Antibiotic tolerance and persistence in clinical isolates of Escherichia coli evaluated by high-resolution time-kill assays

Bacteria showing tolerance to antibiotic killing are characterized by a slower kill rate compared with other isolates from the same species. Antibiotic persistence is defined by a subpopulation of bacteria showing a reduced kill rate when an isolate is exposed to a bactericidal antibiotic. Both are best studied using time-kill experiments and have been associated with treatment failure and the development of antibiotic resistance. Despite their potential clinical importance, quantitative data for the variation in antibiotic tolerance and persistence are lacking. Here, we performed time-kill experiments using 15 clinical E. coli isolates with or without acquired TEM-1 or OXA-1 β-lactamases and E. coli ATCC25922 against four antibiotics (piperacillin-tazobactam, cefotaxime, meropenem, and ciprofloxacin) at 10× minimum inhibitory concentration (MIC) in a rich medium. We used mathematical modeling to obtain quantitative estimates of tolerance and persister frequency. For β-lactams, the time-kill curves showed three distinct phases: an initial bacteriostatic phase, a second phase with rapid killing, and finally, a phase with slow killing characteristic of the persister fraction. For ciprofloxacin, the initial bacteriostatic phase was absent. Estimates of the duration of the bacteriostatic phase and fast kill rates were correlated between piperacillin-tazobactam and cefotaxime, but not with meropenem or β-lactamase content. Using MDK₉₉, that is, the duration of antibiotic exposure to obtain a 2 log reduction in CFU count, as a measure of tolerance, we did not identify highly tolerant isolates. MDK₉₉ was the shortest for ciprofloxacin. The proportion of cells undergoing slow second-phase killing correlated between all β-lactams and was highest for meropenem. Only 1/15 clinical isolates displayed an increased proportion of persister cells.IMPORTANCEStudies of the clinical impact of antibiotic tolerance and persistence have lacked standardized protocols and definitions. This study uses reference methodology and mathematical modeling to provide quantitative measures in clinical E. coli isolates of metrics commonly used to describe tolerance and persister cell frequency. We observe that the β-lactam-mediated killing is preceded by a bacteriostatic phase that was absent for ciprofloxacin. The data indicate that tolerance may be specific for the mechanism of antibiotic action, and persister cell frequency varies for different drug classes. Using the data obtained from the study isolates, we do not observe increased tolerance for any isolate for the antibiotics tested, and only a single isolate displayed a high persistence phenotype. This study provides a basis for obtaining quantitative definitions of isolates showing high degrees of tolerance and/or persistence.