Metabolic switching of astringent and beneficial triterpenoid saponins in soybean is achieved by a loss‐of‐function mutation in cytochrome P450 72A69

Summary

Triterpenoid saponins are major components of secondary metabolites in soybean seeds and are divided into two groups: group A saponins, and 2,3‐dihydro‐2,5‐dihydroxy‐6‐methyl‐4H‐pyran‐4‐one (DDMP) saponins. The aglycone moiety of group A saponins consists of soyasapogenol A (SA), which is an oxidized β‐amyrin product, and the aglycone moiety of the DDMP saponins consists of soyasapogenol B (SB). Group A saponins produce a bitter and astringent aftertaste in soy products, whereas DDMP saponins have known health benefits for humans. We completed map‐based cloning and characterization of the gene Sg‐5, which is responsible for SA biosynthesis. The naturally occurring sg‐5 mutant lacks group A saponins and has a loss‐of‐function mutation (L164*) in Glyma15g39090, which encodes the cytochrome P450 enzyme, CYP72A69. An enzyme assay indicated the hydroxylase activity of recombinant CYP72A69 against SB, which also suggested the production of SA. Additionally, induced Glyma15g39090 mutants (R44* or S348P) lacked group A saponins similar to the sg‐5 mutant, indicating that Glyma15g39090 corresponds to Sg‐5. Endogenous levels of DDMP saponins were higher in the sg‐5 mutant than in the wild‐type lines due to the loss of the enzyme activity that converts SB to SA. Interestingly, the genomes of palaeopolyploid soybean and the closely related common bean carry multiple Sg‐5 paralogs in a genomic region syntenic to the soybean Sg‐5 region. However, SA did not accumulate in common bean samples, suggesting that Sg‐5 activity evolved after gene duplication event(s). Our results demonstrate that metabolic switching of undesirable saponins with beneficial saponins can be achieved in soybean by disabling Sg‐5.

Introduction

Triterpenoid saponins form a structurally diverse class of natural compounds widely distributed in plants (Vincken et al., 2007; Sawai and Saito, 2011). They contain a triterpene aglycone moiety derived from 2,3‐oxidosqualene and often accumulate conjugated to sugars and other macromolecules. In addition to their important agronomic functions such as in pathogen defense (Papadopoulou et al., 1999; Geisler et al., 2013), triterpenoid saponins influence a wide variety of physiological responses in humans (Shibata, 2001; Kitagawa, 2002; Seki et al., 2011). Some saponins are used as pharmacological reagents or as natural sweeteners. Owing to the importance of triterpenoid saponins for commercial applications in the food, cosmetics, and pharmaceutical industries, their biosynthetic pathway has been extensively studied in both crop and medicinal plants using advanced genomics technologies (Muranaka and Saito, 2013).

