Controlled X‐chromosome dynamics defines meiotic potential of female mouse in vitro germ cells

The mammalian germline is characterized by extensive epigenetic reprogramming during its development into functional eggs and sperm. Specifically, the epigenome requires resetting before parental marks can be established and transmitted to the next generation. In the female germline, X‐chromosome inactivation and reactivation are among the most prominent epigenetic reprogramming events, yet very little is known about their kinetics and biological function. Here, we investigate X‐inactivation and reactivation dynamics using a tailor‐made in vitro system of primordial germ cell‐like cell (PGCLC) differentiation from mouse embryonic stem cells. We find that X‐inactivation in PGCLCs in vitro and in germ cell‐competent epiblast cells in vivo is moderate compared to somatic cells, and frequently characterized by escaping genes. X‐inactivation is followed by step‐wise X‐reactivation, which is mostly completed during meiotic prophase I. Furthermore, we find that PGCLCs which fail to undergo X‐inactivation or reactivate too rapidly display impaired meiotic potential. Thus, our data reveal fine‐tuned X‐chromosome remodelling as a critical feature of female germ cell development towards meiosis and oogenesis.

Epigenetic X‐chromosome remodelling facilitates germ cell development towards meiosis and oogenesis.

The germ cell lineage is unique in its critical function to transmit genetic and epigenetic information from one generation to the next. In mice, primordial germ cells (PGCs), the precursors of eggs and sperm, are specified during early postimplantation development from somatic precursors in the proximal epiblast by inductive signals (Lawson et al , 1999 ; Ohinata et al , 2005 , 2009 ). Thereafter, PGCs migrate and enter the future gonads where they receive sex‐specific somatic signals, which determine the germ cell sex and promote differentiation towards a spermatogenic or oogenic fate (Miyauchi et al , 2017 ; Spiller et al , 2017 ). While in males, germ cells enter mitotic arrest and differentiate into prospermatogonia, in females, germ cells instead progress into meiosis and oogenesis. A hallmark feature of early germ cell development is the extensive epigenetic reprogramming (Kurimoto & Saitou, 2019 ), characterized by global changes in histone marks (Seki et al , 2005 ; Hajkova et al , 2008 ), DNA demethylation and erasure of genomic imprints (Hajkova et al , 2002 ; Seisenberger et al , 2012 ; Shirane et al , 2016 ). This establishes an epigenetic naive state (Ohta et al , 2017 ), which is required in order for PGCs to progress towards gonadal germ cell fate (Hill et al , 2018 ) and to control their timing to enter female meiosis (Yokobayashi et al , 2013 ). Ultimately, this erasure of parental information allows the reestablishment of new paternal and maternal marks during spermatogenesis and oogenesis, respectively, which are critical for the competence of egg and sperm to facilitate embryonic development in the next generation (Reik & Surani, 2015 ; Ohta et al , 2017 ). In addition to these global changes, another important epigenetic reprogramming event takes place in the female germline; the reversal of silencing of the inactive X chromosome by X‐chromosome reactivation. While X‐chromosome inactivation (Lyon, 1961 ; Payer & Lee, 2008 ; Galupa & Heard, 2018 ) is the process by which female mammals (XX) achieve X‐linked gene dosage parity with males (XY), X‐reactivation takes place specifically in pluripotent epiblast cells of the mouse blastocyst (Mak et al , 2004 ; Borensztein et al , 2017 ) and in PGCs during their migration and upon their entry into the gonads (Sugimoto & Abe, 2007 ; Chuva de Sousa Lopes et al , 2008 ). Therefore, while X‐inactivation is associated with pluripotency exit and the differentiated state (Schulz et al , 2014 ), X‐reactivation is a key feature of naive pluripotency and germ cell development (Pasque et al , 2014 ; Payer, 2016 ; Janiszewski et al , 2019 ; Panda et al , 2020 ; Bauer et al , 2021 ; Talon et al , 2021 ). X‐reactivation in mouse PGCs is a multistep process, which initiates during PGC migration with downregulation of Xist, the long non‐coding master regulator RNA of X‐inactivation and concomitant loss of the associated histone H3K27me3 mark from the inactive X (Sugimoto & Abe, 2007 ; Chuva de Sousa Lopes et al , 2008 ). This process is regulated by repression of the Xist gene by the germ cell transcription factor PRDM14 (Payer et al , 2013 ; Mallol et al , 2019 ) and potentially by other members of the pluripotency network such as NANOG or OCT4 (Navarro et al , 2008 ), which are all expressed during PGC development. Subsequently, X‐linked genes become progressively reactivated during migration, with the process being completed after PGCs have reached the gonads, and following the initiation of oogenesis and meiosis (Sugimoto & Abe, 2007 ; Sangrithi et al , 2017 ). X‐linked gene reactivation is thereby thought to be enhanced by a female‐specific signal from gonadal somatic cells (Chuva de Sousa Lopes et al , 2008 ). Although the molecular nature of the X‐reactivation‐promoting signal is currently unknown, the timing of X‐linked gene reactivation around meiotic entry and the dependency of both processes on a female somatic signal suggest a potential mechanistic link. Until now it has not been formally tested, if, and to which degree, the X‐inactivation status might impact the meiotic and oogenic potential of germ cells. Furthermore, previous studies on the X‐inactivation and ‐reactivation dynamics during mouse germ cell development have been limited to few individual genes (Sugimoto & Abe, 2007 ) or have not been allelically resolved and therefore been unable to discriminate between transcripts expressed from either one or two X chromosomes (Sangrithi et al , 2017 ). Therefore, a comprehensive analysis of X‐inactivation and ‐reactivation kinetics and its functional relation to germ cell developmental progression is necessary to gain mechanistic insight. Based on in vitro germ cell differentiation from mouse embryonic stem cells (ESCs) (Hayashi et al , 2011 , 2012 ; Nakaki et al , 2013 ), we developed an X‐chromosome reporter system (XRep) to study the kinetics of X‐inactivation and ‐reactivation during germ cell development. We thereby provide a high‐resolution allelic analysis of X‐chromosome dynamics and discovered that germ cells with high meiotic and oogenic competence are characterized by a moderate degree of X‐inactivation and gradual X‐reactivation kinetics. In contrast, germ cells that failed to undergo X‐inactivation or which reactivated the X chromosome too rapidly displayed abnormal gene expression and differentiation characteristics. Thus, we found first evidence that a controlled sequence of X‐inactivation followed by X‐reactivation to be a characteristic hallmark of normal female germ cells. This suggests that both dosage control and epigenetic reprogramming of the X chromosome may be critical indicators for female germ cells' developmental potential to progress towards meiosis and oogenesis.


XRep, a tailor‐made system for tracing X‐chromosome dynamics during in vitro germ cell development
In order to achieve a better understanding of the X‐chromosome dynamics during mouse germ cell development, we created a tailor‐made in vitro model system called XRep (Fig 1A ). XRep combines the following features. First, it is based on a hybrid female embryonic stem cell (ESC) line containing one Mus musculus (X mus ) and one Mus castaneus (X cas ) X chromosome (Lee & Lu, 1999 ; Ogawa et al , 2008 ), allowing allele‐specific determination of gene expression. Moreover, this line was shown to be karyotypically highly stable (Lee & Lu, 1999 ; Bauer et al , 2021 ), therefore preventing X‐loss, a crucial prerequisite for X‐inactivation and ‐reactivation studies. Additionally, the cell line contains a Tsix truncation (TST) on X mus , forcing non‐random X‐inactivation of the X mus upon cell differentiation (Luikenhuis et al , 2001 ; Ogawa et al , 2008 ). This enabled us to study the X‐inactivation and ‐reactivation dynamics specifically of the X mus chromosome, while the X cas would remain constitutively active. Second, primordial germ cell‐like cells (PGCLCs) can be obtained highly efficiently from XRep cells by doxycycline‐inducible overexpression of the germ cell fate specifier transcription factors BLIMP1 (also known as PRDM1), PRDM14 and TFAP2C (also known as AP2γ) (Nakaki et al , 2013 ), therefore bypassing the need for addition of cytokines. Last, the X‐chromosome status of XRep cells can be traced by dual X‐linked reporter genes placed in the Hprt locus (Wu et al , 2014 ), a GFP reporter on X mus (XGFP) and a tdTomato reporter on X cas (XTomato). This allows us to isolate distinct populations of cells, harbouring either two active X chromosomes (XGFP+/XTomato+) or one inactive and one active X (XGFP−/XTomato+), using fluorescence‐activated cell sorting (FACS). This gives us a unique advantage over in vivo studies, as it enables us to test the importance of X‐inactivation and ‐reactivation for germ cell development by isolating and further culturing cells of different X‐inactivation states. Taken together, this tailor‐made system allows us to assess X‐chromosome dynamics and its importance for female mouse germ cell development in vitro . We first set out to assess competence for PGCLC differentiation of our XRep cell line. We slightly adapted published protocols (Hayashi & Saitou, 2013 ; Nakaki et al , 2013 ), by differentiating ESCs into epiblast‐like cells (EpiLCs) for 4 days, as differentiation for 2 days, as demonstrated in said previous studies, did not yield PGCLCs with our XRep cells likely due to their specific genetic background (Fig EV1A ). Furthermore, we extended the induction time of PGCLC generation from 4 to 5 days to ensure sufficient time to undergo X‐inactivation (Fig 1B ). We quantified PGCLC induction efficiency by FACS analysis, using SSEA1 and CD61 double‐positive staining to mark successfully induced PGCLCs (Fig 1C ). At PGCLC day 5, we found ~ 60% of the cell population to be double‐positive for SSEA1/CD61, indicating a very high PGCLC induction efficiency when compared to the cytokine‐based protocol (Hayashi & Saitou, 2013 ) and in line with previous observations on transcription factor‐based PGCLC induction (Nakaki et al , 2013 ). To further assess the quality of our PGCLCs, we stained cryosections of PGCLC bodies at day 5 of induction for SOX2 and TFAP2C, both germ‐line expressed transcription factors. We observed that > 50% of cells were double‐positive for SOX2 and TFAP2C (Fig 1D ), confirming PGCLC cell identity. We next wanted to assess X‐inactivation kinetics using our XGFP and XTomato reporters. As expected, XTomato stayed active throughout the differentiation (Fig 1E and F ). In contrast, we observed downregulation of the XGFP reporter at day 2 of PGCLC differentiation, with the XGFP− population gradually increasing until day 5 (Fig 1E and F ). Nevertheless, even at day 5, up to 40% of PGCLCs remained XGFP+ in our system (Fig EV1B ). Despite this, the large majority of EpiLCs showed H3K27me3 foci (Fig EV1D–F ), which suggests that both XGFP− and XGFP+ PGCLCs originated from EpiLCs that had initiated X‐inactivation. We do, however, note that EpiLCs still retained XGFP protein staining (Figs 1E and F, and EV1D and F ), while XGFP transcripts were being downregulated (Fig EV1C ), indicating that protein stability of GFP gives a delayed read‐out of X‐inactivation kinetics. Nevertheless, at the PGCLC stage, the XGFP signal faithfully reflected the X‐inactivation state, as XGFP was only detected in cells without the X‐inactivation‐specific H3K27me3 spot (Fig EV1D–F ). In summary, using our tailor‐made XRep cell line, we could show that X‐inactivation initiates early during PGCLC differentiation. Additionally, our system enables the isolation of distinct PGCLC populations, either having undergone X‐inactivation or harbouring two active X chromosomes, suggesting that PGCLC specification can occur in the absence of X‐inactivation as well.

