XACT does not control X chromosome activity in naive hES cells

XISTʼs capacity to attenuate X chromosome transcription instead of triggering complete inactivation made us question whether XACT antagonizes XISTʼs silencing activity. We first validated that two XACT clouds were detected in most (97%) naive cells, as previously described in human preimplantation embryos¹¹ (Fig. 2a,b and Extended Data Fig. ​Fig.2a).2a). Biallelic XACT expression in naive hES cells was not systematically correlated with increased XACT levels compared to primed hES cells (Fig. ​(Fig.2c2c).

Open in a separate window

Fig. 2

XACT does not control X chromosome activity in naive hES cells.

a, Scheme of XACT deletion using CRISPR–Cas9 approach in primed H9 hES cells. Primed WT and XACT KO H9 cells were converted to the naive state using NaiveCult protocol. b, Representative RNA FISH images for XIST and XACT in naive (NaiveCult) and primed H9 hES cells. Percentages of cells displaying the representative pattern are indicated. Scale bar=10µm. c, XACT expression levels in naive (PXGL and NaiveCult) and primed H9 cells obtained from RNA-seq data (n=3). Data are presented as the mean values±s.d. of replicates. d, XACT expression level in XACT WT and KO naive hES cells obtained from RNA-seq data (n=3). Data are presented as the mean values±s.d. of replicates. e, XIST expression level in XACT WT and KO naive cells obtained from RNA-seq data (n=3). Data are presented as the mean values±s.d. of replicates. f, Distribution of the dispersion of XIST RNA FISH signal in XACT WT and KO naive cells (n=50 cells per condition examined over one experiment). The horizontal line represents the median dispersion of each group; Wilcoxon P value≥0.05 (NS). g, X:A ratio from RNA-seq data of WT and XACT KO naive H9 cells (n=3). Data are presented as the mean values±s.d. of replicates. h, Percentage of biallelically expressed SNPs from the X chromosome and chromosome 7 in XACT WT (n=3) and KO (n=3) naive H9 hES cells, obtained from RNA-seq data. Box plots represent the median (center), first and third quartiles (hinges) and ±1.5 IQR (whiskers). Unless otherwise specified, P values were calculated by a two-sided unpaired t-test. Levels of significance: NS (P≥0.05), *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

Source data

Open in a separate window

Extended Data Fig. 2

Characterization of XACT KO clones in naïve hESCs.

a, Quantification of RNA-FISH data showing the percentage of cells with monoallelic or biallelic XIST (green) and/or X-linked gene (XACT; red) signals in naïve (NaiveCult) and primed H9 hESCs. b, Top: Electrophoretic gel pictures showing the PCR amplicons obtained with different screening primers on genomic DNA from XACT targeted primed H9 hESCs. Black rectangles represent 3 independent XACT KO clones which were used for subsequent experiments. Bottom: Schematic representation of XACT deletion strategy showing the positions of the CRISPR sgRNAs (yellow triangles) and screening primers used to characterize the clones (black arrows). c, Expression levels of core and naïve specific pluripotency factors in naïve H9 WT and XACT KO, from RNA-seq data (n=3). Data are presented as mean values +/- SD of replicates. d, PCA displaying the first two principal components of the WT and XACT KO RNA-seq data in naïve and primed H9 hESCs. e, Volcano plot representing the significance versus Log2 fold change for all differentially expressed genes (n=30, FDR<0.05, log2FC > |1|) upon XACT deletion (n=3). f, Representative images of XIST and XACT RNA-FISH in WT and XACT KO naïve (NaiveCult) H9 hESCs. Scale bar= 10µm. Unless otherwise specified, p-values were calculated by two-sided unpaired t-test. Levels of significance: ns≥0.05; * < 0.05; ** < 0.01; *** < 0.001; **** < 0.0001.

Source data

We deleted a 90-kb region encompassing XACT transcription start sites (TSSs) in primed H9 cells (Fig. 2a,d and Extended Data Fig. ​Fig.2b).2b). Three independent homozygous XACT KO clones were successfully converted to a naive pluripotent state. Neither expression of naive-specific and core pluripotency factors nor cell morphology and growth were impacted by XACT deletion (Extended Data Fig. ​Fig.2c).2c). Furthermore, PCA revealed that XACT KO and WT naive H9 cells were grouped together and clustered away from primed hES cells, indicating that loss of XACT does not significantly alter the naive hES cell transcriptome (Extended Data Fig. ​Fig.2d).2d). This was confirmed through differential expression analysis, which revealed few genes with altered expression levels (14 upregulated and 16 downregulated; log₂FC>|1| and FDR<0.05) in XACT KO compared to WT naive H9 cells (Extended Data Fig. ​Fig.2e2e and Supplementary Table 4). Gene Ontology analysis did not reveal any enriched pathways in the list of the 30 DEGs. This indicates that XACT is dispensable for primed to naive resetting and has no major role in naive hES cells.

XIST RNA levels were not altered in XACT KO clones (Fig. ​(Fig.2e).2e). Similarly, the distribution of the XIST signal, which is more dispersed in naive hES cells compared to primed hES cells and fibroblasts^11,26, was similar in XACT WT and KO clones (Fig. ​(Fig.2f2f and Extended Data Fig. ​Fig.2f),2f), indicating that XACT does not regulate XIST expression or localization in naive hES cells. As a proxy for global X chromosome expression, we measured the X:A ratio and observed no significant changes in naive XACT KO compared to WT hES cells (Fig. ​(Fig.2g).2g). Furthermore, the percentage of biallelic X-linked SNPs detected by RNA-seq in XACT KO cells remained unchanged compared to WT, indicating that X chromosome silencing did not initiate in the absence of XACT in naive hES cells (Fig. ​(Fig.2h).2h). These results indicate that XACT is not an antagonist of XIST and does not control X chromosome activity in naive hES cells.

XIST accumulates on the dampened X chromosome in naive hES cells

The XIST accumulation pattern in human preimplantation embryos and naive hES cells is thought to differ from that of primed hES cells, being more dispersed, as shown by RNA FISH^11,26. To address whether the change in XIST repressive activity between naive and primed pluripotency could be linked to the redistribution of XIST RNA, we mapped at high resolution the genomic binding sites of XIST on dampened and inactive X chromosomes in naive and primed H9 hES cells, respectively, using RNA antisense purification followed by high-throughput sequencing (RAPseq). XIST was efficiently and specifically captured from the chromatin of both cell types (Fig. ​(Fig.3a3a).

Open in a separate window

Fig. 3

XIST is broadly distributed on the dampened X chromosome in naive hES cells.

a, Analysis by RT–qPCR of XIST and JPX enrichment after XIST pulldown in naive and primed H9 hES cells. JPX was used as a negative control. Values were normalized to the input (n=2). b, Percentage of aligned DNA sequencing reads per chromosome after XIST pulldown and in input material in naive and primed H9 hES cells (n=2). c, Percentage of XIST RAP peak occupancy per chromosome for primed and naive H9 hES cells (n=2). d, XIST RAP profiles (log₂ enrichment over input) on the X chromosome in naive and primed H9 hES cells. e, Profile plots of XIST RAPseq reads over a 10-kb region flanking the TSSs of X-linked genes in naive and primed H9 hES cells.

