Structure-Based Discovery and Machine Learning-Aided Design of a Serine Protease for Biocatalytic Recycling of Polylactic Acid

Polylactic acid (PLA) is a leading biobased plastic, yet its slow degradation in natural environments hinders the development of a circular bioeconomy. While enzymatic recycling of plastics like poly(ethylene terephthalate) has advanced significantly, the discovery of robust and efficient PLA depolymerases suitable for industrial applications remains a critical challenge. While cutinase-type polyester hydrolases are frequently studied for plastic deconstruction, our findings reveal that thermophilic subtilisin-type serine proteases serve as superior scaffolds for PLA depolymerization. Utilizing a workflow combining structure-based mining with machine learning-guided engineering, a serine protease thermitase (designated TvTH-L01) with PLA depolymerization activity was identified, which was subsequently engineered into a robust quadruple variant, L4-1. The variant showed a 2.48-fold increase in depolymerization efficiency over the wild type and was capable of 90–98% conversion of different commercial PLA products into monomeric lactic acid at a high loading of 50–100 g·L–1 within 48 h. Molecular dynamics simulations and structural modeling further elucidate that the structural mutations successfully stabilize the catalytic conformation and remodel the substrate-binding pocket. Overall, this work establishes a generalizable discovery-to-engineering paradigm, positioning serine proteases as a versatile scaffold for the biocatalytic recycling of aliphatic polyester plastics.