XGFP+ and XGFP− PGCLCs define distinct subpopulations
Having identified two distinct PGCLC populations, we set out to characterize the transcriptional changes taking place during differentiation. We induced EpiLCs from ESCs for 4 days and subsequently induced PGCLCs for 5 days, at which stage we isolated XGFP+ and XGFP− PGCLCs by FACS (Fig EV2A ). With these samples, we performed allele‐specific RNA‐sequencing on two biological replicates (different clones) with two technical replicates each. Principal component analysis (PCA) of the expression profiles showed a high coherence between replicates, with ESCs, EpiLCs and PGCLCs occupying distinct clusters (Fig 2A ). Moreover, we observed that XGFP+ and XGFP− PGCLCs clustered separately, indicating distinct expression profiles of the two populations. To exclude the possibility that the distinct clustering of PGCLC populations was influenced by the different X‐status of the two, we repeated the PCA while eliminating X‐chromosome‐linked genes from the analysis. We observed a highly similar clustering of samples with minimal changes in component variances (Fig EV2B ). In order to assess whether transcriptional differences in XGFP+ and XGFP− PGCLCs could be explained by differences in developmental timing, we took advantage of published datasets of female in vivo PGCs from E9.5, E10.5, E11.5 and E12.5 embryos (Nagaoka et al , 2020 ) and compared expression profiles to our in vitro derived PGCLCs (Fig 2B ). PCA revealed a trajectory where PC1 defined the developmental timing of in vivo and in vitro samples, whereas PC2 did not greatly contribute to the separation of our in vitro PGCLCs (PC loadings in Dataset EV4 ). We found that both PGCLC populations clustered around E10.5, with XGFP+ cells corresponding to a slightly advanced developmental stage. Therefore, as XGFP+ and XGFP− PGCLCs seemed to correspond to a similar developmental time point, we wanted to characterize their transcriptional differences in more detail. We performed differential gene expression analysis and could identify 2,684 upregulated and 2,437 downregulated genes in XGFP− PGCLCs, when compared to XGFP+ PGCLCs (Figs 2C–E and EV2C and D ). Among the genes significantly upregulated in XGFP− PGCLCs, we found early germ cell genes including Blimp1 (Prdm1) , Prdm14 and Tfap2c (Figs 2D and EV2C ). In contrast, in XGFP+ PGCLCs, we observed higher expression of pluripotency genes such as Esrrb and Zfp42 and a subset of late germ cell genes like Dazl . Moreover, when we performed functional annotation by gene ontology (GO) term analysis, we observed enrichment for genes involved in urogenital system development, MAPK regulation and WNT signalling in XGFP− PGCLCs, while genes upregulated in XGFP+ PGCLCs were enriched for DNA methylation involved in gamete generation, meiotic cell cycle and response to LIF signalling (Figs 2F and EV2D ). MAPK signalling is known to be inhibited by double X‐dosage (Schulz et al , 2014 ; Song et al , 2019 ; Genolet et al , 2021 ), which might explain enrichment of this pathway in our XGFP− PGCLCs. LIF signalling on the other hand, which is enriched in our XGFP+ PGCLCs, is known to enable expression of the naive pluripotency network, which represses Xist, thereby promoting the active X state (Payer & Lee, 2014 ; Panda et al , 2020 ). Furthermore, enrichment for meiotic cell cycle genes in XGFP+ PGCLCs such as Aurkc , Dazl and Piwil2 (Fig EV2D ), suggests a premature activation of a subset of meiotic genes in XGFP+ PGCLCs. One characteristic feature of PGCLCs is changes in cell cycle progression and proliferation upon differentiation (Ohta et al , 2017 ), both of which are known to be affected by MAPK, as well as LIF signalling pathways (Meloche & Pouysségur, 2007 ; Onishi & Zandstra, 2015 ). We therefore performed cell cycle analysis using DAPI and found that ESCs, EpiLCs and XGFP+ PGCLCs shared highly similar profiles, with the majority of cells (> 60%) residing in S phase. In contrast, XGFP− PGCLCs showed a decreased number of cells in S phase, concomitant with an increase in cells in G1, suggesting a slower proliferation of this population (Fig 2G ). As our transcriptomics and cell cycle analysis suggested that XGFP+ PGCLCs could correspond to an aberrant PGCLC state with properties related to ESCs, we set out to address if this would also lead to an advantage in growth and survival under physiological conditions favouring ground‐state pluripotent stem cells. We therefore isolated XGFP+ and XGFP− PGCLCs at day 5 and seeded them (1,000 cells per six‐well) on irradiated mouse embryonic fibroblasts in 2i/LIF medium (Fig 2H and I ), which previously has been reported to allow the establishment of pluripotent embryonic germ cell (EGC) lines from in vivo mouse PGCs (Leitch et al , 2010 ). When we then compared EGC colony formation capacity, we found that while almost no colonies (2 from 1,000 seeded cells) originated from XGFP− PGCLCs, we observed a substantially higher number of colonies ( n = 84) from XGFP+ PGCLCs, albeit still fewer than when re‐plating ESCs (633 colonies). Importantly, both ESCs and XGFP+ PGCLCs retained two active X chromosomes, while only a subset of XGFP− PGCLCs had undergone XGFP‐reactivation during EGC colony formation (Fig EV2E and F ). In summary, RNA expression analysis of XGFP+ and XGFP− PGCLCs showed a PGC‐like transcriptome of both populations, further suggesting that X‐inactivation and PGCLC induction can be uncoupled in our system. However, we observed that XGFP+ PGCLCs displayed higher expression of several naive pluripotency genes as well as premature expression of a subset of meiotic genes and a rapid cell cycle. Moreover, considering their higher ability to form EGC colonies under ground‐state pluripotency conditions, this suggests that XGFP+ PGCLCs may correspond to an aberrant PGCLC state with pluripotent stem cell‐related features. This indicates that X‐inactivation could be necessary for correct PGCLC maturation, or alternatively, be a sensitive biomarker of high‐quality PGCLCs.

