Biosynthesis of the anti-diabetic metabolite montbretin A: glucosylation of the central intermediate mini-MbA.

Type 2 diabetes (T2D) affects over 320 million people worldwide. Healthy lifestyles, improved drugs and effective nutraceuticals are different components of a response against the growing T2D epidemic. The specialized metabolite montbretin A (MbA) is being developed for treatment of T2D and obesity due to its unique pharmacological activity as a highly effective and selective inhibitor of the human pancreatic α-amylase. MbA is an acylated flavonol glycoside found in small amounts in montbretia (Crocosmia × crocosmiiflora) corms. MbA cannot be obtained in sufficient quantities for drug development from its natural source or by chemical synthesis. To overcome these limitations through metabolic engineering, we are investigating the genes and enzymes of MbA biosynthesis. We previously reported the first three steps of MbA biosynthesis from myricetin to myricetin 3-O-(6'-O-caffeoyl)-glucosyl rhamnoside (mini-MbA). Here, we describe the sequence of reactions from mini-MbA to MbA, and the discovery and characterization of the gene and enzyme responsible for the glucosylation of mini-MbA. The UDP-dependent glucosyltransferase CcUGT3 (UGT703E1) catalyzes the 1,2-glucosylation of mini-MbA to produce myricetin 3-O-(glucosyl-6'-O-caffeoyl)-glucosyl rhamnoside. Co-expression of CcUGT3 with genes for myricetin and mini-MbA biosynthesis in Nicotiana benthamiana validated its biological function and expanded the set of genes available for metabolic engineering of MbA.

Introduction

In the last decade, diabetes mellitus has become one of the top 10 leading causes of death in the world, with 1.6 million people dying in 2016, compared with less than a million in 2000 (https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death). Type 2 diabetes (T2D) is often associated with obesity and is characterized by the body's inefficient use of insulin and hyperglycaemia causing long‐term damage to various organs. Sugar‐rich diets and lack of physical activity are among the drivers of a global T2D epidemic, which is increasingly affecting children and young adults in both developing and developed countries. In contrast, healthy diets and physical activity, as well as effective nutraceuticals and improved drugs, are some of the means of reducing the risk of T2D or controlling the disease at the pre‐diabetic stage.

Drugs that reduce postprandial carbohydrate breakdown are critical for controlling blood glucose levels in T2D patients. Current drugs against T2D commonly cause unpleasant gastrointestinal side effects that may lead to patient non‐compliance (Scheen, 1997, 2003; Scott and Spencer, 2000). In 2008, a large‐scale screen of 30 000 natural product extracts identified the specialized metabolite montbretin A (MbA) as a highly efficient (Ki = 8 nm) and highly selective inhibitor of human pancreatic α‐amylase (HPA) (Tarling et al., 2008; Williams et al., 2015). HPA catalyzes the initial step of starch digestion in the human gut. MbA has been successfully tested in rats towards drug development for treatment of T2D (Yuen et al., 2016). However, the only known source of MbA is the corm of montbretia (Crocosmia × crocosmiiflora) (Asada et al., 1988; Tarling et al., 2008), where MbA is produced in small amounts during a narrow window of time during seasonal corm development (Irmisch et al., 2018). Due to the rare occurrence and low abundance of MbA in nature and due to its complex chemical structure, extraction of MbA from montbretia corms and chemical synthesis, respectively, are insufficient to produce the metabolite for full‐scale drug development and application. MbA production using a synthetic biology approach through metabolic engineering of a heterologous plant or microbial system is an attractive alternative, but this approach requires knowledge of the genes and enzymes of MbA biosynthesis.

MbA is an acylated flavonol glycoside described as myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside 4′‐O‐α‐l‐rhamnosyl 1,4‐β‐d‐xyloside (Figure 1). MbA biosynthesis occurs almost exclusively in young developing corms and involves the formation and stepwise assembly of seven individual building blocks: the flavonol core myricetin, two units of UDP‐rhamnose (UDP‐Rha), two units of UDP‐glucose (UDP‐Glc), UDP‐xylose (UDP‐Xyl) and caffeoyl‐CoA (Irmisch et al., 2018). The first three steps of MbA assembly, starting with myricetin, proceed via myricetin 3‐O‐α‐l‐rhamnoside (MR) and myricetin 3‐O‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside (MRG) to myricetin 3‐O‐(6′‐O‐caffeoyl)‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside (mini‐MbA) (Irmisch et al., 2018) (Figure 2). These steps require two different UDP‐sugar dependent glycosyltransferases (UGTs), CcUGT1 (UGT77B2) and CcUGT2 (UGT709G2), and a BAHD‐acyltransferase (AT) reaction catalyzed by CcAT1 or CcAT2 (Irmisch et al., 2018). Transcripts of CcUGT1, CcUGT2 and CcAT1/2 showed differential and transient expression with highest transcript levels in young developing corms in the early summer, and low expression in old corms of the previous year matching the dynamics of MbA accumulation (Irmisch et al., 2018). These patterns of differential gene expression have enabled the discovery of the first three genes and enzymes of MbA biosynthesis, CcUGT1, CcUGT2 and CcAT1/2, as well as montbretia genes affecting myricetin biosynthesis (Irmisch et al., 2018, 2019). Using CcUGT1, CcUGT2, CcAT1 or CcAT2, together with montbretia genes of myricetin biosynthesis, specifically a MYB‐transcription factor (CcMYB4), flavonol synthase (CcFLS) and flavonol 3′5′‐hydroxylase (CcCYP2), we recently reconstructed the formation of mini‐MbA in Nicotiana benthamiana (Irmisch et al., 2019). The final three steps and the sequence of reactions of MbA biosynthesis beyond mini‐MbA remain to be identified and are proposed to involve three additional glycosylations with UDP‐Glc, UDP‐Xyl and UDP‐Rha (Figure 2).

Figure 1

Structure of montbretin A (MbA). The building blocks of MbA are labelled: myricetin, rhamnose (Rha), glucose (Glc), xylose (Xyl) and caffeic acid. Position of hydroxyl groups of the myricetin core and A,B,C‐designation of aromatic rings are shown. Dotted line marks the mini‐MbA structure. Montbretin A (MbA), myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside 4′‐O‐α‐l‐rhamnosyl 1,4‐β‐d‐xyloside.