Source data

The X chromosome showed the highest percentage of mapped DNA sequencing reads in both naive (60%) and primed (34%) hES cells, while the remaining reads were distributed across all autosomes (Fig. ​(Fig.3b).3b). PCA and matrix correlation analysis showed that the naive RAP XIST replicates clustered together and away from the primed RAP XIST replicates. As the comparison was made in an isogenic framework, we can conclude that the variation in XIST distribution was because of the cellular context (Extended Data Fig. 3a,b). Indeed, XIST was broadly distributed on the X chromosome in naive hES cells, occupying >50% of the chromosome, while XIST contacted a reduced fraction (<30%) of the primed X chromosome, mostly on the long arm (Fig. 3c,d). Differential enrichment analysis confirmed that most of the X chromosome was enriched for XIST in naive compared to primed hES cells (Extended Data Fig. ​Fig.3c).3c). However, XIST was overall depleted from X-linked TSSs compared to flanking sequences in naive hES cells. This pattern was in sharp contrast to that of primed hES cells, where XIST accumulated over TSSs (Fig. ​(Fig.3e).3e). These observations support a role for XIST distribution in the regulation of X-linked genes, where XIST RNA molecules relocate from X-linked promoters to flanking regions upon primed to naive transition.

Open in a separate window

Extended Data Fig. 3

Characterization of XIST enrichment on the X chromosome in naïve and primed hESCs.

a, PCA plot displaying the first two principal components of the XIST RAPseq data in naïve (PXGL) and primed H9 hESCs. b, Heatmap showing the Pearson correlation between naïve and primed XIST RAPseq replicates on the X chromosome. c, XIST RAPseq profiles showing the log2 enrichment over the X chromosome in naïve compared to primed H9 hESCs. d, Scatter plots showing the linear regression line and Pearson correlation for the comparison of XIST coverage scores with gene, LINE, retrotransposon, and LTR densities in naïve and primed H9 hESCs. e, Expression levels of XIST from naïve: PXGL (n=3) and NaiveCult (n=3), and primed (n=3) H9 RNA-seq data. Data are presented as mean values +/- SD of replicates. f, CUT&RUN H3K4me3 profiles across the XIST gene in naïve and primed H9. XIST spliced isoforms detected in naïve and primed H9 along with the A-F repeats are shown. Arbitrary name of each isoform (XIST1.1, XIST1.2, etc) is indicated on the left, together with their expression pattern (Primed and/or Naïve). Unless otherwise specified, p-values were calculated by two-sided unpaired t-test. Levels of significance: ns≥0.05; * < 0.05; ** < 0.01; *** < 0.001; **** < 0.0001.

Source data

We then probed the correlation of XIST distribution with several genomic features in both naive and primed hES cells. In naive hES cells, XIST coverage was moderately correlated with the density of transposable elements (TEs) such as LINEs (long interspersed nuclear elements) and retrotransposons, in contrast to primed hES cells (Extended Data Fig. ​Fig.3d).3d). This could reveal a role for these elements in XIST spreading upon initial XIST upregulation, as previously suggested^29,30. The difference in XIST distribution was not linked to differential XIST expression levels (Extended Data Fig. ​Fig.3e),3e), suggesting that broader coverage of the X chromosome in naive hES cells was not because of an increased number of XIST RNA molecules. XIST transcript reconstruction revealed the presence of multiple XIST isoforms (Extended Data Fig. ​Fig.3f).3f). The XIST 1.1 isoform, with six exons, corresponded to the annotated XIST structure and was expressed in naive and primed hES cells. Other isoforms were specific to the primed or naive contexts and differed in their first and last exon. Importantly, all isoforms were produced from the same TSS and harbored all six repeat elements (A–F) important for XIST function and localization (Extended Data Fig. ​Fig.3f3f)³¹. Altogether, these data reveal that XIST establishes distinct contacts on dampened and inactive X chromosomes, in line with the distinct activities of XIST in naive and primed pluripotent contexts.

PRC histone marks accumulate on the dampened X chromosome

Because XIST is known to indirectly recruit PRC complexes and given its ability to control X chromosome activity in naive hES cells, we interrogated the chromatin landscape of dampened and inactive X chromosomes using CUT&RUN (cleavage under targets and release using nuclease). Levels of H3K27me3, H2AK119Ub and, to a lesser extent, H3K9me3 were higher on the X chromosomes than on autosomes in naive hES cells (Fig. ​(Fig.4a),4a), even though this tendency was less pronounced than in primed hES cells, at least for H3K27me3 and H3K9me3.

Open in a separate window

Fig. 4

PRC-associated repressive histone modifications accumulate on dampened X chromosome in an XIST-dependent manner in naive hES cells.

a, Mean enrichment of H3K27me3, H2AK119Ub and H3K9me3 per chromosome in naive and primed H9 hES cells (n=2). All individual data points are shown. Box plots represent the median (center), first and third quartiles (hinges) and ±1.5 IQR (whiskers). BPM, bins per million mapped reads. b, CUT&RUN profiles (log₂ enrichment over IgG) for H3K27me3, H2AK119Ub and H3K9me3 on the X chromosome in naive and primed H9 hES cells. c, Percentage of H3K27me3, H2AK119Ub and H3K9me3 peak occupancy per chromosome for primed and naive H9 hES cells. Black and green dotted lines represent the median genomic occupancy. d, Representative immunofluorescence images for H3K27me3 and H2AK119Ub in primed and naive H9 hES cells. Percentages of cells displaying the representative pattern are indicated. Scale bar=10µm. e, Representative immunofluorescence images for H3K27me3 coupled with XIST RNA FISH in naive H9 hES cells. Percentages of cells displaying the representative pattern are indicated. Scale bar=10µm. f, Immunofluorescence images for H3K27me3 and NANOG in human preimplantation blastocysts (stages B3 to B5).

Source data

Coverage profiles showed a distinct distribution of H3K27me3, H2AK119Ub and H3K9me3 on the X chromosome in naive compared to primed hES cells (Fig. ​(Fig.4b).4b). Notably, H3K27me3 and H2AK119Ub covered around 60% of the X chromosome in naive hES cells and 40–50% in primed hES cells, while most of the X chromosome (>60%) was covered by H3K9me3 in primed hES cells versus 33% in naive hES cells (Fig. ​(Fig.4c).4c). H3K27me3 and H2AK119Ub patterns were highly correlated together and anticorrelated with H3K9me3 in both primed and naive hES cells (Extended Data Fig. ​Fig.4a).4a). H3K27me3 and H2AK119Ub signals accumulated into several smaller domains that spanned most of the XIST-coated dampened X chromosome in naive hES cells (Fig. ​(Fig.4b).4b). We observed similar profiles of H3K27me3 and H2AK119Ub using other published datasets of naive H9 hES cells (Extended Data Fig. 4b,c). In contrast, H3K27me3 and H2AK119Ub accumulated over several large domains of the Xi, with the major one spanning around 50Mb in the middle of the long arm, while the H3K9me3 signal spanned most of the short arm, as well as the most proximal and distal parts of the long arm (Fig. ​(Fig.4b).4b). This is reminiscent of what was described in previous studies^32,33. Altogether, this showed that X chromosomes of naive hES cells are enriched for repressive histone modifications, which are massively redistributed along the X chromosome upon transition from primed to naive hES cells.

