations in the Xist 5' region, initially using conventional methylation sensitive restriction enzyme site (MSRE) analysis. The first mutation, Δ5', is a deletion of a 9 kb region 1.1 kb upstream of the Xist transcriptional start site (TSS). The second is an insertion of a transcriptional termination site SPA-MAZ₄ in the region -1.1 kb relative to the TSS 12 (Figure 1A). Both of these mutations show enhanced sense transcription in undifferentiated XY ES cells correlating with preferential inactivation of the mutant X chromosome in vivo 12. In accordance with previous findings 28, the Xist promoter was found to be highly methylated in undifferentiated wild-type (wt) XY ES cells (Figure 1A and 1B). Interestingly, both the Δ5'+neo and SPA+neo XY ES cell lines showed significant hypomethylation at all of the CpG sites analysed, that is, HpaII, HaeII, HhaI, MluI and SacII (Figure 1B). Quantification of the bands using ImageQuant software demonstrated methylation loss within a range of 20–45% for various CpG sites, with the SPA+neo mutant affected more severely (Figure 1C). This indicates that enhanced sense transcription across the Xist promoter can lead to CpG hypomethylation.

We then analysed another Xist mutation, XT67E1 29, a deletion of the most part of Xist exon 1 and the minimal promoter region on the 129 allele in Pgk12.1 XX ES cells (Figure 1A). Whilst the deletion in XT67E1 cells removes a number of methylatable CpG sites in the Xist promoter, sites 36 bp upstream of the TSS are retained. Moreover, the deletion gives rise to a change in size of a BamHI fragment in the Xist 5' region, and we were therefore able to discriminate between wt and mutant alleles (Figure 1A and 1D). MSRE analysis revealed complete hypomethylation on the mutant allele. The wt allele was mosaically methylated, similar to the parental XX ES cell line Pgk12.1 29.

As the Xist TSS is deleted on the mutant allele in XT67E1 cells we anticipated that sense transcription would not be detectable but that antisense Tsix transcription would be unaltered. However, analysis by strand specific polymerase chain reaction (PCR) revealed unexpectedly that the mutant allele transcribes both sense and antisense RNAs and moreover that sense transcripts are abundant relative to the parental Pgk12.1 cell line (Figure 1E, amplicon 4). To verify this result we designed primers that were able to discriminate between mutant and wt alleles; the forward primer was the same for both alleles and was located 59–87 bp upstream of the Xist TSS. The reverse primers were located either in exon 1 (TN51, wt allele) or at the 3'end of the neomycin selective cassette (neoTN9 ≡ 51mut, mutant allele). Strand- and allele-specific reverse transcription (RT) PCR with these primers clearly demonstrates ectopic sense transcription on the mutant allele (Figure 1E). The PGK promoter that drives expression of the neomycin resistance gene cannot be the origin of ectopic transcripts as it is located about 1.7 kb downstream from the region analysed and in a reverse orientation relative to Xist. This suggests that sense transcription is activated from a minor upstream Xist TSS, in line with a result reported previously 13. Importantly this result reinforces the conclusion from analysis of Δ5'+neo and SPA+neo mutant ES cells that enhanced sense transcription initiated upstream of Xist antagonises Xist promoter methylation.

To investigate Xist promoter methylation in more detail we used SEQUENOM matrix-assisted laser desorption/ionisation time of flight (MALDI-TOF) mass spectrometry analysis of bisulphite-modified DNA 30. This approach allowed us to analyse methylatable CpGs more widely and also to obtain accurate quantitative measurement of CpG methylation levels. To validate the method we first analysed the methylation patterns in control XX and XY somatic and ES cell lines (Figure 2A and 2B). As expected, CpG methylation was close to 100% in XY male somatic cells and about 50% in XX female somatic cells, representing the average of a fully methylated inactive Xist locus and a fully unmethylated active locus on the inactive X chromosome 28. In XY ES cells methylation was close to 100% in Xist region 1, although somewhat lower in region 2. Xist was significantly hypomethylated in XX ES cells, consistent with our previous observations 31.

