The Basic Helix–Loop–Helix Transcription Factor GubHLH3 Positively Regulates Soyasaponin Biosynthetic Genes in Glycyrrhiza uralensis

Abstract

Glycyrrhiza uralensis (licorice) is a widely used medicinal plant belonging to the Fabaceae. Its main active component, glycyrrhizin, is an oleanane-type triterpenoid saponin widely used as a medicine and as a natural sweetener. Licorice also produces other triterpenoids, including soyasaponins. Recent studies have revealed various oxidosqualene cyclases and cytochrome P450 monooxygenases (P450s) required for the biosynthesis of triterpenoids in licorice. Of these enzymes, β-amyrin synthase (bAS) and β-amyrin C-24 hydroxylase (CYP93E3) are involved in the biosynthesis of soyasapogenol B (an aglycone of soyasaponins) from 2,3-oxidosqualene. Although these biosynthetic enzyme genes are known to be temporally and spatially expressed in licorice, the regulatory mechanisms underlying their expression remain unknown. Here, we identified a basic helix–loop–helix (bHLH) transcription factor, GubHLH3, that positively regulates the expression of soyasaponin biosynthetic genes. GubHLH3 preferentially activates transcription from promoters of CYP93E3 and CYP72A566, the second P450 gene newly identified and shown to be responsible for C-22β hydroxylation in soyasapogenol B biosynthesis, in transient co-transfection assays of promoter–reporter constructs and transcription factors. Overexpression of GubHLH3 in transgenic hairy roots of G. uralensis enhanced the expression levels of bAS, CYP93E3 and CYP72A566. Moreover, soyasapogenol B and sophoradiol (22β-hydroxy-β-amyrin), an intermediate between β-amyrin and soyasapogenol B, were increased in transgenic hairy root lines overexpressing GubHLH3. We found that soyasaponin biosynthetic genes and GubHLH3 were co-ordinately up-regulated by methyl jasmonate (MeJA). These results suggest that GubHLH3 regulates MeJA-responsive expression of soyasaponin biosynthetic genes in G. uralensis. The regulatory mechanisms of triterpenoid biosynthesis in legumes are compared and discussed.

Introduction

Plants produce a huge variety of natural compounds called specialized metabolites, which are biosynthesized from primary metabolites (Weng et al. 2012). Among these specialized metabolites, triterpenoids are a highly diverse group, and some of these compounds have pharmacological effects (Sawai and Saito 2011). Glycyrrhiza uralensis (licorice) is one of the major medicinal plants belonging to the Fabaceae. An oleanane-type triterpenoid saponin, glycyrrhizin, is a characteristic compound accumulated in the roots and stolons of several Glycyrrhiza species, including G. uralensis, G. glabra and G. inflata (Hayashi and Sudo 2009). Due to its pharmaceutical properties and sweet taste, glycyrrhizin or the roots and stolons of licorice plants are widely used as medicines and natural sweeteners (Hayashi and Sudo 2009). In addition to glycyrrhizin, licorice also produces structurally different triterpenoids, such as soyasaponins, betulinic acid and oleanolic acid (Hayashi et al. 1988, Hayashi et al. 1990, Hayashi et al. 1993, Kojoma et al. 2010).

Recent studies have revealed many biosynthetic enzymes involved in the production of plant triterpenoids. The first committed step for triterpenoid biosynthesis is the cyclization of 2,3-oxidosqualene by oxidosqualene cyclases (OSCs), forming a branch point with the sterol biosynthetic pathway. Based on the cyclization mechanisms, OSCs produce either sterols or various triterpene scaffolds from the last common precursor, 2,3-oxidosqualene (Thimmappa et al. 2014). Although sterols and brassinosteroids, which are biosynthesized from cycloartenol, are mainly classified as primary metabolites, some specialized metabolites such as steroidal glycoalkaloids in the Solanaceae family or cucurbitacins in the Cucurbitaceae family are also derived from sterol precursors (Thimmappa et al. 2014). A huge variety of triterpenoids, which can be classified as specialized metabolites, are derived from various triterpene scaffolds such as β-amyrin, α-amyrin and lupeol. These triterpene scaffolds are subjected to site-specific modification reactions such as oxidation, glycosylation and acetylation to produce structurally diverse triterpenoids (Thimmappa et al. 2014). There are many reports on cytochrome P450 monooxygenases (P450s) and UDP-dependent glycosyltransferases (UGTs) involved in the oxidation and glycosylation of triterpene scaffolds (Thimmappa et al. 2014, Seki et al. 2015).

Most of the enzymes involved in the production of major triterpenoids in licorice have been characterized (Fig. 1). Glycyrrhizin, soyasaponins and oleanolic acid are derived from β-amyrin, which is biosynthesized from 2,3-oxidosqualene by β-amyrin synthase (bAS) (Hayashi et al. 2001). In glycyrrhizin biosynthesis, CYP88D6 and CYP72A154 catalyze oxidation reactions at the C-11 and C-30 positions of β-amyrin, respectively, to produce glycyrrhetinic acid, an aglycone of glycyrrhizin (Seki et al. 2008, Seki et al. 2011). Recently, a single UGT (GuUGAT) was reported to transfer two glucuronic acid moieties to the C-3 position of glycyrrhetinic acid to produce glycyrrhizin (Xu et al. 2016). In soyasaponin biosynthesis, CYP93E3 has been characterized as a β-amyrin C-24 hydroxylase (Seki et al. 2008); however, β-amyrin C-22β hydroxylase, required for the biosynthesis of soyasapogenol B, a common aglycone of soyasaponin I and soyasaponin II, the two soyasaponins previously identified in cultured licorice cells (Hayashi et al. 1990), remains unknown in licorice. Soyasapogenol B is expected to be further glycosylated by UGTs to produce soyasaponins; however, no UGTs involved in soyasaponin biosynthesis have been reported in licorice. In oleanolic acid biosynthesis, CYP716A179 was shown to catalyze oxidation reactions at the C-28 position of β-amyrin to produce oleanolic acid (Tamura et al. 2017a). Betulinic acid is biosynthesized from lupeol, produced by lupeol synthase (LUS) (Hayashi et al. 2004). CYP716A179 has also been shown to oxidize the C-28 position of lupeol to produce betulinic acid (Tamura et al. 2017a). In addition to the enzymes for triterpenoid biosynthesis mentioned above, cycloartenol synthase (CAS) was also identified in licorice (Hayashi et al. 2000).

