The compleX balancing act of controlling X-chromosome dosage and how it impacts mammalian germline development

The emergence of distinct sex chromosomes from autosome pairs during evolution, and the subsequent degradation of one sex chromosome by meiotic recombination in the heterogametic sex, would lead to a potential gene dosage imparity between autosomes and sex chromosomes and between males and females [1,2]. Different species have developed diverse dosage compensation mechanisms to overcome this problem [3,4]. In placental mammals, at least three mechanisms are at play; X-chromosome upregulation (XCU), X-chromosome inactivation (XCI) and X-chromosome reactivation (XCR). To balance X-chromosome gene dosage with autosomal levels, genes on the single active X in XY males and XX females are transcriptionally upregulated by XCU [5]. X-dosage parity between males and females is achieved through epigenetic silencing of one of the two X chromosomes in females, in a process called XCI [6]. Although the mechanisms of XCI demonstrate variation between mammalian subclasses [7,8], in eutherians the long noncoding RNA X-inactive specific transcript (XIST) is the indispensable master regulator of the process [9–12]. XIST becomes monoallelically upregulated and acts in cis to coat the inactivating X chromosome, resulting in the recruitment of a cascade of effectors that elicit heritable large-scale epigenetic gene silencing and heterochromatization [13]. Finally, this leads to the emergence of the Barr body, with the inactive X chromosome becoming a morphologically characteristic hallmark of the female cell nucleus [14]. XCI is maintained in somatic cell lineages, however, in the germline it is reversed, resulting in the emergence of female germ cells with two active X chromosomes, by a process coined XCR [15–17].

XCR co-occurs with genome-wide epigenetic reprogramming in the female germline, where a clean slate is generated for the propagation of genetic material to the next generation [18–20]. This, in turn, prepares the cells for the initiation of meiosis during oogenesis, and ultimately the acquisition of totipotency upon fertilisation [21]. Currently, the mechanisms and biological roles of XCI, XCU and XCR in female germ cell development and oogenesis remain elusive.

In this review, we aim to discuss the significance of the intricate interplay of the different X-chromosome dosage compensation mechanisms and their unique dynamics and role in female embryonic and germ cell development. Particularly, we want to address to which degree the status of the X chromosome is a cause, or a consequence, of normal germ cell development. We discuss the current understanding within the field with a focus on the mouse model system, and suggest gaps in our knowledge that would benefit from further investigation, including in humans.

In female mice, XCI is first initiated at the 2–4 cell stage of embryonic development and occurs specifically on the paternal X chromosome, while the maternal X chromosome remains active [22–25]. This imprinted form of XCI (iXCI in Figure 1) is continuously maintained in the mouse extraembryonic lineages of the primitive endoderm and trophectoderm and later on in the placenta, but is reversed in the developing epiblast of the inner cell mass of the E(mbryonic day) 4.5 blastocyst, which gives rise to all embryonic cell lineages including germ cells [23,26–28]. Imprinted XCI is controlled by the paternal expression of Xist and essential for female embryonic development, as paternal Xist deletion leads to postimplantation lethality [12]. A non-canonical H3K27me3 imprinting mark is established during oocyte growth at the Xist locus, which ensures that Xist remains off on the maternal X chromosome during preimplantation development [29–31]. Subsequently, Xist remains repressed in the extraembryonic lineages by maternal expression of Tsix, the noncoding antisense repressor of Xist [32,33]. In addition to suppression of Xist on the maternal X chromosome, imprinting of the paternal X chromosome in the male germline might be at play, which mediates Xist-independent silencing of repeats [24] and promotes Xist expression during mouse preimplantation development [34]. The process of imprinted XCI is not conserved among all placental mammals and does not occur in humans [35] (Figure 2), whereas it is the only form of XCI in marsupials [39], where it is controlled by a different noncoding RNA called Rsx, the marsupial equivalent to Xist [8,40].

