What makes the maternal X chromosome resistant to undergoing imprinted X inactivation?

Abstract

In the mouse, while either X chromosome is chosen for inactivation in a random fashion in the embryonic tissue, the paternally derived X chromosome is preferentially inactivated in the extraembryonic tissues. It has been shown that the maternal X chromosome is imprinted so as not to undergo inactivation in the extraembryonic tissues. X-linked noncoding Xist RNA becomes upregulated on the X chromosome that is to be inactivated. An antisense noncoding RNA, Tsix, which occurs at the Xist locus and has been shown to negatively regulate Xist expression in cis, is imprinted to be expressed from the maternal X in the extraembryonic tissues. Although Tsix appears to be responsible for the imprint laid on the maternal X, those who disagree with this idea would point out the fact that Tsix has not yet been expressed from the maternal X when Xist becomes upregulated on the paternal but not the maternal X at the onset of imprinted X-inactivation in preimplantation embryos. Recent studies have demonstrated, however, that there is a prominent difference in the chromatin structure at the Xist locus depending on the parental origin, which I suggest might account for the repression of maternal Xist in the absence of maternal Tsix at the preimplantation stages.

This article is part of the themed issue ‘X-chromosome inactivation: a tribute to Mary Lyon’.

1. Imprinted and random X inactivation in the mouse

X inactivation is one of the most important epigenetic gene regulations that operate in early embryogenesis in female mammals [1] and its defect results in embryonic lethality at the periimplantation stages [2]. It has been assumed that X inactivation and cellular differentiation are closely linked and mutually affect their respective processes [3]. There are two waves of X inactivation in the mouse. In the extraembryonic lineages giving rise to the future placenta and some extraembryonic membranes, the paternally derived X chromosome is preferentially inactivated [4], whereas in the embryonic lineage, which will produce all tissues of the fetus, either of the two X chromosomes is inactivated in a random fashion regardless of the parental origin. X inactivation first takes place around the 4–8 cell stages in all the blastomeres in favour of the paternal X [5,6]. By the time the embryos have developed to the blastocyst stage, the trophectoderm has differentiated to form the outer layer of a blastocyst with the undifferentiated inner cell mass (ICM) cells inside, a part of which differentiate into the primitive endoderm 1 day later. While imprinted paternal X inactivation is maintained in the trophectoderm and primitive endoderm, the inactive paternal X tentatively restores its transcriptional activity in the remaining ICM cells that form the epiblast [5,7]. Subsequently, random X inactivation occurs as the epiblast cells differentiate. Several lines of evidence suggest that the paternal X is preferentially inactivated due to the presence of an imprint laid on the maternal X [8–13], which makes it extremely resistant against chromosome silencing, although the possible presence of an imprint laid on the paternal X has not been completely excluded. An X-linked long noncoding Xist (X inactive-specific transcript) RNA plays a pivotal role in both imprinted and random X inactivation. Xist RNA is about 18 kb in length, and becomes upregulated on the future inactive X chromosome and coats it in cis, probably to form a platform for proteins involved in heterochromatinization and gene silencing. Targeted disruption of Xist clearly demonstrated that Xist is essential for X inactivation to occur in cis, and therefore an X chromosome deficient for Xist does not undergo X inactivation [2].

2. Noncoding RNAs in the X inactivation centre

The Xist gene is mapped on the X chromosome in the X inactivation centre (Xic/XIC), which was identified as a master controlling region required for the X chromosome to be inactivated by classic cytogenetic studies using translocations between an X chromosome and an autosome [14,15]. There are several noncoding RNAs transcribed in Xic besides Xist, which include Jpx/Enox, Ftx, and Xist-AR (figure 1 and table 1). Jpx transcribed 10 kb upstream of Xist in an opposite direction relative to Xist has been reported to interact with and evict CTCF bound at the promoter region of Xist and consequently to activate transcription of Xist [16]. Ftx was also suggested to positively regulate Xist via a DNA methylation-mediated mechanism at the Xist promoter [17] although it was shown afterward that Ftx is dispensable for imprinted X inactivation [18]. Xist-AR was recently identified as an antisense RNA occurring at the 5′ region of Xist exon1, and was suggested to upregulate Xist transcription [19]. In addition to these noncoding RNAs involved in Xist activation, Tsix has been identified as an RNA that is antisense to Xist and that covers the entire transcription unit of Xist and negatively regulates Xist in cis [22–24]. The transcription of Tsix is regulated by the Xite locus producing an enhancer RNA encoded upstream of the major promoter of Tsix [20] (figure 1). Another recently identified noncoding RNA, Linx, located 100 kb upstream of Xite, has been suggested to play a role as a long-range transcriptional regulator for Tsix [21] (figure 1). Tsix appears to exert its negative effect on Xist transcription at the onset of X inactivation when the decision of whether or not Xist is monoallelically upregulated is made. Xist and Tsix are reciprocally imprinted in the extraembryonic lineages to be expressed from the paternal and maternal allele, respectively. Targeted disruption of Tsix, when maternally inherited, causes ectopic expression of Xist on the maternal X after implantation and its subsequent inactivation in the extraembryonic tissues of both female and male embryos, which never occurs in wild-type, suggesting that Tsix is required for Xist to stay repressed on the X that remains active [22,24]. Although it is still unknown which of Tsix RNA or the action of antisense transcription is important for Tsix's function, it has been shown that transcription of Tsix has to cover the promoter region of Xist to exert Tsix's negative effect on Xist [25].

