Xist (X-inactive specific transcript) plays a crucial role in X-inactivation. This non-coding RNA becomes upregulated on the X chromosome that is to be inactivated upon differentiation. Previous studies have revealed that although maintenance-type DNA methylation is not essential for X-inactivation to occur, it is required for the stable repression of Xist in differentiated cells. However, it is unknown whether differential de novo methylation at the Xist promoter, which is mediated by Dnmt3a and/or Dnmt3b, is a cause or a consequence of monoallelic expression of Xist. We show that Xist expression is appropriately regulated in the absence of Dnmt3a and Dnmt3b and that a single X chromosome undergoes proper inactivation in mutant females. Our results indicate that a mechanism(s) other than DNA methylation plays a principal role in initiating X-inactivation. We also demonstrate that delayed upregulation of Xistdoes not induce X-inactivation, consistent with a crucial developmental window for the chromosomal silencing.
It is well known that DNA methylation plays an important role(s) in the regulation of gene expression, maintenance of genomic integrity, parental imprinting and X chromosome inactivation (X-inactivation). It is thought that the methylation pattern of the genome is created and stably maintained by a combination of de novo DNA methyltransferases, Dnmt3a and Dnmt3b, and maintenance-type DNA methyltransferase Dnmt1. Targeted disruption of these DNA methyltransferases revealed that the establishment and maintenance of the genomic methylation pattern is essential for normal mammalian development(Li et al., 1992; Okano et al., 1999).
The dose difference of the X-linked genes between males and females is functionally equalized by inactivating one of the two X chromosomes in female somatic cells during early development(Lyon, 1961). Although X chromosomes inherited from father and mother are both active in undifferentiated cells of preimplantation embryos, one of them becomes transcriptionally silenced upon cellular differentiation. This process is regulated by a cytogenetically defined region on the X chromosome called X chromosome inactivation center (Xic) (reviewed by Avner and Heard, 2001). Xic is essential for X-inactivation to occur in cis and involved in both `choice' of X chromosome to be inactivated and subsequent `initiation' of chromosomal inactivation. The inactivated state then spreads both proximally and distally from the Xic. Non-coding RNAs, Xist and its antisense Tsix,which are mapped in Xic, play a crucial role in this process(Brown et al., 1991; Brockdorff et al., 1991; Borsani et al., 1991; Penny et al., 1996; Marahrens et al., 1997; Lee et al., 1999; Lee and Lu, 1999; Lee, 2000; Sado et al., 2001). At the onset of X-inactivation, Xist is upregulated on the future inactive X chromosome and subsequently accumulates on it in cis. However, expression of Xist is repressed on the X chromosome that remains active. Tsix is a negative regulator of Xist in cis. Extinction of Tsix induces Xist accumulation on the X chromosome being subject to inactivation, whereas continued expression of Tsixprevents upregulation of Xist.
Although the mechanism how Xist mediates chromosomal silencing is not fully understood, specific association of Xist RNA along the entire region of the inactive X chromosome implies that it may recruit proteins required for heterochromatinization (Brockdorff, 2002). Recent studies showed that a polycomb group (PcG) protein complex containing Eed and Enx1 (also known as Ezh2), which harbors an activity of methylating histone H3 at lysine 9 and 27 (Cao et al.,2002; Kuzmichev et al.,2002; Müller et al.,2002; Czermin et al.,2002), is localized to X chromosome in Xist-dependent manner in the early phase of inactivation(Mak et al., 2002; Silva et al., 2003; Plath et al., 2003; Erhardt et al., 2003).
