A new Xist allele driven by a constitutively active promoter is dominated by Xist locus environment and exhibits the parent-of-origin effects

Among the many long noncoding RNAs (lncRNA) implicated in gene regulation, the X-inactive-specific transcript (Xist) is one of the most extensively studied lncRNAs and its biological significance has been well described (Brockdorff et al., 1992; Brown et al., 1991, 1992; Marahrens et al., 1997; Penny et al., 1996). Xist RNA plays an indispensable role in X chromosome inactivation (X-inactivation) in female mammals, a process by which the dosage difference of X-linked genes between the sexes is compensated for by transcriptionally silencing one of the two X chromosomes during early development (Lyon, 1961). At the onset of X-inactivation, Xist becomes upregulated from one or the other X chromosome and its lncRNA associates in cis with the X chromosome from which it originates, to induce chromosome-wide heterochromatinization.

In the mouse, X-inactivation is imprinted in favor of the paternal X (Xp) in the extra-embryonic lineage, which gives rise to the placenta and some extra-embryonic membranes (Takagi and Sasaki, 1975), whereas it takes place in an essentially random fashion regardless of the parental origin in the embryonic lineage, which gives rise to all tissues of the fetus. Xist becomes monoallelically upregulated on Xp at the 4- to 8-cell stages and this paternal allele-specific expression is maintained in the trophectoderm and primitive endoderm of the blastocyst, both of which belong to extra-embryonic lineages. Xist clouds are lost in cells that have contributed to the epiblast lineage of the inner cell mass (ICM), and the previously inactivated Xp becomes transiently reactivated prior to the subsequent inactivation of either X in differentiating epiblast cells (Mak et al., 2004; Okamoto et al., 2004). Reactivation of the inactive X has also been observed in the primordial germ cells (PGCs) of female embryos (Monk and McLaren, 1981), which takes several days around the time they enter meiosis and accompanies the loss of Xist clouds (Sugimoto and Abe, 2007). The inactive X is thought to be reactivated for the proper progression of meiosis, especially for the pairing of two homologous X chromosomes in the prophase of meiosis I (Sugimoto and Abe, 2007). Although the downregulation of Xist has commonly been observed in the ICM and PGCs, to what extent they share the mechanism for reactivation is not clear. Williams et al. previously showed that reactivation of X-linked genes sometimes precedes the loss of Xist clouds in the ICM, which implies that the downregulation of Xist is not necessarily a primary cause of reactivation, at least in the ICM (Williams et al., 2011).

Although the biological significance of X-reactivation and of the accompanying downregulation of Xist in the ICM and PGCs have not been fully established, it is assumed that these events are associated with global reprogramming, which resets the previous epigenome in the ICM and PGCs and induces the expression of pluripotency factors including those reported to negatively regulate Xist, such as Oct3/4 (Pou5f1 – Mouse Genome Informatics), Sox2 and Nanog (Navarro et al., 2008). These factors may directly or indirectly downregulate Xist. A recent study, however, showed that the deletion of binding sites for these factors present in Xist intron 1 did not appear to affect this process in the ICMs (Minkovsky et al., 2013). A series of events in the process of reprogramming in addition to the expression of pluripotency factors might also contribute to the downregulation of Xist by causing changes in the chromatin structure of the Xist locus or the expression of transcription factors required for continuous Xist expression. Although the downregulation of Xist may not be a prerequisite for X-reactivation, as suggested by Williams et al. (2011), its biological significance cannot be addressed until the effect of continuous presence of Xist clouds in the ICMs and PGCs is examined.

In contrast to Xist, its antisense negative regulator, known as Tsix, is only expressed from the maternal X (Xm) from the morula/blastocyst stage until the midgestation stage in cells that maintain imprinted X-inactivation, namely in the extra-embryonic lineages (Lee, 2000; Sado et al., 2001). The targeted disruption of Tsix, when maternally inherited, was shown to result in the ectopic upregulation of normally silent maternal Xist and subsequent inactivation of Xm in the extra-embryonic lineages of both sexes, which demonstrated the antagonizing effect of Tsix on Xist in cis. Xm was shown to acquire an imprint during the prophase of meiosis I, which rendered Xm resistant against inactivation in the extra-embryonic lineages (Goto and Takagi, 1998; Tada et al., 2000; Takagi and Abe, 1990). By contrast, whether Xp possesses an imprint that facilitates its inactivation in the extra-embryonic lineages remains unclear. The maternal imprint by itself might be potentially sufficient enough to drive imprinted X-inactivation in these particular lineages by leaving Xp as the sole X capable of undergoing inactivation. As Xm carrying the Tsix deficiency has lost the ability to resist inactivation in the extra-embryonic tissues, Tsix expression could be considered to represent a primary output of the imprint laid on Xm. One can argue against this point of view, however, if the maternal imprint is supposed to be responsible for the repression of maternal Xist at the 4- to 8-cell stages as well because maternal Tsix has not yet been expressed at these stages.