Open in a separate window

Extended Data Fig. 4

Characterization of the X chromatin landscape in WT, XIST KO and XIST CRISPRi UT and Dox-treated clones in naïve hESCs.

a, Heatmap showing the Pearson correlation between X chromosome H3K27me3, H2AK119Ub, and H3K9me3 CUT&RUN replicates from naïve and primed H9. b, X chromosome H3K27me3 and H2AK119Ub profiles from H9 naïve cells reprogrammed using PXGL⁶² or t2iLGO⁶³. c, Scatter plots showing the Pearson correlation, for the X chromosome, between our CUT&RUN dataset and the other datasets. d, Scatter plots showing the Pearson correlation between RAP XIST and repressive histone modifications (H3K27me3, H2AK119Ub, H3K9me3) on the X chromosome in naïve and primed H9. e, CUT&RUN profiles (log2 enrichment over IgG) for H3K27me3, H2AK119Ub and H3K9me3 on the X chromosome in naïve WT, XIST KO and XIST CRISPRi H9 cells. f, Representative immunofluorescence images for H3K27me3 and H2AK119Ub in naïve H9 WT, XIST KO, XIST CRISPRi UT or treated with Dox. Percentages of cells with the representative pattern are indicated. Scale bar= 10µm. g, Quantification of immunofluorescence data showing the percentage of cells with H3K27me3 and H2AK119Ub foci in XIST KO and XIST CRISPRi naïve hESCs. h, H3K27me3 immunofluorescence of one human early blastocyst (B4). i, Quantification of immunofluorescence data from human embryos showing the percentage of cells with 0, 1, 2, or ≥3 H3K27me3 foci.

Source data

Patterns of H3K27me3 and H2AK119Ub in naive hES cells were highly correlated to that of XIST in naive as in primed hES cells, suggesting that the deposition of PRC-associated histone modifications on the X chromosome is mediated by XIST in both contexts (Extended Data Fig. ​Fig.4d).4d). We confirmed this hypothesis by showing a global loss of H3K27me3 and H2AK119Ub on the X chromosome in XIST KO and Dox-treated CRISPRi clones compared to their WT counterpart (Extended Data Fig. ​Fig.4e),4e), while H3K9me3 distribution and levels remained unchanged.

Immunofluorescence approaches confirmed the CUT&RUN data (Fig. ​(Fig.4d)4d) as one focus of H3K27me3 and H2AK119Ub was observed in >60% of naive H9 hES cells, with frequent co-occurrence of the two marks (Fig. ​(Fig.4d).4d). This H3K27me3 focal accumulation corresponded to the XIST-coated dampened X chromosome, as shown by immuno-RNA FISH (Fig. ​(Fig.4e).4e). In the absence of XIST, H3K27me3 and H2AK119Ub coaccumulation was lost on the X chromosome in more than 90% of cells (Extended Data Fig. 4f,g).

Because conflicting results existed regarding the accumulation of PRC-mediated repressive histone modifications on the X chromosomes before XCI establishment^10,11,26,34, we re-examined the accumulation of H3K27me3 in human preimplantation embryos. In four blastocyst-stage embryos, we could observe cells with one or two H3K27me3 foci, in both NANOG⁺ and NANOG⁻ cells (Fig. ​(Fig.4f4f and Extended Data Fig. 4h,i). The presence of two foci in a fraction of NANOG⁺ cells in three of four embryos strongly suggested that active X chromosomes were indeed enriched for H3K27me3 in a fraction of epiblast cells of female blastocysts. Embryo 4, in which only one H3K27me3 focal enrichment was detected in around 10% of cells, was likely male (Extended Data Fig. 4h,i). Overall, these data indicate that PRC-mediated repressive histone modifications are enriched on and broadly cover XIST-associated X chromosomes in most cells of female preimplantation embryos and naive hES cells.

Highly expressed genes escape XIST-mediated dampening

The broad distribution of XIST and repressive histone modifications along the dampened X chromosome led us to explore whether all X-linked genes are equally responsive to XIST-mediated regulation in naive hES cells, according to their allelic expression profiles. Most of the X-linked genes that could be analyzed using SNPs were ‘XIST-sensitive’ genes (47 of 60; 78%), in agreement with XIST covering most of the X chromosome in naive hES cells. These genes displayed lower expression from the XIST-associated X chromosome and became equally expressed from both X chromosomes in the absence of XIST (Fig. ​(Fig.5a5a and Supplementary Table 5). In addition, we could identify a fraction of ‘XIST-resistant’ genes (13 of 60; 22%), with a pXi allelic ratio of ~0.5 in the presence or absence of XIST, indicating equal expression from the dampened (pXi) and active (pXa) X chromosomes in naive hES cells (Fig. ​(Fig.5a5a and Supplementary Table 5).

Open in a separate window

Fig. 5

Highly expressed genes escape XIST-mediated dampening.

a, Heat map showing the pXi allelic ratio of XIST-sensitive and XIST-resistant genes in XIST WT, KO, UT CRISPRi (Dox⁻) and Dox-treated CRISPRi (Dox⁺) naive H9 cells. Constitutive and variable escapees from the Tukiainen study⁴³ are highlighted in light blue and brown, respectively. b, Distance to closest escapee of XIST-sensitive (n=47) and XIST-resistant (n=13) genes. c, Percentage of escapees (constitutive and variable) and inactivated X-linked genes from the Tukiainen study⁴³ among XIST-sensitive and XIST-resistant genes; chi-square test P value<0.01 (**). d, XIST RAPseq counts on TSS±5kb of XIST-sensitive (n=47) and XIST-resistant (n=13) genes in naive H9 cells. e, H3K27me3, H2AK119Ub and H3K9me3 counts on TSS±5kb of XIST-sensitive (n=47) and XIST-resistant (n=13) genes in the presence (XIST WT and UT CRISPRi) or absence (XIST KO and Dox-treated CRISPRi) of XIST in naive H9 hES cells. f, ATAC-seq, H3K27ac and H3K4me3 counts on TSS±5kb of XIST-sensitive (n=47) and XIST-resistant (n=13) genes in the presence (XIST WT and UT CRISPRi) or absence (XIST KO and Dox-treated CRISPRi) of XIST in naive H9 hES cells. g, Violin plots showing the log CPM expression level of XIST-sensitive (n=47) and XIST-resistant (n=13) genes in XIST WT, KO, UT CRISPRi and Dox-treated CRISPRi naive H9 cells, obtained from RNA-seq and fastGRO-seq data. All box plots represent the median (center), first and third quartiles (hinges) and ±1.5 IQR (whiskers). Unless otherwise specified, P values were calculated by a two-sided unpaired t-test. Levels of significance: NS (P≥0.05), *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

Source data

While no significant difference was observed in the distance to the XIST locus between XIST-resistant and XIST-sensitive genes (Extended Data Fig. ​Fig.5a),5a), XIST-resistant genes tended to be located in the vicinity of XCI escapees and were enriched on the short arm, while XIST-sensitive genes appeared to be scattered along the X chromosome (Fig. ​(Fig.5b5b and Extended Data Fig. ​Fig.5b).5b). Of note, the majority of XIST-resistant genes in naive H9 cells are known constitutive or variable escapees (Fig. ​(Fig.5c),5c), suggesting that this set of genes is refractory to XISTʼs repressive activities (dampening or inactivation), whatever the context.