Upon differentiation of the mouse postimplantation epiblast, ultimately giving rise to the embryo proper, a second wave of so-called random XCI (rXCI) is initiated, where either the maternal or paternal X is inactivated and clonally maintained throughout life in somatic cell lineages [6,41–43] (Figures 1 and 2). The timing of rXCI initiation upon differentiation of the epiblast indicates a link between naive pluripotency and the status of the X chromosomes. The presence of two active X chromosomes in mouse embryonic stem cells (ESCs) favours pluripotency through inhibition of GSK3 and MAPK signalling, thereby stabilising the pluripotency network [44]. Dusp9, an X-linked negative regulator of the MAPK pathway, has been shown to play an important role in establishing lower global DNA-methylation levels in XX mouse ESCs [45], compared with XY or XO cells, through its increased dosage from the two active X chromosomes [46–48]. A second X-linked gene, Klhl13, has been recently identified to be involved in promoting pluripotency factor expression, thereby causing delayed differentiation of XX cells due to higher Klhl13 dosage when compared with XO or XY cells [48]. Thus, XCI of dosage-sensitive pluripotency regulators contributes to the differentiation kinetics during female pluripotency exit. In the absence of rXCI due to Xist mutation, XX ESCs show delayed pluripotency exit, cell death and frequent X-chromosome loss [44,49]. In vivo, surprisingly, rXCI might not be an absolute requirement for mouse embryonic development according to a study using conditional Xist deletion in the postimplantation epiblast, in which some embryos survived to term [50]. Nevertheless, partial dosage compensation has been observed, indicating either delayed deletion of Xist after initiation of XCI, or potential alternative dosage compensation mechanisms at play, such as a lack of XCU (see chapter below). It will be important to address in future studies the importance and biological function of rXCI for mouse development in detail and why imprinted XCI in the mouse placenta is essential for viability and cannot be compensated for by other mechanisms [12].

After down-regulation of the naive pluripotency programme during the transition from the pre- to postimplantation epiblast, the programme becomes re-established in the nascent primordial germ cell (PGC) lineage, which is specified in the proximal epiblast around embryonic days 6 and 7 in mice [51–53]. The proximal epiblast cells with competence for PGC fate thereby show high expression of pluripotency genes and a more heterogeneous degree of XCI compared with non-germ cell-competent epiblast cells, with proximal epiblast cells displaying either an absence or incomplete level of X-chromosomal gene silencing [43,54]. In addition to the expression of pluripotency factors such as PRDM14, OCT4, SOX2 and NANOG [55,56], PGCs reacquire an underlying pluripotent-like capacity, which can be uncovered by their potential to give rise to embryonic germ cells (EGCs) in culture. Similar to ESCs, EGCs exhibit self-renewal in addition to the ability to contribute to the three germ layers; ecto-, meso- and endoderm when differentiated in vitro or when injected into mouse blastocysts to form chimaeras [57,58]. Furthermore, female XX EGCs display DUSP9-mediated DNA-hypomethylation when compared with male XY EGCs, indicating the same X dosage-dependent mechanisms being at play as in ESCs [46]. Recently we found that similar to the germ cell-precursors in the epiblast in vivo [43,54], female in vitro-derived primordial germ cell-like cells (PGCLCs) undergo attenuated and heterogenous XCI, with part of the PGCLCs retaining two active X chromosomes (XaXa cells) and another population showing incomplete XCI (XaXi cells) when compared with somatic cells [54]. XaXa PGCLCs exhibited increased expression of naive pluripotency genes such as Zfp42/Rex1 or Esrrb, and a shortened ESC-like cell cycle, when compared with X-inactivated XaXi PGCLCs [54]. Furthermore, XaXa PGCLCs displayed a higher EGC derivation capacity under naive 2i + LIF culture conditions compared with their XaXi counterparts. These data may suggest an enhanced readiness for pluripotency reacquisition as a result of having two active X chromosomes, which would align with the stabilisation of the pluripotency network in ESCs by double X-dosage [44,47,48]. When we further differentiated XaXa and XaXi PGCLCs in vitro towards early meiotic stages and the onset of oogenesis, unexpectedly the cells, which had two active X chromosomes at the PGCLC stage, demonstrated a lower meiotic competence compared with their XaXi counterparts [54]. Specifically, they were less likely to enter and progress through meiotic prophase I under culture conditions that promote meiosis of PGCLCs, either by aggregation with gonadal somatic cells or by culture on feeders with the addition of meiosis-inducing cytokines [59,60]. These data hint towards a link between the X-chromosome status in PGCLCs and their later developmental competence. Further functional testing will be necessary to establish if a lack of XCI in PGCLCs is the cause or a consequence of abnormal germ cell development and how mechanistically temporal control of X-dosage might be important. Furthermore, while PGC-competent precursor cells frequently show incomplete or a lack of XCI also in vivo [43,54], it has not yet been determined by allele-specific RNA-Seq, at which frequency XaXa PGCs might exist. Stainings for X-inactivation marks such as H3K27me3 or Xist RNA suggest that they could comprise ∼20% of PGCs at day E7.5 of development, after PGCs are specified [16,17]. Such XaXa PGCs might disappear eventually, perhaps due to selection mechanisms such as apoptosis. However, in the case that such cells can persist in vivo, we would speculate that XaXa PGCs could be a source of germ cell tumours, due to the rapid self-renewal and pluripotency properties that are conferred by possessing two active X chromosomes [54]. It remains to be tested whether all PGCs in vivo eventually undergo XCI, and if not, what would be the fate of such aberrant cells.