Soybean (Glycine max (L.) Merr.) is one of the most economically important crop plants worldwide and triterpenoid saponins are major components of secondary metabolites in its mature seeds (Tsukamoto et al., 1993; Takada et al., 2010, 2013). Saponins comprise more than 2% of the dry weight of the hypocotyl in developed mature seeds (Fenwick et al., 1991), and are characterized as group A saponins or 2,3‐dihydro‐2,5‐dihydroxy‐6‐methyl‐4H‐pyran‐4‐one (DDMP) saponins based on their chemical structures. Group A saponins are bisdesmoside‐type saponins that have two sugar chains at the C‐3 and C‐22 hydroxyl groups of the aglycone moiety, soyasapogenol A (3β, 21β, 22β, 24‐tetrahydroxyolean‐12‐ene; SA) (Shiraiwa et al., 1991; Figures 1(a) and S1). The DDMP saponins are monodesmoside‐type saponins that contain a sugar chain at the C‐3 hydroxyl group of the aglycone molecule, in which a DDMP residue forms a hemiacetal linkage with soyasapogenol B (3β, 22β, 24‐trihydroxyolean‐12‐ene; SB) at the C‐22 hydroxyl group (Kudou et al., 1992, 1993). Importantly, DDMP saponins and their derivatives (i.e., group B and group E saponins; Figure S1) have health‐promoting functions such as the prevention of dietary hypercholesterolemia (Fenwick et al., 1991; Murata et al., 2005, 2006), suppression of colon cancer cell proliferation (Ellington et al., 2005, 2006), and antiperoxidation of lipids and liver‐protecting activities resulting from accelerated thyroid hormone secretion (Ishii and Tanizawa, 2006). In contrast, acetylated group A saponins cause bitter and astringent aftertastes in soy products (Okubo et al., 1992). Both SA and SB are synthesized from 2,3‐oxidosqualene via β‐amyrin through cyclization and oxidation steps (Kushiro et al., 1998; Shibuya et al., 2006; Takada et al., 2013); however, the C‐21 hydroxyl group is specific to SA and thereby distinguishes SA from SB (Figure 1a). Because SA has the aglycone moiety commonly found in all of group A saponin compounds, the hydroxylation of C‐21 of the β‐amyrin skeleton is considered essential not only for the biosynthesis of SA but also for that of all group A saponins (Takada et al., 2013).

Figure 1

Changes in saponin levels in the sg‐5 cultivar ‘Tohoku 152’.

(a) Chemical structures of soyasapogenols and saponins. While DDMP saponins are synthesized by DDMP conjugation and glycosylation of soyasapogenol B, group A saponins are synthesized by glycosylation of soyasapogenol A. A hydroxyl group attached to the C‐21 position of the β‐amyrin skeleton is specific to soyasapogenol A (red circle). ‘Tohoku 152’ is deficient in all group A saponin compounds because they lack aglycone soyasapogenol A (Takada et al., 2013). Different combinations of sugar moieties conjugated to the R₁ or R₂ positions are also presented in Figure S1.

(b) Changes in saponin levels in the ‘Tohoku 152’ cultivar. Endogenous saponin levels in the hypocotyls of mature seeds in Sg‐5 (‘Williams 82’, ‘Enrei’, and ‘Ohsuzu’) and sg‐5 (‘Tohoku 152’) cultivars were analyzed by liquid chromatography‐mass spectrometry. Asterisks indicate significance of the change in saponin levels in mutants relative to that in the Sg‐5 cultivars (Welch's t‐test, *: max[p] ≤ 0.01). Details of the sugar compositions for group A saponins (Ab, Ac, Af), deacetylated group A saponins (dAb + 1Ac, dAb + 2Ac, and dAb + 3Ac), and DDMP saponins (αg and βg) are shown in Figure S1. Error bars represent the standard deviation of three biological replicates. n.d.: not detected.

The diversity of sugar moieties in soybean saponins is influenced by seven alleles at three different loci: Sg‐1, Sg‐3, and Sg‐4 (Shiraiwa et al., 1990; Tsukamoto et al., 1993; Kikuchi et al., 1999; Takada et al., 2010, 2012). Sg‐1 (Glyma07g38460) was recently cloned and encodes a uridine diphosphate (UDP)‐sugar‐dependent glycosyltransferase (Sayama et al., 2012). The loss‐of‐function sg‐1⁰ allele results in the absence of the second sugar moiety at the C‐22 position of SA that prevents the generation of acetylated forms of group A saponins. The sg‐1⁰ allele was used to develop the ‘Kinusayaka’ commercial soybean cultivar, in which bitter and astringent group A saponins are absent (Kato et al., 2007). The biosynthesis of group A saponins can be also regulated by a single recessive mutant allele, designated sg‐5. This allele was first identified in wild soybean and can disable SA accumulation and eliminate all group A saponins (Sasama et al., 2010; Takada et al., 2013). Because endogenous levels of DDMP saponins are also elevated by the sg‐5 mutant allele, this allele may be useful for generating soybean varieties relevant for the production of food products with improved taste and enhanced health benefits.

