Chromosome compartments on the inactive X guide TAD formation independently of transcription during X-reactivation

Abstract

A hallmark of chromosome organization is the partition into transcriptionally active A and repressed B compartments, and into topologically associating domains (TADs). Both structures were regarded to be absent from the inactive mouse X chromosome, but to be re-established with transcriptional reactivation and chromatin opening during X-reactivation. Here, we combine a tailor-made mouse iPSC reprogramming system and high-resolution Hi-C to produce a time course combining gene reactivation, chromatin opening and chromosome topology during X-reactivation. Contrary to previous observations, we observe A/B-like compartments on the inactive X harbouring multiple subcompartments. While partial X-reactivation initiates within a compartment rich in X-inactivation escapees, it then occurs rapidly along the chromosome, concomitant with downregulation of Xist. Importantly, we find that TAD formation precedes transcription and initiates from Xist-poor compartments. Here, we show that TAD formation and transcriptional reactivation are causally independent during X-reactivation while establishing Xist as a common denominator.

Fig. 1. A tailor-made reprogramming system to efficiently trace X-chromosome reactivation.

a Schematic representation of the PaX reprogramming system. b X-reactivation efficiency of indicated reprogramming intermediates isolated on day 5 and then reprogrammed for an additional 4 days. Shown are representative histograms gated on SSEA1+ cells. Numbers indicate the percentages of X-GFP+ and X-GFP– cells. c PCA of gene expression dynamics during reprogramming (n = 12,318 genes). Solid grey arrow, hypothetical trajectory. Dashed grey arrow, samples deviating from main reprogramming trajectory. d Allelic expression ratio (mus/(mus+cas)) of protein-coding genes expressed from chromosome X (n = 335). For biallelic expression, ratio = 0.5. Box plots depict the first and third quartiles as the lower and upper bounds of the box, with a band inside the box showing the median value and whiskers representing 1.5x the interquartile range. e Average gene expression kinetics of naive pluripotency genes Esrrb, Nanog, Prdm14, and Zfp42 (relative to the levels in iPSC) and Xist^mus (relative to the levels in NPC) during reprogramming (n = 2). f Average endogenous gene expression kinetics of the reprogramming factor genes Klf4, Oct4, and Sox2 during reprogramming (n = 2, relative to the levels in iPSC). Endogenous expression assessed via the genes’ 3’-UTR.

Fig. 2. The inactive X chromosome exhibits A/B-like compartmentalization.

a Allele-specific Hi-C maps of chromosome X^mus at the inactive state in NPCs (left), intermediate state during reprogramming in D5 P-RFP+ cells (middle) and in the active state in ESCs (right). Top: Entire chromosome is shown at 200-kb resolution. Bottom: Zoom-in of the mega-domain boundary is shown at 100-kb resolution. Scale is shown in mega-bases (Mb). The mega-domain boundary Dxz4 is indicated by a red arrow. White-shaded areas, unmappable regions. b A/B compartments of chromosome X at 100-kb resolution obtained with principal component analysis of matrices split at the Dxz4 mega-domain boundary. Positive PC1 values represent A-like compartments (red); negative PC1 values represent B-like compartments (blue). Top: when matrices are not split at the Dxz4 mega-domain boundary, then the PC1 corresponds to the two mega-domains for the inactive X chromosome. c Clustering of PC1 values to compare A/B-like compartmentalization of the X^mus and X^cas at different stages using UMAP (uniform manifold approximation and projection) (n = 1,406 bins). d Mega-domain strength on X^mus depicted by the PC1 of the Hi-C matrices without splitting at Dxz4. Lines show smoothed mean from a fitted loess curve with span 0.25. Shading denotes 95% confidence interval. Dotted line indicates position of Dxz4. e Saddle plots showing the interactions within (AA, BB) and between (AB, BA) compartments (small numbers in the corners) of chromosome X. Data are presented as the log2 ratio of observed versus expected aggregated contacts between bins of discretized eigenvalues (50 categories, bin size = 100 kb). Overall compartmentalization strengths for X^mus and X^cas at different stages are shown as large numbers in the centre.

