Crucial role of antisense transcription across the Xist promoter in Tsix-mediated Xist chromatin modification

An unexpected new view of the function of RNA involved in many biological phenomena has emerged from recent advances in transcriptome analysis revealing that much of the genome in humans, mice, and other organisms is transcribed into RNAs. Although these RNAs are often subject to splicing and polyadenylation, like common protein-coding mRNAs, most of them do not encode proteins. In addition, these noncoding RNAs are often transcribed in an antisense orientation relative to known genes or transcription units,suggesting that they play an important role in transcriptional regulation of the sense partner.

The noncoding Xist (X-inactive specific transcripts) gene(Borsani et al., 1991; Brockdorff et al., 1991; Brown et al., 1991) and its antisense Tsix gene (Lee et al.,1999) comprise one of the sense-antisense pairs that has been most extensively studied so far. These two noncoding genes play a pivotal role in X inactivation in mammals via a mechanism by which the dosage differences of X-linked genes between males and females are equalized by inactivating one of the two X chromosomes in females during early development(Lyon, 1961). Although either the paternal or maternal X chromosome is silenced in a basically random fashion (random X inactivation) in the embryonic tissues, which give rise to the fetus, X inactivation is imprinted in the extraembryonic tissues, from which the placenta and a part of the extraembryonic membranes are derived, in such a way that the paternally derived X is preferentially inactivated(imprinted X inactivation) (Takagi and Sasaki, 1975). At the onset of X inactivation, Xistexpression is induced on one of the two X chromosomes, and its 17-kb noncoding transcripts associate in cis along the chromosome and make it transcriptionally inactive by recruiting protein complexes involved in heterochromatinization. Xist and Tsix are reciprocally imprinted in the extraembryonic tissues so as to be expressed from the paternal and the maternal X chromosome, respectively(Kay et al., 1993; Lee, 2000; Sado et al., 2001). Since Xist is essential for X inactivation to occur in cis, an Xist-deficient X never undergoes inactivation in either embryonic or extraembryonic tissues (Marahrens et al.,1997; Penny et al.,1996). Paternal transmission of the mutated X, therefore, results in the failure of imprinted X inactivation and eventually a selective loss of female embryos soon after implantation owing to the presence of two active X chromosomes in the extraembryonic tissues(Marahrens et al., 1997). By contrast, Tsix deficiency, when maternally inherited, induces the expression of the normally silent maternal copy of Xist and subsequent inactivation of the mutated maternal X in the extraembryonic tissues (Lee, 2000; Sado et al., 2001). This ends up causing functional nullisomy of the X chromosome in these particular tissues of both male and female embryos, and eventually in death at the early postimplantation stage. It therefore seems likely that at the onset of X inactivation, while cessation of Tsix expression allows upregulation of Xist and triggers the inactivation process on the future inactive X, continued expression of Tsix is required for the future active X to preclude the activation of Xist on it. This is consistent with the fact that the Tsix-deficient X is inactivated in all cells of the embryonic tissues or differentiating female ES cells heterozygous for the mutation (Lee and Lu, 1999; Sado et al., 2001). Interestingly, mutant male embryos, even though they inherit the Tsixdeficiency from the mother, can survive if the wild-type extraembryonic tissues are provided by an experimental reconstruction using wild-type tetraploid embryos, implying the existence of a Tsix-independent mechanism for Xist silencing in the embryonic tissues(Ohhata et al., 2006). All these facts taken together indicate that the extraembryonic tissues appear to depend more on Tsix to regulate Xist than the embryonic tissues do.