Figure 2

Sequence of steps in the biosynthesis of montbretin A (MbA). (a) Schematic of the proposed pathway of MbA biosynthesis starting with myricetin. The six steps of MbA assembly are indicated with numbers below the arrows. Steps 1–3 have previously been elucidated and the enzymes catalyzing these reactions are shown as CcUGT1, CcUGT2 and CcAT1 or CcAT2, respectively (Irmisch et al., 2018). The sequence of glycosylation steps 4–6 was identified in this study. The cDNA encoding CcUGT3 (UGT703E1) (bold) and the function of its enzyme are described here. M, myricetin; R, rhamnose; G, glucose; X, xylose; C, caffeic acid. (b) Protein extracts from young corms (yC) catalyze the conversion of mini‐MbA and UDP‐Glc into MbA‐XR2 (peak 1, m/z 949). Formation of peak 1 was over 35‐fold lower in assays with protein extracts from the same amount of old corms (oC). (c) Total protein extracts from yC also catalyze the conversion of MbA‐XR2 and UDP‐Xyl into MbA‐R2 (peak 3, m/z 1081). Formation of peak 3 was over 20‐fold lower in assays with protein extracts from the same amount of oC. Samples were analyzed using LC‐MS in negative ionization mode and extracted ion chromatograms (EIC) 949 in panel (b) and EIC 1081 in panel (c) are shown. Identification of compounds corresponding to peaks 1 and 3 in the LC‐MS analysis is based on the comparison to authentic standards. (d) MS/MS fragmentation pattern of enzyme assay products corresponding to peak 1 and peak 3 shown in panels (a) and (b), respectively. Asterisks indicate mother ions for MSn fragmentation. Peak 1, MbA‐XR2 (myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl rhamnoside); peak 3, MbA‐R2 (myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl rhamnoside 4′‐O‐xyloside). MR, myricetin 3‐O‐rhamnoside; MRG, myricetin 3‐O‐glucosyl rhamnoside; mini‐MbA, myricetin 3‐O‐(6′‐O‐caffeoyl)‐glucosyl rhamnoside; MbA, myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside 4′‐O‐α‐l‐rhamnosyl 1,4‐β‐d‐xyloside.

Glycosylation of specialized metabolites in plants is typically catalyzed by UGTs of the GT‐1 family according to CAZy‐based glycosyltransferase classification (Caputi et al., 2012; Lombard et al., 2014). Plant UGTs use UDP‐activated sugar donors and contain a highly conserved plant secondary product glycosylation (PSPG) motif involved in binding of the UDP moiety (Gachon et al., 2005). UGTs are among the largest gene families in plants and can be divided into 16 different clades (clades A−P) based on sequence phylogeny (Ross et al., 2001; Caputi et al., 2012). The plant UGT family contains members in several clades that catalyze the glycosylation of flavonoids at various hydroxyl groups (Caputi et al., 2012) as well as UGTs that catalyze glycosylation of the sugar moiety of glycosides (Bowles et al., 2005; Frydman et al., 2013; Yonekura‐Sakakibara et al., 2014). In montbretia, we identified 159 different UGTs in the corm transcriptome including CcUGT1 (UGT77B2) and CcUGT2 (UGT709G2) involved in the formation of mini‐MbA (Irmisch et al., 2018).

Here we established the sequence of the final three steps in the biosynthesis of MbA, specifically steps in the conversion of mini‐MbA to MbA (Figure 2). Using a combined approach of transcriptome‐based gene discovery, cDNA cloning, enzyme characterization and gene expression in N. benthamiana, we functionally characterized CcUGT3 (UGT703E1) as a myricetin 3‐O‐(6′‐O‐caffeoyl)‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside 1,2‐β‐d‐glucosyltransferase catalyzing the glycosylation of mini‐MbA to myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside (MbA‐XR2) as the fourth step in MbA biosynthesis.

Results

Identification of the sequence of glycosylation from mini‐MbA to MbA

To identify the sequence of glycosylation from mini‐MbA to MbA, we tested protein extracts prepared from young corms (yC) or old corms (oC) that were collected on 10 June 2016 for UDP‐dependent glycosyltransferase (UGT) activity. This time point was previously shown to have high activity of MbA biosynthesis in yC and low activity in oC, enabling a screen for differentially expressed enzyme activities involved in MbA formation (Irmisch et al., 2018, 2019). Protein extracts of yC converted mini‐MbA (m/z 787) and UDP‐Glc into a compound with m/z 949 (peak 1 in Figure 2), which was identified by liquid chromatography‐tandem mass spectrometry (LC‐MS/MS) as myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside (MbA‐XR2, abbreviated as MbA minus xylose and second rhamnose) based on comparison of retention time and fragmentation pattern to an authentic standard (Figure 2, Figure S1). The MS/MS spectra of m/z 949 showed the loss of the caffeoyl moiety (minus 162 with its weaker ester bond), and MS/MS (MS3) analysis on the resulting m/z 787 showed the loss of the trisaccharide chain yielding myricetin (m/z 316). Assays with protein extracts from yC produced over 35‐fold more MbA‐XR2 than assays with protein extracts from the same amount of oC. These results supported a biosynthetic pathway in which UDP‐Glc dependent formation of MbA‐XR2 from mini‐MbA represents the fourth step in MbA biosynthesis (Figure 2).

As MbA possesses a xylose at the myricetin 4′‐hydroxy group (Figure 1), we also tested for the alternative scenario of a UDP‐Xyl dependent conversion of mini‐MbA. Protein extracts of yC converted mini‐MbA and UDP‐Xyl into a compound with m/z 919 (Figure S2). The fragmentation pattern of m/z 919 was indicative of the position of Xyl at the end of a trisaccharide chain, suggesting the incorporation of Xyl instead of Glc in MbA‐XR2. In addition to m/z 919, these assays produced a compound with m/z 1051 (Figure S2). The MS/MS fragmentation pattern of m/z 1051 showed the initial loss of 132 indicative of a second Xyl attached to the myricetin ring. However, these two compounds (m/z 919 and m/z 1051) with a Xyl at the end of the trisaccharide chain would not be relevant as precursors to MbA, which contains a Glc, but not Xyl, in this position.

Incubation of yC protein extracts with MbA‐XR2 and UDP‐Xyl resulted in the formation of a compound of m/z 1081 (peak 3 in Figure 2), which was identified as myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐1,2‐β‐d‐glucosyl 1,2‐α‐l‐rhamnoside 4′‐O‐β‐d‐xyloside (MbA‐R2, abbreviated as MbA minus second rhamnose) based on comparison of the retention time and fragmentation pattern with an authentic standard (Figure 2, Figure S1). The prominent daughter ion in the MS/MS spectra, m/z 949 agreed with the loss of the Xyl on the 4′‐hydroxyl group of myricetin (Figure 2). Assays with protein extracts of yC produced over 20‐times more MbA‐R2 compared with protein extracts derived from the same amount of oC. These results confirmed MbA‐R2, produced by UDP‐Xyl dependent transformation of MbA‐XR2 in step 5, as the penultimate intermediate in the MbA biosynthesis (Figure 2).

