Determining quaternary allostery of the spastin motor through bioinformatics and graph networks

Microtubule severing enzymes are multimeric nanomachines that are essential for regulating the cytoskeleton of a cell. Spastin is one of these motors that is found in particularly high concentration in neurons. Key loss of function mutations in this motor have been associated with human disease. Thus, its function and dynamics are intricately connected and necessary to understand. We set out to probe this connection by focusing on allosteric networks in the quaternary states of spastin. While recent approaches, which take advantage of evolutionary information and machine learning methods, investigated allostery in single chain proteins, only few studies have probed large multimeric proteins. We employed a coevolutionary mutual information analysis, integrated with our molecular dynamics simulations of the bound and unbound ligand states, and we built a mathematical graph to visualize the allosteric networks based on learned parameters. We found that in addition to other allosteric regions previously identified in experiments, one of the central pore loops is a crucial hot spot. Moreover, we found a novel allosteric region in the helical bundle domain, important for nucleotide binding and protein-protein binding between protomers. These findings establish the power of our approach in characterizing the allostery of quaternary structures.