Identification of a Unique Type of Isoflavone O-Methyltransferase, GmIOMT1, Based on Multi-Omics Analysis of Soybean under Biotic Stress

Abstract

Isoflavonoids are commonly found in leguminous plants. Glycitein is one of the isoflavones produced by soybean. The genes encoding the enzymes in the isoflavone biosynthetic pathway have mostly been identified and characterized. However, the gene(s) for isoflavone O-methyltransferase (IOMT), which catalyzes the last step of glycitein biosynthesis, has not yet been identified. In this study, we conducted multi-omics analyses of fungal-inoculated soybean and indicated that glycitein biosynthesis was induced in response to biotic stress. Moreover, we identified a unique type of IOMT, which participates in glycitein biosynthesis. Soybean seedlings were inoculated with Aspergillus oryzae or Rhizopus oligosporus and sampled daily for 8 d. Multi-omics analyses were conducted using liquid chromatography–tandem mass spectrometry and RNA sequencing. Metabolome analysis revealed that glycitein derivatives increased following fungal inoculation. Transcriptome co-expression analysis identified two candidate IOMTs that were co-expressed with the gene encoding flavonoid 6-hydroxylase (F6H), the key enzyme in glycitein biosynthesis. The enzymatic assay of the two IOMTs using respective recombinant proteins showed that one IOMT, named as GmIOMT1, produced glycitein. Unlike other IOMTs, GmIOMT1 belongs to the cation-dependent OMT family and exhibited the highest activity with Zn²⁺ among cations tested. Moreover, we demonstrated that GmIOMT1 overexpression increased the levels of glycitein derivatives in soybean hairy roots when F6H was co-expressed. These results strongly suggest that GmIOMT1 participates in inducing glycitein biosynthesis in response to biotic stress.

Accession numbers: RNA-sequencing reads are available in the DDBJ Sequence Read Archive database with accession number (DRA010476).

Introduction

Flavonoids are an important class of plant specialized metabolites derived from phenylalanine. They have various biological functions, including antioxidant action and interaction with other organisms. Isoflavonoids are a group of flavonoids that are mostly produced by leguminous plants. In soybean, three types of isoflavone aglycones, such as genistein (5,7,4′-trihydroxyisoflavone), daidzein (7,4′-dihydroxyisoflavone) and glycitein (7,4′-dihydroxy-6-methoxyisoflavone) are produced via two different pathways (Fig.�1). Genistein is synthesized via the pathway shared with other flavonoids, in which chalcone synthase (CHS) and type I and type II chalcone isomerase (CHI) are involved. Daidzein and glycitein are produced via the specific pathway involving polyketide reductase, also called chalcone reductase, and CHS and type II CHI (Welle et�al. 1991, Akashi et�al. 1997, Mameda et�al. 2018). Soybean accumulates genistein, daidzein, glycitein and their glucosides and malonylglucosides (Fig.�1). Glycitein and its derivatives are accumulated exclusively in the hypocotyl, and the others are found in the whole body but most accumulated in the seed (Kudou et�al. 1991). To date, many functions of daidzein and genistein have been clarified for soybean. For example, daidzein and genistein are required for symbiosis with rhizobia (Subramanian et�al. 2006) and daidzein is a precursor of glyceollin, a major phytoalexin in soybean (Fig.�1). However, the physiological function of glycitein is almost unknown. The amount of glycitein and its derivatives is very low and varies among cultivars when compared to other isoflavones (Artigot et�al. 2013). Glycitein derivatives are accumulated only in the hypocotyl under normal conditions (Kudou et�al. 1991), and their accumulations are notably induced by biotic and abiotic elicitors (Landini et�al. 2003, Al-Tawaha et�al. 2006). These observations suggest that glycitein plays a role in defense reaction against pathogen infection.

Fig. 1.

Isoflavone biosynthesis pathway in soybean. Acetylated isoflavone glucosides are considered artifacts produced from malonylated isoflavone glucosides (Horowitz and Asen 1989). PAL, phenylalanine ammonia-lyase; C4H, cinnamate 4-hydroxylase; 4CL, 4-coumarate-CoA ligase; PKR or CHR, polyketide reductase or chalcone reductase, respectively; IF7GT, isoflavone 7-O-glucosyltransferase; IF7MaT, isoflavone 7-O-glucoside 6′′-O-malonyltransfrease; ICHG, isoflavone conjugates hydrolyzing beta-glucosidase.

