Dosage differences of X-linked genes between male and female mammals are compensated for by a mechanism known as X-inactivation, and the noncoding Xist gene plays a crucial role in this process. The expression of Xist is regulated in cis by its noncoding antisense gene, Tsix, whose transcripts (though a fraction of them stay unspliced), are processed like common proteincoding RNAs. It has been suggested that certain classes of sense-antisense pairs of RNA are causally involved in not only gene regulation but also higher order chromatin structure in various organisms. In fact, recent studies demonstrated that Tsix modulates Xist expression through modification of the chromatin structure. It is still unknown, however, whether the RNA product is important for the function of Tsix or whether the antisense transcription is sufficient. To obtain insight into this issue, we eliminated the splicing products of Tsix in the mouse and explored the effects of this elimination on Tsix-mediated Xist silencing. To our surprise, the Xist locus was stably repressed on the X carrying the splicing-defective Tsix allele. Moreover, the repressive chromatin configuration was properly established at the Xist locus. These unexpected results indicate that the splicing products are dispensable for Tsix-mediated Xist silencing.
Recent progress in transcriptome analysis has revealed that a large proportion of the genome can be transcribed from both strands, producing sense and antisense transcripts (Carninci et al., 2005; Kiyosawa et al., 2003; Yelin et al., 2003). The available evidence suggests that certain classes of RNA are causally involved in not only gene regulation but also higher order chromatin structure in various organisms (Bernstein and Allis, 2005). In addition, antisense transcription has often been implicated in either transcriptional or post-transcriptional regulation of the sense partner or the neighboring genes, although the exact molecular mechanisms have yet to be determined (Katayama et al., 2005).
It is known that X-inactivation in mammals, which ensures dosage equivalence of X-linked genes between males and females (Lyon, 1961), is regulated by the noncoding Xist (X-inactive specific transcript) gene (Brockdorff et al., 1992; Brown et al., 1992) and its antisense non-coding Tsix gene (Lee et al., 1999) mapped in the X chromosome inactivation center (Xic), a cytogenetically identified master regulatory region of X-inactivation (Avner and Heard, 2001). Xist is essential for the initiation of X-inactivation to occur in cis (Marahrens et al., 1997; Penny et al., 1996), and its expression is regulated by Tsix on the same chromosome in a negative fashion (Lee, 2000; Lee and Lu, 1999; Sado et al., 2001). The Xist/Tsix locus is probably one of the best-studied loci harboring a sense-antisense pair of transcripts, and study of this locus should allow us to obtain further insight into the molecular mechanisms underlying antisensemediated gene regulation, which appears to be rather common in a variety of systems.
The paternal X chromosome is selectively inactivated by imprinting in the extra-embryonic tissues, which give rise to the placenta and part of the extra-embryonic membranes (Takagi and Sasaki, 1975), where the paternal copy of Xist is always activated and the maternal one is never expressed (Kay et al., 1993). Targeted disruption of Tsix, when maternally inherited, induces ectopic activation of the normally silent copy of Xist on the same X chromosome in these tissues (Lee, 2000; Sado et al., 2001). Subsequent inactivation of the mutated maternal X results in functional nullisomy of the X chromosome in both males and females, and eventual embryonic death at early postimplantation stages. Although the targeted disruption of Tsix revealed that Tsix is a negative regulator of Xist effective in cis, until recently little was known about the molecular mechanisms underlying the antisense regulation at the Xist locus. We recently showed that in the absence of Tsix, the Xist locus fails to establish the repressive chromatin configuration, and instead manifests the active modification in embryos, suggesting that Tsix mediates Xist-silencing through modification of the chromatin structure (Sado et al., 2005). Others also demonstrated Tsix-mediated chromatin effects using undifferentiated and differentiating embryonic stem (ES) cells as an in vitro model system (Navarro et al., 2005; Sun et al., 2006).
