An empirical long-term competition among natural yeast isolates reveals that short-term fitness largely but not entirely predicts long-term outcomes

In this study, we investigate the relative contribution of initial fitness to the long-term success of a genotype competing in a naturally diverse population. Specifically, we compete over 300 genetically barcoded S. cerevisiae isolates in a pooled setting for over 700 generations. We found that the strains that remain at detectable frequency until the end of the competition uniformly come from the top 95th percentile in the initial fitness values, making initial fitness the most significant predictor of long-term success. However, we occasionally see heterogeneity in the competition outcomes, which suggests a role of stochastic adaptation, clonal interference, and possibly frequency-dependent changes in strains’ fitness. We demonstrate that the “finalists” of our competition change on the genetic level, and that the spectrum of de novo mutations depends both on the strains’ genotype and environment. Finally, we show that gene targets of the novel mutations are specific to the combination of strain identity and environment, even among the genetically similar strains and environments that select for the same strains in the beginning of the competition.