On the metabolic basis and predictability of global epistasis.

Fitness effects of mutations are epistatic, i.e., they depend on the genetic background in which they occur. Despite microscopic idiosyncrasies of mutational effects, laboratory evolution experiments showed consistent patterns of declining adaptability, indicating some underlying 'rule' that dictates mutational effects. Recently, global epistasis showed that mutations become less beneficial as the fitness of the genetic background in which they occur increases, thus making epistasis predictable. However, there exists no mechanistic explanation of this fitness dependence of fitness effects. In this work, we use a prokaryotic cell growth and division model based on metabolic sizing to study the mechanistic basis, and predict the likelihood of global epistasis. We report two important results. One, although the average extent of global epistasis does not change with genetic variation, the likelihood of strong global epistasis decreases as genetic variation increases. Two, the fitness effects of mutations in the slow reaction modules are well predicted using the global epistasis approach. Therefore, when the fitness of the cell is contingent on a rate-limiting step, mutations in that step are well-predicted by the fitness of the background in which they occur. Thus, declining adaptability is a consequence of evolution acting in a step-wise fashion to improve the slowest modules in a cell. Overall, our work offers a first-principle explanation of global epistasis and declining adaptability.