Tsix transcripts are subject to splicing and polyadenylation, like common protein-coding RNAs (Sado et al., 2001; Shibata and Lee, 2003). The major products of spliced Tsix RNA in undifferentiated ES cells, which are about 2.7 kb and 4.3 kb in length on northern blots, appear to be produced by differential usage of several polyA signals found in exon 4 of Tsix (Sado et al., 2001). RT-PCR-based methods also revealed the presence of some variants produced by alternative splicing (Sado et al., 2001; Shibata and Lee, 2003). In addition, Tsix RNA exists in spliced and unspliced forms, although the functional relevance of splicing remains unknown. Because a significant fraction of the RNA is processed, and the splicing variants identified so far all share exon 4, which is the only exon possessing complementarity with the Xist sequence, it had been postulated that the splicing products (especially the sequence in exon 4) play an important role in the function of Tsix by masking the A-repeats in Xist, a region known to be required for Xist-mediated silencing (Wutz et al., 2002), through base pairing (Shibata and Lee, 2003).
To obtain further insight into this issue, we studied the functional significance of spliced Tsix RNA produced by antiparallel transcription through the endogenous Xist locus. Here we eliminated the splicing products of Tsix in embryos by disturbing the splicing between proximal exons and exon 4. Loss of spliced Tsix RNA, however, compromised neither the Tsix-mediated Xist silencing nor the establishment of the proper chromatin configuration in the Xist promoter region. Our findings suggest that chromatin modulation at the Xist locus is not mediated by spliced Tsix RNA but rather nascent/unspliced Tsix RNA or the antisense transcription per se.
MATERIALS AND METHODS
A plasmid containing a 4.4-kb EcoRI fragment and a 6.5-kb XbaI fragment from the Xist genomic locus (Sado et al., 2005) was digested with PmaCI and SmaI to release a fragment harboring the splicing acceptor site of Tsix exon 4, which was then replaced with a fragment containing an IRES-EGFP and a puromycin resistance gene similarly to how a targeting vector for the Xist2lox allele was produced (Sado et al., 2005). The targeting vector thus produced was electroporated into R1 ES cells (Nagy et al., 1993), and chimeras were produced as described previously (Sado et al., 2005). Briefly, out of 96 colonies, three harbored the expected homologous recombination (TsixΔSA2lox), and they were serially injected into blastocysts to generate chimeras. In this particular case, two female chimeras crossed with C57Bl/6 males happened to transmit the TsixΔSA2lox allele to both male and female pups. Those females carrying TsixΔSA2lox were crossed with males that ubiquitously expressed cre recombinase to derive animals carrying TsixΔSA.
RT-PCR and northern blotting
RT-PCR was carried out on cDNA prepared from 1 μg of total RNA using either random primer or strand-specific primers. For strandspecific RT-PCR, cDNA was synthesized using R1910J as a primer for Tsix and Xist7(-)20 as a primer for Xist. Subsequent amplification of a common sequence in Tsix/Xist was performed using primers R700P2 and Xist6(-)20. A Gapd-specific primer, GapdR, was also included in the cDNA reaction to amplify a Gapd sequence using GapdF and GapdR2 for normalization in real-time PCR analysis. Real-time PCR was performed on cDNA thus produced using SYBR Premix Ex Taq (Takara) and a LightCycler (Roche) according to the manufacturers' instructions. The expression levels of each sequence were the value normalized by Gapd based on a mean value of the respective sequence obtained from two or three independent experiments. Primer sequences used in this study were as described previously (Sado et al., 2005; Sado et al., 2001) except for the following: 8111F, ctg cca cct gct ggt tta tt; 8420R, cca cat gaa aga gat cag ac; EGFP1, ctt ctt caa gga cga cgg ca; EGFP2, ttg tac agc tcg tcc atg cc; Tsix2F, caa tct cgc aag atc cgg tga; Tsix2R, agt gga tgc agg act caa gat.