To test if yC protein extracts catalyze the complete conversion of mini‐MbA into MbA, we incubated protein extracts with mini‐MbA, UDP‐Glc and UDP‐Xyl with or without UDP‐Rha. These assays resulted in the formation of MbA‐XR2 (m/z 949), MbA‐R2 (m/z 1081) and MbA (m/z 1227) (Figure S3). MbA is present in yC protein extracts and could not be entirely removed in the extraction process. An increase (1.8 ± 0.15, n = 3, P = 0.001) in the formation of MbA was observed when UDP‐Rha was included as a substrate, along with a decrease in MbA‐XR2 (m/z 949) and MbA‐R2 (m/z 1081) indicating the efficient conversion of those intermediates into MbA. These results support a UDP‐Rha‐dependent reaction as the final step 6 of MbA biosynthesis (Figure 2).

The MbA pathway intermediates mini‐MbA, MbA‐XR2 and MbA‐R2 are present in young corms

We previously showed that the first two intermediates in MbA biosynthesis, MR and MRG, accumulate in higher amounts in yC compared with oC (Irmisch et al., 2018). Following the enzyme assays with yC and oC that established the sequence of steps in the MbA biosynthesis, we screened corm extracts for the proposed intermediates downstream of MRG that were previously not detected, specifically mini‐MbA, MbA‐XR2 and MbA‐R2 (Figure 2). We performed targeted metabolite analysis by LC‐MS of corm extracts of the 16 June 2017 time point. We used a long‐run LC method to achieve separation of the more complex intermediates and performed both full scan and MS/MS on masses m/z 787 corresponding to mini‐MbA, m/z 949 corresponding to MbA‐XR2, or m/z 1081 corresponding to MbA‐R2 (Figure 3). These analyses revealed the presence of proposed MbA pathway intermediates mini‐MbA (peak 2 in Figure 3), MbA‐XR2 (peak 5) and MbA‐R2 (peak 8) in extracts of yC but not in oC. Additional compounds with m/z 787 (peak 1,3), m/z 949 (peak 4,6) or m/z 1081 (peak 7) were present in yC or oC extracts. Their fragmentation patterns suggest that they represent various myricetin glycosides, based on the loss of one or multiple m/z 162 or m/z 146 (putatively Glc or Rha, respectively) and the occurrence of the daughter ion m/z 316 or 317.

Figure 3

Intermediates in MbA biosynthesis in young corms. (a) Detection of metabolites with m/z 787, m/z 949 and m/z 1081 in young corms (yC, grey) and old corms (oC, black) analyzed using a long‐run LC‐MS. Peaks 2, 5 and 8 represent mini‐MbA, MbA‐XR2 and MbA‐R2, respectively, which are known or proposed intermediates in the biosynthesis of MbA. These three metabolites were detected in yC extracts but not in oC extracts. Extracted ion chromatograms are shown. (b) Total ion chromatograms (TIC) from MS/MS experiments of m/z 787, m/z 949, m/z 1081 of yC extracts shown in (a) and EIC of authentic standards for mini‐MbA, MbA‐XR2 and MbA‐XR2. (c) MS/MS fragmentation patterns of metabolites corresponding to peaks 1–8 in panels (a) and (b). Peak 1, unidentified; peak 2, mini‐MbA (myricetin 3‐O‐(6′‐O‐caffeoyl)‐glucosyl rhamnoside); peak 3 and 4, unidentified; peak 5, MbA‐XR2 (myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl rhamnoside); peak 6 and 7, unidentified; peak 8, MbA‐R2 (myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl rhamnoside‐4′‐O‐xyloside). Asterisks indicate mother ion for MS/MS fragmentation.

Identification of UGT candidates for the glucosylation of mini‐MbA to MbA‐XR2 by transcriptome co‐expression analysis

We previously established co‐expression analysis as a successful strategy for discovery of genes in the formation of mini‐MbA (Irmisch et al., 2018). Here, we used Haystack co‐expression analysis across montbretia flowers, stems, leaves, stolons, yC and oC with CcUGT1, CcUGT2, CcAT1 and CcAT2 as baits to screen for candidate UGTs for the fourth step in the MbA biosynthesis, specifically the glucosylation of mini‐MbA to MbA‐XR2 (Figure 2). In this analysis, the UGT transcript DN67918_c7_g7_i4 showed a strong correlation (R ≥ 0.98) with CcUGT1, CcUGT2, CcAT1 and CcAT2, as well as over 16‐fold higher transcript abundance in yC compared with oC (>logFC4). The open reading frame (ORF) of DN67918_c7_g7_i4 translated into a protein of 361 amino acids (aa). Compared with known UGTs, the translated ORF lacked the C‐terminal region. A partly matching transcript TR29056_c3_g1_i1 covering the ORF for the C‐terminus was detected by screening a separate montbretia transcriptome made of corms harvested in July 2013 (corm_2013) with the DN67918_c7_g7_i4 sequence (Figure S4). PCR with primers spanning the complete ORF of both transcripts and cDNA template from yC of the 10 June 2016 time point generated two different full‐length UGT cDNAs, UGT703E1 and UGT703E2. These two sequences shared 88% nucleotide (nt) identity and 79% predicted aa identity. They shared 88% and 95% nt identify and 80% and 91% aa identity with DN67918_c7_g7_i4, respectively. In a phylogenetic analysis, both UGT703E1 and UGT703E2 fall into clade D of family 1 UGTs (Figure 4) while the previously characterized CcUGT1 and CcUGT2 belong to clade F and clade P, respectively (Irmisch et al., 2018). The closest characterized UGT in other species is UGT703B1 from crocus (Crocus sativus) (Figure 4), which glucosylates the 7‐hydroxyl group of kaempferol (Moraga et al., 2009).

Figure 4

Phylogeny of montbretia UGT703E1 (CcUGT3) and UGT703E2. Amino acid sequences of CcUGT1, CcUGT2, UGT703E1 (CcUGT3) and UGT703E2 were aligned with selected UGTs from other plant species using ClustalW and a neighbour‐joining tree was constructed using MEGA6. Alignments and phylogeny were used to cluster montbretia UGTs with known UGT clades. Montbretia UGTs characterized in this work are in bold and clades are labelled. Ac, Allium cepa; At, Arabidopsis thaliana; Bp, Bellis perennis; Ca, Catharanthus roseus; Cc, Crocosmia × crocosmiiflora; Cm, Citrus maxima; Cs, Crocus sativus; Db, Dorotheanthus bellidiformis; Gm, Glycine max; Gt, Gentiana triflora; Fa, Fragaria × ananassa; Ip, Ipomoea purpurea; Lb, Lycium barbarum; Mt, Medicago truncatula; Nt, Nicotiana tabacum; Pf, Perilla frutescens; Pg, Panax ginseng; Ph, Petunia hybrid; Pm, Pueraria montana; Sb Scutellaria baicalensis; Sr, Stevia rebaudiana; Vh, Verbena hybrid; Zm, Zea mays.

