Transcriptomes of spirooxindole alkaloids-producing plants (M. speciosa, Rauwolfia serpentina, and G. sempervirens), MIAs-producing but spirooxindole alkaloids-free plants (Camptotheca acuminata, Amsonia hubrichtii, Cinchona ledgeriana, Nothapodytes nimmoniana, Ophiorrhiza pumila, and Catharanthus roseus) and a MIAs-free plant (Arabidopsis thaliana) are publicly available from the Medicinal Plant Genomic Resources (http://mpgr.uga.edu/), the PhytoMetaSyn database (https://bioinformatics.tugraz.at/phytometasyn/), TAIR10 (https://www.arabidopsis.org/), and previous studies (Rather et al., 2018; Rai et al., 2021). TransDecoder was used to generate the proteomes, which were subsequently subjected to OrthoFinder for orthogroups analysis. Candidates belonging to the CYP71 family that act on the B ring of the indole moiety were selected to test for activities in the spirooxindole scaffold formation. We focused on the orthogroups containing CYP orthologues unique to M. speciosa (MsCYP). The phylogenetic tree of the MsCYP candidates with other reported CYPs involved in the biosynthesis of MIAs from different species was constructed by the Geneious Tree Builder program in the Geneious Prime software package (Biomatters). The amino acid sequence alignment of the MsCYP candidates was performed by the Geneious Prime software package (Biomatters). The open reading frames of MsCYP candidates were obtained from the available transcriptome (Brose et al., 2021). The sequences combining overhangs of SpeI and NotI restriction sites at the multiple cloning site 1 of the pESC-Leu2d plasmid were synthesized by TwistBioscience (CA, USA) (Ro et al., 2008; Nguyen et al., 2021). The constructs were inserted into pESC already containing the required redox partner cytochrome P450 reductase (CPR) (Ro et al., 2008) by using 5X In-Fusion cloning system (Takara Bio USA Inc.). The yeast strain YPL 150 C:PEP4KO was used for heterologous expression of the CYP candidates following the procedure described before (Nguyen et al., 2022).

The in vivo CYP activity screening assays were conducted using the established protocol for yeast whole-cell assays in the 96-well plate (Nguyen et al., 2022). Various MIA alkaloid substrates were fed at a final concentration of 10µM into the yeast cultures for 48 h ( Supplementary Figure 13 ). The in vitro assays were conducted in different buffers: 1 M citrate pH 4, 1 M citrate pH 5, 1 M HEPES pH 6, 1 M HEPES pH 7, 1 M HEPES pH 8, 1 M Tris pH 9, and 1 M Tris pH 10. The in vitro reaction condition was performed with 100 µL of 100 µM buffer, 250 µM NADPH, 10 mg total microsomal protein, and 10 µM hirsuteine at 37 °C for 1 h. The yeast cells containing plasmid without CYP construct were used as empty vector controls. The reaction supernatants were collected by centrifugation and filtration with a 0.2 µm syringe filter (Sartorius). The supernatants were injected to ultra-performance liquid chromatography (UPLC) coupled with a Xevo TQ-S Cronos Triple Quadrupole Mass Spectrometer (MS). All UPLC-MS analyses were conducted on an XBridge BEH XP (50 x 2.1 mm, 1.7 μm) column at a flow rate of 0.6 mL.min^–1. The column was pre-equilibrated in 90% solvent A (water + 0.1% formic acid), and 10% solvent B (acetonitrile + 0.01% formic acid). The eluting conditions were: 0–8 min, 10–50% B; 8.0–8.5 min, 50–100% B; 8.5–9.5 min, 100% B; and 9.5–11 min, 100–10% B to re-equilibrate the column. Immunoblotting experiment of recombinant MsCYP72056 enzyme was conducted as described before (Nguyen et al., 2021). Steady-state enzyme kinetics was conducted by varying the concentration of hirsuteine substrate from 0 to 300 µM in HEPES pH 7.5, at a fixed concentration of NADPH at 250 µM and analyzed using GraphPad Prism 9.4.1 (GraphPad software).