XCR, the reverse process to XCI, by which the inactive X chromosome gets reactivated into an active state, occurs naturally at two points in mouse development (Figures 1 and 2). Firstly, imprinted XCI is reversed by XCR in the epiblast of the E4.5 blastocyst [23,26–28], and the subsequent rXCI is erased by XCR specifically in the germline [15–17]. Reversal of XCI related to pluripotency can also be studied in vitro, for example by cell fusion of somatic cells with pluripotent stem cells (reviewed in [61]), or through the reprogramming of somatic cells into induced pluripotent stem cells (iPSCs) [62], which has become an effective model for investigation of XCR dynamics and mechanisms [28,63–67]. Furthermore, in vitro model systems for X-inactivation and its subsequent reactivation in the germ cell lineage have been developed [59,68] and recently applied to gain insight into the biological role of X-dosage compensation in the germline [54].

A critical step during XCR, common to all systems in which it occurs, is the down-regulation of Xist expression, which is a prerequisite for the structural and epigenetic remodelling of the inactive X into an active state. Multiple control elements feed into Xist regulation at the X-inactivation centre (Xic), the region which harbours the Xist locus itself and its cis-regulators. The Xic contains both Xist activators, such as the noncoding Jpx [69,70], Ftx [71,72], Xert [73] and the protein-coding Rnf12/Rlim [74,75], but also Xist repressors such as its antisense transcript Tsix [33,76] and the noncoding element Linx [77]. It is, therefore, not surprising that Xist down-regulation during XCR may be achieved by a combination of different mechanisms. During iPSC-reprogramming, the down-regulation kinetics of Xist and its activator Jpx correlate tightly, and Jpx knockdown in somatic cells is sufficient to decrease Xist expression, suggesting Jpx down-regulation to be one of the mechanisms to shut down Xist during XCR [65]. Recently, GATA transcription factors have been reported as activators of Xist expression by binding to distal enhancers, whereby Xist upregulation is impaired in GATA1/4/6 triple mutant embryos during preimplantation development [78]. As GATA factors are specifically down-regulated in the mouse blastocyst epiblast [79], it will be interesting to examine if this could contribute to the Xist down-regulation during XCR. Furthermore, Xist extinction is mirrored by the upregulation of the naive pluripotency network during reprogramming, which mediates both direct and indirect repression of Xist [63–65]. Pluripotency factors such as PRDM14, NANOG, ZFP42/REX1, SOX2 and OCT4 bind in different combinations at regulatory hubs along the Xic [80], and thereby cause repression of the Xist activator Rnf12 [28,81], upregulation of the Xist repressor Tsix [82,83] and repression of Xist itself [28,75,84]. Indeed, deletion of Nanog or Prdm14 resulted in XCR defects in mouse blastocysts and during iPSC-reprogramming [28,85], indicating the importance of the pluripotency network in shutting down Xist during XCR.

In the mouse germline in vivo, XCR is initiated in PGCs in the hindgut as they migrate towards the developing gonads. Xist expression and the associated H3K27me3 accumulation on the inactive X thereby both get erased gradually in a cell-heterogeneous manner between E8.5 and E11.5 [16,17,86,87]. Like in blastocysts and during iPSC-reprogramming [28], the germ cell and pluripotency factor PRDM14 plays an important role in this process in addition to its function for genome-wide epigenetic reprogramming, which occurs concomitantly during PGC migration [88]. This includes global upregulation of the H3K27me3 and down-regulation of the H3K9me2 histone marks and repression of the DNA-methyltransferase machinery in PGCs, thereby contributing to their low global DNA-methylation state [56,88–91]. Reactivation of X-linked genes occurs subsequently between ∼E10.5 and E14.5 as mouse PGCs colonise the gonads and enter meiosis, and based on the few X-linked genes studied thus far, the reactivation kinetics differs from gene to gene and is heterogenous between cells [16,92]. An X chromosome-wide study using allele-specific single-cell RNA-Seq, as it has been done for XCR in mouse blastocysts [27] or for XCI and XCU kinetics in pre- and postimplantation embryos [43,93,94], will be needed to obtain a full picture of the XCR kinetics in mouse PGCs in vivo. Studies on XCR in mouse blastocysts during iPSC-reprogramming and germ cell in vitro differentiation have shown that XCR is a progressive process where genes are reactivated at different times [27,54,64–66]. During iPSC-reprogramming, early reactivated genes are located within a 3D-cluster in close proximity to constitutively active escapee genes, suggesting that reactivation initiates nearby already active open chromatin regions [64,65]. Furthermore, it has been reported that genes reactivating early during XCR in mouse epiblast show enrichment in MYC binding sites [27], and proposed that the pluripotency factor ZFP42/REX1 may also facilitate reactivation of X-linked genes during iPSC-reprogramming [66]. Functional studies will be needed to confirm if pluripotency factors, besides their established role in repressing Xist, also play a direct role in reactivating X-linked genes during XCR. Furthermore, it will be interesting to find out to which degree the order of genes, in which they are reactivated, is conserved between the different systems where XCR occurs. For example, the cell type of origin, from which iPSCs are derived, could influence the epigenetic state of the inactive X and additionally express different sets of transcription factors during reprogramming, thereby affecting the kinetics and order (early vs. late) in which X-linked genes are reactivated. It might be different again for XCR in PGCLCs, which are derived from early epiblast cells, which show incomplete and heterogenous XCI [43,54] and might be, therefore, more prone to reactivation than somatic cells used for iPSC-reprogramming experiments. Comparing the kinetics of XCR in these different systems will reveal if gene reactivation propensity is hard-wired into the X-chromosome sequence and mediated by a common set of trans-acting factors, or if it depends on the cell type-specific context.

