Extensive differential gene expression and regulation by sex in human skeletal muscle

Abstract

The identification of sex-differential gene regulatory elements is essential for understanding sex-differential patterns of health and disease. We leveraged bulk and single-nucleus RNA sequencing (RNA-seq) and single-nucleus ATAC-seq data from 281 skeletal muscle biopsies to characterize sex differences in gene expression and regulation at the cell-type and whole-tissue levels. We found highly concordant sex-biased expression of over 2,100 genes across the three muscle fiber types and bulk tissue. Gene pathways related to mitochondrial activity and energy metabolism were enriched for male-biased expression, whereas those related to signal transduction and cell differentiation were enriched for female-biased expression. We found widespread sex-biased chromatin accessibility enriched in proximal and distal gene regulatory states; in gene promoters, sex-biased chromatin accessibility was positively associated with sex-biased expression. Long noncoding RNAs (lncRNAs) and microRNAs (miRNAs) also showed extensive sex-biased expression in the fiber-type and bulk data, respectively. Together, these results highlight nuclear and cytoplasmic mechanisms for sex-differential gene regulation in skeletal muscle.

Keywords: chromatin accessibility; gene expression; gene regulation; miRNA; sex differences; single nucleus; skeletal muscle; snATAC-seq; snRNA-seq.

Graphical abstract

Figure 1

Sex differences in cell-type composition of human skeletal muscle (A) Uniform manifold approximation and projection (UMAP) of 429,569 RNA and ATAC nuclei across 279 individuals. (B) Cell-type proportions for each individual sorted by sex and proportion of type 1 muscle fiber nuclei. (C) Mean cell-type proportions by sex. Error bars correspond to ±1 standard deviation calculated separately in each sex. (D) Fold change and 95% confidence intervals for the combined number of RNA and ATAC male nuclei compared to the number of female nuclei in each cell type. Cell types with significantly more nuclei at a FDR < 0.05 in females are colored red and in males are colored blue. See also Figures S1–S3 and Tables S2 and S3.

Figure 2

Sex differences in cell-type-specific gene expression in human skeletal muscle (A) The number and proportion of genes significantly differentially expressed by sex (FDR < 5%) in 10 muscle cell types. The proportions of genes significantly differentially expressed by sex (FDR < 5%) in type 1 muscle fiber data downsampled to match the sample size and total UMI are indicated with dashed white or black lines. Sample sizes varied by cell type; see Table S1. (B) Histograms and smooth curves with 95% confidence intervals for the proportion of lncRNAs, protein coding genes, and pseudogenes significantly differentially expressed by sex (FDR < 5%) by mean UMI calculated across all samples (117 females and 162 males) in type 1 muscle fiber. (C) The set of GO terms that are highly enriched (FDR < 0.1%) for genes expressed higher in males (odds ratio > 1) or females (odds ratio < 1) in at least one muscle fiber type. (D) The −log10 p values across the three muscle fiber types colored by direction of effect for the top 40 autosomal genes in the biological process GO term oxidative phosphorylation. (E) The −log10 p values across the three muscle fiber types colored by direction of effect for the top 40 autosomal genes in the cellular component GO term caveola. See also Figures S4–S11 and Tables S4, S5, S6–S8, and S9.

Figure 3

Comparison of sex differences in gene expression from bulk vs. single-nucleus RNA-seq (A) Scatterplot of the signed −log10 p values (>0 higher in males; <0 higher in females) of differential expression of 15,722 genes between the bulk (n = 268) and single-nucleus pseudobulk (n = 279). (B) The set of GO terms that are most highly enriched for genes expressed higher in males (odds ratio > 1) or females (odds ratio < 1) in the bulk. (C) The number of genes with significant sex-biased expression in at least one of the fiber types (type 1, type 2a, or type 2x) and/or bulk muscle. The sex-biased genes identified in at least one muscle fiber type and in bulk are further divided into those expressed more highly in the same sex in all muscle fiber types for which sex-biased expression was identified and in bulk (concordant) and those with mixed directions of effect (discordant). (D) The number and direction of sex-biased genes identified in at least one non-muscle fiber cell type by the joint sex-biased expression status in the fiber types and bulk. The number above each set of two bars indicates the number of sex-biased genes identified in at least one non-muscle fiber cell type that are a subset of the gene grouping in (C) directly above each set. For each group, when applicable, the proportions of sex-biased genes that are expressed more highly in the opposite sex in the non-fiber cell type compared to the fiber types (green) and bulk (purple) are shown. See also Figures S8 and S12–S16 and Tables S10, S11, S12, S13, and S14.

Figure 4

Sex differences in cell-type-specific chromatin accessibility in human skeletal muscle (A) The number and proportion of peaks significantly differentially accessible by sex (FDR < 5%) in 12 muscle cell types. The proportions of peaks significantly differentially accessible by sex (FDR < 5%) in type 1 muscle fiber data downsampled to match the sample size and total UMI of the other cell types are indicated by dashed white or black lines. Sample sizes varied by cell type; see Table S1. (B) The smooth curve and 95% confidence intervals for the proportion of peaks significantly differentially expressed by sex (FDR < 5%) by chromatin state by mean peak count across all samples and histograms of the number of peaks by state and mean peak count in type 1 muscle fiber. (C) The set of transcription factor binding site motifs (transcription factors may bind multiple motifs) that are most highly enriched for peaks with higher counts in males (odds ratio > 1) or females (odds ratio < 1) in type 1 muscle fiber in active TSS, strong transcription, quiescent, and enhancer consensus chromatin states. The columns are sorted by the mean odds ratio across the chromatin states. (D) The proportion of autosomal genes with 0 sex-biased peaks, ≥1 male-biased peaks, or ≥1 female-biased peaks <1 kb upstream of gene TSSs colored by their differential expression status and direction in the three fiber types, pseudobulk, and bulk. The single-nucleus pseudobulk and bulk genes are annotated with sex-biased peaks from the type 1 fiber. The histogram on the right shows the number of genes in bulk with each number/direction of sex-biased peaks; the distribution is similar for the three fiber types and pseudobulk. See also Figures S7 and S17–S21 and Tables S15, S16, S17, S18, and S19.