Open in a separate window

Extended Data Fig. 5

Genetic and epigenetic features of XIST-sensitive and XIST-resistant genes.

a, Distance to XIST locus of XIST-sensitive (n=47) and resistant (n=13) genes. b, Chromosomal map indicating the positions of XIST-sensitive and resistant genes on the X chromosome. The centromere is indicated by a grey rectangle. c, Comparison of MeD-seq methylation scores in TSS (right) and gene body (left) between XIST-sensitive (n=47) and resistant (n=13) genes. d, Profile plots of H3K27me3, H2AK119Ub, and H3K9me3 over XIST-sensitive and resistant genes in XIST positive (WT, XIST CRISPRi UT) and XIST negative (XIST KO, Dox-treated XIST CRISPRi) naïve H9 cells. e, Expression levels of XIST-resistant (n=13) and sensitive (n=47) genes in human pre-implantation and post-implantation embryos, based from scRNA-seq^20,25. Purple and orange dotted lines indicate the median expression of XIST-resistant and XIST-sensitive genes, respectively. BLASTO: blastocyst; EPI: epiblast; PrE: primitive endoderm; TE: trophectoderm. Numbers correspond to the embryonic days. All boxplots represent the median (center) first and third quartile (hinges) and ±1.5×IQR (whiskers). Unless otherwise specified, p-values were calculated by two-sided unpaired t-test. Levels of significance: ns≥0.05; * < 0.05; ** < 0.01; *** < 0.001; **** < 0.0001.

Source data

XIST-sensitive genes were significantly more contacted by XIST RNA and had a higher accumulation of H3K27me3 and H2AK119Ub compared to XIST-resistant genes, while H3K9me3 and DNA methylation levels were similar in both categories (Fig. 5d,e and Extended Data Fig. ​Fig.5c).5c). Accordingly, only XIST-sensitive genes displayed reduced H3K27me3 and H2AK119Ub levels in the absence of XIST (XIST KO and Dox-treated CRISPRi), with levels reaching that of XIST-resistant genes, while H3K9me3 remained constant across gene categories and genotypes (Fig. ​(Fig.5e5e and Extended Data Fig. ​Fig.5d5d).

Conversely, the TSSs of XIST-sensitive genes were less accessible than those of XIST-resistant genes and were marked by lower levels of active histone modifications (H3K27ac and H3K4me3; Fig. ​Fig.5f).5f). H3K27ac but not H3K4me3 levels were elevated specifically at XIST-sensitive genes upon loss of XIST but remained lower compared to XIST-resistant genes. This indicates that, while XIST is responsible for the deposition of repressive histone marks at target genes, it only mildly prevents the acquisition of active promoters’ features. In agreement with these observations, XIST-sensitive genes were expressed at lower levels than XIST-resistant genes in naive hES cells and did not reach the expression levels of XIST-resistant genes in the absence of XIST (Fig. ​(Fig.5g).5g). Interestingly, a similar tendency was observed in preimplantation and postimplantation human embryos (Extended Data Fig. ​Fig.5e).5e). Altogether, these results show that, in naive hES cells and possibly in preimplantation embryos, XIST attenuates the expression of the majority of X-linked genes, with highly expressed genes resisting XIST targeting and dampening activity.

Human preimplantation embryos

The use of human embryos donated to research as surplus of in vitro fertilization (IVF) treatment was allowed by the French embryo research oversight committee (Agence de la Biomédecine) under approval number RE18-010R. Embryos used were initially created in the context of an assisted reproductive cycle with a clear reproductive aim and then voluntarily donated for research once the patients had fulfilled their reproductive needs or tested positive for the presence of monogenic diseases. Informed written consent was obtained from both parents of all couples that donated spare embryos following IVF treatment. French legislation does not include the research project in the consent form; the embryos are donated to research, in general. The Agence de la Biomédecine oversight committee rules which project can use which embryos. Donor compensation is forbidden under French law.

All human preimplantation embryos used in this study were obtained from and cultured in the Assisted Reproductive Technology unit of the University Hospital of Nantes, France, which is authorized to collect embryos for research under approval number AG110126AMP of the Agence de la Biomédecine. Molecular analysis of the embryos was performed in compliance with the embryo research oversight committee and the International Society for Stem Cell Research (ISSCR) guidelines⁴⁴.

Human embryos were thawed following the manufacturer’s instructions (Sydney IVF Thawing Kit for slow freezing, from Cook Medical; RapidWarmCleave or RapidWarmBlast for vitrification, from Vitrolife). Human embryos frozen at the eight-cell stage were loaded in a 12-well dish (Embryoslide, from Vitrolife) with nonsequential culture medium (G-TL, from Vitrolife) under mineral oil (liquid paraffin, from Origio), at 37°C in 5% O₂ and 6% CO₂.

Generation of XACT and XIST KO cell line

Single-guide RNA (sgRNA) sequences flanking the XACT and XIST promoters and first exon were obtained using the web-based tool CRISPOR (http://crispor.tefor.net/) and are provided in Supplementary Table 9. sgRNAs were cloned under a U6 promoter into the pSpCas9(BB)-2A-GFP (green fluorescent protein) (a gift from Feng Zhang; Addgene 48138) and the pSpCas9(BB)-2A-mCherry (generated in house, by replacing GFP with an mCherry reporter using the NEBuilding HiFi DNA Assembly Cloning Kit (New England Biolabs)⁴⁵. Using the Amaxa 4D-Nucleofector system (Lonza), one million primed H9 hES cells were transfected with 2.5µg of each plasmid (to a total of 5µg). Cells were sorted by fluorescence-activated cell sorting (INFLUX 500-BD, BioSciences) 48h after transfection. Double-positive cells were plated onto a Matrigel-coated 6-cm Petri dish in mTeSR medium supplemented with 1× CloneR (StemCell Technologies). Individual colonies were manually placed into 96-well plates ~10days after transfection. Deletions and inversion events were screened by PCR. Primer sequences can be found in Supplementary Table 10. XACT and XIST KO H9 primed hES cells were reprogrammed to naive using the NaiveCult and PXGL protocols, respectively, as described above.

