Using population-level whole-genome sequencing to profile colistin resistance evolution dynamics in diverse clinical Pseudomonas aeruginosa lineages

Abstract

Colistin (CST) is a critical last-line antibiotic for treating multidrug-resistant Pseudomonas aeruginosa infections, especially in chronic respiratory disease. Despite its clinical importance, the P. aeruginosa CST resistome remains incompletely characterised, with most studies focusing on the PAO1 prototypic strain. Here, we used experimental evolution to apply stepwise CST selective pressure to nine P. aeruginosa lineages representing different clinical presentations and diverse genetic backgrounds. Population-level whole-genome sequencing of broth cultures exposed to 2-fold increasing CST (0.0625 to 512 μg/mL) was undertaken at multiple concentrations. We identified several known and novel mutational drivers of low- and high-level CST resistance, alongside associated compensatory mutations at allele frequencies (AFs) as low as 5%. Despite diverse clinical and genetic backgrounds, all lineages convergently evolved mutations within the two-component system pmrAB, with most accompanied by phoPQ variants. When canonical TCS remodelling plateaued, populations acquired novel secondary driver mutations, most notably in the lipid A-modifying gene, lpxO2. In addition to AMR driver mutations, strains evolved complex adaptive strategies to mitigate severe CST-induced oxidative and metabolic stress. Adaptations included the emergence of hypermutators and alterations to targeted DNA and protein repair networks. Furthermore, we observed convergent regulatory rewiring of the cyclic-di-GMP network and central metabolism pathways, likely acting as compensatory mechanisms to promote biofilm formation and offset the fitness costs of CST resistance. Using a mixture-aware population genomics approach, we captured intense clonal interference and transient tolerance mutations that failed to reach fixation, illustrating dynamic evolutionary complexities that would be missed with traditional end-point analysis methods.