Incomplete X-inactivation initiated by a hypomorphic Xist allele in the mouse

Mammals have evolved a mechanism to compensate for the difference in the number of X chromosome genes between XX females and XY males by inactivating one of the two X chromosomes in females (X-inactivation) (Lyon, 1961). In the mouse, despite some contradictory reports (Huynh and Lee, 2003; Mak et al., 2004; Okamoto et al., 2004; Okamoto et al., 2005), several lines of evidence support the idea that X-inactivation takes place at around the 4- to 8-cell stage, with the paternal X chromosome being invariably inactivated in every blastomere (imprinted X-inactivation). At the blastocyst stage, whereas imprinted X-inactivation is stably maintained in the trophectoderm and primitive endoderm, which give rise to the future placenta and some extra-embryonic membranes (Takagi and Sasaki, 1975), the hitherto inactivated paternal X becomes transiently reactivated in those cells that have contributed to the inner cell mass (ICM) (Mak et al., 2004; Okamoto et al., 2004). These epiblastic cells, which give rise to all tissues of the fetus, subsequently undergo inactivation in a random fashion as they differentiate (random X-inactivation) (Monk and Harper, 1979).

An X-linked non-coding gene, Xist (X-inactive-specific transcript) (Borsani et al., 1991; Brockdorff et al., 1991; Brown et al., 1991; Brockdorff et al., 1992; Brown et al., 1992), plays a crucial role in both imprinted and random X-inactivation. Xist RNA expressed from the X chromosome to be inactivated coats almost the entire X chromosome in cis (Clemson et al., 1996) and induces chromosome-wide silencing, most probably by recruiting proteins involved in heterochromatinization. Targeted disruption of the Xist gene has unequivocally shown that the X chromosome deficient in Xist never undergoes X-inactivation (Penny et al., 1996; Marahrens et al., 1997). Interestingly, conditional deletion of Xist revealed that X-inactivation is maintained even in the absence of Xist RNA in embryonic fibroblasts, indicating that Xist is dispensable for the maintenance of the inactive state once established (Csankovszki et al., 1999). A subsequent study further demonstrated that there is a developmental window within which X-inactivation shifts from reversible to irreversible and that the former depends on the association of Xist RNA with the inactive X whereas the latter no longer requires it (Wutz and Jaenisch, 2000). None of the targeted mutations of Xist created to date allows further analysis of how Xist RNA induces chromosome silencing as they are defective in initiating the X-inactivation process (Penny et al., 1996; Marahrens et al., 1997; Csankovszki et al., 1999; Sado et al., 2005; Hoki et al., 2009). One reason why the molecular mechanism of Xist RNA-mediated silencing is still largely unclear nearly two decades after the discovery of the Xist gene is probably the lack of a mutation that compromises the function of Xist RNA. Previously, we attempted to express Xist RNA defective in silencing in mice by simply removing the A-repeat, which has been shown to be essential for the silencing function of the RNA (Wutz et al., 2002) and is encoded in exon 1 of Xist. It turned out, however, that deletion of the A-repeat unexpectedly abolished transcription of the Xist gene and did not allow expression of dysfunctional Xist RNA (Hoki et al., 2009).

Here we describe a partial loss-of-function mutation of Xist (Xist^IVS) found among the targeted Xist alleles that we have generated (Ohhata et al., 2008). Although Xist^IVS is differentially upregulated and its transcript coats the X chromosome in cis in embryonic and extra-embryonic tissues, X-inactivation thus initiated does not seem to be fully established. The RNA expressed from Xist^IVS therefore seems to be dysfunctional. Xist^IVS is likely to be a novel mutation and provides an opportunity to dissect the molecular mechanisms by which Xist RNA induces chromosome-wide silencing.

The generation of mice carrying the Xist^IVS allele has been described previously (Ohhata et al., 2008). These mice have been maintained by crossing females heterozygous for Xist^IVS with C57Bl/6 males.

Total RNA prepared from the trophoblast of individual E7.5 embryos or 1 μg of total RNA prepared from trophoblast stem (TS) cells was converted into cDNA using random primers. Subsequent PCR was carried out using gene-specific primer sets.

For the half-life assay of Xist RNA, total RNA was prepared from the respective TS cell lines treated with 5,6-dichloro-1-β-d-ribofuranosylbenzimidazole (DRB) and converted into cDNA using the oligo(dT) primer. cDNA thus synthesized at each time point was subsequently used for quantitative PCR as described previously (Hoki et al., 2009).