Fig. 1

Proposed biosynthetic pathways of triterpenoids and sterols in licorice. A P450 enzyme functionally characterized in this study is indicated in red. Arrows indicate a cyclization, single oxidation or single glycosylation step. Numbers on β-amyrin and lupeol indicate carbon positions. Abbreviations: Ara, arabinose; bAS, β-amyrin synthase; CAS, cycloartenol synthase; Gal, galactose; GlcA, glucuronic acid; LUS, lupeol synthase; Rha, rhamnose.

It is known that the accumulation patterns of glycyrrhizin, soyasaponins, oleanolic acid and betulinic acid in licorice plants or tissue cultures differ (Hayashi et al. 1988, Hayashi et al. 1993, Kojoma et al. 2010, Tamura et al. 2017a). Generally, biosynthesis of plant specialized metabolites is temporally and spatially regulated. This can often be achieved by regulating biosynthetic genes at the transcriptional level. Key elements of transcriptional regulation are transcription factors (TFs) that bind to promoter regions of target genes or form complexes with other DNA-binding proteins to modulate gene expression (Yang et al. 2012). The identification of TFs that regulate specific biosynthetic pathways of specialized metabolites is useful not only for understanding the regulatory mechanisms of specialized plant metabolic pathways but also as a tool for manipulating multiple biosynthetic genes for biotechnological applications (Grotewold 2008). There are several reports on TFs involved in specialized metabolites derived from sterol precursors, such as basic helix–loop–helix (bHLH) TFs regulating cucurbitacin biosynthesis in the cucumber (Shang et al. 2014), ethylene response factor (ERF) TFs regulating steroidal glycoalkaloid biosynthesis in the potato and the tomato (Cárdenas et al. 2016, Thagun et al. 2016), and a WRKY TF regulating the biosynthesis of withanolides in Withania somnifera, which are thought to be derived from primary sterols (Singh et al. 2017). However, reports on TFs regulating triterpenoid biosynthesis are very limited. Recently, two bHLH TFs, TSAR1 and TSAR2 from the model legume Medicago truncatula, were shown to regulate the biosynthesis of non-hemolytic triterpenoid saponins (soyasaponins; possess a hydroxyl group at the C-24 position of β-amyrin) and hemolytic triterpenoid saponins (possess a carboxyl group at the C-28 position of β-amyrin), respectively (Mertens et al. 2016). In licorice, biosynthetic genes for major triterpenoids have been elucidated, and these genes are differentially expressed (Seki et al. 2008, Seki et al. 2011, Ramilowski et al. 2013, Tamura et al. 2017a). However, no TFs involved in the regulation of triterpenoid biosynthesis have yet been reported in licorice.

In this study, we identified a bHLH TF, GubHLH3, that positively regulates the expression of soyasaponin biosynthetic genes in G. uralensis. Overexpression of GubHLH3 in transgenic hairy roots of G. uralensis up-regulated the gene expression of β-amyrin synthase (bAS) and CYP93E3. We also newly identified that CYP72A566 functions as a β-amyrin C-22β hydroxylase in G. uralensis, as the second P450 necessary for the biosynthesis of soyasapogenol B from β-amyrin, by heterologous expression in engineered yeast. Gene expression of CYP72A566 was also up-regulated by overexpression of GubHLH3 in hairy roots, suggesting that GubHLH3 co-ordinately regulates soyasaponin biosynthetic genes. Moreover, accumulation of soyasapogenol B and its biosynthetic intermediate, sophoradiol (22β-hydroxy-β-amyrin), was enhanced in transgenic hairy roots overexpressing GubHLH3. Our results contribute to the elucidation of the molecular mechanisms underlying triterpenoid biosynthetic pathways in licorice, as well as the shared and distinct evolution of triterpenoid biosynthesis in legumes.

Results

Mining of candidate TFs

We performed two different approaches using model plant species to identify candidate TFs regulating glycyrrhizin or soyasaponin biosynthetic pathways in G. uralensis. The first approach used the publicly available gene co-expression analysis tool for M. truncatula (https://mtgea.noble.org/v3/) (Benedito et al. 2008, He et al. 2009) to find TFs co-expressed with soyasaponin biosynthetic genes. Although M. truncatula does not produce glycyrrhizin, soyasaponin I (soyasapogenol B glycoside) has been identified in this plant (Huhman et al. 2005, Carelli et al. 2011). In addition to bAS, CYP93E2 and CYP72A61v2 have been characterized as P450s for the production of soyasapogenol B from β-amyrin (Fukushima et al. 2013). In co-expression analysis with a correlation threshold of 0.6, no TFs were identified when bAS (Probeset ID: Mtr.32384.1.S1_s_at) was used as a query. However, when CYP93E2 (Mtr.8618.1.S1_at) or CYP72A61v2 (Mtr.43117.1.S1_at) were used as queries, a sequence annotated as a bHLH TF (Mtr.1885.1.S1_at) was identified as a unique TF by both searches. According to the database, Mtr.1885.1.S1_at corresponds to Medtr0246s0020.1 in the Medicago v4.0 annotation.

The second approach was comprehensive screening of an Arabidopsis TF library through high-throughput yeast one-hybrid (Y1H) assays (Mitsuda et al. 2010) against the promoter sequences of bAS, CYP88D6 and CYP93E3 (Supplementary Table S1). Promoter sequences of approximately 2 kb for bAS, CYP88D6 and CYP93E3 isolated from G. uralensis (strain 308-19) were divided into approximately 500 bp segments (Supplementary Fig. S1), and used as bait. Of 1,428 Arabidopsis TFs, 32 TFs showed positive interactions with at least one fragment of the bAS, CYP88D6 or CYP93E3 promoters (Supplementary Table S2). Since a bHLH TF was identified in the first approach, we focused on the three bHLH TFs: AtMYC2 (At1g32640), AtMYC4 (At4g17880) and AtbHLH018 (At2g22750) (Supplementary Table S2). Among the 1,428 Arabidopsis TFs included in the Y1H assays, AtbHLH018 had the highest alignment score with Medtr0246s0020.1 (46% amino acid sequence identity), the M. truncatula TF identified in the first approach (Supplementary Table S3).

