Structural insights into a dual-substrate O-methyltransferase reveal the enzymatic mechanism of parallel paeonol biosynthesis

Abstract

Paeonol, the medicinal indicator component of the traditional Chinese medicine ‘Mudanpi’, has an incompletely elucidated biosynthetic pathway, particularly its post-modification steps. Specifically, the post-modification processes involve 4′-O-methylation and 2′-hydroxylation, thus the formation can occur either sequentially or in parallel. Here, using integrated transcriptomic and metabolomic analyses, we identify a key dual-substrate 4′-O-methyltransferase, PoOMT1, in Paeonia ostii. In vitro assays demonstrate that PoOMT1 exhibits high affinity, catalytic efficiency, and strict regioselectivity toward two paeonol precursors: 4′-hydroxyacetophenone and 2′,4′-dihydroxyacetophenone. Heterologous expression in tobacco and silencing in P. ostii of PoOMT1 further confirm its role in paeonol biosynthesis, supporting a parallel pathway. Structural analysis of the PoOMT1 ternary complex, combined with mutagenesis studies, uncover a His-Asn-Gln catalytic triad essential for the 4′-O-methylation, and reveal a T-shaped substrate binding pocket. This pocket operates through dual-anchor and steric constraint mechanisms, which provide the structural basis for strict regioselectivity toward the 4′-OH group and facilitate the parallel biosynthetic pathway. Moreover, the divergent catalytic profiles of ten PoOMT1 homologs that recognize the same substrates further corroborate the proposed catalytic mechanism. Our work elucidates the molecular basis of 4′-O-methylation and parallel biosynthetic pathway of paeonol, and lays a foundation for the complete elucidation of its pathway.

Paeonol is a bioactive compound with an incompletely elucidated biosynthetic pathway. Here, the authors identify the dual-substrate methyltransferase PoOMT1, establish a parallel biosynthetic pathway, and uncover the structural basis for its strict regioselectivity.

Introduction

Tree peony, belonging to the genus Paeonia (family Paeoniaceae), is a popular ornamental native to China, valued for its floral colors and patterns. Historically, it is originally utilized for the medicinal properties¹. The root phloem of tree peony is processed into Mudanpi (Fig. 1a), a traditional Chinese medicine whose pharmacologically active indicator component is paeonol². Paeonol exhibits a broad spectrum of bioactivities, including anti-inflammatory, anti-tumor, antibacterial, hypoglycemic, and neuroprotective effects^3,4. Currently, paeonol is primarily extracted from tree peony roots. However, conventional extraction methods suffer from low yield and purity⁵. Chemical synthesis routes, meanwhile, face challenges such as high substrate toxicity and undesirable by-product formation⁶. In recent years, synthetic biology has enabled the efficient and sustainable production of numerous medicinal natural products^7–12, offering a promising alternative for paeonol supply. Nonetheless, the complete biosynthetic pathway of paeonol remains unelucidated.

Fig. 1. The biosynthetic pathway of paeonol in Paeonia ostii.

a Photographs of P. ostii plant, root system, and traditional Chinese medicine Mudanpi. b Compounds containing an acetophenone skeleton identified in P. ostii root phloem. c Proposed biosynthetic pathway of paeonol starting from phenylalanine. PAL phenylalanine ammonia lyase, C4H cinnamate 4-hydroxylase, 4CL 4-coumarate-CoA ligase, CHD cinnamoyl CoA hydratase-dehydrogenase, TE peroxisomal thioesterase.

As a phenolic compound, paeonol is derived from the phenylpropanoid pathway. Isotope tracing experiments using ¹⁴C-labeled phenylalanine and cinnamic acid in P. suffruticosa confirmed their efficient incorporation into paeonol, establishing these as the intermediates in its biosynthesis¹³. Cinnamic acid is sequentially converted to 4-coumaroyl-CoA by cinnamate 4-hydroxylase and 4-coumarate-CoA ligase¹⁴. This intermediate is then processed by cinnamoyl-CoA hydratase-dehydrogenase, peroxisomal thioesterase, and spontaneous decarboxylation to yield 4-hydroxyacetophenone¹⁵, a structural analog of paeonol that accumulates in tree peony¹⁶.

Comparative structural analysis indicates that paeonol possesses additional C2′-hydroxyl and C4′-methoxy groups. Therefore, the remaining unidentified steps in paeonol biosynthesis are 2′-hydroxylation and 4′-O-methylation.

O-methylation is catalyzed by O-methyltransferases (OMTs), which transfer a methyl group from S-adenosyl-L-methionine (SAM) to a hydroxyl group of their substrate. It has been reported that the O-methylation can improve the pharmacological properties of natural products, including stability, lipophilicity, oral bioavailability, and biological activity^17–19. Plant OMTs that act on phenolic compounds are categorized into two major types according to their molecular weight and enzymatic characteristics¹⁹. Although these enzymes often display substrate promiscuity, individual OMTs typically exhibit specificity toward both a preferred substrate and a defined methylation site^20–23. This specificity is governed by the three-dimensional spatial conformation formed among the OMT, SAM, and the substrate molecule^24–27. To date, no plant OMT has been reported to utilize two distinct preferred substrates simultaneously in the biosynthesis of a single phenolic compound. The specific OMT responsible for paeonol biosynthesis has not been characterized, and its enzymatic properties, protein structure, and catalytic mechanism remain unknown.

In this work, we report a dual-substrate 4′-O-methyltransferase PoOMT1 from P. ostii that participates in a parallel pathway for paeonol biosynthesis. Through crystal structure analysis and mutagenesis studies, we characterize a T-shaped substrate binding pocket that confers strict regioselectivity toward the 4′-OH group and underlies the parallel biosynthetic pathway. A unique His-Asn-Gln catalytic triad essential for 4′-O-methylation is identified. Additionally, we characterize ten PoOMT1 homologs from other plant species that recognize the same substrates but exhibit distinct catalytic efficiencies and key amino acid residues, thereby corroborating the proposed catalytic mechanism.