Although genistein and daidzein are biosynthesized from naringenin and liquiritigenin, respectively, the reactions are catalyzed by both 2-hydroxyisoflavanone synthase (IFS) and 2-hydroxyisoflavanone dehydratase (HID) (Fig.�1). IFS belongs to the CYP93C subfamily of the cytochrome P450 family and is essential in isoflavonoid biosynthesis. Two IFS cDNAs, such as CYP93C1 (IFS1) and CYP93C5 (IFS2), have been identified in soybean (Steele et�al. 1999, Jung et�al. 2000). An HID cDNA has also been identified in soybean (Akashi et�al. 2005). Isoflavone biosynthesis from liquiritigenin diverges to enter the glycitein biosynthetic pathway in a reaction catalyzed by flavonoid 6-hydroxylase (F6H; CYP71D9). An F6H cDNA was first isolated as the elicitor-inducible gene (Schopfer and Ebel 1998). Later, an in vitro assay using a heterologously expressed protein in yeast revealed that F6H recognizes liquiritigenin as a substrate, but not daidzein (Latunde-Dada et�al. 2001). Following the production of 6-hydroxyliquiritigenin by the F6H, IFS and HID convert it to 6-hydroxydaidzein. Glycitein has been hypothesized to be produced by isoflavone O-methyltransferase (IOMT) but has not been experimentally confirmed so far.

The plant S-adenosyl-l-methionine (SAM)-dependent OMTs are grouped into two major types: cation-independent type (type 1) and cation-dependent type (type 2) (Noel et�al. 2003). Type 1 OMTs form a large family that comprised the OMTs involved in specialized (secondary) metabolism, such as phenylpropanoid and flavonoid biosynthesis. This family contains all the OMTs identified in the isoflavonoid biosynthesis, namely: (+)-6a-hydroxymaackiain 3-OMT (Wu et�al. 1997), daidzein 7-OMT (He et�al. 1998), 2-hydroxyisoflavanone 4′-OMT (Akashi et�al. 2003), and isoflavone 3′-OMT (Li et�al. 2016). Most of the type 2 OMTs identified in plants so far are the caffeoyl-CoA OMTs (CCoAOMTs), which are the key enzymes of lignin biosynthesis. Among the OMTs involved in flavonoid biosynthesis, only two kinds of OMTs belong to the type 2 family; they are the phenylpropanoid and flavonoid OMTs (PFOMTs) from ice plant (Mesembryanthemum crystallinum) (Ibdah et�al. 2003), rice (Oryza sativa) (Lee et�al. 2008) and sweet basil (Ocimum basilicum) (Berim and Gang, 2013) and anthocyanin OMTs (AOMTs) from grape (Vitis vinifera) (Hugueney et�al. 2009), soybean (Kovinich et�al. 2011), tomato (Gomez Roldan et�al. 2014), Paeonia (Du et�al. 2015) and Nemophila menziesii (Okitsu et�al. 2018).

In this study, we aimed to understand isoflavonoid biosynthesis under biotic stress and identify undiscovered enzymes/genes in the pathway. We analyzed transcriptomic and metabolomic changes in soybean seedlings inoculated with Aspergillus oryzae or Rhizopus oligosporus based on the finding that isoflavonoids were increased by the inoculation of fungi in soybean (Simons et�al. 2011a, 2011b, Aisyah et�al. 2013). We clarified that the biosynthesis of glycitein and its derivatives was induced by the inoculation of these fungi. In addition, we identified an IOMT catalyzing the last step of glycitein biosynthesis, which has a unique characteristic in terms of cation dependency. The findings in this study will contribute to the understanding of the physiological importance of glycitein.

Results

Changes in isoflavone contents and gene expression in fungal-inoculated soybean seedlings

We analyzed time-dependent changes in isoflavonoid contents in soybean seedling, inoculated with A. oryzae or R. oligosporus, for 8 d. Targeted metabolome analysis using liquid chromatography–tandem mass spectrometry (LC–QqQ-MS) was used to analyze genistein, daidzein, glycitein and their derivatives. Under normal conditions without inoculation, a developmental increase in daidzein was observed (Fig.�2A). When the seedlings were inoculated with fungi, particularly R. oligosporus, the metabolic flow was redirected to the biosynthesis of glyceollin I, glycitein and glycitin derivatives (Fig.�2A). Of these compounds, glyceollin I was detected only in fungal-inoculated soybean, and its amount increased in a time-dependent manner (Fig.�2B). In addition, glycitein accumulation was induced by fungal inoculation. Although two glycitin derivative levels increased following inoculation, glycitin level did not. No obvious trend was observed in other metabolic pathways (Supplementary Tables S1, S2).

Fig. 2.

Time-dependent changes in isoflavonoid contents in fungal-inoculated soybean seedlings. (A) Heatmap representation, (B) peak area value of glycitein derivatives and glyceollin analyzed by LC–QqQ-MS and the mean and SD of three bulked (approximately five individuals) replicates are shown. Inoculation was conducted after 2 d of soaking. n.d., not detected.

