An Effective Computational Strategy for UGTs Catalytic Function Prediction

The GT-B type glycosyltransferases play a crucial post-modification role in synthesizing natural products, such as triterpenoid and steroidal saponins, renowned for their diverse pharmacological activities. Despite phylogenetic analysis aiding in enzyme family classification, distinguishing substrate specificity between triterpenoid and steroidal saponins, with their highly similar cyclic scaffolds, remains a formidable challenge. Our studies unveil the potential transport tunnels for the glycosyl donor and acceptor in PpUGT73CR1, by molecular dynamics simulations. This revelation leads to a plausible substrate transport mechanism, highlighting the regulatory role of the N-terminal domain (NTD) in glycosyl acceptor binding and transport. Inspired by these structural and mechanistic insights, we further analyze the binding pockets of 44 plant-derived UGTs known to glycosylate triterpenes and sterols. Notably, sterol UGTs are found to harbor aromatic and hydrophobic residues with polar residues typically present at the bottom of the active pocket. Drawing inspiration from the substrate binding and product release mechanism revealed through structure-based molecular modeling, we devised a fast sequence-based method for classifying UGTs using the pre-trained ESM2 protein model. This method involved extracting the NTD features of UGTs and performing PCA clustering analysis, enabling accurate identification of enzyme function, and even differentiation of substrate specificity/promiscuity between structurally similar triterpenoid and steroidal substrates, which is further validated by experiments. This work not only deepens our understanding of substrate binding mechanisms but also provides an effective computational protocol for predicting the catalytic function of unknown UGTs.