Heterogeneous and moderate X‐inactivation is a feature of germ cell fate in vitro and in vivo
To this point, due to the lack of an allele‐specific transcriptomic analysis, the X‐inactivation and ‐reactivation dynamics during mouse PGC development in vivo and in vitro have not been assessed on a chromosome‐wide level. Therefore, to determine X‐chromosome‐wide gene inactivation kinetics during PGCLC differentiation, we assessed the allelic expression ratio between the inactive X mus and the active X cas . We performed PCA of the allelic ratio of our samples in addition to neural progenitor cells (NPCs) from the same parental clone (Bauer et al , 2021 ) to include a cell type shown to have undergone complete X‐inactivation (Fig 3A ). We observed that the PC1 of the PCA defined the degree of X‐inactivation, separating samples with two active Xs on the left (ESCs and XGFP+ PGCLCs), and with one inactive X on the right (XGFP− PGCLCs and NPCs). Moreover, we noticed that EpiLCs were positioned at the centre, suggesting an intermediate degree of X‐inactivation. We next determined X‐inactivation kinetics, while focussing on genes biallelically expressed in ESCs (Fig EV3A ) (allelic expression ratio > 0.3 and < 0.7) and established an X‐inactivation cut‐off of an allelic ratio of 0.135, according to the distribution of genes in NPCs and the local minimum (Fig EV3B ), and similar to cut‐offs used in previous studies (Peeters et al , 2014 ; Borensztein et al , 2017 ; Xu et al , 2017 ; Janiszewski et al , 2019 ; Bauer et al , 2021 ). As a control, we assessed the allelic expression ratio of the fully hybrid chromosome 13, which maintained biallelic expression throughout the time course (Fig EV3C ). In contrast, we observed initiation of X‐linked gene silencing in EpiLCs, progressing further in XGFP− PGCLCs, while XGFP+ PGCLCs showed biallelic expression, similar to ESCs (Fig 3B ). To assess X‐inactivation dynamics in more detail, we grouped X‐linked genes according to their silencing kinetics (Fig 3C ). We found 62 genes to have undergone X‐inactivation (XCI) in EpiLCs (termed early XCI), and 138 genes to have undergone inactivation in PGCLCs (late XCI). Moreover, to our surprise, we observed a large number of genes (93) to still be active in XGFP− PGCLCs (escapees). In comparison, we observed 46 genes escaping X‐inactivation in NPCs (Bauer et al , 2021 ), out of which 37 were also found to be escapees in PGCLCs (Fig 3D ). While a certain degree of escape from X‐inactivation is expected, the percentage of escapees we observed for PGCLCs here is at 32%, which is considerably higher than reported for other cell types (Fig EV3F ) (Peeters et al , 2014 ; Marks et al , 2015 ; Balaton et al , 2021 ). Given these results, we wondered how this large degree of escape from X‐inactivation might be explained. We assessed Xist expression levels and could observe high levels in EpiLCs, reaching levels comparable to those in NPCs (Fig 3E ). However, expression levels in XGFP− PGCLCs were considerably decreased, which might be explained by the high expression of Prdm14 in PGCLCs, a known repressor of Xist (Payer et al , 2013 ). This is also in line with in vivo data (Sugimoto & Abe, 2007 ), where Xist has been shown to be completely downregulated in E10.5 PGCs of equivalent stage (Fig 2B ). Furthermore, we wanted to know which features might distinguish escapees from inactivated genes in our system. We measured gene expression levels from the X mus allele in ESCs and found escapees to be significantly higher expressed, while early inactivating genes, in contrast, showed the lowest expression levels (Fig EV3D ). Similarly, expression of escapees from the X cas allele was also elevated in XGFP− PGCLCs (Fig EV3E ). When we compared our gene categories of early, late and escaping X‐inactivation from XGFP− PGCLCs with published data of X‐inactivation kinetics during embryoid body (EB) differentiation (Marks et al , 2015 ), the overlap of genes was relatively modest (Fig EV3F ). This suggests that the timing of inactivation of individual X‐linked genes varies between different cell types and differentiation systems. Having observed that both their precursors, EpiLCs, as well as PGCLCs themselves show a distinct moderate X‐inactivation with a high number of escapees, we wondered if our in vitro findings were reflecting the situation in vivo during mouse development. We therefore re‐analysed publicly available allelically resolved single‐cell RNA‐seq data of E6.5 embryos (Cheng et al , 2019 ) which corresponds to the developmental time point that includes the in vivo equivalents of EpiLCs (Fig 3F and G ; Han et al , 2010 ; Hayashi et al , 2011 ). We focussed on female cells from a CAST(F) X C57(M) cross, which after filtering for sufficient allelic information (see methods) left 239 cells to be included in our analysis. We then performed uniform manifold approximation and projection for dimension reduction (UMAP) on genome‐wide single‐cell expression data, and then applied shared nearest‐neighbour (SNN) modularity optimization‐based clustering (Fig 3G ). This yielded six clusters, which according to marker gene expression (Fig EV3G ) corresponded to the extraembryonic cell types of anterior visceral endoderm ( Amn +, Cer1 +, Gata4 +, Otx2 +; cluster 1), posterior visceral endoderm ( Amn + , Gata4 + , Wnt3 +; cluster 2) and parietal endoderm ( Fst +; cluster 3) and the embryonic cell types of proximal epiblast ( Dppa4 +, Ifitm3 +/ Fragilis +, Pou5f1 +; cluster 4), transition epiblast ( Otx2 +, Pou5f1 +; cluster 5) and posterior epiblast ( Gata4 +, Hand1 +, Pou5f1 +, T +, Wnt3 +; cluster 6). To assess the degree of X‐inactivation, we calculated the average allelic ratio per cell (Fig 3H ), which, as expected, highlighted imprinted X‐inactivation in the extraembryonic cell types; where the paternal C57 allele was always silenced, and showed random X‐inactivation in the epiblast clusters. However, within the epiblast we noticed striking differences, with a considerably larger fraction of cells from the proximal epiblast—the Ifitm3/Fragilis‐positive region with competence for PGC specification (Lawson & Hage, 1994 ; Saitou et al , 2002 )—not having undergone X‐inactivation (Fig 3I ), when compared with the other non‐PGC‐competent epiblast clusters. Specifically, 25% of proximal epiblast cells displayed an average allelic ratio between 0.4 and 0.6 indicating a lack of X‐inactivation compared to < 10% for transition and posterior epiblast cells. Moreover, when we then compared the degree of gene silencing for cells with an average allelic ratio < 0.4 (Fig 3J ), to allow comparison with imprinted XCI cells, we found that 56% of proximal epiblast genes escaped / displayed incomplete X‐inactivation at that stage. This strongly contrasts to < 8% for all other cell types, including extraembryonic and other epiblast clusters (Fig 3K ). This could partially be explained by the higher pluripotency factor expression in the proximal epiblast when compared with the posterior epiblast (Fig EV3H ), which may contribute to the reduced degree of X‐inactivation specifically in PGC‐competent precursor cells. Taken together, we find that EpiLCs and PGCLCs undergo a moderate degree of X‐inactivation, characterized by a large percentage of escapees / incompletely silenced genes. Moreover, our analysis suggests that low expression of Xist in PGCLCs might lead to a failure of gene silencing of highly expressed genes, leading to a large percentage of escapees. Furthermore, our analysis of in vivo E6.5 data shows heterogeneity and a moderate degree of X‐inactivation specifically in the PGC‐competent cells of the proximal epiblast, highlighting how this distinct X‐inactivation state is a conserved hallmark of germline competence both in vitro and in vivo .

Single‐cell RNA‐seq analysis of meiotic entry of in vitro ‐derived germ cells reveals clusters of distinct developmental progression
After having established the degree of X‐inactivation during PGCLC specification, we wanted to address the further developmental progression of PGCLCs depending on their X‐chromosome status. Having identified and isolated distinct PGCLC types with either two active X‐chromosomes (XGFP+ PGCLCs) or one active and one inactivated X‐chromosome (XGFP− PGCLCs) (Fig 2 ), we were able to assess whether the X‐inactivation status of PGCLCs had an impact on germ cell maturation. Furthermore, we sought to investigate to which degree X‐reactivation and meiotic entry were intrinsically coupled processes. To this end, we differentiated XGFP+ and XGFP− PGCLCs using an adapted in vitro reconstituted ovary (rOvary) protocol (Hayashi & Saitou, 2013 ) and performed single‐cell RNA‐sequencing (scRNA‐seq) using the SMART‐Seq v5 Ultra Low Input RNA (SMARTer) kit for Sequencing (Takara Bio) (Karimi et al , 2021 ). Briefly, we aggregated in vitro derived PGCLCs, originating from either XGFP+ or XGFP− populations, for 6 days with somatic cells isolated from E13.5 female embryonic gonads plus mesonephros in order to mimic the female urogenital environment and provide in vitro ‐derived germ cells with the appropriate signalling niche (Chuva de Sousa Lopes et al , 2008 ; Hayashi et al , 2012 ) to facilitate their meiotic entry and X‐reactivation (Fig 4A and B ). We then sorted single cells of the following populations on which we performed scRNA‐seq. Derived from XGFP− PGCLC rOvaries, we collected three populations: XGFPhigh reactivated (144 cells, XTomato+/XGFP+), XGFPintermediate (XTomato+/XGFPint., 144 cells) and XGFPlow (XTomato+/XGFPlow, 136 cells). From the constitutively active XGFP+ PGCLC rOvaries, we collected one population: XGFP+ constitutive (XTomato+/XGFP+, 188 cells) (Fig 4A and B ). In total, we obtained 391 million reads, with an average of 740,000 reads per cell. Next, to ensure that our analysis focussed on germ cells of appropriate quality, we only included cells with the germ cell marker Dazl expression > 1 (log2 counts per 10,000) and with sufficient allelic information (see methods). This left us with 379 cells in total and 15,583 informative genes. To characterize cellular heterogeneity, we performed dimensionality reduction using UMAP on genome‐wide single‐cell expression data, and projected the four FACS populations (Fig 4B and C ) on our UMAP plot. We observed a clear separation into two major groups, which coincided well with the levels of XGFP fluorescence. One group predominantly included the XGFPlow and XGFPintermediate germ cells (originating both from the XGFP− PGCLCs) on the left, and another group was constituted from the XGFPhigh reactivated and XGFP+ constitutive germ cells on the right. We then set out to answer whether our XGFP+ and XGFP− PGCLCs, which were the starting material for our rOvaries (Fig 4A and B ), showed a differential developmental profile, and in particular, if the meiotic germ cells originated preferentially from XGFP+ or XGFP− PGCLCs. We applied SNN modularity optimization‐based clustering which returned five clusters (Fig 4D ) that showed distinct patterns according to the expression of mitotic and meiotic germ cell marker genes (Fig 4E and F ). We identified two mitotic clusters termed “Mitotic 1” and “Mitotic 2,” showing expression of the PGC marker Stella (also known as Dppa3 ) as well as mitotic PGC markers Morc1 and Nanog . Moreover, we identified two pre‐meiotic clusters termed “Pre‐meiotic 1” and “Pre‐meiotic 2,” defined by the initial expression of both Stella and Ddx4 , and lastly one meiotic cluster termed “Meiotic'' characterized by expression of the meiotic genes Prdm9 and Sycp3 . Next, we wanted to assess whether a directionality within the clusters and eventually among the two groups could be observed. Pseudo‐time analysis using RNA velocity (La Manno et al , 2018 ) placed the meiotic cluster at the apex of a path which revealed a differentiation trajectory directed towards meiosis, initiating from the pre‐meiotic clusters (Fig 4F ). Moreover, comparison to in vivo data (Zhao et al , 2020 ) showed that our mitotic clusters corresponded to E12.5 germ cells, whereas pre‐meiotic and meiotic clusters corresponded to later time points; E14.5 and E16.5 (Fig EV4A–C ), confirming that our in vitro clusters followed an in vivo ‐like developmental trajectory. Intriguingly, both pre‐meiotic and meiotic germ cells almost exclusively originated from XGFP− PGCLCs, whereas mitotic germ cells consisted of XGFPhigh reactivated and XGFP+ constitutive germ cells. Considering this, we wanted to assess whether our XGFP− PGCLCs could therefore mature further and initiate oogenesis. We took advantage of a published in vitro differentiation protocol and aggregated XGFP− PGCLCs with embryonic‐derived somatic gonadal cells, forming an rOvary, followed by the culture of the rOvary onto a transwell to allow in vitro differentiation (IVDi) of PGCLCs (Fig 4G ) (Hayashi et al , 2017 ). However, to perform the experiment in a more physiological niche, without external cues, no retinoic acid was added to the IVDi culture and the IVDi tissue was cultured for 11 days until primary follicles had formed. We then stained the entire whole‐mount tissue for DAZL and SYCP3 to identify mature (DAZL+) and meiotic (SYCP3+) germ cells and could observe on average 200 oocytes in cysts per aggregate and moreover, around 50 primary follicles (Fig 4H and I ), similar to what has been observed previously (Hamada et al , 2020 ), showing that our XGFP− PGCLCs could indeed mature further. Taken together, germ cells seem to adopt highly similar transcriptomes when two active X chromosomes are present, irrespective of their parental condition of origin and hence regardless of whether cells underwent X‐inactivation followed by X‐reactivation (XGFPhigh), or were constitutively X‐active (XGFP+). Moreover, germ cells can undergo X‐reactivation in the absence of the meiotic gene expression programme, suggesting that X‐reactivation is not dependent on meiotic entry. However, our data suggest that X‐inactivation may be either functionally important, or, alternatively, a predictive indicator for subsequent germ cell maturation and entry into meiosis, as germ cells originating from constitutively active XGFP+ PGCLCs failed to acquire a meiotic transcriptional profile.