Although it is clear that Tsix is a negative regulator of Xist effective only in cis, until recently the molecular mechanism of antisense regulation by Tsix at the Xist locus remained poorly understood. Recent analyses of the Tsix-deficient X in embryos or ES cells, however, suggested that Tsix silences Xistthrough modification of the chromatin structure(Navarro et al., 2005; Sado et al., 2005; Sun et al., 2006). Furthermore, our recent study demonstrated that splicing products of Tsix RNA are dispensable for Tsix-mediated Xistsilencing in mouse embryos (Sado et al.,2006). Therefore, we are particularly interested in the biological significance of the antisense transcription across the Xist promoter. To address this issue, in the present study we studied a mutation in mouse causing termination of Tsix transcription in exon 4 before the transcription proceeded across the Xist promoter region. This mutation, which essentially eliminated antisense activity in the upstream region of Xist, caused inappropriate activation of Xist and aberrant chromatin modification in the 5′ region of Xist, like the loss of function mutation of Tsix previously reported(Sado et al., 2005). These findings demonstrate that the antisense transcription across the Xistpromoter plays a crucial role in Tsix-mediated silencing of the Xist gene.

The second intron of the human γ-globin gene (IVS2) was amplified by PCR using primers Hu.Gl.IVS2-1(+)35SalI and Hu.Gl.IVS2-1(-)33SalI and cloned into pBluescriptII-SK(-) to produce pIVS2. A fragment containing a polyadenylation sequence that originated from the mouse Pgk1 gene and a triple polyadenylation sequence (a multiple polyadenylation sequence, mpA)amplified from the genome of a Rosa26R mouse using tpA-1(-)41NotI and tpA-2(+)37XhoI as primers was placed 3′ to a floxed puromycin-resistance gene driven by the mouse Pgk1 promoter (PgkPuro cassette) to produce pfloxPgkPuro-mpA. The floxPgkPuro-mpA cassette thus created was inserted at the unique PvuII site present in the IVS2 fragment in an antisense orientation with respect to IVS2. A 9.0 kb genomic EcoRI-BamHI fragment (104,306-113,284 in GenBank J421479)was cloned into pBluescriptII-SK (-) lacking an XhoI and a SalI site (pΔXS-EB9.0). The IVS2 fragment containing the floxPgkPuro-mpA cassette was subsequently introduced into pΔXS-EB9.0 at the unique XhoI site present in exon 1 of Xist to generate pTsix-mpA. Similarly, the IVS fragment containing only the PgkPuro cassette was inserted at the XhoI site in pΔXS-EB9.0 to generate pCont-X.

Each of the targeting vectors (pTsix-mpA and pCont-X) was electroporated into R1 ES cells (Nagy et al.,1993), and selection with puromycin was carried out as previously described (Sado et al., 2005). The expected homologous recombination was found in 1 out of 192 colonies and 4 out of 240 colonies when pTsix-mpA or pCont-X was introduced, respectively. Chimeric males were produced and serially crossed with females heterozygous for the Tsix-deficient X(XX^ΔTsix) to transmit the mutated paternal X to female offspring as previously described(Sado et al., 2005). The floxed puromycin-resistance gene was subsequently removed by crossing females heterozygous for either Tsix^pA2lox or Xist^IVS2lox with males, which ubiquitously expressed cre recombinase, to derive the Tsix^pA and Xist^IVS allele, respectively. Removal of the floxed puromycin cassette in ES cells was carried out by transfection of cre-expressing plasmid pBS185.

The visceral endoderm was separated from the yolk sac mesoderm according to Dandolo et al. (Dandolo et al.,1993). RNA was prepared from the embryo proper, visceral endoderm,placenta and ES cells using RNeasy (Qiagen). Thermal conditions for all the following PCRs will be provided upon request.

For the analysis of antisense transcription in the region distal to the mpA sequence in ES cells, cDNA was synthesized from 5 μg total RNA using Thermoscript (Invitrogen) at 56°C with each of the following three different primers: (1) AS90RS, (2) R1910J and (3) R371P1 for Tsix-specific priming in combination with GapdR for Gapd,and subsequently PCR was carried out on 1/25 of the cDNA thus prepared. Primers Xist1175F and Xist1472R were used for cDNA primed by (1) AS90RS, and primers R700P2 and Xist-6(-)20 were used for cDNA primed by (2) R1910J and (3)R371P1. A Gapd sequence was amplified as a control using GapdF/GapdR2.

