De novo biosynthesis of eriocitrin in Saccharomyces cerevisiae through deep learning-guided enzyme screening and systematic metabolic engineering

Eriocitrin, a flavanone-7-<i>O</i>-disaccharide known for its antioxidant and anti-inflammatory properties, holds considerable promise for use in functional foods and pharmaceuticals. However, its large-scale production is constrained by the limitations of conventional plant extraction, including low abundance of active compounds and seasonal variability. Here, we established a <i>de novo</i> biosynthetic pathway for eriocitrin from glucose in engineered <i>Saccharomyces cerevisiae</i> through combinatorial metabolic engineering. First, to verify the feasibility of glycosylation in the eriocitrin synthesis process, two highly efficient flavonoid-7-<i>O</i>-glucosyltransferases were identified using an integrated bioinformatics and deep learning approach. Subsequently, through the reconstruction of the UDP-rhamnose synthesis pathway and enhancement of the key glycosylation precursor UDP-glucose supply, an eriocitrin titer of 147.9 mg/L was achieved following supplementation with 300 mg/L eriodictyol. To further achieve the <i>de novo</i> synthesis of eriocitrin, an eriodictyol-producing strain was developed by screening the optimal F3'H/CPR pair, engineering promoters, and pathway integration, achieving an eriodictyol titer of 237.8 mg/L. Finally, combining the eriodictyol-producing chassis strain with the previously optimized glycosylation module enabled the <i>de novo</i> production of eriocitrin, reaching a titer of 30.5 mg/L. This work demonstrates a proof-of-concept for microbial production of flavonoid disaccharides, providing a foundation for a sustainable and scalable supply of eriocitrin.