In this study, we identified Glyma15g39090 as the Sg‐5 gene that is essential for SA biosynthesis in soybean. The natural sg‐5 mutant was observed to carry a premature stop codon in Glyma15g39090 (L164*). This gene encodes a cytochrome P450 (CYP) 72A enzyme (CYP72A69), and using an enzyme activity assay, we show that the soybean Sg‐5 protein has catalytic activity in a heterologous yeast system that supports the use of SB as a substrate for the production of SA. Furthermore, two additional independent lines with induced mutations in Glyma15g39090 (R44* and S348P) exhibited decreased accumulation of group A saponins and increased accumulation of DDMP saponins, similar to the natural sg‐5 allele. Additionally, the R44* mutation was observed to be allelic to sg‐5 in terms of saponin phenotype, indicating that Glyma15g39090 is the gene responsible for the sg‐5 phenotype. Interspecies or inter‐chromosomal comparisons of synteny in soybean, common bean (Phaseolus vulgaris), and barrel medic (Medicago truncatula) suggest that Sg‐5 activity has evolved through neofunctionalization, a process by which a duplicated gene mutates to develop a new function (Rastogi and Liberles, 2005; Conrad and Antonarakis, 2007). Our results demonstrate that the biosynthesis of undesirable group A saponins and beneficial DDMP saponins can be simultaneously controlled in soybean plants by disabling Sg‐5 function.

Results

Effect of sg‐5 on endogenous saponin accumulation

A previous study determined that the naturally occurring sg‐5 allele decreases the accumulation of group A saponins in developed mature seeds (Takada et al., 2013) (Figure 1b). To investigate whether sg‐5 influences the accumulation of group A saponins in other tissues, we analyzed the tissue‐specific accumulation of saponins in 2‐week‐old plants and developing seeds. The DDMP saponins accumulated in all tissues of the analyzed Sg‐5 cultivars, ‘Williams 82’, ‘Enrei’ and ‘Ohsuzu’ (Figure S2). They also accumulated in the sg‐5 cultivar ‘Tohoku 152’, a soybean breeding line in which the sg‐5 mutation was introduced. There was a greater abundance of DDMP saponins in the ‘Tohoku 152’ hypocotyls than in the Sg‐5 hypocotyls, consistent with the results of an earlier study (Takada et al., 2013). The elevated DDMP levels in the ‘Tohoku 152’ hypocotyls were associated with increased levels of saponin αg (Figure 1b). Unlike the DDMP saponins, the group A saponins accumulated only in the hypocotyls, cotyledons, and roots of the Sg‐5 cultivars. They were undetectable in the leaves and stems of 2‐week‐old plants (Figure S2). However, ‘Tohoku 152’ seeds did not accumulate detectable amounts of group A saponins in any tissue. These results indicate that the sg‐5 allele affects the accumulation of group A saponins not only in seeds, but also in other plant tissues.

