Adaptive randomness: How does phenotypic stochasticity fuel adaptive evolution?

Biological systems are inherently stochastic, hence any trait varies stochastically even when genotype and environment are held constant. Although such phenotypic stochasticity has been thought to affect the adaptation process, we lack population genetic theory and experimental systems to study it in a rigorous quantitative manner. Here, we tackle this problem by combining theoretical and experimental approaches. First, we develop a novel population genetics simulator that implements genotype-phenotype maps with a genetic control of phenotypic stochasticity in addition to phenotypic expectation. This model allows us to predict under what evolutionary conditions mutations with high phenotypic stochasticity are beneficial. Second, we experimentally validate these theoretical predictions taking advantage of well-characterised “arrhythmic” mutants of circadian clock genes in Drosphila melanogaster, whose eclosion time is more stochastic than wild-type individuals. Specifically, we conduct a selection experiment on eclosion time for populations consisting of both wild-type and arrhythmic alleles, and compare the empirical allele frequency change over generations to the theoretical prediction under matching population genetic parameters. Our study is one of the first attempts to shed light on the under- appreciated roles that randomness plays in adaptive evolution.