UGT703E1 converts Mini‐MbA into MbA‐XR2

To test UGT703E1 and UGT703E2 for glucosyltransferase activity, the cDNAs were expressed in E. coli and the recombinant proteins were Ni‐purified and assayed using mini‐MbA and UDP‐Glc followed by LC‐MS and LC‐MS/MS analysis of products. Assays with UGT703E1 resulted in the formation of a single product peak with m/z 949 identified as MbA‐XR2 based on matching retention time and fragmentation pattern with an authentic standard (Figures 5 and S1). No activity was observed with UGT703E2 or the empty vector control (Figure 5).

Figure 5

Enzyme activity of recombinant UGT703E1 (CcUGT3) and transcript expression profiles of UGT703E1 in young and old corms. (a) UGT703E1 and UGT703E2 were heterologously expressed in E. coli and Ni‐purified proteins were assayed for activity with mini‐MbA and UDP‐Glc. The control assay was performed with protein from E. coli containing the empty vector. Products were analyzed using LC‐MS and the extracted ion chromatograms (EIC, m/z 949) are shown. (b) The MSn fragmentation patterns of peak 1 are shown. Asterisks indicate mother ion for MSn fragmentation. Peak 1, MbA‐XR2 (myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl‐rhamnoside). (c) RNA was isolated from yC and oC harvested at different time points of corm development. Transcript abundance was determined by qRT‐PCR. Means and standard errors are shown (n = 3). Different letters above the data points indicate significant differences between harvest points. Asterisks (*) indicate the statistical significance between yC and oC for all time points.

We tested UGT703E1 for substrate specificity with UDP‐Glc as the sugar donor against different flavonoids and phenolics as potential sugar acceptors, specifically myricetin, MR, MRG, MbA‐G (MbA minus the terminal Glc), epicatechin, salicin and caffeic acid (Table S1). UGT703E1 showed no activity with any of those acceptor substrates (Table S1). It is noteworthy that MRG, which is missing the caffeoyl moiety compared with mini‐MbA, did not serve as a substrate for UGT703E1. In addition to UDP‐Glc, UGT703E1 also accepted UDP‐Xyl and UDP‐Rha as sugar donors with mini‐MbA as the acceptor substrate (Figure S5). The fragmentation pattern of the respective products with m/z 919 and m/z 933 showed the initial loss of m/z 162 (caffeoyl moiety) indicating the formation of a trisaccharide side chain with Xyl or Rha, respectively, instead of the terminal Glc in MbA‐XR2 (Figure S5). The turnover rate for mini‐MbA and UDP‐Glc was 1.2‐fold higher compared with the turnover rate of mini‐MbA with UDP‐Xyl and 19‐fold higher compared with mini‐MbA and UDP‐Rha (Figure S5). Similar patterns of different preference for UDP‐Glc over UDP‐Xyl and UDP‐Rha were observed when UGT703E1 was assayed simultaneously with UDP‐Glc, UDP‐Xyl and UDP‐Rha and mini‐MbA (Figure S5). These results established UGT703E1 as being CcUGT3, which catalyzes the conversion of mini‐MbA to MbA‐XR2 in step 4 of the MbA biosynthetic pathway (Figure 2).

Transcript expression profiles of UGT703E1 (CcUGT3) in yC and oC support a role in MbA biosynthesis

We measured profiles of transcript abundance of UGT703E1 (CcUGT3) during corm development in yC and oC from 10 June to 6 October 2016 using quantitative real‐time PCR (qRT‐PCR) (Figure 5). As with the previously described genes in mini‐MbA biosynthesis (Irmisch et al., 2018), transcript levels of UGT703E1 (CcUGT3) were low and did not significantly change across all time points in oC and were significantly higher in yC at all time points (Table S2). Transcript abundance was highest in young corms harvested in June, with over 28‐fold higher transcript levels compared with oC of the same time point and significantly lower in samples harvested after June, (Figure 5) matching the reported profiles of MbA accumulation and earlier pathways genes (Irmisch et al., 2018, 2019).

N. benthamiana leaves co‐expressing UGT703E1 (CcUGT3) produce MbA‐XR2 and MbB‐XR2

We previously showed that mini‐MbA (also abbreviated as MRG‐Caff) and mini‐MbB (MRG‐Cou), which contains a coumaroyl moiety instead of the caffeoyl moiety, could be produced in N. benthamiana by combined transient co‐expression of the three montbretia myricetin biosynthesis genes CcMYB4, CcFLS and CcCYP2 and the three montbretia mini‐MbA biosynthesis genes CcUGT1, CcUGT2 and CcAT1 (Irmisch et al., 2019). Here, we built upon this expression system to validate the function of UGT703E1 (CcUGT3) in planta. N. benthamiana leaves were infiltrated with Agrobacterium tumefaciens containing the construct 35S pro:UGT703E1 and A. tumefaciens harbouring individual 35S ‐gene constructs for the above mentioned six genes for myricetin and mini‐MbA biosynthesis (Figure 6). Plants expressing the six myricetin and mini‐MbA biosynthesis genes served as controls. Leaves were collected 5 days after infiltration, and MeOH/H2O extracts analyzed by LC‐MS, LC‐MS/MS and LC‐MS/QTOF. Control leaves expressing myricetin and mini‐MbA biosynthesis genes produced mini‐MbA (MRG‐Caff, m/z 787, peak 2 in Figure 6), mini‐MbB (MRG‐Cou, m/z 771, peak 3) and myricetin 3‐O‐(6′‐O‐coumaroyl)‐glucosyl glucoside (MGG‐Cou m/z 787, peak 1) consistent with previous work (Irmisch et al., 2019) (Figure S6, Table S3). A decrease in the peak areas corresponding to these products was observed when UGT703E1 (CcUGT3) was co‐expressed with genes for myricetin and mini‐MbA biosynthesis, indicative of UGT703E1 (CcUGT3) dependent conversion of mini‐MbA, mini‐MbB and MGG‐Cou. To detect potential products of these reactions, we screened leaf extracts for CcUGT3‐dependent occurrence of compounds with m/z 949 or m/z 933 corresponding to glucosylation (+Glc m/z 162) of mini‐MbA and MGG‐Cou or mini‐MbB, respectively. Leaves co‐expressing UGT703E1 (CcUGT3) produced a single unique m/z 933 compound (peak 6 in Figure 6) tentatively identified as MbB‐XR2. The fragmentation of the mother ion m/z 933 into m/z 787 (MRGG) and m/z 316 (M) indicates the initial loss of the coumaroyl moiety followed by the loss of the trisaccharide chain (Figure 6). Three m/z 949 compounds (peaks 3, 4 and 5 in Figure 6) were specific to the co‐expression of UGT703E1 (CcUGT3). Peak 5 was identified as MbA‐XR2, the fourth intermediate in MbA biosynthesis, based on comparison of fragmentation pattern and retention time to an authentic standard (Figures 6 and S7). The fragmentation of m/z 949 (peak 3) into m/z 803 (MGGG) and m/z 316 (M) as well as the decrease of MGG‐Cou (m/z 787 peak 1, Figure S6) suggests the identity of peak 3 as myricetin 3‐O‐(glucosyl‐6′‐O‐coumaroyl)‐glucosyl glucoside (MGGG‐Cou). Peak identities were additionally supported by accurate masses (Table S3). The identity of m/z 949 peak 4 could not be resolved. As UGT703E1 (CcUGT3) accepts UDP‐Xyl in vitro we also screened for m/z 903 reflecting the addition of a xylose to mini‐MbB. Trace amounts of an m/z 903 peak, possessing less than 1% of the peak area of MbB‐XR2 (m/z 933) could be detected. None of the above mentioned peaks was detected in control samples.