For northern blot analysis, 20 μg of total RNA was electrophoresed, blotted onto a nylon membrane and probed with an Xist cDNA fragment corresponding to nucleotides 37-416 (GenBank accession NR_0014633) and a Gapdh cDNA fragment.

A Cy3-labeled probe was prepared by nick translation using pR97E1 (Sado et al., 1996). Cytological preparations of blastocysts (Okamoto et al., 2000) and the embryonic ectoderm (Takagi et al., 1982) were made as described previously. Those of TS cells were made by the conventional air-drying method. Hybridization and subsequent washing were carried out by standard methods.

For immunofluorescence microscopy, TS cells grown on a coverslip were fixed with 4% paraformaldehyde, permeabilized in PBS containing 0.5% Triton X-100, and placed in PBS containing 0.5% BSA for blocking. The distribution of H3K27me3 was visualized using a mouse monoclonal antibody against H3K27me3 (Ohhata et al., 2008) and either an Alexa Fluor 488-labeled (Invitrogen) or a CF 594-labeled (Biotium) secondary antibody against mouse IgG. For immunostaining acetylated histone H4 and H3K4me2, rabbit polyclonal antibodies against acetylated histone H4 (Upstate, 06-598) and H3K4me2 (Upstate, 07-030) were used, respectively, and subsequently visualized by an Alexa Fluor 488-labeled secondary antibody against rabbit IgG.

E6.5 and E7.5 embryos in the decidua and E13.5 placentas fixed in Bouin's fixative were dehydrated, embedded in paraffin, sectioned at 6 μm and stained with Hematoxylin and Eosin.

E12.5 placentas fixed in 4% paraformaldehyde were dehydrated, embedded in paraffin and sectioned at 6 μm. Both sense and antisense RNA probes labeled with Digoxygenin were prepared using plasmids (pPL1R and pPL1, respectively) containing an 844 bp cDNA fragment of Pl-1 by in vitro transcription with T7 RNA polymerase.

E7.5 embryos from crosses between C57Bl/6 females and males carrying either an X-linked lacZ transgene (Tan et al., 1993) or Xist^IVS in combination with the same lacZ transgene on the same X chromosome were stained for lacZ expression using X-gal as described previously (Sado et al., 2000).

TS cells were derived from blastocysts from crosses between DBA/2 females and either C57Bl/6 or Xist^IVS/Y males according to Oda et al. (Oda et al., 2006). One wild-type female TS cell line (DD1) and three +/Xist^IVS TS cell lines (T3, T5 and T9) were established. For induction of differentiation, TS cells that had been cultured in conditioned medium (Oda et al., 2006) in the presence of FGF4 and heparin on a gelatin-coated dish overnight were fed with TS medium without FGF4 and heparin for 6 days. All TS cell lines were used for RNA-FISH and immunofluorescence. For analysis of the timing of chromosomal replication, analysis of the half-life of Xist RNA and the microarray, DD1 and T5 were used.

TS cells were labeled with 100 μg/ml 5-bromodeoxyuridine (BrdU) for 13.5-14 hours, with the final hour in the presence of 100 ng/ml colcemid. Metaphase spreads were prepared by a routine air-drying method and chromosomes were stained with freshly prepared Acridine Orange.

For the microarray analysis, a custom 44K DNA microarray (Agilent Technologies, Santa Clara, CA, USA) was used. Probe sequences of this array are described in our previous study (Kobayashi et al., 2010). Total RNA was isolated from TS cells using TRIzol (Invitrogen) and labeled with Cy3-CTP using the QuickAmp labeling kit (Agilent Technologies). Hybridized slides were scanned using a microarray scanner (Agilent Technologies) and the hybridization signals were converted to numerical data with Feature Extraction software version 10.5.1.1 (Agilent Technologies). Normalization of the data was achieved using Gene Spring GX11 software (Agilent Technologies). Statistical analysis of the data was performed with the R statistical package (www.r-project.org/). Microarray data have been deposited at Gene Expression Omnibus with accession number GSE27373.