CRISPR inhibition

A sgRNA targeting the XIST promoter was designed using the web-based tool CRISPOR (http://crispor.tefor.net/) and cloned into the PB_rtTA_BsmBI vector (a gift from Mauro Calabrese; Addgene 126028)⁴⁶. The sgRNA sequence can be found in Supplementary Table 9. Using the Amaxa 4D-Nucleofector system (Lonza), one million primed H9 hES cells were transfected with 1.5µg of PB_tre_dCas9_KRAB (a gift from Mauro Calabrese; Addgene 126030), 0.75µg of PB_rtTA_BsmBI and 1.5µg of piggyBac transposase⁴⁷. Cells were then treated with G418 (350µgml⁻¹) and hygromycin (350µgml⁻¹) until separate colonies were obtained. The number of random insertions in the genome was verified by qPCR and clones with the lowest insertion number (n=2) were used for further experiments. For XIST depletion, PXGL-reprogrammed CRIPSRi naive hES cells were treated with 1µgml⁻¹ Dox for 10days.

CUT&RUN

CUT&RUN was performed as previously described⁴⁸. Briefly, 0.3 million cells per replicate were bound to 20µl of concanavalin A-coated beads (Bangs Laboratories) in binding buffer (20mM HEPES, 10mM KCl, 1mM CaCl₂ and 1mM MnCl₂). The beads were washed and resuspended in DIG Wash buffer (20mM HEPES, 150mM NaCl, 0.5mM spermidine and 0.05% digitonin). The primary antibodies (1:50) were added to the bead slurry and rotated at room temperature for 1h. The beads were washed with DIG Wash buffer and protein A–MNase (micrococcal nuclease) fusion protein (1:400, produced by the Institut Curie Recombinant Protein Platform, 0.785mgml⁻¹) was added, before rotating at room temperature for 15min. After two washes, the beads were resuspended in 150µl of DIG Wash buffer and the MNase was activated with 2mM CaCl₂, before incubating for 30min at 0°C. MNase activity was terminated with 150µl of 2× Stop buffer (200mM NaCl, 20mM EDTA, 4mM EGTA, 50µgml⁻¹ RNase A and 40µgml⁻¹ glycogen). Cleaved DNA fragments were released by incubating for 20min at 37°C, followed by centrifugation for 5min at 16,000g at 4°C and collection of the supernatant from the beads on a magnetic rack. The DNA was purified by phenol–chloroform; libraries were prepared using the TruSeq ChIP Library Preparation Kit from Illumina following the manufacturer’s protocol and sequencing was performed on a NovaSeq 6000 instrument (ICGex next-generation sequencing (NGS) platform) to generate 2×100 paired-end reads. All the antibodies used in this study are listed in Supplementary Table 11.

RAPseq

RAPseq for human XIST was performed as previously described⁴⁰. The human XIST oligo pool was ordered from GenScript and amplified as previously described to generate single-stranded DNA (ssDNA) biotinylated oligos⁴⁹. The oligo sequences are listed in Supplementary Table 12. Briefly, 20 million cells were harvested and incubated with 10ml of freshly made 2mM DSG at room temperature for 45min. Cells were further cross-linked with 10m of 3% formaldehyde for 10min at 37°C and the reaction was stopped by adding glycine to a final concentration of 500mM. Cells were pelleted at 4°C and stored at 80°C. Nuclei were isolated in cold lysis buffer (20mM HEPES pH 7.5, 50mM KCl, 1.5mM MnCl₂, 1% IGEPAL CA-630 (NP-40), 0.4% sodium deoxycholate and 0.1% N-lauroylsarcosine). Chromatin was solubilized by sonication and fragmented by TURBO DNase digestion (Thermo Fisher Scientific). XIST pulldown was performed on 10 million cells using 1µg of Dynabeads MyOne Streptavidin C1 (Thermo Fisher Scientific) and 50pmol of biotinylated oligos. DNA was eluted by RNase H digestion and cross-linking was reversed by proteinase K digestion at 65 °C for 1h. Finally, DNA was purified using the GeneJET Gel Extraction kit (Thermo Fisher Scientific); libraries were prepared using the TruSeq ChIP Library Preparation Kit from Illumina and sequencing was performed on a NovaSeq 6000 instrument (ICGex NGS platform) to generate 2×100 paired-end reads.

Assay for transposase-accessible chromatin with sequencing (ATAC-seq)

ATAC-seq was performed as previously described⁵⁰. Briefly, 50,000 cells were resuspended in 50µl of cold lysis buffer (10mM Tris–HCl pH 7.4, 10mM NaCl, 3mM MgCl₂ and 0.1% IGEPAL CA-630) and centrifuged for 10min at 500g at 4°C. The nuclear pellets were resuspended in 50μl of transposase reaction mix (25µl of 2× TD buffer, 2.5µl of transposase and 22.5µl of H₂O) and incubated at 37°C for 30min in a ThermoMixer at 1,000 r.p.m. Reactions were cleaned up using the MinElute PCR purification kit (QIAGEN) and DNA was eluted in 10µl of elution buffer. Transposed DNA was preamplified for five cycles in 50µl of reaction mix (2.5μl of 25μM primer Ad1, 2.5μl of 25μM primer Ad2, 25μl of 2× Master Mix, 10µl of H₂O and 10μl of transposed elution) with the following cycling conditions: 5min at 72°C, 30s at 98°C and five cycles of 10s at 98°C, 30s at 63°C and 1min at 72°C. Then, 15μl of qPCR amplification reaction (5μl of preamplified sample, 0.5μl of 25μM primer Ad1, 0.5μl of 25μM primer Ad2, 5μl of 2× NEBNext Master Mix, 0.24μl of 25× SYBR Green in DMSO and 3.76μl of H₂O) was carried out with the following cycling conditions: 30s at 98°C and 20 cycles of 10s at 98°C, 30s at 63°C and 1min at 72°C. The required number of additional cycles for each sample was calculated. After the final amplification, double-sided bead purification was performed with AMPure XP beads. Final ATAC-seq libraries were eluted in 20μl of nuclease-free H₂O from the beads and were sequenced on a NovaSeq 6000 instrument (Novogene) to generate 2×150 paired-end reads.

