Multidomain enzymes orchestrate two or more catalytic activities to carry out metabolic transformations with increased control and speed. Here, we report the design and development of a genome-mining approach for targeted discovery of biochemical transformations through the analysis of co-occurring enzyme domains (CO-ED) in a single protein. CO-ED was designed to identify unannotated multifunctional enzymes for functional characterization and discovery based on the premise that linked enzyme domains have evolved to function collaboratively. Guided by CO-ED, we targeted an unannotated predicted ThiF–nitroreductase di-domain enzyme found in more than 50 proteobacteria. Through heterologous expression and biochemical reconstitution, we discovered a series of natural products containing the rare oxazolone heterocycle and characterized their biosynthesis. Notably, we identified the di-domain enzyme as an oxazolone synthetase, validating CO-ED-guided genome mining as a methodology with potential broad utility for both the discovery of unusual enzymatic transformations and the functional annotation of multidomain enzymes.

多域酶协调两个或多个催化活动，以提高控制和速度进行代谢转化。在这里，我们报告了一种基因组挖掘方法的设计和开发，用于通过分析单个蛋白质中的共现酶域 (CO-ED) 来有针对性地发现生化转化。CO-ED 旨在识别未注释的多功能酶，以进行功能表征和发现，其前提是连接的酶域已经进化为协同发挥作用。在 CO-ED 的指导下，我们针对在 50 多种变形菌中发现的未注释的预测 ThiF-硝基还原酶双结构域酶。通过异源表达和生化重组，我们发现了一系列含有稀有恶唑酮杂环的天然产物，并对它们的生物合成进行了表征。值得注意的是，我们将双结构域酶鉴定为一种恶唑酮合成酶，验证了 CO-ED 引导的基因组挖掘是一种具有潜在广泛用途的方法，可用于发现不寻常的酶促转化和多结构域酶的功能注释。