The following primer sets were used for PCR: R700P2 and F1063AS (Sado et al., 2005) for the allelic expression analysis of Xist^IVS and wild-type Xist; Rbm3F5 (5′-GCCTTGGTGCTAATTATTGCC-3′) and Rbm3R5 (5′-CAAGGACATCGCAATCCTTTA-3′) for Rbm3; HprtF3 and HprtR2 for Hprt, Fmr1F7 and Fmr1R7 for Fmr1, G6pdF3 and G6pdR3 for G6pd, all as described by Sugimoto and Abe (Sugimoto and Abe, 2007); and XistEx7F31 and XistEx7R20 for the analysis of Xist expression by quantitative PCR (Hoki et al., 2009).

We previously generated the Xist^IVS allele by gene targeting for a functional study of Tsix (Ohhata et al., 2008). This allele harbors a ~1 kb intron derived from the human γ-globin gene ~0.9 kb downstream of the major transcription start site of mouse Xist (GenBank accession NR_001463) (Fig. 1A). Since wild-type females crossed with males hemizygous for Xist^IVS (X^IVSY) gave birth to few female pups (Fig. 1B), we initially concluded that the Xist^IVS allele was dysfunctional, like the Tsix^pA allele created in parallel (Ohhata et al., 2008), which harbored the same intron containing multiple polyadenylation signals in a sense orientation relative to Tsix at the same position (Fig. 1A). It turned out, however, that female embryos carrying the paternal Xist^IVS allele (XX^IVS), when examined at embryonic day (E) 7.5 and E9.5, were essentially indistinguishable from wild-type male littermates (Fig. 1C). There would have been no opportunity for these embryos to develop to these stages if imprinted X-inactivation had been completely abolished in the extra-embryonic tissues. This finding therefore strongly suggested that, in contrast to the null mutant alleles reported previously (Marahrens et al., 1997), Xist^IVS could initiate imprinted inactivation of the paternal X. It was interesting, however, that morphologically abnormal embryos started to be recovered from E10.5 onwards and some of them still exhibited fairly normal morphology even at E12.5, whereas others were severely affected by this stage (Fig. 1C). It should be noted that the placentas of XX^IVS female fetuses were significantly smaller than those of wild-type male littermate fetuses. These phenotypes were never observed in heterozygous females carrying the maternal Xist^IVS allele (X^IVSX), indicating that there were some defects in imprinted X-inactivation in XX^IVS female fetuses. It is therefore likely that although Xist^IVS is apparently able to initiate imprinted X-inactivation, it is not fully functional and compromises proper development of the placenta at the midgestation stage.

Upon transcription of the Xist^IVS allele, the intron introduced into exon 1 was expected to be spliced out to produce transcripts (Xist^IVS RNA) essentially the same as those transcribed from the wild-type allele, except for the addition of an unrelated 16 nucleotide sequence derived from the targeting vector at the junction after splicing (Fig. 2A). To examine whether the Xist^IVS allele was transcribed and the introduced intron was removed by splicing in vivo, we utilized the E7.5 trophoblast, in which X-inactivation is imprinted and, therefore, Xist is expressed only from the paternal allele, for RT-PCR amplification of the region spanning the intron. The fragment amplified in XX^IVS was slightly longer than those amplified in the wild type and X^IVSX (Fig. 2B). Direct sequencing confirmed the presence of the additional 16 nucleotide sequence in the longer fragment unique to XX^IVS. In addition, the introduction of the additional intron in exon 1 did not perturb splicing between the endogenous introns (data not shown). The same results were obtained in the visceral yolk sac endoderm of E12.5 fetuses, in which X-inactivation is also imprinted (data not shown). We also carried out a northern blot analysis using XX and XX^IVS trophoblast stem (TS) cells (see below), in which Xist is exclusively expressed from the paternal allele, to examine whether the introduced intron causes improper splicing. We did not detect any bands of unexpected size in XX^IVS TS cells (Fig. 2C). These results demonstrated that the Xist^IVS allele was not only transcribed to produce RNA that underwent the expected splicing, but was also subject to proper imprinting in the tissues in which X-inactivation is imprinted.

We also carried out RNA-FISH in the blastocyst to see whether Xist expression was appropriately upregulated during the preimplantation stage. It is known that Xist expression is initiated at around the 4- to 8-cell stage (Okamoto et al., 2004) and is confined to the paternal allele in preimplantation embryos (Kay et al., 1994). The Xist clouds observed in XX^IVS blastocysts were indistinguishable in shape and intensity from those observed in wild-type blastocysts (Fig. 2D). It is likely, therefore, that imprinted X-inactivation is properly initiated by upregulating the Xist^IVS allele in XX^IVS embryos during the preimplantation stage.