DNA methylation has been implicated in the regulation of Xist in differentiated cells (Beard et al.,1995). The promoter region of the transcriptionally active Xist allele on the inactive X chromosome is unmethylated, whereas that of the transcriptionally inactive Xist allele on the active X chromosome is highly methylated (Norris et al., 1994). Studies on Dnmt1-deficient ES cells and embryos suggested that although X-inactivation can occur in the absence of DNA methylation, maintenance of the methylation at the Xist promoter is necessary for its stable repression in differentiated cells(Beard et al., 1995; Panning and Jeanisch, 1996). It is possible, however, that the intact de novo methyltransferases present in Dnmt1-deficient embryos generate minimal methylation patterns required for differential activation of Xist. We previously showed that the Xist locus is extensively demethylated in ES cells deficient for both Dnmt3a and Dnmt3b (Okano et al.,1999). This suggests that the locus is one of the targets for these de novo methyltransferases, which create the differential methylation pattern at the Xist promoter after the blastocyst stage(McDonald et al., 1998). It is of particular interest whether the differential methylation underlies the mechanism of monoallelic expression of Xist at the onset of X-inactivation.
We have studied X-inactivation in mouse embryos deficient for de novo methyltransferases Dnmt3a and Dnmt3b. Although the promoter of Xistis extremely hypomethylated in these embryos, expression of Xist was not affected in the majority of cells. Cytogenetic and molecular analyses indicated that one of the two X chromosomes was properly inactivated,suggesting that de novo methylation is dispensable for the initiation and propagation of X-inactivation. We further demonstrate that delayed upregulation of Xist caused by hypomethylation at the promoter does not induce X-inactivation, supporting the importance of the developmental window for the chromosomal inactivation(Wutz and Jeanisch, 2000).
Materials and methods
ES cells and mice
The male ES cell line deficient for Dnmt3a and Dnmt3b is described previously (Okano et al.,1999). ES cells were induced to differentiate as embryoid bodies by suspension culture in the absence of LIF(Sado et al., 1996).
The conditional alleles of Dnmt3a and Dnmt3b were introduced into mice, which harbored 2 loxP sites on both sides of the catalytic domain (M.O. and E.L., unpublished; a detailed description of these mice will appear elsewhere). These mice were crossed with CAG-cre transgenic mice to uniformly excise the catalytic domain of each methyltransferase. The null mutant alleles thus produced were functionally equivalent to the disrupted alleles previously reported(Okano et al., 1999).[Dnmt3a-/-, Dnmt3b-/-] embryos were obtained at E9.5 from intercrosses between double heterozygotes. For genotyping of the embryos, DNA was prepared from the yolk sac and all the analyses were performed on the embryos proper.
Bisulfite genomic sequencing
Genomic DNA prepared from the embryos proper was digested with EcoRI and treated with urea/bisulfite essentially as previously described (Paulin et al.,1998). Two round PCR was performed for amplification of both Xist and Hprt. For the first cycle, a Pr1/Pr2 pair and an Hprt-bs1/Hprt-bs3 pair were used for Xist and Hprt, respectively. For the second round, Pr2 and Pr3, and Hprt-bs2 and Hprt-bs3 were used, accordingly. Amplified products were cloned using TOPO-TA-cloning (Invitrogen). The thermal conditions for PCR was 94°C for 5 minutes, followed by 25 cycles of 94°C for 30 seconds, 60°C for 30 seconds, 72°C for 1 minute, and final extension at 72°C for 5 minutes. Primers for Xist, Pr1, Pr2 and Pr3, are described elsewhere (McDonald et al., 1998). Primers for Hprt are as follows: Hprt-bs1, 5′-gtg att att tgg gaa ttt ttt ggg aga-3′; Hprt-bs2, 5′-gta tgg tta gta tta ttt ttt tta gaa-3′; and Hprt-bs3, 5′-act cta cta aaa tcc cct taa ctc acc-3′.
RNA-FISH and replication timing analysis
RNA-FISH was performed as described previously(Sado et al., 2001). BrdU was incorporated into E9.5 embryos for 7-8 hours and cytological preparations were made in the same manner as RNA-FISH and stained with Acridine Orange.