To obtain enzymatic products at sufficient yields for structural elucidation, we performed multiple in vitro assays of MsCYP72056 containing 10 mL 100 µM HEPES pH 7.5, 250 µM NADPH, 10 mg microsomal protein, and 50 µM hirsuteine at 37 °C for 1 h. Reactions were stopped by adding 1 mL of methanol. The in vitro assay supernatants were combined after centrifugation. The crude enzymatic products mixture was extracted from the supernatant by liquid-liquid extraction with chloroform, which was removed in vacuo by GeneVac. Concentrated samples were subjected to a Varian semi-preparative HPLC equipped with a Kinetex^® 5 μm EVO C18 column (100 Å, 100 × 250 mm) at a flow rate of 1.5 mL.min^–1. The column was equilibrated in 90% solvent A (water, 0.1% formic acid) and 10% solvent B (acetonitrile, 0.1% formic acid). The eluting conditions were conducted: 0–5 min, 10–20% B; 5–25 min, 20–70% B; 25–27 min, 70–90% B; 27–30 min, 90% B; 30–31 min, 90–10% B; and 31–34 min, 10% B to re-equilibrate the column. Approximately 0.2 mg of each product was dissolved in 600 μL CDCl₃ and subjected to 1D NMR (¹H, ¹³C) and 2D NMR (HSQC, HMBC, COSY NOESY) analyses on a Bruker Avance 600 MHz NMR spectrometer. CD analyses were performed with 0.2 mg/mL samples in CH₃OH on the Jasco J-815 CD spectrophotometer from 200–400 nm.

Using OrthoFinder, we generated the orthogroups from the publicly available transcriptomes of ten species, including non-MIA-producing plants, MIA-producing and spirooxindoles-free plants, and spirooxindole-producing plants (see Materials and method ). We focus on orthogroups containing CYP71 orthologues specific for spiroxindole alkaloid-producing plants such as M. speciosa ( Supplementary Figure 2A ) (Bindra, 1973; Ahmad and Salim, 2015; Manwill et al., 2022) as we speculated that member(s) of the CYP71 subfamily could oxidize the seco-/hetero-yohimbine alkaloids to oxygenated intermediates, which would be rearranged to spirooxindole pairs ( Figure 1 ). Our analysis identified six candidates, namely MsCYP53813, MsCYP72054, MsCYP72056, MsCYP9580, MsCYP9583, and MsCYP9585 from the orthogroup OG0016157.

From the spirooxindole structures reported in M. speciosa, we traced back to the plausible corynanthe- and ajmalicine-type precursors. Among these, ajmalicine, tetrahydroalstonine, mitragynine, 9-hydroxycorynantheidine, yohimbine, corynanthine, and hirsuteine were available and used for the functional validation of CYP candidates. To test the enzyme activities, 10 μM of the putative substrates were fed to 100-μL YPL154C:PEP4KO yeast cultures for 48 hr. Only yeast cultures harbouring the construct pESC-Leu2d::CPR/MsCYP72056 showed the consumption of hirsuteine ([M+H]⁺ m/z 367.5) and the formation of two new products ([M+H]⁺ m/z 383.5), 1 at 2.4 min and 2 at 2.6 min, as analyzed by LC-MS/MS ( Supplementary Figure 5A ). A 16-amu difference between the products and the substrate indicated that MsCYP72056 catalyzed an oxygenation/oxidation reaction. No enzymatic product was observed when hirsuteine was incubated with yeast transformed with an empty vector or constructs containing other CYP candidates. In vitro assays with microsomal protein of yeast expressing pESC-leu2d::CPR/MsCYP72056 also showed that in the presence of NADPH, hirsuteine was consumed, resulting in the formation of products 1 and 2 ( Figure 2A ). Michaelis–Menten kinetics characterization of MsCYP72056 with hirsuteine revealed a K _M value of 68.33 µM. We also investigated the in vitro activity of MsCYP72056 in the pH range of 4–10 and found that the enzymatic reaction was more favourable at pH 6–10 ( Figure 2B ). There was a 20-fold increase of isocorynoxeine levels from the same amount of substrate as the pH increased from 4 to 10, of which the most dramatic increase (5% to 50%) was observed as the pH increased from 5 to 6 ( Figure 2D ).