For the analysis of allelic expression of Xist, cDNA was prepared from total RNA of E13.5 embryo proper and visceral endoderm in a strand-specific manner as described above using primers Xist-7(-)20 and GapdR,and PCR was carried out using primer sets R371P1 and Xist-6(-)20, and GapdF and GapdR2 for Xist and Gapd, respectively. An amplified product of Xist was digested with PvuII and electrophoresed on a 4% NuSieve 3:1 agarose gel.

For the analysis of Tsix expression in the placenta, RT-PCR was carried out on cDNA prepared by random priming as previously described(Sado et al., 2005). A proximal and a distal part of Tsix were amplified using primer sets Tsix2F and Tsix2R, and Xist (-540) and Xist-10(-)20, respectively(see below).

A northern blot was prepared using 20 μg total RNA from the embryo proper at E13.5 as previously described(Sado et al., 2005) and serially probed with pR97E1 for Xist(Sado et al., 1996) and a cDNA fragment for Gapd.

Genomic DNA isolated from the embryo proper and visceral endoderm at E13.5 was digested with BclI in combination with either methylation-sensitive restriction endonuclease SacII, HhaI, PmaCI or HpaII, electrophoresed on a 0.7% agarose gel,blotted on a nylon membrane and hybridized with either HS0.7 or BE0.6 as a probe (Sado et al., 2005).

About 2×10⁶ undifferentiated ES cells were used as a source of chromatin for ChIP assays using an antibody against the CTD of RNA polymerase II large subunit phosphorylated at serine 2 (H5) (Abcam) and normal mouse IgG (Sigma) as a mock control. Sonicated soluble chromatin was incubated with 3 μg of either H5 or mouse normal IgG overnight at 4°C. Chromatin precipitated with H5 was purified by incubating for another 3 hours at 4°C with anti-mouse IgM antibody raised in goat (Sigma), which had been previously bound to Dynabead-Protein-G (Invitrogen). Washing and elution of immunoprecipitated chromatin and extraction of DNA were carried out as previously described (Sado et al.,2005).

The visceral endoderm separated from the yolk sac mesoderm was individually suspended in 10 mM Tris-HCl, pH 7.4, 3 mM CaCl₂, 2 mM MgCl₂ including proteinase inhibitor Complete (Roche), to which an equal volume of 10 mM Tris-HCl, pH 7.4, 3 mM CaCl₂, 2 mM MgCl₂,1% NP-40, Complete was added. Following crosslinking by formaldehyde, cells were collected by centrifugation and resuspended in SDS lysis buffer. Subsequent procedures for the preparation of soluble chromatin were carried out in the same manner as previously described(Sado et al., 2005). To avoid potential variation in the efficiency of immunoprecipitation among different lots of commercially available polyclonal antibodies, we used monoclonal antibodies raised against H3K4me2, H3K4me3, H3K9me2 and H3K27me3 (gifts from Hiroshi Kimura and Naohito Nozaki), each of which had been previously bound to anti-mouse IgG-Dynabeads 280. Normal mouse IgG was used as a mock control.

For quantitative analysis, real-time PCR was carried out as previously described (Sado et al., 2006). The primers used were: Tsix2F/Tsix2R (for region d in Fig. 2A), Xist (-540)F/8692R(for region g in Fig. 2A),8111F/8418R (for region h in Fig. 2A), and Xist1F/Xist101R (for Fig. 6).