Figure 1.

Positive regulation of Xist and Tsix by noncoding RNAs mapped in the X inactivation centre region. Noncoding RNAs transcribed in the same direction as Xist are delineated on the top and those transcribed in the same direction as Tsix are at the bottom of the line representing the chromosomal DNA. Centromere (cen) on the left and telomere (tel) on the right. (Online version in colour.)

Table 1.

List of noncoding RNAs that regulate each of Xist and Tsix at Xic.

positive regulator for Xist
Jpx/Enox	Sun et al. [16]
Ftx	Chureau et al. [17]; Soma et al. [18]
Xist-AR	Sarkar et al. [19]
positive regulator for Tsix
Xite	Ogawa et al. [20]
Linx	Nora et al. [21]

3. An imprint laid on the maternal X to resist inactivation

The presence of an imprint that renders the maternal X resistant to undergoing inactivation in the extraembryonic tissues was originally pointed out by Lyon & Rastan [26] and a series of studies by Takagi and colleagues experimentally revealed using embryos with a supernumerary maternal X chromosome that the X chromosomes derived from the mother indeed bear such an imprint effective only in the extraembryonic tissues, while this imprint is apparently erased or ineffective in the embryonic tissue [8,10–13]. The presence of two active X chromosomes results in substantial loss of the polar trophoblast lineages such as the ectoplacental cone and extraembryonic ectoderm, suggesting that these tissues are probably most susceptible to the presence of two active X chromosomes. In contrast, the visceral endoderm, a derivative of the primitive endoderm, appears to be less affected by the presence of two active X chromosomes.

It should be pointed out, however, that a small fraction of cells in the extraembryonic tissues of parthenogenetic embryos somehow manage to inactivate one of the two maternal X chromosomes. This seems to be achieved by ectopic upregulation of Xist on either of the maternally derived X chromosomes at around the morula/blastocyst stages [27,28]. Although the timing of Xist upregulation in parthenogenetic embryos is relatively late as compared to the upregulation of the paternal Xist in normal female embryos, these observations suggest that the imprint of the maternal X can be overridden in parthenogenetic embryos. In fertilized embryos, however, this imprint appeared to be so rigid that the maternal X chromosome does not undergo inactivation in the extraembryonic lineages even under circumstances in which the paternal X never becomes inactivated due to the lack of Xist [2]. Interestingly, embryonic lethality caused by paternal deficiency of Xist is sometimes fully rescued if combined simultaneously with maternal deficiency of Tsix. This results in an inverse situation of imprinted X inactivation in the extraembryonic tissues, where the maternal X is inactivated and the paternal X remains active [24]. This demonstrates that it does not necessarily have to be the paternal X that undergoes inactivation in the extraembryonic tissues as long as dosage compensation of the X-linked genes is achieved. It has been shown that the maternal imprint of the X chromosome is established during the period when the primary oocyte grows and increases in volume during prophase of meiosis I [9]. Although neither X chromosome undergoes inactivation in the majority of cells in the extraembryonic tissues of parthenogenetic embryos [10], which contain two sets of the genome derived from the fully grown oocyte, if they are reconstituted using the genome of the nongrowing and fully grown oocyte, the X chromosome derived from the nongrowing oocyte undergoes inactivation in the extraembryonic tissues. This clearly demonstrates that the imprint laid on the maternal X is acquired during oocyte growth.

It would be reasonable to assume that this imprint prevents upregulation of maternal Xist in the extraembryonic tissues. Given the fact that the strict repression of maternal Xist is abolished if Tsix is disrupted on the maternal X [22,24], it seems likely that the maternal expression of Tsix is responsible for preventing upregulation of Xist on the maternal X, and thereby mediates the maternal imprint against X inactivation. Those who disagree this idea would point out the fact that Tsix expression has not yet been initiated when only the paternal Xist becomes preferentially upregulated at the 4- to 8-cell stages [24], and would therefore claim that Tsix would have nothing to do with the imprint preventing Xist upregulation on the maternal X at these stages. A possible difference in the chromatin structure between the maternal and paternal Xist locus may, however, account for the monoallelic upregulation of Xist only on the paternal X at these stages. Since the sperm genome undergoes global exchange from protamine to histone and becomes decondensed in the zygote, there would be some differences in the chromatin structure, and therefore in the accessibility of transcription machinery, between the maternal and paternal genome. It is possible that these differences cause a bias in the timing when a subset of genes starts to be expressed in the early preimplantation embryos in favour of the paternal alleles. Although a recent study demonstrated that such parental biases in the timing of gene activation were not evident at the early preimplantation stages [29], this does not necessarily rule out the existence of rare loci that behave in that way. Since Tsix expression initiates on the maternal X at around the morula stages, if the influence of differences in the chromatin structure between the two parental Xist alleles lasts for the first three cell divisions or so, maternal Xist would stay repressed without Tsix during the early preimplantation stages. Tsix upregulated on the maternal X by the time when the chromatin structure at the Xist locus becomes equivalent between the paternal and maternal X via multiple rounds of DNA replication would, in turn, take over the role in preventing the upregulation of Xist on the maternal X in the cells that contribute to the extraembryonic lineages.

