Landscape of X chromosome inactivation across human tissues

Abstract

X chromosome inactivation (XCI) silences transcription from one of the two X chromosomes in female mammalian cells to balance expression dosage between XX females and XY males. XCI is, however, incomplete in humans: up to one-third of X-chromosomal genes are expressed from both the active and inactive X chromosomes (Xa and Xi, respectively) in female cells, with the degree of 'escape' from inactivation varying between genes and individuals. The extent to which XCI is shared between cells and tissues remains poorly characterized, as does the degree to which incomplete XCI manifests as detectable sex differences in gene expression and phenotypic traits. Here we describe a systematic survey of XCI, integrating over 5,500 transcriptomes from 449 individuals spanning 29 tissues from GTEx (v6p release) and 940 single-cell transcriptomes, combined with genomic sequence data. We show that XCI at 683 X-chromosomal genes is generally uniform across human tissues, but identify examples of heterogeneity between tissues, individuals and cells. We show that incomplete XCI affects at least 23% of X-chromosomal genes, identify seven genes that escape XCI with support from multiple lines of evidence and demonstrate that escape from XCI results in sex biases in gene expression, establishing incomplete XCI as a mechanism that is likely to introduce phenotypic diversity. Overall, this updated catalogue of XCI across human tissues helps to increase our understanding of the extent and impact of the incompleteness in the maintenance of XCI.

Extended Data Figure 1. Assessment of skew in XCI in GTEx female samples (V3 analysis release)

a) Shows the estimated skew in XCI by tissue across individuals and b) shows the skew in XCI by individual across tissue samples available. Number in brackets after tissue or sample name gives the number of individuals or tissues, respectively, contributing to each boxplot. Details of the analysis is given in the Supplementary Note.

Extended Data Figure 2. Comparison of expression characteristics between reported genic XCI categories in the GTEx data

a) Table showing the statistics for the comparison of the proportion of significantly biased (FDR<1%) genes by reported XCI status. Distributions are illustrated in Fig. 2b. N = 29 for all comparisons. b) Table showing the statistics for the comparison of the consistency in effect sizes across tissues. Distributions are illustrated in Fig. 2c. Only genes expressed in at least five of the 29 tissues are included. c) Number of tissues showing significant sex bias (FDR<1%) per gene by reported XCI status. d) Statistics for the comparison illustrated in c). e) Number of tissues in which genes are expressed by reported XCI status. f) Statistics for the comparison illustrated in e). All P-values are from two-sided Wilcoxon tests, except for a) where paired, two-sided Wilcoxon test was applied. Only genes assessed for sex bias in at least one tissue are included unless otherwise stated.

Extended Data Figure 3. Change in the proportion of discovered sex-biased genes by XCI category with varying q-value cut-offs

a) The proportion of sex-biased genes across tissues. Here a gene is classified as sex-biased if the q-value for association falls below the given threshold in at least one tissue. b-f) Examples of the change in the proportion of sex-biased expression in individual tissues. The dashed black line indicates the FDR<1% cut-off applied in the analyses to determine sex-biased expression.

Extended Data Figure 4. Heatmap representation of male-female expression differences in all assessed X-chromosomal genes (N=681) across 29 GTEx tissues

The color scale displays the direction of sex bias with red color indicating higher female expression. Genes that were too weakly expressed in the given tissue type to be assessed in the sex bias analysis are colored grey. Dots mark the observations where sex bias was significant at FDR<1%

Extended Data Figure 5. Comparison of expression characteristics between Xp and Xq, the evolutionary newer and older regions of chrX, respectively, by XCI status and for the whole chromosome

a) and b) show level of median expression across GTEx tissues in log2 RPKM units, and c) and d) show the breadth of expression measured as the number of tissues (max = 29) in which genes are expressed (median expression across samples > 0.1 RPKM and expressed in more than 10 individuals at >1 counts per million). P-values are calculated using the Wilcoxon Rank Sum test. All genes expressed in at least one tissue are included in the comparisons.

Extended Data Figure 6. X-chromosomal RNA-seq and WGS data in the GTEx donor with fully skewed XCI (GTEX-UPIC)

a) Allelic expression in chrX in 16 RNA-sequenced tissue samples available from the donor. Dashed red lines indicate PAR1 and PAR2 boundaries. b) Allele balance and allele depth across chrX in WGS for GTEX-UPIC and randomly chosen two female and one male GTEx WGS samples.

Extended Data Figure 7. Expressed alleles at biallelically expressed ASE sites in scRNA-seq

a) X-chromosomal genes repeatedly biallelic in scRNA-seq (see Methods for details). b) Illustration of the relative expression from the two alleles at all X-chromosomal ASE sites that were repeatedly biallelically expressed across cells in either of the two scRNA-seq samples that showed random XCI (Y035 and 24A). Narrow white lines separate observations from individual cells.

