The Genetic Landscape of a Metabolic Interaction

Interactions between proteins provide the basis for cells to perform metabolism, grow, divide, move, and appropriately respond to external stimuli. Because proteins do not act as independent entities, the genetic background influences the effect of a mutation in unexpected ways. This context-dependence of mutational effects is epistasis. Extensive progress has been made in our ability to identify epistasis between proteins. However, how the epistasis between a pair of proteins is distributed across the amino acid sequence is less clear. Previous work characterized this sequence-level epistasis between proteins that bind to form a physical complex. Until now, the structural pattern and magnitude of epistasis between pairs of mutations spanning interacting metabolic enzymes remained uncharacterized. In this work, we deeply examined the context dependence of mutations for two essential enzymes in the bacterial folate metabolic pathway, Dihydrofolate Reductase (DHFR) and Thymidylate Synthase (TYMS). To achieve this goal, we used deep mutational scanning assays on DHFR in the context of varying activities of TYMS. The result is a rigorous dataset with epistasis measurements over the entire amino acid sequence of DHFR. We found that the positions with the greatest magnitude of epistasis within the structure of DHFR lied at the active site. However, the sign of epistasis at the DHFR active site was dependent on whether TYMS was active. Beyond the active site, the distribution of positive epistasis among the positions of DHFR was also context- dependent on the state of TYMS. Therefore, we can think of the active site as a non-physical interface between protein pairs that do not form a physical complex but share an intermediate. The potential consequences of this dataset on the epistasis between DHFR and TYMS are profound. This dataset is fundamental towards our understanding of how epistasis mechanistically emerges in nonlinearities between catalytic activity in enzymes, protein abundance, and cellular growth rate. This experimental dataset is also necessary to credibly validate predictions of epistasis from models of statistical co-evolution.