The evolution of context-specific dominance during selective sweeps

Dominance is a widespread feature of genetic variants which affects life-history traits and fitness. Although dominance is generally thought to be an intrinsic property of genetic variants, it can sometimes evolve, as in the classic case of melanism in the peppered moth. The broader question of how likely dominance is to evolve is, however, controversial, because conditions favouring dominance evolution are often restrictive. Here, we revisit Haldane's classic hypothesis that dominance might evolve during the spread of beneficial mutations to fixation (i.e., during selective sweeps). We first confirm results of earlier models that sweeps of unconditionally beneficial mutations generate little potential for dominance to evolve, even in cases where modifier alleles segregate prior to selective sweeps. However, when sweeping beneficial alleles trade off between different environments --- which we explore with the illustrative case of sexually antagonistic selection --- the scope for dominance evolution expands. This occurs because modifier alleles can alter dominance separately in each environment, increasing the mean fitness of heterozygotes, prolonging the sojourn time of the sweep, and generating more heterozygosity upon which the modifier can act. In extreme cases, beneficial mutations that were initially destined for fixation can undergo a "dominance reversal" as a result of dominance evolution, converting them to balanced polymorphisms. We quantify how regularly dominance reversals of sweeping sexually antagonistic alleles can be expected to evolve. Overall, our results highlight conditions that allow the dominance of beneficial mutations to evolve, which we discuss in light of data on the frequency of selective sweeps, standing genetic variation for modifiers, and plasticity of modifier effects.