Embryos incubated for 3-4 hours in the presence of colcemid were trypsinized and treated with 0.075 M KCl for 8 minutes at room temperature. Cells were fixed by an addition of equal volume of 4% paraformaldehyde, which were then subjected to cytospin for 10 minutes at 800 rpm (Cytospin 2,Shandon). The specimens thus prepared were permeabilized in KCM buffer (120 mM KCl, 10 mM NaCl, 20 mM Tris-HCl pH 7.7) containing 0.1% Triton X-100 for 10 minutes at room temperature. The first antibody against acetylated histone H4(Upstate) and the second antibody against rabbit IgG conjugated with FITC were diluted 100-fold with KCM containing 0.1% Triton X-100.
Total RNA (2.5 μg) was converted to cDNA using random primers. For quantitative study, real-time PCR was carried out using LightCycler-FastStart DNA Master SYBR Green I (Roche). Concentration of MgCl2 was determined at 3 mM for all primer pair sets. Standard curves for each primer pair were drawn using a series of dilution of cDNA generated from wild-type male ES cells. Expression levels of each message were determined as a value relative to the abundance of Gapd. The thermal conditions of LightCycler were as follows: for Hprt, Rps4 and Gapd,94°C for 10 minutes, followed by 34 cycles of 94°C for 10 seconds,55°C for 10 seconds, 72°C for 24 seconds; for G6pd and Pgk1, 94°C for 10 minutes, followed by 32 cycles of 94°C for 10 seconds, 58°C for 10 seconds, 72°C for 24 seconds. Primer sequences are as follows: Pgk1F, 5′-gtt gca gac aag atc cag ac-3′, Pgk1R, 5′-aca ttg ctg aga gca tcc ac-3′; G6pdF,5′-atc atc gtg gag aag ccc tt-3′, G6pdR, 5′-ttc ttc aca tag agg aca gc-3′; Rps4F, 5′-tca tca gca ttg aca aga cc-3′, Rps4R, 5′-ggattc ctt ttc ctc tgg ga-3′; Gapd/F, 5′-atg gcc ttc cgt gtt cct ac-3′, and Gapd/R, 5′-tgt gag gga gat gct cag tg-3′. MusHprtF and MusHprtR have been previously described(Sado et al., 1996).
Amplifications of Xist and Tsix were performed on 1/50 of the cDNA prepared from ES cells and embryoid bodies using Mx23b and MIX20 as described previously (Kay et al.,1993; Sado et al.,1996). Oct3/4 sequence was amplified using the following primers: Oct3/4F, 5′-tgg gtg gat tct cga acc tg; Oct3/4R, 5′-cct tct gca ggg ctt tca tg-3′.
Xist is extensively hypomethylated in[Dnmt3a-/-, Dnmt3b-/-] embryos
Expression of Xist is developmentally regulated(Panning et al., 1997; Sheardown et al., 1997). It is expressed at a low level from all active X chromosomes in undifferentiated cells in both sexes. Following differentiation, it disappears on the single X chromosome in males and the future active X chromosome in females, whereas it becomes upregulated on the future inactive X chromosome and accumulates on it in cis. This transition of Xist expression is a key event at the onset of X-inactivation and de novo methylation at the Xist promoter may prevent upregulation of Xist on the X chromosome that remains active. We previously showed that the promoter of Xist is extensively demethylated in male ES cells deficient for Dnmt3a and Dnmt3b (Okano et al.,1999).
To explore the role for de novo DNA methylation in X-inactivation and also in the regulation of Xist at the onset of X-inactivation, we analyzed mouse embryos homozygous for the disruptions of Dnmt3a and Dnmt3b. All the embryos examined were recovered at embryonic day (E)9.5, when [Dnmt3a-/-, Dnmt3b-/-]embryos are readily identified by gross morphology(Fig. 1A)(Okano et al., 1999). The methylation profile at the Xist promoter was analyzed in[Dnmt3a-/-, Dnmt3b-/-] female embryos by bisulfite genomic sequencing. As expected, the region differentially methylated in wild-type female embryos was extremely hypomethylated in[Dnmt3a-/-, Dnmt3b-/-] females(Fig. 1B). Similarly, the CpG island containing the Hprt promoter was almost completely unmethylated in these embryos, whereas in wild-type female embryos, it showed variable levels of methylation in about half of the molecules(Fig. 1B). The differential methylation at this locus is still being established at E9.5(Lock et al., 1987).