RNA extracted from ES cells harboring the respective mutation was subjected to northern hybridization as described previously (Sado et al., 2001). An RNA probe was prepared from p10 and p10R for detection of Tsix and Xist, respectively (Sado et al., 2001). For Gapd, a cDNA fragment was used as a probe.
Methylation analysis and DNaseI hypersensitive site assay
For methylation analysis of the Xist promoter region, Southern hybridization was performed in exactly the same manner as described previously (Sado et al., 2005).
The DNaseI hypersensitive site assay was carried out as previously described (Sado et al., 2005), with the exception of a 0.4 kb AflIII genomic fragment (Af0.4) being used as a probe.
Generation of a new Tsix mutant allele defective in splicing
In a previous study, we created a new Xist mutant allele, Xist1lox, by replacing a part of exon 1 with a promoter-less IRES-EGFP cassette (Sado et al., 2005). This allele was created so as not to delete the splicing acceptor site for exon 4 of the overlapping antisense Tsix transcript (Fig. 1C). Although the Xist gene was disrupted by the replacement, primary Tsix transcripts were still appropriately processed to yield the major 2.7-kb and 4.3-kb RNAs with a polyA tail (Fig. 2D). We demonstrated that the function of Tsix was not impaired by this alteration, and the repressive chromatin configuration was properly established at the Xist promoter on the active mutated X chromosome (Sado et al., 2005). To elucidate the functional significance of spliced Tsix RNA in the proposed Tsix-mediated chromatin regulation at the Xist promoter, we attempted to disturb the splicing event by deleting the splicing acceptor site for exon 4 of Tsix. Accordingly, another targeting vector, which was very similar to the one used to derive the Xist1lox allele, was created (Fig. 1A). The difference between these vectors was that the 5 ′ arm either did or did not contain a 0.6-kb PmaCI-EcoRI fragment harboring the splicing acceptor site. Upon a correct targeting event, the splicing acceptor site was successfully deleted on the single X chromosome in male ES cells by replacement with the IRES-EGFP cassette (TsixΔSA2lox) (Fig. 1B,C). A puromycin resistance gene was subsequently removed by transient expression of cre recombinase in ES cells to generate the TsixΔSA allele (data not shown).
Splicing variants are not produced from the TsixΔSA allele in ES cells
We examined whether or not spliced Tsix RNA was eliminated in TsixΔSA/Y ES cells by performing RT-PCR using cDNA prepared by random priming. Unlike in XY and Xist1lox/Y ES cells, a primer set spanning the introns (Xist1175F and 21b80F) failed to amplify the product derived from spliced Tsix RNA in TsixΔSA/Y as well as in XdcY ES cells, which carry the Xist1lox allele on a Tsix-deficient X chromosome (Sado et al., 2005) (Fig. 2A,B). The presence of antisense transcripts was, however, evident in XY, Xist1lox/Y, and TsixΔSA/Y ES cells when another primer set (8111F and 8420R) located in exon 4 was used, suggesting that the antisense transcription itself was not disrupted. The relative abundance of the antisense transcripts in each ES cell line was further analyzed by real-time PCR using two primer sets designed to hybridize within either exon 2 (Tsix2F and Tsix2R) or exon 4 (8111F and 8420R) (Fig. 2C). In XdcY, antisense expression was detected in exon 2 at a level comparable to that in XY but was barely detected in exon 4, which is consistent with the fact that Tsix is truncated on Xdc (Sado et al., 2005). On the other hand, the level of antisense transcripts in Xist1lox/Y and TsixΔSA/Y turned out to be about twice as high as that in wild-type XY ES cells in both regions. It is unclear whether the transcription is upregulated or the stability of the antisense transcripts is increased by the alteration introduced on the mutated X in Xist1lox/Y and TsixΔSA/Y. The absence of the spliced forms in TsixΔSA/Y ES cells was further confirmed by northern blotting using polyA RNA (Fig. 2D). Although an RNA probe specific to Tsix detected 2.7-kb and 4.3-kb bands in XY and Xist1lox/Y ES cells, neither of these bands was detected in TsixΔSA/Y or XdcY ES cells. The hybridization signal observed in the high molecular weight region in XY, Xist1lox/Y, and TsixΔSA/Y, which was missing in XdcY, probably represented unspliced Tsix RNA and/or perhaps an incompletely processed one that contains the third intron. These results confirmed that the TsixΔSA allele was defective in splicing between proximal exons and exon 4 as expected, with the antisense expression activity being essentially unaffected.