4. Maternal Xist may simply not be ready for transcription in the early preimplantation embryo

A recent study by Fukuda et al. showed that overexpression of Kdm4b histone H3K9me3 demethylase [30] by introduction of its mRNA in the oocyte prior to parthenogenetic activation results in upregulation of Xist at the 4-cell cell stage, which is rather early for parthenogenetic embryos [31]. This experiment suggests that decondensation of the maternal genome by demethylation of H3K9me3 facilitates the expression of Xist on the maternally derived X chromosomes. They further showed by DNA-FISH that the chromosome region spanning the Xist locus becomes condensed during the growth of oocytes at the prophase of meiosis I, when the maternal imprint against X inactivation is established [32,33]. Although the condensed maternal Xist locus relaxes during early preimplantation development, the extent of the decondensation across the Xist loci derived from normally developed oocytes is significantly smaller than the extents across the Xist loci derived from nongrowing oocytes as well as those on the paternal X in androgenetic embryos. These studies support the idea that the Xist locus on the maternal X is more condensed than that on the paternal X in normal embryos and may need a few rounds of DNA replication to become permissive to transcription. Recent ultra-sensitive ChIP-seq analysis revealed widespread and relatively long-lasting parent-of-origin asymmetries in preimplantation mouse embryos [34,35].

This may account for our recent finding that the Xist^CAG allele, which is generated by replacing the endogenous Xist promoter with a heterologous CAG promoter to facilitate its selective monoallelic expression in many types of cells, exhibits a parent-of-origin-specific difference in the timing of its upregulation in the preimplantation embryos [36]. Xist^CAG starts to be expressed from the 4- to 8-cell stages upon paternal transmission, whereas it does not become upregulated until the blastocyst stage upon maternal transmission. The differential behaviour of the CAG promoter introduced at the Xist locus could possibly be due to the difference in the chromatin structure depending on the parental origin. The CAG promoter in this context appears to be controlled in the same way as the endogenous Xist promoter in the early preimplantation stages. It is tempting to speculate that the difference in the chromatin structure at the Xist promoter between the paternal and maternal X underlies the differential behaviour of the Xist allele according to the parental origin during the preimplantation development. Given the finding that the endogenous Xist promoter is enriched in H3K9me3 when transmitted to the zygotes [32], it is likely that the CAG promoter is also modified in the same manner as the endogenous one during oogenesis. If this is the case, it raises the question of how the promoter region of Xist becomes targeted by the machinery for the repressive modifications such as H3K9me3 in a sequence-independent manner.

5. Concluding remarks

The presence of an imprint laid on the maternal X chromosome was predicted based on the cytogenetic observation by Takagi and colleagues that when multiple maternal X chromosomes are present, they do not undergo inactivation in the extraembryonic tissues of the early postimplantation embryo. This imprint is established in the primary oocytes while they are growing in prophase of meiosis I. Since functional loss of Tsix results in ectopic inactivation of the maternal X, it is likely that the imprinted maternal expression of Tsix underlies the molecular basis for the imprint, which makes the maternal X resistant against inactivation in the extraembryonic lineages. I propose that the mechanisms underlying the resistance of the maternal X against inactivation in the extraembryonic tissues of the postimplantation embryo pointed out by Takagi and colleagues should be distinguished from those that operate at the preimplantation stages, when only the paternal but not maternal Xist is expressed. Available evidence suggests that the chromatin structure at the Xist locus is different between the paternal and maternal X chromosomes during very early stages of preimplantation development. I envision that such a difference in the chromatin structure allows the imprinted paternal expression of Xist in the preimplantation embryos while Tsix has not yet been expressed. Although the chromatin structure of the Xist locus would become equivalent on the maternal X and paternal X through multiple rounds of DNA replication, the imprinted expression of Tsix occurring on the maternal X by then should take over the role of keeping Xist repressed afterward. It is tempting to speculate that the lack of imprinted expression of Xist in the preimplantation embryos in non-rodents [37] may be related to the fact that they undergo more cell division or DNA replication by the time XIST starts to be expressed [38].

Data accessibility

This article has no additional data.

Competing interests

I declare I have no competing interests.

Funding

T.S. is funded by a Grant-in-Aid for Scientific Research on Innovative Areas from the Ministry of Education, Science, Sports, and Culture of Japan (16H01320, 17H05606).