Expression of Xist is not affected in the majority of cells in [Dnmt3a-/-, Dnmt3b-/-] embryos
It has been reported that expression of Xist is regulated by CpG methylation at the promoter in differentiated cells(Beard et al., 1995). We therefore examined the expression of Xist in double mutant male and female embryos. RNA-FISH revealed that ectopic accumulation of XistRNA occurred on one and two X chromosomes in males and females, respectively(Fig. 2A-F). However, such accumulation was found only in a subset of the cells in[Dnmt3a-/-, Dnmt3b-/-] embryos (0-4.9%in males and 3.9-17.7% in females) and the majority of the cell population in female embryos manifested a normal pattern of Xist accumulation despite extensive hypomethylation at the promoter on both alleles(Fig. 2G). This implies that monoallelic expression of Xist can be induced in the absence of differential methylation at the promoter. It is therefore likely that methylation at the Xist promoter is not essential for preventing upregulation of Xist on the future active X chromosome at the onset of X-inactivation. It should be noted that the proportion of cells exhibiting ectopic Xist expression in [Dnmt3a-/-, Dnmt3b-/-] embryos was comparable with that detected in Dnmt1-/- embryos (2-3.5%, data not shown)(Panning and Jeanisch 1996). Although it has been suggested that ectopic expression of Xist may lead to X-inactivation in Dnmt1-/- embryos(Panning and Jeanisch, 1996),we believe, based on the study in [Dnmt3a-/-, Dnmt3b-/-] male ES cells, that it does not induce X-inactivation in [Dnmt3a-/-, Dnmt3b-/-] double mutant embryos (see below).
One of the two X chromosomes is appropriately inactivated in[Dnmt3a-/-, Dnmt3b-/-] embryos
The expression pattern of Xist observed by RNA-FISH implied that X-inactivation took place normally in most cells of[Dnmt3a-/-, Dnmt3b-/-] female embryos. Cytogenetic and molecular analyses were, therefore, carried out to address the activity of each X chromosome in double mutant females. As the inactive X chromosome replicates late in S phase of the cell cycle, we analyzed the replication timing of X chromosomes in [Dnmt3a-/-, Dnmt3b-/-] embryos in both sexes. A single X chromosome in two [Dnmt3a-/-, Dnmt3b-/-] male embryos replicated synchronously with the autosomes in every metaphase spread examined (Fig. 3B). In four[Dnmt3a-/-, Dnmt3b-/-] females,however, one of the two X chromosomes unanimously replicated late in S phase as seen in wild-type female embryos (Fig. 3A,B). One of the two X chromosomes in[Dnmt3a-/-, Dnmt3b-/-] female embryos thus behaved like an inactive X chromosome.
It is also known that acetylated histone H4 is excluded from the inactivated X chromosome (Jeppesen and Turner, 1993), which is a relatively late event in the X-inactivation process (Keohane et al.,1996). Immunostaining with an antibody against acetylated histone H4 revealed that one of chromosomes was hypoacetylated in the mutant female embryos (n=4) as seen in wild-type females(Fig. 3C). Chromosome painting with an X-specific probe confirmed that the unlabeled chromosome was indeed an X chromosome (data not shown).
Finally we performed real time PCR to compare the expression levels of X-linked genes, Rps4, Pgk1, G6pd and Hprt, all of which are subject to X-inactivation, in [Dnmt3a-/-, Dnmt3b-/-] embryos. The expression level of each gene relative to that of Gapd did not show a significant difference between males and females (Fig. 4), suggesting that one of the two alleles at each locus were silenced in [Dnmt3a-/-, Dnmt3b-/-]females. It is worth mentioning that a relative expression level of Pgk1, although appropriately compensated between males and females,was significantly higher in double mutant than wild-type embryos. It is likely that the regulation of basal expression of Pgk1 has been affected by a loss of functional de novo DNA methyltransferases.