Results and discussion

Metabolomics-based prediction of paeonol biosynthetic pathway

According to our previous study¹, paeonol is mainly distributed in the root phloem of P. ostii. Therefore, we conducted a widely targeted metabolomic analysis of this tissue to identify structurally related metabolites and infer the biosynthetic pathway of paeonol. In total, 539 metabolites were detected using UPLC-MS/MS platform (Supplementary Data 1). Phenolic acids were the largest group (216), followed by flavonoids (141), terpenoids (54), alkaloids (53), tannins (31), others (24), lignans and coumarins (19), and steroid (1). Among these, 11 compounds, including paeonol, contain an acetophenone skeleton (Fig. 1b). Structurally, paeonol is characterized by a hydroxyl group at the C2′ position and a methoxy group at the C4′ position of the acetophenone core. Based on the positions of these functional groups, we identified t (1), 2’,4’-dihydroxyacetophenone (2) and t (6) as potential biosynthetic precursors to paeonol. This suggests two plausible post-modification pathways (Fig. 1c). Pathway I, in which compound 1 is initially methylated at the C4′ position to form 6, followed by 2′-hydroxylation to produce 7; Pathway II, in which 1 is first hydroxylated at the C2′ position to form 2, followed by 4′-O-methylation to yield 7. Thus, based on metabolite profiling and enzymatic logic, the post-modification steps of paeonol biosynthesis may proceed through one of these two pathways or operate in parallel. The realization of parallel biosynthesis critically depends on the substrate specificity and flexibility of the key modifying enzymes, including both O-methyltransferases and hydroxylases. In fact, parallel biosynthesis of natural products mediated by hydroxylases and O-methyltransferases has been previously documented²⁸.

Mining and enzymatic characterization of 4’-OMT in paeonol biosynthesis

Given the structural features of paeonol (with hydroxyl groups at both the C2′ and C4′ positions), it is notable that methylation occurs exclusively at the 4′-hydroxyl group, whereas no methylation is observed at the 2′-hydroxyl group. This observation suggests that, at least for paeonol, the O-methylation modification exhibit greater specificity than hydroxylation. Therefore, identifying and characterizing the responsible 4′-OMT is critical for elucidating the post-modification steps of paeonol biosynthesis. Our previous study revealed that paeonol accumulation in P. ostii is tissue-specific¹, and this distinct accumulation pattern makes the phloem an ideal system for investigating paeonol biosynthesis. In this study, we collected RNA from petals, leaves, stems, and phloem of P. ostii and performed sequencing on a DNBSEQ platform. For the twelve libraries, the average of total clean reads was 69.90 million, and the average Q30 percentage was 89.30% (Supplementary Table 1). The reliability of the RNA-seq data was validated by qRT-PCR on a random set of differentially expressed genes (DEGs), showing a strong positive correlation (R² = 0.9576) (Supplementary Table 2, Supplementary Fig. 1). All sequencing data have been deposited in the Genome Sequence Archive in National Genomics Data Center under accession number CRA019310.

To identify OMT genes potentially involved in paeonol biosynthesis, we first performed an HMMER search (Biosequence analysis using profile hidden Markov Models, https://www.ebi.ac.uk/Tools/hmmer) against the P. ostii genome using domain models PF00891 and PF01596, identifying 102 OMT genes. The phylogenetic tree showed that 78 genes belonged to Type I OMT subfamily and 24 genes belonged to Type II subfamily (Supplementary Fig. 2). Further, the expression patterns of Type I OMT genes across the four tissues were analysed based on the RNA-Seq data (Supplementary Data 2), and they exhibited different FPKM values. According to the differences of paeonol contents in different tissues¹, we selected 7 genes with preferentially higher expression in phloem than in other tissues as further candidates, and named them PoOMT1 to 7 (Fig. 2a).

Fig. 2. Identification and enzymatic characterization of PoOMT1.

a Expression patterns of PoOMT1 ~ 7 in the transcriptomes of different tissues of P. ostii. Three replicates were tested for each tissue (n = 3). b Conversion of PoOMT1 ~ 7 using 1 and 2 as the substrates. Reactions were performed with 20 μg purified PoOMTs in Tris-HCl buffer (pH 8.0) under 37 °C for 2 h. c Catalytic function of PoOMT1 using 1 as substrate. d (+)-ESI-MS and MS² spectra of 6. e Catalytic function of PoOMT1 using 2 as substrate. f (+)-ESI-MS and MS² spectra of 7. g Michaelis-Menten plot of PoOMT1 was fitted using 1 as substrate. h Michaelis-Menten plot of PoOMT1 was fitted using 2 as substrate. Data are presented as mean values ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

To determine the catalytic functions, these candidate genes were amplified from the phloem cDNA library, cloned into the pET28a(+) vector, and heterologously expressed in Escherichia coli Rosetta (DE3). To investigate the biosynthetic pathway of paeonol, the catalytic activities of candidate PoOMTs were tested using substrates 1 and 2 as methyl acceptors (Fig. 2b, Supplementary Fig. 3). HPLC analysis showed that only PoOMT1 efficiently catalyzed 4′-O-methylation of both substrates, producing product 6 (conversion rate >80%) (Fig. 2c, d) and product 7 (conversion rate >75%) (Fig. 2e, f), respectively. PoOMT2 and PoOMT4 exhibited minimal activities toward both substrates (conversion rate <5%) (Fig. 2b, Supplementary Fig. 4). Further enzymatic characterization revealed that PoOMT1 reached maximum conversion within 2 hours, with optimal pH of 8.0 and temperature of 37 °C, and its catalytic activity was independent of divalent metal ions (Supplementary Fig. 5).

Kinetic analysis (Fig. 2g, h) demonstrated that PoOMT1 has high affinity for both substrates, with K_m values of 56.30 ± 8.17 μM for substrate 1 and 33.38 ± 6.50 μM for substrate 2. Biosynthetic pathway specific OMTs exhibited significantly lower K_m values for substrates than promiscuous OMTs. Specifically, CcOMT8, the benzylisoquinoline alkaloid biosynthetic OMT, has a K_m value of 48.61 μM for (S)-scoulerine²⁹, whereas the promiscuous methyltransferase SmOMT showed K_m values exceeding 200 μM for multiple substrates³⁰. The K_m values of PoOMT1 toward substrate 1 and substrate 2 are only 56.30 μM and 33.38 μM, respectively, implying that these are likely the specific substrates of PoOMT1. Moreover, the enzyme also displayed high catalytic efficiency toward both substrates, with k_cat/K_m values of 60.46 mM⁻¹s⁻¹ and 44.88 mM⁻¹s⁻¹ for substrate 1 and substrate 2, respectively. Collectively, these results establish PoOMT1 as the key enzyme responsible for the 4′-O-methylation in paeonol biosynthesis and support its role in a parallel biosynthetic pathway.