Despite its importance, Xist down-regulation by itself is not sufficient for XCR to take place. In addition, multiple layers of epigenetic silencing including DNA methylation and repressive histone marks such as H3K27me3 need to be erased and properties of active chromatin such as histone acetylation and active RNA-Polymerase II need to be regained [27,63,64,95]. Another important step during XCR is the remodelling of the 3D X-chromosome structure, from a compact and transcriptionally inactive state into an open active chromatin configuration [65,96]. Cohesin is an essential factor in this process, as knockdown of the cohesin complex member Smc1a during iPSC-reprogramming results in abnormalities during the restructuring of the X chromosome and in defects in X-linked gene reactivation [67]. XCR occurs with markedly different kinetics in the blastocyst epiblast (within 1 day) [27] in contrast with the germ cell lineage (several days) [16], when considering the time it takes from Xist down-regulation to X-linked gene reactivation. This suggests important mechanistic differences between the reactivation of imprinted XCI in the epiblast versus the reversal of rXCI in germ cells. Potential explanations could be a different epigenetic status of the inactive X after imprinted and rXCI, or a difference in the factors mediating the epigenetic remodelling. For example, DNA methylation of X-linked gene promoters is critical for the maintenance of random, but not imprinted, XCI [63,95,97], and the DNA-methylation states between the active and inactive X chromosomes are equally low in mouse preimplantation embryos already prior to Xist down-regulation and XCR [98]. In the germ cell lineage, however, DNA-demethylation occurs slowly and gradually on the reactivating inactive X chromosome mostly after Xist has been down-regulated, and is only completed once germ cells have entered the gonads. This is in line with the slower X-linked gene reactivation dynamics when compared with the epiblast [99]. Further studies will be needed to dissect the specific mechanism of DNA-demethylation on the inactive X chromosome and its contribution to gene reactivation during XCR in the germ cell lineage.

In addition to how XCR is regulated mechanistically, it remains an open question of what may be its biological function in female germ cell development. After germ cells complete XCR, they enter oogenesis and initiate the meiotic programme until the diplotene stage of prophase I, at which point the early oocytes are held in a state of the meiotic arrest until oocyte maturation and meiosis resume during adolescence. It has been demonstrated that the presence of two X chromosomes contributes to the proper progression through meiosis in oogenesis. In the case of cells with a single X-dosage, either as XO or XY, meiotic progression is perturbed [100–102]. This, at least in part, is due to the mispairing of sex chromosomes during prophase I, and may also be a result of differential dosage of X-transcripts [87,102,103], although more evidence is required before this can be confirmed. In relation to the XCR process, a question that remains to be answered is whether an active, euchromatic X chromosome could pair with an inactive, heterochromatic X in the case that XCR fails, or if this would lead to meiotic pairing defects and trigger a meiotic check-point, leading to a potential loss of oocytes. Our recent study on X-chromosome dynamics during mouse germ cell development in vitro, further corroborated an association between the XaXa state and the meiotic programme [54]. We observed that germ cells, which had gradually reactivated the X chromosome subsequently initiated meiosis, whereas cells that had undergone XCR too rapidly rather showed a mitotic phenotype. This suggests that XCR in itself is not sufficient to drive cells through a meiotic programme, but that the correct kinetics of XCR may play a role in order for meiosis to succeed. Thus, precocious reactivation of certain X-linked dosage-sensitive genes might be detrimental to normal germ cell development and meiotic entry, and might need to be controlled in sync with meiosis-inducing signals from gonadal somatic cells. As in pluripotent stem cells, double X-dosage-dependent repression of DNA methylation [46–48,104] may also aid the activation of meiotic genes, and thereby contribute to the right timing of meiotic entry [105,106]. Our understanding of these processes would benefit greatly from further allele-specific single-cell RNA-Seq of the reactivating germline in vivo, to better resolve the temporal control of X-linked genes during XCR. With this, we may be able to pinpoint key X-linked genes required for the correct establishment of meiotically competent germ cells.

