🚀David Baker preprint from @UWproteindesign 👇👇 What makes cutting protein bonds so challenging that nature evolved specialized enzymes with metal cofactors to accomplish it?@biorxivpreprint @ETH_en "Computational design of cysteine proteases" • Amide bonds in proteins have half-lives of hundreds of years under physiological conditions, making them far more stable than ester bonds (with amine leaving groups having pKa >35 versus <8 for activated esters), and previous computational enzyme design has only succeeded with activated small molecule ester substrates rather than the energetically demanding peptide bond cleavage required for proteases. • The researchers used RoseTTAFold Diffusion 2 for Molecular Interfaces (RFD2-MI) to design zinc proteases from minimal catalytic motifs, constructing an ideal active site with five functional groups (three zinc-binding residues H1, H2, E1, one catalytic base E2, and one oxyanion-stabilizing tyrosine Y) based on aminopeptidase N and astacin structures, performing two-sided design of both protease and substrate sequences to ensure precise positioning of the target amide bond. • Of 135 designs tested in a single design round, 36% showed activity (14.7% for Zn-only models and 87.5% for Zn-water models), with all active designs cleaving precisely at intended sites confirmed by mass spectrometry; the most active design (Zn45 cleaving ZnO36 substrate) achieved kcat of 0.025 ± 0.002 s⁻¹, KM of 26 ± 5 μM, and kcat/KM of 900 ± 200 M⁻¹s⁻¹, representing >10⁸-fold rate acceleration over uncatalyzed hydrolysis; designs exhibited zinc binding with Kd between 10⁻¹⁰ and 10⁻⁸ M, showed substrate specificity in 4 of 5 scaffolds, and could be reprogrammed to cleave disease-relevant human TDP-43 protein with 4 variants achieving ≥80% cleavage at 5 hours. Authors: Hojae Choi et. al Donald Hilvert, Samuel J. Pellock, David Baker Link: