Computer-aided directed evolution ofl-threonine aldolase for asymmetric biocatalytic synthesis of a chloramphenicol intermediate.

l-Threonine aldolases (LTAs) employing pyridoxal phosphate (PLP) as cofactor can convert low-cost achiral substrates glycine and aldehyde directly into valuable β-hydroxy-α-amino acids such as (2R,3S)-2-amino-3-hydroxy-3-(4-nitrophenyl) propanoic acid ((R,S)-AHNPA), which is utilized broadly as crucial chiral intermediates for bioactive compounds. However, LTAs' stereospecificity towards the β carbon is rather moderate and their activity and stability at high substrate load is low, which limits their industrial application. Here, computer-aided directed evolution was applied to improve overall activity, selectivity and stability under desired process conditions of a l-threonine aldolase in the asymmetric synthesis of (R,S)-AHNPA. Selectivity and stability determining regions were computationally identified for structure-guided directed evolution of LTA-variants under efficient biocatalytic process conditions using 40% ethanol as cosolvent. We applied molecular modeling to rationalize selectivity improvement and design focused libraries targeting the substrate binding pocket, and we also used MD simulations in nonaqueous process environment as an effective and promising method to predict potential unstable loop regions near the tetramer interface which are hot-spots for cosolvent resistance. An excellent LTA variant EM-ALDO031 with 18 mutations was obtained, which showed ∼ 30-fold stability improvement in 40% ethanol and diastereoselectivity (de) raised from 31.5% to 85% through a three-phase evolution campaign. Our fast and efficient data-driven methodology utilizing a combination of experimental and computational tools enabled us to evolve an aldolase variant to achieve the target of 90% conversion at up to 150 g/L substrate load in 40% ethanol, enabling the biocatalytic production of β-hydroxy-α-amino acids from cheap achiral precursors at multi-ton scale.