Map‐based cloning of Sg‐5

Previous linkage analyses indicated the Sg‐5 locus was positioned between the simple sequence repeat (SSR) markers Satt117 and GMES0332 on chromosome 15 (Takada et al., 2013; Figure 2a). To more precisely determine the genomic position of the Sg‐5 locus, 148 BC₄F₂ individuals derived from a cross between ‘Ohsuzu’ (Sg‐5) and ‘Tohoku 152’ (sg‐5) soybean cultivars were genotyped using SSR markers (Data S1). When the saponin phenotype (i.e., the presence or absence of group A saponins) was compared with the corresponding genotype, the Sg‐5 locus was mapped to the 3.4‐Mb genomic region between the SSR markers Gm15.43503k and GMES0332 (Figure 2a). We further delimited the position of the Sg‐5 locus using 342 BC₄F₃ progenies of the BC₄F₂ recombinant individual. The Sg‐5 locus was eventually mapped to a 138‐kb region between the SSR markers Gm15.45648k and Gm15.45786k. According to the Glyma1.1 soybean reference genome sequence (Schmutz et al., 2010), there are 10 predicted genes (Glyma15g39090 to Glyma15g39172) in this region. Of these genes, two encode putative Pif1 DNA helicases, one encodes a putative peptidase, three encode proteins with unknown function, and four encode putative CYP72A proteins. Three of the four CYP72A genes, Glyma15g39090, Glyma15g39150, and Glyma15g39160, encode CYP72A proteins of a canonical length (>400 amino acids). However, Glyma15g39098 is a pseudo‐gene that includes a premature stop codon in the Sg‐5 cultivars (‘Williams 82’, ‘Enrei’, and ‘Ohsuzu’). This stop codon results in the premature termination of the protein sequence after 163 amino acids (L164*; Figure S3). Interestingly, both the Glyma15g39090 and Glyma15g39098 genomic DNA fragments were not amplified by polymerase chain reaction (PCR) in ‘Tohoku 152’ (Figure 2b). Additionally, Southern blot analysis indicated that the genomic region in the vicinity of Glyma15g39090 and Glyma15g39098 was shorter in ‘Tohoku 152’ than in the Sg‐5 cultivars (Figure 2c). Using dideoxy DNA sequencing of a PCR fragment encompassing the Glyma15g39090 and Glyma15g39098 genomic region revealed that a 6.7‐kb fragment spanning from the middle of Glyma15g39090 to the middle of Glyma15g39098 was absent in the ‘Tohoku 152’ genome (Figure S4). This mutant gene consisted of the anterior and posterior portions of Glyma15g39090 and Glyma15g39098, respectively, in ‘Tohoku 152’ (Figures 2d and S3). Thus, similar to Glyma15g39098 in the Sg‐5 cultivars, Glyma15g39090 in the ‘Tohoku 152’ genome included a premature stop codon (L164*; Figure S3). These results indicate that the sg‐5 mutant contains a loss‐of‐function nonsense mutation in Glyma15g39090.

Figure 2

Map‐based cloning of Sg‐5.

(a) Genetic analysis of sg‐5. Genetic linkage analysis was completed using 148 BC₄F₂ (i) and 342 BC₄F₃ (ii) individuals derived from the ‘Ohsuzu’ (Sg‐5) × ‘Tohoku 152’ (sg‐5) cross. ‘Ohsuzu’ and ‘Tohoku 152’ genomic regions are indicated with red or blue filled bars, respectively, while heterozygous regions are indicated with purple filled bars. ‘Saponin type’ indicates the zygosity in terms of group A saponin accumulation, which was determined in the descendants of each line. Genes annotated as putative CYP72A genes (Glyma15g39090, Glyma15g39098, Glyma15g39150, and Glyma15g39160) are indicated with gray filled boxes (bottom).

(b) PCR detection of CYP72A genes. Glyma15g39090, Glyma15g39098, Glyma15g39150, and Glyma15g39160 genomic DNA fragments were amplified in the Sg‐5 cultivars ‘Williams 82’ (W) and ‘Ohsuzu’ (O), and in the sg‐5 cultivar ‘Tohoku 152’ (T).

(c) Southern blot analysis of the Sg‐5 genomic region. Genomic DNA was digested with the restriction enzyme EcoRI. The genomic DNA fragments containing Glyma15g39090 and Glyma15g39098 were detected in ‘Williams 82’ (W), ‘Ohsuzu’ (O), and ‘Tohoku 152’ (T) cultivars. The positions of the EcoRI site are indicated in (d).

(d) Diagram illustrating changes in Glyma15g39090 gene structure in ‘Tohoku 152’. ‘Tohoku 152’ contains a premature stop codon in Glyma15g39090, likely resulting from the deletion of a 6.7‐kb genomic DNA fragment between Glyma15g39090 and Glyma15g39098. The Sg‐5 cultivars, ‘Williams 82’, ‘Ohsuzu’, and ‘Enrei’, carry a premature stop codon in Glyma15g39098 (Figure S3).

