Identification and modification of amine dehydrogenases by genomic and structural approaches for biocatalytic amine synthesis

Facing the great demand of amine intermediates in pharmaceutical, agrochemical as well as in fine and bulk chemical industries, new sustainable processes for amine synthesis need to be developed considering the current economic and environmental challenges. Biocatalysis, i.e. the use of enzymes in synthesis, represents a greener alternative to traditional chemical processes. It enables to work catalytically in mild conditions and generally with shorter reaction times thanks to the high chemo- and stereoselectivity of enzymes. First mainly limited to lipases and transaminases, the biocatalytic alternatives for amine synthesis have expanded since the 2010s with the discovery of enzymes catalyzing the reductive amination of carbonyl-containing compounds. These solutions encompass several enzyme families such as opine dehydrogenases, imine reductases, reductive aminases (RedAms) and amine dehydrogenases (AmDHs). The latter, previously restricted to engineered amino acid dehydrogenases, has grown in recent years thanks to the identification of genes coding for native AmDHs (nat-AmDHs) that catalyze the imine formation from a carbonyl compound and ammonia and its subsequent reduction by NAD(P)H cofactor. These enzymes harbor a high (S)-enantioselectivity rarely obtained from other enzyme families catalyzing the same reaction. The nat-AmDH family has been extended over the years through several iterations of (meta)genome mining but still lacked in active site diversity. The studies described in this research work aimed at filling this missing diversity using structure-guided approaches based on homology models or 3D structures obtained from an external collaborative work. First, a deep structural analysis of the family members coupled with targeted protein engineering highlighted the importance of some residues in the active site. The family was furthered characterized for reductive amination of still undescribed key substrates such as short (hydroxy)ketones or aromatic aldehydes. Some approaches were carried out to expand the substrate scope of the native AmDHs, restricted to short carbonyl substrates. One strategy relied on the in silico identification of two bulky residues highly conserved within the family which were considered as hot spots for the active site expansion. These two residues were mutated in ten nat-AmDHs bearing similar scaffolds but slightly different active sites. The majority of the mutants were active towards still unreached substrates such as aldehydes from hexanal (non-optimized conversions up to 99.5%) to nonanal (44.5%), ketones from hexan-2-one (39.5%) to nonan-2-one (16%) and 4-phenylbutan-2-one (52.5%) while still retaining their high (S)-enantioselectivity (>90% ee). The best hit CfusAmDH-W145A was furthered characterized and used in a 100-mg scale synthesis of (2S)-octan-2-amine and heptan-1-amine. Secondly, an approach of continuous in vivo evolution using automated devices available at the Genoscope was attempted to enhance the activity of some nat-AmDHs and AspRedAm. The selection screens were based on the activity of these enzymes in the oxidative deamination way that was hypothetically too low to enable sufficient initial cells growth thus preventing good starting point for evolution. Eventually, within the scope of the MODAMDH project to find AmDHs with high sequence diversity among 2.6 billions sequences, the family was drastically extended from 2,011 members to 17,039 members. Thus, taking advantage of all the structural and experimental data collected on the characterized nat-AmDHs, a promising structure-guided selection of 92 new enzymes harboring specific features has been done. Thanks to this work, we have a nearly complete structural picture and understanding of the nat-AmDHs diversity that will help further improvements and applications in synthetic chemistry.