We observed unexpected activation of the Xist locus on the mutated X chromosome in undifferentiated ES cells. As shown in Fig. 2D, an RNA probe specific to Xist detected the chimeric transcripts of 4.5 kb in XdcY and Xist1lox/Y and those of 4.0 kb in TsixΔSA/Y. Despite the activation of the Xist locus, neither of the DNaseI hypersensitive sites specifically found in the 5 ′ region of the transcriptionally active Xist allele in somatic cells (see Fig. 6) was observed on the mutated X in undifferentiated ES cells (Fig. 2E). While the ectopic activation of the Xist locus in XdcY was sustained even after differentiation, expression of the chimeric transcripts in Xist1lox/Y and TsixΔSA/Y became downregulated upon differentiation as embryoid bodies for 12 days (Fig. 2F), suggesting that developmental regulation of the Xist gene was compromised on the Xdc chromosome, as previously described (Sado et al., 2005) but not on the mutated X in Xist1lox/Y and TsixΔSA/Y. It is likely that the repressive action of Tsix occurs after differentiation.
Splicing variants of Tsix are eliminated in the mouse
Females heterozygous for TsixΔSA2lox were produced through germline transmission from chimeric mice and subsequently crossed with males expressing cre recombinase to derive animals carrying TsixΔSA by excision of the puromycin resistance gene (see Materials and methods) (Fig. 1D). Since the Xist gene on the mutated X was consequently destroyed, males failed to transmit the TsixΔSA allele to female offspring when crossed with wild-type females, as previously described (Marahrens et al., 1997; Sado et al., 2005). The mutant allele was, however, transmitted to both male and female pups from the mother at the expected ratio (data not shown).
Taking advantage of embryos harboring maternally derived TsixΔSA, we studied the role of spliced Tsix RNA in Tsix-mediated Xist silencing. We first asked whether the splicing products had been eliminated in mice as in TsixΔSA/Y ES cells. It is known that Tsix is imprinted to be expressed from the maternal allele in the extraembryonic tissues such as the placenta (Lee, 2000; Sado et al., 2001). RNA isolated from the placenta at embryonic day (E) 13.5 was subjected to RT-PCR. While the presence of the spliced products was evident in both wild-type males and females and those carrying maternal Xist1lox with primers spanning exon 2 and exon 4 (Xist1175F and 21b80F), such products were not observed in TsixΔSA/+ females or TsixΔSA/Y males (Fig. 3). In contrast, the expected fragment was amplified in all cases with a primer pair (8111F and 8420R) located in exon 4. These results suggest that TsixΔSA impaired the splicing event but not the antisense transcription per se during embryonic development.
Xist is properly repressed in the absence of spliced Tsix RNA
In previous studies, loss of function mutation of Tsix caused activation of Xist in cis (Lee, 2000; Sado et al., 2001), which was attributed to the failure to establish the repressive chromatin configuration at the Xist promoter (Sado et al., 2005; Sun et al., 2006). We first examined if the elimination of spliced Tsix RNA abrogated the proper regulation of Xist on the mutated X chromosome carrying the TsixΔSA allele. In female embryos heterozygous for the respective mutation, as shown in Fig. 4, the mutated X stays active in every cell because of Xist-deficiency on it. In our assay system, therefore, the expression of chimeric RNA from the mutated active X would demonstrate defects in the mechanism of silencing of Xist. As previously shown, XdcX female embryos expressed chimeric RNA due to functional deficiency of Tsix (Fig. 4). Such ectopic activation of the Xist promoter on the mutated X was not, however, observed in either Xist1lox/+ or TsixΔSA/+ embryos (Fig. 4), suggesting that the Xist promoter was appropriately repressed on the mutated X chromosome carrying TsixΔSA. Similarly, the expression of chimeric RNA was detected in XdcY but not in TsixΔSA/Y. These results demonstrated that the Xist locus was properly silenced even in the absence of spliced Tsix on the mutated X.
