Divergence in genetic (co)variances and the alignment of gmax with phenotypic divergence

To better understand the sources of biological diversity in nature, we need information on the mechanisms underlying population divergence. Biological systems with patterns of naturally occurring adaptive variation among populations can provide insight into the genetic architecture of diverging traits and the influence of genetic constraints on responses to selection. Using a system of reproductive character displacement in the North American mushroom-feeding fly Drosophila subquinaria, we assessed patterns of genetic (co)variance among a suite of chemical signaling traits and divergence in this pattern among populations. D. subquinaria exhibits stronger reproductive isolation against the closely related Drosophila recens in sympatry, where both female mating preferences and male chemical signalling traits have diverged from the ancestral allopatric populations. We collected three wild populations from each region and, in the lab, characterized the phenotypic divergence in these traits as well as the additive genetic (co)variance structure (G-matrix) with replicate breeding designs. We found divergence between allopatric and sympatric D. subquinaria in the shape and size of the G-matrix, and that the leading axis of genetic variance (gmax) had changed in sympatry to come into alignment with the primary axis of phenotypic divergence between the sympatric and allopatric regions.

Methods
This dataset was generated by phenotyping Drosophila subquinaria offspring from several large breeding designs. Lab stocks were established by collecting flies from several natural populations, which were then used to conduct blocked breeding designs in order to track the relatedness of offspring. Phenotyping was primarily the assessment of an individuals CHC (cuticular hydrocarbon) profile using gas chromatography. This profile is the relative abundance of a collection of chemical compounds that are necessary for various biological functions. These were transformed to give nine logcontrast traits for each individual. This phenotype, along with an individuals parentage codes and fixed effects of block and gas chromatograph channel, were used in all subsequent analyses.