Putative orthologs of Medtr0246s0020.1 and the three Arabidopsis bHLH TFs in G. uralensis were searched using blast (Altschul et al. 1997) in the transcriptome database of G. uralensis (http://ngs-data-archive.psc.riken.jp/Gur/blast.pl) (Ramilowski et al. 2013) (Supplementary Table S4). Among the G. uralensis unigenes in the database, Unigene29516 had the highest alignment score with Medtr0246s0020.1 (65% amino acid sequence identity), and a relatively high alignment score with AtbHLH018 (42% amino acid sequence identity). Unigene 21689 had the highest alignment score with AtMYC2 and AtMYC4 (56% and 53% amino acid sequence identities, respectively), and Unigene723 also had a comparably high alignment score with AtMYC2 and AtMYC4 (54% and 51% amino acid sequence identities, respectively). We cloned these three unigenes from the cDNA library obtained from roots of G. uralensis as candidate TFs, and named them GubHLH1 (Unigene21689), GubHLH2 (Unigene723) and GubHLH3 (Unigene29516). The alignments of these three candidate TFs are shown in Supplementary Fig. S2.

CYP72A566 is β-amyrin C-22β hydroxylase, involved in soyasaponin biosynthesis

Soyasapogenol B, the aglycone of soyasaponins identified in licorice, possesses hydroxyl groups at the C-22β and C-24 positions of the β-amyrin skeleton. In a previous study, we showed that CYP93E3 had β-amyrin C-24 hydroxylase activity (Seki et al. 2008); however, C-22β hydroxylase has not yet been identified in licorice. To prepare for the functional analysis of candidate TFs in the soyasaponin biosynthetic pathway, we sought to identify β-amyrin C-22β hydroxylase in soyasaponin biosynthesis.

In M. truncatula, co-expression of CYP93E2 and CYP72A61v2 produced soyasapogenol B in engineered yeast endogenously producing β-amyrin (Fukushima et al. 2013). CYP72A155 isolated from G. uralensis was shown to be phylogenetically close to CYP72A61v2; however, this enzyme did not show catalytic activities on β-amyrin (Seki et al. 2011). Previously, we performed RNA sequencing (RNA-Seq) analysis on tissue-cultured stolons of G. uralensis (Tamura et al. 2017a). When we searched for a contig corresponding to CYP72A155 in this RNA-Seq data, the best hit exhibited many mismatches with CYP72A155 (96% nucleotide identity). Therefore, we suspected that this CYP72A155-like contig was the putative β-amyrin C-22β hydroxylase gene in G. uralensis. To test this hypothesis, we first cloned this contig (designated CYP72A566 by the P450 nomenclature committee) and compared amino acid sequences with CYP72A61v2 and CYP72A155 (Supplementary Fig. S3). The amino acid sequence identity between CYP72A566 and CYP72A61v2 was 85%, slightly higher than the identity between CYP72A155 and CYP72A61v2 (82%). Next, the enzymatic activities of CYP72A566 against β-amyrin and 24-hydroxy-β-amyrin were tested in an engineered yeast strain endogeneously producing β-amyrin and 24-hydroxy-β-amyrin (strain 1, Supplementary Table S5). In strain 1, production of β-amyrin (1) and 24-hydroxy-β-amyrin (2) was observed (Fig. 2; mass spectra in Supplementary Fig. S4). When CYP72A566 was introduced into this strain to generate strain 2 (Supplementary Table S5), new peaks corresponding to sophoradiol (22β-hydroxy-β-amyrin) (3) and soyasapogenol B (4) were confirmed (Fig. 2). Therefore, CYP93E3 and CYP72A566 are two P450s involved in the biosynthesis of soyasapogenol B from β-amyrin. Phylogenetic analysis of the P450s involved in triterpenoid biosynthesis in legumes also supports the function of CYP72A566 as a β-amyrin C-22β oxidase (Supplementary Fig. S5).

Fig. 2

Gas chromatography–mass spectrometry (GC-MS) analysis of products isolated from engineered yeast strains for in vivo enzyme assay of CYP72A566. β-Amyrin (1), 24-hydroxy-β-amyrin (2), sophoradiol (3) and soyasapogenol B (4) were identified through comparison with authentic standards. Cytochrome P450 reductase (CPR) was co-expressed with CYP93E3 by galactose-inducible promoters. To avoid overlapping chromatograms, the baselines for strains 1 and 2 were moved 2.5 × 10⁷ and 3.7 × 10⁷ total ion counts, respectively. Mass spectra of identified peaks are shown in Supplementary Fig. S4.

Transient co-transfection assays of candidate TFs

To find potential transactivation of triterpenoid or sterol biosynthetic genes of G. uralensis by the three candidate TFs (GubHLH1–GubHLH3), we performed transient co-transfection assays with these candidate TFs against promoter sequences of three OSC and five P450 genes in G. uralensis (Fig. 3). Promoter regions of each gene (1.5–2.5 kb; Supplementary Table S1) were fused with a β-glucuronidase (GUS) reporter gene and transiently introduced into protoplasts isolated from tobacco BY-2 cells, together with an effector construct (35S:TF) and an internal control (35S:LUC). GubHLH1 and GubHLH2 significantly transactivated bAS, LUS, CYP88D6, CYP93E3, CYP72A566 and CYP716A179 promoters compared with a control vector. Transactivation of the CYP93E3 promoter by GubHLH1 and GubHLH2 was 19- and 17-fold, respectively, and that of the CYP72A566 promoter was 45- and 50-fold, respectively. GubHLH3 also significantly transactivated bAS (3.6-fold), CYP93E3 (44-fold) and CYP72A566 (254-fold) promoters, but did not transactivate LUS, CYP88D6, CYP72A154 or CYP716A179 promoters. The CAS promoter was also significantly transactivated by GubHLH3 (1.8-fold); however, the fold change was much lower than the transactivation of the CYP93E3 and CYP72A566 promoters. Therefore, transactivation by GubHLH3 was highly selective for promoters of soyasaponin biosynthetic genes.

Fig. 3

Transient co-transfection assays of promoters of triterpenoid biosynthetic genes of G. uralensis and GubHLH1–GubHLH3. Error bars indicate the SE of four transfections. Asterisks indicate significant differences between the control and each effector determined by Dunnett’s test (*P < 0.05, **P < 0.01, ***P < 0.001).

Identification of GubHLH3-binding sites on the CYP93E3 promoter

Various TFs have been reported to regulate transcription through binding to cis-acting regulatory elements on promoter regions (Yamaguchi-Shinozaki and Shinozaki 2005). To narrow down the binding sites of GubHLH3 on the CYP93E3 promoter, we first generated deletion constructs of the CYP93E3 promoter and performed transient co-transfection assays (Fig. 4A, C). GubHLH3 transactivated all three deletion constructs by >30-fold (Fig. 4C). Therefore, putative binding sites were predicted to be downstream of position −370 [relative to the transcription start site (TSS)] on the CYP93E3 promoter (Fig. 4A).