Fig. 3. Subcompartmentalization of the inactive X chromosome.

a Identification of spatial clusters (numbers 1 to 12 on the axes) and their associated subcompartments (text in colour next to cluster labels) on the inactive X using k-means clustering on a balanced matrix of chromosome X^mus D5 P-RFP+ at 50-kb resolution. Red areas interact more while blue areas interact less. b Allele-specific Hi-C map of chromosome X^mus in NPCs at 100-kb resolution. Scale is shown in mega-bases (Mb). Mega-domain boundary Dxz4 is indicated by a red arrow. White-shaded areas, unmappable regions. Position of spatial clusters is shown below. Position of ATAC peaks in NPC X^mus is shown in black, genes escaping X-inactivation in NPCs are shown in green. Xist RNA binding pattern in NPCs (CHART-seq, composite scaled tracks) taken from ref. . c Distribution of PC1 values in NPC X^mus of the spatial clusters. Box plots depict the first and third quartiles as the lower and upper bounds of the box, with a band inside the box showing the median value and whiskers representing 1.5x the interquartile range. n is given in brackets and indicates number of 50 kb bins. d Polar chart showing the coverage of the subcompartments on chromosome X (fraction of linear sequence occupied by each subcompartment). e Gene density of subcompartments as number of genes per 50 kb bin. Sample sizes are given in brackets. f Xist RNA enrichment of subcompartments in NPCs (composite scaled data). CHART-seq data from ref. . g H3K27me3 enrichment of subcompartments in NPC^mus. ChIP-seq data from ref. . h CBX1 enrichment of subcompartments in NPC^mus. DamID data from ref. . e–h The numbers above the bars indicate p-values (two-sample unpaired Wilcoxon-Mann-Whitney test with R defaults). Box plots depict the first and third quartiles as the lower and upper bounds of the box, with a band inside the box showing the median value and whiskers representing 1.5x the interquartile range. n is given in brackets and indicates number of 50 kb bins. i Network of spatial clusters on chromosome X^mus in NPCs obtained by applying the ForceAtlas2 algorithm to Hi-C interaction patterns of spatial clusters. Each cluster represents a single node of the network. Line width correlates with interaction strength. j Inter-mega-domain interactions of clusters (across the mega-domain boundary) in NPC^mus. The numbers above the bars indicate p-values (unpaired two-samples t-test with R defaults). n = 2 biologically independent replicates.

Fig. 4. Initiation of chromatin opening and gene expression from a distinct 3D cluster.

a Allelic expression ratio (= mus/(mus+cas)) of X-linked genes in spatial clusters. Cutoff >0.14 defines escapees (Supplementary Fig. 4a). For biallelic expression, ratio = 0.5. Only protein-coding genes with sufficient allelic information and expression for chromosome X^cas are counted (see methods). b Chromatin accessibility of each spatial cluster in NPCs shown as number of ATAC peaks per 1 Mb. c Dynamics of chromatin opening of spatial clusters. Only new peaks differential from NPCs were used. Relative differential ATAC peaks were then obtained by dividing the sum of peaks of each cluster at a given time point, by the sum of peaks in iPSC. Therefore NPCs will have a value of 0 and iPSCs a value of 1. Zoom-in shows early chromatin opening from NPCs until D6 P-RFP+. d ATAC-seq profiles of chromatin opening at two representative X-linked regions of 4 Mb. Position of ATAC peaks is shown in black (except for NPCs, differential new peaks compared to NPCs are shown). Genes either escaping X-inactivation in NPCs or being reactivated based on RNA expression are shown in green. Position of spatial clusters is shown at the bottom. e Dynamics of gene reactivation of gene-rich A-like clusters. Fractions of reactivated genes per cluster are shown. 0, no reactivated gene. 1, all genes reactivated. Threshold for gene reactivation, allelic expression ratio >0.14. f Violin plots showing the linear distance of genes to the closest escapee. Distances were calculated between the transcriptional start sites (TSS). Numbers of genes per cluster are given at the bottom. g Violin plots showing the linear distance to the closest escapee for genes of cluster 5 reactivating early, at D4 P-RFP+, compared to genes reactivating after that (“main reactivation”). Distances were calculated between the TSS. h Violin plots showing the promoter accessibility of genes of cluster 5. ATAC signal in a window of ±2 kb around the TSS was summed. g, h The numbers above the plot indicate p-values (two-sample unpaired Wilcoxon-Mann-Whitney test with R defaults).

