Natural and experimental evolution as a novel tool to identify genetic mechanisms underlying the complex phenotype of multidrug tolerance

During the golden age of antibiotic discovery, the fight against bacterial infections was believed to be won. Indeed, the introduction of antibiotics has changed health-care by curing previously untreatable infectious diseases and allowing for more efficient prophylaxis. Yet, recently, the evolution of antibiotic resistance has become a serious danger. The problem is even worse as antibiotic tolerance in many bacteria not only relies on the build-up of resistance-conferring mutations but also on a peculiar phenotype called bacterial persistence. Cells within antibiotic-sensitive populations can phenotypically switch to an antibiotic-tolerant 'persister' state. By allocating part of its cells to this dormant persister state, a population spreads its risks. Persisters impede successful therapy outcome by causing relapse of infection and by catalyzing resistance development.

Despite the clinical relevance of persistence, almost no empirical data exist on how persistence evolves in the face of antibiotic treatment. Therefore, we tested evolutionary hypotheses proposed by risk-spreading theories. They predict that the extent of bet-hedging or allocation to dormant life stages evolves to inversely correlate with the length of good conditions. We used in vitro evolution experiments under intermittent antibiotic treatments to show that daily treatments with aminoglycosides quickly select for mutants with extremely high persister levels, up to 100%, both in a lab strain and a pathogenic Escherichia coli isolate. The increased tolerance also protects the cells better against treatment with another class of antibiotic, thereby demonstrating cross-tolerance, a hallmark of persistence.

Next, we examined the costs and benefits of this high-persistence phenotype. The extreme benefit of high persistence under daily antibiotic treatment appears to be also attributable to a better responding to environmental signals, e.g. a faster resumption of growth following treatment and dilution in fresh medium. Furthermore, although high-persistence mutants suffer from a cost in the absence of antibiotics as persisters are cells that do not grow, this cost is minimized, as the onset of persister generation upon entry into stationary phase, is delayed. This cost is however sufficient for a (partial) rewinding of evolution to decreased persister levels when high-persistence mutants are evolved in the absence of antibiotics. We developed a persistence-evolution model that predicts an inverse correlation between the allocation to persistence and the length of absence of antibiotic treatments, similarly to more general models on bet-hedging and persistence. Evolution experiments under different antibiotic treatment frequencies confirm the predicted correlation and result in genuinely intermediary persister levels.

Stress responses and toxin-antitoxin modules have been reported to be involved in persistence, and the tolerant nature of persisters has been attributed to either general dormancy, the inactivity of antibiotic-specific targets or a reduced antibiotic uptake. Still, persistence mechanisms remain poorly understood. For the high-persistence phenotype in this study, we identified five target genes that were not previously connected with persistence: oppB, nuoN, gadC, mscL and yaaU. The identified mutations are all gain-of-function mutations. Furthermore, all targeted genes encode membrane-spanning transporters or channels. We confirmed, however, that a reduced antibiotic uptake by a reduction in membrane potential, is not involved. Decreases in intracellular pH or ATP:ADP ratios were found in some of the mutants as was a downregulated translation activity. However, this is no general explication for the high persistence in all mutants. Also at the regulatory level, we found various responses in the mutants. Most of them depend critically on sigma factor RpoS and the stringent response and display a stationary phase-specific increase in persistence. RpoS and the stringent response play, however, no role in the nuoN mutant and for the yaaU mutant an increased persister level was even observed in exponential phase. Therefore, the high levels of persistence selected for under identical conditions, may be caused by different mechanisms. 

This thesis contributes to a better understanding of the evolutionary versatility of persistence and its underlying mechanisms and will eventually lead to an improvement of patients' outcome in the face of bacterial infection. Moreover, our results will impact future persistence research as well as research in the field of antibiotic resistance and bet-hedging theory.