Pleiotropy, or the influence of a single locus on multiple traits, is thought to be an important mechanism of evolutionary constraint and could be an important deterministic factor during adaptive evolution. Theoretical models and work in quantitative genetics has long suggest that pleiotropy strongly impacts the probability that a given locus will be used in adaptation. Despite these theoretical predictions, we have lacked explicit tests of how pleiotropy contributes to evolutionary predictability. The overarching goal of this project was to test whether pleiotropy is a source of evolutionary constraint that underlies the predictability of evolutionary responses. Specifically, we sought to identify the genomic loci associated with repeated divergence between stickleback adapted to different habitats and test whether pleiotropy levels explained the repeated use of the same loci during adaptation. Illumina sequence data from multiple independent ecotype pairs of threespine stickleback were analyzed to estimate the genome-wide patterns of genetic divergence. We found that a large portion of the genome was evolving in a repeatable manner among independently derived population pairs of stickleback. Furthermore, ecologically relevant traits mapped to these repeatedly diverged genomic regions, and ecological similarity was an important predictor of the magnitude of parallel evolution. Using two proxies for pleiotropy, gene connectivity and number of traits with mapped QTL, we estimated the relationship between the level of pleiotropy and probability of parallel evolution. We found that parallel genomic regions contained genes with significantly more pleiotropy than uniquely evolving (non-parallel) regions. The increased mean pleiotropy of parallel windows could not be explained by other genomic factors, as there was no significant difference in mean gene count, mutation or recombination rates between parallel and non-parallel windows. Interestingly, although non-parallel windows contained genes that were on average less connected and influencing fewer mapped traits than parallel windows, these windows also tended to contain the gene that were the most pleiotropic. Taken together, our findings are consistent with the idea that low or intermediate levels of pleiotropy may be beneficial for adaptation, and that it is only at high levels that pleiotropy becomes constraining. These findings may help to inform our expectations about the genetic architecture of rapid evolutionary responses in nature or in agricultural and medical settings.