Since morphologically abnormal XX^IVS fetuses started to be recovered at ~E10.5, when the growth of the fetus starts to primarily depend upon maternal nutrition via the placenta, it seemed reasonable to assume that the lethality was due to a malfunction in the placenta. We carried out a histological analysis of the mutant placentas, which were moderately and severely affected at E13.5. Trophoblast giant cells were sporadically present between the spongiotrophoblast layer and maternal decidua in the placenta of wild-type male littermate fetuses (Fig. 3A). By contrast, some of the trophoblast giant cells in the moderately affected placentas appeared to be necrotic and, furthermore, those in the severely affected placentas were essentially missing (Fig. 3A). The lack of trophoblast giant cells in the severely affected placentas apparently resulted in detachment of the inner spongiotrophoblast from the outer maternal decidua of the placenta. A significant loss of trophoblast giant cells in the mutant placenta was also confirmed by in situ hybridization using the trophoblast giant cell-specific marker placental lactogene-1 (Pl-1; Prl3d1 – Mouse Genome Informatics) (Faria et al., 1991) (Fig. 3B). There were significantly fewer cells positive for Pl-1 in the mutant than wild-type placenta. Other trophoblast subtypes, such as the spongiotrophoblast and labyrinthine trophoblast, did not show any discernible defects by this stage. It is therefore likely that the lethality among XX^IVS fetuses is primarily due to a substantial loss of trophoblast giant cells in the placenta.

The fact that placental anomalies were observed only when the Xist^IVS allele was transmitted from the father strongly suggested defects in imprinted X-inactivation in XX^IVS fetuses. Given the gross morphology of XX^IVS embryos at E7.5, however, it would be unlikely that imprinted X-inactivation in the extra-embryonic tissues was severely compromised in XX^IVS embryos from the beginning. It is possible that imprinted X-inactivation initiated by the Xist^IVS allele was incomplete but could be tolerated during the initial differentiation of the extra-embryonic tissues and allowed successful implantation and subsequent development until the midgestation stage. Alternatively, imprinted X-inactivation was properly initiated but reactivation occurred in a subset of cells in the extra-embryonic tissues, such as the trophoblast giant cells, owing to a failure to maintain the inactive state. A combination of these two scenarios is also possible.

To explore these possibilities, we took advantage of a lacZ transgene on the paternal X chromosome (Tan et al., 1993). This lacZ transgene, although not expressed in the trophoblast, has been shown to be a useful reporter to monitor the activity of the X chromosome at this stage in embryos (Tan et al., 1993; Tam et al., 1994; Sado et al., 2000; Sugimoto et al., 2000). Upon paternal transmission, the lacZ transgene on the wild-type X was repressed in the extra-embryonic tissues of E7.5 embryos (Fig. 4A), consistent with imprinted inactivation of the paternal X. By contrast, the same lacZ transgene failed to be repressed when transmitted on the mutated X^IVS from the father (Fig. 4A). This suggested that the mutated X^IVS was not fully inactivated in extra-embryonic tissues.

We examined the expression of some endogenous X-linked genes on the paternal X^IVS by RT-PCR using the trophoblast of E7.5 F1 hybrids from crosses between JF1 females and either wild-type or X^IVSY males, so as to distinguish the parental origin of the expressed allele using restriction site polymorphisms between JF1 and laboratory strains such as C57BL/6 and 129/SvJ. The expression of Hprt, Fmr1 and G6pd on the paternal X^IVS had not been properly repressed in X^JFX^IVS by the time that those on the wild-type X had been largely repressed in X^JFX^lab (Fig. 4B). The expression analyses of these endogenous X-linked genes and of the X-linked lacZ transgene suggest that the inactivation status of the mutated X^IVS is incomplete in the extra-embryonic tissues.

Since the extra-embryonic tissues provide a good opportunity to assess the silencing function of Xist^IVS by introducing it from the father, we derived TS cells from XX and XX^IVS blastocysts for further analysis. TS cells, which originate from the trophectoderm, can be induced to differentiate into polyploidal trophoblast in vitro (Tanaka et al., 1998). It has been shown that, in contrast to the situation in female embryonic stem (ES) cells, which retain two active X chromosomes, the paternal X in female TS cells is preferentially inactivated (Mak et al., 2002; Kunath et al., 2005).