Finally, XCR might also be needed for creating an epigenetic clean slate in preparation for the transfer of genetic material to the next generation, thereby ensuring offspring survival. Without XCR, maternal inheritance of an inactive X chromosome to male embryos, which have only a single X chromosome, and as well to female embryos, which undergo imprinted XCI of the paternal X, might potentially lead to lethality due to the absence of X-linked gene expression. After XCR has occurred, H3K27me3-based non-canonical imprinting of the Xist gene during oocyte maturation further safeguards the activity of the maternally inherited X chromosome during preimplantation development [29,30] and later in extraembryonic tissues by Xist's antisense regulator Tsix [32]. Direct functional perturbation of XCR will be needed to reveal its importance for meiosis specifically and for germ cell maturation, oogenesis and fertility in general.

Apart from the dosage compensation mechanism of XCI, which ensures X-linked gene dosage parity between XX females and XY males, another level of dosage control exists for the regulation of gene expression from the active X. First proposed by Susumo Ohno in his seminal book ‘Sex Chromosomes and Sex-linked Genes’ [1], XCU is the result of transcriptional hyperactivation of the active X chromosome in order to attain X-to-autosome dosage balance, and occurs in both males and females. According to Ohno's hypothesis, XCU developed during the evolution of distinct sex chromosomes from an autosome pair, which are the X and Y chromosomes in placental mammals. Degradation and the associated gene-loss from the Y chromosome, therefore, would lead to a potential underdosage of genes that are only expressed on one X chromosome in males, in comparison with autosomal genes expressed from two chromosome copies. While XCU of the single X in males would lead to the dosage balance with autosomal genes, it could potentially lead to an overdose in XX females, which led to the evolution of XCI to counteract XCU [2,3]. Although XCU has been extensively studied as the dosage compensation mechanism of the single male X chromosome in the fruit fly Drosophila melanogaster [107], its existence has been under debate in mammals, due to a previous lack of allele-specific analysis methods and differences in approaches when comparing X-chromosomal with autosomal gene expression levels [5,108–112]. The discrepancy between studies providing evidence in favour and against Ohno's hypothesis might have also stemmed from the fact that species-specific differences exist in the fraction of dosage-sensitive genes on the sex chromosomes, which are dosage compensated depending on their evolutionary history [113,114]. However, many recent studies investigating the epigenetic and gene expression states of individual X chromosomes with allelic resolution, by which chromatin accessibility and transcripts can be assigned to the active or inactive X, have brought major advances to our understanding of X-chromosome regulation, confirming that XCU in mammals indeed exists [66,94,115–119]. Interestingly, the new data suggest that XCU has an elastic relationship with XCI and XCR, whereby the level of upregulation of the active X chromosome during XCU is tuned to the overall X-dosage, acting as a counterbalance to XCI, and therefore in direct relationship with the status of the inactivating X chromosome (Figures 1 and 2). During early mouse development, ESC-differentiation and iPSC-reprogramming, when dynamic regulation of X-chromosome activity by XCI and XCR occurs, XCU in response to XCI ensures maintenance of the overall balance of X-linked and autosomal gene expression [66,94,117–119].

Mechanistically, XCU regulation has been demonstrated to operate through increased transcriptional burst frequency, i.e. the frequency at which a gene switches between transcriptionally active and inactive states, thus resulting in higher X-linked gene expression [66,94,117,119]. This is associated with increased chromatin accessibility on the active X chromosome [66], although this has not been detected on the single-cell level [94]. Furthermore, enhanced transcription on the active X is correlated with an enrichment of active chromatin marks and phosphorylated active RNA-Pol II [115,116], although it is not clear if this is cause or consequence of the hyperactive state during XCU. Moreover, it has been proposed that the half-life of X-chromosomal transcripts may be enhanced, as they would be less prone to nonsense-mediated decay [116,120], suggesting that also potential posttranscriptional mechanisms may be involved in XCU. In Drosophila, the male X chromosome becomes epigenetically upregulated [107] by the histone acetyltransferase MOF, which is part of the male-specific lethal (MSL) complex. MSL gets recruited to the single male X chromosome through two long noncoding RNAs; roX1 and roX2 [121] and acts to deposit H4K16ac marks, resulting in the upregulation of X-linked genes [122–124]. Also in mouse cells, enrichment of MSL and MOF has been observed at specific X-chromosomal regions and has been put forward to play a role in regulating X-linked genes [116,125]. Interestingly, MSL2 has been shown to bind and regulate Tsix expression in order to repress Xist in ESCs, demonstrating an involvement in X chromosome regulation [125]. With regards to XCU, MOF and its associated H4K16ac mark have been shown to be enriched at a subset of X-linked gene promoters, and MOF depletion led to their down-regulation [116]. A continued dissection of the mammalian homologues of the Drosophila dosage compensation system would be invaluable, and may be the key to achieving a greater understanding of the mechanism of mammalian XCU. Additional insight into the XCU mechanism comes from a recent study, which identified the chromatin factor BRD4 as a critical regulator of XCU during mouse preimplantation development [126]. Embryos treated with BRD4 inhibitor or injected with siRNAs to knock down BRD4 showed defective XCU and reduced developmental potential after being transferred as blastocysts into recipient mothers [126]. Due to its dual function both as a reader and writer of histone lysine acetylation marks [127], BRD4 could thereby potentially provide a downstream mediator of histone acetylation by MOF homologues in mammals. Overall, however, as it is the case for XCI and XCR, the mechanisms of XCU remain far from being understood.

