Loss of One X and the Y Chromosome Changes the Configuration of the X Inactivation Center in the Genus Tokudaia

In Eutheria, females possess two X chromosomes, whereas males carry a single X and a Y chromosome. To address the gene dosage imbalance on the X chromosome between females and males, mammalian cells have evolved a mechanism called X chromosome inactivation (XCI), wherein one of the X chromosomes in females is transcriptionally silenced [1‒3]. This process ensures equitable X-linked gene expression levels in somatic cells [4]. The XCI mechanism is governed when the X chromosome-to-autosome ratio (X/A) is ≥1, and one of the X chromosomes is inactivated to achieve a balanced ratio [5].

The mechanism underlying XCI, resulting from the XX/XY constitution, is conserved and widespread among Eutheria. Nevertheless, a few exceptional mammalian species possess unique sex chromosomes, and the genus Tokudaia (Muridae, Rodentia) stands out as a prime example. The genus includes three species, each endemic to a single island in southernmost Japan, and is faced with significant conservation challenges [6]. The Tokunoshima spiny rat (Tokudaia tokunoshimensis (TTO), 2n = 45, XO/XO) and Amami spiny rat (Tokudaia osimensis (TOS), 2n = 25, XO/XO) lack one X chromosome and the Y chromosome. The Sry gene, a sex-determining gene in eutherian and marsupial mammals, was also completely lost [7‒9]. Both males and females share a single X chromosome, and their genome sequences are nearly identical with only a male-specific 34 kb duplication upstream of Sox9 on chromosome 3 in TOS [10], resulting in equal gene dosage of the X chromosome in the XO/XO species. By contrast, the Okinawa spiny rat (Tokudaia muenninki (TMU), 2n = 44) exhibits the typical XX/XY sex chromosome system found in most mammals [11]. However, the sex chromosomes have fused with a pair of autosomes, forming recently acquired regions referred to as the “neo-X” and “neo-Y” on the X and Y chromosomes, respectively. These neo-X and neo-Y regions display partial genetic differentiation, indicating recombination suppression between them [12].

Transcaucasian mole vole, Ellobius lutescens (2n = 17), also has been known as an XO/XO species. The closely related species, Ellobius talpinus, has a 54,XX karyotype in both females and males, and the XCI mechanism is most likely still intact [13, 14]. The previous report showed that the X remained monovalent without chiasma formation in female and male meiosis in E. lutescens, XO/XO [13]. The genome sequencing of Xic in E. lutescens indicated that all genes known to control XCI are still present, and the whole region showed >90% conservation between E. lutescens and E. talpinus [15].

To achieve XCI, a complex interplay of multiple long non-coding RNAs (lncRNAs) and proteins occurs at a pivotal locus known as the X inactivation center (Xic). In both mice and humans, the Xic encompasses an approximately 660 kb region [16]. Within the Xic, critical sequences govern the counting, selection, and initiation of silencing of the future inactive X (Xi), while maintaining gene expression from the active X (Xa) [17]. Several genes reside in the Xic, including the lncRNAs Tsx (also known as protein-coding gene [18]), Xist, Tsix, Jpx, and Ftx, and protein-coding genes Cdx4, Chic1, Slc16a2, and Rlim. The cornerstone of XCI regulation is the lncRNA Xist [1, 19, 20]. Mouse Xist has six conserved repeat sequences, A-, B-, C-, D-, E-, and F-repeats, and two promoters, P1 and P2 [21‒26]. These repeats are distributed across exon 1 (A–D and F) and exon 7 (E). The pivotal role of Xist in XCI is attributed to the presence of these highly conserved repeats, particularly the A-repeat [27]. Xist lacking the A-repeat is expressed but unable to form the characteristic Xi-associated structure due to the lack of initial coiling [28]. The A-repeat includes 7–8 copies of a conserved sequence known as the AUCG tetraloop [29]. The other repeats function as binding sites for chromatin regulatory complexes, loop maintenance, localization, and late stages of XCI [22, 30, 31].