In this study, we generated a new Xist allele (Xist^CAG) that was driven by a CAG promoter, which is considered one of the strongest and most ubiquitous promoters (Niwa et al., 1991). We expected that this allele, when paternally inherited, would allow us to unanimously inactivate the targeted Xp in the embryonic lineage by the constitutive expression of Xist^CAG upon the choice of X-inactivation and to also sustain Xist expression in the ICM and PGCs. By contrast, when maternally inherited, we expected that this allele would result in a phenocopy of embryos with maternal deletion of Tsix in both sexes. We found that Xist^CAG behaved as expected, except for the failure to sustain Xist clouds in the ICM and PGCs. Although the unexpected downregulation of Xist^CAG in these cells prevented further investigations into the effects of the continuous expression of Xist in cells undergoing reprogramming and X-reactivation, we identified interesting parental origin specific-differences in the timing of Xist^CAG upregulation in preimplantation embryos. Its expression started at the 4- to 8-cell stages when paternally inherited, whereas Xist^CAG was not upregulated until the blastocyst stage when maternally inherited. Here, we discussed what these results may indicate in terms of the differential regulation of Xist in preimplantation-stage embryos.

A targeting vector was constructed to replace the endogenous Xist promoter with a cassette containing a CAG promoter (Niwa et al., 1991), which is known to be active in many types of cells including germ cells (Okabe et al., 1997; Sakai and Miyazaki, 1997), and a floxed selection marker (Fig. S1A). Following homologous recombination in embryonic stem cells (ESCs), those harboring the mutant allele referred to as Xist^CAG2L were isolated (Fig. S1B). Because the transcription initiated from the CAG promoter could essentially be terminated by the presence of polyadenylation signals located 3′ to the selection marker, we expected that the Xist^CAG2L allele would not produce functional Xist RNA and would behave in the same way as a null allele that was created previously (Marahrens et al., 1997). We therefore crossed the male chimeras with females heterozygous for the Tsix deletion (ΔTsix) (Sado et al., 2001) (see Materials and Methods) and successfully obtained female offspring carrying the Xist^CAG2L allele in combination with the ΔTsix allele (Fig. S1C). They were subsequently crossed with males carrying a Pgk2-cre transgene (Kido et al., 2005), which allowed the expression of cre recombinase specifically in the spermatocyte under the control of the Pgk2 promoter. We expected that males carrying Xist^CAG2L in combination with Pgk2-cre would transmit the Xist^CAG allele to female pups via cre-mediated conversion in the spermatocytes when crossed with wild-type females. As shown in Fig. S1D, we successfully recovered females heterozygous for Xist^CAG, which demonstrated that spermatogenesis was not compromised by the conversion of Xist^CAG2L into Xist^CAG in the spermatocytes. The presence of the respective alleles thus introduced into mice was confirmed by Southern blotting (Fig. S1C). A 197 bp portion surrounding the major transcription start site of Xist (chrX: 100,678,537-100,678,733 on mm9) was replaced with the CAG promoter in the Xist^CAG allele.

It was reasonable to expect the X chromosome undergoing inactivation to be confined to the wild-type X in X^CAG2LX females and the mutated X^CAG in XX^CAG females. (Throughout this article, the maternal X precedes the paternal one, and X^CAG2L and X^CAG refer to X chromosomes carrying Xist^CAG2L and Xist^CAG, respectively.) However, because RNA expressed from the Xist^CAG allele lacked the first 20 nucleotides present in wild-type Xist RNA and possessed an additional unrelated sequence derived from the CAG cassette instead, we examined whether these changes affected the function and processing of RNA produced from the Xist^CAG allele. We derived mouse embryonic fibroblasts (MEFs) heterozygous for the respective Xist mutations using mice with a Robertsonian translocation, Rb(X.9)6H, in which the X chromosome and chromosome 9 were fused at the centromere [X^Rb(X.9)] (Tease and Fisher, 1991). X^CAG2LX^Rb(X.9) and X^Rb(X.9)X^CAG MEFs thus derived allowed us to easily identify the mutated X by its morphology. We carried out RNA fluorescence in situ hybridization (RNA-FISH) to examine the expression of Xist in these mutant MEFs as well as in control XX^Rb(X.9) MEFs (Fig. S2A,B). The results demonstrated that whereas the Xist^CAG2L allele was functionally null and did not produce functional Xist RNA, the Xist^CAG allele constitutively expressed functional RNA coating the X^CAG in cis.

