GoKawiilWe then set out to find the gene responsible for the conversion of 12 to the next known biosynthetic intermediate cinchonamimal 7. Hydroxylation of cinchonium would directly produce 13, the cyclic form of 7. Therefore, we searched for biosynthetic genes predicted to encode oxidases. All annotated cytochrome P450 enzymes and oxoglutarate-dependent dioxygenases were ranked on the basis of Pearson’s correlation coefficients with MCC using the snRNA-seq data (Extended Data Fig. 2a). A candidate annotated as a 2-oxoglutarate-dependent dioxygenase was confirmed to catalyse this transformation in both N. benthamiana and in vitro enzymatic assays, and was subsequently designated as cinchonaminal synthase (CiS) (peak A, Extended Data Fig. 2b,c and Supplementary Fig. 21a). Notably, CiS further oxidized cinchonaminal to form a compound with spectral data consistent with the carboxylic acid (herein named as cinchonaminic acid (15); peak B, Extended Data Fig. 2b–d). This compound was not observed in Cinchona, which indicated that the next pathway intermediate rapidly intercepts the initial enzymatic product before overoxidation can occur. Conversely, however, Cinchona is postulated to reduce 7 to form cinchonamine 10 (Fig. 1). Screening of candidate reductases, obtained through coexpression correlation with CiS (Supplementary Fig. 21b), led to the identification of an alcohol dehydrogenase that converts 7 to 10, which we named cinchonaminal reductase (CiR) (peak C, Extended Data Fig. 2e–g). CiS and CiR were also enriched in the same clusters (clusters 4–6; Fig. 3b) as the upstream biosynthetic genes (for example, MAT and MCC) according to the snRNA-seq data. This result suggests that cell-type localization is maintained in these downstream pathway steps.
Conversion of cinchonaminal 7 into a quinoline moiety probably involves oxidative opening of the indole, followed by cyclization and dehydration to form the quinoline scaffold. As the phylogenetically related plant M. speciosa uses a cytochrome P450 enzyme (family CYP71) to oxidize an indole to a spirooxindole35,37, we speculated that a Cinchona CYP71 P450 enzyme might also be involved in the oxidative transformation of this indole substrate. Because the feeding studies and metabolite accumulation levels indicated that these late-stage steps were more highly expressed in root and stem tissues (Supplementary Figs. 4 and 22–27), we performed a clustering analysis to prioritize gene candidates based on this expression pattern (Supplementary Fig. 28). Among the top CYP71 P450 candidates selected on the basis of the coexpression analysis with CiS, a functional enzyme was identified, which we named cinchonaminal oxidase (CiO) (Extended Data Fig. 3a,b). CiO catalysed the oxidation of 7, which produced the two ketone quinoline isomers cinchonidinone (8) and cinchoninone (9). These stereoisomers, which are known to exist in equilibrium38,39, were observed in a 1:0.23 ratio, which is approximately the same ratio observed in C. pubescens. The enzymatic activity of CiO was further confirmed through in vitro assays using yeast microsomes (Supplementary Fig. 29), which demonstrated that this single P450 enzyme can catalyse the indole-to-quinoline ring expansion. VIGS of CiO was also conducted in C. pubescens. Leaves in which CiO was silenced showed a significant accumulation of 13 and 10 along with decreased biosynthesis of both ketone quinoline alkaloid stereoisomers (8 and 9). This result provides evidence for the physiological function of CiO in planta (Extended Data Fig. 3c–e).
The final step in cinchona alkaloid biosynthesis involves the reduction of the ketone moiety that is generated after the formation of the quinoline scaffold. Previous studies have suggested that this transformation is probably catalysed by a NADPH-dependent oxidoreductase40. A total of 60 NADPH-dependent reductase genes that were coexpressed with CiO in the bulk transcriptomic dataset and/or snRNA-seq were cloned and tested (Supplementary Dataset 1). Screening in N. benthamiana did not lead to an active candidate. Therefore, we prioritized the selected candidate genes on the basis of the analysis of expression trends (Supplementary Fig. 28a–c) and re-expressed 15 putative aldo-keto reductases and alcohol dehydrogenases (including the above mentioned KR1–KR8) in E. coli for assays in vitro. Only one enzyme, KR4, displayed clear reductase activity, converting both non-methoxylated and methoxylated quinoline ketone stereoisomers into the corresponding alcohols (Extended Data Fig. 4a,b). This functional enzyme also reduced the aldehyde function of corynantheal scaffolds (Extended Data Fig. 1a), a result that further supported its involvement in the biosynthesis of cinchona alkaloids. Notably, despite robust activity in vitro, KR4 did not show detectable activity when assayed in N. benthamiana leaves after transient expression and infiltration of quinoline ketones. We designed a fusion protein comprising CiO and KR4 for transient expression in N. benthamiana. Although oxidation activity of CiO was observed, no reduced product was obtained. To test whether N. benthamiana contained endogenous components that inhibit this aldo-keto reductase, diluted crude extracts from N. benthamiana leaves were added to in vitro reactions, substituting 10% of the reaction volume. The addition of N. benthamiana extract abolished KR4 activity in vitro, a result that implicates the presence of an unknown inhibitory factor that suppresses KR4 function (Extended Data Fig. 4a,b and Supplementary Fig. 30a,b). These results provide a reasonable explanation for the lack of detectable activity in N. benthamiana and highlights a potential limitation in using this heterologous system for functional validation of reductases.
Although the two late-stage genes CiO and KR4 were selectively and highly expressed in stem and root tissues, these genes also showed clear enrichment in clusters 4–6 of the leaf snRNA-seq dataset, clusters previously associated with nearly all upstream alkaloid biosynthetic genes (Fig. 3b). Notably, the expression of monoterpene indole alkaloid pathway genes in the Apocynaceae species C. roseus seems to switch from epidermal cells to specialized idioblast cells in the late stages of alkaloid biosynthesis32. By contrast, in Cinchona, a member of the Rubiaceae family, all alkaloid biosynthetic genes identified in this study were selectively enriched in epidermal cells (clusters 4–6). This consistent cell-type-specific expression pattern of the biosynthetic genes in Cinchona, along with high coexpression in bulk tissue (Supplementary Fig. 31), provides further support that these newly identified genes function in the same metabolic pathway and contribute to the biosynthesis of cinchona alkaloids. However, we note that KR4 did not exhibit absolute stereoselectivity with the ketone quinoline alkaloids (8, 9, 8a and 9a). In particular, with the methoxylated analogues 8a–9a, a stereoisomer that is not observed in the plant was formed in higher levels than the naturally observed stereoisomers (Supplementary Fig. 30c,d). Although this reductase showed clear activity with the quinolines, it is possible that a reductase with higher stereoselectivity remains to be identified.