XCU has been shown to result in the transcriptional hyperactivation of the single active X chromosome of around 1.6 fold of average expression compared with one copy of autosomal genes (resulting in an approximate 0.8 X:autosome ratio) in mouse embryos and somatic cells [5,66,94,108,116,117,119,128]. This incomplete dosage compensation might be due to the fact that not all X-linked genes are dosage-sensitive and therefore compensated by XCU, and might explain why some previous studies did not detect XCU in mammalian cells when looking at X-chromosomal gene expression globally [109,112,119,126,129]. It also explains why, after XCR in mouse pluripotent stem cells, in which XCU is absent from females but present in XO females and XY males, X-chromosome dosage-dependent differences regarding DNA methylation and pluripotency exit kinetics become apparent between XX and XY or XO pluripotent stem cells. XX cells display an X to autosome ratio of around 1 while XY or XO cells reach around 0.7–0.8 [44,46–48,66,108]. Finally, the incomplete X-dosage compensation level achieved by XCU compared with full double X-dosage might also provide an explanation as to why there are deleterious effects on female development in Xist-knockout embryos in the absence of XCI and XCU [12,94]. However, it still remains to be tested if XCU, while it is absent from female Xist-knockout preimplantation embryos [94], is also missing during postimplantation stages when the Xist-KO lethality phenotype emerges. In the case that XCU might occur, an absence of XCI in the presence of XCU could in turn generate an even higher overdose of X-linked genes. In particular, it would be interesting to examine if XCU is regulated differently between the placenta and in embryonic tissues. This might explain why a failure in imprinted XCI is tolerated in preimplantation embryos [94] but is lethal in the placenta [12], and why a lack of rXCI in postimplantation embryonic tissues can be partially compensated for [50].

In the mouse germ cell lineage in vivo, the current lack of allele-specific datasets only allows inference of the contributions of XCI, XCR and XCU to the overall X-chromosomal gene dosage from the dynamic changes of X:autosomal ratio in XX, XO and XY data. The X:A ratios in early XX, XO and XY mouse germ cells start out at similar levels around or below 1 due to the combined action of XCI and XCU in XX females and XCU in XO females and XY males [87,103]. Interestingly, XCU in the germline appears to be down-regulated only after the occurrence of XCR once germ cells have entered the gonad, resulting in an X-to-autosome ratio that exceeds 1 in XX females due to the combined expression from both X chromosomes in addition to XCU (Figure 1) [87]. Gonadal germ cells are believed to remain in this state for approximately 3 days, coinciding with meiotic entry and not returning to an X-to-autosome ratio of 1 until the pachynema stage of prophase I. Allelic RNA-Seq analysis would be required to disentangle the exact kinetics of XCR in relation to XCU, and to measure their contribution to the overall X-dosage. Whether this elevated dosage of the X chromosome in the XX germline is functionally important for female meiosis to occur remains to be tested. Further investigations will deepen our understanding of the relationship between the X-chromosome status and the switch from mitosis to meiosis in developing germ cells.

While a lot has been learned about X-chromosome dosage mechanisms from the mouse model system, it is becoming increasingly apparent that not all findings from the mouse are applicable to primates and humans in particular [130,131]. For example, the dynamics of XCI during human embryonic development differs from mice (Figure 2) due to the absence of imprinted XCI and a later onset of gene silencing in humans [35,37,132].

As outlined above, in the mouse, the noncoding RNA Tsix acts as an antisense repressor of Xist on the active X chromosome, thereby preventing Xist expression and XCI of the active X both during rXCI in the embryo and imprinted XCI in the placenta [32,33,76]. In primates, however, TSIX does not seem to play such a role, as its transcript terminates prior to overlapping with the XIST promoter [133,134], explaining why in human placentas random, but not imprinted XCI has been found [135,136]. In preimplantation mouse embryos, imprinted XCI is not controlled by Tsix but rather by a H3K27me3 imprinting mark of the Xist locus [30,137]. Also, this imprint of the XIST locus seems to be absent from human preimplantation embryos, as XIST RNA becomes expressed after embryonic genome activation from both female X chromosomes and the single male X, although this does not lead to immediate gene silencing [35,37,132]. XACT, a hominoid-specific noncoding RNA, which coats the active X chromosome in human pluripotent stem cells [138], is initially co-expressed with XIST in early embryos from both X chromosomes and has been proposed to potentially interfere with XIST spreading or in recruiting its silencing effector proteins [37,132,139]. However, the function of XACT RNA during early development still needs to be tested, as a recent study did not detect effects on X-linked gene expression after the deletion of XACT in human pluripotent stem cells [140]. Single-cell RNA-Seq analysis of female blastocysts suggested a biallelic reduction in X-linked gene expression, termed dampening [37]. Dampening conceptually resembles the dosage compensation mechanism in the nematode Caenorhabditis elegans, where gene expression is down-regulated from both X chromosomes by half in XX hermaphrodites compared with XO males [141]. However, when different analysis parameters were used, a switch from biallelic to monoallelic expression was observed in human embryos, rather suggesting the onset of rXCI instead of dampening [132,142,143]. In an implantation model of in vitro cultured human embryos, XCU in males and rXCI and XCU in females occurred progressively and asynchronously in different cell lineages of epiblast, trophectoderm and primitive endoderm [38]. A recent study used Cynomolgus monkeys (macaques) as a non-human primate model system for X-chromosome dynamics that more closely resemble humans than mice [144]. Similar to humans, in monkey preimplantation embryos XIST becomes expressed from both X chromosomes in females and the single X in males, and does not lead to the silencing of X-linked genes. This is despite the recruitment of polycomb repressive marks H3K27me3 and H2AK119ub, which has not been observed in human embryos at that stage [35]. After implantation, XIST expression becomes monoallelic in female monkey embryos and eventually leads to gene silencing [144]. Another recent paper comparing single-cell RNA-Seq data from mouse, monkey and human foetal oocyte development concluded that overall X-dosage regulation is similar across these species [145]. It remains an open question, why XIST expression in early human and monkey embryos does not lead to immediate gene silencing and how the switch from bi- to monoallelic X-linked expression is mediated.

The germ cell lineage features commonalities and differences between the X-chromosome dynamics in mice, monkeys and humans. In mice, early PGCs show XCI [15–17] but to a lesser degree when compared with somatic cells or early embryonic cells without PGC competence [43,54], and subsequently go through XCR during migration and colonisation of the gonads. Also in monkeys, early PGCs show partial XCI after specification, with some genes being more efficiently silenced than others, and most female PGCs initially showing monoallelic XIST expression [144]. Afterwards, during PGC migration, XIST gets down-regulated, H3K27me3 erased and X-linked genes progressively reactivated. Once monkey germ cells have entered the gonads, XCR is complete and X-linked genes biallelically expressed, however, XIST becomes reexpressed from both X chromosomes in females and the single X in males, without fully coating the X chromosomes and without leading to gene silencing, similar to the state during preimplantation development.

In the human germline, the X-chromosome dynamics have been studied so far during the migratory (week 4–5 of development) and gonadal germ cell stages (week 5 onwards). Allele-specific single-cell RNA-Seq analysis of X-linked genes initially suggested that biallelic expression and thereby XCR have already occurred at the first time point observed (4 weeks) [146]. However, a later study with further in-depth single-cell analysis concluded that XCR in human germ cells was rather a gradual process, which appears to occur in a heterogeneous manner [147] and is accompanied by erasure of H3K27me3 and DNA-demethylation of X-linked gene promoters [147,148]. As currently data on the X-status of human PGCs at their specification time (week 2) are missing, it is not clear to which extent XCI has already occurred in the early human germline before XCR takes place. Interestingly, XIST RNA has been detected in both male and female germ cells and its overall expression levels do not predict the extent of X-linked gene reactivation during XCR [147,149]. A recent study reported that X-linked gene expression is reduced in female germ cells prior to their meiotic entry when XIST becomes down-regulated [150]. In RNA-FISH experiments, XIST frequently showed a biallelic dispersed expression pattern and was accompanied by biallelic coexpression with XACT RNA in premeiotic germ cells, resembling the situation in human preimplantation embryos [132,150]. While XIST and XACT are expressed, X-linked gene expression is attenuated compared with autosomal genes, which has been interpreted as an XIST-dependent dampening mechanism [150]. Nevertheless, it still remains unclear, if biallelic dampening really occurs and what the contributions of XIST and XACT RNA may be. An alternative hypothesis for reduced X-linked gene dosage in female germ cells could be incomplete XCR, as it has been previously proposed [147]. From the currently available data, many different possible interpretations for X-linked gene dosage regulation in human germ cells exist; ranging from the possibility of continuous dampening in the absence of XCI and XCR, over XCI and XCU followed by XCR, to a sequence of all XCI, XCU, XCR and dampening [103,151]. Needless to say, further investigations including allele-specific single-cell RNA-Seq and multi-omics analyses will be necessary to bring clarity to the complex control of X-linked gene dosage during human germ cell development in vivo.