Our cytological and molecular studies strongly suggest that[Dnmt3a-/-, Dnmt3b-/-] female embryos properly undergo X-inactivation in almost all cells and that a loss of functional Dnmt3a and Dnmt3b does not impair either the initiation of X-inactivation or the propagation of the inactivated state along the X chromosome.
Prolonged culture of [Dnmt3a-/-, Dnmt3b-/-] male ES cells causes derepression of Xist
In the above experiments, however, a small proportion of cells in[Dnmt3a-/-, Dnmt3b-/-] embryos did show ectopic accumulation of Xist(Fig. 2A-F). As previous studies on Dnmt1-deficient ES cells suggested that DNA methylation is required for stable repression of Xist(Panning and Jeanisch, 1996),the ectopic Xist accumulation in the [Dnmt3a-/-, Dnmt3b-/-] embryos may have arisen from delayed activation, not from failure in initial repression. To address this issue, we made use of male ES cells deficient for both Dnmt3a and Dnmt3b (Okano et al.,1999). RNA-FISH was performed on these male ES cells before and after induction of differentiation. Although the single X chromosome was never coated with Xist RNA in undifferentiated state, ectopic Xistaccumulation was observed in about 3.2% and 16.8% of cells at day 2 and day 5 of differentiation, respectively (Fig. 5B; data not shown). This is in agreement with the results obtained with the E9.5 [Dnmt3a-/-, Dnmt3b-/-] embryos (0-17.7%). At day 12 of differentiation, however, a surprisingly high percentage (68%) of cells from[Dnmt3a-/-, Dnmt3b-/-] embryoid bodies showed ectopic Xist accumulation(Fig. 5A,B), suggesting progressive activation of the unmethylated Xist locus in the differentiated [Dnmt3a-/-, Dnmt3b-/-]cell population. In control experiments, Xist accumulation was never detected in either Dnmt3a-/-, Dnmt3b-/- or parental wild-type ES cells, but a subset of Dnmt1-/- cells showed ectopic Xist accumulation(Fig. 5B), in agreement with the previous report (Panning and Jeanisch,1996). These observations confirm the importance of DNA methylation in stable repression of Xist(Panning and Jeanisch,1996).
Ectopic accumulation of Xist in differentiated[Dnmt3a-/-, Dnmt3b-/-] cells does not induce X-inactivation
A previous study on Dnmt1-/- ES cells suggested that ectopic Xist accumulation causes aberrant X-inactivation and subsequent cell death and that this may explain why ectopic Xistaccumulation was detected only in a small proportion of the mutant cells(Panning and Jeanisch, 1996). However, in our experiment, cells with ectopic Xist accumulation predominated in the differentiated [Dnmt3a-/-, Dnmt3b-/-] male ES cell population, implying that X-inactivation was not induced. If ectopic X-inactivation took place in the mutant male ES cells, these cells would die upon differentiation because of functional nullisomy for X chromosome. We addressed the issue by examining whether selective death of [Dnmt3a-/-, Dnmt3b-/-] cells occurs in a 1:1 mixture of wild-type and the mutant cells. We took advantage of the deletion introduced at the Dnmt3b locus to compare the proportion of wild-type and mutant cells,which was visualized by Southern blotting as a restriction fragment length difference. The intensities of the bands derived from the wild-type and[Dnmt3a-/-, Dnmt3b-/-] ES cells appeared almost identical before and after differentiation (up to day 12)(Fig. 6A), indicating no selective elimination of the mutant cells.