Fig. 5. Remodelling of the X-inactivation centre leading to Xist downregulation.

a ATAC-seq profiles of chromatin opening on X^mus at a region encompassing the Tsix and Xist genes (mm10; 103,416,500 bp–103,490,000 bp). Position of ATAC peaks is shown in black (except for NPCs, differential peaks compared to NPCs are shown). b Chromatin accessibility on X^mus at Xist promoter 1 (mm10; 103,482,600 bp–103,483,800 bp) and Xist intron 1 (mm10; 103,470,900 bp–103,474,200 bp) as depicted in a. c Insulation score at 10-kb resolution at a region encompassing Tsix TAD-D (mm10; 103.18 Mb–103.45 Mb) and Xist TAD-E (mm10; 103.47 Mb–104.0 Mb). Dotted lines show TAD borders. Only genes with implicated roles in X-inactivation or X-reactivation are shown. d Allele-specific Hi-C maps of chromosome X^mus at 10-kb resolution at a region encompassing TAD-D (left) and TAD-E (right) in NPCs, D5 cells and ESCs. e Differential allele-specific Hi-C maps between D5 and NPCs (left), and ESCs and D5 cells (right) of the region shown in d. d, e Dotted lines show TAD borders and additionally separate TAD-E in two regions at the TSS of Ftx (mm10; 103.62 Mb) for quantification in f. f Sum of intra-domain interactions of X^mus are shown. TAD-E was separated in two regions at the TSS of Ftx. n = 2 biologically independent replicates. g Expression of Jpx^mus and Xist^mus relative to the levels in NPCs. R and p-values calculated by Pearson’s correlation are shown. n = 2 biologically independent replicates. h NPCs at day 9 of differentiation were nucleofected with LNA GapmeRs and expression of Jpx and Xist analysed by quantitative RT-PCR one day later normalized to Gapdh. The numbers above the bars indicate p-values (unpaired two-samples t-test with R defaults). Error bars denote SEM. Ctrl., control LNA. n = 4 biologically independent replicates. Source data are provided as a Source Data file.

Fig. 6. Structural changes during X-reactivation in the absence of chromatin opening and transcription.

a Two representative X-linked regions of 10 Mb for early TAD formation are shown. Allele-specific Hi-C map of chromosome X^mus at 20-kb resolution. Scale is shown in mega-bases (Mb). Insulation scores at 50-kb resolution, dashed line indicates cut-off for TAD borders at -0.086. ATAC-seq profiles with ATAC peaks shown in black (except for NPCs, differential peaks compared to NPCs are shown). Genes either escaping X-inactivation in NPCs or being reactivated based on RNA expression are shown in green. Position of subcompartments is shown at the bottom. b Meta region plot of insulation score at TAD boundaries at each time-point. Lines show mean. n = 116 TAD boundaries. c Comparison of insulation scores for chromosome X^mus. Interquartile range of insulation scores is shown on the right. n = 2,785 50 kb bins. d Comparison of domain scores for chromosome X^mus. n = 100 TADs. e Degree of change in domain score of X^mus of subcompartments on D5. The relative domain score at D5 = (D5-NPC)/(ESC-NPC). n = 100 TADs. c–e Box plots depict the first and third quartiles as the lower and upper bounds of the box, with a band inside the box showing the median value and whiskers representing 1.5x the interquartile range. d, e The numbers above the lines indicate p-values (two-sample unpaired Wilcoxon–Mann–Whitney test with R defaults). f Correlation between Xist RNA CHART-seq enrichment in NPCs and the relative domain score at D5 is shown. Points represent TADs. Colours of points indicate subcompartments. R and p-values calculated by Pearson’s correlation are shown. Black line represents linear regression fitting. Shading denotes 95% confidence interval of the fit. g As f for H3K27me3 ChIP-seq in NPCs. h Comparison of domain dynamics and chromatin opening. Relative domain score is shown. Relative sum of ATAC peaks per TAD is shown. Only TADs with a minimum of 15 peaks in ESC were used. Early, TADs that changed from NPC to D5 (Supplementary Fig. 6g). Late, TADs that did not change from NPC to D5. Line shows mean. Error bars denote SEM. Shading denotes 95% confidence interval. Right panel shows comparison of chromatin opening dynamics (relative ATAC peaks) between early and late TADs. i Comparison of domain dynamics and gene reactivation. Relative domain score is shown. Relative mean expression per TAD is shown. Early, TADs that changed from NPC to D5. Late, TADs that did not change from NPC to D5. Line shows mean. Error bars denote SEM. Shading denotes 95% confidence interval. Right panel shows comparison of gene reactivation dynamics (relative RNA expression) between early and late TADs. (h, i) The numbers above the lines indicate p-values (two-sample unpaired Wilcoxon–Mann–Whitney test with R defaults). n indicates number of TADs.

Fig. 7. Graphical summary.

a The inactive mouse X chromosome displays an underlying compartment structure characterized by distinct epigenetic signatures. b Partial X-reactivation occurs first in an A-like subcompartment near escapee genes. Early formation of TADs occurs in B-like compartments, while mega-domains are still maintained, and precedes transcriptional reactivation. c Full X-reactivation then occurs rapidly in sync with Xist downregulation, which is promoted by downregulation of Jpx.