Recycling Upstream Redox Enzymes Expands the Regioselectivity of Cycloaddition in Pseudo-Aspidosperma Alkaloid Biosynthesis

Nature uses cycloaddition reactions to generate complex natural product scaffolds. Dehydrosecodine is a highly reactive biosynthetic intermediate that undergoes cycloaddition to generate several alkaloid scaffolds that are the precursors to pharmacologically important compounds such as vinblastine and ibogaine. Here we report how dehydrosecodine can be subjected to redox chemistry, which in turn allows cycloaddition reactions with alternative regioselectivity. By incubating dehydrosecodine with reductase and oxidase biosynthetic enzymes that act upstream in the pathway, we can access the rare pseudoaspidosperma alkaloids pseudo-tabersonine and pseudo-vincadifformine, both in vitro and by reconstitution in the plant Nicotiana benthamiana from an upstream intermediate. We propose a stepwise mechanism to explain the formation of the pseudo-tabersonine scaffold by structurally characterizing enzyme intermediates and by monitoring the incorporation of deuterium labels. This discovery highlights how plants use redox enzymes to enantioselectively generate new scaffolds from common precursors.

A lkaloid-producing plants in the Apocynaceae family have evolved cyclases that catalyze the cycloaddition of a highly reactive substrate, dehydrosecodine (1), into distinct alkaloid scaffolds. 1,2 We and others recently discovered and characterized the only known enzymes that catalyze cycloaddition of 1: tabersonine synthase (TiTabS and CrTS), which catalyzes the formation of (−)-tabersonine (2) (precursor to anticancer drugs vinblastine and vincristine), catharanthine synthase (CrCS), which catalyzes formation of (+)-catharanthine (3) (precursor to vinblastine and vincristine), and coronaridine synthase (TiCorS), which catalyzes formation of (−)-coronaridine (4) (precursor to antiaddiction agent ibogaine) ( Figure 1). 2−5 TiTabS/CrTS and CrCS directly yield 2 and 3 from 1 by a formal [4+2] cycloaddition, 2,6 while TiCorS initially forms a hitherto uncharacterized unstable intermediate, which is then enzymatically reduced to 4. The 1 substrate could, in principle, undergo alternative cycloaddition reactions to yield additional scaffolds, but extensive mutagenesis of these cyclases did not result in an expansion of the enzymatic product profile. 2 Here we show that the cyclase TiCorS can, in addition to generating 4, also produce the alternative pseudo-aspidosperma (Ψ-aspidosperma)-type alkaloid pseudo-tabersonine (Ψ-tabersonine) (5) (Figure 1). We show the mechanistic basis behind this transformation by first characterizing the unstable intermediate produced by the cyclase TiCorS. This intermediate can be intercepted by a reductase to generate 4 or by both a reductase and an oxidase to generate the alternative scaffold 5. Deuterium labeling provides evidence for the mechanism of these enzymatic transformations. In short, the chemical reactivity of 1 is exploited by both a cyclase and a pair of redox enzymes that can isomerize the alkene moieties, which in turn facilitates new cyclization regioselectivity.
Tabernanthe iboga, a plant that produces 2 and 4 via enzymes TiTabS and TiCorS, respectively, 3 also produces Ψaspidosperma-type alkaloid 20-epi-ibophyllidine ( Figure 1). 7−9 We hypothesized that 5 would be the precursor to 20-epiibophyllidine. 10 The natural product 5, rarely observed in nature, has only been isolated from Tabernaemontana calcarea, a species closely related to T. iboga. 11,12 To identify T. iboga enzymes that could form (+)-Ψ-tabersonine 5, we performed coupled in vitro biochemical assays in which the unstable 1 substrate is enzymatically generated from the upstream intermediate stemmadenine acetate (6). 6 is first oxidized by the flavin-dependent enzyme precondylocarpine acetate synthase (PAS) 5 to generate precondylocarpine acetate (7) and then undergoes a 1,4-iminium reduction by the medium chain alcohol dehydrogenase dihydroprecondylocarpine acetate synthase (DPAS) (Figure 1). 13 The resulting reduced unstable product, dihydroprecondylocarpine acetate (8), undergoes elimination of an acetoxy group to yield 1, which is then captured by one of the cyclases (Figure 1).
