PROJECT SUMMARY (30 lines) Adaptive prediction (AP) is a strategy utilized by all organisms to predict and prepare for a future selective pressure. E. coli and M. tuberculosis (MTB), for instance, utilize neutral cues such as a rise in temperature or nutrient starvation to prepare in advance for a hostile host environment. There is growing evidence that the drug/immune tolerant phenotype resulting from AP gives pathogens a window of opportunity to evolve antimicrobial resistance (AMR)—a catastrophic problem that could cause >10 million deaths by 2050. Knowledge of how AP is encoded within the genome and gene networks of an organism will enable strategies to disrupt and prevent drug tolerance to potentiate complete killing by frontline drugs. We’ve demonstrated proof-of-concept for this strategy by potentiating bedaquiline killing of MTB through rational disruption of the starvation-induced, bedaquiline-specific tolerance network with a second drug—pretomanid (Peterson et al, Nature Micro 2016). To further advance this approach, we established a laboratory evolution framework to dissect dynamics and mechanisms of AP (Lomana et al, Genome Biol Evol 2017). Using this set up we have demonstrated that when subjected to laboratory evolution in an artificially structured environment, novel AP emerges within 50 generations to enable Saccharomyces cerevisiae (yeast) to use caffeine as a cue to anticipate and elicit a protective response to subsequent challenge with a sub-lethal dose of 5-fluoroorotic acid. Based on evolutionary dynamics, genetic variation, and phenotypic heterogeneity of evolved lines, we hypothesize that three factors govern emergence and retention of AP: (1) cost vs. benefit of AP vis-à-vis frequency and predictability of coupled environmental changes, including period between exposures, energy required for advanced preparedness, and overall fitness benefit; (2) coordinated changes in metabolic and regulatory networks to adaptively trigger a tolerant state upon sensing a cue; and (3) evolutionary game strategies (bet-hedging) arising from population heterogeneity. The two specific aims to test these hypotheses will make use of a systems approach to study and manipulate complex phenotypes, including, (i) an integrated network model for predicting phenotypic consequences of regulatory and metabolic mutations; (ii) a technology for phenotyping >10,000 colonies, (iii) a technology to sort translationally active and dormant sub-populations; and (iv) laboratory evolution and genome engineering capabilities to generate and manipulate AP. Through iterative computational prediction and experimentation, we will characterize how structure and dynamics of environmental change influences emergence and retention of AP (Aim 1); and elucidate and rationally manipulate interplay of metabolic, regulatory, and evolutionary game strategies for AP (Aim 2). This project will advance theory of AP with implications on strategies to preempt AMR; advance tools to predict and manipulate complex phenotypes; and track and isolate rare strains within heterogeneous populations. 1

PROJECT NARRATIVE Adaptive prediction (AP) is a strategy utilized by organisms from all domains of life to predict and prepare in advance for a future selective pressure. There is growing evidence that the drug/immune tolerant phenotype resulting from AP gives pathogens a window of opportunity to select mutations and evolve antimicrobial resistance—a catastrophic problem that is projected to cause up to 10 million deaths by 2050. Knowledge of how AP is encoded within the genome and gene networks of an organism will enable strategies to disrupt and prevent drug tolerance to potentiate complete killing by frontline drugs. 1