Transcriptome analysis was also performed via RNA sequencing (Supplementary Table S3). We focused on the expression patterns of the genes annotated to be involved in flavonoid biosynthesis (Fig.�3 and Supplementary Table S3). We found that several genes were induced by fungal inoculation. Among the phenylpropanoid and flavonoid biosynthetic genes, phenylalanine ammonia-lyase 2.3, CHS7 and CHS8, which are involved in isoflavone production (Dhaubhadel et�al. 2007), were induced. IFS1 and IFS2, which are essential for isoflavone biosynthesis, were also induced. Their expression peaked at 3 d after soaking (DAS) in A. oryzae-inoculated seedlings (Fig.�3). F6H was also induced by fungal inoculation, but its expression peaked at 8 DAS in R. oligosporus-inoculated seedlings (Fig.�3). The expression of F6H was suppressed to a certain low level under the non-inoculated condition (Supplementary Fig. S1). As F6H is the key enzyme at the branching point of the glycitein biosynthetic pathway (Fig.�1), the induction of F6H might be responsible for redirecting isoflavonoid biosynthesis in response to fungal inoculation.

Fig. 3.

Time-dependent changes in the expression of putative isoflavonoid biosynthetic genes. References of gene names are described in Supplementary Table S9.

Identification of candidate OMTs responsible for glycitein biosynthesis by transcriptome co-expression analysis

Among the enzymes responsible for soybean isoflavonoid biosynthesis, only IOMT, which is the last step of glycitein biosynthesis, was unidentified. We planned to isolate the candidates of IOMT genes based on the ‘guilt-by-association’ principle (Saito et�al. 2008, Gillis and Pavlidis 2012) using transcriptome data. This principle is based on the assumption that the genes involved in the same metabolic pathway are coordinately expressed when the product metabolite(s) is accumulated. To this end, we carried out a weighted gene co-expression network analysis (WGCNA) (Langfelder and Horvath 2008). After removing the transcripts with the sum of transcripts per million (TPM) <1 in all samples, the remaining 59,343 transcripts were subjected to WGCNA for clustering based on the expression pattern. The transcripts were classified into 382 classes (Supplementary Table S4). The F6H gene belonged to the class number 229. This class contained three transcripts annotated as methyltransferase (Glyma.05G147000.1, Glyma.05G147000.2 and Glyma.07G048900.1), and their coding sequences (CDSs) were predicted to encode 236, 185 and 372 amino acids, respectively. Since Glyma.05G147000.2 encoded the shorter amino acid sequence, lacking two exons found in Glyma.05G147000.1, we selected Glyma.05G147000.1 and Glyma.07G048900.1 as the IOMT candidates and tentatively named them OMT1 and OMT2, respectively. OMT1 and OMT2 showed a very similar expression pattern to that of F6H (Supplementary Fig. S1). We conducted a phylogenetic analysis with the known OMTs, which were involved in the flavonoid and isoflavonoid biosynthesis. The known OMTs were classified into type 1 (cation-independent) OMT clade and type 2 (cation-dependent) OMT clade. The previously identified isoflavonoid OMTs formed a subclade within the type 1 OMT clade (Fig.�4A). Unexpectedly, OMT1 belonged to the type 2 OMT clade, while OMT2 was a member of the type 1 OMT clade but did not belong to the isoflavonoid OMT subclade. In addition, OMT1 had a catalytic triad (–Lys–Asn–Asp–), which is essential for the catalysis of the methyl transfer in cation-dependent OMT (Fig.�4B) (Brandt et�al. 2015).

Fig. 4.

Comparison of GmIOMT1 with other OMTs. (A) Phylogenetic tree of representative type 1 (cation-independent) and type 2 (cation-dependent) OMTs. The tree was created using the neighbor-joining method with an amino acid sequence and 1,000 bootstrap replicates. Percentages next to enzyme names indicate the amino acid identity with GmIOMT1. (B) Multiple alignments of cation-dependent OMTs. *The residues of the catalytic triad. COMT, caffeic acid O-methyltransferase; D7OMT, daidzein 7-O-methyltransferase; F3′OMT, flavonoid 3′-O-methyltransferase; F3′5′OMT, flavonoid 3′,5′-O-methyltransferase; HI4′OMT, 2-hydroxyisoflavanone 4′-O-methyltransferase; HMM, (+)-6a-hydroxymaackiain 3-O-methyltransferase; IF3′OMT, isoflavone 3′-O-methyltransferase; NOMT, naringenin 7-O-methyltransferase.

Enzyme assay of candidate OMTs

The CDSs of OMT1 and OMT2 were cloned and expressed with Escherichia coli expression system. Recombinant proteins were extracted as the cell-free extract, and the enzyme assay was carried out using 6-hydroxydaidzein and SAM as substrates. Reaction products were analyzed using LC-ultraviolet (UV) detection. Interestingly, the peak coinciding with glycitein was detected only when using OMT1 in the reaction (Fig.�5). The reaction product was chromatographically purified and confirmed to be glycitein by Nuclear Magnetic Resonance (NMR) analysis (Supplementary Methods). This result indicated that OMT1 has the activity of converting 6-hydroxydaidzein to glycitein. Thus, we renamed OMT1 and referred to it henceforth as GmIOMT1.

