Discovery and Design of Thermostable PET-Degrading Enzymes

Poly(ethylene terephthalate), or PET, is an aliphatic-aromatic copolymer of ethylene glycol and terephthalic acid. More than 80 million tons of PET-based plastic products are produced on an annual basis, 72% of which are not recycled and instead are destined for landfills or incineration. Furthermore, most PET recycling facilities require relatively pure feedstreams that may not be feasible. PETases—enzymes that can readily degrade the polymer into its oligomers and monomers—remediate challenges in PET waste management and offer a sustainable way to biodegrade the polymer without releasing carbon dioxide. Targeted PET degradation past the polymer’s glass transition temperature (T_g = 75°C), where amorphous regions are readily accessible to catalysts, is achievable with thermostable enzymes.

Bioinformatics was used to discover naturally occurring PETase candidates, while computational and structural insights guided the engineering of thermostable PET-degrading enzymes. Homologs from the three classical PETases—hydrolase from P. sakaiensis, cutinase from T. fusca, and leaf compost cutinase from an unknown organism—were scanned for their species of origin. Mass sequence alignment was used to confirm the presence of a hydrolytic catalytic triad. Consequently, several PETase candidates from organisms living at up to 65°C were found. Structural analysis with both AlphaFold and crystal structures guided the installation of disulfide bonds to increase enzyme thermostability. The resulting enzymes retained activity beyond 90°C. Subsequently, MutCompute, a convolutional neural network that was trained on crystal structures from the Protein DataBank, guided single-point enzyme mutagenesis. Variants of Leaf Compost Cutinase (LCC) were created from MutCompute recommendations. When compared with the wild-type, it was found that the top MutCompute result, LCC^L66A, resulted in an increase in PET oligomer degradation at 70°C and 80°C by 21% and 55%, respectively; LCC^L66A also released 120% more PET monomers when incubated with PET at 37°C. The second most favorable variant predicted by MutCompute, LCC^N246T, had little activity at most temperatures but outcompeted wild-type LCC at and above the glass transition temperature. Structural insights validated the MutCompute recommendations; L66A reduces the hydrophobicity of a solvent-accessible microenvironment, while N246T improves substrate docking in the active site. The results emphasize the importance of using both quantitative models, such as neural networks and bioinformatics, as well as structural insights for the generation and validation of beneficial enzyme mutations.