From a functional perspective, in vitro germ cell derivation from human pluripotent stem cells could shed light on the requirement of X-dosage control for the different steps from PGC specification to meiotic entry and oogenesis. At the moment human in vitro germ cell development is less far advanced than the mouse model system [152] and only early meiotic stages have been achieved so far [153]. Nevertheless, in two recent studies, it has been tested whether the original X-chromosome status of the pluripotent stem cells used as the starting material would affect human PGCLC-specification efficiency [154,155]. While the first study did not observe any significant impact, but rather that cell type of origin might affect PGCLC-differentiation efficiency [154], the second reported that stem cells with an eroded XCI state showed reduced germ cell competence when compared with cells with more complete XCI [155]. In the study with the most advanced in vitro development into oogonia-like cells, progressive demethylation of X-linked gene promoters and partial XCR has been observed [153]. At this point, it is difficult to assess if the X-status is deterministic or rather diagnostic for the germ cell differentiation potential of pluripotent cell lines. In particular, the general cell line-to-cell line variability can affect germ cell development, which is influenced by many confounding parameters beyond the often ill-defined X-status, such as genetic background, cell type of origin and culture conditions. Functional experiments modulating the X-status in a well-controlled, isogenic cell background, will be needed to directly address the potential roles of X-chromosome dosage for human germ cell development.

Despite all the recent progress in deciphering the dynamics of the X chromosome in the germline, studies that allelically resolve the contribution of the different dosage compensation mechanisms and functionally interrogate their importance are still missing. Here we have highlighted some of the key findings that help us gain better understanding of this relationship, however, significant questions remain to be addressed (Box 1). With these questions addressed, we would hope to better understand the role of the X chromosome in the establishment of the germline, and whether its dynamics are a cause, or simply a consequence, of correct female germ cell development. Importantly further understanding of what is normal during germ cell development in vivo will help recapitulate these processes during germ cell development from pluripotent stem cells in vitro. Achieving faithful epigenetic reprogramming will be a critical step towards the generation of high-quality human oocytes in a dish.

• The interplay and roles of different X-chromosome dosage compensation mechanisms in the mammalian female germline remain elusive.

• Strict temporal control of X-linked gene expression dynamics suggests a link between the X-chromosome status and germ cell developmental competence, but the extent of this is unknown.

• Further work aiming to resolve the allele-specific regulation of X-chromosome dosage and the functional perturbation of the X-chromosome status during germline development would help to better resolve the details of this relationship.

The authors declare that there are no competing interests associated with the manuscript.

Work on germ cells in the B.P. laboratory has been supported by the Spanish Ministry of Science, Innovation and Universities; the Agencia Estatal de Investigacion (AEI, PID2021-123383NB-I00); the Agencia de Gestio d'Ajuts Universitaris i de Recerca (AGAUR, 2021 SGR 01222) and contributions from Clinica EUGIN. We would like to thank the Spanish Ministry of Economy, Industry and Competitiveness (MEIC) to the EMBL partnership and to the “Centro de Excelencia Severo Ochoa”. We also acknowledge support of the CERCA Programme of the Generalitat de Catalunya.

Bernhard Payer: Conceptualization, Funding acquisition, Writing — original draft, Writing — review and editing. Tom Mattimoe: Conceptualization, Visualization, Writing — original draft, Writing — review and editing.

We want to thank all past and present members of the Payer laboratory for their contributions.

ESCs

embryonic stem cells

iPSCs

induced pluripotent stem cells

PGCLCs

primordial germ cell-like cells

PGCs

primordial germ cells

Xa

active X chromosome

XACT

X-active coating transcript

XCI

X-chromosome inactivation

XCR

X-chromosome reactivation

XCU

X-chromosome upregulation

Xi

inactive X chromosome

XIST

X-inactive specific transcript