Figure 6

Acylated myricetin trisaccharides produced by N. benthamiana transiently co‐expressing UGT703E1 (CcUGT3). N. benthamiana leaves were infiltrated with A. tumefaciens transformed with plasmids carrying the 35S‐promoter‐gene constructs for myricetin biosynthesis (CcMYB4, CcFLS, CcCYP2), mini‐MbA biosynthesis (CcUGT1, CcUGT2, CcAT1) with or without the additional construct 35S :CcUGT3 (UGT703E1) (a). Leaves were collected at day 5 after infiltration. Metabolites were extracted with MeOH/H2O and analyzed by LC‐MS and LC‐MS‐TOF utilizing the long LC run. Product identification was done based on their fragmentation patterns, accurate masses or/and comparison to authentic standards. The extracted ion chromatograms (EIC) m/z 933 (black) and m/z 949 (blue) (b), MS/MS fragmentation patterns of products formed in (a) (c), and selected product structures (d) are shown. Peaks 1 and 2, unidentified; peak 3, tentatively identified as myricetin 3‐O‐(glucosyl‐6′‐O‐coumaroyl)‐glucosyl glucoside; peak 4, unidentified; peak 5, myricetin 3‐O‐(glucosyl‐6′‐O‐caffeoyl)‐glucosyl rhamnoside (MbA‐XR2); peak 6, myricetin 3‐O‐(glucosyl‐6′‐O‐coumaroyl)‐glucosyl rhamnoside. M, myricetin (pink); R, rhamnose (yellow); G, glucose (green); Cou, coumaric acid (light blue); Caf, caffeic acid (dark blue). Asterisks indicate mother ion for MS/MS fragmentation.

Discussion

Building upon our previous work on the biosynthesis of mini‐MbA (Irmisch et al., 2018), we have now established the complete sequence of steps in the MbA biosynthetic pathway (Figure 2), which is supported by results from enzyme assays with corm extracts and detection of the previously missing intermediates mini‐MbA, MbA‐XR2 and MbA‐R2 in yC. This knowledge of the sequence of steps in the MbA biosynthesis, allowed us to focus our search for the next UGT in the biosynthetic pathway, CcUGT3, by targeting candidates that met the following criteria: (i) differential and higher expression in yC relative to oC; (ii) matching the temporal expression profiles of CcUGT1, CcUGT2, CcAT1 and CcAT2 with an early summer peak in yC; and (iii) enzyme activity with mini‐MbA and UDP‐Glc as substrates. The discovery and characterization of UGT703E1 met these criteria, which defines CcUGT3 as a glycoside glucosyltransferase (GGT) that catalyzes the formation of MbA‐XR2 as the fourth intermediate in MbA biosynthesis. The function of CcUGT3 was validated with the pathway reconstruction to MbA‐XR2 and MbB‐XR2 in N. benthamiana, extending the metabolic engineering of MbA to four out of six required steps starting from myricetin.

MbA biosynthesis proceeds via step‐wise assembly of individual building blocks

Enzymes that catalyze glycosylations of complex specialized metabolites in plants may be promiscuous with different intermediates of the biosynthetic system, which establishes a metabolic grid. Alternatively, these enzymes may be specific for individual intermediates in a linear pathway. As examples for the former, in steviol glycoside biosynthesis in stevia (Stevia rebaudiana), the initial glycosylation occurs at the C‐13 hydroxyl group, but subsequent glycosylations apparently do not follow a strict sequence (Richman et al., 2005). Similarly, in flavonoid biosynthesis in Arabidopsis thaliana, the sequence of xylosylation and glucosylation at C‐2″ or C‐6″, respectively, of cyanidin 3‐O‐glucoside, can vary (Morita et al., 2005; Sawada et al., 2005). In contrast, acylated anthocyanins in lobelia (Lobelia erinus) known as lobelinins appear to be synthesized through decorations of the anthocyanin core that follow a specific sequence starting with the formation of a coumaroyl rutinoside at the 3‐OH group of the anthocyanin core, followed by the glucosylation and malonylation on the 5‐OH before the anthocyanin B‐ring is modified (Hsu et al., 2017).

The results presented here, together with our previous work (Irmisch et al., 2018), suggest that biosynthesis of MbA proceeds through a linear pathway defined by a sequence of six steps involving five different glycosylations and an acylation (Figure 2). Previous work showed that CcUGT1, CcUGT2 and CcAT1 or CcAT2 function in this sequence to produce mini‐MbA from myricetin (Irmisch et al., 2018). Here, we showed that CcUGT3 extends the disaccharide moiety attached to the myricetin 3‐OH group yielding the final trisaccharide functionality. Completion of the trisaccharide moiety does not precede the acylation that produces mini‐MbA. Formation of the trisaccharide moiety at the myricetin C‐ring (Figure 1) is followed by the decoration of the myricetin B‐ring, which are the later steps similar to other flavonoid biosynthetic pathways (Yamazaki et al., 2002; Hsu et al., 2017). The montbretia UGTs responsible for the remaining formation of the rhamnosyl xyloside disaccharide moiety at the 4′‐hydroxy of the B‐ring of MbA remain to be identified. Given that the complete set of glycosylation of MbA was achieved in corm protein extracts with UDP‐sugars as sugar donors, we propose that the final formation of the 4′‐disaccharide in MbA biosynthesis is catalyzed by two remaining UGTs, in contrast with the alternative scenarios of acyl‐glucose‐dependent glycosyltransferase or possible transglycosylation reactions (Matsuba et al., 2010).

