Understanding how genes interact is a central challenge in biology. Non-additive genetic (epistatic) interactions are crucial in constraining evolutionary trajectories and underlie many complex traits. Yet high-throughput methods for uncovering epistatic interactions are lacking. The current “gold standard” approach for characterizing epistasis is to engineer all possible combinations of evolved mutations and to measure the fitness of each genotype. Though powerful, this strategy is not practical for populations that contain more than a few mutations. To overcome this limitation we developed a new high-throughput method for assaying fitness and epistasis in genetic crosses and applied this method to twelve populations from a previously published yeast evolution experiment.
By crossing each evolved clone with its ancestor, we generated large pools of recombinant progeny containing random combinations of evolved mutations. For each of the twelve crosses we determined the distribution of fitness among 192 randomly selected recombinant individuals. Next, we determined the background-averaged fitness effect for all 111 mutations across our twelve evolved populations by propagating the recombinant pools and quantifying the change in frequency of each mutation over time. We developed a theoretical framework that combines these two data sets to reveal the patterns of epistasis among the mutations in each population.
Our analysis shows that most interactions between evolved mutations follow the model of “global diminishing returns” epistasis. This method also succeeded in identifying several new examples of idiosyncratic epistasis. We are beginning to explore the biological basis underlying these interactions. This is the first step towards understanding how epistatic interactions arise in the context of evolution and how epistasis influences the dynamics of adaptation.