Functional characterization of PoOMT1 in planta

Given that the cellular environment in tobacco may more closely resemble that of P. ostii than heterologous expression system, we further evaluated the catalytic activity of PoOMT1 toward 1 and 2 using an in vivo tobacco assay (Fig. 3a). Notably, none of the compounds 1, 2, 6 or 7 were detected in wild-type tobacco leaves. Upon confirming PoOMT1 expression (Fig. 3b), we injected substrate 1 into leaves and observed production of 6 exclusively in PoOMT1-overexpressing leaves (Fig. 3c). Similarly, injection of substrate 2 led to specific accumulation of 7 only in transgenic leaves (Fig. 3d). These results demonstrate that PoOMT1 catalyzes the conversion of both 1 to 6 and 2 to 7 in a plant cellular environment.

Fig. 3. Functional analysis of transient expression and silencing of PoOMT1.

a Fluorescence images of Nictiana benthamiana leaf under blue light after substrate injection. WT, wild-type; 1301-GFP, empty vector control; 1301-PoOMT1-GFP, PoOMT1-overexpressing leaf. b qRT-PCR analysis of PoOMT1 expression in different leaf treatments. Different letters indicate significant differences at P = 0.0044 using one-way ANOVA. c HPLC profiles of leaf extracts after injection of substrate 1. d HPLC profiles of leaf extracts after injection of substrate 2. e Fluorescence images of P. ostii root under blue light after agroinfiltration. TRV2-GFP, empty vector control; TRV2-PoOMT1-GFP, PoOMT1-silenced root. f qRT-PCR analysis of PoOMT1 expression in different root treatments. Different letters indicate significant differences at P = 0.000096 using one-way ANOVA. g Contents of 6 and 7 measured in different root treatments. Different letters indicate significant differences at P = 0.000161 for 6 and P = 0.0000688 for 7 using one-way ANOVA. Data are presented as mean values ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

The establishment of mature genetic transformation systems in tree peony has been challenging, largely due to its long life cycle and difficulties in tissue culture³¹. Virus-induced gene silencing (VIGS) offers a rapid alternative for transient transcript suppression of homologous genes, bypassing the need for stable transformation³². This approach has been successfully applied in tree peony for validating gene function in vivo^33,34. To examine the role of PoOMT1 in paeonol biosynthesis in P. ostii, we employed a VIGS system to suppress its expression in root. A tobacco rattle virus 2 (TRV2) vector was constructed with green fluorescent protein (GFP) as a reporter³⁵.

At 6 days after infiltration, green fluorescence was visible under blue light in either empty vector or PoOMT1-silenced roots, but not in wild-type roots (Fig. 3e), confirming successful vector delivery. qRT-PCR analysis revealed no significant difference in PoOMT1 expression between wild-type and empty vector roots; however, PoOMT1 transcript level in silenced roots was reduced to 6.50% of that in wild-type roots (Fig. 3f). Consistent with this, the content of 7 was comparable in wild-type and empty vector roots, but decreased markedly in PoOMT1-silenced roots, reaching only 8.64% of wild-type level. Similarly, the level of 6 remained unchanged between wild-type and empty vector roots, but declined significantly in silenced roots to 33.02% of the wild-type level (Fig. 3g). In addition, when PoOMT2 and PoOMT4 were silenced, the contents of 6 and 7 in silenced roots did not significantly decrease compared to the wild-type and empty vector roots, indicating that they were not potential candidates (Supplementary Fig. 6). These results confirm that PoOMT1 is essential for paeonol biosynthesis in P. ostii phloem and functions in both proposed pathways. Together with in vitro assays, heterologous expression in plant (tobacco) system and VIGS silencing in vivo, our findings demonstrate that PoOMT1 mediates paeonol biosynthesis in P. ostii via a parallel pathway.

Substrate regioselectivity of PoOMT1 in paeonol biosynthesis

Metabolite profiling of P ostii phloem identified several hydroxyacetophenone derivatives in addition to compounds 1 and 2 (Fig. 1b). PoOMT1 catalyzed 4′-O-methylation of both 1 and 2 with similar efficiency, indicating that the presence of a 2′-OH group does not affect its regioselectivity (Fig. 2g, h). To systematically assess the regioselectivity of PoOMT1 toward hydroxyacetophenones, we selected a series of naturally occurring analogs from P. ostii for activity assays. First, 3′,4′-dihydroxyacetophenone (12) (Supplementary Fig. 7), which contains an ortho-hydroxyl group introduced into compound 1, was tested. PoOMT1 showed markedly reduced relative activity toward this substrate, declining to 16.8% (Fig. 4a, Supplementary Fig. 8). Next, 2′,4′,6′-trihydroxyacetophenone (5) and 2′,3′,4′-trihydroxyacetophenone (4), which bear meta- and ortho-hydroxyl group introductions into compound 2, respectively, were evaluated. Both structural modifications led to decreased enzymatic activity. Introducing a meta-hydroxyl group reduced activity to approximately 33.5% (Fig. 4a, Supplementary Fig. 9), whereas the ortho-hydroxylated analog retained 84.7% relative activity (Fig. 4a, Supplementary Fig. 10). Comparison with reference standards confirmed that, regardless of the position of additional hydroxyl groups on 1 or 2, methylation occurred exclusively at the 4′-OH position.

Fig. 4. Substrate regioselectivity profiling of PoOMT1.

a Relative activity of PoOMT1 toward structural analogs of substrate 1. b Relative activity of PoOMT1 toward 3′/5′-hydroxyacetophenone analogs normalized to activity with substrate 1. c Structures of 2′/6′-hydroxyacetophenone analogs tested in this study. The value within red background indicates the relative activity toward the 4′-hydroxyl group of different substrates, whereas the value within blue background indicates the relative activity toward the 3′/5′-hydroxyl groups of different substrates. Data are presented as mean values ± SD (n = 3 independent experiments).

To further probe whether PoOMT1 could methylate non-4′-hydroxyl positions, we tested compounds lacking the 4′-OH group, including 3′-hydroxyacetophenone (14), 2′,3′-dihydroxyacetophenone (13), 2′,5′-dihydroxyacetophenone (3), 2′-hydroxyacetophenone (15), and 2′,6′-dihydroxyacetophenone (16). When a 3′- or 5′-hydroxyl group was present (14, 13, and 3), only trace catalytic activity ( ≤ 5%) was detected (Fig. 4b, Supplementary Figs. 11–13). To further elucidate the precise chemical structures of the products and identify the methylation sites catalyzed, products 6, 7, 17, 18, 20, 21 and 22 were purified and structurally characterized using nuclear magnetic resonance (NMR) spectroscopy, including ¹H NMR, ¹³C NMR, and heteronuclear multiple bond correlation (HMBC) (Supplementary Figs. 14–34). The structural identification was consistent with the bioactivity results. In contrast, no activity was observed toward compounds containing only a 2′- or 6′-OH group (15 and 16; Fig. 4c). These results demonstrate that PoOMT1 exhibits strict regioselectivity for the 4′-OH group of hydroxyacetophenones, providing a biochemical basis for the paralled biosynthesis of paeonol in P ostii.

