Machine-Learning-Guided Discovery of Cytochrome P450 Enzymes for Bioproduction of Jolkinolides and Other...

Plant biosynthetic genes often scatter in plant genomes. Cytochrome P450 enzymes (CYPs) play pivotal roles in the oxidation of plant terpenoids, such as hydroxylation and oxidation-mediated formation of other functional groups. Hundreds of CYP genes are usually present in a plant genome, which results in a great bottleneck for accurately identifying target CYPs. Jolkinolides, a group of labdane-related diterpenoids (LRDs), are the characteristic metabolites with potent anticancer activities in Euphorbia herbs, featuring an α,β-unsaturated γ-lactone ring as their vital pharmacophore. Herein, we constructed a model under a supervised paradigm to predict the compatible pairs of CYPs and LRD olefins through navigating both sequence and substrate spaces, which dramatically increases the hit rate. Ten CYPs involved in LRD biosynthesis were characterized from Euphorbia fischeriana Steud., including the ones within five subfamilies (CYP82BU, CYP80C, CYP71BF, CYP82J, and CYP71AN) in which no members have hitherto been reported to be able to oxidize LRD scaffolds. Their discovery enables enzymatic oxidation at 12 sites of four tested LRD olefins for the first time. Notably, the combination of EfCYP71BF25 and EfCYP82BU7 results in the formation of an α,β-unsaturated γ-lactone ring. By coexpressing them with a C-3 oxidase EfCYP76A211 in Nicotiana benthamiana, two minor jolkinolides (helioscopinolides A and E) with potent anticancer activities were produced, and after metabolic optimization, their titers reached 331.0 and 102.2 mg/kg weight of plant material, 103.3 and 30.1 folds of their amounts in E. fischeriana, respectively. These findings provide a high-performance approach for identifying CYPs from plant genomes and enrich the enzymatic catalysts for bioproduction of jolkinolides and other valuable LRDs.