Sex-biased microRNA expression in mammals and birds reveals underlying regulatory mechanisms and a role in dosage compensation

Abstract

Sexual dimorphism depends on sex-biased gene expression, but the contributions of microRNAs (miRNAs) have not been globally assessed. We therefore produced an extensive small RNA sequencing data set to analyze male and female miRNA expression profiles in mouse, opossum, and chicken. Our analyses uncovered numerous cases of somatic sex-biased miRNA expression, with the largest proportion found in the mouse heart and liver. Sex-biased expression is explained by miRNA-specific regulation, including sex-biased chromatin accessibility at promoters, rather than piggybacking of intronic miRNAs on sex-biased protein-coding genes. In mouse, but not opossum and chicken, sex bias is coordinated across tissues such that autosomal testis-biased miRNAs tend to be somatically male-biased, whereas autosomal ovary-biased miRNAs are female-biased, possibly due to broad hormonal control. In chicken, which has a Z/W sex chromosome system, expression output of genes on the Z Chromosome is expected to be male-biased, since there is no global dosage compensation mechanism that restores expression in ZW females after almost all genes on the W Chromosome decayed. Nevertheless, we found that the dominant liver miRNA, miR-122-5p, is Z-linked but expressed in an unbiased manner, due to the unusual retention of a W-linked copy. Another Z-linked miRNA, the male-biased miR-2954-3p, shows conserved preference for dosage-sensitive genes on the Z Chromosome, based on computational and experimental data from chicken and zebra finch, and acts to equalize male-to-female expression ratios of its targets. Unexpectedly, our findings thus establish miRNA regulation as a novel gene-specific dosage compensation mechanism.

Figure 1.

Sex-biased miRNA expression in mouse, opossum, and chicken. Male-to-female (M:F) ratios were calculated for three male and three female replicates per tissue and the significance assessed with DESeq2 (Love et al. 2014), with Benjamini-Hochberg correction for multiple tests (Benjamini and Hochberg 1995) performed separately for somatic and gonadal samples. For display purposes, the log₂-transformed M:F ratios were capped at −3 and 3, with more extreme ratios replaced by these values. Each data point corresponds to a multimap group (MMG) comprised of one or more mature miRNAs (see main text), and MMGs were considered sex-linked if they included at least one sex-linked member. Expression values correspond to the mean normalized counts provided by DESeq2. The most sex-biased miRNAs per tissue are listed in Table 1. The Z-linked and W-linked miRNA miR-122-5p is visible as the rightmost data point in the chicken liver panel. All data shown in this figure are included in Supplemental Table S4.

Figure 2.

Transcriptional mechanisms underlying sex-biased miRNA expression. (A) Correlation between sex-biased expression of intronic miRNAs (y-axis) and their host genes (x-axis). Male-to-female (M:F) expression ratios were calculated with DESeq2. For each tissue, the Spearman correlation coefficient (rho) is given, together with its Benjamini-Hochberg corrected P-value (Benjamini and Hochberg 1995). (B) Overview of the two miRNA loci that were associated with sex-biased DHS regions (Ling et al. 2010) in mouse liver. In both cases, a sequence of 1 kb is depicted. The displayed miRNA precursor region corresponds to the 5p and 3p sequences with the intervening loop sequence. (C) Expression levels in mouse liver of the 5p and 3p mature miRNAs shown in B, for three female and three male replicates. Raw read counts were normalized with DESeq2. Statistical significance: (*) P < 0.05; (**) P < 0.01; (***) P < 0.001.

Figure 3.

Somatic expression of ovary-biased and testis-biased miRNAs. (A) Proportion of autosomal MMGs with ovary-biased or testis-biased expression that are detected in at least one somatic tissue. (B) Somatic sex bias of ovary-biased and testis-biased miRNAs. If an MMG was detected in more than one somatic tissue, the most extreme M:F ratio was chosen. Statistical significance: (**) P < 0.01; (***) P < 0.001; (n.s.) not significant.

Figure 4.

Role of miR-2954-3p in dosage compensation. (A) Expression of miR-2954-3p in chicken tissues. Raw read counts were normalized with DESeq2. (B) Proportion of Z-linked genes among the genes that were predicted to be targets of miR-2954-3p and autosomal miRNAs. Target prediction was performed with TargetScan and only 8-mer sites were considered. (C) Log₂-transformed M:F expression ratios for Z-linked target genes of miR-2954-3p or autosomal miRNAs. (D) Proportion of annotated ohnologs between Z-linked targets of miR-2954-3p and autosomal miRNAs. (E) Log₂-transformed expression ratios for protein-coding genes following miR-2954-3p knockdown compared to control. Genes that are down-regulated by miR-2954-3p are expected to have positive ratios. The first panel shows Z-linked genes in purple and autosomal genes in gray. The following two panels show Z-linked genes predicted to be more (light purple) or less (dark purple) dosage-sensitive based on chicken expression data and ohnolog annotations. Only genes that were 1-to-1 orthologs and located on the Z Chromosome in both chicken and zebra finch were included in the analyses. Statistical significance: (*) P < 0.05; (**) P < 0.01; (***) P < 0.001.