The nature of Xist RNA expressed from X^CAG in MEFs was examined by northern blotting. As shown in Fig. S2C, X^CAG produced apparently the same species of Xist RNA as the wild-type X, which indicated that there were no aberrant transcripts resulting from unexpected splicing or termination. The expression level of Xist in XX^CAG was, however, about tenfold higher than that in XX, which is consistent with the strong activity of the CAG promoter (Fig. S2C). We further compared the stability of the Xist RNA expressed from X^CAG with that from the wild-type X using XX^CAG and XX MEFs, and found that the half-life of RNA expressed from the Xist^CAG allele was essentially the same as that of wild-type Xist RNA (Fig. S2D).

As the inactive X is known to replicate late in the S phase of the cell cycle, we examined when replication of the respective X chromosomes occurred in these MEFs. In XX^Rb(X.9) MEFs, the morphologically normal wild-type X replicated late in 58% of cells, whereas X^Rb(X.9) did so in the remaining 42%, consistent with random X-inactivation. In mutant MEFs, by contrast, the late replicating X was confined to X^Rb(X.9) in X^CAG2L X^Rb(X.9) MEFs, and to X^CAG in X^Rb(X.9)X^CAG MEFs (Fig. S2E,F). These results demonstrated that both mutations caused a skewing of inactivation in favor of the wild-type X in X^CAG2LX and the mutated X in XX^CAG.

We further examined gene expression on X^CAG2L and X^CAG in MEFs prepared from F1 fetuses of the JF1 strain and respective mutant mice. These MEFs allowed us to study the allelic expression of X-linked genes and the result demonstrated that the genes on X^CAG2L were active, whereas those on X^CAG had been uniformly inactivated in the respective mutant MEFs (Fig. S2G). Taking all these results together, we concluded that both of the mutant alleles exerted the expected effect on the inactivation pattern of the X chromosome: the Xist^CAG2L allele rendered X^CAG2L incompetent to undergo inactivation, whereas the Xist^CAG allele constitutively produced functional RNA to silence the chromosome in cis.

Selective inactivation of a genetically manipulated X chromosome was shown to occur in the somatic cells of females heterozygous for X^ΔTsix of a paternal origin. Although these females were apparently normal, both male and female embryos conceived by them died in utero when X^ΔTsix was transmitted owing to the ectopic expression of maternal Xist from the mutated X^ΔTsix in the extra-embryonic tissues after implantation (Lee, 2000; Sado et al., 2001). We crossed XX^CAG females and wild-type C57Bl/6 males to see if Xist^CAG behaved in a manner similar to the Tsix-deficient allele upon maternal transmission. Although XX^CAG females gave birth to healthy pups of both sexes, X^CAG had been barely transmitted to them, which suggested that most of the embryos that had inherited Xist^CAG died in utero (Fig. 1A). In a single female survivor with maternal Xist^CAG, it is possible that the imprinted Xp inactivation in the extra-embryonic tissues was somehow reversed by shutting off paternal Xist at the early postimplantation stage, allowing the embryo to survive despite the expression of maternal Xist^CAG. When embryos were dissected out at embryonic day (E) 7.5 from XX^CAG females, both X^CAGX and X^CAGY embryos, although found at the expected ratio, were stunted with abnormal morphology, which was reminiscent of embryos carrying maternal X^ΔTsix (Fig. 1A,B). The distal part of these embryos, which was enriched with embryonic ectoderm, and their trophoblast were subsequently examined for Xist expression by RNA-FISH. X^CAGX embryos were manifested by single Xist clouds in the embryonic ectoderm, but by two clouds in a large proportion of the nuclei in the trophoblast (Fig. 1C; Fig. 2B). The single clouds in the former were probably derived from maternal Xist^CAG. In X^CAGY embryos, single Xist clouds were found in most of the nuclei in both tissues (Fig. 1C; Fig. 2B). These results demonstrated that X^CAG, when maternally inherited, recapitulated the behavior of the X deficient for Tsix. It is likely that the maternal transmission of Xist^CAG caused embryonic lethality owing to functional nullisomy of the X chromosome in the extra-embryonic lineage of X^CAGX embryos and in both embryonic and extra-embryonic lineages of X^CAGY embryos.

