Host-specific plasmid evolution explains the variable spread of clinical antibiotic-resistance plasmids

Antibiotic resistance encoded on plasmids is a pressing global health problem. Predicting which plasmids spread in the long term remains very challenging, even though some key parameters influencing plasmid stability have been identified, such as plasmid growth costs and horizontal transfer rates. Here, we show these parameters evolve in a strain-specific way among clinical plasmids and bacteria, and this occurs rapidly enough to alter the relative likelihoods of different bacterium–plasmid combinations spreading. We used experiments with Escherichia coli and antibiotic-resistance plasmids isolated from patients, paired with a mathematical model, to track long-term plasmid stability (beyond antibiotic exposure). Explaining variable stability across six bacterium–plasmid combinations required accounting for evolutionary changes in plasmid stability traits, whereas initial variation of these parameters was a relatively poor predictor of long-term outcomes. Evolutionary trajectories were specific to particular bacterium–plasmid combinations, as evidenced by genome sequencing and genetic manipulation. This revealed epistatic (here, strain-dependent) effects of key genetic changes affecting horizontal plasmid transfer. Several genetic changes involved mobile elements and pathogenicity islands. Rapid strain-specific evolution can thus outweigh ancestral phenotypes as a predictor of plasmid stability. Accounting for strain-specific plasmid evolution in natural populations could improve our ability to anticipate and manage successful bacterium–plasmid combinations.