We then examined whether X-inactivation occurs in[Dnmt3a-/-, Dnmt3b-/-] male ES cells upon induction of differentiation. We amplified X-linked Pgk-1, G6pdand Rps4 by RT-PCR and, as shown in Fig. 6B, none of them showed drastic decrease in expression over the course of differentiation, despite the fact that a significant proportion of the cells exhibited Xistaccumulation with downregulation of antisense Tsix. This suggests that X-inactivation did not occur in the differentiated mutant ES cells. The absence of a late replicating X chromosome in these differentiated cells further supported this notion (n=25, data not shown). These results strongly suggest that delayed accumulation of Xist RNA ectopically expressed in [Dnmt3a-/-, Dnmt3b-/-]male ES cells does not trigger aberrant X-inactivation, and this is most probably the case for [Dnmt3a-/-, Dnmt3b-/-] embryos. These findings support the previous proposal that Xist RNA needs to be expressed in a critical window of differentiation to initiate X-inactivation(Wutz and Jeanisch, 2000).
We studied the role for de novo methylation in X-inactivation using mouse embryos and ES cells deficient for de novo methyltransferases Dnmt3a and Dnmt3b. Our cytological and molecular studies demonstrated that one of the two X chromosomes was properly inactivated in [Dnmt3a-/-, Dnmt3b-/-] female embryos, suggesting that X-inactivation can initiate and propagate along the chromosome in the absence of de novo DNA methylation. The de novo methyltransferases thus appear to be dispensable for the conversion of X chromosome into heterochromatin. Recent studies showed that the PcG proteins, Eed and Enx1/Ezh2, are implicated in X-inactivation(Mak et al., 2002; Silva et al., 2003; Plath et al., 2003; Erhardt et al., 2003). Interestingly, Enx1/Ezh2 has a histone methyltransferase activity specific for lysine 9 and 27 in the N-terminal tail of histone H3, which is involved in heterochromatin formation and gene silencing(Cao et al., 2002; Kuzmichev et al., 2002; Müller et al., 2002; Czermin et al., 2002). As methylation of histone H3 at lysine 9 (and probably 27) becomes evident at relatively early phase of the X-inactivation process in female ES cells(Heard et al., 2001; Mermound et al., 2002), it is possible that histone H3 methylation at these residues plays a more pivotal role than DNA methylation in spreading of the inactive state from the Xic.
Previous studies on Dnmt1-/- embryos and ES cells showed that initiation of monoallelic Xist expression and X-inactivation can occur in the absence of maintenance-type DNA methylation(Beard et al., 1995; Panning and Jeanisch, 1996). They did not, however, address whether de novo methylation at the Xist promoter, which should occur in the mutant cells, contributes to the initiation of X-inactivation. Some residual methylation was, in fact,detected at the Xist promoter in Dnmt1-/- embryos(Sado et al., 2000), which was most probably mediated by the de novo DNA methyltransferases. In[Dnmt3a-/-, Dnmt3b-/-] female embryos,despite the lack of de novo methylation at the Xist promoter,accumulation of Xist RNA was confined to a single X chromosome in the great majority of nuclei. The present study, therefore, provides the first evidence that a mechanism(s) other than DNA methylation is responsible for causing the differential expression of Xist. It is clear,nevertheless, that DNA methylation is important for the stable repression of Xist on the active X chromosome in differentiated cells. In agreement with the previous reports on Dnmt1-deficient cells(Beard et al., 1995; Panning and Jeanisch, 1996),the present study indicated that upon differentiation, hypomethylation at the Xist promoter caused by a functional loss of de novo DNA methyltransferases resulted in progressive derepression of Xist. It should be noted, however, that our finding was at variance with the study by Panning and Jaenisch (Panning and Jaenisch, 1996) in that ectopically expressed Xist did not lead to silencing of X-linked genes. While available evidence suggests that Xist expressed in differentiated cells does not cause X-inactivation (Tinker and Brown,1998; Wutz and Jaenisch,2000) (this study), Hall et al.(Hall et al., 2002) recently demonstrated that a human XIST transgene could induce chromosome inactivation in postdifferentiation human HT-1080 cells. Perhaps, ectopic expression of Xist might exert its effect on silencing in some particular cell types.