We first examined the timing of X chromosome replication in XX and XX^IVS TS cells by BrdU incorporation and staining with Acridine Orange. The inactive X replicates late in S phase and is visualized as a pale chromosome by this method. One of the two X chromosomes replicated asynchronously in 96.8% of XX TS cells and in 80.0% of XX^IVS TS cells (Fig. 5A). The X^IVS in the majority of female TS cells was, therefore, classified as a typical inactive X by conventional cytogenetic analysis.

We then carried out RNA-FISH to examine the expression of Xist in TS cells before and after induction of differentiation. Xist clouds observed in XX TS cells were essentially unchanged upon differentiation (Fig. 5B). By contrast, although the Xist clouds in XX^IVS TS cells were indistinguishable from those of XX TS cells before differentiation, they were lost or diminished in a subset of nuclei after differentiation for 6 days (Fig. 5B,C). Similar results were obtained in other, independently established XX^IVS TS cells (data not shown). These results suggest that upon differentiation, a subset of differentiated XX^IVS TS cells lose Xist RNA that had initially coated the mutated X^IVS.

Immunofluorescence analysis was also carried out in TS cells using an antibody against histone H3 trimethylated at lysine 27 (H3K27me3), a known epigenetic modification on the inactive X (Plath et al., 2003; Silva et al., 2003). Whereas there was no change in the proportion of cells with H3K27me3 on the inactive X among XX TS cells before and after differentiation, a significant population of differentiated XX^IVS TS cells lost the H3K27me3 mark for the inactive X (Fig. 5D,E). This loss of H3K27me3 on the X^IVS might be relevant to the reduction in the levels of PRC2 (polycomb repressive complex 2) on the inactive X that is normally observed during the differentiation of TS cells (Silva et al., 2003). The observed loss of Xist RNA and H3K27me3 in differentiated XX^IVS TS cells implies that the apparently inactive X^IVS in undifferentiated TS cells becomes reactivated in a subset of the population during differentiation.

It was important to determine whether the inactive state of the mutated X^IVS had been fully established in undifferentiated TS cells. We compared the expression of Xist in XX and XX^IVS TS cells by quantitative RT-PCR (qRT-PCR). Although the level of Xist RNA did not differ significantly between XX and XX^IVS TS cells before differentiation, it was diminished in the latter following differentiation (Fig. 5F), consistent with the findings obtained by RNA-FISH.

Since Xist RNA expressed from the Xist^IVS allele contains an additional 16 nucleotide sequence, which could conceivably affect the stability of the RNA, we compared the half-life of the Xist RNA expressed in XX and XX^IVS TS cells using DRB, an inhibitor for RNA polymerase II. There was no significant difference in the stability of the wild-type and Xist^IVS RNAs (Fig. 5G).

We examined the global expression levels of X-linked genes in XX and XX^IVS TS cells. A microarray analysis demonstrated that there was no significant difference in the average expression levels of X-linked genes between the cells before differentiation (Fig. 6A). Forty-five out of 572 probes (7.9%) on the array, however, showed a more than twofold increase in expression in XX^IVS as compared with the wild type (see Table S1 in the supplementary material; all data can be accessed at GEO with accession GSE27373). Among these genes, Car5b and Sms were further examined for their expression levels in wild-type and mutant TS cells before and after differentiation by qRT-PCR. Consistent with the microarray analysis, their expression levels were in fact higher in XX^IVS than XX TS cells both before and after differentiation (see Fig. S1 in the supplementary material). In addition, those X-linked genes that were affected in XX^IVS TS cells did not form a cluster in a particular region on the X chromosome, such as in the vicinity of the Xist locus, but were dispersed along the chromosome (Fig. 6B).

Since the increased expression of X-linked genes in the mutant TS cells could result from a failure to establish hypoacetylation of histone H4 or hypomethylation of H3K4 on X^IVS, we carried out immunostaining for these modifications in combination with an antibody against H3K27me3 to see whether an H4 acetylation or H3K4 methylation hole underlying the inactive X domain is detected in undifferentiated cells. In both XX and XX^IVS cells, the H3K27me3 domain was essentially excluded from both modifications (see Fig. S2 in the supplementary material).