Fig. 5.

LC–UV chromatogram of the reaction mixture of OMT1 assay. The assay was conducted using the cell-free extract from OMT1-expressing Escherichia coli with 6-hydroxydaidzein as a substrate. The cell-free extract from β-glucuronidase (uidA)-expressing E. coli was used as a negative control.

Biochemical properties of GmIOMT1

The pH dependence of GmIOMT1 was examined in the reaction mixture containing 20 mM MgCl₂. The optimum pH range was neutral (pH 7.0–7.6) (Fig.�6A) as it was with the other OMTs reported previously. Cation dependency was analyzed according to a previous study (Hugueney et�al. 2009). Generally, cation-dependent OMTs contain cation inside, and their enzymatic activities are decreased in the presence of EDTA by the chelating effect. GmIOMT1 activity was decreased in the presence of EDTA (Fig.�6B). GmIOMT1 activity was enhanced by all the divalent cations tested at 5 mM: Zn²⁺, Mg²⁺, Ca²⁺ and Mn²⁺ by 3-, 2-, 2-, and 1.5-fold, respectively; then, we checked Zn²⁺ concentration dependency. The GmIOMT1 activity was almost the same at the range of 2.5–10 mM Zn²⁺ (Fig.�6C). Finally, the kinetic parameters against 6-hydroxydaidzein was determined at 10 mM Zn²⁺ as follows: K_m = 2.1 � 0.5 μM, V_max = 411 � 20 pkat mg⁻¹, k_cat = 1.1 � 10⁻² s⁻¹ and k_cat/K_m = 5.2 s⁻¹mM⁻¹ (Fig.�6D).

Fig. 6.

Characteristics of GmIOMT1. (A) pH dependence, (B) cation dependence, (C) Zn²⁺-concentration, (D) Michaelis–Menten kinetics and (E) substrate specificity of GmIOMT1 and substrate structure. The mean � standard error (n = 3 technical replicates) are shown.

Since the known cation-dependent OMTs showed broad substrate specificity against the catechol structure-containing compounds, we analyzed the specific activity of GmIOMT1 against catechol structure-containing isoflavones: 6-hydroxydaidzein, 8-hydroxydaidzein and 3′-hydroxydaidzein. The result showed that GmIOMT1 displayed almost the same activity with 6-hydroxydaidzein and 8-hydroxydaidzein; however, the activity with 3′-hydroxydaidzein was 2-fold weaker (Fig.�6E).

Functional analysis of GmIOMT1 in hairy roots

We further analyzed the in vivo function of GmIOMT1 using soybean hairy roots. Since F6H is not expressed in hairy roots, 6-hydroxydaidzein was expected to be absent in the substrate of GmIOMT1; thus we expressed not only GmIOMT1 but also F6H. F6H-GmIOMT1- or GFP-overexpressed hairy roots were analyzed by real-time PCR and LC–QqQ-MS. Real-time PCR revealed that the expression level of F6H and GmIOMT1 in F6H-GmIOMT1-overexpressed lines was approximately 15,000–35,000- and 1,000–3,500-fold higher than that in GFP-overexpressed lines (Fig.�7A). Glycitein derivatives were detected in hairy roots by LC–QqQ-MS, and 6″-O-malonylglycitin and glycitin were significantly increased in the F6H-GmIOMT1 lines except for F6H-GmIOMT1-2 (Fig.�7B). In contrast, 6″-O-acetylglycitin and glycitein showed no significant difference. Despite the significant increase in the expression level, the increase in glycitein-related metabolites was only about twice that of the control.

Fig. 7.

Analysis of GmIOMT1 function in cultured hairy roots. (A) transcript levels of F6H and GmIOMT1 in hairy roots. The means � standard errors (n = 3, biological replicates) are shown. Transcript levels were analyzed using the ΔΔCt method, and skip16 was used as the internal standard. Transcript levels were normalized against the GFP-1 value, which was set at 1. (B) Results of LC–QqQ-MS analysis of glycitein derivatives in hairy roots. The means � SD (n = 3, biological replicates) are shown. Different letters indicate significant differences according to Tukey–Kramer test (P < 0.05).

Discussion

Usefulness of integrated omics analysis for the identification of novel genes in soybean

Although glycitein was first isolated from soybean in 1973 (Naim et�al. 1973), it was not until 2001 that the involvement of F6H in its biosynthesis was identified (Latunde-Dada et�al. 2001). IOMT catalyzing the last step in glycitein biosynthesis had not been revealed for >40 years. The characteristics of the soybean genome make it difficult to analyze genes: the genome size is relatively large (approximately 1.1 Gbp); half of the genome has highly repetitive sequences; and >90% of the genes, which do not have repetitive sequences, have over two copies (Schmutz et�al. 2010). Identification of the IOMT gene by biochemical assay screening or functional prediction from genome sequences was particularly difficult because soybean has >500 genes annotated as methyltransferase in Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html). In this study, we could efficiently narrow down the candidate genes to two from >500 putative methyltransferase genes based on the co-occurrence analysis of metabolome and transcriptome using F6H, a key gene in glycitein biosynthesis, as the query. The omics data obtained in this study can be further used for the isolation of other unknown genes such as the transcription factor involved in glycitein and glyceollin biosynthesis and the elucidation of the mechanism of biotic stress response in soybean.