X^CAGX and X^CAGY embryos, although stunted, apparently had undergone gastrulation. We reasoned that if maternal X^CAG became silenced during the preimplantation stages, these mutants would have suffered from the functional nullisomy of the X chromosome and would not have survived until the gastrulation stage. To investigate when maternal Xist^CAG became upregulated and formed the cloud in X^CAGX and X^CAGY embryos during development, we carried out RNA-FISH using embryos recovered at earlier stages from XX^CAG females crossed with wild-type males. The sex of each embryo was subsequently identified by Y chromosome painting (the results of Y-painting were noted visually after staining, but were not recorded as image data). Xist expression was examined in 15 female and 15 male 8-cell-stage embryos recovered at E2.5. We found that whereas the nuclei of one class of female embryos contained only a single large Xist cloud, those of the other class contained very faint scattered signals in addition to a large Xist cloud (Fig. 2A,B; Fig. S3). Similarly, the nuclei of half of the male embryos were negative for Xist, whereas those of the remaining half contained very faint scattered signals (Fig. 2A,B; Fig. S3). Although we could not verify the genotype of the embryos, we suspected that the very faint scattered signals represented very weak expression of Xist^CAG, which contrasted well with the intense signal of the Xist cloud in females. When E3.5 blastocysts were examined, a small subset of the nuclei in 3/7 females and 6/11 males contained two and one Xist cloud(s), respectively (Fig. 2A,B). It would be reasonable to assume that they represented X^CAGX and X^CAGY embryos, respectively. It is, therefore, likely that maternal Xist^CAG had not been fully upregulated in the majority of the blastomeres at this stage. This contrasted well with female blastocysts carrying Xist^CAG on Xp (XX^CAG), which exhibited clear single Xist clouds in a large proportion of the nuclei (Fig. 3A). In fact, Xist^CAG in every blastomere of XX^CAG embryos (n=3) had been upregulated in the same manner as Xist on Xp in wild-type female embryos by the 8-cell stage (Fig. 3A). The prevalence of the nuclei containing an extra Xist cloud, although increased to some extent, was still as low as 20% in presumptive X^CAGX and X^CAGY embryos recovered at E4.5, but reached a level comparable to that seen in the embryonic and extra-embryonic tissues of E7.5 mutant embryos by E6.5 (Fig. 2B). These results demonstrated that the timing of Xist^CAG upregulation differed depending on its parental origin and, when maternally inherited, it started to be expressed at around the blastocyst stage despite the allele being under the control of the CAG promoter. It is likely that ectopic inactivation of Xm did not take place until around the time of implantation in X^CAGX and X^CAGY embryos. This is again reminiscent of the timing when Xm carrying the Tsix deficiency started to be ectopically inactivated (Sado et al., 2001).

Paternal Xist upregulated in every blastomere of the female embryos during the early cleavage stages was shown to be silenced in cells contributing to a part of the ICM (Mak et al., 2004; Okamoto et al., 2004). This accompanied transient reactivation of the hitherto inactive Xp prior to the onset of random X-inactivation taking place in the epiblast lineage. It was of interest whether the expression of Xist^CAG inherited from the father, which was initiated at around the 4- to 8-cell stages, was sustained in cells allocated to the ICM with respect to understanding the biological significance of Xist downregulation in Xp reactivation and the reprogramming of cells in the ICM. Accordingly, we carried out RNA-FISH for Xist expression in combination with immunostaining for ICM lineage-specific markers, either Nanog or Oct3/4, using blastocysts recovered from wild-type females crossed with [X^CAG2LY; Pgk2-cre] males in comparison with wild-type. Among the cells positive for Nanog, ∼10% had already lost Xist clouds in E3.5 wild-type blastocysts. Contrary to our expectation, the proportion of those negative for Xist were comparable in XX^CAG blastocysts and wild-type blastocysts recovered at E3.5 (Fig. 3B,C). When examined at E4.5, the population of cells negative for Xist RNA had become >40% of those positive for Oct3/4 in both cases (Fig. 3B,C). It was likely, therefore, that Xist^CAG upregulated by the 8-cell stage failed to sustain its expression in a subset of ICM cells. These results suggested that paternally derived Xist^CAG, although driven by the CAG promoter, was downregulated in the epiblast lineage in a manner similar to that in wild-type. Although unexpected silencing of Xist^CAG precluded a further analysis for the effect of the continuous association of Xist RNA with Xp on X-reactivation and the reprogramming in the ICM, it raised an interesting possibility that a mechanism involved in the repression of the endogenous Xist promoter on Xp in the epiblast lineage might also be effective on the CAG promoter.

