Computational Design of Periplasmic Binding Protein Biosensors Guided by Molecular Dynamics

Abstract

Periplasmic binding proteins (PBPs) evolved to scavenge materials from the environment by binding substrate and delivering it to the host bacteria. PBPs are common scaffolds for biosensors: effector proteins, such as circularly permuted green fluorescent protein (cpGFP), can be inserted into a PBP sequence such that the effector protein’s output is changed by conformational changes upon PBP substate binding. The site of insertion for maximum output change is often determined by comparison of PBPapo/holocrystal structures, but random insertion libraries have shown that this method can miss the best sites. Here, we present a computational method for identifying insertion sites for cpGFP in the maltose binding PBP that creates functional biosensors. Our method, which uses protein contact analysis, identifies the best-known insertion regions, as well as additional regions that have not been identified previously. We experimentally characterised biosensors with insertions at these regions and found that 75% of designs were functional. This method is generalisable beyond this system and PBPs more generally, and so could be applied to other binding proteins or effector domains.

Allosteric modulation of fluorescence revealed by hydrogen bond dynamics in a genetically encoded maltose biosensor

“…16,17 Thus, instead of relying on the single protein conformations for interpreting the functioning of the biosensor, using dynamics to disclose the range of available motions is a more rational approach 18 and is already being used as a strategy in the computational design of GEFBs. 19 With the advent of AlphaFold predicted structures 20 , new designs will inevitably be increasingly proposed, and sampling the functionally relevant conformations of the predictions will also be an issue to consider. For example, a recent report revealed the working mechanism of a single FP based camp biosensor by using metadynamics molecular dynamics (MD) simulations to sample the apo form starting from the holo crystal structure.…”

“…Efficient sampling of protein conformations and disclosing allosteric regions then becomes a priority in the computational design process 16 . Thus, instead of relying on the single protein conformation for interpreting the functioning of the biosensor, using dynamics to disclose the range of available motions is a more rational approach 17 and is already being used as a strategy in the computational design of GEFBs 18 . With the advent of AlphaFold predicted structures, 19 new designs will inevitably be increasingly proposed, and sampling the functionally relevant conformations of the predictions will also be an issue to consider.…”