PoOMT1 architecture and SAH binding

To elucidate the three-dimensional structure of PoOMT1, we purified the recombinant protein to high homogeneity via Ni‑NTA affinity chromatography and size‑exclusion chromatography (Supplementary Fig. 35). To improve the probability of protein crystallization, by co-incubating with the methyl donor product S-adenosyl-L-homocysteine (SAH), we solved the crystal structure of the PoOMT1/SAH complex at 2.03 Å resolution (PDB ID: 9WAZ; Supplementary Table 3). The electron density for SAH was well-defined in the structure (Fig. 5a). The PoOMT1/SAH complex crystallized as a monomer, contrasting with the dimeric organization typically seen in plant Type I OMTs^36,37. This monomeric form is not unprecedented. A few other Type I OMTs such as IOMT8 from Medicago sativa (PDB ID: 6CIG)³⁸ and 6OMT from Thalictrum flavum (PDB ID: 5ICC)³⁹ have also been reported as monomers (Supplementary Fig. 36). However, size-exclusion chromatography analysis indicated an apparent molecular weight of approximately 70 kDa for PoOMT1/SAH (Fig. 5b), suggesting a dimeric state in solution. Further analysis of crystal packing revealed a homodimer formed through symmetric interface interactions (Fig. 5c). Structural alignment identified the top three structural homologs of PoOMT1 as emodin 1-O-methyltransferase Plu4891 from Photorhabdus luminescens (RMSD = 2.6 Å)⁴⁰, caffeic acid 3-O-methyltransferase MsCOMT from Medicago sativa (RMSD = 3.0 Å)³⁶, and emodin 14-O-methyltransferase GedA from Aspergillus terreus (RMSD = 3.5 Å)⁴¹ (Supplementary Fig. 37). Therefore, structural features indicate that PoOMT1 belongs to a typical plant Type I methyltransferase. The catalytic activity of SAM-dependent methyltransferases is generally inhibited by SAH^30,42. Accordingly, when equimolar amounts of SAH molecules were supplemented into the reaction system, the catalytic activity of PoOMT1 toward substrates 1 and 2 was significantly inhibited (Supplementary Fig. 38), further confirming that PoOMT1 is a typical SAM-dependent methyltransferase.

Fig. 5. Crystal structures of PoOMT1 and analysis of SAH binding.

a Cartoon representation of the PoOMT1/SAH complex. The F_o–F_c electron density map (contoured at 2.0σ) of SAH is shown as blue mesh. b Size-exclusion chromatography profile of PoOMT1 and corresponding calibration curve of standard proteins. c Ribbon diagram of the PoOMT1 homodimer in complex with SAH. d Left: Domain architecture of the PoOMT1/SAH complex, with the N-terminal domain (NTD) in pink and C-terminal domain (CTD) in gold. Right: Interactions between SAH and PoOMT1 residues. Black, yellow, and red dashed lines represent hydrogen bonds, hydrophobic interactions, and salt bridges, respectively. e Relative catalytic activities of the PoOMT1 mutants toward substrates 1 and 2. Data are presented as mean values ± SD (n = 3 independent experiments). The WT control data are shared with Fig. 6e as the mutants presented in both panels were analyzed concurrently with a common WT control group. Source data are provided as a Source Data file.

PoOMT1 consists of two functional domains: a C‑terminal catalytic domain housing the substrate‑binding and active sites (colored gold in Fig. 5d, left), and an N‑terminal domain involved in dimerization that forms the back wall of the methyl‑acceptor binding pocket (colored pink in Fig. 5d, left). SAH is stabilized in the C‑terminal binding pocket through multiple hydrogen bonds and hydrophobic interactions with residues G203, L225, D244, M245, K258, and I260 (Fig. 5d, right), all highly conserved among plant Type I OMTs. Site‑directed mutagenesis showed that the M245A mutant substantially reduced catalytic activity toward substrates 1 and 2, whereas the D244A, K258A, and I260A mutants abolished activity entirely (Fig. 5e).

Structural basis of regioselectivity and dual-substrate specificity in PoOMT1

To elucidate the catalytic mechanism of PoOMT1 and the structural basis for its regioselectivity toward hydroxyacetophenones, we solved the crystal structures of two ternary complexes: PoOMT1/SAH/1 and PoOMT1/SAH/2, at 1.83 Å and 2.28 Å resolutions, respectively (Supplementary Table 3). Both complexes crystallized as monomers (Supplementary Fig. 39) and were modeled as homodimers (Fig. 6a, b, top). Well-defined electron densities for SAH/1 and SAH/2 confirmed the same binding pose (Fig. 6a, b, down, Fig. 6c). In plant Type I OMTs, a conserved His-Asp-Glu triad typically mediates methylation^39,43. Sequence alignment revealed that in PoOMT1, Asp and Glu are replaced by Asn263 and Gln324, respectively (Supplementary Fig. 40). Structural analysis of the PoOMT1/SAH/2 complex showed that the NE2 atom of His262 and the backbone amide of Asn263 form hydrogen bonds with the 4′-OH group of substrate 2, while Gln324 hydrogen-bonds via its Nε atom to the NE1 atom of His262 (Fig. 6d). This interaction network resembles the S_N2-like mechanism (Supplementary Fig. 41) mediated by the canonical His-Asp-Glu triad in other Type I OMTs^39,43. Mutations H262A, N263A, and N263D completely abolished catalytic activity toward substrate 2. Q324A mutation nearly eliminated activity, whereas the Q324E mutation enhanced activity over 2-fold (Fig. 6e), likely due to the stronger electron donating capacity conferred by the increased acidity of Glu. Identical trends were observed in the catalytic activities of mutations toward substrate 1, suggesting that the unique His262-Asn263-Gln324 triad in PoOMT1 may employ a similar S_N2-like mechanism but operate through distinct cooperativity. Moreover, leveraging the stronger methylation activity of Q324E, the heterologous synthesis of paeonol can be enhanced through the integration of metabolic engineering strategies. Similarly, gene editing of PoOMT1 is also expected to achieve efficient production of paeonol. Further, when Gln324 was mutated to the acidic Asp, the catalytic activity did not increase as expected, but instead slightly decreased (Fig. 6e). This is attributed to the shortened side chain length, which caused the interaction distance to exceed 4 Å, thereby greatly weakening the electron donating capacity of the group (Supplementary Fig. 42). Therefore, leveraging the electronegativity of key catalytic residues to engineer methyltransferases, attention need be paid to the size of the group and the interatomic distance to improve the likelihood of successful engineering.