X‐reactivation occurs progressively during germ cell maturation and is almost complete with meiotic entry
Having shown that cells could undergo X‐reactivation in the absence of meiosis, we nevertheless wanted to assess if X‐reactivation was a prerequisite for meiotic entry. In order to analyse X‐reactivation dynamics in more detail, we again took advantage of the hybrid background of our XRep cell line and performed allele‐specific RNA expression analysis, which allowed us to successfully detect allele‐specific expression of 220 X‐linked genes (see methods). To first assess the X‐status on a chromosome‐wide level, we calculated the average allelic ratio of all X‐linked genes (Fig 5A ). As expected, we observed biallelic X‐linked gene expression of the XGFP+ and XGFPhigh mitotic clusters 1 and 2, as reflected by an average allelic ratio of 0.5. However, cells of the pre‐meiotic and meiotic clusters, despite originating from mostly XGFPlow and XGFPintermediate populations, showed close to biallelic expression at an average allelic ratio of ~ 0.4, as the sensitivity of the XGFP reporter was insufficient to mark cells as reactivated if they had low levels of X‐inactivation (Fig 5B ). We therefore assessed the X‐status on a gene‐by‐gene level and compared it to the data of ESCs, EpiLCs and XGFP− PGCLCs (Fig 5C and D ). In addition to 78 escapees, being active throughout the differentiation, we observed early X‐chromosome reactivation (early XCR) of 58 genes in pre‐meiotic cells. Therefore, the vast majority of genes (85%) had escaped X‐inactivation in the first place, or had undergone reactivation, before the onset of meiosis. Moreover, 17 genes reactivated as cells underwent meiosis (late XCR), with only 8 genes still being inactive in meiotic cells (no XCR). Furthermore, we observed that early reactivating genes displayed higher allelic ratios in XGFP− PGCLCs compared to late reactivating genes (Fig 5D and E ), suggesting that the degree of silencing could influence X‐reactivation timing. Taken together, in contrast to our observations based on the XGFP reporter, we observed by allele‐specific RNA‐Seq that reactivation of X‐linked genes was almost complete by the time of meiotic entry. In detail, X‐reactivation in germ cells seems to occur in two waves. First, before the onset of meiosis for the majority of genes and second, concomitantly with meiotic entry for a small subset of genes.

X‐inactivation in PGCLCs is associated with an increased meiotic competence
Our single‐cell RNA‐seq analysis showed an exclusive ability for XGFP− over XGFP+ PGCLCs to differentiate into mature germ cells with a meiotic transcriptional profile in the rOvary system. We therefore wanted to functionally compare their ability to enter meiosis and their capability to differentiate to more mature stages. To be able to assess in more detail the extent of prophase I progression, we cultured XGFP− and XGFP+ PGCLCs for an additional 9 days on immortalized m220 stromal feeder cells in the presence of BMP2 and retinoic acid (Fig 6A ), which was previously shown to facilitate entry into meiosis (Miyauchi et al , 2017 ). During this expansion culture, we observed a progressive accumulation of SYCP3+ meiotic cells (Fig 6B and C ). In particular, cells originating from XGFP− PGCLCs formed a significantly higher number of meiotic cells than XGFP+ PGCLCs, confirming our previous observations in the rOvary system. Intriguingly, by day 5 of the expansion culture, all SYCP3+ meiotic cells were XGFP+ (Fig 6B and D ), indicating the co‐occurrence of XGFP‐reactivation with meiotic entry in this system as well. We then prepared chromosomal spreads from the expansion cultures and performed immunostaining for SYCP3, which shows a distinctive pattern according to the different prophase stages (Fig 6E ). Moreover, to aid the correct recognition of the different stages, we stained for the double‐strand break marker γH2AX, a phosphorylated form of the histone variant H2AX (Mahadevaiah et al , 2001 ). This showed that the majority of SYCP3+ cells could successfully enter the zygotene stage by day 9 of expansion culture (Fig 6E and F ). While the efficiency of meiotic entry was significantly higher from XGFP− PGCLCs (Fig 6C ), XGFP+ PGCLCs frequently formed abnormal, extensively proliferating colonies (Fig 6G ) in the expansion system, with significantly larger aggregates (Fig 6H ). Nevertheless, once committed to meiosis, XGFP+ derived cells were able to progress to the Zygotene stage as well (Fig 6E and F ), suggesting that the major developmental bottleneck for XGFP+ cells appears to be early during PGCLC development and meiotic entry. Taken together, while both XGFP+ and XGFP− PGCLCs are able to reach Zygotene stage, we found that XGFP− originating cells were able to enter meiosis at significantly higher efficiency, corroborating results obtained in the rOvary system that X‐inactivation in PGCLCs is associated with increased meiotic and oogenic potential.