Our results suggested that X-inactivation was not induced despite the cis-association of Xist on the X chromosome. Taking advantage of an inducible expression system of Xist in transgenic ES cells, Wutz and Jaenisch (Wutz and Jaenisch,2000) previously showed that Xist can initiate chromosomal silencing only during an early phase of ES cell differentiation(up to 48 hours after induction of differentiation), which indicates that there is a crucial developmental window for X-inactivation. Ectopic expression of Xist, therefore, most probably occurs later than this window in[Dnmt3a-/-, Dnmt3b-/-] embryos and male ES cells, thereby resulting in no induction of X-inactivation.
It has been suggested that DNA methylation is important for the stable repression of genes on the inactivated X chromosome. Our previous study on Dnmt1-deficient mouse embryos suggests that substantial loss of DNA methylation leads to partial reactivation of X-linked lacZ transgenes in the embryonic lineage of E9.5 embryos(Sado et al., 2000). By contrast, hypomethylation caused by the failure of de novo DNA methylation did not appear to reactivate the silent copy of the four endogenous X-linked genes in [Dnmt3a-/-, Dnmt3b-/-] females at E9.5. This may be ascribed to different susceptibility to hypomethylation between the endogenous genes and the transgenes. It is possible that the regulation of multicopy exogenous sequences such as a tandem array of the lacZ transgenes is more sensitive to the loss of DNA methylation than the endogenous genes. It will be important to address whether or not the maintenance mechanism of the X-inactivated genes is affected in[Dnmt3a-/-, Dnmt3b-/-] background.
It is known that Xist is exclusively expressed from the paternal allele in each blastomere of early preimplantation embryos(Kay et al., 1993; Sheardown et al., 1997),although the promoter region of Xist shows a low level of CpG methylation on both parental alleles(McDonald et al., 1998). It seems, therefore, possible that distinctive chromatin structures inherited from the parents are responsible for the upregulation of the paternal allele and stable repression of the maternal allele in the preimplantation embryos. It is therefore likely that the modification of chromatin is capable of inducing the differential expression of Xist on its own. Methylation of histone H3 at lysine 9 and 27 and histone H4 at lysine 20 is implicated in repression of gene expression in yeast and fruit fly(Nakayama et al., 2001; Noma et al., 2001; Cao et al., 2002; Nishioka et al., 2002a; Fang et al., 2002), both of which essentially lack DNA methylation. It is tempting to speculate that these histone modifications in combination with methylation of histone H3 at lysine 4, which is involved in transcriptional activation(Strahl et al., 1999; Noma et al., 2001; Wang et al., 2001; Nishioka et al., 2002b),primarily regulate the differential expression of Xist(Fig. 7), which would not be affected in [Dnmt3a-/-, Dnmt3b-/-]embryos. CpG methylation and hypoacetylation(Gilbert and Sharp, 1999) then follow these events to fix the established epigenetic states. As Tsixappears to regulate the expression of Xist negatively at the onset of X-inactivation (Lee, 2000; Sado et al., 2001), it should also play a role in monoallelic upregulation of Xist. It should be noted that Tsix became downregulated in the absence of de novo methyltransferases in differentiating ES cells. The CpG island found in the vicinity of the major transcription start site of Tsix, which became methylated upon differentiation in male ES cells, stayed unmethylated after differentiation in [Dnmt3a-/-, Dnmt3b-/-] male ES cells, suggesting that regulation of Tsix is also independent of DNA methylation (data not shown). Further studies on the epigenetic regulation and function of Tsix may clarify the initial event that triggers X-inactivation.
We thank Minako Kanbayashi for genotyping and breeding of mice, and Chikako Suda for sequencing. This work was supported by grants from PRESTO, Japan Science and Technology Agency (JST) and Grant-in Aid for Scientific Research on Priority Area `Genome Biology' from the Ministry of Education, Science,Sports, and Culture of Japan to T.S.; and by grants CA82389 and GM52106 from the National Institutes of Health to E.L.