Fig. 4

Determination of binding sites of GubHLH1–GubHLH3 on the CYP93E3 promoter. (A) Deletion constructs of the CYP93E3 promoter and the location of E-box (E1) and N-box (N1 and N2) motifs. Numbers indicate the position relative to the transcriptional start site (TSS; +1). the sequence between +138 and +158 is the coding region of CYP93E3. (B) Mutations at E1, N1 and N2 introduced in this study. (C) Transient co-transfection assays of deletion series of the CYP93E3 promoter and GubHLH1–GubHLH3. (D) Transient co-transfection assays of the mutated CYP93E3 promoter and GubHLH1–GubHLH3. Error bars indicate the SE of four transfections. Asterisks indicate significant differences between the control and each effector determined by Dunnett’s test (*P < 0.05, **P < 0.01, ***P < 0.001). The values for the wild type CYP93E3 promoter are the same as those described in Fig. 3.

bHLH TFs are known to bind to the E-box motif (CANNTG) (Toledo-Ortiz et al. 2003). Moreover, GubHLH3 is similar to TSAR2 in M. truncatula (53% amino acid identity), and the N-box motif (CACGCG or CACGAG) (Pires and Dolan 2010) is reportedly recognized by TSAR2 (Mertens et al. 2016). Therefore, we focused on E-box and N-box motifs on the shortest promoter fragment (deletion 3) of CYP93E3 and introduced mutations into these motifs (Fig. 4A, B). In transient co-transfection assays using these mutated promoter fragments and 35S:GubHLH3, mutations at E1 and N2 reduced transactivation by GubHLH3 to 18- and 8-fold, respectively, but mutation at N1 showed almost the same level of transactivation as the wild-type promoter (Fig. 4D). When both E1 and N2 motifs were mutated simultaneously, the level of transactivation dropped to 3.4-fold. Therefore, E-box (E1) and N-box (N2) motifs are predicted to be involved in the binding of GubHLH3 to the CYP93E3 promoter. Similar results were obtained when GubHLH1 or GubHLH2 were used as effector constructs (Fig. 4C, D), suggesting that the E-box (E1) and the N-box (N2) are necessary for the binding of these TFs to the CYP93E3 promoter.

Overexpression of GubHLH3 enhances the expression of soyasaponin biosynthetic genes

To characterize GubHLH3 functionally in G. uralensis, we generated transgenic hairy roots overexpressing GubHLH3 under the Cauliflower mosaic virus (CaMV) 35S promoter. We obtained four empty vector control lines (Control) and five GubHLH3 overexpression (GubHLH3-OX) lines, and analyzed the transcript levels of triterpenoid or sterol biosynthetic genes and GubHLH1–GubHLH3 by quantitative real-time PCR (qPCR) (Fig. 5). In GubHLH3-OX lines, the transcript levels of bAS, CYP93E3 and CYP72A566, the three genes required for the biosynthesis of soyasapogenol B from 2,3-oxidosqualene, were all significantly enhanced compared with control lines. The relative transcript levels of bAS, CYP93E3 and CYP72A566 in GubHLH3-OX lines were more than three times higher than in control lines. In addition to these soyasaponin biosynthetic genes, CYP72A154 was also significantly up-regulated to approximately twice the level of the control lines. The transcript level of CYP88D6 in GubHLH3-OX lines was >30 times higher than in the control lines; however, the transcript levels of CYP88D6 and of GubHLH3 did not seem to be correlated with each other in individual hairy root lines (Supplementary Fig. S6). Therefore, it is unclear whether GubHLH3 up-regulates the expression of CYP88D6. The transcript levels of the two remaining triterpenoid biosynthetic genes, LUS and CYP716A179, and the sterol biosynthetic gene CAS were not enhanced in GubHLH3-OX lines compared with control lines (Fig. 5). We also generated two lines expressing a chimeric repressor of GubHLH3 (GubHLH3–SRDX), and analyzed the transcript levels of triterpenoid or sterol biosynthetic genes and GubHLH1–GubHLH3. The transcript levels of bAS and CYP93E3 were slightly lower than those of control lines; however, the CYP72A566 level was nearly the same as in the control lines.

Fig. 5

Quantitative real-time PCR (qPCR) analysis of triterpenoid biosynthetic genes and GubHLH1–GubHLH3 in transgenic hairy roots. Error bars indicate the SD of four (Control), five (GubHLH3-OX), and two (GubHLH3-SRDX) independent hairy root lines. Asterisks indicate significant differences between control lines and GubHLH3-OX or GubHLH3-SRDX lines determined by Dunnett’s test (*P < 0.05, **P < 0.01, ***P < 0.001).

To elucidate whether the enhanced expression of soyasaponin biosynthetic genes in GubHLH3-OX hairy root lines affects metabolite accumulation, we compared the relative quantities of sapogenins accumulated in the control line and GubHLH3-OX hairy root lines (Fig. 6). Compared with the control line, the relative amounts of soyasapogenol B and sophoradiol, a possible intermediate between β-amyrin and soyasapogenol B, were more than doubled in two of the three GubHLH3-OX lines. In particular, GubHLH3-OX-5 accumulated 5.6 times more soyasapogenol B and 12 times more sophoradiol than the control line. On the other hand, the amount of oleanolic acid in the GubHLH3-OX lines was less than the amount in the control line, suggesting that oleanolic acid biosynthesis competes with soyasaponin biosynthesis. We did not detect glycyrrhetinic acid or possible intermediates between β-amyrin and glycyrrhetinic acid, even in the GubHLH3-OX lines (Supplementary Fig. S7). Therefore, the observed increase in the expression level of CYP88D6 in several GubHLH3-OX lines (Supplementary Fig. S6) is not likely to contribute to glycyrrhizin biosynthesis.

Fig. 6

Relative quantification of sapogenins accumulated in transgenic hairy root lines. Analysis was performed for one control line (Control-2) and three GubHLH3-OX lines (GubHLH3-OX-3, 4 and 5). Line numbers correspond to qPCR analysis of individual lines shown in Supplementary Fig. S6. The peak area of each compound was normalized to an internal standard (echinocystic acid) and values are shown relative to the value of Control-2. Error bars indicate the SE of two technical replicates. In addition to the indicated compounds, a trace amount of lupeol was identified in the three GubHLH3-OX lines.