Fig. 6. Structural basis for regioselectivity and dual-substrate specificity in PoOMT1.

a Cartoon representation of the PoOMT1/SAH/1 ternary complex. The F_o–F_c omit electron density maps (contoured at 2.0σ) of SAH and 1 are shown as blue mesh. b Cartoon representation of the PoOMT1/SAH/2 ternary complex. The F_o–F_c omit electron density maps (contoured at 2.0σ) of SAH and 2 are shown as blue mesh. c Structural superposition of SAH/1 and SAH/2 from PoOMT1 complex structures. d Catalytic center and essential residues in PoOMT1/SAH/2 ternary complex. e Relative catalytic activities of the region A residues mutants toward substrates 1 and 2. The WT control data are shared with Fig. 5e as the mutants were analyzed concurrently with a common WT control group. f Interactions between the 1/2 and the substrate binding pocket. g Relative catalytic activities of the region C residues mutants toward substrates 1 and 2. h Four functional regions in the substract binding pocket. i The cross-section of the T-shaped architecture pocket of PoOMT1 with substrate 2. j Relative catalytic activities of the region D residues mutants toward substrates 1 and 2. Data are presented as mean values ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Superposition of the PoOMT1/SAH/1, PoOMT1/SAH/2, and PoOMT1/SAH structures revealed nearly identical conformations, indicating minimal structural change upon substrate binding (Supplementary Fig. 43). Both substrates bind in almost the same position (Fig. 6c), with their 4′-hydroxyl groups oriented toward and in close proximity (3.0 Å) to the catalytic residue His262 (Fig. 6d), rationalizing the preferential methylation at this position. Analysis of substrate interactions showed that the 4′-hydroxyl forms hydrogen bonds with His262 and Asn263, and mutagenesis studies (Fig. 6e) confirmed their roles in anchoring 4′-hydroxyl. The acetyl group is stabilized by hydrophobic interactions with Leu126 and Leu120, and its carbonyl oxygen hydrogen-bonds with Asn121 (Fig. 6f). The single-site mutations reduced activity by ∼50% due to compensatory effects. However, double mutations led to <10% residual activity, and the triple mutant was inactive (Fig. 6g), indicating these residues are essential for anchoring the acetyl group. PoOMT1 thus employs a dual-anchor mechanism that stabilizes both the acetyl group and the 4’-hydroxyl group of the substrates. No interactions were observed between the 2′-hydroxyl of substrate 2 and the binding pocket (Fig. 6f), consistent with similar catalytic efficiencies toward substrates 1 and 2.

To further explore the mechanism of dual-substrate specificity, we divided the substrate binding pocket of PoOMT1 into four regions (Fig. 6h). Regions A (H262 and N263) and C (L120, N121, and L126,) facilitate the dual-anchor mechanism, while regions B (F153 and Y154) and D (L117, F312, and L315) are predominantly hydrophobic and contain bulky residues. Cross-sectional analysis revealed a characteristic T-shaped topology (Fig. 6i). Region B, at the top of the “T”, imposes steric constraints via F153 and Y154, accounting for the markedly reduced activity against substrates with additional C3′ or C6′ hydroxyl groups (e.g., substrates 5 and 12). Region D, at the bottom, provides space to accommodate the 2′-hydroxyl group. Mutating L117, F312, and L315 in region D to bulkier tryptophan abolished activity toward substrate 2 but largely preserved activity toward substrate 1 (Fig. 6j). This T-shaped binding pocket thus enables similar affinity for both substrates 1 and 2, and provides a structural basis for the parallel biosynthesis of paeonol by PoOMT1.

Functional divergence of PoOMT1 homologs underpins its unique catalytic mechanism

To further substantiate the catalytic mechanism, we conducted a homology search using PoOMT1 as query and identified 89 OMT homologs with >50% sequence identity (Supplementary Fig. 44). For functional and sequence analyses, we selected three OMTs from the phylogenetically proximal order Saxifragales: BsOMT from Bergenia scopulosa, CjOMT from Cercidiphyllum japonicum, and HvOMT from Hamamelis virginiana, along with TgOMT from Ternstroemia gymnanthera (Ericales), AtOMT from Akebia trifoliata (Ranunculales), MgOMT from Magnolia grandiflora (Magnoliales), TvOMT from Tetrastigma voinierianum (Vitales), MmOMT from Matricaria matricariodes (Asterales), AaOMT from Ailanthus altissima (Sapindales) and GsOMT from Gylcine soja (Fabales) as evolutionarily distant controls (Supplementary Table 4). Sequence comparisons revealed high identity among Saxifragales OMTs (CjOMT: 75%, HvOMT: 71%, BsOMT: 66%), whereas the maximum identity to PoOMT1 observed in any other OMT was 64%. (Fig. 7a). Catalytic assays with excess substrates 1 and 2 revealed significant functional divergence among the OMTs. PoOMT1 exhibited the highest catalytic activity toward substrates 1 and 2. The other Saxifragales OMTs also maintained appreciable activity toward both substrates, whereas OMTs from other plant orders showed low or no detectable activity. Notably, although both TvOMT and MmOMT share 57% sequence identity with PoOMT1, they displayed markedly different catalytic activities (Fig. 7b, c; Supplementary Fig. 45). Sequence alignment identified the catalytic triad as a key variable region responsible for the observed differences in catalytic activity. PoOMT1 possesses a His-Asn-Gln triad, and the other three Saxifragales OMTs have His-Asn-Glu, while seven OMTs from other plant orders contain the conserved His-Asp-Glu triad (Fig. 7d). This highlights the importance of the His-Asn-Gln triad for the parallel biosynthesis of paeonol by PoOMT1. Moreover, Saxifragales have diverged evolutionarily from other plant orders at the catalytic triad, with Asn263 replacing the conserved Asp residue. Additionally, in the substrate binding pocket, they carry Tyr154 and Leu117 in regions B/D, whereas bulky Trp and Phe residues occupy these positions in other plants (Supplementary Fig. 46).