We previously sequenced the Xist gene and assessed Xist RNA expression and localization in two species with distinct sex chromosome systems, TOS (XO/XO) and TMU (XX/XY) [32]. In TOS, we observed loss-of-function mutations in the Xist gene. Notably, RT-PCR and Northern blotting analyses confirmed the absence of Xist RNA expression in both male and female TOS cells. By contrast, TMU exhibited female-specific Xist RNA expression, with localization predominantly to the long arm of the original X chromosome and partial localization to the neo-X region of Xi [33]. Interestingly, although the neo-X region in TMU does not exhibit heterochromatinization and does not contribute to Barr body formation, it forms a slightly condensed structure [33]. Other genes essential for XCI within the genus Tokudaia have not been characterized.

In this study, we compared the Xic structure in three Tokudaia species and the mouse model. We detected nine key genes within the Xic, namely, Cdx4, Chic1, Tsx, Xist, Tsix, Jpx, Ftx, Rlim, and Slc16a2, in all Tokudaia species. Through a detailed sequence comparison, we identified remarkable structural changes in lncRNA genes, Xist and Tsix, and the intergenic region Jpx-Ftx that were specific to the two XO/XO species. Furthermore, transcripts per million (TPM) estimated from RNA-seq data showed female-specific expression of Xist and Tsix in the XX/XY species, but not in the two XO/XO species. Our findings provide evidence that the lack of XCI in the XO/XO species can be explained by major changes in the Xic, shedding new light on the interplay between sex chromosome evolution and XCI functionality.

The high-quality genome assemblies and gene annotations of the genomes of three Tokudaia species were previously reported [34]; however, these annotations do not include ncRNA genes. Therefore, we predicted here the exon-intron structure of nine genes in Xic. We first predicted gene structure for TMU using the following methods: (1) an RNA-seq-based method that predicts gene structures on the basis of transcriptome sequencing results and (2) a homology method that predicts gene structures on the basis of gene sequences of related species, mouse. The two prediction methods are described below.

Gene prediction was performed using both mapping and de novo methods using RNA-seq data (TMU, DRR059296-302; TTO, DRR496898, DRR496899; and TOS, DRR426688, DRR426689) and genome sequence data [34] (TMU, PRJDB16411; TTO, PRJDB16412; and TOS, PRJDB16413). In the mapping method, HISAT2 v2.2.1 [35] was used to map RNA-seq reads to genome sequences. The mapped reads were assembled into gtf files using StringTie v2.2.1 [36]. In de novo method, RNA-seq reads were used to assemble into transcripts by Trinity v2.9.1 [37]. Redundant sequences were removed from the assembled contigs using CD-Hit v4.8.1 [38], and the contigs were aligned to the genome sequence using GMAP v2023.10.10 [39].

The mouse sequence of each gene (Cdx4, Chic1, Tsx, Tsix, Xist, Jpx, Ftx, Slc16a2, and Rlim) was used as the query of GMAP v2023.10.10 with cross-species parameter to annotate the TMU genome. Accession numbers of mouse genes are shown in online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000539294). The gene structure results predicted in TMU were used for the prediction of TTO and TOS genes according to the same methodology.

The prediction results of the two methods were manually curated. For the gene predictions of TTO and TOS, the predicted genes of TMU were aligned to the TTO and TOS genomes by GMAP, partially corrected based on the RNA-seq mapping results.

Xist repeats of Tokudaia species were predicted by comparing the sequences of all repeats in mouse [25] to the Xist sequence of Tokudaia species by using BLAT (v.36) [40]. A customized command was used to improve the sensitivity (blat -noHead -minIdentity = 70 -stepSize = 1 -tileSize = 9).

To generate the synteny plot of Xic loci, the lylaplot (https://github.com/mokuno3430/emu/tree/main/lyla_plot) tool was used. To run this tool, sequence alignment and comparison were performed by BLAST (v2.9.0) [41]. The mouse Xic genome sequence of mm39 X chromosome (from 5 kb upstream Cdx4 to 5 kb downstream Rlim, 102,360,005-103,029,892) was used for comparison with Tokudaia Xic.