While X‐chromosome inactivation has been a long‐studied phenomenon (Lyon, 1961 ) and has been shown to play an important biological role for embryonic development (Marahrens et al , 1997 ) and pluripotency exit (Schulz et al , 2014 ), its reversal by X‐reactivation and its biological function during germ cell development have remained elusive to date. Previous studies on X‐chromosome dynamics during female mouse germ cell development have been hampered by a lack of allelic resolution, a low number of genes assessed, as well as an inability to directly trace the X‐chromosome status of single cells (Sugimoto & Abe, 2007 ; Chuva de Sousa Lopes et al , 2008 ). To overcome these limitations, we generated with XRep an in vitro system, which allowed us to reveal the X‐chromosome inactivation and reactivation cycle and its functional relation to germ cell development and meiotic progression. Although our XGFP marker has its limitations by showing a slight delay in response during kinetic changes, such as downregulation during X‐inactivation and upregulation during X‐reactivation, it allowed us to isolate PGCLC populations with distinct developmental capacity. We thereby uncovered that X‐inactivation is an important hallmark of proper PGCLC differentiation in order to progress at later stages towards meiotic entry (Fig 7 ). X‐reactivation, on the other hand, coincides temporally with meiotic maturation. This is in line with the timing of X‐reactivation in mouse germ cells in vivo (Sugimoto & Abe, 2007 ; Chuva de Sousa Lopes et al , 2008 ; Sangrithi et al , 2017 ), where it takes place gradually, initiating during germ cell migration and peaking after colonization of the gonads around the time of meiotic entry. Additionally, our in vitro system enabled the isolation of PGCLCs harbouring two active X, a unique advantage over in vivo systems, as it allowed us to compare the differentiation potential of PGCLCs with and without X‐inactivation. While our results suggest that PGCLC specification can occur in the absence of X‐inactivation, we found that germ cells, which had never undergone X‐inactivation in the first place, or in which X‐reactivation occurred pre‐emptively, displayed a mitotic germ cell character, did not enter a normal meiotic trajectory on a transcriptomic level and showed a significantly reduced meiotic differentiation capacity. This further highlights how timely X‐inactivation and ‐reactivation might be necessary for proper germ cell maturation (Fig 7 ). Moreover, while we acknowledge that our findings are based on data generated in vitro , we note that allele‐specific single‐cell RNA‐seq of E5.5‐E6.5 epiblast cells, the precursors of PGCs, revealed a considerable heterogeneity in X‐inactivation progression at this developmental time window (Mohammed et al , 2017 ; Cheng et al , 2019 ; preprint: Naik et al , 2021 ; Lentini et al , 2022 ), which could potentially allow cells to give rise to XaXa PGCs, similar to our XGFP+ PGCLCs. Indeed, when we analysed data from E6.5 epiblast in vivo (Cheng et al , 2019 ), we detected specifically in the proximal epiblast, which contains the competent precursor cells for PGC development (Lawson & Hage, 1994 ; Saitou et al , 2002 ), a high fraction of cells, which have not gone through X‐inactivation (Fig 3H and I ). Furthermore, proximal epiblast cells which did go through X‐inactivation displayed a lower degree of gene silencing than other cells of the epiblast or extraembryonic tissues at that stage (Fig 3J and K ). Thus, our data support the idea of a potential functional link between appropriate X‐chromosome dosage compensation kinetics and developmental progression during mammalian germ cell maturation. Nevertheless, we acknowledge that it would be important to validate our results during later stages of germ cell development in vivo . Particularly, whether PGCs that have failed to go through X‐inactivation exist, and to what fate these cells would commit. It remains an open question, what could be the potential role of X‐inactivation for proper PGCLC development and if it is a driver or, alternatively, a diagnostic mark for meiotic competence of germ cells. We observed that XGFP+ PGCLCs, which failed to undergo X‐inactivation, differed from XGFP− PGCLCs on multiple accounts. Albeit sharing an overall similar transcriptome signature with their XGFP− germ cell counterparts, XGFP+ PGCLCs displayed ESC‐like features including a higher expression of naive pluripotency genes, shortened cell cycle and propensity to form pluripotent EGC colonies when cultured under 2i/LIF conditions. An explanation for this pluripotency‐related phenotype could be the two‐fold expression of critical X‐linked dosage‐sensitive genes, which need to be silenced by X‐inactivation to allow normal pluripotency exit during ESC differentiation (Schulz et al , 2014 ). For example, Dusp9 , an X‐linked regulator of MAPK signalling, has been shown to be responsible for the lower DNA‐methylation levels of XX pluripotent stem cells, when compared with XY and XO cells (Choi et al , 2017 ; Song et al , 2019 ; Genolet et al , 2021 ). In germ cell development, DNA methylation safeguards repression of late germ cell / meiotic genes during early germ cell stages and demethylation of their promoters is required for their upregulation during germ cell maturation and meiotic entry (Yamaguchi et al , 2012 ; Hill et al , 2018 ). Along those lines, we observed that XGFP+ PGCLCs also displayed precocious expression of a subset of late germ cell markers, which remained repressed in XGFP− PGCLCs. Importantly, demethylation of late germ cell genes alone has been shown to only lead to partial activation of some germ cell genes, while not being sufficient for their full expression in the absence of meiosis‐inducing signals (Miyauchi et al , 2017 ; Ohta et al , 2017 ). This would explain our observation of a relatively mild upregulation of late germ cell genes in our XGFP+ PGCLCs and why this was not sufficient to aid entrance of XGFP+ cells into a full meiotic trajectory after their aggregation with gonadal somatic cells. Klhl13 , another X‐linked MAPK pathway regulator, has been recently described to promote pluripotency factor expression, thereby delaying differentiation when expressed at double dose (Genolet et al , 2021 ). The counterbalance between pluripotency vs. differentiation‐promoting signalling responses was also observed in our gene expression analysis, in which we found “MAPK regulation” and “WNT signalling” to be enriched GO terms in XGFP− PGCLCs, while “response to LIF” was enriched in XGFP+ PGCLCs (Fig 2F ). Apart from being involved in pluripotency, MAPK inhibition (Kimura et al , 2014 ) as well as WNT‐ and LIF‐signalling pathways (Ohinata et al , 2009 ; Hayashi et al , 2011 ) play facilitating roles during PGCLC induction, therefore differential enrichment of these pathways in our XGFP+ and XGFP− PGCLCs might contribute to their distinct developmental potentials. Taken together, the combination of these differential features might lead to a reduced mitotic propensity of XGFP− PGCLCs, which might prime them for meiotic entry, while XGFP+ PGCLCs rather remain mitotic and do not enter meiosis. To which degree this may be a cause or consequence of the X‐inactivation status in PGCLCs and how X‐linked gene dosage might affect female germ cell development will need to be functionally addressed by future studies, for example, by testing the meiotic potential of Xist knockout cells. Two recent studies used human in vitro PGCLC differentiation systems to investigate if the X‐inactivation state of human pluripotent stem cells would affect their propensity to differentiate towards the germ cell lineage. While the first study did not find a significant effect (Chang et al , 2021 ), the second study reported that human iPSCs with eroded X‐inactivation showed a lower efficiency in forming PGCLCs when compared to cells with a higher degree of X‐inactivation (Yokobayashi et al , 2021 ). This suggests that faithful X‐dosage control might be an important feature of both mouse and human germ cell development. While we found that X‐inactivation marked PGCLCs of full potential for subsequent meiosis and oogenesis, X‐reactivation occurred progressively during their transition from pre‐meiotic into meiotic stages. Evidently, X‐reactivation is not dependent on meiotic entry as it occurred completely in mitotic germ cells as well, and X‐reactivation by itself was also not sufficient for germ cells to enter a meiotic trajectory. However, it remains to be tested whether X‐reactivation is a requirement for female germ cells to progress through meiosis, or if the two processes are functionally unrelated. As in the case of pluripotency, reactivation of dosage‐sensitive X‐linked genes could enable the initiation of the meiotic gene expression programme by promoting the derepression and upregulation of meiotic genes (Yamaguchi et al , 2012 ; Hill et al , 2018 ). The absence of double X dosage and/or abnormalities in meiotic pairing ability greatly diminishes the success rate of XO and XY germ cells to pass through meiotic prophase due to delay of meiotic initiation and meiotic arrest when compared to XX germ cells (Hamada et al , 2020 ). Therefore, equalizing the chromatin state between the heterochromatic inactive X and euchromatic active X by X‐reactivation could be a necessary step in order to allow X‐X chromosome pairing during meiotic prophase. Our XRep system will provide a unique tool to test the potential requirement of X‐reactivation for meiotic progression and thereby reveal the biological function of the intriguing epigenetic yoyo of X‐inactivation and ‐reactivation in the mammalian germ cell lineage.

Materials and Methods

Cell culture

Embryonic stem cell culture: Serum/LIF
Embryonic Stem Cells (ESCs) were maintained and expanded on 0.2% gelatin‐coated dishes in DMEM (Thermo Fisher Scientific, 31966021) supplemented with 10% Fetal Bovine Serum (FBS) (ES‐qualified, Thermo Fisher Scientific, 16141079), 1,000 U/ml LIF (ORF Genetics, 01‐A1140‐0100), 1 mM Sodium Pyruvate (Thermo Fisher Scientific, 11360070), 1x MEM Non‐Essential Amino Acids Solution (Thermo Fisher Scientific, 11140050), 50 U/ml penicillin/streptomycin (Ibian Tech, P06‐07100) and 0.1 mM 2‐mercaptoethanol (Thermo Fisher Scientific, 31350010). Cells were cultured at 37°C with 5% CO 2 . Medium was changed every day and cells were passaged using 0.05% Trypsin‐EDTA (Thermo Fisher Scientific, 25300054) and quenched 1:5 in DMEM supplemented with 10% FBS (Life Technologies, 10270106). Cells were monthly tested for mycoplasma contamination by PCR.

Embryonic stem cell culture: 2i/LIF
ESCs were cultured for 24 h prior to the start of the primordial germ cell‐like cell induction in 2i/LIF medium. Briefly, a home‐made version of the N2B27 medium was prepared based on previous reports (Ying et al , 2008 ) with additional modifications reported in Hayashi and Saitou ( 2013 ) containing two chemical inhibitors 0.4 µM PD032591 (Selleck Chemicals, S1036) and 3 µM CHIR99021 (SML1046, SML1046) together with 1,000 U/ml LIF (ORF Genetics, 01‐A1140‐0100). ESCs were seeded on a dish coated with 0.01% poly‐L‐ornithine (Sigma‐Aldrich, P3655) and 500 ng/ml laminin (Corning, 354232).

XRep cell line generation
We used the female F2 ESC line EL16.7 TST (obtained from Jeannie Lee, Massachusetts General Hospital (Boston, USA)), derived from a cross of Mus musculus musculus with Mus musculus castaneus (Ogawa et al , 2008 ). As a result, cells contain one X chromosome from M.m. musculus (X mus ) and one from M.m. castaneus (X cas ). Moreover, EL16.7 TST contains a truncation of Tsix on X mus (Tsix TST/+ ), which abrogates Tsix expression and leads to the non‐random inactivation of X mus upon differentiation. XGFP and XtdTomato vectors were integrated first, followed by integration of rtTA and last of germ cell transcription factor vectors.

XGFP and XtdTomato dual‐colour reporter
A GFP reporter construct (Wu et al , 2014 ) was targeted in the second exon of Hprt on X mus as described in Bauer et al ( 2021 ). The same strategy was used to simultaneously target a tdTomato reporter construct in the second exon of Hprt on X cas and a GFP reporter on X mus . Briefly, 5 × 10 6 EL16.7 TST ESCs were nucleofected with the AMAXA Mouse Embryonic Stem Cell Nucleofector kit (LONZA, VPH‐1001) using program A‐30 with 1.6 µg each of GFP and tdTomato circularized targeting vectors and 5 µg single gRNA vector PX459 (5′‐TATACCTAATCATTATGCCG‐3′) (Addgene, 48139, a gift from Feng Zhang). Homology arms flanking the target site were amplified from genomic DNA and cloned into pBluescript II SK(+) (Addgene, 212205) by restriction enzyme‐based cloning and the cHS4‐CAG‐nlstdTomato‐cHS4 and cHS4‐CAG‐nlsGFP‐cHS4 constructs, kindly provided by J. Nathans (Wu et al , 2014 ), were cloned between the two homology arms. 7.5 µM of RS‐1 (Merck, 553510) was added to enhance homology‐directed repair. To select for the homozygous disruption of Hprt , cells were grown in the presence of 10 µM 6‐thioguanine (Sigma‐Aldrich, A4882‐250MG) for 6 days, and GFP+ / tdTomato+ cells were isolated by FACS using a BD Influx (BD Biosciences). Single clones were screened by Southern blot hybridization as described in (Bauer et al , 2021 ).