Having validated the SEQUENOM method we went on to analyse Xist promoter methylation for the mutant XY ES cell lines described above, and additionally in Δhs XY mutant ES cell lines where ectopic transcription-dependent skewing of X inactivation has also been reported to occur 13. We found that deletion of 9 kb of the Xist upstream region (Δ5') leads to around 20% loss of methylation in both region 1 and region 2. This was also the case for mutants carrying the PGKneo cassette, as well as for Δneo ES cells (Figure 2C). More severe loss of methylation was observed for SPA+neo mutant (Figure 2D), consistent with the MSRE analysis above (Figure 1B and 1C). Moderate hypomethylation was observed for the Δhs+neo, but not for the Δhs Δneo mutant (Figure 2E). Strikingly, all of the mutants that show hypomethylation of the Xist CpG island also show preferential X inactivation of the mutant allele in female heterozygotes 1213. Conversely, the Δhs Δneo mutation that does not affect the randomness of X inactivation shows no hypomethylation. Collectively, these results demonstrate a direct correlation between hypomethylation of the Xist promoter and an increased probability of that chromosome being chosen as the inactive X in heterozygous females.

It was previously reported that abolition of Tsix transcription does not cause Xist promoter hypomethylation in undifferentiated ES cells, although analysis was restricted to two MSRE sites in region I 6. To further address this issue we used the SEQUENOM assay to assess Xist promoter methylation in two different Tsix mutant ES cell lines, pSS1Δ2.7 and pAA2Δ1.7 11. In the first mutant, pSS1Δ2.7, Tsix exon 1 was deleted, but this had no effect on Tsix transcription or function. In the second mutant, pAA2Δ1.7, Tsix transcription through the Xist locus was abolished, causing primary non-random inactivation of the targeted allele in female embryos. We analysed three independent pAA2Δ1.7 XY ES cell clones, and in all cases we observed clear hypomethylation of the Xist promoter (Figure 2F). Hypomethylation in CpG region I was moderate compared with CpG region 2, possibly accounting for the fact that Sun et al. 6 did not observe this result. No methylation differences were observed for the pSS1Δ2.7 mutant, consistent with normal Tsix transcription and random X inactivation in female heterozygotes. This result suggests that prior to the onset of X inactivation, Tsix transcription, possibly together with physiological levels of sense Xist transcription, determines Xist promoter CpG methylation levels that, in turn, have an impact on the probability of that X chromosome being selected as the inactive X during the onset of random X inactivation. Increasing sense transcription, or possibly utilisation of a heterologous upstream promoter, antagonises promoter CpG methylation such that there is an increased probability of that X chromosome being selected as the inactive X in a heterozygous female.

What is the mechanism by which sense and antisense transcription impact on CpG methylation and repress the Xist promoter? One possibility is that sense and antisense RNAs trigger an RNAi response, similar for example to RNA-dependent DNA methylation in higher plants 26. Alternatively, an RNAi-independent mechanism involving the antisense RNA Tsix, or both sense and antisense RNAs, may operate. The fact that increased sense transcription reduces Tsix-dependent Xist promoter methylation could be interpreted to indicate that silencing does not depend on dsRNA production. However, this does not rule out that the low levels of endogenous sense transcription from the Xist promoter co-operates with antisense Tsix RNA in RNAi-mediated silencing. To investigate this further we set out to analyse ES cells deficient for the RNase III enzyme Dicer, which is essential for the RNAi pathway in mammalian cells.

We set out to derive an ES cell line in which the gene encoding Dicer can be deleted conditionally using CRE/loxP, allowing us to discriminate between primary effects attributable to Dicer deletion and secondary effects resulting from derivation and long-term culture of ES cells deficient for this essential factor. We established the D41 XY ES cell line, in which the RNase III domain was flanked with loxP sites on both Dicer alleles 27. Initially, Dicer-deficient ES cell lines were derived from the D41clone by transfection with a pCAG-Mer-Cre-Mer, tamoxifen-inducible Cre recombinase plasmid (see Methods for details), followed by hydroxytamoxifen (4-OHT) treatment (Figure 3A). The efficiency of Cre-recombination was low, and Dicer^Δ/Δ colonies were quickly overgrown by heterozygous Dicer^lox/Δ colonies. Nevertheless, we isolated three independent sub-clones, S5, S6 and E5, that had deleted the Dicer RNase III domain on both alleles (Figure 3B).