Expression of soyasaponin biosynthetic genes and GubHLH3 is induced by methyl jasmonate treatment

The biosynthesis of soyasaponins in cultured cells of G. glabra was up-regulated following the application of exogenous methyl jasmonate (MeJA) (Hayashi et al. 2003). To analyze the effects of exogenously applied MeJA on the expression of triterpenoid or sterol biosynthetic genes and GubHLH1– GubHLH3, we treated tissue-cultured stolons of G. uralensis with MeJA and analyzed the time-course transcriptional response of these genes by qPCR (Fig. 7). The three soyasaponin biosynthetic genes, bAS, CYP93E3 and CYP72A566, were all up-regulated by MeJA, and the three exhibited similar responses to each other. They were strongly up-regulated 3 h after MeJA treatment, then their expression levels dropped 12 h after MeJA treatment. Other triterpenoid biosynthetic genes and CAS showed different responses from those of the soyasaponin biosynthetic genes, suggesting that bAS, CYP93E3 and CYP72A566 are co-ordinately regulated under MeJA treatment. The expression of GubHLH3 was induced 0.5−1 h after MeJA treatment, peaking 2 h after treatment. This earlier response of GubHLH3 to MeJA than the responses of the soyasaponin biosynthetic genes suggests that GubHLH3 stimulates the expression of downstream soyasaponin biosynthetic genes in response to MeJA treatment. On the other hand, GubHLH1 and GubHLH2 were up-regulated 12 h after MeJA treatment. This is much later than the up-regulation of the soyasaponin biosynthetic genes. To analyze the effects of other elicitors, we applied yeast extract (YE) or salicylic acid (SA) to tissue-cultured stolons of G. uralensis, and analyzed the transcriptional responses (Supplementary Fig. S8). In previous studies, YE was shown to suppress the production of soyasaponins in cultured cells of G. glabra (Hayashi et al. 2005), and SA was shown to have little effect on the expression of bAS in M. truncatula (Suzuki et al. 2005). Therefore, these two elicitors were not expected to up-regulate the soyasaponin biosynthetic genes. Our results supported this prediction, as these two elicitors failed to up-regulate the expression of the soyasaponin biosynthetic genes or GubHLH3 (Supplementary Fig. S8). From these results, we suggest that GubHLH3 controls MeJA-responsive expression of soyasaponin biosynthetic genes in G. uralensis.

Fig. 7

Time-course qPCR analysis of triterpenoid biosynthetic genes and GubHLH1– GubHLH3 after methyl jasmonate (MeJA) treatment. Tissue-cultured stolons were treated with 0.1% ethanol (mock) or 100 μM MeJA. Relative transcript levels at 0 h were set equal to 1. Error bars indicate the SE of three technical replicates.

Discussion

Plants produce various classes of specialized metabolites, including triterpenoids. Although recent studies have revealed various enzyme genes involved in plant triterpenoid biosynthesis, their regulatory mechanisms remain largely unknown. Here, we identified GubHLH3, the bHLH TF regulating the expression of bAS, and the two P450 genes required for the biosynthesis of soyasaponins, in an important medicinal plant, G. uralensis.

The regulation of soyasaponin biosynthetic genes by GubHLH3 in G. uralensis is supported by a series of experimental results. First, in transient co-transfection assays of promoter:GUS and 35S:TF constructs, GubHLH3 significantly transactivated bAS, CYP93E3 and CYP72A566 promoters, but not the promoters of other triterpenoid biosynthetic genes (CYP88D6 and CYP72A154 for glycyrrhizin biosynthesis; CYP716A179 for oleanolic acid and betulinic acid biosynthesis; and LUS for betulinic acid biosynthesis) (Fig. 3). Secondly, in GubHLH3-OX hairy root lines of G. uralensis, the transcript levels of soyasaponin biosynthetic genes (bAS, CYP93E3 and CYP72A566) were clearly enhanced in all five GubHLH3-OX lines compared with control lines (Fig. 5; Supplementary Fig. S6). Metabolite analysis of representative transgenic hairy root lines showed higher levels of soyasapogenol B and sophoradiol in GubHLH3-OX lines compared with the control line (Fig. 6). Thirdly, expression of GubHLH3 was strongly up-regulated by the plant hormone MeJA, and the expression levels of bAS, CYP93E3 and CYP72A566 were also strongly up-regulated by MeJA, with similar responses among the three (Fig. 7). To compare the expression patterns of GubHLH3 and soyasaponin biosynthetic genes in field-grown intact G. uralensis plants, we performed qPCR analysis using RNA samples from the previously reported transcriptome analysis of G. uralensis (Supplementary Fig. S9). This transcriptome analysis was performed in four libraries to examine the differences between organs (roots or leaves), seasons (summer or winter) and strains (glycyrrhizin high-producing or low-producing strains), which were predicted to affect the expression of glycyrrhizin biosynthetic genes (Ramilowski et al. 2013). The expression levels of the three soyasaponin biosynthetic genes were particularly high in the roots of the high glycyrrhizin-producing strain harvested in summer (Supplementary Fig. S9). This expression pattern is very similar to that of GubHLH3, further supporting our conclusion that soyasaponin biosynthetic genes are under the control of GubHLH3. In contrast, GubHLH1 and GubHLH2 were expressed at substantial levels in the other samples, suggesting that these two TFs are not responsible for the regulation of soyasaponin biosynthetic genes.

CYP88D6 is a key enzyme in the production of glycyrrhizin in G. uralensis (Seki et al. 2008). In our qPCR analysis of GubHLH3-OX hairy root lines, four of the five lines showed higher transcription levels of CYP88D6 compared with the control lines. However, the transcription levels of GubHLH3 and CYP88D6 did not correlate with each other (Supplementary Fig. S6). Moreover, the expression pattern of CYP88D6 after MeJA treatment was completely different from that of GubHLH3 (Fig. 7), and CYP88D6 and GubHLH3 also showed different responses to SA or YE (Supplementary Fig. S8). Therefore, it is not reasonable to suggest that GubHLH3 directly regulates CYP88D6 expression. There is a possibility that GubHLH3 overexpression indirectly affects the expression of CYP88D6. The other two bHLH TFs cloned in this study, GubHLH1 and GubHLH2, showed significant transactivation against the CYP88D6 promoter (Fig. 3). However, their responses to MeJA, SA and YE differed markedly from those of GubHLH3 (Fig. 7; Supplementary Fig. S8), and the expression patterns in intact plants were also quite different (Supplementary Fig. S9). We did not detect glycyrrhetinic acid or possible intermediates between β-amyrin and glycyrrhetinic acid in any of the three GubHLH3-OX lines in the metabolite analysis (Supplementary Fig. S7). Further screening, such as a comprehensive gene co-expression analysis of licorice samples, is necessary to find the regulators of glycyrrhizin biosynthetic genes including CYP88D6.