Rosa26 rtTA
One microgram of R26P‐M2rtTA‐targeting vector (Addgene, 47381) and 5 µg of PX459 gRNA vector (5′‐GACTCCAGTCTTTCTAGAAGA‐3′) were nucleofected with the AMAXA Mouse Embryonic Stem Cell Nucleofector kit (LONZA, VPH‐100) using program A‐30 in the XRep. Cells were selected with 3 μg/ml puromycin (Ibian tech., ant‐pr‐1) for 5 days, with medium being changed daily. Single clones were screened for rtTA expression by quantitative RT–PCR and by Southern blot hybridization, with genomic DNA being digested by EcoRV.

Germ cell transcription factors overexpression
PB‐TET vectors containing key germ cell factors Blimp1, Tfap2c and Prdm14 (Nakaki et al , 2013 ) were kindly given by F. Nakaki. Cells were transfected with 3 µg each of PB‐TET vectors, pPBCAG‐hph and a PiggyBac transposase vector using the AMAXA Mouse Embryonic Stem Cell Nucleofector kit (LONZA, VPH‐1001). Transfected cells were selected with 200 μg/ml hygromycin B Gold (Ibian Tech., ant‐hg‐1) for 10 days and genotyped by PCR for transgenes. The primer sequences are shown in Table 1 . Copy number integration was estimated by Southern blot hybridization. Briefly, 15 µg of genomic DNA were digested with BamHI. DNA fragments were electrophoresed in 0.8% agarose gel and transferred to an Amersham Hybond XL membrane (GE Healthcare, RPN303S). The b‐geo probe was designed downstream of the BamHI site, obtained by digesting the PB‐TET‐Avi‐Blimp1 plasmid with CpoI/SmaI, labelled with dCTP [α‐32P] (Perkin Elmer, NEG513H250UC) using High Prime (Roche, 11585592001), purified with an Illustra ProbeQuant G‐50 Micro Column (GE Healthcare, 28903408) and hybridization performed in Church buffer. Radioisotope images were captured with a Phosphorimager Typhoon Trio.

Epiblast‐like cell and primordial germ cell‐like cell induction
XRep ESCs were induced into PGCLCs as described previously (Hayashi & Saitou, 2013 ) with the following modifications as this condition was most efficient in generating PGCLCs. ESCs were thawed on 0.2% gelatin in serum/LIF and after 24 h seeded at a density of 0.6 × 10 5 cells/cm 2 in 2i/LIF medium on a dish coated with 0.01% poly‐L‐ornithine (Sigma‐Aldrich, P3655) and 500 ng/ml laminin (Corning, 354232). Twenty‐tour hours later, ESCs were dissociated with TrypLE express for 5 min at 37°C and induced into EpiLCs by addition of human recombinant basic fibroblast growth factor (bFGF) (Invitrogen, 13256‐029) and activin A (Peprotech, 120‐14P) and seeding on 16.7 µg/ml human plasma fibronectin‐coated plates (Merck Millipore, FC010). After 48h, EpiLCs were split using TrypLE Express (Life Technologies 12604013) and re‐seeded at 0.2 × 10 5 cells/cm 2 on 16.7 µg/ml human plasma fibronectin‐coated plates. After an additional 48 h, EpiLCs were aggregated in U‐bottom 96‐well Lipidure‐Coat plate (Thermo Fisher Scientific, 81100525) at 2,000 cells per aggregate in GK15 medium (GMEM (Life Technologies, 11710035), 15% KnockOut Serum Replacement (KSR) (Thermo Fisher, 10828028), 0.1 mM nonessential amino acids (NEAA) (Thermo Fisher Scientific, 11140050), 1 mM sodium pyruvate (Thermo Fisher Scientific, 11360), 2 mM Glutamax (Life Technologies, 35050061), 0.1 mM 2‐mercaptoethanol (Thermo Fisher Scientific, 21985‐023) and 100 U/ml penicillin and 0.1 mg/ml streptomycin (Thermo Fisher Scientific, 15140) with 1.5 µg/ml doxycycline (Tocris, 4090/50) for 5 days.

Embryonic Germ Cell (EGC) colony‐forming assay
Primordial germ cell‐like cells were sorted by FACS at day 5 of differentiation for their XGFP status and seeded onto immortalized mouse embryonic fibroblasts at 1,000 cells/well of a six‐well plate. Cells were cultured for 7 days in 2i/LIF medium thereby facilitating their transition into pluripotent EGC colonies (Leitch et al , 2010 ), changing the medium every 24 h. EGC colony‐forming capacity of XGFP+ and XGFP− PGCLCs was compared to ESCs replated at equal numbers and scored by alkaline phosphatase (AP) staining. Cells were fixed in 4% paraformaldehyde before adding the AP solution (1 mg/ml Fast Red (Sigma, F8764) and 0.01% w/v Naphthol AS‐MX phosphate (Sigma, 855)). The reaction was stopped after 10 min by removing the AP solution and washing the cells with MillQ water. Each well was imaged with a digital camera and EGC colonies were counted manually.

PGCLCs mitotic expansion
PGCLC mitotic expansion culture was performed as previously described (Ohta et al , 2017 ) with few modifications. Briefly, 5 days after PGCLC induction, SSEA1+/CD61+ PGCLCs were sorted by flow cytometry onto m220 feeder cells, which constitutively express a membrane‐bound form of mouse stem cell factor (Dolci et al , 1991 ; Majumdar et al , 1994 ) on 0.1% gelatin‐coated optical bottom plates (Nunc, 165305). The expansion culture was maintained for a total of 9 days. The first 3 days in GMEM containing 100 ng/ml SCF (Peprotech, 250‐03), 10 µM forskolin (Sigma‐Aldrich, F3917), 10 µM rolipram (Abcam, ab120029), 2.5% FBS (Capricorn Scientific, FBSES12B), 10% KSR, 0.1 mM NEAA, 1 mM sodium pyruvate, 2 mM Glutamax (Life Technologies, 35050061), 0.1 mM 2‐mercaptoethanol, 100 U/ml penicillin, 0.1 mg/ml streptomycin and 100 nM all‐trans retinoic acid (RA) (Enzo Life Sciences, BMLGR100).

PGCLCs meiosis induction
Meiosis was induced after 3 days of mitotic expansion culture as previously reported (Miyauchi et al , 2017 , 2018 ) by a combined treatment of 300 ng/ml BMP2 (R&D Systems, 355‐BM) and 100 nM RA. Medium was replaced completely every 2 days until the end of the culture period.

rOvary reconstitution
A total of 10,000 sorted SSEA1+/CD61+ PGCLCs were mixed with 75,000 freshly thawed E13.5 female somatic gonadal and mesonephric cells (SSEA1−/CD31−) or E12.5 female somatic gonadal cells from CD1/ICR strain mice and cultured in Lipidure‐Coat plates at 37°C in a 5% CO 2 incubator for 6 days for the scRNAseq protocol or for 2 days for the IVDi as described in Hayashi et al ( 2017 ). Mouse care and procedures were conducted according to the protocols approved by the Ethics Committee on Animal Research of the Parc de Recerca Biomèdica de Barcelona (PRBB) and by the Departament de Territori i Sostenibilitat of the Generalitat de Catalunya (Ref. No. 10469).

Oocyte in vitro differentiation (IVDi) culture
IVDi culture was performed as previously described (Hayashi et al , 2017 ). Briefly, one single rOvary was placed in the middle of a 24‐well Transwell‐COL membrane (Corning, CLS3470‐48EA) and cultured in alpha‐MEM (Life Technologies, 12571063) with 0.15 mM ascorbic acid (Sigma‐Aldrich, A7506), 2% FBS, 2 mM Glutamax (Life Technologies, 35050061), 0.1 mM 2‐mercaptoethanol and 50 U/ml penicillin/streptomycin under normoxic condition (20% O 2 and 5% CO 2 at 37°C) for 11 days, changing IVDi medium every other day.

Fluorescence‐activated cell sorting (FACS)
After 5 days of culture, PGCLC aggregates were dissociated using TrypLE Express (Thermo Fisher Scientific, 12604021) for 8 min at 37°C, with periodical tap mixing. The reaction was quenched 1:5 with wash buffer DMEM/F12 (Thermo Fisher Scientific, 11320‐082) containing 0.1% bovine serum albumin (BSA) fraction V (Thermo Fisher Scientific, 15260‐037) and 30 mM HEPES (Gibco, 15630‐056) containing 0.1 mg/ml of DNAse I (Sigma‐Aldrich, DN25‐10MG). The cell suspension was centrifuged at 300 g for 5 min, resuspended in FACS buffer (0.1% BSA in PBS) and passed through a 70 µm cell strainer (Corning, 352350). Cells were stained with 1:100 SSEA1‐eFluor 660 (Thermo Fisher Scientific, 50‐8813‐42) and 1:10 CD61‐PE‐Vio770 (Miltenyi Biotec, 130102627) for 1h at 4°C. Cells were washed thrice with FACS Buffer, stained with 1:1,000 DAPI (Thermo Fisher Scientific, D1306) and then FACS sorted using a BD FACSAria II or a BD Influx. Double‐positive population of PGCLCs was collected in GK15 medium. Data were analysed with Flowjo (Tree Star) software.

Cell cycle analysis
Identification of G1, S and G2/M cell cycle phases was based on DNA content and performed as described previously (Bonev et al , 2017 ) with minor modifications. Briefly, ESCs, EpiLC and PGCLCs were dissociated and quenched as described above. Cells were then fixed for 10 min at room temperature with freshly prepared 1% formaldehyde in PBS (Sigma‐Aldrich, F8775‐4X25ML) and the reaction then quenched by addition of 0.2 M glycine (NZYTech, MB01401) for 15 min on ice. 1 × 10 6 cells/ml were permeabilized using 0.1% saponin (Sigma‐Aldrich, 47036‐50G‐F) containing 10 µg/ml DAPI (Thermo Fisher Scientific, D1306) and 100 µg/ml RNase A (Thermo Fisher Scientific, EN0531) for 30 min at room temperature, protected from light with slight agitation. After washing once with cold PBS, samples were resuspended in cold 0.5% BSA in PBS at a concentration of 1 × 10 6 cells/ml and immediately analysed using a BD LSRFortessa.

