Dual Catalytic Activity of a Cytochrome P450 Controls Bifurcation at a Metabolic Branch Point of Alkaloid Biosynthesis in Rauwolfia serpentina

Abstract Plants create tremendous chemical diversity from a single biosynthetic intermediate. In plant‐derived ajmalan alkaloid pathways, the biosynthetic intermediate vomilenine can be transformed into the anti‐arrhythmic compound ajmaline, or alternatively, can isomerize to form perakine, an alkaloid with a structurally distinct scaffold. Here we report the discovery and characterization of vinorine hydroxylase, a cytochrome P450 enzyme that hydroxylates vinorine to form vomilenine, which was found to exist as a mixture of rapidly interconverting epimers. Surprisingly, this cytochrome P450 also catalyzes the non‐oxidative isomerization of the ajmaline precursor vomilenine to perakine. This unusual dual catalytic activity of vinorine hydroxylase thereby provides a control mechanism for the bifurcation of these alkaloid pathway branches. This discovery highlights the unusual catalytic functionality that has evolved in plant pathways.

The plant Rauwolfia serpentina (Apocynaceae), commonly known as Indian snakeroot, has traditionally been used for medicinal purposes in South and South-East Asia for its antihypertensive and calming effects.T his plant is also one of the fifty fundamental herbs in traditional Chinese medicine. [1] Rauwolfia spp.p roduce aw ide variety of compounds, including approximately 150 monoterpene indole alkaloids (MIAs) such as reserpine,yohimbine,a nd raubasine. [2] One of the best known MIAs of Rauwolfia is the ajmalan-type MIA ajmaline,aclass Ia antiarrhythmic agent often used in diagnosis of patients suspected of having Brugada syndrome. [3] Like all MIAs, ajmaline is produced via the versatile intermediate strictosidine,w hich is subject to multiple catalytic steps to yield the structurally diverse members of the MIA natural product family. [2] All biosynthetic transformations leading from strictosidine to ajmaline have been detected in enzyme fractions from Rauwolfia cell culture,t hough only six biosynthetic genes have been cloned and characterized ( Figure 1). [4][5][6][7][8][9][10][11] Several studies showed that vomilenine is an ajmaline biosynthetic intermediate that is produced through selective hydroxylation of vinorine at the C-21 position by Rauwolfia cell culture extracts (Figure 1). [12][13][14] However,n oe nzyme isolated from R. serpentina has shown this activity. [11] Comparative transcriptomics and metabolomics have been widely and successfully used to elucidate the genes of plant alkaloid metabolism. [15][16][17][18] Here we report the use of an available Rauwolfia transcriptome (Medicinal Plant Genomics Resource (MPGR), http://medicinalplantgenomics.msu. edu/index.shtml) to identify and characterize ac ytochrome P450 enzyme (CYP82S18) that hydroxylates vinorine to generate vomilenine,w hich we show exists as rapidly interconverting epimers in solution. Surprisingly,t his cytochrome P450 also catalyzes the redox neutral isomerization of vomilenine to form the alkaloid perakine.T herefore,t his second, non-oxidative function of the cytochrome P450 provides an unexpected route to the diversification of the ajmalan alkaloid scaffold.
Early studies using R. serpentina protein extracts suggested the involvement of ac ytochrome P450 (CYP) in the conversion of vinorine into vomilenine. [19] Thee xpression profiles of genes encoding CYPs from an available Rauwolfia transcriptome database were used to search for candidates. [20] Approximately 270 transcripts are annotated as cytochrome P450s in the R. serpentina transcriptome.O ft hese,1 10 transcripts are expressed in young and mature root tissues, where ajmaline is highly accumulated. Tr anscripts from the dataset were clustered using as elf-organizing map (SOM; Figure S1 in the Supporting Information) for efficient candidate identification. [21] Thes earch for vinorine hydroxylase focused on 1) candidates with gene expression profiles correlated with the accumulation of ajmaline and 2) candidates within or adjacent to nodes containing previously identified MIA enzymes.A vailable metabolomic data suggested that while ajmaline is found in the highest levels in young roots,a erial organs display trace to undetectable ajmaline levels ( Figure 2A). Correspondingly,t ranscripts encoding known enzymes in the ajmaline pathway were most highly expressed in roots ( Figure 2B). Eight CYP candidates from different families ( Figure 2C)that displayed similar expression profiles to known ajmaline genes were identified from analysis of the SOM.
