Predicting evolutionary dynamics of multi-drug resistance

There is an urgent need to develop novel approaches to halt the evolution and spread of antimicrobial resistance. Combination therapies (multiple drugs given as a single prescription) are promising, both for preventing resistance and for optimising treatments for specific infections. However, recent experimental work has shown that combination therapies can select for multi-drug resistance in conditions experienced by natural microbial populations (e.g. temporally-varying drug concentrations and elevated mutation rates).

To establish whether combination therapies are a viable strategy, we need predictive models for how well drug combinations prevent resistance under real-world conditions where antibiotics are present (including infection and agriculture). A combined modelling and experimental approach is required because testing more than a handful of drugs is a considerable logistical challenge-5 antibiotics screened at 10 doses requires nearly 10 million growth assays, beyond the limits of high-throughput technologies. However, models need to account for the basic biology of microbial growth under temporally-varying antibiotic levels, which requires experimental measurement. Model predictions will be validated by experimentally exposing bacterial populations to the best and worst identified combinations to see if multi-drug resistance evolves. This work is crucial for establishing combination therapies as a viable solution to the antibiotic resistance crisis.

The aim of this project will be to combine mathematical modelling and experimental approaches to understand the origin and evolution of multi-drug resistance evolution under biologically-realistic conditions. The programme of work will follow three lines of investigation.

I will first develop a stochastic model of multi-drug resistance evolution incorporating in-host pharmacokinetics and bacterial growth models. I will investigate factors leading to environments with sub-inhibitory antibiotic concentrations, such as differences in pharmacokinetic profiles, or as during the use of antibiotics as growth promoters in animals. This will be modelled by varying bacterial growth rates as a function of time. This will create predictions for the probability that multi-drug resistance can evolve during the course of combination therapy.

I will parametrise the model by developing a predictive framework for epistatic interactions between arbitrary combinations of resistance mutations, using high-throughput transcriptomics together with growth rate and competitive fitness assays. This will allow us to model the effects of multi-drug resistance on bacterial fitness, without needing to exhaustively screen all possible combinations.

Finally, predictions from the model will be tested by experimentally evolving E. coli populations in environments with multiple antibiotics. Combinations will be evaluated for their ability to reduce the likelihood of multi-drug resistance evolving.