Primer sequences were as follows: Hu.Gl.IVS2-1(+)35SalI,5′-GGGTCGACTCAAGGTGAGTCCAGGAGATGTTTCAG-3′; Hu.Gl.I -VS2-1(-)33SalI, 5′-GGGTCGACAGGAGCTGTTGAGATGAAAGGAGAC-3′;tpA-1(-)41NotI, 5′-TTATATTAAGGCGGCCGCATCAGCTTGATGGGGATCCAGAC-3′;tpA-2(+)37XhoI, 5′-TTCTAACTCGAGCAGAAGCTTGCAGATCTGCGACTCT-3′;AS90R2, 5′-TGCGGGATTCGCCTTGATTT-3′; Xist1472R,5′-AGGGCAGGTCACATGACTTC-3′; Xist-10(-)10,5′-ACAAAATGGCTCCTTGGTTC-3′; 21b80F,5′-CCTGCAAGCGCTACACACTT-3′; F748P1,5′-CAGGTAGTGCAATAACTCACAA-3′; 8111F,5′-CTGCCACCTGCTGGTTTATT-3′; 8418R,5′-ACATGAAAGAGATCAGACAC-3′; Xist1F,5′-TGTTTGCTCGTTTCCCGTGGAT-3′; Xist101R,5′-CATAAGGCTTGGTGGTAGGG-3′; GapdF, GapdR, GapdR2, R1910J, R371P1,Xist21F, Xist(-540)F, Xist1175F, Xist1554R, Xist-6(-)20, Xist-7(-)20, R700R2 and 8692R were previously described (Sado et al., 2005). Tsix2F and Tsix2R were described by Sado et al.(Sado et al., 2006).

We attempted to address the impact of premature termination of the antisense transcription on the effect of Tsix by placing a multiple polyadenylation sequence (mpA) in exon 4 of Tsix(Fig. 1A). However,introduction of the mpA cassette and a floxed puromycin-resistance gene would inevitably disrupt the Xist gene. We therefore embedded these sequences in an intron derived from the human γ-globin locus (IVS2) and inserted them in exon 1 of Xist in an appropriate orientation in the targeting vector, as shown in Fig. 1B. We assumed that the inserted fragment would be spliced out upon transcription as an intron from the primary transcript of Xistand that the resultant processed transcript would behave as a functional Xist RNA. ES cells harboring this allele(Tsix^pA2lox) were produced by gene targeting(Fig. 1B,C). In parallel, we also used gene targeting to generate another mutant allele, Xist^IVS2lox, which harbored the same intron without the mpA cassette at exactly the same position as the Tsix^pA2lox allele (see Fig. S1 in the supplementary material). The selection marker was subsequently removed from both mutated alleles by transient expression of cre recombinase to generate Tsix^pA and Xist^IVS, which were subjected to further analyses.

To test whether Tsix transcription was effectively terminated before reaching the Xist promoter on the X chromosome carrying the Tsix^pA allele (X^pA) but not on the X carrying the Xist^IVS allele (X^IVS), strand-specific RT-PCR was carried out in the respective mutant ES cells. As shown in Fig. 2A,B, the expected fragment (PCR product e in Fig. 2A) was amplified in all cases in which cDNA was synthesized with primer a located proximal to the mpA cassette. By contrast, PCR product f was barely detectable in ES cells carrying Tsix^pA when either primer b or c located downstream of the mpA cassette was used for cDNA synthesis. The antisense transcription was evident in the regions proximal and distal to the inserted intron in ES cells harboring the Xist^IVS allele (Fig. 2B), demonstrating that the presence of the intron itself did not affect the transcriptional elongation of Tsix in the relevant region. These results suggested that the antisense transcription beyond the inserted mpA cassette was dramatically reduced on X^pA.

Effective attenuation of Tsix transcription was further examined by chromatin immunoprecipitation (ChIP) assays using a monoclonal antibody against the C-terminal domain (CTD) of RNA polymerase II (pol II)phosphorylated at Ser2, which has been widely used for immunoprecipitating the elongation form of RNA pol II (Ahn et al.,2004; Morris et al.,2005). Immunoprecipitated chromatin prepared from ES cells was subjected to PCR using a primer set located in the proximal region of exon 2 of Tsix (PCR product d) and primer sets located in the promoter region of Xist (PCR products g and h)(Fig. 2A,C). In addition to wild-type XY, X^pAY and X^IVSY male ES cells, male ES cells carrying X^dc, on which Tsix is truncated owing to the insertion of an IRESβgeo cassette at the distal region of exon 2(Sado et al., 2005; Sado et al., 2001), were included in the assay. The elongation form of RNA pol II was evenly distributed in exon 2 of Tsix in all types of ES cells, whereas it was significantly reduced in the promoter region of Xist in X^pAY to a level comparable with that in X^dcY, which essentially lacks Tsix transcription(Fig. 2C). The distribution of RNA pol II phosphorylated at Ser2 was higher in the distal region of Tsix (region g and h) than in the proximal region (region d) in XY and X^IVSY ES cells, a pattern that is commonly observed on transcriptionally active genes (Ahn et al.,2004; Morris et al.,2005). It therefore seems reasonable to conclude that the majority of the antisense transcription beyond the mpA cassette was significantly diminished as expected.

