Adaptation of the yeast gene knockout collection is near-perfectly predicted by fitness and diminishing return epistasis

Abstract

Adaptive evolution of clonally dividing cells and microbes is the ultimate cause of cancer and infectious diseases. The possibility of constraining the adaptation of cell populations, by inhibiting proteins that enhance their evolvability has therefore attracted substantial interest. However, our current understanding of how individual genes influence the speed of adaptation is limited, partly because accurately tracking adaptation for many experimental cell populations in parallel is challenging. Here we use a high throughput artificial laboratory evolution (ALE) platform to track the adaptation of >18.000 cell populations corresponding to single gene deletion strains in the haploid yeast deletion collection. We report that the fitness of gene knockout near-perfectly (R2=0.91) predicts their adaptation dynamics under arsenic exposure, leaving virtually no role for dedicated evolvability functions in the corresponding proteins. We tracked the adaptation of another >23.000 yeast gene knockout populations to a diverse range of selection pressures and generalised the almost perfect (R2=0.72 to 0.98) capacity of initial fitness to predict the rate of adaptation. Finally, we reconstruct mutations in the genes FPS1, ASK10, and ARR3, which together account for almost all arsenic adaptation in wildtype cells, in gene deletions covering a broad fitness range. We show that the predictability of arsenic adaptation can be understood almost entirely as a global epistasis phenomenon where excluding arsenic from cells, through these mutations, is more beneficial in cells with low arsenic fitness regardless of what causes the arsenic defects. The lack of genes with a meaningful effect on the adaptation dynamics of clonally reproducing cell populations diminishes the prospects of developing adjuvant drugs aiming to slow antimicrobial and chemotherapy resistance.