Dot plot alignment for pairwise comparison of Xic was generated using the nucmer command in MUMmer (V.3.23) [42]. All output files generated were transformed into matching coordinates by using the mummerplot standard command. The matching coordinate files were plotted using gnuplot (V.5.9) [43].

The expression level of genes in Xic was confirmed. The RNA-seq reads were mapped to each species’ genome using STAR (v.2.7.9) [44]. Whole genome annotation results were used in which the Xic region was replaced with nine curated transcripts. Then, the TPM was calculated by using RSEM (V.1.3.3) [45] with default parameters for paired-end reads. A bar plot of the TPM per gene was performed using RStudio (V.2012.15.2) [46] and the package ggplot2. TATA box binding protein (Tbp) gene was used as an internal control, and its accession number of mouse is shown in online supplementary Table S2.

We compared the gene content and order in the Xic between Tokudaia species and mouse. In mouse, Cdx4, Chic1, Tsx, Tsix, Xist, Jpx, Ftx, Rlim, and Slc16a2 were present in the 660 kb Xic (102,365,005–103,024,892 bp from Cdx4 to Rlim) on the X chromosome. All of these genes were annotated to the X chromosome in all Tokudaia species and the gene order was conserved (Fig. 1). The estimated lengths of the Xic were 636, 582, and 595 kb in TMU, TTO, and TOS, respectively. Sequence identities between species are shown in Table 1. The sequence identity of Tokudaia species and mice was >86%; TTO and TOS showed the highest sequence identity of 99.12%.

Owing to its importance as the master regulator of XCI, we performed a comparative sequence analysis of Xist. We newly determined the Xist sequence of TTO and updated the results of a previous study that reported the Xist sequences of TMU and TOS [32]. All repeat regions were conserved in TMU, however, with a partial deletion in B- and C-repeat (Fig. 2; Table 2). Although A- and F-repeats were confirmed in TTO and TOS, the D-repeat was completely lost and C- and E-repeats were partially lost as a result of deletions in exons 1 and 7. A partial deletion in B-repeat was confirmed in TOS, but it was completely lost in TTO. Tsix, which is the antisense lncRNA to Xist, showed significant structural changes in the two XO/XO species (Fig. 3a). The sequence includes mouse exon 3 that was absent in all Tokudaia species, and the sequence of exon 2 was additionally lost in the XO/XO species, TTO and TOS.

By comparison of the gene structure between mouse and Tokudaia species, Tokudaia-specific changes were found in three ncRNA genes, Tsx, Jpx, and Ftx (Fig. 3b–d). Although the sequence of mouse Tsx exon 6 was present in all Tokudaia genome sequences, it was not transcribed (Fig. 3b). In Jpx, the sequence which includes mouse exon 2 was absent in the genome sequence of all Tokudaia species (Fig. 3c). The size of exon 3 was longer than that of a mouse; therefore, the total length of exons in Tokudaia species is about 800–1,000 bp longer than that of the mouse (online suppl. Table S1). Ftx had the same number of exons as a mouse, but with Tokudaia-specific features in several exons (Fig. 3d). The sequences of mouse exons 3, 5, and 10 were absent in Tokudaia species genomes. The sequences of Tokudaia exons 3, 4, and 5 showed the homologies with the mouse exons 4, 6, and 7, respectively. Tokudaia exons 6, 7, and 10 were the genus specific sequence. The exon 11 was longer than that of a mouse; therefore, the total length of exons in Tokudaia species was about 1,300 bp longer than that of the mouse (online suppl. Table S1). While these changes were observed in ncRNA genes, we confirmed well conservation in protein-coding genes, Cdx4, Chic1, Slc16a2, and Rlim (online suppl. Fig. S1; Table S1).