Fig. 7. Functional characterization of PoOMT1 homologs.

a The phylogenetic analysis of PoOMT1 homologs. b Relative catalytic activities of PoOMT1 homologs toward substrate 1. c Relative catalytic activities of PoOMT1 homologs toward substrate 2. d Sequence alignment of key residues of PoOMT1 homologs. Saxifragales OMTs are indicated by light blue shading, while those from other orders are marked with pink shading. Data are presented as mean values ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Residue conservation analysis among PoOMT1 homologs further revealed that catalytically important residues (N263 and Q324) and substrate binding residues (L117, N121, and Y154) are poorly conserved (Supplementary Fig. 47), aligning with their phylogenetic distribution. These findings reinforce the proposed catalytic mechanism of PoOMT1 and provide additional enzymatic tools for the synthetic biological production of paeonol.

In conclusion, we identified PoOMT1, a 4′-O-methyltransferase from P. ostii that catalyzes dual substrates in a parallel paeonol biosynthetic pathway and exhibits strict regioselectivity for the 4′-OH group of hydroxyacetophenones. Structural and mutagenesis analyses revealed a T-shaped substrate binding pocket that underlies this distinctive catalytic capability. In this architecture, regions A and C form a dual-anchor system that positions the substrates and dictates regioselectivity, whereas regions B and D impose steric constraints that enable dual-substrate recognition. A unique His-Asn-Gln catalytic triad in region A mediates 4′-O-methylation. Furthermore, we characterized ten PoOMT1-like enzymes from other plant species. Comparison of their catalytic profiles and sequences reinforced the essential role of the T-shaped binding pocket. Our study systematically elucidates the parallel 4′-O-methylation step in paeonol biosynthesis and establishes a foundation for fully resolving its pathway and enabling its production by synthetic biology.

Methods

Plant materials and chemicals

P. ostii plants used in this study were collected from the Peony Germplasm Resource Nursery of Northwest A&F University (108.07° E, 34.26° N) and identified by Prof. Yanlong Zhang. Fresh petal, leaf, stem, and root phloem were sampled, immediately frozen in liquid nitrogen, and stored at −80 °C. All samples were processed in three biological replicates. N. benthamiana was used for transient expression assays.

Reference standards were purchased from Yuanye Bio-Technology (Shanghai, China). HPLC-grade acetonitrile and formic acid were purchased from Thermo Fisher Scientific (Waltham, MA, USA). All crystallization materials and reagents were purchased from Hampton Research (Laguna Niguel, CA, USA).

Metabolome analysis

Metabolites of P. ostii phloem were extracted and analyzed by Metware (Wuhan, China) using a UPLC-ESI-MS/MS system (UPLC, SHIMADZU Nexera X2; MS, Applied Biosystems 4500 Q TRAP) equipped with an Agilent SB-C18 column (100 mm × 2.1 mm, 1.8 μm). Separation was performed using a gradient of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile) at 0.35 mL/min and 40 °C: 0–9 min, 5%–95% B; 9–10 min, 95% B; 10–11.1 min, 95–5% B; and 11.1–14 min, 5% B. The injection volume was 4 μL. MS detection used an ESI source in positive/negative mode with ion spray voltages of 5500 V/−4500 V, source temperature 550 °C, and curtain gas 25 psi. Metabolites were identified using the Metware database.

RNA-seq analysis

Total RNA was individually extracted from petal, leaf, stem, and phloem using RNA Prep Pure Plant Plus kit (Tiangen, Beijing, China). The quality and quantity of RNA were assessed with a NanoDrop 2000 (Thermo, CA, USA) and an Agilent Bioanalyzer 2100 (Agilent Technologies, CA, USA), respectively. Libraries were constructed by BGI (Wuhan, China) and sequenced on a BGISEQ-500 platform. Clean reads were mapped to P. ostii genome⁴⁴ using HISAT2⁴⁵ and Bowtie2⁴⁶. Gene expression level was calculated by RSEM⁴⁷ and expressed as the FPKM value⁴⁸. Differential expression was analyzed using the DESeq2 (v1.22.2)⁴⁹ (Q value ≤ 0.05). Functional annotation was performed via BLASTx against Nr and KEGG databases, and GO terms were assigned with Blast2GO.

Transient expression in tobacco

The freeze-thaw method⁵⁰ was used to introduce pCAMBIA1301-PoOMT1-GFP and the empty pCAMBIA1301-GFP vector into Agrobacterium tumefaciens strain GV3101. Transformed bacteria were cultured in LB medium supplemented with kanamycin at 28 °C shaking at 175 rpm for 3 h, followed by additional incubation for 12 h. Bacterial cells were centrifuged and resuspended in infiltration buffer (10 mM MgCl₂, 250 mM MES, and 10 mM acetosyringone) to a final OD₆₀₀ of 1.0. The suspension was injected into N. benthamiana leaves using a 1 mL syringe. Infiltrated plants were kept in darkness at 22 °C for 24 h, then transferred to light for 48 h. GFP fluorescence was visualized using a UV lamp. After fluorescence confirmation, 1 mL of substrate 1 or 2 (0.5 mg/mL) was injected into the fluorescent leaf areas, followed by incubation in darkness for 24 h and under light for 48 h. Leaves were then harvested for qRT-PCR and analysis of product 6 or 7.

Virus-induced gene silencing assay

The silencing of PoOMT1 in P. ostii root by VIGS was conducted as previously described³⁵. TRV1 and TRV2-GFP plasmids were maintained in our laboratory. To generate TRV2-PoOMT1-GFP, a 353 bp fragment of PoOMT1 was PCR-amplified using specific primers (Supplementary Data 3). All TRV constructs were transformed into A. tumefaciens GV3101. Transformed strains were cultured overnight at 28 °C with shaking at 220 rpm. Bacterial cells were centrifuged and resuspended in infiltration buffer (10 mM MgCl₂, 250 mM MES, and 10 mM acetosyringone) to an OD₆₀₀ of 1.0. Cultures containing TRV1 and TRV2 constructs were gently shaken for 4 h at 28 °C and mixed in a 1:1 ratio (v/v). For vacuum infiltration, whole seedlings were submerged in the infiltration buffer and subjected to 0.8 MPa pressure for 1 h. After infiltration, seedlings were rinsed, replanted, and placed in an artificial climate chamber at 4 °C overnight in darkness before returning to room temperature. Roots from non-infiltrated seedlings and seedlings infiltrated with TRV2-GFP served as controls. At 3 d after infiltration, the seedlings were collected for GFP analysis. Seedlings were collected at 3 days post-infiltration for GFP analysis, and roots were harvested at 6 days for qRT-PCR and quantification of 6 and 7.

qRT-PCR analysis

First-strand cDNA was prepared from 1 μg of total RNA per sample using the FastKing RT Kit with gDNase (Tiangen, Beijing, China). qRT-PCR was performed on a LightCycler Real-Time PCR System using SYBR^® Premix Ex Taq^TM Kit (Takara). Each 20 μL reaction contained 10 μL SYBR Premix Ex Taq II, 0.4 uL ROX, 0.8 μL each of forward and reverse primers, 0.8 μL cDNA, and 7.2 μL RNase-free ddH₂O. Amplification conditions were: 95 °C for 30 s; 40 cycles of 95 °C for 5 s, 55 °C for 30 s, and 72 °C for 30 s; 95 °C for 15 s; 60 °C for 60 s; and 95 °C for 15 s. All reactions were run in triplicate. Relative gene expression was calculated using the 2^-ΔΔCt method, with 18S rRNA as the internal reference. Data are presented as mean values ± SD of three biological replicates. Gene-specific primers used in this study are listed in Supplementary Data 3.