RNA-seq

Total RNAs were collected using an RNeasy Mini Kit (QIAGEN) and extracted following the manufacturer’s recommendation. Libraries were generated using the Illumina Stranded Total RNA Prep Ligation with the Ribo-Zero Plus kit according to the manufacturer’s recommendation and sequenced on a NovaSeq 6000 instrument (ICGex NGS platform) to generate 2×100 paired-end reads.

fastGRO-seq

Low-input fastGRO-seq was performed as previously described²⁸. Nuclei were isolated from five million cells and a nuclear run-on assay was performed by adding 25µl of 2× nuclear run-on buffer (10mM Tris–HCl pH 8, 5mM MgCl₂, 300mM KCl, 1mM DTT, 500μM ATP, 500μM GTP, 500μM 4-thio-UTP, 2μM CTP, 200 μgml⁻¹ SUPERase-In and 1% Sarkosyl) at 30°C for 7min. RNA was extracted with TRIzol LS reagent (Invitrogen) and quantified using the NanoDrop 2000. A total of 30µg of RNA was fragmented by sonication and the efficiency was analyzed using the Agilent High-Sensitivity RNA ScreenTape Assay. Fragmented RNA was incubated in biotinylation solution (25mM HEPES pH 7.4, 10mM EDTA pH 8.0 and 50µg of MTS-biotin (Biotium)) for 30min in the dark at 24°C and 800r.p.m. RNA was then precipitated using ethanol and resuspended in nuclease-free water. After DNase treatment, biotinylated RNA was enriched using M280 Streptavidin Dynabeads (Invitrogen) and precipitated using ethanol. Libraries were prepared using the Illumina Stranded Total RNA Prep Ligation with the Ribo-Zero Plus kit according to the manufacturer’s recommendation and sequencing was performed on a NovaSeq 6000 instrument (ICGex NGS platform) to generate 2×100 paired-end reads.

MeD-seq

MeD-seq assays were essentially performed as previously described⁵¹. Briefly, 10µl of genomic DNA (input 90ng) from naive hES cells was digested with LpnPI (New England Biolabs) generating 32-bp fragments around the fully methylated recognition site containing a CpG. These short DNA fragments were further processed using the ThruPLEX DNA-seq 96D Kit (Rubicon Genomics). Stem-loop adaptors were blunt-end ligated to repaired input DNA and amplified to include dual-indexed barcodes using a high-fidelity polymerase to generate an indexed Illumina NGS library. The amplified end product was purified on a Pippin HT system with 3% agarose gel cassettes (Sage Science). Multiplexed samples were sequenced on Illumina NextSeq 2000 systems for single-end reads of 50bp according to the manufacturer’s instructions. Dual-indexed samples were demultiplexed using bcl2fastq software (Illumina).

RNA FISH

Cell preparation: Primed hES cells were grown on coverslips. Naive hES cells were centrifuged onto Superfrost Plus slides (VWR) using the Cytospin 3 Cytocentrifuge (Shandon). The cells were fixed for 10min in a 3% paraformaldehyde solution (Electron Microscopy Science) and permeabilized for 5–10min in ice-cold cytoskeleton (CSK) buffer (10mM PIPES, 300mM sucrose, 100mM NaCl and 3mM MgCl₂ pH 6.8) supplemented with 0.5% Triton X-100 (Sigma-Aldrich) and 2mM vanadyl ribonucleoside complex (VRC; New England Biolabs).

Probe preparation: RNA FISH probes were obtained after Nick translation of fosmids and BAC constructs purified using phenol–chloroform. Then, 1μg of purified DNA was labeled for 3h at 15°C with fluorescent dUTPs (SpectrumOrange and SpectrumGreen from Abott Molecular and Cy5-UTPs from GE HealthCare Life Science).The templates used in this study were as follows human XIST fosmid (BacPac Resources Center, WI2-3059D20), human POLA1 BAC (BacPac Resources Center, RP11-11104L9), human XACT BAC (BacPac Resources Center, RP11135D3) and human HUWE1 BAC (BacPac Resources Center, RP11-975N19).

Hybridization: First, 100ng of probes were supplemented with 1μg of Cot-I DNA (Invitrogen) and 3μg of Sheared Salmon Sperm DNA (Invitrogen). After precipitation, the probes were resuspended in deionized formamide (Sigma-Aldrich), denatured for 7min at 75°C and further incubated for 10min at 37°C. Probes were mixed with an equal volume of 2× hybridization buffer (4× SSC, 20% dextran sulfate, 2mgml⁻¹ BSA and 2mM VRC). Coverslips were dehydrated in 80–100% ethanol washes and incubated with the hybridization mix at 37°C overnight in a humid chamber. Next, the coverslips were washed for 4min at 42°C three times with 50% formaldehyde and 2× SSC (pH 7.2) and three times with 2× SSC. The coverslips were mounted in VECTASHIELD PLUS DAPI (Vector Laboratories).

Immunofluorescence staining

Primed and naive hES cells were prepared as described above and fixed for 10min in a 3% paraformaldehyde solution (Electron Microscopy Science). Cells were then permeabilized for 7min with ice-cold PBS supplemented with 0.5% Triton X-100 (Sigma-Aldrich). Cells were blocked in 1× PBS with 1% BSA (Sigma-Aldrich) for 15min at room temperature and incubated for 45min at room temperature with primary antibody diluted in 1× PBS with 1% BSA. After three PBS washes, cells were incubated for 40min at room temperature with Alexa Fluor 488 anti-rabbit or Alexa Fluor 594 anti-mouse secondary antibodies (Life Technologies). Finally, coverslips were mounted in VECTASHIELD PLUS DAPI (Vector Laboratories). For combined Immunofluorescence and RNA FISH, immunofluorescence staining was first performed and then RNA FISH was performed as described above.

Human embryos were fixed at the B3, B4 or B5 stages according to the grading system proposed by Gardner and Schoolcraft⁵². Embryos were fixed with 4% paraformaldehyde (Electron Microscopy Sciences) for 5min at room temperature and washed in 1× PBS with 0.1% BSA. Embryos were permeabilized and blocked in 1× PBS with 0.2% Triton and 10% FBS at room temperature for 60min. Samples were incubated overnight at 4°C with primary antibodies. After three PBS washes, embryos were incubated for 2h at room temperature with Alexa Fluor 488 anti-mouse or Alexa Fluor 568 anti-mouse and Alexa Fluor 647 anti-goat secondary antibodies (Life Technologies) along with DAPI counterstaining (Invitrogen). All primary antibodies used in this study are listed in Supplementary Table 11.

Microscopy and image analysis

Fluorescent microscopy images for hES cells were taken on a fluorescence DMI-6000 inverted microscope with a motorized stage (Leica), equipped with a charge-coupled device (CCD) HQ2 camera (Roper Scientifics) and an HCX PL APO ×64–100 oil objective (numerical aperture, 1.4; Leica) using the MetaMorph software (version 7.04, Roper Scientifics). Approximately 40 optical z-sections were collected at 0.3-μm steps at different wavelengths depending on the signal (DAPI, 360–470nm; FITC, 470–525nm; Cy3, 550–570nm; Cy5, 647–668nm). We computed the dispersion of XIST RNA by comparing the cumulative volume of the signal to the theoretical spherical volume it could occupy on the basis of the maximal radial distance. Embryo confocal immunofluorescence images were acquired with an A1-SIM Nikon confocal microscope and a ×20 oil objective. Whole embryos were imaged with 0.5–1-μm optical sections. Stacks were processed using ImageJ and are represented as a two-dimensional ‘maximum projection’ throughout the manuscript.