The Tsix^pA2lox and Xist^IVS2loxalleles were introduced into mice through chimeric males crossed with females heterozygous for Tsix deficiency, and the puromycin-resistance gene was subsequently removed to generate the Tsix^pA and Xist^IVS allele, respectively, as previously described(Sado et al., 2005). We first examined whether these alleles behaved in the same manner as in the mutant ES cells. Since Tsix is imprinted in the extraembryonic tissues to be expressed from the maternal allele in both sexes(Lee, 2000; Sado et al., 2001), we used total RNA isolated from embryonic day (E) 12.5 placentas of male embryos recovered from Tsix^pA/+ and Xist^IVS/+females crossed with wild-type males for RT-PCR. Transcription of Tsix was evident in the proximal region of Tsix on the mutated X carrying either Tsix^pA or Xist^IVS (Fig. 2D). Similarly, spliced products of Tsix were also produced from these mutated X chromosomes (data not shown). When PCR was performed using a primer set located in the promoter region of Xist,by contrast, the expected fragment was readily detectable in X^IVSY as well as wild-type XY placenta, but was significantly reduced in X^pAY (Fig. 2D). Similarly, antisense transcription was barely detected in the promoter region of Xist in the placenta isolated from X^pAX females. These results indicate that the antisense transcription runs across the Xist promoter on X^IVS as on the wild-type X, but is effectively attenuated on X^pA in vivo as expected.

If the primary transcript of Xist from X^pA was processed as we expected, the splicing product would be functional and competent to induce X inactivation. To examine whether or not X^pAcould undergo inactivation, males hemizygous for Tsix^pAwere crossed with wild-type females. Recovery of live female pups from this cross would indicate that Xist on the mutated paternal X is functional, because Xist deficiency should cause failure of imprinted X inactivation and result in a selective loss of female embryos at early postimplantation stages (Marahrens et al.,1997). Of 176 offspring obtained from 43 litters, 174 were males and 2 were females (Fig. 3A). One of the two females was XO and the remaining one had inherited the mutated X. The strong bias toward male pups suggested that Xist on the mutated X was dysfunctional and therefore unable to induce X inactivation. When examined at E7.5, female embryos, although found at a reasonable ratio,exhibited abnormal morphology typical of those suffering from the failure of imprinted X inactivation (Fig. 3B,C) (Marahrens et al.,1997). In the reciprocal crosses, the mutated X was maternally transmitted to both male and female pups with the Mendelian ratio(Fig. 3D). Taking advantage of an X-linked GFP transgene (X^GFP), the X inactivation pattern was examined in female embryos heterozygous for Tsix^pA. As shown in Fig. 3E,X^pAX^GFP females recovered at E8.5 were virtually negative for GFP fluorescence, indicating that X^GFP had been uniformly inactivated even in the embryonic lineage, where X inactivation should occur basically at random. Taking all these findings together, we concluded that the insertion of the extra intron unexpectedly made the Xist gene dysfunctional and that as a result, X^pA was essentially defective in undergoing inactivation. Similarly, paternal transmission of Xist^IVS resulted in embryonic lethality(data not shown), demonstrating that the function of Xist on X^IVS was also impaired.