Characteristics of GmIOMT1

Although soybean isoflavonoid OMT is expected to be a member of leguminous plant-specific IOMTs, GmIOMT1 belongs to the cation-dependent OMT (type 2 OMT) clade (Fig.�5). The internal cation of cation-dependent OMTs is considered necessary for binding with the hydroxyl group of the substrate (Ibdah et�al. 2003, Hugueney et�al. 2009). Cation type affects substrate specificity and kinetic parameters (Lukacin et�al. 2004). The assay using the purified GmIOMT1 indicated that the optimum pH was similar to that for other cation-dependent OMTs, and the activity decreased in the absence of cations and presence of EDTA (Fig.�6A, B). Although the activity of previously reported cation-dependent OMTs, such as AOMT and PFOMT, was inhibited by Zn²⁺ (Ibdah et�al. 2003, Hugueney et�al. 2009), GmIOMT1 activity was rather increased by Zn²⁺ to a greater extent than Mg²⁺ (Fig.�6B, C). The mechanism, by which the enzyme activity changes, depending on the type of cation, is still unknown. Crystal structure analysis of GmIOMT1, which has different cation requirements, would help elucidate the underlying mechanism in future studies.

GmIOMT1 showed activity not only toward 6-hydroxydaidzein, which is the direct precursor of glycitein, but also to other isoflavones possessing catechol structure (Fig.�6E). Although 3′-hydroxydaidzein and 8-hydroxydaidzein were isolated from fungal-fermented soybean, they have not been detected in non-fermented soybean (Chang 2014). Since fungi have cytochrome P450, which catalyzes the hydroxylation of isoflavone, 3′-hydroxydaidzein and 8-hydroxydaidzein cannot be produced by soybean but by fungi; thus, the reaction toward these isoflavones is not expected to occur in soybean. Such a wide range of substrate specificities would be useful in metabolic engineering for material production.

The expression levels of F6H and GmIOMT1 in their overexpressing hairy root lines were much higher than those in GFP control (Fig.�7A), and 6″-O-malonylglycitin and glycitin were increased in lines numbers 1 and 3 (Fig.�7B). However, despite the significant increase in the expression level, the increase in glycitein-related metabolites was only about twice that of the control. This may be because the translational efficiency of those transcripts was very low and/or that there was an unknown mechanism for adjusting the amount of glycitein, such as feedback regulation.

Glycitein biosynthesis probably involves different enzymes depending on the situation

In this study, we identified GmIOMT1 as the gene induced by fungal inoculation along with increased glycitein biosynthesis using metabolome and co-expression analysis. Glycitein is induced by conditions, such as fungal infection, and localizes specifically in the hypocotyl of soybean (Kudou et�al. 1991, Al-Tawaha et�al. 2006). In our study, glycitein was detected via highly sensitive LC–QqQ-MS (Fig.�2B), although GmIOMT1 and F6H were hardly expressed in non-fermented soybean (Supplementary Fig. S1). This result suggests that glycitein biosynthesis occurs via two routes, inducible and constitutive routes, which consist of different sets of enzymes with the same activities. In fact, previous research suggests a possibility that F6H paralogs, CYP71D101 and CYP71D102, participate in constitutive glycitein biosynthesis (Artigot et�al. 2013). In addition, CYP71D101 and CYP71D102 were not or were hardly expressed in all samples of soybean seedlings regardless of fungal inoculation (Supplementary Fig. S1). A recent study reported a quantitative trait locus for 6″-O-malonylglycitin accumulation (Watanabe et�al. 2019) but did not determine its match with the positions of GmIOMT1, F6H and their paralogs. Thus, the genes participating in glycitein biosynthesis might be adequately used depending on organs, growth stage, stress response and so on. Identification and characterization of other enzymes and transcription factors involved in glycitein biosynthesis will help understand the biological functions of glycitein and its derivatives, which should be different from those of other isoflavonoids, such as genistein, daidzein and their derivatives.

Materials and Methods

Chemicals

6-Hydroxydaidzein and 3′-hydroxydaidzein were purchased from INDOFINE Chemical Company, Inc. (Hillsborough, NJ, USA). 8-Hydroxydaidzein and glycitein were purchased from Nagara Science (Gifu, Japan) and Sigma-Aldrich (St. Louis, MO, USA), respectively. Glyceollin I was purchased from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan). SAM solution (32 mM) was purchased from New England Biolabs (Ipswich, MA, USA).

