A rugged yet easily navigable fitness landscape of antibiotic resistance

Abstract

A fitness landscape is a biological analogue of a physical landscape, in which each genotype occupies a location whose elevation corresponds to fitness. Theoretical models predict that rugged fitness landscapes with multiple peaks should impair Darwinian evolution, because natural selection prevents evolving populations from traversing the valleys that lie between peaks. Experimental tests of this prediction are very limited. Here we combine CRISPR-Cas9 genome editing and deep sequencing to map the fitness landscape of more than 260000 genotypes of the E. coli folA gene in an environment harboring the antibiotic trimethoprim. The folA gene encodes the key metabolic enzyme dihydrofolate reductase (DHFR), which is also a target of this antibiotic. With 514 mostly low fitness peaks, the DHFR fitness landscape is rugged. Despite this ruggedness, its highest fitness peaks are easily accessible to evolving populations. Fitness-increasing paths to high fitness peaks are abundant, and individual peaks have large basins of attractions. The basins of different peaks overlap, which renders the outcome of adaptive evolution highly contingent on chance events. In sum, ruggedness need not be an obstacle to Darwinian evolution but can reduce its predictability. If true in general, evolutionary biology and other fields of sciences in which landscapes play an important role may have to re-appraise the complexity of optimization problems on realistic landscapes.