Evolution and genetic architecture of disassortative mating at a locus under heterozygote advantage

The evolution of mate preferences may depend on natural selection acting on the mating cues and on the underlying genetic architecture. While the evolution of assortative mating with respect to locally adapted traits has been well-characterized, the evolution of disassortative mating is poorly characterized. Here we aim at understanding the evolution of disassortative mating for traits under strong balancing selection, by focusing on polymorphic mimicry as an illustrative example. Positive frequency-dependent selection exerted by predators generates local selection on wing patterns acting against rare variants and promoting local monomorphism. This acts across species boundaries, favouring Mullerian mimicry among defended species. In this well-characterized adaptive landscape, polymorphic mimicry is rare but is observed in a butterfly species, associated with polymorphic chromosomal inversions. Because inversions are often associated with recessive deleterious mutations, we hypothesize they may induce heterozygote advantage at the color pattern locus, putatively favoring the evolution of disassortative mating. To explore the conditions underlying the emergence of disassortative mating, we modeled both a trait locus (colour pattern for instance), subject to mutational load, and a preference locus. We confirm that heterozygote advantage favors the evolution of disassortative mating and show that disassortative mating is more likely to emerge if at least one allele at the trait locus is free from any recessive deleterious mutations. We modelled different possible genetic architectures underlying mate choice behaviour, such as self referencing alleles, or specific preference or rejection alleles. Our results showed that self referencing or rejection alleles linked to the color pattern locus can be under positive selection and enable the emergence of disassortative mating. However rejection alleles allow the emergence of disassortative mating only when the color pattern and preference loci are tightly linked. Our results therefore provide relevant predictions on both the selection regimes and the genetic architecture favoring the emergence of disassortative mating and a theoretical framework in which to interprete empirical data on mate preferences in wild populations.