Extended Data Figure 8. Assessment of the level of Xi expression at escape genes and in different regions of the X chromosome

a) The ratio of Xi-to-Xa expression in the single cell samples (left panel; each circle represents a sample) and in the skewed XCI donor from GTEx (middle panel; each circle represents a tissue), and the female-to-male ratio in expression (right panel, each circle represents a tissue) at reported escape genes. Genes are ordered according to their location in the X chromosome with genes in the pseudoautosomal region residing in the top part of the figure. Dark border around a circle indicate there was significant evidence for Xi expression greater than the baseline in the given sample or tissue (left and middle panels) or significant sex-bias in the given tissue (right panel). Given some outliers, e.g. XIST, the Xi-to-Xa ratio is capped at 1.75 and female-to-male ratio at 2.25. b) The relative expression arising from the X and Y chromosome at PAR1 genes in skeletal muscle in eight males. The allelic expression at these genes was assigned to the two chromosomes utilizing parental genotypes available for these samples (see Methods for details). The dashed line at 0.5 indicates the point where expression from X and Y chromosomes is equal. The error bars give the 95% confidence intervals for the observed read ratio. c) Heatmap representation of the change in pattern of sex-bias at 13 X-Y homologous gene pairs (see Methods for details) in nonPAR from only including the X-chromosomal expression (heatmap on the left) to accounting for the Y-chromosomal expression (heatmap on the right). The color scale displays the direction of sex-bias with red color indicating higher female expression. Genes that were too lowly expressed in the given tissue type to be assessed in the sex-bias analysis are colored grey. Dots mark the observations where sex-bias was significant at FDR<1%. The grey bars on top of the heatmaps indicate the location of the gene in the X chromosome: dark grey indicating Xp and lighter grey Xq.

Figure 1

Schematic overview of the study. Previous expression-based surveys of XCI^, have established the incomplete and variable nature of XCI, but these studies have been limited in the tissue types and samples assessed. To investigate the landscape of XCI across human tissues, we combined three approaches: 1) sex biases in expression using population-level GTEx data across 29 tissue types, 2) allelic expression in 16 tissue samples from a female GTEx donor with fully skewed XCI, and 3) validation using single cell RNA-seq by combining allelic expression and genotype phasing. WGS, whole genome sequencing; WES, whole exome sequencing; scRNA-seq, single-cell RNA-seq.

Figure 2

Assessment of tissue-sharing and population-level impacts of incomplete XCI in GTEx data. a) Male-female expression differences in reported XCI-escaping genes (N=82) across 29 GTEx tissues. b) Proportion of significantly biased (FDR<1%) genes in each tissue by reported XCI status. c) Proportion of tissues where the bias direction is shared by reported XCI status. Genes expressed in at least five tissues are included. d) Sex bias pattern of nine genes not classified as full escape genes that follow a similar profile to established escape genes. e) Chromatin state enrichment between escape and inactive genes in the Roadmap Epigenomics female samples.

Figure 3

Assessment of tissue-sharing of XCI in a GTEx donor with highly skewed XCI. a) Distribution of the skewness of XCI in GTEx female samples (N=62, V3 release). Each data point shows the mean skew in XCI across tissue samples per individual. b) Classification of X-chromosomal genes (N=186) into full or incomplete and tissue-shared or heterogeneous XCI based on the analysis of ASE patterns across tissues. Error bars show the 95% credible interval. c-e) Examples of genes where the ASE-based assessment of XCI status match previously reported assignments (TSR2, inactive; XIST, escape; ZBED1, escape). Note that XIST is, unusually for an escape gene, expressed monoallelically, only from Xi. f) KAL1 shows strong evidence for tissue-specific escape. g-k) Genes without previous or conclusive evidence for escape from XCI but classified as incompletely inactivated in this sample. In c-k asterisks indicate that the Xi expression in the given tissue was significant at FDR < 1% (one-sided binomial test) and errors bars show the 95% confidence interval.

Figure 4

Analysis of XCI using scRNA-seq. a) Proportion of genes demonstrating full and partial XCI in the ASE analysis in single cell RNA-seq data, and the concordance with previously reported XCI status. b-l) Examples of genes with different XCI patterns in scRNA-seq: previously reported inactive gene (b), known escape gene in PAR1 (c), escape gene with known exclusive expression from Xi (d) new candidates for escape genes that demonstrate incomplete XCI in only a subset of samples (e-k), and a known escape gene that shows escape of varying degrees in the three samples (Pearson's Chi-squared test for equal proportions, P=3.80×10^-7) (l). Asterisk above a bar indicates that the proportion of Xi expression, i.e. blue bar, in a given sample is significantly greater than the expected baseline (FDR < 1%, one-sided binomial test). Error bars show the 95% confidence interval.