UGT703E1 (CcUGT3) is a clade D glycoside glucosyltransferase (GGT)

CcUGT3 catalyzes the fourth step in MbA biosynthesis, the conversion of mini‐MbA into MbA‐XR2, which involves formation of a 1,2‐linked trisaccharide chain using an acylated flavonol disaccharide substrate. A variety of GGTs mediating the formation of acceptor‐disaccharides, mainly flavonoid disaccharides, have been described (Yonekura‐Sakakibara et al., 2014; Di et al., 2015; Li et al., 2016; Rodas et al., 2016; Hsu et al., 2017). Less is known about GGTs catalyzing the formation of more complex glycosyl chains. For example, in soybean (Glycine max), GmF3G2″Gt (UGT79B30) is involved in the formation of the branched chain flavonoid trisaccharide kaempferol 3‐O‐glucosyl‐(1,2‐rhamnosyl‐1,6‐glucoside) (Di et al., 2015). In steviol glycoside biosynthesis, UGT76G1 forms the branch chain trisaccharide of rebaudioside A from stevioside or steviolbioside (Richman et al., 2005). Certain glycosylated triterpene saponins in soybean contain a trisaccharide side chain, produced by GGTs UGT73P2 (GmSGT2) and UGT91H4 (GmSGT3) (Shibuya et al., 2010). In Madagascar periwinkle (Catharanthus roseus), CaUGT3 exhibits a unique glucosyl chain elongation activity forming di‐, tri‐ and tetra saccharides with 1,6‐linkages in a sequential manner (Masada et al., 2009).

Montbretia CcUGT3 does not accept the flavonol disaccharide MRG as substrate, but uses the acylated mini‐MbA to catalyze formation of the trisaccharide chain, suggesting a substrate binding mechanism involving the recognition of the acyl group. In contrast, the common daisy (Bellis perennis) GGT BpUGAT involved in anthocyanin formation glucosylates both cyanidin 3‐O‐glucoside and the acylated cyanidin 3‐O‐(6″‐O‐malonyl) glucoside (Sawada et al., 2005).

Subfamily A of family 1 GTs has been described to contain GGTs catalyzing glycosylations of sugars attached to flavonoids (Bowles et al., 2005), and this clade contains the large majority of known GGTs that act on a variety of substrates including flavonoid glycosides as well as terpenoid and lignan glycosides (Noguchi et al., 2008; Shibuya et al., 2010; Jung et al., 2014; Di et al., 2015). In addition, a few GGTs have been reported belonging to clade H (stevia UGT76G1), clade E (crocus UGT707B1) and clade D (soybean GmSGT2, GmUGT73F4, GmUGT73F2) (Richman et al., 2005; Shibuya et al., 2010; Sayama et al., 2012; Trapero et al., 2012; Irmisch et al., 2018). The two characterized GGTs in MbA biosynthesis, CcUGT2 (Irmisch et al., 2018) and CcUGT3 (this work) fall into clades P and D, respectively, highlighting that prediction of GGT function by clade association is not currently possible (Yonekura‐Sakakibara and Hanada, 2011). Clade D is one of the largest groups of plant UGTs covering a wide array of different functions (Fukuchi‐Mizutani et al., 2003; Moraga et al., 2009; Shibuya et al., 2010; Caputi et al., 2012), and clade D UGTs of the same species may also cover diverse functions. Overall, most montbretia UGTs are members of clade D comprising about 35% of corm‐expressed UGTs (56 members) (Irmisch et al., 2018). In soybean, clade D is also dominant with 43 UGT members (Caputi et al., 2012).

Utility of UGT703E1 (CcUGT3) for MbA metabolic engineering

Similar to the previously characterized enzymes in MbA biosynthesis (Irmisch et al., 2018), CcUGT3 is highly stereo‐ and regio‐specific for product formation. However, when expressed in N. benthamiana, CcUGT3 can act in the formation of both MbA‐XR2 and MbB‐XR2. In the N. benthamiana system availability of caffeoyl‐CoA as opposed to coumaroyl‐CoA appears to be limiting, which favours biosynthesis towards MbB instead of MbA (Irmisch et al., 2018, 2019). These results confirm that enhancing access to caffeoyl‐CoA remains a critical issue for metabolic engineering of MbA biosynthesis in N. benthamiana. Both MbA and mini‐MbA are potent and selective inhibitors of HPA, while MbB is not an effective inhibitor (Tarling et al., 2008; Williams et al., 2015). MbA inhibited HPA with a Ki of 8 nm; the Ki of mini‐MbA was 93 nm. The CcUGT3 product MbA‐XR2 has a Ki of 42 nm making it a substantially more efficient inhibitor relative to mini‐MbA due to the addition of a glucose unit (Williams et al., 2015). Building upon our recent work that showed that mini‐MbA and mini‐MbB production can be engineered in N. benthamiana (Irmisch et al., 2019), we have now accomplished the CcUGT3 dependent production of MbA‐XR2 and MbB‐XR2. Ongoing and future work aims to further extend this system to the heterologous production of the complete MbA molecule, which will require two additional UGTs.

Experimental procedures

Plant material

Montbretia (Crocosmia × crocosmiiflora) plants of the variety Emily McKenzie were obtained, propagated and harvested (2016 plant material) as previously described (Irmisch et al., 2018). Additional, plants harvested on 16 June 2017 originated from young corms which had been separated from the old corm in November 2016. Nicotiana benthamiana plants were grown from seed in potting soil in a controlled environment chamber (day, 26°C; night, 22°C; 16 h/8 h light/dark cycle).

Metabolite extraction for detection of MbA pathway intermediates

Young corms (newly developing during the 2017 growing season) and old corms (1 year old) were harvested June 16, 2017, flash frozen and ground into a fine powder, of which 100 mg per sample was extracted with 400 μl 50% MeOH/H2O (v/v) (2 h shaking at 21°C). The undiluted supernatant was used for metabolite analyses. Extracts were analyzed using the long‐run LC‐MS method (described below).

UGT assays with corm protein extracts

UDP‐sugar dependent glycosyltransferase assays with corm protein extracts were done as previously described (Irmisch et al., 2018). In brief, 500 mg of powdered corm was extracted with 2.5 ml of buffer (100 mm NaPi, pH 7.4, 5 mm ascorbic acid, 5 mm sodium bisulfite, 5 mm dithiothreitol, 1 mm EDTA, 10% (v/v) glycerol, 1% (w/v) PVP, 4% (w/v) PVPP, 4% (w/v) Amberlite XAD‐4, 0.1% (v/v) Tween) for 1 h at 4°C. Following centrifugation at 4300  for 30 min (4°C), the supernatant was collected and desalted three times into assay buffer (10 mm Tris‐HCl, pH 7.5, 1 mm dithiothreitol, 10% (v/v) glycerol) using NAP‐5 columns (GE Healthcare, Little Chalfont Buckinghamshire, UK). Desalted protein extract (75 μl) was used in 150 μl total assay volume with 50 μm mini‐MbA or 50 μm MbA‐XR2 as acceptor and one or more of the following UDP‐sugars, 1 mm UDP‐Glu (Sigma‐Aldrich, Steinheim, Germany), 1 mm UDP‐Xyl (CarboSource Services, Athens, GA, USA) and 50 μl UDP‐RhaP. UDP‐RhaP was prepared as described in Irmisch et al. (2018). Assays were incubated at 21°C for 6 h in a Teflon‐sealed, screw‐capped 1‐ml GC glass vial and stopped by placing vials on ice after the addition of 150 μl MeOH. After centrifugation at 4300  for 20 min (4°C) supernatant was transferred into a fresh vial and analyzed for product formation by LC‐MS.

