Widespread cryptic variation in genetic architecture between the sexes

Abstract

The majority of the genome is shared between the sexes, and it is expected that the genetic architecture of most traits is shared as well. This common architecture has been viewed as a major source of constraint on the evolution of sexual dimorphism (SD). SD is nonetheless common in nature, leading to assumptions that it results from differential regulation of shared genetic architecture. Here, we study the effect of thousands of gene knockout mutations on 202 mouse phenotypes to explore how regulatory variation affects SD. We show that many traits are dimorphic to some extent, and that a surprising proportion of knockouts have sex-specific phenotypic effects. Many traits, regardless whether they are monomorphic or dimorphic, harbor cryptic differences in genetic architecture between the sexes, resulting in sexually discordant phenotypic effects from sexually concordant regulatory changes. This provides an alternative route to dimorphism through sex-specific genetic architecture, rather than differential regulation of shared architecture.

Keywords: Between‐sex genetic correlation; genetic architecture; knockout; rFM; sexual dimorphism.

Figure 1

(A) Estimates and associated uncertainty for sexual dimorphism for each trait analyzed. Each horizontal line displays the credible intervals for one trait, where traits have been arranged by the posterior median. Shaded regions indicated the credible intervals of 50%, 80%, and 95% of the posterior densities from a multilevel model. Sexual dimorphism is averaged across the wild‐type genotypes, and defined as the ratio of female and male means. (B) As in panel A, but depicting the between‐sex genetic correlation $r_{\mathrm{fm}}^{\mathbf{K}}$. Note that the traits have been arranged independently in each panel.

Figure 2

The between‐sex genetic correlation does not depend on sexual dimorphism in the trait. Each point is a trait, with error bars indicating the 95% credible interval (CI) in the estimates. The red line represents the model fit of a linear model on the Fisher‐transformed $r_{\mathrm{fm}}^{\mathbf{K}}$, with the shaded region indicating the 95% CI, including propagation of trait level uncertainty. Sexual dimorphism is expressed as the SD ratio.

Figure 3

The between‐sex genetic correlation decreases as size dimorphism increases over development. (A) Estimates for sexual dimorphism in body mass for wild‐type mice. Points indicate the posterior median with wide and narrow line segments denoting the 66% and 95% credible intervals, respectively, and the density gradient represents the posterior density. (B) As in panel A, but depicting the between‐sex genetic correlation. (C) Association of sexual size dimorphism and the $r_{\mathrm{fm}}^{\mathbf{K}}$ during development. Points are posterior medians with 95% credible intervals, as in panels A and B, with lines connecting subsequent week. Weeks 3 through 7 are numbered.

Figure 4

Identifying genotypes with consistent sexually discordant effects. Each point is a genotype, having been tested for at least 50 traits, with error bars denoting 95% credible intervals (CIs). The average percentile rank for the absolute sexually discordant effect of a genotype is plotted along the x‐axis. The y‐axis shows the average percentile tank for the absolute concordant effect. Red points indicate genotypes that tend to have more sexually discordant effects than other genotypes, whereas blue points are genotypes that have less discordant effects (CI does not overlap 50th percentile).