Both licorice and M. truncatula produce soyasaponins, which have soyasapogenol B as their aglycone (Hayashi et al. 1990, Huhman et al. 2005, Carelli et al. 2011). The three enzymes required for the biosynthesis of soyasapogenol B from 2,3-oxidosqualene in licorice (bAS, CYP93E3 and CYP72A566) are the corresponding orthologs of those in M. truncatula (bAS, CYP93E2 and CYP72A61v2) (Supplementary Fig. S5). In M. truncatula, bHLH TF TSAR1 is reported to boost soyasaponin (non-hemolytic saponin) biosynthesis and activate the expression of relevant enzyme genes, bAS, CYP93E2 and CYP72A61v2, whereas TSAR2 is reported to boost hemolytic saponin biosynthesis, and activate the expression of the relevant enzyme genes, bAS, CYP716A12 and CYP72A68v2 (Mertens et al. 2016). Interestingly, phylogenetic analysis indicated that GubHLH3 is phylogenetically closer to TSAR2 (53% amino acid sequence identity) than to TSAR1 (43% amino acid sequence identity) (Supplementary Fig. S10). However, GubHLH3 activated the expression of bAS, CYP93E3 and CYP72A566, the three corresponding orthologous genes of which were activated by TSAR1 in M. truncatula. Therefore, although biosynthetic enzymes for the production of soyasaponins are shared between M. truncatula and G. uralensis, the regulatory mechanism of soyasaponin biosynthesis is not conserved in these two species. By conducting phylogenetic analyses of putative bHLH TFs encoded by unigenes identified in the previous G. uralensis transcriptome analysis (Ramilowski et al. 2013), we found one putative ortholog of TSAR1 (Unigene1680) and another phylogenetically close unigene (Unigene16153) (Supplementary Fig. S10). Further studies on the comparative analysis of these TFs as well as GubHLH3 may provide a better understanding of the evolutionary mechanisms of the regulation of triterpenoid biosynthesis in legumes.

Materials and Methods

Plant materials

Tissue-cultured stolons of G. uralensis (Hokkaido-iryodai strain) were maintained in Murashige and Skoog (MS) medium (Duchefa Biochemie) supplemented with 6% sucrose and 0.01 μM 1-naphthaleneacetic acid (NAA) as reported previously (Kojoma et al. 2010). Leaves of G. uralensis strain 308-19 (Mochida et al. 2017) were used for genomic DNA extraction. Roots of G. uralensis strain 308-19 harvested in June 2011 (Ramilowski et al. 2013) were used for RNA extraction. Glycyrrhiza uralensis (Hokkaido-iryodai strain) seeds were collected in 2007. Suspension cultures of tobacco BY-2 cells (Nicotiana tabacum L. cultivar Bright Yellow 2) were maintained at 26°C with rotary shaking at 125 r.p.m. in 100 ml of MS plant salt mixture (Wako Pure Chemical Industries) supplemented with 3% sucrose, 0.2 g l^–1 potassium dihydrogen phosphate, 1 mg l^–1 thiamine hydrochloride, 0.1 g l^–1myo-inositol and 0.2 mg l^–1 2,4-D, and subcultured every 7 d.

Chemicals

β-Amyrin, α-amyrin, lupeol, oleanolic acid, ursolic acid, glycyrrhetinic acid (18β-glycyrrhetinic acid) and echinocystic acid were purchased from Extrasynthese. Soyasapogenol B was purchased from Tokiwa Phytochemical. Sophoradiol, 24-hydroxy-β-amyrin, 11-oxo-β-amyrin, 30-hydroxy-β-amyrin and 11-deoxoglycyrrhetinic acid were a kind gift of Dr. Kiyoshi Ohyama (Tokyo Institute of Technology). Betulinic acid, methyl-β-cyclodextrin and gibberellin A₃ (GA₃) were purchased from Tokyo Chemical Industry. MeJA, NAA and SA were purchased from Sigma-Aldrich.

Isolation of genomic DNA and total RNA from plant tissues

Genomic DNA was isolated using Nucleon Phytopure Genomic DNA Extraction Kits (GE Healthcare) according to the manufacturer’s instructions. Total RNA was extracted as previously described (Tamura et al. 2017a).

Isolation of promoter regions

To design primers to isolate the promoter regions of bAS, CYP88D6 and CYP93E3, we previously performed PCR-based genome walking using genomic DNA isolated from tissue-cultured stolons of G. uralensis (Hokkaido-iryodai strain) (Supplementary Methods S1). Based on the promoter sequences obtained, we isolated promoter regions from the genomic DNA of strain 308-19 with primers 1–6 (Supplementary Table S6) by PCR. The promoter regions of CAS, LUS, CYP72A154, CYP72A566 and CYP716A179 were isolated from genomic DNA of strain 308-19 by PCR with primers 7–16 (Supplementary Table S6) designed from the genomic information of strain 308-19 (Mochida et al. 2017). All of the amplified promoter regions were cloned into pENTR/D-TOPO (Thermo Fisher Scientific) to make entry clones.

Determination of the transcriptional start sites

The TSSs were determined by the 5′-rapid amplification of cDNA ends (RACE) PCR method using the SMARTer RACE cDNA Amplification Kit (Clontech/TAKARA BIO) with primers 17–32 (Supplementary Table S6). First-strand cDNA was synthesized from 1 μg of total RNA according to the manufacturer’s instructions. Total RNA obtained from the roots of strain 308-19 was used to determine the TSSs of bAS, CAS, CYP88D6, CYP72A154 and CYP93E3, and total RNA obtained from tissue-cultured stolons of the Hokkaido-iryodai strain was used for LUS, CYP72A566 and CYP716A179.

High-throughput Y1H screening of the Arabidopsis TF library

Fragments (∼500 bp) of promoter regions of bAS, CYP88D6 and CYP93E3 were amplified by primers 33–56 and adaptor primers 57 and 58 (Supplementary Table S6) to be cloned into pDONRGm-P4P1R (Oshima et al. 2011) using Gateway BP Clonase II Enzyme mix (Thermo Fisher Scientific), then transferred into R4L1pDEST_HISi (Mitsuda et al. 2010) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific). The obtained bait constructs were integrated into the URA3 locus of yeast strain YM4271 by homologous recombination to make Y1H yeast strains. Y1H screening was performed as previously described (Oda-Yamamizo et al. 2016), with 376 mini-pools covering 1,428 Arabidopsis TFs.