Although Xist on these mutated X chromosomes turned out to be dysfunctional, the Tsix^pA allele successfully terminated the antisense transcription as expected. This allowed us to fully address the effects brought about by the loss of Tsix transcription in the Xist promoter region on Xist silencing by using a previously described approach (Sado et al.,2005; Sado et al.,2006).

We first examined whether the Xist locus became ectopically activated by the truncation of Tsix in E13.5 male embryos. Northern blotting of total RNA demonstrated that although wild-type male embryos did not express Xist at all, a significant amount of Xist RNA was expressed in X^pAY embryos(Fig. 4A). We also carried out strand-specific RT-PCR using RNA isolated from the embryos proper (embryonic lineage) and the visceral endoderm (extraembryonic lineage) in both sexes recovered at E13.5 from Tsix^pA/+ females crossed with males carrying a Mus. m. molossinus (JF1)-derived X chromosome(X^JF1). A PvuII site polymorphism between the laboratory strains and JF1 allowed us to assess the allelic expression of Xistin females. As shown in Fig. 4B, Xist on the X chromosome of the laboratory strain(X^lab) was predominantly expressed in the embryo proper of wild-type females, probably owing to the Xce effect, which is known to cause some bias in the choice of chromosome for X inactivation, whereas the great majority of Xist detected in the visceral endoderm was derived from X^JF1, in agreement with imprinted paternal X inactivation. In both X^pAX^JF1 and X^IVSX^JF1 females,the mutated X should remain active in every cell because of the inability of Xist to induce inactivation. Xist on the active X is usually repressed in somatic cells. However, although the Xist locus on the mutated X in X^IVSX^JF1 females was appropriately repressed, it was ectopically activated in X^pAX^JF1females in both embryonic and extraembryonic tissues(Fig. 4B). This was also the case in X^pAY male embryos (Fig. 4B). By contrast, such ectopic activation of Xist was not observed on X^IVS in either embryonic or extraembryonic lineages in either sex (Fig. 4B). It should be noted that although Xist on the mutated X^pA was incompetent for inducing chromosomal inactivation, as mentioned above, the introduced intron was spliced out from the primary transcript as expected(data not shown). Nonetheless, these results suggest that the premature termination of Tsix transcription compromises the Tsix-mediated Xist silencing.

Since the truncation of Tsix might have distinct impact on the embryonic and extraembryonic tissues, we compared the levels of ectopically expressed Xist in the embryo proper and in the visceral endoderm of X^pAY male embryos by real-time PCR. Intriguingly, ectopic expression of Xist was reproducibly higher in the visceral endoderm than in the embryo proper (Fig. 4C), suggesting that the extraembryonic tissues were more susceptible than the embryonic tissues to the truncation of Tsix in terms of Xist silencing. However, the ectopic expression in the visceral endoderm was about 30-40% of the level of Xist expression in the visceral endoderm of wild-type females. The reason why the Xistlocus on X^pA was not fully activated is currently unknown.

Next, we studied the effect of the premature termination of Tsixin exon 4 on the chromatin in the Xist promoter region. It is known that CpG sites in the Xist promoter of the transcriptionally active allele on the inactive X are largely unmethylated, whereas the transcriptionally inactive allele on the active X is highly methylated(Norris et al., 1994). Our previous study demonstrated that Tsix is required for the establishment of CpG methylation in the Xist promoter region(Sado et al., 2005). We therefore examined whether this Tsix function is impaired by the premature termination. The methylation profile of the Xist promoter region was examined in the embryo proper and in the visceral endoderm at E13.5 by Southern blotting using four different methylation-sensitive restriction enzymes in combination with methylation-insensitive BclI. Although this region on X^IVS, which stays active, was completely methylated,like that on the wild-type active X, in both the embryo proper and visceral endoderm (see Fig. S2 in the supplementary material), a reduction in methylation was evident on X^pA(Fig. 5), which also stays active, especially in the extraembryonic tissues. Intriguingly, the decrease in CpG methylation was much more prominent in the visceral endoderm than in the embryo proper in both sexes, which is consistent with the difference in the level of Xist expressed ectopically in these tissues. These results suggest that the loss of Tsix transcription across the Xist promoter region diminishes the function of Tsix and causes a reduction in CpG methylation in the 5′ region of Xist,as was seen on X^dc (Sado et al., 2005).

