Machine Learning and Metabolic Model Guided CRISPRi Reveals a Central Role for Phosphoglycerate Mutase in Persistence.

Upon nutrient starvation, serovar L2 (CTL) shifts from its normal growth to a non-replicating form, termed persistence. It is unclear if persistence is an adaptive response or lack of it. To understand that transcriptomics data were collected for nutrient-sufficient and nutrient-starved CTL. Applying machine learning approaches on transcriptomics data revealed a global transcriptomic rewiring of CTL under stress conditions without having any global stress regulator. This indicated that CTL's stress response is due to lack of an adaptive response mechanism. To investigate the impact of this on CTL metabolism, we reconstructed a genome-scale metabolic model of CTL (CTL278) and contextualized it with the collected transcriptomics data. Using the metabolic bottleneck analysis on contextualized CTL278, we observed phosphoglycerate mutase () regulates the entry of CTL to the persistence. Later, was found to have the highest thermodynamics driving force and lowest enzymatic cost. Furthermore, CRISPRi-driven knockdown of and tryptophan starvation experiments revealed the importance of this gene in inducing persistence. Hence, this work, for the first time, introduced thermodynamics and enzyme-cost as tools to gain deeper understanding on CTL persistence.