To ensure that deletion of the RNase III domain completely abolished Dicer function we performed Northern blot hybridisation of RNA from floxed and Dicer-deficient clones with a probe for the micro RNA, miR-292 (Figure 3C). The result confirmed the absence of miR-292 and enrichment of the corresponding pre-miRNA in the S5 and S6 Dicer mutant clones relative to control floxed cell lines.

We went on to characterise Dicer-deficient ES cell lines. In accordance with previous observations 32, Dicer^Δ/Δ clones overexpress major satellite repeats (data not shown) and are not capable of differentiation. We attempted to differentiate the cells by withdrawing LIF. In contrast to parental Dicer^lox/loxcells, D3 Dicer^Δ/Δ clones did not form embryoid bodies, but stayed in irregular-shaped clumps which subsequently attached and continued to grow. After 11 days of differentiation levels of the pluripotent ES cell markers Oct4, Nanog, Fgf4 and Errβ expression levels remained unchanged (Additional file 1). Interestingly, T/Brachyury, which is usually expressed at low levels in wt ES cells, presumably due to a small number of differentiated cells, was completely absent in Dicer^Δ/Δ clones, and did not appear even after culturing cells for 11 days under differentiation conditions. This result suggests that Dicer-deficient ES cells are either unable to differentiate or alternatively that differentiated cells present in the cultures do not survive.

To determine whether Dicer deficiency affects Xist promoter methylation, we performed MSRE analysis of DNA from control and Dicer^Δ/Δ clones (Figure 3D and 3E). Interestingly, all Dicer^Δ/Δ clones demonstrated partial loss of methylation at all of the restriction sites analysed. It should be noted, however, that different clones showed varying degrees of methylation loss, with S5 showing the highest percentage of hypomethylation and S6 the lowest (Figure 3E and 3F).

The cell lines used in this preliminary analysis underwent several rounds of cloning and selection during the derivation process, so we proceeded to derive further lines, in this case making use of ES cells carrying floxed Dicer alleles and a tamoxifen inducible Cre recombinase gene targeted into the Rosa26 locus 33. In this system Cre-recombinase-mediated deletion of the floxed cassette was very efficient following addition of tamoxifen, and we were able to select several individual clones from two different parental Dicer^lox/lox cell lines, DTCM23 and DTCM49. In addition, we established Dicer-deficient cells from pools of 200–250 colonies of tamoxifen treated cells (Figure 3A). All further analyses were performed in parallel on clones derived by both approaches.

To determine Xist promoter hypomethylation quantitatively we analysed bisulphite-modified DNA from the D3^lox/lox and S5, S6, E5 Dicer^Δ/Δ clones using SEQUENOM MALDI-TOF mass spectrometry analysis. D3-derived Dicer^Δ/Δ clones showed substantial hypomethylation both in region 1 and region 2 (Figure 4A). The D3^lox/lox parental clone also showed moderate loss of methylation in region 1 and significant loss in region 2, but less than in Dicer^Δ/Δ clones. The reason for this remains unknown, but could be explained by a mutation or rearrangement that occurred during the several rounds of cell selection to which the D3^lox/lox cells were subjected. To exclude the possibility of such an unrelated mutation causing the observed hypomethylation phenotype, we performed mass spectrometry analysis for the DTCM23^lox/lox parental cell line and for Dicer^Δ/Δ clones 23ΔE3, 23ΔF4 and the 23Δpool. The DTCM23^lox/lox cell line showed a methylation pattern similar to the 129/1 XY ES control, while all Dicer^Δ/Δ clones demonstrated hypomethylation most notably in region 2 (Figure 4B). A similar result was obtained for another set of clones, DTCM49^lox/lox and Dicer^Δ/Δ derivatives (Additional file 2A).

To determine the dynamics of methylation loss, we treated the DTCM23^lox/lox and DTCM49^lox/lox cell lines with 4-OHT and collected DNA for SEQUENOM analysis 50 and 168 hours later. The result shows that Xist promoter hypomethylation occurs rapidly following the deletion of Dicer and that further hypomethylation occurs with continued cell passaging, again most notably in region 2 (Figure 4C and Additional file 2B).