To visualize detailed Xic structure differences between species, dot plot alignment of all loci in mouse and each Tokudaia species was generated (Fig. 4). The dot plot illustrates the high conservation of genomic structure between mouse and TMU in the Xic (Fig. 4a). By contrast, deletions and inversions were identified in TTO and TOS (Fig. 4b, c). We focused on the two regions indicated by gray color in Figures 4b and c, and the enlarged images are shown in Fig. 4d–g. An insertion (6.9 kb) was observed in TOS Tsx (Fig. 4f), supporting the result of gene structure comparison in Figure 3b. Deletions were confirmed in Xist and Tsix (Fig. 4d and f), and these results supported the results in Figures 2 and 3a. Additionally, three deletions and one inversion were confirmed in the intergenic region between Jpx and Ftx (Fig. 4e, g). The size of each deletion was 10.2 kb, 1.8 kb, and 10.9 kb, and the size of inversion was 11.4 kb. The location and size of deletions and inversion were mostly identical in TTO and TOS.

Since we detected notable structural changes in the two genes, Xist and Tsix, in the XO/XO species, we estimated TPM values using RNA-seq data to evaluate expression patterns (Fig. 5). The expression levels of Tbp as an internal control were mostly the same between sexes in all species (Fig. 5a), supporting the validity of this TPM calculation. TMU Xist showed a high female-specific expression, while TTO and TOS showed no expression in either sex (Fig. 5b). We previously reported that the Xist lncRNA was not expressed in TOS [32]; therefore, our result supported the previous data. We here newly confirmed that the Xist lncRNA was not expressed even in TTO.

Tsix expression was quite low level in TMU (Fig. 5c). The reason for the extremely low expression level of Tsix was thought to be that this lncRNA is mainly transcribed in undifferentiated cells such as ES cells [47], and we used differentiated somatic cells (fibroblasts) from each Tokudaia species for RNA-seq analysis. Although the expression level was quite low, TMU showed female-specific expression. By contrast, different expression patterns were confirmed in the XO/XO species (Fig. 4c). Neither XO/XO species showed a female-specific expression pattern, suggesting that Tsix expression is not regulated in a sex-dependent manner observed in general XX/XY species.

We observed that the gene content and order in the Xic are largely conserved between the genus Tokudaia and mice, however, with partial structural variation in TTO and TOS. Within Tokudaia species, TTO and TOS showed the highest sequence identity. A previous molecular phylogenetic analysis revealed that TMU (XX/XY) diverged first among the three species, and TOS and TTO (XO/XO) are phylogenetically closely related [9]. Therefore, the patterns of sequence identity likely reflect the phylogenetic relationships within the genus.

Protein-coding genes showed a consistent size and high identity between Tokudaia species and mouse, whereas partial variation was discovered in the regions within ncRNA genes. The variation in the three ncRNA genes, Tsx, Ftx, and Jpx, represent a distinct mark of the Tokudaia species, and these variations are not related to sex chromosome constitution.

Only in the XO/XO species, remarkable structure changes were observed in Xist and Tsix that are a key regulator of XIC and an integral part of its downregulation, respectively. All functional repeats were completely conserved in TMU, suggesting that the Xist RNAs of TMU are functionally conserved. This finding supports the results of our previous study of XCI in TMU [32, 33]. By contrast, TTO and TOS exhibited a partial loss of exons 1 and 7, including C-, D-, and E-repeats, and B-repeat is lost in TTO. Notably, the C-repeat recruits the polycomb group complex. Polycomb group works by maintaining the silent state of the Xi through the action of B- and C-repeats [48]. In the same way, the E-repeat is essential for anchoring Xist RNA molecules to the Xi territory by facilitating cis-localization [49]. The function of the D-repeat is not well-characterized; the region is more heavily degenerated in rodents than in other mammals, suggesting that the mouse D-repeat is not functionally relevant [22, 25]. Based on these previous findings, the deletions in Xist in TTO and TOS could affect the recognition and maintenance of the future Xi. The Xist lncRNAs are not expressed in the XO/XO species; however, our findings indicate that even if they are expressed, they are incapable of producing a successful and lasting XCI in the XO/XO species. The deletions of Tsix in TOS and TTO could also cause XCI regulation failure.