All these results suggest that although the inactive X^IVS in the mutant TS cells manifests many characteristics of an inactive X chromosome in an undifferentiated state, the inactive state is not comparable to that in wild-type TS cells. Furthermore, an increase in the global expression levels of the X-linked genes became evident in XX^IVS cells upon differentiation for 6 days (Fig. 6A). Although there was no propensity for a particular distribution on the X chromosome (Fig. 6B), 20% of the probes (117/572) were upregulated more than twofold in XX^IVS cells as compared with XX cells. Since a subset of these genes were expressed at equivalent levels prior to differentiation, it seemed likely that genes initially repressed on the X^IVS progressively regained their transcriptional activity (see Table S2 in the supplementary material; all data can be accessed at GEO with accession GSE27373). We further examined the expression levels of two of these genes, Tsc22d3 and Dmrtc1a, by qRT-PCR and confirmed that whereas the differences between XX and XX^IVS cells were less significant before differentiation, they became obvious after differentiation (see Fig. S2 in the supplementary material). Taking all these results together, we suggest that the Xist^IVS RNA expressed from the Xist^IVS allele is not fully functional for establishing the inactive state.

We previously showed that Xist is exclusively expressed from the wild-type X in E13.5 heterozygous female fetuses carrying maternal Xist^IVS (X^IVSX) (Ohhata et al., 2008), suggesting that random X-inactivation is extremely biased in favor of the wild-type X. The same was true for those heterozygous females carrying paternal Xist^IVS that exhibited fairly normal development (data not shown). It should be noted, however, that the GFP-positive cells detected in E7.5 X^GFPX^IVS embryos, which carried a GFP transgene on the wild-type maternal X (Nakanishi et al., 2002) in combination with Xist^IVS on the paternal X, appeared to be progressively lost toward the midgestation stage (Fig. 7A). This contrasted with wild-type embryos carrying the same transgene on the maternal X (X^GFPX), where a significant population of the cells remained positive for GFP, reflecting stable maintenance of the X-inactivation pattern determined early on in a random fashion. It was therefore likely that the nonrandom inactivation of the wild-type X in female fetuses heterozygous for Xist^IVS was a result of selection against those cells that have chosen X^IVS to be inactivated at the onset of random X-inactivation, suggesting that X-inactivation initiated by Xist^IVS in the embryonic tissue was also defective.

We then asked whether X-inactivation was initiated in the embryonic tissue of the X^IVSX^IVS homozygous mutant. The gross morphology of E7.5 X^IVSX^IVS embryos showed severe developmental anomalies compared with wild-type and XX^IVS embryos (Fig. 7B). The histology of the presumptive homozygous mutants further revealed that the embryonic ectoderm, which is derived from the epiblast, was only poorly formed at E6.5 and had not developed further when examined at E7.5 (Fig. 7C). The developmental anomalies observed in X^IVSX^IVS embryos were most probably due to improper X-inactivation.

We examined the expression of Xist^IVS by RNA-FISH using the distal part of the embryo recovered at E7.5, which mainly consists of the embryonic ectoderm. Although apparently normal Xist clouds were detected in the nuclei of embryonic ectoderm cells in X^IVSX^IVS embryos, there were significantly fewer nuclei positive for Xist RNA in X^IVSX^IVS than in the wild type (Fig. 7D). By contrast, only a slight reduction in nuclei positive for an Xist cloud was observed in the trophoblast of X^IVSX^IVS embryos (Fig. 7D). This was consistent with the fact that the defects caused by the paternal transmission of Xist^IVS in the extra-embryonic tissues of XX^IVS embryos had not yet become evident at this stage. The presence of single Xist clouds in the majority of the nuclei in the embryonic ectoderm demonstrates that X-inactivation has been initiated in the embryonic lineage of X^IVSX^IVS embryos. The absence of Xist clouds in 30% of the population in the embryonic ectoderm of the homozygous mutant implies that the Xist^IVS RNA that had accumulated on the X chromosome early on was lost over time, as was seen in differentiated XX^IVS TS cells, and seems to be causally involved in the developmental defects of the embryonic ectoderm. We cannot rule out the possibility, however, that transcriptional initiation of Xist is reduced in those cells that lack Xist clouds, which fail to initiate X-inactivation and result in the observed developmental anomalies.