In addition to hirsuteine, 16 other structurally various indole alkaloids were used to assess the substrate scope of MsCYP72056 ( Supplementary Figure 13 ). Both in vivo and in vitro assays using cultures of yeast containing pESC-leu2d::CPR/MsCYP72056 and its microsomal fractions, respectively, showed that MsCYP72056 accepted hirsutine beside hirsuteine but not the other alkaloids ( Supplementary Figures 5 , 13 ). Similar to hirsuteine, hirsutine is also a (3R) secoyohimbine alkaloid with an ethyl group at C-20 in place of a vinyl group in hirsuteine ( Supplementary Figure 5 ). LC-MS analysis showed the consumption of hirsutine by MsCYP72056, yielding two products 3 and 4 with 16 amu (m/z 385.5) greater than the substrate (m/z 369.5). Multiple reaction monitoring (MRM) and daughter scan analyses were performed to detect the oxindole scaffold of the enzymatic products. The specific daughter ion of oxindole ([M+H]⁺ m/z 160.0) was observed, confirming that hirsutine was also converted into two spirooxindole products (Avula et al., 2015). Based on MsCYP72056 activity with hirsuteine, we speculated that the products from hirsutine were (3-epi-) rhynchophylline-type spirooxindoles, some of which have recently been isolated from M. speciosa (Flores-Bocanegra et al., 2020).

To elucidate the stereochemistry of the enzymatic products 1 and 2, large-scale in vitro reactions were conducted. Approximately 0.2 mg of the two products were purified and subjected to 1D NMR (¹H, ¹³C), 2D NMR (HSQC, HMBC, COSY, NOESY), and circular dichroism (CD) analyses ( Supplementary Figures 7 - 9 ). The 1D NMR data confirmed the spirooxindole skeleton of two enzymatic products compared to previously reported compounds corynoxeine and isocorynoxeine ( Supplementary Tables 1 , 2 ; Supplementary Figures 7 , 10 ) (Kitajiina et al., 2001; Flores-Bocanegra et al., 2020). In particular, the spectra of compound 2 resemble those of corynoxeine, and 1 had identical spectra with isocorynoxeine. In 2D NMR analysis, key NOE signals of isocorynoxeine, such as H-3/H-6, H-5/H-21, and H-20/H-21 were observed in NOESY spectrum of 1 ( Supplementary Table 2 ; Supplementary Figures 11B ). These signals initially indicated the 3S and 20R configurations of C-3 and C-20 in 1 (Seki et al., 1993; Qi et al., 2015). Based on the previous studies in conformational analysis of spirooxindole alkaloids, 1 possessed the normal-type conformation as that of mitraphylline and formosanine (Shamma et al., 1967; Seki et al., 1993). As the deshielding effect at H-9 has been reported to differentiate the 7S and 7R isomers in the normal-type spirooxindole alkaloids (Seki et al., 1993), the downfield shift of H-9 in 1 (δ 7.45) depicted the 7S configuration ( Supplementary Table 2 ). Although 2 had similar 1D NMR data with corynoxeine, we could not observe the key NOE signals of the normal-type conformation (3S, 20R) in the spectra of 2 ( Supplementary Table 1 ; Supplementary Figure 8B ). However, we could detect the NOE signals of the pseudo-type alkaloids (3R, 20R) as recently described in two compounds 3-epi-rhynchophylline and 3-epi-corynoxine B (Flores-Bocanegra et al., 2020). The NOE correlations were among H-3/H-9, H-9/H-6, H-5/H-3, H-18/H-20 ( Supplementary Figure 8B ). Therefore, we proposed that 2 was an epimer of corynoxeine, which was 3-epi-corynoxeine (3R, 7R, 20R).

To further confirm our proposed structures, we examined the C-3 and C-7 stereocenters of the enzymatic products of MsCYP72056 and the authentic standards of corynoxeine and isocorynoxeine using circular dichroism (CD) spectroscopic analysis ( Figure 2C ). Intriguingly, the positive Cotton effect at 250 nm confirmed the 3R configuration of product 2. On the contrary, a negative Cotton effect at 250 nm suggested the 3S configuration of product 1. By comparing the CD and NMR spectra of product 1 with isocorynoxeine standard data ( Figure 2C , Supplementary Table 2 ), we concluded that 1 was isocorynoxeine (3S, 7S) ( Figure 2A ). Meanwhile, 2 was a 3R spirooxindole since it has the opposite Cotton effect at 250 nm, as compared to the corynoxeine standard ( Figure 2C ). Therefore, product 2 (3R, 7R) was 3-epi-corynoxeine ( Figure 2A ). Based on the product profile of the enzymatic reaction, we named this enzyme 3-epi-corynoxeine/isocorynoxeine synthase (3eCIS).