Protein expression and purification

For heterologous expression, pET28a(+)-sumo-OMT plasmids were transformed into E. coli Rosetta (DE3). Cells were cultured in 4 L LB medium (1% Tryptone, 0.5% Yeast extract, and 1% NaCl) with 50 μg/mL kanamycin at 37 °C with shaking at 220 rpm until OD₆₀₀ reached 0.6. Protein expression was induced with 0.4 mM isopropyl β-D-1-thiogalactopyranoside, followed by incubation at 18 °C for 18 h. Cells were harvested by centrifugation for 10 min. The pellets were resuspended in 120 mL of lysis buffer (20 mM Tris-HCl, 300 mM NaCl, pH 8.0) and disrupted using a high-pressure homogenizer (AH-NANO, ATS, China) at 800 bar. Lysates were clarified by centrifugation at 12,000 × g for 60 min, and the supernatant was loaded onto a gravity column pre-equilibrated in lysis buffer containing 5 mL Ni-NTA affinity resin (GenScript, Nanjing, China).

The column was washed sequentially with lysis buffer containing 25 mM imidazole (10 column volume, CV), 50 mM imidazole (6 CV), then the SUMO tag proteins were eluted with lysis buffer containing 100 mM imidazole (4 CV) and 300 mM imidazole (4 CV). Eluted proteins were digested with Ubl-specific protease 1(Ulp1) at 4 °C for 8 h to remove SUMO tag. The sample was reapplied to Ni-NTA resin and washed with lysis buffer containing 10 mM, 20 mM, and 50 mM imidazole (3 CV each). SUMO tag and other proteins were eluted with 300 mM imidazole (4 CV). Tag-free OMT proteins were further purified by size-exclusion chromatography using a HiLoad 16/600 Superdex 75 column (GE Healthcare) in a buffer containing 20 mM Tris-HCl pH 8.0, 150 mM NaCl, and 1 mM dithiothreitol. Peak fractions containing target protein were collected and checked by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Purified proteins were concentrated to 10.0 mg/mL using an Amicon Ultra-30 K filter (Millipore), flash-frozen in liquid nitrogen, and stored at −80 °C.

Site-directed mutagenesis

Site-directed mutagenesis of PoOMT1 was performed by PCR using primers (Supplementary Data 3) designed to generate desired mutations, with the wild-type pET28a(+)-sumo-PoOMT1 plasmid as template. PCR was conducted using Phanta Max Super-Fidelity DNA Polymerase (Vazyme, Nanjing, China) in a Biorad C1000 Thermal Cycler. Template DNA was digested with DpnI at 37 °C for 2 h, and PCR products were transformed into E. coli TOP10 competent cells. All mutations were verified by the Sanger sequencing. Mutant proteins were purified following the same protocol as the wild type.

Enzyme activity assay in vitro

Reactions were performed in a total volume of 100 μL containing 20 μg purified proteins, 2 mM SAM, 1 mM methyl acceptor substrate, and 50 mM Tris-HCl buffer (pH 8.0). After incubation at 37 °C for 2 h, reactions were terminated by adding 100 μL methanol. Samples were centrifuged at 12,000 × g for 30 min, and supernatants were filtered through a 0.45 μm membrane for HPLC and LC−MS analysis. All experiments were performed in triplicate.

To determine the optimal reaction time for PoOMT1 using 1 as the methyl acceptor, 15 time points (5, 10, 15, 30, 45, 60, 75, 90, 105, 120, 135, 150, 165, 185, and 240 min) were test. To test the dependence of divalent metal ions on PoOMT1 activity, MgCl₂, BaCl₂, CuCl₂, FeCl₂, CaCl₂ and divalent metal ion chelating agent EDTA were used individually in the final concentration of 5 mM. To optimize the reaction temperature, reactions were performed at different temperatures (4, 10, 18, 25, 37, 45, 50, and 60 °C).

To optimize the pH value, different reaction buffers with pH from 4.0−6.0 (Citric acid-Sodium citrate buffer), 6.0−8.0 (Na₂HPO₄-NaH₂PO₄ buffer), 7.0−9.0 (Tris-HCl buffer), and 9.0−11.0 (Na₂CO₃-NaHCO₃ buffer).

Kinetic parameters determination

Kinetic parameters of PoOMT1 with substrates 1 and 2 were determined in 100 μL reactions containing 50 mM Tris-HCl buffer (pH 8.0), 2 μg purified PoOMT1, 10 mM SAM, and varying substrate concentrations (1−400 μM). Reactions were incubated at 37 °C for 10 min, quenched, centrifuged, and analyzed by HPLC. The Michaelis–Menten constant (K_m) values and turnover number (k_cat) were calculated by following the Lineweaver−Burk plot method. All experiments were performed in triplicate.

HPLC and LC−MS analysis

HPLC analysis was performed on a Shimadzu-LC-20AT (Japan) equipped with an Ultimate^® XB-C18 column (4.6 mm×250 mm, 5 μm, Welch Materials, Inc., China) at 0.8 mL/min, using mobile phases of 0.1% formic acid in water (A) and acetonitrile (B). The gradient program was: 0−20 min, 10−100% B; 20−25 min, 100% B; 25−28 min, 100−10% B; and 28−35 min, 10% B. HPLC chromatogram was recoeded at 254 nm. Products were further confirmed by LTQ XL Orbitrap mass spectrometer (Thermo Fisher Scientific Inc.) in positive ion mode with 35% collision energy. Conversion rate in percentage was calculated from HPLC peak areas of product and substrate. Mutant relative activity was calculated as the ratio of the product peak area to that of the wild type. All experiments were performed in triplicate.

