Lineage-specific dynamics of loss of X upregulation during inactive-X reactivation

Abstract

In mammals, X chromosome dosage is balanced between sexes through the silencing of one X chromosome in females. Recent single-cell RNA sequencing analysis demonstrated that the inactivation of the X chromosome is accompanied by the upregulation of the active X chromosome (Xa) during mouse embryogenesis. Here, we have investigated if the reactivation of inactive-X (Xi) leads to the loss of Xa upregulation in different cellular or developmental contexts. We find that while Xi reactivation and loss of Xa upregulation are tightly coupled in mouse embryonic epiblast and induced pluripotent stem cells, that is not the case in germ cells. Moreover, we demonstrate that partial reactivation of Xi in mouse extra-embryonic endoderm stem cells and human B cells does not result in the loss of Xa upregulation. Finally, we have established a mathematical model for the transcriptional coordination of two X chromosomes. Together, we conclude that the reactivation of Xi is not always synchronized with the loss of Xa upregulation.

Keywords: B-cells; Epiblast; X-chromosome inactivation; X-chromosome upregulation; XEN; dosage compensation; extra-embryonic endoderm stem cells; germ cells; iPSC; pre-implantation embryos; scRNA-seq.

Figure 1

Erasure of Xa upregulation upon reactivation of Xi in mouse embryonic epiblast (A) Comparison of allelic X:A ratio (top) and allelic expression of X and autosomes (bottom) between EPI:ICM (Xa^matXi^pat) and EPI:E5.5 (Xa^matXa^pat) cells. A comparison between EPI:ICM and EPI:E5.5 epiblasts of male cells is also plotted. (B) Comparison of allelic X(escapees):A ratio (left) and allelic expression of escapees and autosomes (right) between EPI:ICM (Xa^matXi^pat) and EPI:E5.5 (Xa^matXa^pat) cells. (C) Heatmap representing the allelic expression changes of Xa^mat and Xi^pat genes upon transition of EPI:ICM (Xa^matXi^pat) cells to EPI:E5.5 (Xa^matXa^pat) cells. Xa^mat, maternal active-X; Xi^pat, paternal inactive-X. In all boxplots, the line inside each of the boxes denotes the median value, the red circle denotes the mean value, and the edges of each box represent 25% and 75% of the dataset, respectively (Wilcoxon rank-sum test: p values < 0.0001; ^∗∗∗∗, < 0.01; ^∗∗, NS: not significant).

Figure 2

Erasure of Xa upregulation upon reactivation of Xi during mouse iPSCs reprogramming (A) Schematic showing different stages of reprogramming of female MEF to iPSC (bulk RNA-seq dataset). (B) The plot represents allelic X:A ratio throughout the different stages of iPSC reprogramming. (C) Schematic showing different stages of reprogramming of female MEF to iPSC (scRNA-seq dataset). (D) Top: comparison of allelic X:A ratio between X-inactivated (Xa^castXi^mus) vs. X-reactivated (Xa^castXa^mus) cells during iPSC reprogramming. Bottom: allelic expression (log2 allelic TPM+1) of X-linked and autosomal genes in X-inactivated vs. X-reactivated cells during iPSC reprogramming (Wilcoxon rank-sum test: p value < 0.0001; ^∗∗∗∗, NS: not significant). (E) Heatmap representing the allelic expression changes of Xa^cast and Xi^mus genes upon conversion of X-inactivated (Xa^castXi^mus) cells to X-reactivated (Xa^castXa^mus) cells during iPSC reprogramming. (F) Plots representing non-allelic X:A ratio throughout the different stages of reprogramming of male MEF to iPSC. In all boxplots, the line inside each of the boxes denotes the median value, the red circle denotes the mean value, and the edges of each box represent 25% and 75% of the dataset, respectively.

Figure 3

Reactivation of Xi and loss of Xa upregulation is not tightly coupled in mouse germ cells (A) Schematic showing differentiation of XGFP- PGCLCs toward meiotic cells using rOvary system as described in the study by Severino et al., 2022 (Severino et al., 2022). (B) Comparison of allelic X:A ratio between X-inactivated (Xa^castXi^mus) and X-reactivated (Xa^castXa^mus) cells. (C) Allelic expression (log2 allelic TPM+1) of X-linked and autosomal genes in X-inactivated (Xa^castXi^mus) vs. X-reactivated cells (Xa^castXa^mus). (D) Scatterplots showing the correlation between Xa^cast upregulation (Xa^cast:A^cast) and reactivation of the Xi^mus (fraction expression from X^cast allele) in mitotic, pre-meiotic1, pre-meiotic2, and meiotic (labeled with different colors) germ cells. R is Pearson’s correlation. In all boxplots, the line inside each of the boxes denotes the median value, the red circle denotes the mean value, and the edges of each box represent 25% and 75% of the dataset, respectively (Wilcoxon rank-sum test: p values < 0.0001; ^∗∗∗∗, < 0.01; ^∗∗ and < 0.05; ^∗; NS: not significant).