The results described so far suggest that the premature termination of Tsix, which apparently compromised the function of Tsix, had more profound effects on Xist silencing in the extraembryonic lineages than in the embryonic lineage. We therefore used chromatin prepared from the visceral endoderm of the male yolk sac to analyze the histone modifications in the 5′ region of Xist using the ChIP assay. We used monoclonal antibodies raised against histone H3 di- and trimethylated on Lys4 (H3K4me2 and H3K4me3, respectively), dimethylated on Lys9 (H3K9me2) and trimethylated on Lys27 (H3K27me3), whose specificities in the ChIP assay have been extensively studied elsewhere (Kimura et al., personal communication). H3K4me2 and H3K4me3 are known to be associated with transcriptionally active regions, and H3K9me2 and H3K27me3 with transcriptionally inactive regions. Real-time PCR was performed to amplify the region immediately downstream of the transcription start site of Xist (+1 to +101 in GenBank Acc. No. L04961). On the wild-type active X in the visceral endoderm, this region was highly enriched in H3K9me2 and H3K27me3 but largely devoid of H3K4me2(Fig. 6), consistent with the fact that Xist is repressed on the active X. The same region on the mutated active X in X^pAY showed an enrichment of H3K4me2 and exclusion of H3K9me2 and H3K27me3 (Fig. 6). By contrast, this region in X^IVSY possessed histone modifications similar to those on the wild-type X(Fig. 6). The distribution of H3K4me3 did not appear to be affected by the presence or absence of Tsix, and stayed at a low level. These results suggest that it is the truncation of Tsix, not the presence of the extra intron in exon 1 of Xist, that facilitated the active modifications instead of the repressive ones in the 5′ region of Xist.

In this study we examined the functional significance of the antisense transcription across the Xist promoter in Tsix-mediated Xist silencing. Premature termination of Tsix transcription in the Xist gene resulted in inappropriate activation of Xist in cis on the mutated active X in embryos and MEFs (data not shown), as was the case on a Tsix-deficient X(Sado et al., 2005). This was accompanied by reduced CpG methylation and aberrant histone modifications in the 5′ region of Xist, especially in the extraembryonic tissues. By contrast, neither ectopic activation of Xist nor aberrant chromatin modifications were observed on the active X bearing Xist^IVS, which retains the antisense transcription across the Xist promoter. These findings strongly suggest that it was the transcriptional attenuation of Tsix, not the introduction of the exogenous DNA sequences, that compromised the function of Tsix in establishing the transcriptionally repressive chromatin configuration in the 5′ region of Xist. However, we cannot completely exclude the possibility that the observed phenotype was caused by disruption of not only Tsix function but also Xist function.

The transcription of Tsix primarily initiates about 13 kb downstream of Xist and extends as far as 50 kb along the entire Xist gene beyond the promoter region. Previous studies demonstrated that the truncation of Tsix in the region downstream of Xistabolishes the function of Tsix in both embryos and differentiating ES cells (Luikenhuis et al.,2001; Sado et al.,2001). Shibata and Lee further suggested that Tsix-mediated Xist silencing requires concurrent antiparallel transcription across the Xist gene(Shibata and Lee, 2004). Here we have demonstrated that Tsix^pA, which still retained the transcription of the proximal 93% of the region encoding the Tsixgene, failed to establish Xist silencing, suggesting that Tsix has to be transcribed across the Xist promoter to establish silencing of Xist. Although Air and Kcnq1ot1 noncoding antisense RNAs have also been suggested to be involved in chromatin modification, truncation of these RNAs impairs not only expression of the respective sense partner but also allelic expression of neighboring imprinted genes in the respective imprinted domains(Mancini-Dinardo et al., 2006; Sleutels et al., 2002). This implies that the actions of Tsix and these noncoding genes are distinct in terms of mediating the chromatin modifications.