In vivo enzyme assay of CYP72A566

CYP72A566 was amplified from the first-strand cDNA obtained from tissue-cultured stolons of G. uralensis (Hokkaido-iryodai strain) with primers 59 and 60 (Supplementary Table S6) and cloned into pENTR/D-TOPO (Thermo Fisher Scientific) to make an entry clone. cDNA for CYP72A566 was transferred into pYES-DEST52 (Thermo Fisher Scientific) and a Gateway compatible version of pESC-HIS (Agilent Technologies) using the Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific) to generate galactose-inducible expression of CYP72A566. These constructs were introduced, using the Frozen-EZ Yeast Transformation II Kit (Zymo Research), into Saccharomyces cerevisiae INVSc1 (MATa his3Δ1 leu2 trp1-289 ura3-52/MATα his3Δ1 leu2 trp1-289 ura3-52; Thermo Fisher Scientific) pre-transformed with pYES3-ADH-OSC1 (Fukushima et al. 2011) for constitutive expression of bAS and pELC-CYP93E3 (Seki et al. 2008) for galactose-inducible dual expression of cytochrome P450 reductase (Lotus japonicus CPR1) and CYP93E3. The engineered yeast strains made for this study are listed in Supplementary Table S5. For in vivo enzyme assay, each yeast strain was cultured as previously described (Tamura et al. 2017b), but the culture time after galactose induction was prolonged for 5 d. The yeast culture was extracted as previously described (Tamura et al. 2017a). For derivatization, 100 μl of the extract was evaporated and trimethylsilylated with 100 μl of N-methyl-N-(trimethylsilyl)trifluoroacetamide (Sigma-Aldrich) at 80°C for 30 min before gas chromatography–mass spectrometry (GC-MS) analysis.

Reporter constructs for transient co-transfection assays

To produce promoter:GUS reporter constructs, each promoter region was transferred to pGEM-GW-GUS-NOS (Supplementary Methods S2) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific). To generate deletion constructs for the CYP93E3 promoter, primers 61–64 (Supplementary Table S6) were used to obtain deletions 1 and 2 (Fig. 4A) by inverse PCR, and primers 65 and 6 (Supplementary Table S6) were used to obtain deletion 3 (Fig. 4A) by cloning into pENTR/D-TOPO (Thermo Fisher Scientific). To introduce a mutation at the E-box (E1) and N-box (N1 and N2) (Fig. 4A, B), inverse PCR was performed using primers 66–71 (Supplementary Table S6). The obtained deletion and mutation constructs were transferred to pGEM-GW-GUS-NOS using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific).

Cloning of TF genes and construction of effector constructs for transient co-transfection assays

First-strand cDNA obtained from total RNA isolated from the roots of strain 308-19 was used as a PCR template. The full-length coding sequences (CDSs) without stop codons of GubHLH1, GubHLH2 and GubHLH3 were amplified by PCR with primers 72–77 (Supplementary Table S6) and cloned into pENTR/D-TOPO (Thermo Fisher Scientific) to make entry clones. Each CDS was transferred into pDEST_35S_HSP_GWB5 (Fujiwara et al. 2014) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific) to make GubHLH1–3/pDEST_35S_HSP_GWB5 constructs, and used as effector constructs for transient co-transfection assays.

Transient co-transfection assays

Transient co-transfection assays were performed using polyethylene glycol (PEG)-mediated transient transformation of protoplasts isolated from tobacco BY-2 cells as previously described (Sakamoto et al. 2012), with some modifications. Briefly, 12.5 ml of BY-2 cells were collected in a 50 ml tube 7 d after subculturing and incubated with 20 ml of digestion buffer (Sakamoto et al. 2012) for 2 h at 25°C with reciprocal shaking at 80 r.p.m. in the dark. After digestion, the protoplasts were treated accordingly (Sakamoto et al. 2012). For co-transfection assays, 35 μl of protoplast suspension (3–5×10⁵ cells ml^–1), 45 μl of the PEG solution (Sakamoto et al. 2012) and 10 μl of plasmid solution containing 1 μg of effector construct, 1 μg of reporter construct and 0.5 μg of 35S:LUC internal control (pAM-PAT-LUC; Supplementary Methods S2) were mixed in a 96-well plate (Thermo Fisher Scientific) by vortexing at 900 r.p.m. for 15 s. After incubation at room temperature for 10 min in the dark, the mixture was suspended with 200 μl of W5 buffer (Sakamoto et al. 2012), then centrifuged at 100×g for 5 min. After washing three times with the W5 buffer, 200 μl of W5 buffer was added and the protoplasts were incubated at 25°C for 24 h in the dark. The incubated protoplasts were vortexed at 900 r.p.m. for 15 s, then 20 μl of Luciferase Cell Culture Lysis 5× Reagent (Promega) was added. The mixture was vortexed for another 15 s and centrifuged at 100×g for 1 min. The supernatant was used for luciferase (LUC) and GUS reporter assays. For LUC assays, 50 μl of Luciferase Assay Reagent (Promega) were added to 10 μl of supernatant, and LUC activities were determined by luminescence measurement using a Wallac 1420 multilabel counter (PerkinElmer). For GUS assays, 100 μl of MUG substrate mix for GUS assay (Yoo et al. 2007) were added to 10 μl of the supernatant or Luciferase Cell Culture Lysis Reagent diluted with the W5 buffer as a blank. After incubation at 37°C for 90 min, GUS activities were measured with a Wallac 1420 multilabel counter (PerkinElmer) by excitation at 355 nm and measuring emission at 460 nm. The average value of the blanks (four wells) was subtracted from the measured GUS activities of the samples to calculate the GUS/LUC activities. Each assay was conducted at least twice, and gave similar results.

Generation of transgenic hairy root lines

Constructs for dual expression of GubHLH3(–SRDX) and green fluorescent protein (GFP) were used to generate transgenic hairy root lines. To construct the plasmid for dual expression of GubHLH3 and GFP, the GubHLH3-HSPter cassette of GubHLH3/pDEST_35S_HSP_GWB5 was amplified with primers 78 and 79 (Supplementary Table S6) and cloned into pENTR/D-TOPO (Thermo Fisher Scientific). The cassette was then transferred into pBI-OX-GW (GFP selection) (Inplanta Innovations) to make the GubHLH3 and GFP dual expression plasmid. To construct the plasmid for dual expression of GubHLH3-SRDX and GFP, the CDS of GubHLH3 was transferred into pDEST_35S_SRDX_HSP_GWB5 (Fujiwara et al. 2014) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific) to make GubHLH3/pDEST_35S_SRDX_HSP_GWB5. Then, the GubHLH3-SRDX-HSPter cassette was amplified with primers 78 and 79 (Supplementary Table S6) and transferred into pBI-OX-GW (GFP selection) (Inplanta Innovations) accordingly. pBI-OX-GW (GFP selection) lacking a Gateway cassette was used as an empty vector control. These three constructs were introduced into Agrobacterium rhizogenes strain ATCC15834 by electroporation.

