Compatibility of evolutionary responses to constituent antibiotics drive resistance evolution to drug pairs.

Antibiotic combinations are considered a relevant strategy to tackle the global antibiotic resistance crisis since they are believed to increase treatment efficacy and reduce resistance evolution (World Health Organization and Global Tuberculosis Programme, 2016). However, studies of the evolution of bacterial resistance to combination therapy have focused on a limited number of drugs and have provided contradictory results (Hegreness et al., 2008; Lipsitch and Levin, 1997; Munck et al., 2014). To address this gap in our understanding, we performed a large-scale laboratory evolution experiment, adapting 8 replicate lineages of Escherichia coli to a diverse set of 22 different antibiotics and 33 antibiotic pairs. We found that combination therapy significantly limits the evolution of de novo resistance in E. coli, yet different drug combinations vary substantially in their propensity to select for resistance. In contrast to current theories, the phenotypic features of drug pairs are weak predictors of resistance evolution. Instead, the resistance evolution is driven by the relationship between the evolutionary trajectories that lead to resistance to a drug combination and those that lead to resistance to the component drugs. Drug combinations requiring a novel genetic response from target bacteria compared to the individual component drugs significantly reduce resistance evolution. These data support combination therapy as a treatment option to decelerate resistance evolution and provide a novel framework for selecting optimized drug combinations based on bacterial evolutionary responses.