Abstract
<title>Abstract</title> <p> Colibactin is a microbiome-derived genotoxin that damages the genomes of intestinal epithelial cells and has been linked to colorectal cancer. Its biosynthesis is encoded by a large gene cluster ( <italic>pks</italic> / <italic>clb</italic> ) that functions as a hybrid nonribosomal peptide synthetase - polyketide synthase assembly line, yet the structural basis of its assembly remains poorly understood. Here we show cryo-electron microscopy structures of colibactin biosynthetic enzyme ClbB that uncover a noncanonical architecture for a fully-reducing polyketide synthase module. Unlike metazoan fatty acid synthases and iterative polyketide synthases, ClbB arranges its condensing and reducing regions in a crossed configuration. Within its reducing region, the enoylreductase forms an inverted dimer and the dehydratase is positioned peripherally as monomers. Moreover, the dehydratase-enoylreductase unit is embedded within the ketoreductase fold. Bioinformatic analyses combined with deep-learning-based structure prediction indicate that this domain-embedding logic is broadly distributed across diverse bacterial phyla. These findings define a previously unrecognized class of reducing polyketide synthase modules and provide an expanded structural framework for understanding how modular polyketide synthases coordinate multi-step chemistry. The work also extends design principles for engineering megasynthases for chemical biosynthesis and synthetic biology. </p>