Thec oding regions of the CYP candidates were inserted into the multiple cloning site of ayeast dual-expression vector with the required cytochrome P450 reductase CPR. To determine the enzyme activity of CYP candidates,1 0mm of the substrate,vinorine,was fed to 1mLyeast cultures for 48 h. Only yeast cultures that harbored the construct encoding candidate 5437 (pESC-leu2d:CPR/5437) showed the consumption of vinorine and the formation of two new products ( Figure 3A). No enzymatic product was observed when vinorine was incubated with yeast cultures harboring empty vector, or any of the other CYP candidates.Similarly,invitro assays with microsomal fractions of yeast harboring pESC-leu2d:CPR/5437 also showed that in the presence of NADPH, vinorine (m/z 335) was consumed by the enzyme,resulting in the formation of two reaction products with m/z 351 as evidenced by LC-MS analysis ( Figure 3A). In vitro assays with all other candidates did not show any consumption of vinorine.T he substrate specificity of 5437 was determined using 10 MIAs representing several different structural subgroups ( Figure S2). Vomilenine,the hydroxylated product of vinorine,w as the only other alkaloid accepted by 5437, albeit with much lower efficiency (13 %o ft he substrate turnover rate compared with vinorine;F igure 3B). This high substrate specificity is consistent with previous work with microsomal extracts of Rauwolfia. [14] Thec losest homologue of 5437 in the related Apocynaceae plant Catharanthus roseus (cra_locus_12789) did not accept vinorine as asubstrate.
Them ajor enzymatic product of 5437 had am ass of m/z 351, an increase of 16 amu compared to the vinorine starting material, which strongly suggests that 5437 catalyzes ah ydroxylation reaction. Am inor enzymatic product with m/z 351, but ad ifferent retention time,w as also observed. To identify the reaction products,alarge-scale reaction was carried out, and the major product was characterized by NMR analysis (COSY,N OESY,H MBC,H SQC). Both 1 Ha nd 13 C chemical shifts were in good agreement with previously reported data for vomilenine ( Figure 1, Figure S6, Table S1). [13,22] 5437 was therefore designated vinorine hydroxylase (VH). This enzyme displays an optimal pH of 6.5 for the hydroxylation reaction ( Figure S5), which is lower than previously reported for the reaction catalyzed by R. serpentina microsomes. [19] Theo ptimal pH for the isomerization reaction, which was not reported to be associated with VH activity in R. serpentina microsomes, [19] was 4.5 ( Figure S5). Under optimal conditions (HEPES-buffered reaction at pH 6.5 and 30 Michaelis-Menten-type reaction kinetics with an estimated V max value of 98 pmol min À1 mg À1 total protein and a K m for vinorine of 6.8 mm (compared to a K m of 23 mm R. serpentina microsomes [19] ;F igure S5). No significant substrate or product inhibition was detected.
Thes tereochemistry of the newly installed hydroxy group at C-21 of vomilenine was assumed to have the (R)c onfiguration,   Figure S8). In both solvents,o nly as ingle signal set was observed, thus indicating ar apid equilibration on an NMR timescale in these solvents.I nc ontrast, two signal sets in ar atio of approximately 4:1w ere observed in [D 6 ]DMSO. Based on NOE data, these signal sets could be clearly assigned to the C-21 epimers,w ith 21-epi-vomilenine being the main component ( Figure S8). This assignment is also supported by the strong deshielding of H-5 of 21-epi-vomilenine (d H = 3.70 ppm), caused by the proximity of the C-21 hydroxy group,i nc omparison to vomilenine (d H = 3.26 ppm;T able S1). Aprevious report showed that this slow equilibration can be accelerated by addition of water, [23] thus suggesting that the epimerization could occur under physiologically relevant aqueous conditions.G iven that plant CYPs in specialized metabolism typically catalyze highly specific reactions,i ti sm ost likely that VH catalyzes ar egio-and stereoselective hydroxylation reaction that is followed by isomerization, although oxidation by VH through hydride abstraction to form the iminium moiety could also be possible.E xperimentally demonstrating that the enzyme is stereoselective was not possible,h owever, since the vomilenine isomers could not be separated using the established method [24] and monitored over time in the LC-based assay (Figure S9). Notably,a jmaline does not equilibrate in asimilar fashion to vomilenine,t hus suggesting that further modifications of the scaffold, presumably reduction of the C19 = C20 double bond, increase the energy barrier between the C-21 epimers. Them inor enzymatic product of VH was identified as perakine,a se videnced by co-elution with an authentic standard of perakine ( Figure 3A), along with 1 Ha nd 13 CNMR analysis (Table S2, Figure S6), although limited availability of vomilenine precluded the determination of kinetic data for this reaction. Perakine has long been known to be apossible product of vomilenine,since the introduction of ahydroxy group at C-21 allows opening of the ring via the newly formed hemiaminal. Ther esulting amine can then undergo aM ichael addition to form perakine (Scheme 1). However,w hile this isomerization from vomilenine to perakine can occur non-enzymatically,h arsh chemical conditions and extended reaction times are required. [25] It has also

Angewandte Chemie
Communications 9442 www.angewandte.org been shown that formation of the perakine-derived product raucaffrinoline is observed only when vomilenine is incubated with crude Rauwolfia enzyme mixtures, [26] thus suggesting that the conversion of vomilenine into perakine is enzymatically catalyzed. However,t he enzyme responsible for this conversion has never been identified.