Reactivation of the inactive X also takes place in PGCs around the time when PGCs enter meiosis. Xist clouds are known to be lost in PGCs by E10.5 prior to reactivation of the inactive X (de Napoles et al., 2007; Sugimoto and Abe, 2007). In XX^CAG female embryos, X^CAG should have been uniformly inactivated in every cell including PGCs by the constitutive expression of Xist^CAG. Gonads isolated from E10.5, E13.5 and E16.5 embryos were trypsinized and the cell suspension containing germ cells and the surrounding somatic cells was cytospun onto a slide for RNA-FISH. A GFP transgene driven by the Oct3/4 promoter (Yoshimizu et al., 1999) was introduced into the embryos so that germ cells could be identified by immunostaining with an antibody against GFP. As shown in Fig. 4A,B, Xist RNA disappeared in most of the wild-type PGCs by E10.5, which is consistent with previous studies. By contrast, >70% of PGCs still retained Xist RNA on X^CAG, although this was lower than that of the surrounding somatic cells, in XX^CAG embryos at E10.5 (Fig. 4A,B). Although we expected that Xist RNA might remain on X^CAG in germ cells throughout development, the population of germ cells positive for Xist RNA declined over time to 20% at E13.5 and 0.4% at E16.5 in XX^CAG embryos. As a significant difference in the number of germ cells was not observed between wild-type and XX^CAG embryos at each stage examined, the reduction in the number of germ cells positive for Xist RNA was unlikely to be due to elimination by cell death, but was attributable to the downregulation of Xist^CAG. We carried out qRT-PCR to examine the expression levels of Xist RNA in PGCs isolated from E13.5 XX and XX^CAG gonads. As shown in Fig. 4C, the relative abundance of Xist RNA stably present in PGCs was markedly lower than that in the surrounding somatic cells of the gonads in both XX and XX^CAG fetuses. This suggests that the absence of Xist clouds in RNA-FISH experiments was not due to a loss of localization to the inactive X. As was the case in the ICM, a mechanism that silences the endogenous Xist promoter in PGCs appeared to be effective on the CAG promoter.

It was unexpected that the CAG promoter could not sustain Xist^CAG expression in the ICM and PGCs, given the fact that many transgenes driven by the CAG promoter are expressed in the ICM and PGCs. As the unexpected silencing of transgenes is often associated with promoter methylation, we carried out bisulfite sequencing to examine the methylation levels of the 16 and 59 CpG sites in the endogenous Xist promoter and CAG promoter, respectively, in E13.5 PGCs isolated from XX and XX^CAG gonads by fluorescence-activated cell sorting (FACS) using the GFP transgene driven by the Oct3/4 promoter. In the gonads of XX females, the Xist promoter was methylated at ∼40% in somatic cells surrounding PGCs, consistent with the Xist promoter in somatic cells being methylated on the active X and unmethylated on the inactive X (Fig. 5A) (McDonald et al., 1998; Norris et al., 1994). Interestingly, this region was essentially unmethylated in PGCs, which suggests the possibility that Xist was downregulated in PGCs by either a mechanism independent of promoter methylation if at the transcriptional level or post-transcriptional processing of Xist RNA.

The Xist promoter on the wild-type X was markedly methylated in the gonads of XX^CAG females, whereas the CAG promoter on X^CAG was unmethylated in somatic cells (Fig. 5B). Interestingly, as was the case in XX germ cells, the endogenous Xist promoter on the wild-type X and the CAG promoter on X^CAG were both essentially unmethylated in XX^CAG PGCs. This suggests that the eventual downregulation of the Xist^CAG allele in XX^CAG PGCs was not mediated by promoter methylation. As was the case in the ICM, the Xist^CAG allele appeared to be repressed in the same manner as endogenous Xist in the PGCs of wild-type females.