Figure 4

Partial reactivation of Xi does not lead to the loss of active-X upregulation in mouse XEN cells (A) Schematic showing the target sites of sgRNAs at the Xist locus and heterozygous deletion of Xist from the paternal Xi^mus allele. (B) Left: detection of Xist/Tsix RNA (white) and Rnf12 (red) through RNA-FISH in WT and ΔXist^mus XEN cells. DAPI stained the nuclei in blue. The scale bar represents 1 μm. Right: quantification of Xist-coated nuclei in WT and ΔXist^mus XEN cells. (C) RNA-seq signals for Xist in WT and ΔXist^mus XEN cells. (D) Plot representing the allelic X:A ratio in WT and ΔXist^mus XEN cells. (E) Plot representing autosomal and X chromosome allelic expression (Log2 allelic TPM+1) in WT and ΔXist^mus XEN cells. (F) Identification of different categories of reactivated genes (Xr-low, Xr-intermediate, and Xr-robust) based on the reactivation status through profiling the fraction Xi^mus allele expression in WT and ΔXist^mus XEN cells. (G) Allelic X:A ratio for reactivated X-linked genes in WT and ΔXist^mus XEN cells. (H) Allelic expression (log2 allelic TPM+1) plot for autosomal and reactivated X-linked genes in WT and ΔXist^mus XEN cells. (I) Heatmap representing the allelic expression changes of Xa^mol and Xi^mus genes (reactivated genes) of WT XEN cells upon partial reactivation of Xi in ΔXist^mus XEN cells (Wilcoxon rank-sum test: p value < 0.0001; ^∗∗∗∗; NS: not significant).

Figure 5

Partial reactivation of Xi in human B cells does not lead to the loss of X upregulation Plots representing (A) allelic X:A ratio and (B) autosomal and X chromosome allelic expression (log2 allelic TPM+1) in Ctrl, sgXIST, inhibitor and sgXIST + inhibitor-treated B cells. (C) Identification of different categories of reactivated genes (Xr-low, Xr-intermediate, and Xr-robust) based on the reactivation status through profiling of the fraction paternal X expression in Ctrl, sgXIST, inhibitor and sgXIST + inhibitor-treated B cells. Plots for (D) allelic X:A ratio for reactivated genes and (E) allelic expression (log2 allelic TPM+1) for autosomal and reactivated X-linked genes. (F) Heatmap representing allelic expression of X-linked genes from Xi^pat and Xa^mat allele in Ctrl, sgXIST, inhibitor and sgXIST + inhibitor-treated B cells. (G) Model representing that the partial reactivation of Xi does not result in loss of Xa upregulation in XEN and B cells. In all boxplots, the line inside each of the boxes denotes the median value and red circle mean value (Wilcoxon rank-sum test: p values < 0.0001; ^∗∗∗∗, < 0.001; ^∗∗∗, < 0.01; ^∗∗ and < 0.05; ^∗).

Figure 6

Phenomenological model to explain partial and full reactivation dynamics (A–F) Plots representing fits obtained from simulations related to X:A ratio of two X chromosomes during iPSC reprogramming: (A) with only cross-inhibition, (B) with cross-inhibition and self-activation, and (C) with cross-inhibition and self-inhibition. Plot representing fits on allelic X:A ratio during partial reactivation: (D) with only cross-inhibition, (E) with cross-inhibition and self-activation, and (F) with cross-inhibition and self-inhibition. (G) Schematic representation of all combinations of possible cross-regulatory links (interactions between the chromosomes) and self-regulatory links (interactions within a chromosome). (H and I) Heatmap representing R² for different fits for testing self-regulatory connections with fixed cross-inhibition on full and partial reactivation data, respectively. (J and K) Heatmaps of R² for fits for testing cross-regulatory connections with fixed self-activation on full and partial reactivation data, respectively. (L and M) Heatmaps of R² for fits testing cross-regulatory connections with fixed self-inhibition on full and partial reactivation data, respectively. (N) Time-course distribution of X level on addition of noise to the model.