Our initial hypothesis was that a dedicated cyclase in T. iboga would catalyze isomerization and cyclization of 1 to 5. The T. iboga transcriptome, which contains the two previously identified cyclases TiTabS and TiCorS (81% sequence identity, Figure S1), does not harbor any additional cyclase homologues that might have alternative cyclization specificity. However, to our surprise, we observed that under certain conditions, 5, rather than 4, was formed in assays using TiCorS ( Figure 2a). Therefore, TiCorS appears to be involved in production of both 4 and 5.
In vitro assays using heterologously produced proteins ( Figure S2) were used to probe the conditions that led to a switch in product selectivity. We first noted that the use of specific homologues of reductase DPAS and oxidase PAS with TiCorS led to changes in the product profile. T. iboga has two homologues of the reductase DPAS (TiDPAS1, TiDPAS2, Figure S3) and three homologues of PAS (TiPAS1, TiPAS2, TiPAS3, Figure S4). 3 Additionally, a PAS homologue (CrPAS) 5 from the taxonomically related plant, Catharanthus roseus, which produces vinblastine, is also available and was tested. Most importantly, the pH of the reaction was critical, with 5 formation being observed at pH 7.5 and 4 production observed at pH 9.5 ( Figures S5 and S6). Overall, 5 formation was favored at pH 7.5, with TiPAS1−3 or CrPAS, TiDPAS1. and TiCorS, 2 production was favored at pH 7.5, with TiPAS1−3 or CrPAS, TiDPAS1, and TiTabS, and 4 formation was favored at pH 9.5 with TiPAS1−3, TiDPAS2, and TiCorS ( Figure 2a and Figures S5 and S6). Additionally, we also could produce the over-reduced version of 5, pseudo-vincadifformine (Ψ-vincadifformine) (9), by using TiPAS1−3 (instead of CrPAS), TiDPAS1, and TiCorS (Figure 2a and Figure S7).
To further substantiate the results obtained in vitro, we reconstituted the biosynthetic enzymes reported here, leading to the production of 2, 5, and 9 in Nicotiana benthamiana. Enzymes were transiently expressed in N. benthamiana leaves, disks were excised from the transformed leaf tissue, and these disks were placed in buffer containing 6 ( Figure 2b). We observed the production of 2, 5, and 9 in the extracts of the leaf disks using this expression system (Figure 2b). 4 was not detected in any of the enzyme combinations tested in N. benthamiana, presumably because of the higher pH conditions required for formation of this product. Additionally, the selectivity observed for the PAS homologues was not observed in planta, because N. benthamiana harbors an endogenous enzyme that is able to oxidize substrate 6 ( Figure 2b). 5 An acid-stable isomer of dehydrosecodine, angryline (1a), can also be isolated and directly used in cyclization assays  Figure 1). 2 1a must be used at a pH value above 8.5, where it will open to generate the reactive 1. When 1a was used in enzymatic assays in place of 6 (pH 9.5), we could observe formation of 5, 2, 4, and 9 ( Figure S8). Notably, both PAS (TiPAS1−3, CrPAS) and TiDPAS1 were required for the formation of 5 and 9, indicating that these enzymes are required for the formation of the Ψ-aspidosperma scaffold. PAS was not required for 4 production, which was favored when reductase TiDPAS2 was used ( Figures S9 and S10). To investigate the mechanism by which TiCorS can act, we first set out to characterize the initial, unstable product that is released from TiCorS in the absence of reductase (Figure 1). We optimized conditions under which this intermediate could be isolated and reductively trapped this compound with NaBH 4 . NMR analysis showed that the compound was 16carbomethoxycleavamine (10), suggesting that the initial cyclization product of TiCorS is 16-carbomethoxycleaviminium (11) (Figure 3 and Figure S11). When the TiCorS product was reduced with NaBD 4 , the deuterium label was incorporated at carbon 21, 10-d, which would be expected for 1,2-reduction of 11 ( Figure 3). Notably, the crystal structure of related cyclase CrCS (70.8% sequence identity, Figure S1) showed that CS initially forms (+)-16-carbomethoxycleaviminium (11a), 14 which then subsequently cyclizes to 3, though unlike TiCorS, the intermediate is not released from the active site. 2 Moreover 3 can open to form 11a under acidic conditions ( Figure S12). 1,2, 15 We used CD spectroscopy to show that TiCorS generates (−)-16-carbomethoxycleavamine (10), which is the opposite enantiomer that is generated by CrCS ( Figure S13).