As eries of control experiments were performed to determine whether VH in fact catalyzes this redox-neutral isomerization. Isolated vomilenine was not converted into perakine in the assay buffer without the presence of VH (buffer only) or with microsomes lacking VH ( Figure 3B), even over aw ide range of pH values from 2.5 to 12.5 ( Figure S4B). Formation of perakine was measured in aseries of buffers at fixed ionic strength under saturating vinorine concentrations,t hus demonstrating that VH is active over alarge range of pH values (6.5 to 8.5;F igure S4). NADPH is not required for the reaction, which consistent with the fact that this is anon-oxidative reaction ( Figure S10). Theoptimal pH for the isomerization of vomilenine to perakine is approximately 4.5, which is unusual for aC YP ( Figure S4). VH does not accept perakine as as ubstrate,t hus suggesting that the formation of perakine from vomilenine is directional. Therelative yield of perakine is higher when vinorine is used as as ubstrate ( Figure 3A), compared to when vomilenine is used as as ubstrate ( Figure 3B). This could suggest that keeping vomilenine in the active site of the CYP is important for more efficient rearrangement of the hydroxylated intermediate.C ollectively,o ur data suggest that in addition to hydroxylating vinorine to form vomilenine,VHalso catalyzes the non-oxidative isomerization of vomilenine to perakine. While it is not clear how VH catalyzes this reaction, we hypothesize that acidic or basic residues within the substrate recognition site of the CYP facilitate either the ring opening or the subsequent Michael addition to form perakine.T he reaction could also proceed via an azetidine intermediate; however, modeling studies suggest that this molecule may be too sterically congested, therefore we favor the stepwise mechanism shown in Scheme 1. Since assays with perakine indicated that the isomerization reaction is irreversible,a nd NMR analysis of the perakine product indicated that no enol is present, we predict that tautomerization of the enol to the aldehyde is likely to be irreversible (solid arrow,S cheme 1). Thee quilibria of the other steps shown in Scheme 1a re not known. This enzymatic isomerization showed as trong pH dependence,with apHoptimum at pH 4.5 and low efficiency above pH 7( Figure S4). Thep K A value of ajmaline is 8.2, [27] and the structurally similar vomilenine likely has as imilar pK A value.W ehypothesize that enzymatic protonation of the nitrogen may play an important role in facilitating opening of the aminal moiety as the first reaction step (Scheme 1), although the lack of structural data for plant CYP enzymes makes it challenging to design mutations to test this.ACYP that catalyzes an on-oxidative reaction is highly unusual:t o the best of our knowledge,o nly as ingle plant CYP that catalyzes ad ehydration reaction [28] along with af ungal CYP that catalyzes an unexpected terpene cyclase reaction have been reported. [29] It is likely that the inherent reactivity of the vomilenine substrate,along with the availability of an enzyme with ab inding pocket that could (even weakly) bind vomilenine,e volved to perform the relatively simple acidbase catalysis to form the perakine product, thereby generating an additional branch of alkaloids.
Thed ual catalytic function of VH suggests that this enzyme may have ad ouble function in MIA biosynthesis in Rauwolfia. While the rates indicated by the in vitro assays suggest that it most efficiently delivers vomilenine as the intermediate in the ajmaline pathway,t he enzyme also gives rise to perakine,w hich is al ess abundant alkaloid in Rauwolfia spp.T he expression profile of VH is consistent with ar ole in the production of both ajmaline and perakine. VH, like other ajmaline biosynthetic enzymes,i sh ighly expressed in root tissues ( Figure S7A), but is also found in leaf,where perakine and perakine-derived alkaloids accumulate ( Figure S7B). [30] Perakine reductase,b eing expressed abundantly in leaf,c aptures the perakine product and drives the pathway toward formation of perakine-derived products in leaves.V omilenine reductase,w hich is highly expressed in young root, provides the driving force for the formation of dihydro-vomilenine and ajmaline in root. Therefore,c ontrol of ajmaline and perakine levels may depend on both the dual catalytic function of VH, and on the tissue expression profile of VH and downstream pathway enzymes.T he catalytic activity of VH, along with the localization of ajmaline and perakine enzymes,g enerates ac omplex metabolic network that allows the production of chemical diversity in aspatially localized manner.
Herein, we report the discovery of vinorine hydroxylase (VH), acytochrome P450 that hydroxylates vinorine to form vomilenine,amissing step in the biosynthesis of the antiarrhythmic agent ajmaline.S urprisingly,V Ha lso catalyzes the formation of perakine through anon-oxidative isomerization reaction of vomilenine.T his cytochrome P450 thereby extends control of the bifurcation into two pathway branches via an unusual dual catalytic function. This discovery highlights the catalytic versatility of plant CYP enzymes.