Appropriate chromatin configuration was established at the Xist promoter on the X carrying TsixΔSA
It is known that the Xist promoter is differentially methylated in female somatic cells, with the active allele on the inactive X being unmethylated and the inactive allele on the active X being methylated (Norris et al., 1994). Our previous study demonstrated that functional loss of Tsix impairs establishment of complete methylation at CpG sites in the Xist promoter in cis. To examine whether CpG methylation was affected by loss of the Tsix splicing products, we carried out Southern blot analysis. Genomic DNA isolated from both male and female embryos carrying TsixΔSA at E13.5 was digested with methylation-sensitive HhaI or SacII in combination with methylation-insensitive BclI. As shown in Fig. 5, the Xist promoter was completely methylated on the mutated active X carrying TsixΔSA at a level comparable to that on the active X carrying Xist1lox as well as the wild-type allele in the embryonic and extra-embryonic tissues in both sexes. This result suggests that CpG methylation is established even in the absence of spliced Tsix RNA.
We went on to study the chromatin structure at the Xist locus by DNaseI hypersensitive site assays. Several DNaseI hypersensitive sites have been found in the Xist promoter region (Sado et al., 2005; Sheardown et al., 1997). It is known that HS1 and HS5 are specific to the inactive X, and HS3 is common to both the active and inactive X. In a previous study, we showed that functional deficiency of Tsix results in the failure to establish the closed chromatin structure in the Xist promoter region. As shown in Fig. 6, neither of the ectopic DNaseI hypersensitive sites seen on the Xdc was detected on the X carrying TsixΔSA in either male or female embryos at E13.5. This result indicates that the closed chromatin structure is established at the Xist promoter even in the absence of spliced Tsix RNA.
This study was carried out to further elucidate the molecular aspects of the Tsix-mediated Xist silencing described in our previous report (Sado et al., 2005). Tsix transcripts are peculiar in that although a fraction of them remain unspliced, various patterns of splicing all remove most of the region complementary to Xist RNA from the 40-kb primary transcript, except for a 1.9-kb region present in the proximal end of Xist, to yield the major processed products of length 2.7 kb and 4.3 kb (Sado et al., 2001; Shibata and Lee, 2003). In addition, the distal part of the processed Tsix RNA thus produced covers the region up to 1.8-kb upstream of the Xist transcription start site, where no transcription has been detected in the sense orientation. These features of the antisense transcripts at the Xist locus imply that the 1.9-kb region common to Tsix RNA and Xist RNA is relevant to the effect of Tsix on the antisense transcription. It has been shown that the function of Tsix is impaired by truncation of Tsix in the downstream region of Xist without deleting any genomic elements (Luikenhuis et al., 2001; Shibata and Lee, 2004). In addition, Shibata and Lee failed to rescue the loss-of-function phenotype of Tsix caused by the truncation even though a Tsix minigene was introduced immediately downstream of Xist so as to be transcribed in the same orientation as Xist under the control of an exogenous constitutively active promoter (Shibata and Lee, 2004). These studies suggest that Tsix action is not merely derived from the processed transcripts but requires concurrent antiparallel transcription through Xist.
