Genetic perturbations that affect bacterial resistance to antibiotics have been characterized genome-wide, but how do such perturbations interact with subsequent evolutionary adaptation to the drug? Here, we show that strong epistasis between resistance mutations and systematically identified genes can be exploited to control spontaneous resistance evolution. We evolved hundreds of Escherichia coli K-12 mutant populations in parallel, using a robotic platform that tightly controls population size and selection pressure. We find a global diminishing-returns epistasis pattern: strains that are initially more sensitive generally undergo larger resistance gains. However, some gene deletion strains deviate from this general trend and curtail the evolvability of resistance, including deletions of genes for membrane transport, LPS biosynthesis, and chaperones. Deletions of efflux pump genes force evolution on inferior mutational paths, not explored in the wild type, and some of these essentially block resistance evolution. This effect is due to strong negative epistasis with resistance mutations. The identified genes and cellular functions provide potential targets for development of adjuvants that may block spontaneous resistance evolution when combined with antibiotics.

New ways of curbing the ability of bacteria to evolve spontaneous resistance could 10 mitigate the looming antibiotic resistance crisis. Progress toward this goal requires a comprehensive understanding of the key factors that contribute to resistance evolvability. Here, we present a systematic approach to identify cellular functions that affect the evolvability of resistance. Using a robotic lab-evolution platform that keeps population size and selection pressure under tight control for hundreds of Escherichia coli populations evolving in parallel, we 15 quantified the effects of a genome-wide selection of pre-existing gene deletions on resistance evolution. Initial resistance of strains with gene deletions differed by more than tenfold but converged toward a hard upper bound for resistance during the evolution experiment, reflecting a global pattern of diminishing returns epistasis. We identified specific cellular functions that drastically curtail the evolvability of resistance; beyond DNA repair, these include membrane 20 transport, LPS biosynthesis, and chaperones. Perturbations of efflux pumps prevented resistance evolution completely or forced evolution on inferior mutational paths, not explored in the wild type. We show that strong negative epistasis generally underlies these phenomena. The identified functions provide new targets for adjuvants tailored to block evolutionary paths to resistance when combined with antibiotics. 25