The reversal of the inactive X chromosome is observed in two different phases in the mouse life cycle. It occurs in cells that have contributed to the ICM in the blastocyst and also in PGCs around the time when they enter meiosis; however, whether the underlying mechanism is common between these cell types is unknown. As reactivation in both cases accompanied the downregulation of Xist, it has been presumed that the loss of Xist RNA might be causally involved in reactivation. A previous study by Williams et al. (2011), however, showed that the reactivation of genes on the inactivated Xp sometimes preceded the disappearance of an Xist cloud in a subset of cells in the ICM, which suggested that X-reactivation was not necessarily coupled with the loss of Xist RNA in the ICM. We assumed that the Xist^CAG allele would allow us to address the effect of the continuous association of Xist RNA with the X chromosome on X-reactivation and the subsequent reprogramming if its expression was sustained in the ICM and PGCs. The results were as expected in somatic cells and its phenotype upon maternal transmission mimicked the phenotype caused by a maternal deficiency in Tsix; however, cells in neither the ICM nor PGCs sustained an Xist cloud formed by the expression of Xist^CAG. Although this was unexpected given the fact that the Xist^CAG allele was driven by the CAG promoter, which is known to be active in many types of cells including those of the ICM and PGCs, these results may imply that even the heterologous CAG promoter, when placed in the context of the Xist locus, comes to be under the control of the same regulatory mechanism as the endogenous Xist promoter. Given the fact that the CAG promoter has been replaced with a 197 bp region surrounding the major transcription start site of Xist, it is likely that the element crucial for the presumed regulatory mechanism for the Xist promoter resides outside the replaced region. It is important to note that both male and female ESCs carrying Xist^CAG could be efficiently derived from blastocysts and that Xist^CAG in such ESCs became upregulated to form an Xist cloud only upon induction of differentiation (Fig. S4), further suggesting that the CAG promoter behaved in a way similar to the endogenous Xist promoter.

Although the transcription of Xist has been correlated with the methylation status of CpG sites in the Xist promoter in somatic cells (Norris et al., 1994), DNA methylation does not seem to be involved in the repression of Xist at the blastocyst stage as the Xist promoter is essentially unmethylated in blastocysts and isolated ICM cells (McDonald et al., 1998). This also appeared to be the case in PGCs as the Xist promoter in E13.5 PGCs, in which an Xist cloud essentially disappeared by E10.5, was largely devoid of DNA methylation. It is tempting to speculate that the presumptive regulatory mechanism, which makes the Xist promoter transcriptionally inert in the ICM and PGCs in a manner independent of DNA methylation, also effectively operates upon the CAG promoter residing at the Xist locus and downregulates Xist^CAG in the ICM and PGCs of XX^CAG embryos. Among the mutated Xist alleles so far introduced into the mouse, that reported by Nesterova et al. (2003), which causes skewed X-inactivation due to selective expression of the targeted alleles, and that reported by Savarese et al. (2006), which is driven by a tetracyclin-inducible promoter, would be of interest to see the behavior of their heterologous promoter at the Xist locus in the ICM and PGCs. The latter allele can be upregulated in undifferentiated ESCs if doxycycline (dox) is administered, whereas the Xist^CAG allele is not as long as the cells are maintained in an undifferentiated state. The presence of abundant transactivator in the dox-inducible cells might explain the difference in the action between these two alleles in undifferentiated ESCs. Nonetheless, it should be mentioned that we could not formally rule out the possibility that Xist clouds were lost in the ICM and PGCs as a result of an alternative mechanism that abolishes the stability of Xist RNA rather than transcriptional repression.

This study also provided another important implication for the asymmetric expression of Xist between Xp and Xm in preimplantation embryos. The timing by which Xist^CAG became upregulated was significantly different depending on its parental origin. Upon paternal transmission, a single Xist cloud derived from Xist^CAG became evident by the 8-cell stage, which was similar to wild-type XX embryos, whereas an extra Xist cloud originating from maternally inherited Xist^CAG in X^CAGX and X^CAGY embryos appeared in a subset of cells at the blastocyst stage. This may be attributed to the difference in the chromatin structure at the Xist locus between Xp and Xm derived from a spermatozoon and oocyte, respectively, in the early cleavage stages of embryos. As the paternal genome undergoes global remodeling of the chromatin soon after fertilization, it may be reasonable to assume that this paternal genome-specific event rendered the Xist locus on Xp more permissive to transcription than that on Xm. If the chromatin structure of the maternal Xist locus needs further cell division to become similar to that of the paternal locus, this would account for the lag in the timing of maternal Xist^CAG being upregulated relative to paternal Xist^CAG. Such a difference in the chromatin structure, if any, may also be relevant to the monoallelic upregulation of Xist in wild-type female embryos in favor of the paternal allele during preimplantation development (Sado and Sakaguchi, 2013). In other words, the maternal Xist allele was repressed during early cleavage stages because it was not yet ready for transcription owing to the compact chromatin structure being inaccessible by transcription factors. Given that the very weak scattered signals of Xist were observed from maternally inherited Xist^CAG, one might raise an alternative possibility that the maternal X chromosome is incompetent for Xist-coating at the early preimplantation stage. Xist-coating detected in parthenogenetically activated embryos, however, argue against this. Recently, Fukuda et al. demonstrated that overexpression of Kdm4b, a histone demethylase for H3K9me3, can induce upregulation of Xist from the 4-cell stage in parthenogenetically activated embryos (Fukuda et al., 2014). This suggests, in fact, that global demethylation of H3K9me3 and subsequent alteration of the chromatin structure facilitate derepression of Xist on Xm otherwise repressed. This is apparently consistent with the idea that maternal Xist is repressed in the early preimplantation embryo primarily owing to the chromatin structure impermissive for transcription. An earlier work demonstrated that Xist expression occurs at the morula/blastocyst stage even in parthenogenones or gynogenones (Kay et al., 1994; Matsui et al., 2001; Nesterova et al., 2001). Although it was originally suggested that this represented erasure of the maternal imprint responsible for Xist silencing by the morula/blastocyst stage, it could be also explained as a result of several rounds of replication or cell division, which could have made the chromatin structure at the maternal Xist locus transcriptionally permissive.

