A single residue determines substrate preference in benzylisoquinoline alkaloid N-methyltransferases

N-methylation is a recurring feature in the biosynthesis of many plant specialized metabolites, including alkaloids. A crucial step in the conserved central pathway that provides intermediates for the biosynthesis of benzylisoquinoline alkaloids (BIAs) involves conversion of the secondary amine (S)-coclaurine into the tertiary amine (S)-N-methylcoclaurine by coclaurine N-methyltransferase (CNMT). Subsequent enzymatic steps yield the core intermediate (S)-reticuline, from which various branch pathways for the biosynthesis of major BIAs such as morphine, noscapine and sanguinarine diverge. An additional N-methylation yielding quaternary BIAs is catalyzed by reticuline N-methyltransferase (RNMT), such as in the branch pathway leading to the taxonomically widespread and ecologically significant alkaloid magnoflorine. Despite their functional differences, analysis of primary sequence information has been unable to accurately distinguish between CNMT-like and RNMT-like enzymes, necessitating laborious in vitro screening. Furthermore, despite a recent emphasis on structural characterization of BIA NMTs, the features and mechanisms underlying the CNMT-RNMT functional dichotomy were unknown. We report the identification of structural variants tightly correlated with function in known BIA NMTs and show through reciprocal mutagenesis that a single residue acts as a switch between CNMT- and RNMT-like functions. We use yeast in vivo screening to show that this discovery allows for accurate prediction of activity strictly from primary sequence information and, on this basis, improve the annotation of previously reported putative BIA NMTs. Our results highlight the unusually short mutational distance separating ancestral CNMT-like enzymes from more evolutionarily advanced RNMT-like enzymes, and thus help explain the widespread yet sporadic occurrence of quaternary BIAs in plants. While this is the first report of structural variants controlling mono-versus di-methylation activity among plant NMT enzymes, comparison with bacterial MT enzymes also suggests possible convergent evolution.