Combining Evolutionary Steering and Coevolutionary Phage Training to Generate Predictable Phage-Antibiotic Synergy

The rapid rise of multidrug-resistant bacterial pathogens has renewed interest in bacteriophage therapy, yet bacterial resistance to phages can evolve rapidly. Two evolutionary strategies have been proposed to address this challenge: coevolutionary phage training, which adapts phages to anticipated resistance mechanisms, and phage-antibiotic synergy, which exploits tradeoffs between phage resistance and antibiotic susceptibility. Here, we show that these strategies intersect in complex but predictably exploitable ways. Using Escherichia coli and bacteriophage λ, we demonstrate that resistance to a trained-but not an untrained-phage generates collateral sensitivity to erythromycin and rifampicin, but not to five other antibiotics. This training-induced synergy is mechanistically linked to repeated disruption of lpcA, a gene involved in lipopolysaccharide core biosynthesis. Leveraging this mechanistic insight, we successfully predicted that novobiocin and rifapentine would also exhibit collateral sensitivity with a trained λ phage. We further show that incorporating sub-lethal erythromycin during phage training reshapes coevolutionary trajectories, increasing the frequency with which phages evolve the ability to use a second host receptor and thereby enhancing bacterial suppression. Together, these results reveal that despite the inherent complexity of phage-bacteria coevolution, key outcomes can be anticipated and directed. Our findings highlight the promise of evolution-informed, predictive strategies for designing therapies.