Plant materials

A soybean cultivar Otofuke-Osodefuri, which is known for its high isoflavonoid accumulation (Tanifuji et�al. 2009), was used. Aspergillus oryzae strain AOK139 and R. oligosporus were purchased from AKITA KONNO CO., LTD (Akita, Japan). Soybean seeds were germinated in a chamber at 26�C and sprinkled with water (26�C) for 12 min every 4 h. After 48 h when germination started, sprinkling was stopped; germinated soybeans were inoculated with 1% (w/w) of R. oligosporus or A. oryzae spore and then incubated under a humidity of 100%. Soybeans seedlings were sampled every 24 h for 8 d from the start of germination and then frozen at −80�C.

Sample preparation for widely targeted metabolomics

Frozen samples were powdered using dry ice powder via a mill cutter (Tube Mill control and MT 40; IKA, Staufen, Germany) and were then lyophilized by a freeze-dryer (dry chamber, DRC-1000; freeze-drying instrument, FDU-2100; EYELA, Tokyo, Japan). Four milligrams of powdered samples were weighed (AP324W, Shimadzu Corporation, Kyoto, Japan) and placed into a 2-ml tube with 5-mm zirconia beads. One milliliter of extraction solvent with 0.1% (v/v) formic acid in 80% (v/v) methanol and internal standards (8.4 nM of lidocaine and 210 nM of 10-camphorsulfonic acid) was added into the tube, and the metabolites were extracted using a bead-shocker (Shake Master NEO, Biomedical Science, Tokyo, Japan) for 2 min at 1,000 rpm, followed by centrifugation at 9,100 � g for 1 min. The extracted solutions were evaporated using a liquid handling system (MicrolabSTARplus, Hamilton Company, Reno, NV, USA), and the residual extract was redissolved with LC–MS grade water (FUJIFILM Wako Pure Chemical Corporation) so that the final diluted solution would be 40-fold. The redissolved solution was filtered (MZHVN0W50; Merck Millipore, Darmstadt, Germany). One microliter of the solution including 100 ng of the sample was subjected to widely targeted metabolomics.

Widely targeted metabolomics

We modified the method of widely targeted metabolomics (Sawada et al., 2009) as follows. The selective reaction monitoring (SRM) conditions of 490 standard metabolites were optimized by flow injection analysis, and the retention time (RT) was determined using LC–QqQ-MS (Nexera MP/LCMS-8050; Shimadzu Corporation) (Supplementary Table S5, SRM and RT; Supplementary Table S6, LC conditions; Supplementary Table S7, QqQ-MS conditions). The raw data of peak area values of 13 target isoflavonoids and 477 metabolites were collected using the LabSolution software (Shimadzu Corporation). The raw data (lcd file) were converted into Abf file using Reifycs Abf (Analysis base file) Converter (https://www.reifycs.com/AbfConverter/). The peak area values of LC–QqQ-MS data were calculated using MRMPROBS (Tsugawa et�al. 2013, http://prime.psc.riken.jp/compms/mrmprobs/main.html). The data matrix (72 samples � 490 metabolites) is provided in Supplementary Tables S1, S2 and S8. The data matrix of metabolome data was analyzed using R script with pheatmap package. Z-score was calculated using the average peak area of biological replicates.

Total RNA isolation and cDNA synthesis

Total RNAs were isolated from the powdered frozen samples using the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). cDNAs for cloning and real-time PCR were synthesized using the SuperScript IV FirstStrand Synthesis System (Thermo Fisher Scientific, Waltham, MA, USA).

Transcriptomics

RNA samples for RNA sequencing were prepared by mixing the RNAs of three biological replicates. The process from library preparation to RNA sequencing was performed by TAKARA BIO INC. (Shiga, Japan). RNA sequencing was carried out using HiSeq 2500 (Illumina, San Diego, CA, USA) under the conditions of read length (100 bp) and paired-end. The obtained reads were cleaned by FaQCs (Lo and Chain 2014). Quantification of transcripts was carried out by Kallisto (Bray et�al. 2016) using the soybean transcript sequence (Williams 82 Assembly 2 Annotation 1). Data of 88,647 transcripts were obtained. For each transcript, the sum of TPM values in all 24 samples (three treatments � eight time points) was calculated. After removing the transcripts with the sum of <1, the remaining 59,343 transcripts were subjected to WGCNA using the R package (Langfelder and Horvath 2008).

Construction of expression vector for plant

The construction methods and simplified schematic of plant expression vectors are described in the Supplementary Methods and Supplementary Fig. S2.