With knowledge of the cyclization product of TiCorS, we proposed a mechanism for the formation of 4. After release from the active site of TiCorS, 11 undergoes a 1,4-reduction by TiDPAS2, which in turn would facilitate a second cyclization to form 4 ( Figure 4). We speculate that at higher pH values, TiDPAS2 favors 1,4-reduction, which when followed by tautomerization, 16 primes the substrate to cyclize to 4.
However, the mechanism by which 11 could be transformed into the 5 scaffold was still not clear. Therefore, we performed the reaction in the presence of isotopically labeled NADPH (pro-(R)-NADPD) to determine where the deuteride is incorporated in 5. 13,17,18 We incubated 6 with CrPAS, TiDPAS1, and TiCorS or TiTabS, along with pro-(R)-NADPD, which is required for DPAS reduction ( Figure  S14). With TiTabS, we saw formation of the 2 product with a mass consistent with incorporation of one deuterium, as expected from the action of DPAS acting on 7 (Figure 4). 13 However, when TiCorS was substituted for TiTabS, the resulting 5 product had a mass consistent with incorporation of two deuterium atoms (Figure 4 and Figure S14), clearly demonstrating that formation of 5 from 6 requires two reduction steps. Furthermore, the 9 product showed a mass incorporation of three deuterium atoms (Figure 4 and Figure  S14). To corroborate these results, we also incubated the trapped isomer of 1, 1a with PAS, TiDPAS1, the labeled NADPD cofactor, and TiTabS or TiCorS. The 2 product produced by TiTabS had no deuterium incorporation, as expected; formation of 5 and 9 from 1a showed incorporation of one and two deuterium atoms, respectively, and was strictly  Figure S15). We isolated isotopically labeled 2, 5, and 9 and showed that the deuterium labels were incorporated at carbon 19 for 2 as expected, 13 and carbons 19 and 21 for 5, carbons 19, 21, and 15 for 9 ( Figure 4). We performed CD spectroscopy of these isolated products and assigned the stereochemistry as (+)-Ψtabersonine (5) and (+)-Ψ-vincadifformine (9), which is the expected stereochemistry based on the downstream ibophyllidine products ( Figure S16). Finally, we incubated the chemically 1,2-reduced TiCorS product, 10, with the oxidase PAS and observed formation of 5 ( Figure 4 and Figure S17), indicating that cyclization occurs after PAS-catalyzed oxidation.
Together, this evidence strongly supports a mechanism for the formation of 5 (Figure 4). TiCorS cyclizes 1 to 11 and releases it from the active site, where it is reduced to 10 via 1,2reduction by TiDPAS1 and then reoxidized by PAS. The resulting intermediate is then primed to spontaneously cyclize to form 5. 9 forms by PAS oxidation of the doubly reduced (−)-16-dihydrocarbomethoxycleavamine (12) (Figure 4 and Figure S7). The switch between 4 and 5 is ultimately controlled by whether DPAS catalyzes a 1,4-reduction or 1,2reduction. The changes in the assay pH conditions or protein− protein interactions may be responsible for favoring 1,2-reduction over 1,4-reduction. Alternatively, an as yet undiscovered reductase that generates 4 at physiological pH may be responsible for the biosynthesis of this compound in T. iboga.

■ ASSOCIATED CONTENT
* sı Supporting Information