To test how hypomethylation affects the transcriptional status of the Xist promoter in Dicer-deficient ES cell lines, we performed RNA fluorescent in situ hybridisation (FISH) analysis on the S5 Dicer^Δ/Δ ES clone using probes detecting Xist and Tsix transcripts. The majority of the S5 cells showed a single pinpoint signal similar to the control 129/1 XY ES cells. However, occasionally there were cells with an upregulated Xist signal, which either 'painted' the X chromosome (Figure 5A, third panel from the top) or was scattered in the vicinity of the chromosome (Figure 5A, bottom panel). On average about 10% of cells showed this upregulation pattern, confirming that hypomethylation of the Xist promoter compromised regulation of Xist expression.

Next we analysed Xist expression quantitatively for all Dicer^lox/lox control and Dicer^Δ/Δ ES clones. The data for Xist expression was normalised to β-actin and then to the level of the Xist transcript in control 129/1 XY ES cells and is presented in Figure 5B. All Dicer^Δ/Δ ES clones showed an elevated level of Xist expression in comparison with their corresponding parental floxed cell lines; however, the absolute level of Xist upregulation varied between individual clones. It is worth noting that while Xist expression in DTCM23^lox/lox and DTCM49^lox/lox was the same as 129/1 control, the D3^lox/lox clone showed elevated expression, consistent with the observed promoter hypomethylation.

Xist promoter hypomethylation in Dicer-deficient ES cells could be due to a direct effect on recruitment of DNA methyltransferases (Dnmts), for example mediated by the RNAi pathway. Alternatively sense and/or antisense transcription may be required to establish other features of the underlying chromatin structure at the Xist promoter, for example specific histone lysine methylation marks, which in turn could have an indirect impact on the recruitment of Dnmts. To test the indirect model we analysed the repressive histone modifications H3K9me2 (data not shown), H3K27me3, H4K20me3 as well as the active mark H3K4me2 over the Xist locus in wt and Dicer-deficient ES cells using chromatin immunoprecipitation (ChIP). None of these histone modifications showed a significant change in Dicer-deficient ES cells (Additional files 3 and 4).

The absence of detectable changes in histone modification in Dicer^Δ/Δ cells relative to floxed parental cells suggested that hypomethylation results from a direct effect on recruitment of Dnmts. It was reported previously that Xist promoter methylation is mediated by de novo DNA methyltransferases Dnmt3a and/or Dnmt3b 34. We therefore went on to analyse the levels of these enzymes, and also of the maintenance methyltransferase Dnmt1, using Western blotting. In agreement with previous observations Dnmt3a and Dnmt3b levels were very low in XX compared with XY control ES cell lines 31. Interestingly we also observed reduced levels of Dnmt3a in Dicer^Δ/Δ clones in comparison with Dicer^lox/lox controls (Figure 6A–C). The most affected clone was S5, which showed approximately five times less Dnmt3a protein in comparison with the D3 control. The DTCM23 and DTCM49 sets of Dicer^Δ/Δ clones demonstrated a depletion of Dnmt3a and also a slight decrease in Dnmt3b levels.

To determine whether Dnmt3a/b depletion resulted from transcriptional or post-transcriptional regulation, we performed quantitative RT-PCR analysis with primers designed for Dnmt1, Dnmt3a2 (the major Dnmt3a isoform in ES cells), Dnmt3b (for all Dnmt3b isoforms) and Dnmt3L (Figure 6D–F). Consistent with the Western data, we found that levels of Dnmt3a2 transcript are consistently lower in Dicer^Δ/Δ clones than in controls. We also found that the level of Dnmt3L, a functional partner of Dnmt3a2, is much reduced in Dicer-deficient clones. The level of Dnmt3b was reduced in the DTCM series of clones, but not in the S5, S6 and E5 Dicer^Δ/Δ clones, in accordance with the Western blot results. The level of Dnmt1 was not significantly different between control and Dicer-deficient clones. Affymetrix microarray analysis of RNA from D3^lox/lox versus S5 Dicer^Δ/Δ clones also showed a 2.4-fold decrease of Dnmt3a and 3.3-fold decrease of Dnmt3L (data not shown).