Total RNA extraction and RT–qPCR

Total RNAs were collected using the RNeasy Mini Kit (QIAGEN) and extracted following the manufacturer’s recommendation. Quantification of the extracted RNAs was performed using the NanoDrop 2000. The DNase step was performed on 1μg of RNA for 30min at 37°C using TURBO DNase (Thermo Fisher Scientific). RNAs were reverse-transcribed using the SuperScript IV kit (Thermo Fisher Scientific) following the manufacturer’s recommendation. Complementary DNAs (cDNAs) were diluted 2.5 times in water and the RNA expression level was assessed by RT–qPCR using the Power SYBR Green Master Mix (Thermo Fisher Scientific) and ViiA-7 Real-Time PCR system (Applied Biosystems). Transcript RNA levels were normalized against the β-actin reference gene following the 2^−ΔΔCt method. All RT–qPCR primers used in this study are listed in Supplementary Table 13.

RIP

The cellular extract from 1 million naive H9 hES cells was prepared by adding 1ml of HNTG buffer (20mM HEPES pH 7.9, 50mM NaCl, 1% Triton X-100, 1mM MgCl₂, 1mM EDTA and 10% glycerol) followed by incubation on ice for 25min. Cells were then centrifuged at 16,000g at 4°C for 25min. The supernatant was collected and 80µl was used as the input fraction. Next, 30μl of Magna ChIP Protein G Magnetic Beads (Merck) per immunoprecipitation were washed with PBS–Tween 0.1% three times. Then, 2.5μg of SPEN antibody (Abcam, AB72266) was added to the beads, followed by 10-min incubation at room temperature with rotation. Beads were washed three times with PBS–Tween 0.1% and resuspended with HNTG buffer. Next, 30μl of beads were added to the lysate prepared above and incubated for 1.5h at 4°C with rotation. After immunoprecipitation, beads were washed three times with 500μl of HNTG buffer. For RNA preparation, input and immunoprecipitation samples were treated with 200μg of proteinase K in a total volume of 200μl for 45min at 65°C. RNA was purified using the RNeasy MinElute Cleanup Kit (Qiagen) according to the manufacturer’s recommendations in a final volume of 14μl of water. The DNase step and RT were performed as described above. The RNA enrichment level was assessed by RT–qPCR using the Power SYBR Green Master Mix (Thermo Fisher Scientific) and ViiA-7 Real-Time PCR system (Applied Biosystems).

siRNA KD SPEN

SPEN KD was achieved using two different mixes of siRNAs. Mix 1 contained one siRNA (CliniSciences, CRH7929) and mix 2 contained two siRNAs (Thermo Fisher Scientific, 1299001 and 4392420). As a negative control, scramble siRNA was used (ThermoFisher Scientific, AM4611). Naive H9 cells were transfected using the lipofectamine RNAiMAX transfection reagent (ThermoFisher Scientific, 13778030) according to the manufacturer’s recommendations. PXGL medium was changed 1h before transfection. After 24h, another round of transfection was performed. After 48h in total, cells were collected and RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s recommendation.

Raw data processing

RNA-seq, fastGRO-seq, CUT&RUN and RAPseq reads were trimmed using trim_galore (version 0.6.5) (https://github.com/FelixKrueger/TrimGalore) with a minimum length of 50bp. Reads were then mapped to the human genome (hg38) and mouse genome (mm10) using Bowtie 2 (version 2.4.4)⁵³ with the following parameters: ‘--local --very-sensitive-local --no-unal --no-mixed --no-discordant --phred33 -L 10 -X 700’. Reads were then deduplicated using MarkDuplicates from Picard (version 2.23.5) (http://broadinstitute.github.io/picard/) with the following options: ‘--CREATE_INDEX=true --VALIDATION_STRINGENCY=SILENT --REMOVE_DUPLICATES=true --ASSUME_SORTED=true’. BAM files were sorted, filtered (minimum mapping quality=10) and indexed with SAMtools (version 1.13)⁵⁴. Reads from MEFs were discarded using the XenofilteR package in R (version 4.1.1)⁵⁵. Replicate reproducibility was assessed by Pearson’s correlation of the signal using deepTools plotCorrelation.

RNA-seq and fastGRO-seq data analysis

For gene expression analysis, read counts were quantified for each gene with htseq-count from htseq (version 0.13.5)⁵⁶ using the following options: ‘--stranded reverse -a 10 -t exon -i gene_id -m intersection-nonempty’. The annotation file used was Homo_sapiens.GRCh38.90.gtf. Reads marked by special counters (no feature, ambiguous, too low aQual, not aligned or alignment not unique) were eliminated. For XIST isoform reconstruction, Scallop (version 0.10.4) was used with the default parameters⁵⁷. XIST isoform quantification was performed using kallisto (version 0.46.2) on the trimmed FASTQ file with the option ‘--rf-stranded’ (ref. ⁵⁸). For that purpose, the transcriptome FASTA file of exons and introns was generated from the hg38 reference genome using the BUSpaRse package in R (version 4.1.1) and used to produce a kallisto index.

Single-cell expression data analysis

FASTQ files from refs. ^20,25 were obtained from the ENA (European Nucleotide Archive), under accession numbers PRJEB11202 and PRJNA431392, respectively. STAR (version 2.7.9a) was used for read alignment to the GRCh38 reference human genome. The STAR index was generated using the following options: ‘--runMode genomeGenerate --runThreadN 30 --genomeFastaFiles HS.GRCh38.fa --limitGenomeGenerateRAM 168632718037 --outFileNamePrefix STAR_INDEX. --sjdbGTFfile Homo_sapiens.GRCh38.90.gtf --genomeDir STAR_INDEX_petropoulos_2016 --sjdbOverhang 42’ for Petropoulos data and ‘--genomeDir STAR_INDEX --sjdbOverhang 149’ for Zhou data.

For Petropoulos data, reads (single-end) were aligned using the following options: ‘--soloType SmartSeq --soloFeatures Gene --runThreadN 12 --genomeDir STAR_INDEX_petropoulos_2016 --readFilesCommand zcat --readFilesIn sample_A.fastq.gz --soloStrand Unstranded --soloUMIdedup NoDedup --outFileNamePrefix petro_2016/STAR_ALIGN_sample_A.STARSOLO --outSAMtype BAM SortedByCoordinate --outSAMattrRGline ID:sample_A’.

For Zhou data reads (paired-end) were aligned using the following options: ‘--soloType CB_UMI_Simple --soloCBwhitelist zhou_2019/whitelist.txt --soloFeatures Gene Velocyto --runThreadN 5 --genomeDir STAR_INDEX --readFilesCommand zcat --readFilesIn multiplex_sample_A_R1.fastq.gz multiplex_sample_A_R2.fastq.gz --outFileNamePrefix zhou_2019/STAR_ALIGN_multiplex_sample_A.STARSOLO --outSAMtype BAM SortedByCoordinate --outSAMattributes NH HI nM AS CR UR CB UB GX GN sS sQ sM --soloCBstart 1 --soloCBlen 8 --soloUMIstart 9 --soloUMIlen 7 --soloCellFilter None --soloBarcodeReadLength 0’.