Immunofluorescence of PGCLC bodies and rOvaries
Immunofluorescence analysis of PGCLC bodies or rOvaries was performed on cryosections prepared as follows: Aggregates were fixed with 4% paraformaldehyde (PFA) (Electron Microscopy Science, 15713) in PBS at room temperature for 30 min, followed by three washes in PBS and submerged in serial concentrations of 10 and 30% of sucrose (Sigma‐Aldrich, S0389) in PBS, 15 mins and overnight at 4°C respectively. The samples were embedded in OCT compound (Sakura Finetek, 4583), snap‐frozen in liquid nitrogen and cryo‐sectioned at a thickness of 10 µm at −20°C on a cryostat (Leica, CM1850). The sections were placed on a coated glass slide (MAS‐GP type A; Matsunami, S9901) and dried completely. For immunostaining, the slides were blocked with PBS containing 10% normal goat serum (NGS) (Abcam, ab7481), 3% BSA (Sigma‐Aldrich, A3311) and 0.2% Triton X‐100 (Sigma‐Aldrich, T9284) for 1 h at room temperature, followed by incubation with the primary antibodies diluted in a 1:1 solution of blocking buffer to PBS with 0.2% Tween (PBST) (Sigma‐Aldrich, P7949) overnight at room temperature. The slides were washed three times with PBST, then incubated with the secondary antibodies diluted as the primary, with DAPI at 1 µg/ml for 1 h at room temperature. Following three washes in PBST, the samples were mounted in VECTASHIELD with DAPI (Vector Laboratories, H1200) and observed under a Leica SP8 confocal microscope. All images were analysed using Fiji/Image J software (Schindelin et al , 2012 ). All antibodies used in this study are listed in Table 2 .

Immunofluorescence of cultured PGCLC‐derived cells
Immunofluorescence analysis of cultured PGCLC‐derived cells was performed as described in (Nagaoka et al , 2020 ). Briefly, PGCLCs were cultured on m220 feeder cells seeded on a 0.1% gelatin‐coated plate used specifically for imaging (Nunc, 165305). PGCLC‐derived cells were fixed at c5, c7 or c9 with 4% PFA (Electron Microscopy Science, 15713) in PBS at room temperature for 30 min, followed by three washes in PBS. Fixed cells were blocked in PBS containing 10% NGS, 3% BSA and 0.2% Triton X‐100 for 1 h, then incubated with the primary antibodies diluted in a 1:1 solution of blocking buffer to PBS with 0.2% Tween (PBST) at room temperature overnight. After three washes in PBST, cells were incubated with the secondary antibodies and DAPI at room temperature for 2 h and washed three times in PBST. Finally, the well was filled with VECTASHIELD without DAPI (Vector laboratories, H1000). Immunostained samples were observed with a Leica SP8 confocal microscope.

Meiotic cell spreads
Cultured PGCLC‐derived cells were harvested by TrypLE Express at 37°C for 5 min, quenched with 1:1 TrypLE wash buffer (DMEM/F12 containing 0.1% BSA fraction V, 30 mM HEPES), filtered through a 70 µM strainer and centrifuged at 300 g for 5 min. Cell pellets were dislodged by tapping and washed once in PBS. Cells were then treated with a hypotonic solution (30 mM Tris–HCl, 50 mM sucrose (Sigma, S0389), 17 mM trisodium citrate, 5 mM ethylenediaminetetraacetic acid (EDTA), 2.5 mM dithiothreitol (DTT) (Sigma, D0632) and 0.5 mM phenylmethylsulfonylfluoride (PMSF) (Sigma, P7626)), pH 8.2‐8.4 at room temperature for 20 min. Cells were spun down 3 min at 300 g , resuspended in 100 mM sucrose and the cell suspension distributed onto slides (Matsunami, S9901) covered with 1% PFA in H 2 O (Electron Microscopy Science, 15713) with 0.2% Triton X‐100 (pH 9.2–9.4). The slides were incubated at room temperature overnight in a humidified chamber. Finally, the slides were air‐dried and washed with 0.5% Kodak Photo‐Flo 200 (Kodak, B00K335F6S) for 2 min at room temperature. The spread slides were blocked in PBS containing 10% NGS, 1% BSA for 1 h and then incubated with the primary antibodies diluted in a 1:1 solution of blocking buffer to PBS with 0.2% Tween (PBST) at room temperature overnight. After three washes in PBST, cells were incubated with the secondary antibodies and DAPI at room temperature for 2 h, washed three times in PBST and mounted in VECTASHIELD mounting medium with DAPI (Vector Laboratories, h1200). Immunostained cells were observed under a Leica SP8 confocal microscope.

Immunofluorescence of IVDi tissues
Day 11 IVDi tissues were treated while still attached to the transwell member as follow: culture medium was carefully removed from the transwell and the whole membrane was fixed in 4% PFA (Electron Microscopy Science, 15713) in PBS for 30 min at room temperature, washed twice with PBS and blocked overnight at room temperature in 10% NGS, 1% BSA and 0.2% Triton X‐100. Primary antibodies were diluted in a 1:1 solution of blocking buffer to PBS with 0.2% Tween (PBST) and incubated overnight. After three washes with PBST, secondary antibodies and DAPI diluted as the primary were incubated an additional overnight, washed thrice and the whole membrane mounted on VECTASHIELD with DAPI (Vector Laboratories, H1200). Immunostained tissues were observed under a Leica SP8 confocal microscope.

Tilescan analysis
All images were analysed using Fiji/Image J software (Schindelin et al , 2012 ). XtdTomato fluorescence was used to determine aggregate size. First, Gaussian Blur with Sigma (Radius) 5.00 was applied. Then, a threshold with 10–255 and settings “dark background” and “B&W” was set. Finally “Analyze Particles” was used to measure aggregate size.

RNA fluorescent in situ hybridization and immunofluorescence
Cells were fixed with 3% paraformaldehyde PFA (Electron Microscopy Science, 15713) for 10 min with 2 mM Ribonucleoside‐Vanadyl Complex RVC (New England Biolabs, S1402S) at room temperature and then permeabilized for 5 min on ice in 0.5% Triton‐X with 2 mM RVC. Cells were then blocked in 3% BSA/PBS with 2mM RVC for 1h at room temperature, incubated with primary antibodies diluted in blocking solution with 2mM RVC overnight at 4°C. The secondary antibodies were diluted in blocking buffer and incubated 1h at room temperature. Cells were then again fixed in 3% PFA for 10 min at room temperature. Strand‐specific RNA FISH was performed with fluorescently labelled oligonucleotides (IDT) as described previously (Del Rosario et al , 2017 ). Briefly, probe mix was prepared by mixing 10 ng/ml equimolar amounts of Cy5 labelled Xist probes BD384‐Xist‐Cy5‐3' (5'‐ATG ACT CTG GAA GTC AGT ATG GAG /3Cy5Sp/ ‐3'), BD417‐5'Cy5‐Xist‐Cy5‐3' (5'‐ /5Cy5/ATG GGC ACT GCA TTT TAG CAA TA /3Cy5Sp/ ‐3'), 0.5 µg/µl yeast t‐RNA (Life Technologies, 15401029) and 20 mM RVC. Probe mix was pre‐annealed at 80°C for 10 min followed by 30 min at 37°C and hybridized in 25% formamide, 10% dextran sulphate and 2xSSC pH 7 at room temperature overnight. Slides were then washed in 25% formamide 2xSSC pH 7 at room temperature, followed by washes in 2xSSC pH 7 and then mounted with Vectashield (Vector Laboratories, H1200). Images were acquired using a Zeiss Cell Observer.

RNA extraction, cDNA synthesis and qPCR analysis
Total RNA was isolated from ESCs, EpiLCs and PGCLCs (two biological replicates each, corresponding to two different clones, with further two technical replicates each) using phenol‐chloroform extraction (Sigma Aldrich, P2069) followed by ethanol precipitation and quantified by Nanodrop. cDNA was produced with a High‐Capacity RNA‐to‐cDNA kit (Thermo Fisher Scientific, 4387406) and was used for qRT–PCR analysis in triplicate reactions with Power SYBR Green PCR Master Mix (Thermo Fisher Scientific, 4367659). The gene expression levels are presented as ∆∆Ct normalized with the mean Ct values of one housekeeping gene, Arbp, in a normalization sample (ESCs). The primer sequences used in this study are listed in Table 1 .

Bulk RNA‐seq analysis
RNA libraries were prepared using the TruSeq Stranded Total RNA Library Preparation kit (Illumina, 20020596) followed by 125 bp paired‐end sequencing on an Illumina HiSeq 2500.