LC‐MS analysis

LC was performed on an Agilent 1100 HPLC (Agilent Technologies GmbH, Waldbronn, Germany) with Agilent ZORBAX SB‐C18 column (50 × 4.6 mm, 1.8 μm particle size) (Merck, Darmstadt, 370 Germany) using aqueous formic acid (0.2% v/v) (mobile phase A) and acetonitrile plus formic acid (0.2% v/v) (mobile phases B). The short‐run elution profile was: 0–0.5 min, 95% A; 0.5–5 min, 5–20% B in A; 5–7 min 90% B in A and 7.1–10 min 95% A. The flow rate was 1 ml × min−1 at a column temperature of 45°C. To improve separation of MbA pathway intermediates a long‐run elution profile was used: 0–7 min, 95% A; 7–9 min, 5–15% B in A; 9–20 min 15–18% B in A; 20–25 min 18–90% B in A; 25–27 min 90% B in A and 27.1–30 min 95% A). Flow rate was 1 ml × min−1 at a column temperature of 40°C. LC was coupled to an Agilent MSD Trap XCT‐Plus mass spectrometer equipped with an electrospray operated in negative ionization mode (capillary voltage, 4000 eV; temp, 350°C; nebulizing gas, 60 psi; dry gas 12 L min−1) and an Agilent 1100 Diode Array Detector (DAD, detection 200–700 nm, J&M Analytik AG, Aalen, Germany). MSn was conducted to obtain fragmentation patterns for compound identification. The LC/MSD Trap Software 5.2 (Bruker Daltonik, GmbH, Billerica, MA, USA) was used for data acquisition and processing. Enzyme assay products were quantified using an external MbA standard curve. Compounds were tentatively identified using their molecular masses and specific fragmentation patterns. Authentic standards were available for mini‐MbA, MbA‐XR2, MbA‐R and MbA (Williams et al., 2015). An enzyme assay using CcAT1, MRG and coumaroyl‐CoA was used to generate mini‐MbB (Irmisch et al., 2018).

Accurate mass measurement was performed on an Agilent 1290 Infinity UHPLC (Agilent Technologies GmbH, Waldbronn, Germany) utilizing the same column, mobile phase and long‐run as described above. The LC was coupled to an Agilent 6530 Accurate Mass Q‐TOF mass spectrometer equipped with an electrospray ion source operated in negative ionization mode (capillary voltage, 4000 eV; temp, 350°C; nebulizing gas, 60 psi; dry gas 12 L × min−1) and an Agilent 1290 DAD, detection 190–400 nm, J&M Analytik AG, Aalen, Germany). Accurate mass MS/MS experiment was conducted to analyse fragmentation patterns for compound identification. Hexakis(1H, 1H, 3H tetrafluoropropoxy)phosphazine/purine/ammonium trifluoroacetate mixture was used as API‐TOF Reference Mass solution. The Mass Hunter Workstation Software, version B.07.00, 2015 (Agilent Technologies) was used for data acquisition and processing.

Identification of candidate UGTs

To identify candidate UGTs the published transcriptome and differential expression (DE) data of yC and oC as well as DE data for other organs were utilized as previously described (Irmisch et al., 2018). Haystack (http://haystack.mocklerlab.org/) was used to filter UGTs whose expression patterns correlated with CcUGT1, CcUGT2 and at least one of the CcATs, CcAT1 or CcAT2. A correlation cutoff, of R ≥ 0.8 was used and results were filtered to meet criteria of at least 16‐fold (logFC(fold change) = 4) greater expression in yC compared with oC. A transcriptome assembly using sequencing data from corms harvested July, 2013 (corm_2013) was processed as described previously for yC/oC data (Irmisch et al., 2018). This data set was used to obtain the complete ORF of the target UGT sequence.

UGT cDNA cloning and heterologous expression in E. coli

Candidate UGTs were amplified from cDNA prepared from yC of the 10 June 2016 time point and cloned into the pJET1.2/blunt vector (ThermoFisher Scientific) for sequencing (Table S4). Complete open reading frames of candidate UGTs were cloned as BsaI or BbsI fragments into the pASK‐IBA37 vector (IBA‐GmbH, Göttingen, Germany). The E. coli TOP10 strain (Invitrogen) was used for heterologous UGT expression. Cultures were grown at 21°C, induced at an OD600 = 0.5 with 200 μg L−1 anhydrotetracycline (Sigma‐Aldrich) and then placed at 18°C and grown for another 20 h. Cells were collected by centrifugation and disrupted by five freeze and thaw cycles in chilled extraction buffer (50 mm Tris−HCl, pH 7.5, 10 mm MgCl2, 5 mm dithiothreitol, 2% (v/v) glycerol, 150 mm NaCl2, 20 mm imidazole, 1× Pierce™ protease inhibitor (EDTA‐free, ThermoFisher Scientific), 25 U benzonase nuclease (Merck, Germany), and 0.2 mg × ml−1 lysozyme). Cell fragments were removed by centrifugation at 14 000  for 20 min at 4°C and the supernatant was loaded onto a Ni‐NTA agarose column (Qiagen, Hilden, Germany). Protein was eluted with elution buffer (10 mm Tris−HCl, pH 7.5, 500 mm imidazole, 1 mm dithiothreitol, 10% (v/v) glycerol) and desalted into assay buffer (10 mm Tris−HCl, pH = 7.5, 1 mm dithiothreitol, 10% (v/v) glycerol) using Illustra NAP‐5 Columns (GE Healthcare). Protein concentrations were determined using UV absorption at 280 nm.

UGT assays with recombinant proteins

UGT assays with recombinant were performed with 1 μg of purified recombinant protein, 50 μm mini‐MbA and 1 mm UDP‐Glc in a Teflon‐sealed, screw‐capped 1‐ml GC glass vial. Unless stated otherwise, assays were performed in assay buffer in a final volume of 50 μl and incubated at 25°C. Assays were incubated for 2 h, and stopped by placing on ice after the addition of an equal volume of MeOH. CcUGT3 was tested for substrate specificity using 0.5 μg of protein, 1 mm UDP‐Glc and 50 μm of the different substrates listed in Table S1, and incubated for 2 h. For relative turnover rates for mini‐MbA and different UDP‐sugars, 0.5 μg purified CcUGT3 was assayed with 50 μm mini‐MbA and 1 mm UDP‐Glc or UDP‐Xyl or HPLC‐purified UDP‐Rha for 60 min. Product quantification was done based on an MbA standard curve.

