Abstract
Pathogen evolution of drug resistance often occurs in a stepwise manner via accumulation of multiple mutations, which in combination have a non-additive impact on fitness, a phenomenon known as epistasis. Epistasis complicates the sequence-structure-function relationship and undermines our ability to predict evolution. We present a computational method to predict evolutionary trajectories that accounts for epistasis, using the Rosetta Flex ddG protocol to estimate drug binding free energy changes upon mutation and an evolutionary model based in thermodynamics and statistical mechanics. We apply this method to predict evolutionary trajectories to known multiple mutations associated with resistant phenotypes in malaria. Resistance to the combination drug sulfadoxine-pyrimethamine (SP) in malaria-causing species Plasmodium falciparum (Pf) and Plasmodium vivax (Pv) has arisen via the accumulation of multiple point mutations in the DHFR and DHPS genes. Four known PfDHFR pyrimethamine resistance mutations are highly prevalent in field-isolates and multiple studies have shown epistatic interactions between these mutations determine the accessible evolutionary trajectories to the highly resistant quadruple mutation N51I,C59R,S108N,I164L. We simulated the possible evolutionary trajectories to this quadruple PfDHFR mutation as well as the homologous PvDHFR mutations. In both cases, our most probable pathways agreed well with those determined experimentally. We also applied this method to predict the most likely evolutionary pathways to observed multiple mutations associated with sulfadoxine resistance in PfDHPS and PvDHPS. This novel method can be applied to any drug-target system where the drug acts by binding to the target.
Significance Statement Antimicrobial resistance (AMR) is a major public health threat, resulting from the overuse and misuse of antimicrobial drugs. Our ability to monitor emerging resistance and direct the most appropriate treatment strategy (stewardship) would be strengthened by the development of methods to accurately predict the mutational pathways that lead to resistance. We present a computational method to predict the most likely resistance pathways that result from mutations that alter drug binding affinity. We utilized the Rosetta Flex ddG protocol and a thermodynamic evolutionary model. This novel approach can accurately capture known resistance trajectories and can be applied to any system where resistance arises via changes in drug binding affinity.
Competing Interest Statement
The authors have declared no competing interest.
Footnotes
Competing Interest Statement: None.