Most of the targeted mutations of Xist reported to date are functionally null, except that in which exon 4 is deleted (Xist^Ex4del), which affects neither the initiation of X-inactivation in cis nor the random choice of the X to be inactivated in heterozygous females (Caparros et al., 2002) and the recently reported mutation that harbors a targeted inversion in exon 1 (Xist^INV) (Senner et al., 2011). In this study, we have provided evidence that one of the Xist mutant alleles we created by gene targeting, Xist^IVS, behaves like a partial loss-of-function mutation or a hypomorphic allele. In contrast to the null mutation reported previously (Marahrens et al., 1997), the paternal transmission of Xist^IVS does not compromise early post-implantation development of XX^IVS embryos, which have apparently healthy extra-embryonic tissues that are indistinguishable from those of wild-type male littermate embryos. These embryos, however, never survive beyond the midgestation stage, most probably owing to placental defects, in particular, the selective loss of trophoblast giant cells.

In addition, a subset of endogenous genes as well as a lacZ transgene on the paternal X^IVS are not appropriately repressed in the extra-embryonic lineages of E7.5 embryos and such misexpression of the paternal alleles becomes more prominent at later stages (data not shown). Further analysis using TS cells, which are known to carry an inactive X chromosome of paternal origin (Mak et al., 2002; Kunath et al., 2005), demonstrated that although the expression of one of the two X chromosomes in XX^IVS cells, most probably X^IVS, was manifested by late replication, and that there was an Xist RNA coating and an enrichment of H3K27me3 in the majority of the population, a subset of X-linked genes was expressed at higher levels in XX^IVS than in wild-type XX TS cells. Differences in the global expression levels of X-linked genes between XX^IVS and XX TS cells became prominent upon induction of differentiation.

Intriguingly, the differentiation of XX^IVS TS cells accompanies a reduction in the prevalence of nuclei positive for Xist RNA and H3K27me3. If the inactive state of the mutated X in undifferentiated TS cells reflects the inactive state promoted by Xist^IVS RNA in the trophectoderm of blastocysts, our observations suggest that Xist^IVS RNA can trigger the X-inactivation process but is defective in establishing the proper heterochromatin configuration required for maintaining the inactivate state. We still cannot rule out the possibility, however, that Xist^IVS RNA initially establishes the proper inactivation status but fails to maintain it owing to reduced expression or transcriptional initiation of Xist. It is tempting to speculate, nonetheless, that although imprinted X-inactivation initiated by Xist^IVS in the extra-embryonic lineages is incomplete, it is probably sufficient for the initial differentiation of the extra-embryonic tissues to support successful implantation and subsequent development. However, owing to the aberrant heterochromatin configuration created by Xist^IVS RNA, the X^IVS loses repressive modifications such as H3K27me3 over time and eventually regains transcriptional activity to a level that is not tolerated in certain types of extra-embryonic cells such as the trophoblast giant cells. Although such a scenario is based on the assumption that Xist^IVS RNA is dysfunctional, it is worth considering that co-transcriptional events, such as splicing, are important in setting up Xist RNA-mediated silencing and that the intron might interfere with these events.

Detailed analyses of female embryos homozygous for Xist^IVS demonstrated that, in contrast to the extra-embryonic tissues, development of the embryonic ectoderm was severely affected even at E6.5, implying that the effect of Xist^IVS on random X-inactivation was much more severe than that on imprinted X-inactivation. When examined by RNA-FISH, however, a substantial fraction of the nuclei in the embryonic ectoderm exhibited single Xist clouds, suggesting that one or other of the Xist^IVS alleles had become upregulated in the homozygous mutants at the onset of X-inactivation. In addition, the gradual loss of GFP-positive cells in X^GFPX^IVS embryos between E7.5 and E10.5 is most likely the result of selection against those cells that have chosen not the wild-type allele but Xist^IVS early on for initiating the X-inactivation process. These observations suggest that Xist^IVS behaves in the same way as wild-type Xist in terms of the choice and initiation of random X-inactivation. It is therefore likely that the defects in random X-inactivation are attributable to the events that follow the accumulation of Xist^IVS RNA on the X chromosome. Our preliminary results using ES cells homozygous for Xist^IVS showed an apparently inactive-X-specific enrichment of H3K27me3, one of the earliest markers on the inactivated X (Plath et al., 2002), as well as Xist clouds in a subset of differentiated cells (Y.S. and T.S, unpublished). Although this suggests that Xist^IVS RNA is competent to recruit PRC2 to the X, we rarely found a late replicating X chromosome in the differentiated population.