Allele‐specific analysis
FastQ files that passed quality control were aligned to the mm10 reference genome containing CAST/EiJ and 129S1/SvImJ SNPs positions masked. The positions of all 36 mouse strains SNPs were downloaded from ftp: https://ftp‐‐1505‐SNPs_Indels/mgp.v5.merged.snps_all.dbSNP142.vcf.gz.tbi . From here, we generated a VCF file containing only the SNPs information for the strains of interest, CAST/EiJ and 29S1/SvImJ. Reads with ≥ 1 SNPs were retained and aligned using STAR (Dobin et al , 2013 ) implementing the WASP method (van de Geijn et al , 2015 ) for filtering of allele‐specific alignments. The generated bam files were used for counting reads using the HTseq tool (v0.6.1) (Anders et al , 2015 ). All of the steps above were performed using a customized Nextflow pipeline (Di Tommaso et al , 2017 ). We obtained between 50x10 6 and 75x10 6 reads per replicate. Coherence among samples, time points and replicates was verified by principal component analysis (PCA). Batch effects in principal component analysis (PCA) for comparison to in vivo samples were corrected using the R package limma (Ritchie et al , 2015 ). Differential expression analysis was performed using the R package DESeq2 (v1.16) (Love et al , 2014 ). Briefly, differentially expressed genes were called by comparing XGFP+ PGCLCs and XGFP− PGCLCs or XGFP+ PGCLCs to ESCs. The DESeqDataSet (dds) was generated considering the dataset in its entirety while the DEseq analysis was conducted on dataset filtered as follows: Read counts were normalized by library size using “estimateSizeFactors, ” were filtered for having a mean across the samples > 10 (a more stringent cut‐off than the sum across the samples > 10) and poorly annotated genes on chromosomal patches were removed. The resulting 16,289 genes were kept for downstream analysis. Log2‐fold change was shrinked using the “normal” parameter. Gene ontology enrichment analysis performed on top and bottom differentially expressed genes defined as FDR < 0.001 e log2‐fold change > |1| using the Gorilla. Over‐represented categories were simplified using Revigo ( ) using a similarity of 0.4 as threshold. As background, all identified genes were used.

Single‐cell RNA‐seq analysis
Full‐length single‐cell RNA‐seq libraries were prepared using the SMART‐Seq v5 Ultra Low Input RNA (SMARTer) kit for Sequencing (Takara Bio). All reactions were downscaled to one quarter of the original protocol and performed following thermal cycling manufacturer’s conditions. Cells were sorted into 96‐well plates containing 2.5 µl of the reaction buffer (1× Lysis Buffer, RNase Inhibitor 1 U/µl). Reverse transcription was performed using 2.5 µl of the RT MasterMix (SMART‐Seq v5 Ultra Low Input RNA kit for Sequencing, Takara Bio). cDNA was amplified using 8 µl of the PCR MasterMix (SMART‐Seq v5 Ultra Low Input RNA kit for Sequencing, Takara Bio) with 25 cycles of amplification. Following purification with Agencourt Ampure XP beads (Beckmann Coulter), product size distribution and quantity were assessed on a Bioanalyzer using a High Sensitivity DNA kit (Agilent Technologies). A total of 140 pg of the amplified cDNA was fragmented using Nextera XT (Illumina) and amplified with double indexed Nextera PCR primers (IDT). Products of each well of the 96‐well plate were pooled and purified twice with Agencourt Ampure XP beads (Beckmann Coulter). Final libraries were quantified and checked for fragment size distribution using a Bioanalyzer High Sensitivity DNA kit (Agilent Technologies). Pooled sequencing of Nextera libraries was carried out using a HiSeq4000 (Illumina) obtaining between 0.5 × 10 6 and 1.5 × 10 6 reads per cell. Sequencing was carried out as paired‐end (PE75) reads with library indexes corresponding to cell barcodes. Allele‐specific alignment was done as described for bulk RNA‐seq analysis using STAR and WASP. Data processing and visualization was performed using the R package Seurat (v4.0) (Stuart et al , 2019 ).

In vitro single‐cell analysis
The non‐allelic gene expression matrix was filtered for protein‐coding and non‐coding transcripts using annotations from mmusculus_gene_ensembl version 67. Low‐quality cells with less than 4,000 identified genes, less than 10,000 RNA molecules or more than 5% mitochondrial reads were removed. Data were log normalized and the top 2,000 highly variable features were selected for downstream analysis. The expression matrix was then scaled and linear dimensional reduction was performed. To ensure that our analysis would not be confounded by in vitro differentiation artefacts, we focussed our analysis on germ cells by subsetting for cells with a normalized and scaled Dazl expression greater than 1 (60 of 460 sorted germ cells did not pass this criterium). Moreover, only cells that passed our allelic expression QC (explained below) were retained. Clusters were subsequently identified using “FindClusters” at a resolution of 0.8 on the first 20 principal components and visualized as UMAP projections using “RunUMAP.” Clusters were annotated based on marker gene expression. Processing of allelic data was performed for all cells that passed the Seurat QC. Cells that passed the following criteria were considered for downstream analysis: More than 3,500 total allelic reads (sum of mus and cas), a minimum of 25 allelically expressed genes as well as a minimum of 3% of total allelic reads from either genotype. Moreover, a gene was considered informative if the sum of its allelic reads was higher than 10 and if it was expressed in at least 25% of cells. This resulted in 379 cells that passed all our quality control steps.

In vivo single‐cell analysis
Allelic single‐cell data of E6.5 embryos were obtained from GEO GSE109071 (Cheng et al , 2019 : Data ref: Deng & Cheng, 2019 ), non‐allelic analysis was performed as described above for in vitro . Processing of allelic data was performed for all cells that passed the Seurat QC. Cells that passed the following criteria were considered for downstream analysis: 400 total allelic reads (sum of cas and C57) and a minimum of 25 allelically expressed genes. Moreover, a gene was considered informative if the sum of its allelic reads was higher than 10 and if it was expressed in at least 20% of cells. This gave 239 cells in total and 16,003 informative genes. Single‐cell data of in vivo female germ cells were obtained from GEO GSE130212 (Zhao et al , 2020a : Data ref: Zhao et al , 2020b ). Non‐allelic analysis was performed as described above for in vitro with the following exceptions: low‐quality cells with less than 2,000 identified genes, less than 2,000 RNA molecules were removed. To ensure that the analysis would not be confounded by somatic cells, we focussed it on germ cells by subsetting for cells with a normalized and scaled Dazl expression > 0.5 and removed somatic cells by subsetting for cells with expression < 0.5 for Nr5a1 , Axl , Fosb , Emx2 and Gata4 . This gave 18,417 cells in total and 14,133 informative genes.

RNA velocity analysis
Non‐allele‐specific RNA velocity analysis was performed as follows: Briefly, loom files only of Dazl‐ positive cells were generated from the non‐allelic‐specific BAM files from STAR using velocyto run‐smartseq2 version v0.17.17 using the default parameters, mouse genome assembly mm10 and the UCSC repeat genome masked regions using custom‐made scripts. Subsequently, the loom files were imported into Python version 3.7 and processed using scVelo v0.2.3 (Bergen et al , 2020 ). The metadata, the clusters and the UMAP dimensionality reduction coordinates from Seurat were imported, then the single‐cell data were filtered and normalized with a minimum of 20 counts and 2,000 top genes. The moments for velocity estimations were computed with 20 principal components and 30 neighbours. The genes' full splicing kinetics were recovered before estimating the velocities using the dynamical model . The RNA velocity was visualized using velocity_embedding_stream colour coding cells by their Seurat cluster.

Integration with in vivo datasets
Single‐cell data of in vivo female germ cells (Zhao et al , 2020 ) were analysed as described above. Normalized and scaled in vivo and in vitro data from this study were merged by canonical correlation analysis (CCA) using the Seurat function RunCCA. UMAP was then performed using CCA.

Statistical analysis
Statistical analysis of replicate data was performed using appropriate strategies in R . Number of independent experiments ( n ), type of statistical test, definition of significance and measurements are defined in figure legends. No statistical method was used to predetermine sample size and no data were excluded from the analyses. Samples were not randomized and investigators were not blinded to group allocation during data collection and analysis.

Author contributions
Jacqueline Serverino: Conceptualization; Data curation; Software; Formal analysis; Investigation; Visualization; Methodology; Writing—original draft; Writing—review & editing. Moritz Bauer: Conceptualization; Data curation; Software; Formal analysis; Investigation; Visualization; Methodology; Writing—original draft; Writing—review & editing. Tom Mattimoe: Data curation; Formal analysis; Investigation; Methodology; Writing—review & editing. Niccolo Arecco: Data curation; Software; Formal analysis; Investigation; Methodology; Writing—review & editing. Luca Cozzuto: Software; Methodology. Patricia Lorden: Investigation. Norio Hamada: Resources; Methodology. Yoshiaki Nosaka: Resources; Methodology. So I Nagaoka: Resources; Methodology. Pauline Audergon: Methodology. Antonio Tarruell: Methodology. Holger Heyn: Resources; Supervision; Methodology. Katsuhiko Hayashi: Resources; Supervision; Methodology. Mitinori Saitou: Resources; Supervision; Methodology. Bernhard Payer: Conceptualization; Resources; Supervision; Funding acquisition; Writing—original draft; Writing—review & editing. In addition to the CRediT author contributions listed above, the contributions in detail are: BP, JS and MB conceived the study and wrote the manuscript with input from TM. JS and TM performed experiments. MB and JS established XRep cell line. JS, MB and NA performed bioinformatic analyses. LC wrote the bioinformatic pipeline. JS, MB and NA integrated and visualized data. PA and AT contributed to the initial experimental design setup. PL and HH prepared the single‐cell sequencing libraries. NH, KH, YN, SIN and MS provided constructs, mouse gonadal somatic and m220 feeder cells, and input on experimental set‐up and manuscript writing. BP acquired funding and supervised the research.

Disclosure and competing interests statement
HH is co‐founder of Omniscope and SAB member of MiRXES.

Supporting information

Data availability
The accession number for the sequencing datasets reported in this study is GEO: GSE169201 ( ). Data analysis code: .

The EMBO Journal ( 2022 ) 41 : e109457

Jacqueline Severino, Moritz Bauer, Tom Mattimoe, Niccolò Arecco, Luca Cozzuto, Patricia Lorden, Norio Hamada, Yoshiaki Nosaka, So I Nagaoka, Pauline Audergon, Antonio Tarruell, Holger Heyn, Katsuhiko Hayashi, Mitinori Saitou, Bernhard Payer
The EMBO Journal
John Wiley and Sons Inc.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Figure 10
Figure 11