Iterative computational and rational design generates diverse PET hydrolases for plastic depolymerization

Evolutionary relationships among homologous enzymes can be leveraged to introduce large numbers of amino acid substitutions for improved functional properties. Here, we combine evolution-informed design (EVdesign) with rational design to engineer improved poly(ethylene terephthalate) (PET) hydrolases (PETases). Designs derived from multiple PETases reveal strong parent-dependent differences in success and performance, with Thermobifida fusca Tf Cut2 emerging as the most design-amenable scaffold. Increasing sequence diversification reduces the number of active enzymes but broadens performance variance, enabling discovery of high-performing outliers. Four rational mutations overcome a thermostability bottleneck in Tf Cut2 and this variant provides a robust scaffold for iterative EVdesign, which further improves activity. Overall, we experimentally test 603 designs from eight parent PETases. More than 400 variants containing 12-68 substitutions were able to depolymerize PET, nearly all outperforming their parent enzymes, and many approach benchmark LCC-ICCG performance in partial conversion assays. These results establish EVdesign as a robust strategy for generating diverse, high-activity enzymes.