For female embryos, single cells were grouped by cell type and developmental day, according to the published sample annotation^20,25 refined by clustering on the UMAP (uniform manifold approximation and projection) of batch-corrected data (‘3_buildModel.R’ in https://gitlab.univ-nantes.fr/E137833T/Castel_et_al_2020/-/blob/master/Code/). The sum of raw counts and counts per million (CPM) normalization were computed for each group. The median gene expression of the X chromosome and autosomes was calculated after removing nondetected genes.

SNP calling

We used a whole-genome sequencing dataset of H9 hES cells to identify informative genomic SNPs along the X chromosome that can be used for allelic expression analysis (GSM1227088). Reads were aligned using Bowtie 2 and PCR duplicates were filtered out using MarkDuplicates from Picard (version 2.22.0) as previously described. For proper tag formatting of BAM files, AddorReplaceReadGroups from Picard was used with the following options: ‘SO=coordinate, RGID=id, RGLB=library, RGPL=platform, RGPU=machine, RGSM=sample’. BAM files were further processed using GATK (version 4.1.2.0)⁵⁹ for SNP identification. BaseRecalibrator and ApplyBQSR were used to generate the recalibrated BAM file using high-confidence SNPs referenced in Mills_and_1000G_gold_standard.indels.hg38.vcf.gz (annotation of indels) and dbsnp_146.hg38.vcf.gz (annotation of SNPs). The Variant Call Format (VCF) file H9_WGS_hg38_filtered.vcf was produced with HaplotypeCaller (options: ‘--dont-use-soft-clipped-bases and -stand-call-conf 10.0’) and filtered using VariantFiltration (options: ‘-window 50 -cluster 3 --filter-name FS -filter FS>30.0 --filter-name QD -filter QD<2.0’). The SNP position file info_snp_heterozygous_h9_wgs_chrX.txt was produced from VCF in R (version 4.1.1), using minimum thresholds of 50 for quality score, 10 for read coverage and 0.25 for allelic ratio to define heterozygous positions. We then used mpileup from SAMtools (version 1.15.1) to produce pileup files from RNA-seq and fastGRO-seq BAM files with the following options: ‘--output-MQ --no-output-ins --no-output-ins --no-output-del --no-output-del --no-output-ends’. The putative Xa haplotype was generated in R (version 4.1.1) from monoallelically expressed SNPs from the X chromosome in H9 primed hES cells. This was used to determine the pXi:Xa allelic ratio across all samples. We considered a transcript as biallelic when at least 25% of reads originated from the second allele.

XIST-sensitive and XIST-resistant gene calling

We applied stringent criteria for defining genes sensitive or resistant to XIST in naive cells. First, we determined the pXi:Xa allelic ratio for all detected SNPs, as described above, in XIST KO and CRISPRi datasets. We then computed the FC of the pXi:Xa allelic ratio between the XIST KO and WT and the XIST Dox-treated and UT CRISPRi datasets. An X-linked gene was defined as sensitive to XIST when the FC was greater than 1.2 in both datasets.

CUT&RUN data analysis

BigWig track files were generated with bamCoverage from deepTools (version 3.5.0) using the following parameters: ‘--normalizeUsing BPM --binSize 20 --smoothLength 40’. Coverage score matrices were generated with computeMatrix from deepTools (version 3.5.0) using either reference-point (‘--referencePoint TSS -R Homo_sapiens.GRCh38.90.gtf --beforeRegionStartLength 1000 --afterRegionStartLength 1000 --binSize 20 --missingDataAsZero’) or scale-regions (‘-R Homo_sapiens.GRCh38.90.gtf --regionBodyLength 20000 --startLabel TSS --endLabel TES --binSize 20 --missingDataAsZero’). For histone modification enrichment, raw counts were generated using multiBigwigSummary from deepTools with ‘--binSize 10000’ and the mean counts per chromosome were calculated. The percentage occupancy was determined by calculating for each chromosome the cumulative coverage over the chromosome size. Regions with a minimum threshold of 1.2 (FC over IgG) were considered covered.

ATAC-seq data analysis

ATAC-seq raw data were analyzed using the atacseq pipeline (version 1.2.1) (https://github.com/nf-core/atacseq/) from nf-core⁶⁰ using the default parameters. Briefly, reads were aligned to the human genome (hg38) using BWA (version 0.7.17). Duplicates were marked by Picard and reads mapping to mitochondrial DNA and blacklisted regions were removed. BigWig files were generated using deepTools as described above.

MeD-seq data analysis

Data processing was carried out using specifically created scripts in Python. Raw FASTQ files were subjected to Illumina adaptor trimming and reads were filtered on the basis of LpnPI restriction site occurrence 13–17bp from either the 5′ or 3′ end of the read. Reads that passed the filter were mapped to hg38 using Bowtie 2. Genome-wide individual LpnPI site scores were used to generate read count scores for the following annotated regions: TSSs (1kb before and 1kb after), CpG islands and gene bodies (1kb after TSS until TES (transcription end site)) and normalized (reads per million mapped reads (RPM)) using the total number of CpG reads after the filter. Gene and CpG island annotations were downloaded from ENSEMBL (www.ensembl.org).

RAPseq data analysis

BigWig tracks were generated as previously described. RAPseq naive versus primed log₂ enrichment was calculated using bigwigCompare from deepTools with the default parameters. The percentage occupancy was determined by calculating for each chromosome the cumulative coverage over the chromosome size. Regions with a minimum threshold of 10 (FC over input) were considered covered. RAPseq coverage scores were calculated with multiBigwigSummary, using the bins option for gene density enrichment and BED-file option for TEs (annotation from RepeatMasker).

Statistics and reproducibility

Statistical analyses were performed in R (version 4.1.1). The statistical analysis method for each experiment is specified in the corresponding figure legend. P values<0.05 were considered statistically significant. Unless otherwise mentioned, most of the data shown are either representative of three or more independent experiments or combined from three or more biologically independent samples and analyzed as the mean±s.d.

We thank lab members for critical evaluation of the work leading to this publication and C. Morey and S. Ait-Si-Ali for critical reading of the manuscript. We thank the epigenomic, microscopy, vectorology and BIBS platforms, all hosted in UMR7216 Epigenetic and Cell Fate, for technical advice and access to instruments. We thank the ICGex NGS platform of the Institut Curie supported by the grants ANR-10-EQPX-03 (Equipex) and ANR-10-INBS-09-08 (France Génomique Consortium) from the Agence Nationale de la Recherche (‘Investissements d’Avenir’ program), the ITMO-Cancer Aviesan (Plan Cancer III) and the SiRIC-Curie program (SiRIC Grant INCa-DGOS-465 and INCa-DGOSInserm_12554) for the high-throughput sequencing. We thank the bioinformatics platform of the Institut Curie for data management, quality control and primary analysis. This project was funded by the Agence Nationale pour la Recherche (ANR-14-CE10-0017 to C.R.), European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (101020423 to C.R.), LabEx ‘Who Am I?’ (ANR-11-LABX-0071 to C.R.) and Université de Paris IdEx (ANR-18-IDEX-0001 to C.R.) funded by the French Government through its ‘Investments for the Future’ program. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.