Glycyrrhiza uralensis (Hokkaido-iryodai strain) seeds were scratched to stimulate germination, and placed on 1/2 MS medium (Duchefa Biochemie) supplemented with 1% sucrose at 22°C with a 16 h light/8 h dark cycle. These temperature and light conditions were kept constant throughout hairy root culture. After 6–8 d, roots were removed and the sectioned surface was coated with pellets of Agrobacterium rhizogenes. The infected seedlings were placed on the same medium and co-cultured for 2 d. Then the seedlings were transferred into 1/2 MS medium supplemented with 1% sucrose and 125 mg l^–1 cefotaxime to eliminate Agrobacterium. The seedlings were transferred into fresh medium twice before isolation of hairy roots. One month after infection, hairy roots strongly expressing GFP were isolated from the seedlings and placed on 1/2 McCown woody plant medium (Duchefa Biochemie) supplemented with 1% sucrose, 0.01 μM GA₃ and 125 mg l^–1 cefotaxime. One day after isolation, hairy roots were harvested, frozen in liquid nitrogen, and used for RNA extraction. The rest of the hairy roots were subcultured several times on the same medium to remove Agrobacterium completely, and then transferred to the same medium lacking cefotaxime. Hairy roots cultured for 1 month after subculture were used for metabolite analysis.

Analysis of sapogenins in hairy roots

Extraction of metabolites, acid hydrolysis and sample preparation for GC-MS analysis were performed as previously described (Tamura et al. 2017a) from 40 mg of freeze-dried powder of hairy roots.

GC-MS analysis

GC-MS analysis was performed as previously described (Tamura et al. 2017b), but the hold time at 300°C was 28 min (extracts from yeasts) or 38 min (extracts from hairy roots), and the m/z range was 50–850. Peaks were identified by comparing their retention times and mass spectra with those of authentic standards.

Treatment of tissue-cultured stolons with plant hormones or elicitors

Tissue-cultured stolons were cultured for >2 weeks in MS medium supplemented with 6% sucrose before treatment. MeJA and SA were diluted with ethanol to make 100 mM stock solutions, then 100 μl was added to 100 ml of medium (final concentration, 100 μM). YE (BD Biosciences) was dissolved in water to make a 10% (w/v) stock solution, and 1 ml was added to 100 ml of culture medium (final concentration, 0.1%). For mock treatment in the time-course qPCR analysis after MeJA treatment, 100 μl of ethanol was added to 100 ml of medium. Untreated tissue-cultured stolons were used as the 0 h time point in each series of experiments.

Quantitative real-time PCR

qPCR was performed as previously described (Tamura et al. 2017a). Relative transcript levels of each target gene were calculated using β-tubulin (Seki et al. 2008) (GenBank accession No. LC318135) as a reference gene. The amplification of each sample was performed three times, using primers 80–107 (Supplementary Table S6) designed using the Primer3 website (http://bioinfo.ut.ee/primer3-0.4.0/) (Koressaar and Remm 2007, Untergasser et al. 2012).

Phylogenetic analysis

Full-length amino acid sequences of previously characterized P450s and bHLH TFs were collected from GenBank (http://www.ncbi.nlm.nih.gov/genbank/) (Supplementary Tables S7, S8). To collect putative bHLH TFs in the previous transcriptome analysis of G. uralensis (Ramilowski et al. 2013), the predicted protein sequences were searched against a hidden Markov model (HMM) of HLH (PF00010) obtained from the Pfam database (Finn et al. 2016) as previously described (Mochida et al. 2009). Phylogenetic trees were generated as previously described (Tamura et al. 2017b).

Statistical analysis

Dunnett’s tests were conducted using the multcomp package in R (Hothorn et al. 2008).

Supplementary Data

Supplementary data are available at PCP online.

Funding

This work was supported by the Ministry of Education, Culture, Sports, Science, and Technology of Japan [JSPS KAKENHI grant Nos. JP26450123 and JP17K07754 to H.S., JP16K08301 to M.K. and JP15H04485 to T.M.]; the Yamada Science Foundation [research funding to H.S.]; and the Yoshida Scholarship Foundation [Doctor 21 scholarship to K.T.].

Acknowledgments

We are very grateful to the Takeda Garden for Medicinal Plant Conservation, Kyoto, Japan (Takeda Pharmaceutical Company Ltd.) for providing G. uralensis 308-19 strain plants. We thank Dr. Shingo Sakamoto (AIST) and Ms. Fumie Tobe (AIST) for technical advice on transient co-transfection assays and Y1H assays. We also thank Dr. Kiyoshi Ohyama (Tokyo Institute of Technology) for providing standard compounds, Dr. Bekir Ülker (Max Planck Institute for Plant Breeding Research) for providing the pAM-PAT-GW vector, and Ms. Keiko Fukamoto (Osaka University) for technical assistance.

Disclosures

The authors have no conflicts of interest to declare.

Abbreviations

Abbreviations
 
• bAS

  β-amyrin synthase

 
• bHLH

  basic helix–loop–helix

 
• CAS

  cycloartenol synthase

 
• CDS

  coding sequence

 
• GA₃

  gibberellin A₃

 
• GC-MS

  gas chromatography–mass spectrometry

 
• GFP

  green fluorescent protein

 
• GUS

  β-glucuronidase

 
• LUC

  luciferase

 
• LUS

  lupeol synthase

 
• MeJA

  methyl jasmonate

 
• MS

  , Murashige and Skoog

 
• NAA

  1-naphthaleneacetic acid

 
• OSC

  oxidosqualene cyclase

 
• OX

  overexpression

 
• P450

  cytochrome P450 monooxygenase

 
• PEG

  polyethylene glycol

 
• qPCR

  quantitative real-time PCR

 
• RNA-Seq

  RNA sequencing

 
• SA

  salicylic acid

 
• TF

  transcription factor

 
• TSS

  transcription start site

 
• UGT

  UDP-dependent glycosyltransferase

 
• YE

  yeast extract

 
• Y1H

  yeast one-hybrid

footnotes

footnotes
 
• The nucleotide sequences isolated in this study have been submitted to the DDBJ under the accession numbers LC318131 (GubHLH1), LC318132 (GubHLH2), LC318133 (GubHLH3) and LC318134 (CYP72A566), respectively