To determine whether hypomethylation occurs at other loci in Dicer^Δ/Δ clones we analysed methylation at the differentially methylated regions (DMRs) of two imprinted genes, H19 and Igf2rAir (Additional file 5). In both examples we observed hypomethylation specifically in Dicer^Δ/Δ clones. It should be noted that two recent studies reported hypomethylation of repetitive and unique sequences in Dicer^Δ/Δ ES cells and attributed this to reduced levels of Dnmts 3536.

Given the requirement for Dnmt3a/b in Xist promoter methylation 313437, we conclude that hypomethylation in Dicer^Δ/Δ clones is most likely attributable to decreased levels of expression of these enzymes rather than deficiency of a dsRNA-mediated transcriptional gene silencing mechanism.

Finally, we wished to test for a role for Dicer in the initiation of random X inactivation in XX cells and also to determine whether the RNAi pathway is important in spreading Xist RNA on the inactive X chromosome. As Dicer-deficient ES cells are not capable of differentiation and we have not been able to isolate a stable Dicer-deficient XX ES cell line, we analysed Dicer-deficient XX embryos produced by mating Dicer^wt/Δ heterozygous mice. In general, Dicer^Δ/Δ embryos survived until approximately E7.5–E8.5 and were smaller than their heterozygous or wt littermates, consistent with a study published previously 38. This afforded the opportunity to analyse the initiation of random X inactivation that commences at approximately E5.5. We carried out whole mount RNA FISH using Xist and Tsix probes on E6.5 embryos (Figure 7). The Dicer^Δ/Δ XY embryos showed a Xist/Tsix pinpoint signal similar to their wt and heterozygote XY littermates, while the Dicer^Δ/Δ female embryos had both pinpoint and accumulated Xist transcripts, suggesting that Dicer is not affecting the initiation step of random X inactivation in the inner cell mass (ICM). The presence of Xist clouds in cells of XX embryos implies that spreading of Xist RNA also does not require Dicer activity. We did note that XX embryos showed weaker and more disrupted Xist signal with stronger general background for both Xist and Tsix probes, and that this varied from embryo to embryo. This is likely due to the onset of embryo lethality and apoptosis in mutant embryos. Overall these observations further support our conclusion that X inactivation can occur independent of the RNAi pathway.

Previously we demonstrated that modified Xist alleles showing increased sense transcription from heterologous promoters in ES cells also show preferential X inactivation in XX heterozygous animals 1213. Here we have extended this finding, showing that these modified Xist alleles are partially hypomethylated over the Xist promoter region, providing a mechanistic basis for preferential expression in XX heterozygotes. Similarly the mutant Xist allele in XT67E1 XX ES cells has ectopic transcription in the sense direction and complete hypomethylation of promoter CpG sites that lie immediately upstream of the deleted region. In this case hypomethylation cannot be correlated with increased probability of expression because the deleted allele is lacking the Xist TSS.

Enhanced sense transcription antagonises Xist promoter methylation even when normal levels of antisense Tsix RNA are present. However, our data show that Tsix transcription is important for the establishment of Xist promoter methylation in ES cells, that is, prior to X inactivation. This contrasts with a previous report in which a different Tsix mutant allele was suggested to have a role in Xist promoter methylation during ES cell differentiation, but not prior to onset of random X inactivation in undifferentiated ES cells 6. This discrepancy is explained in part by the fact that hypomethylation occurs more in region 2 than in region 1 (this study), and Sun et al. 6 analysed only region 1. A second possible factor is that independent Tsix mutant XY ES cells behave differently. Notably the Δ65, 2 lox, pAA2Δ1.7 and Δ34# 1 XY ES cell lines all upregulate Xist inappropriately upon differentiation 791116, presumably at least in part due to promoter hypomethylation, whereas the ΔCpG cell line maintains Xist repression throughout differentiation 8. It is possible that this difference arises because the secondary pathway, Xist repression linked to the pluripotency program 18, plays a more dominant role in the ΔCpG XY ES cell line.