Cloning

All CDSs were amplified by PCR using PrimeSTAR MAX DNA Polymerase (TAKARA) with primer sets and cDNA (Aspergillus-fermented soybean at DAS 5) as the template. Primer sequences were as follows: GmIOMT1 (Glyma.05G147000.1): 5′-CACCATGTCGGGTGATTTAGCATACAAG-3′ and 5′-TCACAGACGTCTACACAGGG-3′; OMT2 (Glyma.07G048900.1): 5′-ATGGCTCCATCATTGGAAACC-3′ and 5′-TTACTTATAAAATTCCATAACCCAG-3′; and F6H (Glyma.18G080400.1): 5′-ATGGATCTTCAACTTCTCTACTTC-3′ and 5′-CTAATTATGAACAGTTTTGGGAATG-3′. Amplified fragments were introduced into pENTR-D-TOPO vector (Thermo Fisher Scientific) for GmIOMT1 or pCR8/GW/TOPO vector (Thermo Fisher Scientific) for OMT2 and F6H. CDSs of OMT were introduced into the pET-53-DEST vector (Merck Millipore) using a Gateway LR Clonase II Enzyme Mix (Thermo Fisher Scientific). For GmIOMT1 and F6H expression in hairy root, the genes were introduced into pASG-GW and pAKG-GW vectors (Supplementary Method) using a Gateway LR Clonase II Enzyme Mix.

Crude enzyme assay of OMT candidates

For the assay using crude protein extract, E. coli strain C41 (DE3) (Lucigen, Middleton, WI, USA) was transformed with the pET-53-DEST vector (Merck Millipore), and protein induction was conducted at an OD₆₀₀ adjusted to approximately 0.4. The culture was supplemented with 0.5 mM IPTG and cultured for 3 h at 37�C. Escherichia coli cells were collected by centrifugation, resuspended in 100 mM sodium phosphate buffer (pH 7.6), disrupted by vortex with glass beads for 5 min and centrifuged at 13,000 � g at 4�C for 5 min. An aliquot of 200 μl of supernatant was incubated with 10 μg 6-hydroxydaidzein, dissolved in 4 μl of 2-methoxyethanol and 0.32 mM SAM at 30�C for 1 h. Reaction products were extracted using ethyl acetate and analyzed using LC–UV (Nexera X2; Shimadzu). LC–UV conditions were as follows: Column, Kinetex 2.6 μm XB-C18 50 � 2.1 mm (Shimadzu); detector, UV/VIS detector (SPD-20A) at 260 nm; solvent, A: water (0.1% formic acid), B: acetonitrile (0.1% formic acid); liner gradient, B: 10–70% (5 min); flow rate, 0.36 ml min⁻¹; and column temperature, 40�C.

Enzyme purification

Escherichia coli strain KRX (Promega, Madison, WI, USA) was transformed with GmIOMT1 (pET-53-DEST) and precultured in 20 ml of liquid LB medium (50 mg l⁻¹ carbenicillin) at 37�C overnight. Two milliliters of pre-culture and final concentration of 0.15% glucose, 0.1% rhamnose and 0.5 mM IPTG were added into a 500-ml Erlenmeyer flask containing 200 ml of the same medium and cultured at 16�C overnight. The cell pellet was collected by centrifuge, resuspended in lysis buffer (300 mM KCl, 50 mM KH₂PO₄, 5 mM imidazole, pH 8.0) and then disrupted by sonication. The supernatant collected by centrifuge and His-tagged protein was purified using Profinia protein purification system (Bio-Rad, M�nchen, Germany) with Profinia IMAC Purification Kit (Bio-Rad) and cOmplete His-Tag Purification Column (Roche Diagnostics, Rotkreuz, Switzerland).

Characterization of enzymatic properties

The enzymatic assay was carried out using 900 ng of purified His-tagged GmIOMT1 protein, 320 μM SAM and 100 mM sodium phosphate buffer of pH 7.6 with divalent cations or EDTA in a total of 200 μl for 30 min at 30�C. The optimum pH of GmIOMT1 was determined with 100 μM 6-hydroxydaidzein and 20 mM MgCl₂ in potassium phosphate buffers (FUJIFILM Wako Pure Chemical Corporation) at pH 6.0, 6.4, 7.0, 7.2, 7.6 and 8.0. To determine the cation requirement of GmIOMT1, the enzyme assay was performed in the presence of 5 mM MgCl₂, CaCl₂, MnCl₂ or ZnCl₂ or 50 mM EDTA. According to a previous study (Hugueney et�al. 2009), 5 mM MgCl₂ was also added to test the effect of Zn²⁺, Ca²⁺ and Mn²⁺. To determine the relative activity of GmIOMT1, 100 μM 6-hydroxydaidzein, 8-hydroxydaidzein or 3′-hydroxydaidzein was incubated in the presence of 10 mM ZnCl₂ with recombinant enzyme for 10 min. To determine the kinetic parameters of GmIOMT1, 0.5, 5, 10, 25 and 50 μM 6-hydroxydaidzeins were used. K_m, V_max and k_cat values were calculated using nonlinear regression in SigmaPlot14 (Systat Software, Inc., San Jose, CA, USA).

