Predicting the structure of GxE and GxGxE interactions in Saccharomyces cerevisiae

Evolutionary dynamics depend on how segregating alleles affect organismal fitness. A major obstacle to understanding these dynamics is that the fitness effect of an allele often changes across genetic backgrounds (GxG interactions) and environments (GxE interactions) or both (GxGxE interactions). The statistics of these interactions are in general poorly characterized, but a recently discovered heuristic rule, called "diminishing returns epistasis" (DRE), predicts a substantial fraction of GxG interactions. Here, we show that a straightforward extension of this rule also predicts the statistics of GxE and GxGxE interactions. To test this prediction, we carried out RB-TnSeq experiments where we measured the fitness effects of ~100 insertion mutations across 42 strains of yeast Saccharomyces cerevisiae in 6 environments that vary by temperature and pH, two stressors with global effects on yeast physiology. We find that our predictions explain a substantial fraction of measured GxE and GxGxE effects. Additionally, we uncover that the environment impacts only the intercept, and not the slope, of the DRE patterns (e.g how selection coefficient changes with strain fitness) for the majority of mutations measured. These findings are a substantial step forward toward the goal of predicting fitness trajectories in variable environments.