The truncation of Tsix had distinct impact on the Xistsilencing mechanism in the embryo proper versus the visceral endoderm, which belong to the embryonic and extraembryonic lineages, respectively. Real-time PCR revealed that the level of ectopic expression of Xist was higher in the visceral endoderm than in the embryo proper. Although a reduction in CpG methylation in the 5′ region of Xist was detected in both lineages, it was more prominent in the visceral endoderm than in the embryo proper in both sexes. Similarly, aberrant histone modifications were much more evident in the visceral endoderm than in MEF (data not shown). It therefore seemed likely that the truncation of Tsix had more profound effects on the Xist silencing mechanism in the extraembryonic tissues than in the embryonic tissues. We recently suggested that the embryonic tissues but not the extraembryonic tissues possess a mechanism to shut off Xistin a Tsix-independent manner if Xist becomes ectopically expressed (Ohhata et al.,2006). Since the majority of Xist RNA transcribed from X^pA is subject to the expected splicing of the intron introduced in exon 1, this transcript is almost the same as that transcribed from the wild-type Xist allele, except for the addition of some flanking sequences of the intron. It is possible that the presence of such transcripts activates the putative Tsix-independent Xist silencing mechanism and induces some repressive modifications in the Xistpromoter region in the embryonic tissues. Since this mechanism is missing in the extraembryonic tissues, the visceral endoderm would exhibit severer effects on Xist silencing than the embryo proper or MEF as a result of the truncation of Tsix.

Although the truncation of Tsix inevitably altered the genomic structure of the overlapping Xist gene, we assumed that the function of Xist would be relieved if the exogenous sequence containing the mpA cassette was removed by splicing. However, the Tsix^pAallele accordingly designed did not act in the expected way. Paternal transmission of X^pA resulted in lethality of female offspring at peri-implantation stages, probably owing to the failure of imprinted paternal X inactivation in the extraembryonic lineages, suggesting that the Xist on X^pA was dysfunctional. Although the expected splicing event was observed in both the mutant male ES cells and the visceral endoderm isolated from the X^pAY embryos (data not shown), it is possible that the excision of this intron was not as efficient as we expected. The splicing products were retained in the nucleus, like those transcribed from the wild-type Xist allele, ruling out the possibility that the exogenous intron derived from the protein-coding gene enhanced export of the transcripts to the cytoplasm. Another possibility is that the ectopic expression of Xist on X^pA was not high enough to induce X inactivation. In fact, the ectopic expression of Xist in the visceral endoderm of X^pAY embryos was about 30-40% of the level of Xist expression in the visceral endoderm of wild-type female embryos,and RNA fluorescent in situ hybridization (FISH) failed to detect accumulation of Xist RNA in the mutant male MEFs (data not shown). Although we cannot rule out the possibility that the antisense transcription across the Xist promoter is not the sole event responsible for Tsix-mediated Xist silencing, the incomplete activation of Xist could be due to the structural alteration of a putative cis element required for the full activation of Xist. The insertion of the intron, in fact, increased the physical distance between the Xistpromoter and the YY1 binding sites recently reported to be necessary for Xist expression (Kim et al.,2006). This could have diminished the appropriate interaction between them required to fully activate Xist. It is also formally possible that the presence of the extra sequence left after the expected splicing event has a negative impact on the function of Xist RNA. Further analyses will shed light on the mechanism by which Xistbecomes fully activated and mediates the chromosomal silencing upon the cessation of Tsix transcription on the same chromosome.

We are grateful to Hiroshi Kimura and Naohito Nozaki for sharing the monoclonal antibodies against histone modifications. We thank Minako Kanbayashi for genotyping of mice and maintenance of the mouse colonies. This work was supported by grants from PRESTO, Japan Science and Technology Agency(JST) and from the Ministry of Education, Science, Sports, and Culture of Japan to T.S.