Phylogenetic tree analysis

Amino acid sequence of OMTs was aligned using ClustalW v2.1 in DDBJ (https://clustalw.ddbj.nig.ac.jp/) with default settings and phylogenetic tree was drawn using MEGA7 (Kumar et�al. 2016) using default setting of neighbor-joining method with amino acid sequence and 1,000 bootstrap replicates. Accession numbers were as follows: HI4′OMT (G. echinata), BAC58011.1; HMM (Pisum sativum), AAC49856.1; D7OMT (G. echinata), BAC58012.1; F3′5′OMT (Chrysosplenium americanum), AAA80579.1; F′3OMT (Arabidopsis thaliana), AAB96879.1; CCoAOMT (Medicago sativa), AAC28973.1; HI4′OMT (G. max), XP_003542715.1; NOMT (O. sativa), XM_015763455.1; IOMT6 (Medicago truncatula), XP_013455257.1; IOMT2 (M. truncatula), ABD83942.1; COMT (M. truncatula), XP_003626614.1; GmIOMT1 (G. max), NP_001237478.1 (Glyma.05G147000); PFOMT (Mesembryanthemum crystallinum), AAN61072.1; IF3′OMT (Pueraria lobata), KP057887.1; OMT2 (G. max), XM_003528712.3 (Glyma.07G048900); AOMT (G. max), NP_001242455.1; and Catechol OMT (Homo sapiens), XP_011528188.1.

Production of transformed hairy root

Soybean (cv. Enrei) seeds were sterilized using the chlorine gas method (Paz et�al. 2006). Sterilized seeds were putted on a Gamborg B5 medium (B5 vitamin, 20 g l⁻¹ sucrose and 7 g l⁻¹ agar) and then cultured at 26�C under the 16-h light and 8-h dark condition for 1 week. Agrobacterium rhizogenes strain K599 was transformed by overexpressing GFP or F6H-GmIOMT1 on LB medium (15 g l⁻¹ agar and 100 mg l⁻¹ spectinomycin); the strain was then cultured in liquid medium at 28�C overnight. The cells were collected by centrifuge and resuspended in the infection buffer [10 mM MES, 10 mM MgCl₂, 250 mg l⁻¹ acetosyringone, and 0.02% (v/v) silwet l-77, pH 5.7] at the OD₆₀₀ of 0.1–0.2. The cotyledons of soybean seedlings were divided, leaving 5 mm hypocotyl, and putted on MS medium (MS vitamin, 30 g l⁻¹ sucrose and 9 g l⁻¹ agar). Two microliters of Agrobacterium-suspension buffer was placed at the cut ends of the cotyledon, followed by culturing under the condition similar to those for seed germination. Three weeks after infection, the hairy root from which GFP fluorescence was confirmed was transferred to Gamborg B5 medium (B5 vitamin, 30 g l⁻¹ sucrose, 2 g l⁻¹ gellite and 250 mg l⁻¹ carbenicillin) and cultured at 25�C under dark condition. After 4 weeks, two or three hairy root tips (approximately 1 cm) were transferred to a 125-ml polycarbonate Erlenmeyer flask with a vent cap (Corning Inc., Corning, NY, USA) containing 50 ml of Gamborg B5 liquid medium and then cultured under gyratory conditions (120 rpm) at 25�C for 2 weeks. Three root tips were transferred into three flasks containing the same medium and gyratory cultured under the same conditions.

Analysis of F6H and GmIOMT1 expression by real-time PCR

Primers for real-time PCR were designed using the Primer 3 program. Real-time PCR was conducted using the StepOnePlus system (Thermo Fisher Scientific) with PowerUp SYBR Master Mix (Thermo Fisher Scientific), cDNAs and 300 nM primers. Relative expression was analyzed using the ΔΔCt method and the housekeeping gene SKIP16 (GenBank accession number NM_001255441) as an internal standard. Primer sequences were as follows: GmIOMT1, 5′-AAAAGGCAGGAATGGAGCAC-3′ and 5′-TCTGCTTTATCCGCATCCAC-3′; F6H, 5′-ACCGAAGCCGCAAAGATTTC-3′ and 5′-TTCAAGCCAGACATGTGCTG-3′; and Skip16, 5′-AGATAGGGAAATGGTGCAGGT-3′ and 5′-CTAATGGCAATTGCAGCTCTC-3′.

Supplementary Data

Supplementary data are available at PCP online.

Funding

Japan Science and Technology Agency-Core Research for Evolutional Science and Technology (JST-CREST) (https://www.jst.go.jp/kisoken/crest/en/index.html) (JPMJCR16O2) and Japan Society for the Promotion of Science KAKENHI Grant-in-Aid for Early-Career Scientists (JP19K15821 to K.U.).