XCI regulatory network is orchestrated by two topologically associating domains (TADs) on the X chromosome, TAD-E and TAD-D, each playing a vital role in governing the XCI process. TAD-E is positioned at the chromosome end, while TAD-D is located closer to the centromere [50]. TAD-E exerts a positive effect, along with Ftx, Jpx, Rlim, and Slc16a2, on Xist production during early development [51, 52]. Conversely, Rlim ensures the rapid and stringent inhibition of gene expression in the future Xi [53]. By contrast, TAD-D, which includes Chic1, Cdx4, and Tsx, participates in negative regulation. Tsx and Tsix, the latter being the antisense lncRNA to Xist, safeguard the future Xa from inactivation by blocking Xist coiling on the Xa [18, 54]. Finally, Chic1 and Cdx4 contribute to the stabilization of the chromatin landscape, which is essential for the accurate silencing of the X chromosome [55].

We confirmed large deletions and inversion in the intergenic region of Jpx-Ftx in the XO/XO species. The intergenic region of Jpx–Ftx, that within TAD-E, upregulates Xist. Notably, the loss of the broader Jpx–Rlim region reduces XCI [53], emphasizing the significance of this region in mediating chromatin interactions in the Xic. The presence of TADs, including TAD-D and TAD-E, makes essential the interactions between genes and intergenic regions for initiating and maintaining XCI [50]. Successful XCI will create regions of heterochromatin and euchromatin, the latter harboring genes that escape XCI in the Xi [1, 56]. There is evidence that Ftx serves as an intermediate in the cascade of events leading from Rlim to Xist activation [51]. We observed three deletions and one inversion in Jpx-Ftx in the XO/XO species, both in terms of size and sequence. Mutations in this region likely originated in a common ancestor, raising the possibility that the loss of functional XCI occurred in the ancestral lineage of XO/XO species, likely due to the loss of one X and the Y. We hypothesized that such structural changes would be more likely to occur as a result of having a single X chromosome and self-synapsis. In XO mice, an XO pachytene oocyte in which an unsynapsed univalent X is silenced shows pachytene arrest. By contrast, an XO pachytene oocyte with a self-synapsed X chromosome remains active [57]. Self-synapsis potentially facilitates chromosomal rearrangements. Since it is not possible to obtain tissues from the endangered spiny rat for observations of synapsis, sequence analyses of whole X chromosomes are an important approach to confirm our hypothesis.

The Xist lncRNA was not expressed in the XO/XO species, and similarly, female-specific expression of Tsix was not observed in them. Since individuals with XX chromosomes have not been confirmed in TOS and TTO, XX is thought to be lethal in the early stages of development due to failure of XCI. We therefore hypothesized that the significant structure changes in the intergenic region of Jpx-Ftx resulted in the inability to perform the XCI, and, as a result, a lack of Xist expression. To examine whether the deletions and inversion identified here affect TAD formation, we would like to confirm chromatin accessibility using the Hi-C method. Our study marks the first documentation of the Xic structure in XO/XO species. Our findings provide a basis for investigating XCI evolution along with sex chromosome evolution and contribute to a better understanding of the mechanism underlying XCI.

The authors would like to thank Yuta Mochimaru and Kentaro Matsuoka (School of Life Science and Technology, Tokyo Institute of Technology) for their valuable work and suggestions to generate the data for this study.

This study did not involve any human participants or specimens. An ethics statement was not required for this study type, and no human or animal subjects were used.

The authors have no conflicts of interest to declare.

This work was supported by JSPS KAKENHI Grant No. 16H06279 (PAGS), 22H02667, and 23K23930.

Conceptualization and project administration: A.K. Data curation: L.M.-C. and M.O. Investigation: L.M.-C. Resources: T.J. Supervision: I.Y., S.M., M.O., T.I., and A.K. Writing and original draft preparation: L.M.-C. and A.K. Writing, review, and editing: L.M.-C., I.Y., S.M., M.O., T.I., and A.K.

The accession numbers of nine genes in Xic of Tokudaia species are LC788376-84 (TMU), LC788394-402 (TTO), and LC788385-93 (TOS).