What the X has to do with it: differences in regulatory variability between the sexes in Drosophila simulans

Abstract

The mechanistic basis of regulatory variation and the prevailing evolutionary forces shaping that variation are known to differ between sexes and between chromosomes. Regulatory variation of gene expression can be due to functional changes within a gene itself (cis) or in other genes elsewhere in the genome (trans). The evolutionary properties of cis mutations are expected to differ from mutations affecting gene expression in trans. We analyze allele-specific expression across a set of X substitution lines in intact adult Drosophila simulans to evaluate whether regulatory variation differs for cis and trans, for males and females, and for X-linked and autosomal genes. Regulatory variation is common (56% of genes), and patterns of variation within D. simulans are consistent with previous observations in Drosophila that there is more cis than trans variation within species (47% vs. 25%, respectively). The relationship between sex-bias and sex-limited variation is remarkably consistent across sexes. However, there are differences between cis and trans effects: cis variants show evidence of purifying selection in the sex toward which expression is biased, while trans variants do not. For female-biased genes, the X is depleted for trans variation in a manner consistent with a female-dominated selection regime on the X. Surprisingly, there is no evidence for depletion of trans variation for male-biased genes on X. This is evidence for regulatory feminization of the X, trans-acting factors controlling male-biased genes are more likely to be found on the autosomes than those controlling female-biased genes.

Keywords: Cis/trans gene regulation; X-chromosome; allele-specific expression; sex-biased expression.

Fig. 1.—

Experimental design. Genotypes used in the experiment were produced from 5 parental D. simulans strains (C167.4, Md106, Md199, NewC, and w501), 5 corresponding X-substitution lines (denoted X^sub), and 1 reference strain (st e). For each parental D. simulans strain, X chromosomes were substituted into the common isogenic (st e) background. Each of the parental strains was crossed to the st e line to produce five F₁ genotypes (X^subX^{st e} A^subA^{st e}, X^subY^{st e} A^subA^{st e}) that were heterozygous (or hemizygous) for both X and the autosomes (panel A). Each of the X-substitution lines was crossed to the st e line creating five X^het st e genotypes (X^subX^{st e} A^{st e}A^{st e}, X^subY^{st e} A^{st e}A^{st e}) that were heterozygous (or hemizygous) for X only (denoted X^het st e; panel B). Six genotypes (X^subX^sub A^{st e}A^{st e}, X^{st e}X^{st e} A^{st e}A^{st e}, X^{st e}Y^{st e} A^{st e}A^{st e}) that were homozygous for both X and autosome (panels C and D) were also included in the experiment.

Fig. 2.—

Allele-specific analysis of cis and trans variation. For each contrast 1–6: the genotypes (only c167.4 shown) used in a contrast are shown in the first column; for a given focal gene, the allele-specific expression measurements used in the test are given in the second column, noted as C (allele derived from the C167.4 parental strain) or S (allele derived from the st e reference strain) in the genotype indicated by subscripts P (parental strain), F1 (F₁), or X1 (X^het st e); the genes that could be tested, X-linked (X) and autosomal (Autosome), are listed in the third column; the sex, Male (M) and Female (F), that the test could be conducted for is listed in the fourth column (with the genes that could be tested in superscript); and the effect tested is listed in the fifth column. Note that the cis by trans test was considered significant only if the cis effect in contrasts 1 or 2 was nonzero. For each cis or trans test, the difference in expression between the two alleles was compared with the allele-specific signal in the appropriate DNA control (supplementary fig. S1, Supplementary Material online).

Fig. 3.—

The distribution of cis and trans variation in transcript abundance. The distribution of the cis (solid line) and trans (dashed line) effect estimates (calculated as the standardized mean difference) for genes with significant regulatory variation are shown for males (blue) and females (red). The left panel shows the distribution for autosomal genes in females (n = 4,103 for cis and n = 1,804 for trans), the middle panel shows the distribution for autosomal genes in males (n = 3,704 for cis and n = 1,399 for trans), and the right panel shows the distribution for X-linked genes in females (n = 501 for cis and n = 250 for trans). For each plot, the Y axis is the frequency and the X axis is the standardized estimate of cis or trans differences between X-substitution parental strain genotypes and the st e reference line.

Fig. 4.—

Depletion of regulatory variation on the X relative to autosomes. The percent of genes that vary in cis (solid) or trans (diagonal shading) regulation (F₁ test) relative to the percent expected (light bars) for genes on X and genes on autosomes. Results are shown for male-biased (blue), female-biased (red), and equally expressed (gray) genes. For trans variation, only female-biased genes had significantly less variation on X than expected (P = 0.0001). Whereas depletion of cis variation on X is unrelated to sex bias (females, P < 0.0001; males, P = 0.0031; and unbiased P = 0.0002).