We wanted to determine whether regulation of Xist promoter methylation by sense and antisense RNAs was mediated by the RNAi pathway. We found that Dicer-deficient ES cells show Xist promoter hypomethylation and moderate upregulation of Xist transcripts, an effect that was observed in a number of independent cell lines, albeit with some variation in degree. However a number of facts lead us to conclude that this is an indirect consequence of Dicer deletion. First and foremost we observed downregulaton of the de novo DNA methyltransferases, Dnmt3a, Dnmt3b and Dnmt3L in Dicer-deficient cells. Several studies have demonstrated that Dnmt3a/3b levels are important for maintaining Xist promoter methylation 313437, indicating that reduced Dnmt levels are sufficient to account for Xist promoter hypomethylation in Dicer-deficient cells. Consistent with this idea we observed hypomethylation at imprinted loci, and two recent studies have reported hypomethylation of subtelomeric repeats 35 and promoters of Oct4, Tsp50 and Sox30 genes 36 in independently isolated Dicer-deficient cell lines. Importantly the hypomethylation phenotype seen in these studies was complemented by ectopic expression of Dnmt transgenes, indicating that the RNAi pathway is not directly involved. These studies also demonstrated that downregulation of Dnmts results from overexpression of Rbl2 that is, in turn, normally subject to negative regulation by the miR-290 cluster miRNAs. In agreement with this conclusion we also found that the level of Rbl2 transcript is upregulated 4.4-fold in the Dicer-deficient cell lines described here (data not shown).

A second line of evidence that argues that the RNAi pathway is not required for Xist gene regulation and random X inactivation comes from our analysis of Dicer-deficient embryos at early post-implantation stages. Here we observed appropriate Xist and Tsix expression patterns in XY and upregulation of Xist from a single allele in XX embryos. The fact that XX embryos show reduced intensity of staining of Xist domains most likely reflects that embryo lethality occurs shortly after the stage we examined, E6.5 38. It should be noted that we cannot formally rule out that in XX embryos the pattern we observe represents persistence of the imprinted X inactivation pattern, that is, that Dicer deficiency results in failure to erase imprinted X inactivation prior to establishing random X inactivation.

Given the evidence that the RNAi pathway does not mediate Xist promoter regulation through sense and antisense transcription, what are the alternative mechanisms? The fact that we observe hypomethylation in undifferentiated Tsix-deficient XY ES cells suggests a direct link between antisense transcription and promoter CpG methylation. That some CpG methylation is retained in the Tsix mutant cells may indicate a redundant mechanism for recruiting DNA methylation to the promoter, for example, relating to Xist repression by the pluripotency program 1820, or alternatively may simply reflect that the maintenance methyltransferase activity, Dnmt1, is sufficient to retain promoter methylation up to a defined level.

Assuming that Xist promoter methylation prior to the onset of X inactivation is indeed a consequence of antisense Tsix expression we can envisage two possible mechanisms. Either Tsix directly recruits de novo Dnmts, Dnmt3a and Dnmt3b, as has been suggested previously for Dnmt3a 6, or alternatively Tsix may mediate other chromatin changes at the Xist promoter, for example reducing H3K4 methylation as suggested previously 3, DNA hypomethylation being a secondary consequence. Whilst we are unable to distinguish between these possibilities at present, it is interesting to note that H3K4 methylation antagonises the binding of Dnmt3a/Dnmt3L dimers to nucleosome 39, providing a possible mechanism for hypomethylation of DNA driven by a Tsix-mediated reduction in H3K4 methylation levels. In the context of this model increased sense transcription from upstream heterologous or cryptic promoters may antagonise Xist promoter methylation by increasing the H3K4 methylation levels locally.

Dicer^lox/lox ES cell lines were derived from the ICM of E3.5 embryos using two approaches. In the first approach, ES cell lines were derived from mice homozygous for a Dicer^lox allele. The established Dicer^lox/lox XY ES cell line D41 was subsequently lipofected with pCAG-Mer-Cre-Mer plasmid, carrying tamoxifen inducible Cre-recombinase. Clone D41D3Cre (D3Cre) was treated with 800 nM 4-hydroxytamoxifen (4-OHT, Sigma), plated at clonal density, and individual colonies were picked, expanded and then tested by genomic PCR for the loss of the Dicer RNase III domain. Three clones S5, S6 and E5 were identified that showed loss of the Dicer RNase III floxed cassette.