Preparation, purification, and structural characterization of methylation products

To obtain sufficient methylated products for structural characterization, the enzymatic methylation reaction of substrates 1–4 and 12–14 was scaled up to a 100 mL system. The reaction mixture contained 50 mM Tris-HCl buffer (pH 8.0), 1 mM substrate, 1 mM SAM, and 60 mg purified PoOMT1. The reaction was incubated at 37 °C for 2 h and then terminated by adding an equal volume of methanol. The mixture was concentrated to 5 mL using a rotary evaporator (RE-52AA, shyarong, Shanghai). Subsequently, the methylated products were purified by Shimadzu LC-20AT HPLC system with a reversed-phase XinC Technologies column (250 mm × 20 mm, 10 μm) and detection at wavelength of 330 nm and a flow rate of 10 mL/min. The elution program employed 0.1% formic acid aqueous solution (A) and acetonitrile (B), with a linear gradient of 10% to 100% B over 0–20 min. The injection volume was 200 μL, and the total time was 25 min. The target fractions were collected, concentrated, and lyophilized to yield solid product powders. The obtained products were dissolved in DMSO-d₆ for sample preparation, and their chemical structures were elucidated by NMR spectroscopy. The NMR spectra were recorded on a Bruker AVANCE III-600 spectrometer at 600 MHz for ¹H and 150 MHz for ¹³C in DMSO-d₆. Chemical shifts (δ) are reported in parts per million (ppm), and coupling constants (J) are given in Hertz (Hz).

Protein crystallization

The initial crystallization screening was performed at 18°C using sitting-drop vapor-diffusion method from commercial crystallization kits (Hampton Research and Rigaku Corporation) with near 1000 different reservoir compositions. In brief, purified PoOMT1 protein (10.0 mg/mL) was incubated with 5 mM SAH or 5 mM SAH and 5 mM of either substrate 1 or 2 on ice for 30 min prior to crystallization. 0.7 μL of incubated protein complex was mixed with 0.7 μL of reservoir solution in 96-well Cryschem plates (Hampton Research), which was equilibrated against 70 μL reservoir solution at 18 °C. Within 3−4 d, observe the crystal growth status and select crystals that the size and shape suitable for X-ray diffraction. The optimized crystallization condition of PoOMT1/SAH was followed by 0.2 M MgCl₂, 0.1 M Hepes pH 7.5, and 25% polyethylene glycol 3350. PoOMT1/SAH/1 crystals grew in the reservoir solution containing 0.2 M Ammonium acetate, 0.1 M Bis-Tris (pH 6.5), and 25% polyethylene glycol 3350. PoOMT1/SAH/2 crystals were obtained in the reservoir solution containing 0.2 M Ammonium sulfate, 0.1 M Bis-Tris (pH 6.5), and 25% polyethylene glycol 3350. All crystals were mounted in a cryo-loop and dipped with cryo-protectants through harvesting in reservoir solution supplemented with 20% glycerol and flash-frozen in liquid nitrogen prior to data collection.

Data collection, structure determination and analysis

Crystallographic data sets were collected on the beamlines BL18U1 at the Shanghai Synchrotron Radiation Facility⁵¹. Diffraction images were processed using XDS⁵². The structure of PoOMT1 complex was solved by molecular replacement using a Phaser from the CCP4 suite⁵³. The predicted structure of PoOMT1 by Alphafold2⁵⁴ was used as the searching model. Initial model building was performed with AutoBuild⁵⁵ and manually refined using Coot⁵⁶. Iterative refinement and structure validation were conducted using Phenix⁵⁷. Structural visualization analysis and figure preparation were carried out with the Protein-ligand interaction profiler⁵⁸ and PyMOL (The PyMOL Molecular Graphics System, v2.0 Schrödinger, LLC). Structures were superimposed using the Structure Similarity Matching (SSM) server⁵⁹.

Phylogenetic analysis

Phylogenetic reconstructions were performed using the “build” function of ETE3 (v3.1.3)⁶⁰ accessed through GenomeNet (https://www.genome.jp/tools/ete/). The tree was generated employing fast-tree with slow NNI and MLACC = 3 parameters for more exhaustive maximum-likelihood NNIs⁶¹. Bootstrap analysis comprised 1000 replicates. Subsequently, the phylogenetic tree was visually refined using ChiPlot (https://www.chiplot.online). Sequence alignments were generated using Clustal Omega (https://www.ebi.ac.uk/jdispatcher/msa/clustalo)⁶², ESPript 3.2 (https://espript.ibcp.fr/ESPript/ESPript)⁶³ and WebLogo (https://weblogo.berkeley.edu/)⁶⁴. GenBank accession numbers of other plant OMTs used are provided in Supplementary Data 4.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

X.Z., Y.Z., and W.H. conceived and designed the experiments. J.S. and Y.W. performed most of the experiments. P.L., and Q.Y. assisted with transient experiments on N. benthamiania. S.Y. and J.Y. assisted with transcriptome analysis. X.X. assisted with enzyme activity assay in vitro. X.Z., J.S., Y.H., and W.H. analysed the data. X.Z., J.S., and W.H. wrote the paper. J.G. and Y.Z. revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Yinghong Pan, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the Key project at central government level: The ability establishment of sustainable use for valuable Chinese medicine resources (No. 2060302 to X.Z.), the Natural Science Basic Research Program of Shaanxi (No. 2024JC-YBQN-0182 to X.Z.), the Open Project of Shanghai Key Laboratory of Plant Functional Genomics and Resources (No. PFGR202203 to X.Z.), the National Natural Science Foundation of China (No. 82304333 to W.H.), the Special Project for Cultivating Young Scientific and Technological Talents of Hubei Province (No. 2025DJA029 to W.H.), the Natural Science Foundation of Hubei Province (No. 2025AFB712 to W.H., and No. 2025AFB251 to J.S.), the Scientific and technological innovation project of CACMS (No. CI2023D002 to J.G.).

Data availability

The atomic coordinates and structure factors have been deposited in the Worldwide Protein Data Bank under accession codes 9WAZ (PoOMT1/SAH); 9WBA (PoOMT1/SAH/1); and 9WB6 (PoOMT1/SAH/2). Previously published structures used are available under accession codes 6CIG (MsIOMT8); 5ICC (Tf6OMT); 1KYW (MsCOMT); 6NEG (TfCOMT); 8BIC (Plu4891); 7WH9 (GedA). RNA-seq data have been deposited in the Genome Sequence Archive under accession number CRA019310. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77639-1.