Reverse transcription and quantitative real‐time PCR (qRT‐PCR)

RNA was extracted and cDNA synthesis done as previously described (Irmisch et al., 2018). For qRT‐PCR, the cDNA was diluted 1:5 with water. For the amplification of a UGT3 fragment of approximately 200‐bp length, a primer pair was designed with a Tm ≥ 60°C, a GC content of 45–60%, and a primer length of 20–25 nt (Table S4). Primer specificity was confirmed by agarose gel electrophoresis, melting curve analysis, standard curve analysis and by sequence verification of cloned PCR products. qRT‐PCR reactions were performed in duplicate on a Bio‐Rad CFX96™ instrument (Bio‐Rad Laboratory, Hercules, CA, USA) in optical 96‐well plates using SsoFast™ EvaGreen® Supermix (Bio‐Rad) with the following PCR conditions: initial incubation at 95°C for 30 sec followed by 40 cycles of amplification (95°C for 5 sec, 60°C for 10 sec). qRT‐PCR analyses were performed with three biological replicates for each of the six different time points of yC and oC collections (10 June, 27 June, 22 July, 16 August, 12 September, 6 October 2016) described in Irmisch et al. (2018)). Serin‐incorperator (MEP) and zinc‐finger protein (ZF) were used as reference genes (Irmisch et al., 2018).

Transient expression in N. benthamiana

For expression in N. benthamiana, the coding region of UGT3 was cloned into the pCAMBiA2300U vector. After sequence verification, the pCAMBiA vector carrying UGT3 as well as pCAMBiA vectors carrying the previously described genes for myricetin biosynthesis: CcFLS, CcCYP2 and CcMYB4 (Irmisch et al., 2019) and the genes for mini‐MbA biosynthesis CcUGT1, CcUGT2 and CcAT1 (Irmisch et al., 2018) and the pBIN:p19 were separately transferred into Agrobacterium tumefaciens strain C58pMP90. One ml of overnight cultures (220 rpm, 28°C) were used to inoculate 10 ml LB‐media containing 50 μg × ml−1 kanamycin, 25 μg × ml−1 rifampicin and 25 μg × ml−1 gentamicin for overnight growth. The following day the cultures were centrifuged (4000 , 5 min) and cells were re‐suspended in infiltration buffer (10 mm MES, pH 5.6, 10 mm MgCl2, 100 μm acetosyringone) to a final OD600 of 0.5. After shaking for 3 h at RT, the following combinations of transformed A. tumefaciens were prepared for leaf infiltration using: (i) A. tumefaciens 35S pro:(CcMYB4 + CcFLS + CcCYP2 + CcUGT1 + CcUGT2 + CcAT1) + A. tumefaciens pBIN:p19; (ii) A. tumefaciens 35S pro:(CcMYB4 + CcFLS + CcCYP2 + CcUGT1 + CcUGT2 + CcAT1 + CcUGT3) + A. tumefaciens pBIN:p19. Equal volumes of each line of transformed A. tumefaciens were used to prepare the mixtures. The leaves of 4‐week‐old N. benthamiana plants were infiltrated with A. tumefaciens solution using a 1‐ml needle‐free syringe to gently push the bacterial mixture into the abaxial surface. Infiltrated leaves were labelled with tape and harvested 5 days after infiltration and directly frozen in liquid nitrogen and stored at −80°C until further analysis. Plant material was ground in liquid nitrogen into a fine powder, and 100 mg was extracted with 1 ml 50% (v/v) MeOH for 2 h at RT. The extracts were analyzed using LC‐MS.

Alignment and phylogenetic tree construction

An amino acid alignment of CcUGT3, UGT703E2 and other plants UGTs (Table S5) was constructed using ClustalW algorithm implemented in MEGA6 (Tamura et al., 2011). Based on this alignment, a phylogenetic tree was reconstructed with MEGA6 using a neighbour‐joining algorithm (Poisson model). A bootstrap resampling analysis with 1000 replicates was performed to evaluate the tree topology.

Statistical analysis

To test for significant differences in CcUGT3 transcript abundance in yC and oC at different time points, data were log transformed to meet statistical requirements and a two‐way analysis of variance (anova) was performed followed by Tukey test using SigmaPlot 11.0 for Windows (Systat Software Inc. 2008) (Table S2).

Accession Numbers

Previously published oC, yC and corm_2013 transcriptome libraries are available in the NCBI/GenBank Sequence Read Archive (SRA) (SRP108844). UGT DE data were previously published (Irmisch et al., 2018). New UGT nucleotide sequences were deposited in GenBank with the accession numbers MN108491 (CcUGT3, UGT703E1) and MN108492 (UGT703E2). Accession numbers for all other genes/proteins used in this work are listed in Table S5.

Author Contributions

SI and JB conceived, designed and supervised the research. SI, SJ, LLM, carried out the experimental work. SI and MMSY analyzed data. SI and JB interpreted results and wrote the paper. All authors read, edited and approved the final version of the manuscript.

Funding Information

The research was supported with funds to JB from the GlycoNet Networks of Centres of Excellence (GlycoNet NCE) and the Natural Sciences and Engineering Research Council of Canada (NSERC, Discovery Grant). SI was supported by the Alexander‐von‐Humboldt Foundation through a Feodor Lynen Research Fellowship. JB is a Distinguished University Scholar.

Conflict of Interest

The authors declare that they have no conflict of interest in accordance with journal policy.

Figure S1. Separation and identification of MbA‐XR2 and MbA‐R2 using long‐run LC‐MS.

Click here for additional data file.

Figure S2. Products of assays with corm protein extract using mini‐MbA and UDP‐Xyl as substrates.

Click here for additional data file.

Figure S3. Sequential glycosylation activity of corm protein extracts with mini‐MbA.

Click here for additional data file.

Figure S4. Identification of a C‐terminal region corresponding to transcriptome contig DN67918_c7_g7_i4 or a closely related paralogous sequence.

Click here for additional data file.

Figure S5. Activity of purified recombinant UGT703E1 (CcUGT3) with mini‐MbA and different UDP‐sugars.

Click here for additional data file.

Figure S6. The formation of m/z 771 and m/z 787 by N. benthamiana transiently expressing montbretia genes.

Click here for additional data file.

Figure S7. MbA‐XR2 produced by N. benthamiana transiently co‐expressing UGT703E1 (CcUGT3).

Click here for additional data file.

Table S1. Activity assays of UGT703E1 (CcUGT3) with different acceptors.

Click here for additional data file.

Table S2. Statistical analysis of transcript expression data.

Click here for additional data file.

Table S3. Accurate mass data.

Click here for additional data file.

Table S4. Oligonucleotides.

Click here for additional data file.

Table S5. NCBI identification numbers of sequences used in the phylogenetic tree.

Click here for additional data file.