It has been well accepted that an imprint laid on Xm during the prophase of meiosis I makes Xm resistant to inactivation in extra-embryonic lineages (Goto and Takagi, 1998; Tada et al., 2000). As a maternal deficiency in Tsix leads to the ectopic expression of maternal Xist in cis and the subsequent inactivation of Xm in extra-embryonic lineages (Lee, 2000; Sado et al., 2001), the monoallelic expression of maternal Tsix appears to be, at least in part, responsible for the output of the imprint laid on Xm. However, one can argue against this, given that maternal Xist is repressed at the early preimplantation stages, when Tsix has not yet been expressed. Our findings, however, suggest that the reason why the maternal Xist^CAG allele is less effectively transcribed than the paternal one at the early preimplantation stage might be because the chromatin state or structure of the Xist locus on Xm at that time is not ready for robust transcription. If maternal Xist is repressed simply because the chromatin structure is inaccessible to transcription factors, one may not need to assume a germline imprint that prevents the upregulation of maternal Xist during the early cleavage stages. In other words, whether or not the maternal Xist is transcribed at the early preimplantation stage may not be a matter of a conventional imprint, such as differential methylation, but rather simply related to the chromatin state or structure. As Tsix starts to be monoallelically expressed from Xm at around the morula/blastocyst stages, Tsix would be able to block upregulation of maternal Xist thereafter if the presumptive transcriptionally incompetent chromatin structure at the Xist locus on Xm lasts for a few rounds of cleavage until the morula stage. It should be noted that the transcription of Xist^CAG appears to dominate the repressive effect of Tsix on Xm as Xist^CAG is upregulated from the blastocyst stage onwards.

The targeting vector, pCAG-CΔM20, was constructed as described in supplementary materials and methods, and introduced into J1 ESCs (Li et al., 1992). Of the 464 colonies picked up, 11 harbored the expected homologous recombination (Xist^CAG2L). Chimeric males were crossed with females heterozygous for the Tsix-deficient X (XX^ΔTsix) (Sado et al., 2001) to facilitate the transmission of Xp carrying the Xist^CAG2L allele (X^CAG2L) to female offspring as previously shown (Hoki et al., 2009; Ohhata et al., 2008; Sado et al., 2005, 2006). Females thus generated (X^ΔTsixX^CAG2L) were crossed with wild-type males to obtain X^CAG2LX and X^CAG2LY. The floxed puromycin-resistance gene was removed through the male germline using a Pgk2-cre transgene to derive females heterozygous for Xist^CAG (XX^CAG). All mice were maintained and used in accordance with the Guidelines for the Care and Use of Laboratory Animals of Kinki University (KDAS-26-0006). MEFs were prepared as described in supplementary materials and methods.

A DNA probe for RNA-FISH was prepared from 1 µg of pR97E1 (Sado et al., 1996), a plasmid containing an Xist cDNA fragment, by nick translation using the Nick Translation Kit (Abbott) and either Cy3-dUTP (GE Healthcare) or green-dUTP (Abbott).

