Simulation-guided engineering of split GFPs with efficient beta-strand photodissociation

Green fluorescent proteins (GFPs) are ubiquitous for protein tagging and live-cell imaging. Split- GFPs are widely used to study protein-protein interactions by fusing proteins of interest to split GFP fragments that create a fluorophore upon complementation. Complementation is typically irreversible, and controlled dissociation of the fragments would be desirable. The quantum efficiency of light-induced photodissociation of split GFPs is low, with extensive mutagenesis and screening using traditional protein engineering approaches proving difficult to implement. To reduce the search space, key states in the dissociation process were modeled by combining classical and QM/MM molecular dynamics and enhanced sampling methods, enabling the rational design and engineering of split GFPs with up to 20-fold faster photodissociation rates using non-intuitive amino acid changes. This demonstrates the feasibility of modeling complex molecular processes using state-of-the-art computational methods, and the potential of integrating computational methods to increase the success rate in protein engineering projects.