Using Evolutionary Statistics to Understand Cellular Systems

Metabolic enzyme function is dependent on the larger context of a biochemical pathway. Despite detailed characterization of the requisite molecular "parts," it remains difficult to predict the adaptive response to a simple perturbation. That is: if the activity or expression of a single enzyme is changed, what other proteins (if any) require compensatory mutation? Comparative genomics and experimental evolution provide two powerful approaches to begin addressing these questions. In my thesis work, I examined adaptive interactions with the essential enzyme dihydrofolate reductase (DHFR). Analyses of gene synteny and co-occurrence across 1445 bacterial genomes indicated that DHFR coevolves with thymidylate synthase (TYMS), but is relatively decoupled from the rest of the folate metabolic pathway (and genome). Through directed evolution of E. coli, I demonstrated that these two enzymes adapt cooperatively in response to antibiotic stress. An allele replacement experiment confirmed that a pair of mutations to DHFR and TYMS were sufficient to reconstitute the entire trimethoprim resistance phenotype, establishing that the two enzymes are capable of independently driving adaptation. In the final component of my thesis, I drew on the 'mirror-tree' method to define a new measure of residue-residue coevolution which corrects for the phylogenetic relationship among species. In summary, my results verify that small groups of genes within larger metabolic pathways can form adaptive modules that evolve as a unit in response to environmental or mutational stress. Moreover, my mirror-tree inspired analysis provides a path forward for understanding how coupled adaptation between genes manifests at the resolution of site specific constraints on the protein sequence.