In the second approach ES cell lines were derived from mice homozygous for a Dicer^lox allele crossed to animals either homozygous or heterozygous for tamoxifen inducible Cre-recombinase targeted into the Rosa26 locus (obtained from Artemis Pharmaceuticals; 33). Two derived parental XY ES cell lines, DTCM23 and DTCM49, were treated with 800 nM 4-OHT and plated at clonal density. Individual clones as well as pools of approximately 200–250 clones were genotyped for the loss of the RNase III floxed cassette. Dicer-deficient clones DTCM23 ΔE3, ΔF4, DTCM49 ΔA1 and ΔE2 as well as a pool of Dicer-deficient clones DTCM23 Δpool were selected for further analysis.

ES cells lines were derived and maintained on a feeder layer (mitomycin-inactivated primary mouse embryonic fibroblasts) in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% foetal calf serum (FCS, Autogen Bioclear), 7% Knockout Serum Replacement (KSR), 2 mM L-glutamine, 1× non-essential amino acids, 50 μM 2-mercaptoethanol, 50 μg/ml penicillin/streptomycin (all from Invitrogen) and LIF-conditioned medium, made in house, at a concentration equivalent to 1000 U/ml. Cells were grown at 37°C in a humid atmosphere with 5% CO₂.

Dicer^lox/lox and deficient ES cells were pre-plated for 30 minutes to minimise feeder cell contamination and then grown for 2–3 days until confluent on plates coated with 0.1% gelatin. Genomic DNA was phenol/chloroform extracted by the standard procedure. The genotype of each preparation was confirmed by PCR using primers SEQ28290 (agtaatgtgagcaatagtcccag), Di31831 (agtgtagccttagccatttgc) and Di32050AS (ctggtggcttgaggacaagac) and the following PCR conditions: 95°C for 4 minutes; (95°C for 30 seconds; 60°C for 30 seconds; 72°C for 30 seconds) × 35 cycles. PCR fragments were resolved on a 2.5% agarose gel, giving a 259 bp fragment for the wt allele, a 390 bp fragment for the floxed allele and a 309 bp fragment for the Dicer null allele (see Figure 3B).

Genomic DNA was digested to completion with either EcoRI or BamHI restriction enzymes according to the manufacturer's instructions, ethanol precipitated and re-dissolved in TE buffer (10 mM Tris, pH 8.0; 1 mM EDTA). Using methylation-sensitive enzymes, 10 μg DNA aliquots were re-digested, separated by electrophoresis on a 1% agarose gel and blotted onto a GeneScreen nylon filter (Perkin Elmer Life Sciences). Hybridisation with Xist probe 3 (from -37 bp to +952 bp relative to Xist start site P₁) was performed as described previously 40. Images were collected on a PhosphorImager instrument (Molecular Dynamics) and fragment intensity quantification was performed using ImageQuant software (Molecular Dynamics).

Genomic DNA was extracted in the same way as for methylation-sensitive Southern analysis. We bisulphite treated 2 μg aliquots of high molecular weight DNA using the EZ DNA methylation kit (Zymo Research). Treatment was performed essentially according to the manufacturer's instructions with a modification in the conversion step, that included 20 cycles of sample treatment with the following conditions (95°C for 30 seconds; 50°C for 15 minutes). Converted DNA was purified on the columns and eluted in 100 μl of water. We used 5 μl of the sample per 25 μl PCR reaction.

HotStarTaq DNA Polymerase kit (Qiagen) was used to amplify modified DNA and PCR primers and the conditions used are described in the Table 1. PCR fragments were sent to SEQUENOM GmbH company (Hamburg, Germany) for in vitro transcription and subsequent MALDI-TOF mass spectrometry analysis using EpiTYPER software 30.

RNA was isolated from ES cells using TRIzol reagent (Sigma) according to the manufacturer's instructions. RNA was routinely treated with Turbo DNA-free reagent (Ambion) to exclude the possibility of DNA contamination. cDNA synthesis was primed from random hexamers (GE Healthcare) with Superscript III reverse transcriptase (Invitrogen). Strand-specific RT-PCR for Xist amplicons 4, 5, 51 and 51mut was performed according to the method described previously 13. Primers and PCR conditions are given in Table 2.

