AI-driven engineering of EgtD enabling high-efficiency ergothioneine synthesis with a multi-enzyme cascade

Ergothioneine (ERG), a sulfur-containing natural antioxidant with significant biomedical potentials, has long been limited by low productivity in microbial fermentation. Here, we report the first high-efficiency in vitro reconstruction of a multi-enzyme cascade for ERG biosynthesis. To address the rate-limiting histidine methylation step, we employed a synergistic strategy integrating machine learning-based kinetic prediction (CataPro, DLkcat), molecular dynamics simulations, and conformational dynamics analysis to guide site-directed mutagenesis of Mycolicibacrterium smegmatis EgtD. This yielded a triple mutant M3 (L53A/T59S/E282S) with 3.4- and 2.8-fold enhancements in catalytic efficiency (k_cat/K_m) toward histidine and S-adenosyl-L-methionine (SAM), respectively. Mechanistically, M3 enhances catalysis by reducing the SAM-histine distance, optimizing the catalytic angle, eliminating non-productive hydrogen bonds, and unlocking a secondary ligand tunnel through a mutation-induced gating mechanism. When coupled with a SAM-regeneration module, the engineered cascade enables preparative-scale ERG synthesis, reaching a titer of 6.05 g·L^-1 within 14.5 h and productivity of 10.01 g·L^-1d^-1-the highest report to date. This work showcases a structure-guided, AI-assisted approach that overcomes evolutionary constraints, unlocks latent reactivity, and accelerates the development of efficient biocatalytic cascades for high-value compound synthesis.