In this study, we attempted to further clarify the significance of processed Tsix RNA produced by antiparallel transcription through the endogenous Xist gene. In contrast to targeted disruption (Lee, 2000; Lee and Lu, 1999; Sado et al., 2001) or transcriptional attenuation of Tsix (Luikenhuis et al., 2001; Shibata and Lee, 2004), loss of spliced Tsix RNA compromised neither the Tsix-mediated Xist silencing nor the establishment of proper chromatin configuration in the Xist promoter region. These results indicate that the splicing products of Tsix RNA are unexpectedly dispensable for the chromatin modulation in the Xist promoter region. It is possible that unspliced Tsix RNA on its own is sufficient for mediating the establishment of the repressive chromatin configuration in the Xist promoter region. Alternatively, the RNA products are not required and the antisense transcription per se, especially across the Xist promoter region, is responsible for the bona fide effect mediated by Tsix. Another Tsix mutant allele, which prematurely terminates the antisense transcription in exon 4 of Tsix before it runs across the Xist promoter, appears to be defective in Tsix function despite the fact that more than 90% of the entire Tsix transcription unit is still transcribed to produce truncated RNA (T. Ohhata, Y.H., H.S. and T.S., unpublished). The antisense transcription across the Xist promoter region, therefore, seems to be crucial for Tsix-mediated Xist silencing. Considering all these findings taken together, we currently speculate that the antisense transcription is more important for the Tsix-mediated regulation than the RNA products. It is still possible, however, that the distal region of unspliced Tsix RNA covering the Xist promoter region is involved in the establishment of the appropriate chromatin conformation, perhaps through the formation of a secondary structure. Further study will help to clarify the role of RNA-antisense transcription in the antisense regulation by Tsix at the Xist locus.
Another unexpected observation in this study was the activation of the Xist locus on the mutated X chromosome in undifferentiated Xist1lox/Y and TsixΔSA/Y ES cells. Unlike in XdcY ES cells, however, Xist becomes downregulated upon differentiation in these mutant cells. This is consistent with the finding in the mutant embryos that Tsix's function is compromised on Xdc but not affected on the mutated X chromosome carrying Xist1lox and TsixΔSA. Nonetheless, the activation of the Xist locus in the undifferentiated state in Xist1lox/Y and TsixΔSA/Y as well as in XdcY ES cells suggests that in these mutant cells, the genetic alteration introduced at the Xist locus impaired the silencing mechanism at the Xist locus, which would be independent of Tsix function, in undifferentiated ES cells. It is possible that the IRES-EGFP fragment introduced at the mutated locus somehow induces the Xist promoter, although the chromatin structure as revealed by the DNaseI hypersensitive site assay showed no difference between the wild-type and mutated alleles. Alternatively, the deleted region in these mutant alleles may contain a cis regulatory element required for Xist-silencing in undifferentiated cells. It has been shown that Xist that has been expressed from the paternal allele since the 2-4 cell stage, becomes downregulated in a cell that has contributed to the inner cell mass (ICM) at the blastocyst stage (Mak et al., 2004; Okamoto et al., 2004). Tsix is expressed only from the maternal allele, most probably in the trophectoderm, at this stage (Lee, 2000; Sado et al., 2001) and, therefore, repression of the paternal Xist in the ICM should be mediated by a Tsix-independent mechanism. It is tempting to speculate that the putative cis element mentioned above, if any, might be involved in the silencing of the hitherto active Xist in undifferentiated ICM cells. It will be of interest to see if the Xist locus on the paternally derived mutated X chromosome carrying either Xist1lox or TsixΔSA is activated at the 2-4 cell stage and subsequently downregulated in the ICM cells at the blastocyst stage.
We thank Minako Kanbayashi for genotyping of mice and Tatsuya Ohhata for valuable discussions throughout the course of this study. This work was supported by grants from Precursory Research for Embryonic Science and Technology (PRESTO), Japan Science and Technology Agency (JST) and from the Ministry of Education, Science, Sports, and Culture of Japan to T.S.
- Accepted October 3, 2006.
- © 2006.