Chromosome spreads of MEFs were prepared by a standard air-dry method, and cytological preparations of blastocysts and postimplantation embryos were made as described by Okamoto et al. (2000) and Takagi et al. (1982), respectively. RNA-FISH was performed as described previously (Sado et al., 2001). Following examination of the Xist signal, chromosome painting was carried out to distinguish the morphologically normal X and X^Rb(X, 9) (Tease and Fisher, 1991) using painting probes for chromosome X and chromosome 9 (Cambio) according to the manufacturer's instructions. The gender of blastocysts was identified using a painting probe for the Y chromosome after RNA-FISH.

Cells were incorporated with 5-bromo-2′-deoxyuridine (BrdU) for 8 h and chromosome spreads were subsequently prepared by a standard air-drying method for staining with Acridine Orange. The late-replicating inactive X chromosome was identified as a pale chromosome by this method.

MEFs were seeded onto dishes and cultured overnight. The medium was replaced with that containing 50 µg/ml 5,6-dichloro-1-β-D-ribofuranosyl benzimidazole (DRB; Calbiochem) on the following day and cells were collected every 3 h and lysed in ISOGEN (Wako). Five-hundred nanograms of total RNA isolated at each time point was converted into cDNA as described below. Real-time PCR was carried out for the quantification of Xist RNA at each time point using primers XistEx7F21 and XistEx7R20. The level of Xist RNA was normalized relative to that of 18S ribosomal RNA amplified using primers 18SRNA1 and 18SRNA2. The abundance of the respective RNA at each time point was shown as a value relative to the abundance at 3 h after the addition of DRB.

Following the DNase I treatment, RNA prepared from MEFs was converted into cDNA using the Prime Script 1st Strand cDNA Synthesis Kit (Takara) with random hexamer as a primer. PCR was carried out using primers specific for each of the X-linked genes, and amplified products were subsequently digested with appropriate restriction enzymes as described (Kalantry et al., 2009; Sugimoto and Abe, 2007).

Blastocysts recovered at E3.5 and E4.5 were fixed in 4% paraformaldehyde containing 0.1% Triton X-100 for 10 min and subjected to hybridization with an Xist probe overnight. Following serial washes, embryos were postfixed with 4% paraformaldehyde and subjected to immunostaining using antibodies against either Nanog (Yamaguchi et al., 2009; 1:250) or Oct3/4 (N-19) (Santa Cruz, sc8628; 1:400), which were visualized by secondary antibodies, Alexa488-conjugated anti-rabbit IgG (Life Technologies, A-11034; 1:1000) for Nanog or Alexa488-conjugated anti-goat IgG (Life Technologies, A-11055; 1:1000) for Oct3/4.

Gonads isolated from E10.5, E13.5 and E16.5 female fetuses carrying an Oct3/4-GFP transgene (Yoshimizu et al., 1999) were trypsinized and the resultant suspension was subjected to a hypotonic treatment in 75 mM KCl for 10 min and fixed in 2% paraformaldehyde for 10 min on ice. Cells were cytospun for 10 min at 800 rpm (Cytospin 2, Shandon), permeabilized with 0.5% Triton X-100 in PBS for 5 min, and washed with PBS for 5 min. Following hybridization with an Xist probe and washing, cells were stained with an anti-GFP rat monoclonal antibody (Nacalai, 04404-26; 1:100) and an Alexa488-conjugated anti-rat IgG (Life Technologies, A-11006; 1:1000). Images were captured using either an Andor iXon^EM CCD camera or an Olympus DP70 color CCD camera.

Genomic DNA prepared from PGCs at E13.5 was treated with bisulfite and purified using the Bisulfast Kit (TOYOBO) and EZ Kit (Zymo Research), respectively. Two-round PCR was carried out using semi-nested primers. First round PCR was carried out for 25 cycles with 8 ng of DNA using primers Bx-CAG1113F and Bx-CAG1686R for the CAG promoter, and XistPr2 and XistPr3 for the Xist promoter. Second round PCR was performed for 25-30 cycles with one-twentieth of the reaction from first round PCR using primers Bx-CAG1113F and Bx-CAG1646R for the CAG promoter, and XistPr2 and XistPr4 for the Xist promoter. The PCR products were subsequently cloned into the pGEM-Teasy vector (Promega) and sequenced.

Primer sequences used in this study are described in supplementary materials and methods.

We would like to thank Hitoshi Niwa for providing the plasmid containing the CAG promoter. We would also like to thank Hirosuke Shiura for technical advice on immuno-RNA-FISH of blastocysts; Takashi Tada for the kind gift of an anti-Nanog antibody; and Minako Kanbayashi and Michiko Arii for maintenance of the mouse colonies.