Real-time PCR was performed with SYBR Green PCR Master Mix (Bio-Rad) on a Chromo4 Real-time PCR System (Bio-Rad). PCR primers and conditions are listed in Table 2. A melting curve test was performed at the end of each experiment to ensure the specificity of amplification. The data was normalised to β-actin and Idh2 and then to one of the control samples in the set. Each amplicon was analysed at least twice in triplicate on independent cDNA preparations.

Total RNA (20–30 μg), isolated using TRIzol reagent, was separated by PAGE on a 15% urea-containing gel. RNA was transferred to Hybond-XL nylon membrane using a Bio-Rad semi-dry blot apparatus at a constant current of 2.1 mA/cm² for 1 hour. The membrane was UV-crosslinked with 1000 μJ in a Stratagene UV crosslinker and hybridised with ³²P-dCTP labelled mi292as probe. The image was acquired on a PhosphorImager instrument.

RNA FISH was performed as described previously 741. pXist, an 18 kb DNA fragment spanning the whole Xist transcript, was labelled using digoxygenin-16-dUTP nick translation mix (Roche) and detected with antidigoxygenin fluorescein isothiocyanate (AD-FITC) antibody raised in sheep (Roche), followed by anti-sheep fluorescein isothiocyanate (FITC) antibody (Vector Laboratories). Images were acquired on a Leica TCS SP5 confocal microscope using LAS AF software.

For whole mount RNA FISH E6.5 embryos were obtained from crosses between mice heterozygous for the Dicer RNase III domain deletion. Embryos were dissected from uteri, rinsed in pre-chilled phosphate buffered saline (PBS), and permeabilised in Cytoskeletal (CSK) buffer for 10 minutes on ice. The washes were carried out in 3 cm Petri dishes kept on ice throughout the whole procedure. The embryos were fixed in 4% formaldehyde for 15 minutes on ice and rinsed in pre-chilled PBS. All aforementioned solutions included 0.1% Tween-20 (Sigma) to prevent the embryos from sticking. Embryos were then dehydrated through an ethanol series (70%, 80%, 90%, 100%). The procedure was performed on a glass slide with a depression (VWR). After the last dehydration wash, ethanol was allowed to evaporate and 15 μl of hybridisation solution, containing DIG-labelled Xist and biotinylated Tsix probes, were added immediately. The slide was covered with a coverslip, sealed with rubber cement, and the hybridisation was performed overnight at 37°C. The Xist probe was full-length Xist cDNA and the Tsix probe was a 4.6 kb EcoRI fragment surrounding Tsix major start site. The post-hybridisation washes were as described previously 41 with a modification, which included addition of 0.1% Tween-20 to all solutions. The Xist probe was detected with AD-FITC antibody raised in sheep (Roche), followed by anti-sheep FITC antibody (Vector Laboratories), and Tsix probe was detected with avidin-Texas red (AV-TR) followed by biotinylated anti-avidin antibody and then again with AV-TR. All antibodies were from Vector Laboratories unless otherwise stated. Images were acquired on a Leica TCS SP5 confocal microscope using LAS AF software. After imaging, each embryo was genotyped by PCR to determine sex and Dicer genotype.

Western blot analysis was performed as described previously 31 with some modifications. Briefly, proteins were separated on 8% SDS-PAGE gels and transferred in 1× Transfer buffer (48 mM Tris, 39 mM Glycine, 0.037% SDS, 20% methanol) at 100 mA/gel for 45 minutes, using Bio-Rad semi-dry blotting apparatus. Dnmt3a antibody (working dilution 1:250) and Dnmt3b antibody (WD 1:300) were from Alexa Biosciences; Dnmt1 antibody (WD 1:250) was from Abcam; and LaminB antibody (WD 1:2000) was from Santa Cruz. Enhanced chemiluminescence detection was performed as recommended by the manufacturer (GE Healthcare).

All mouse work was carried out in accordance with the United Kingdom's Home Office regulations under the Animals (Scientific Procedures) Act 1986.