Our work using a new strategy to transform ubiquitin from a native binder/substrate to a specific tight binding inhibitor was recently accepted in the Journal of Molecular Biology. The paper “What Strengthens Protein-Protein Interactions: Analysis and Applications of Residue Correlation Networks” is now available online. This work is an international collaboration with the Chang group at the University of California, Riverside. During the COVID pandemic, we had frequent online meetings, exchanged ideas and data, and prepared the manuscript. We are excited to share this paper.

It has been shown that native biological protein-protein interactions can be further improved by mutating the key residues in the binding pocket. How to do so is not easy. We previously use AI tools to present a solution to enhance protein-protein interactions. Here we used an integrated approach to perform rational design. We chose the MERS-CoV papain-like protease (PLpro) as a study system to validate this theory. This viral PLpro captures ubiquitin (Ub) and cleaves the poly-Ub chains. We aim to mutate and transform Ub to an ubiquitin variant (UbV). The variants compete with Ub and occupy the PLpro binding cavity. Therefore, PLpro is catalytically inhibited by UbVs.

We generated residue-correlation networks to identify sidechain dihedral correlations between two proteins (MERS PLpro and Ub). Our analysis suggested altering the sidechain charge state, bulkiness, or hydrophobicity could significantly stabilize the interactions. We selected to mutate three key residues, A46, K48, and E64, to Phe, Glu, and Tyr, respectively. The A46F single mutation improves the binding affinity by ~15 times compared to wild-type Ub. The triple mutant (A46F-K48E-E64Y, called UbV3) further demonstrated a super tight binder with its K_d of 2.77 nM (Ub is 40.75 µM). To avoid the cell using UbV3 for native ubiquitination, we further mutated the C-terminal R74-G75 to N74-S75. Thus, the ubiquitin can not be activated by the E1 enzyme. This 5-point mutant (UbV5, A46F-K48E-E64Y-R74N-G75S) is our best inhibitor for MERS PLpro as the IC₅₀ is 9.7 nM and the K_d is 1.48 nM. The improvement of K_d from 40.75 µM to 1.48 nM is greater than 27,500 times.

This work is a beautiful fruit through the close collaboration between computational biologists and experimental biochemists. We demonstrated the analysis and application of the residue-correlation network. It can be broadly applied to other critical biological systems. The two first-authors, David Hung and Yun-Jung Hsieh, are PhD students at UC Riverside and National Taiwan University, respectively. They are the key persons to achieve the results.