Whereas most of the nuclei in the embryonic ectoderm contain a clear Xist cloud in E7.5 wild-type female embryos, ~30% of these nuclei in E7.5 X^IVSX^IVS embryos were negative for Xist RNA. By contrast, an Xist cloud was detected in the nuclei of the trophoblast at a comparable frequency in wild-type and homozygous Xist^IVS mutants. Given the RNA-FISH results for undifferentiated and differentiated XX^IVS TS cells, however, it can be presumed that at least a subset of the trophoblast nuclei in XX^IVS embryos would lose Xist clouds later on. It is likely that the cells that lack an Xist cloud in the embryonic ectoderm had initiated X-inactivation by inducing the expression of one or other Xist^IVS allele but then lost Xist clouds by E7.5, as was seen in differentiated TS cells.

X-inactivation initiated by Xist^IVS appears to be tolerated in the extra-embryonic lineages until the midgestation stage, but rarely allows proper differentiation of the embryonic ectoderm. The trophoblast giant cells and other trophoblast subtypes seem to respond to this mutation differently. This might reflect a lineage-specific difference in the tolerance to incomplete X-inactivation. In other words, the embryonic lineage requires stricter X-inactivation than the extra-embryonic lineages, and there are also some differences in tolerance among tissues even within the extra-embryonic lineages. Alternatively, the degree of X-inactivation induced by Xist^IVS might not be comparable among these lineages.

Previous studies showed that Xist RNA is required for the initiation of X-inactivation but is dispensable for the maintenance of the inactive state once established (Csankovszki et al., 1999; Wutz and Jaenisch, 2000). By contrast, we found that upon differentiation, TS cells heterozygous for Xist^IVS and, perhaps, early epiblastic cells homozygous for Xist^IVS, suffer from apparent reactivation of the mutated X^IVS with accompanying loss of Xist RNA and the H3K27me3 mark. This might represent tissue-specific differences in the requirement for Xist RNA to maintain the inactive state. It should be noted, however, that the loss of H3K27me3 is more prominent than the loss of Xist RNA from the X^IVS in differentiating TS cells, suggesting that the loss of H3K27me3 precedes that of Xist RNA. In this case, the loss of Xist RNA would not be a cause, but rather a result, of the reactivation of X^IVS. It is possible that because the inactive state is not fully established, the aberrant chromatin configuration facilitates spontaneous reactivation of X^IVS in certain circumstances. Further analysis of the presence or absence of chromatin modifications and proteins known to be associated with the inactive X would allow us to more specifically define the defects in the inactivated X^IVS.

Xist^IVS RNA differs from wild-type Xist RNA only in that it contains an unrelated 16 nucleotide insertion ~0.9 kb downstream from the 5′ end. There is no significant difference between Xist^IVS and wild-type Xist RNA in X chromosome coating in the blastocyst, expression level or stability, at least in undifferentiated TS cells. RT-PCR of cDNA from the E7.5 trophoblast demonstrated that the intron introduced into the Xist^IVS allele is efficiently removed by the expected splicing. In addition, the presence of this intron does not seem to perturb splicing between endogenous exons (data not shown). Currently, it is not known why Xist^IVS RNA is defective in inducing stable silencing of the X chromosome. It is possible, however, that the insertion of the 16 nucleotide sequence affects the function of the A-repeat, which has been shown to be essential for the silencing function of Xist RNA (Wutz et al., 2002), perhaps by interfering with the interaction of an unidentified trans factor with the A-repeat or by disrupting the formation of its appropriate secondary structure. Further study to elucidate the defects in X-inactivation initiated by Xist^IVS RNA as well as those in the Xist^IVS RNA itself would provide more insight into the molecular mechanism by which Xist RNA induces chromosome silencing.

We thank Minako Kanbayashi for genotyping of mice and for maintenance of the mouse colonies. This work was supported by a Grant-in-Aid for Scientific Research on Innovative Areas and a Grant-in-Aid for Scientific Research (B) from the Ministry of Education, Science, Sports, and Culture of Japan to T.S.