Recombinant glyma15g39090 exhibits enzymatic activity on soyasapogenol B

The sg‐5 mutation has been shown to disable endogenous SA accumulation, and simultaneously enhance SB accumulation (Takada et al., 2013). Thus, it has been hypothesised that Sg‐5 encodes a C‐21 hydroxylase that catalyzes the production of SA with SB as the substrate. To test this hypothesis, we conducted an enzyme assay using a yeast, Saccharomyces cerevisiae, heterologous protein expression system (Seki et al., 2011). When SB was fed to yeast in which expression of recombinant Glyma15g39090 from ‘Williams 82’ was induced, a reaction product was detected by gas chromatography‐mass spectrometry (Figure 3a). The retention time and mass spectrum of the reaction product were identical to those of SA (Figure 3b). In contrast, when yeasts transformed with cDNA encoding Glyma15g39098 (i.e., a prematurely terminated protein) or Glyma15g39160 (i.e., the Sg‐5 homolog) were subjected to the same assay, the hydroxylated SB was not obtained (Figure 3a). Furthermore, the ‘Tohoku 152’ mutant Glyma15g39090 protein also failed to give the reaction product, although it is still possible that the Glyma15g39098, Glyma15g39160 or Glyma15g39090 proteins in ‘Tohoku 152’ were unstable or expressed at much lower levels relative to Glyma15g39090. These results suggested that recombinant Glyma15g39090 was able to use SB as a substrate to produce SA. This is consistent with the SA deficiency and SB hyperaccumulation phenotypes in the sg‐5 mutant ‘Tohoku 152’ (Takada et al., 2013); therefore, it seemed most likely that Glyma15g39090 is involved in SB C‐21 hydroxylation that is essential for SA biosynthesis.

Figure 3

Enzyme activity of recombinant Glyma15g39090.

(a) Recombinant Glyma15g39090 acts on soyasapogenol B. Recombinant yeasts induced expression of proteins derived from ‘Williams 82’ (Glyma15g39090, Glyma15g39098, and Glyma15g39160) and ‘Tohoku 152’ (Glyma15g39090) were incubated with soyasapogenol B. Reaction products were analyzed by gas chromatography‐mass spectrometry. The asterisk indicates the compound produced by ‘Williams 82’ Glyma15g39090. The chromatograms of standard soyasapogenol A and soyasapogenol B are provided for comparison.

(b) Mass spectrum of the reaction product produced by ‘Williams 82’ Glyma15g39090. The mass spectrum of standard soyasapogenol A is provided for comparison (top).

Induced Glyma15g39090 mutants exhibited simultaneous changes in the levels of group A and DDMP saponins similar to ‘Tohoku 152’

To confirm that a loss‐of‐function mutation in Glyma15g39090 is responsible for the sg‐5 phenotype, we isolated additional Glyma15g39090 mutants by screening the Targeting Local Lesions In Genome (TILLING) library developed for the Sg‐5 cultivar ‘Enrei’ (Tsuda et al., 2015). Of the 1536 randomly mutagenized soybean lines, one line designated EnT‐1376 was observed to carry a premature stop codon mutation (R44*; Figure 4a,b). An additional four lines, EnT‐1339 (S348P), EnT‐0038 (E57K), EnT‐1256 (A80T), and EnT‐0629 (A268T) were identified as amino acid substitution mutant alleles. When endogenous levels of saponins in the hypocotyls of mature seeds were analyzed by liquid chromatography‐mass spectrometry (LC‐MS), EnT‐1376 (R44*) was found to lack group A saponins, similar to ‘Tohoku 152’ (Figure 4c). Furthermore, the group A saponin level in the EnT‐1339 (S348P) amino acid substitution mutant was 4‐fold lower than that of wild‐type ‘Enrei’. Similar to the abundance of saponin αg in ‘Tohoku 152’, the endogenous levels of saponin αg were significantly increased in EnT‐1376 and EnT‐1339 relative to the levels in the wild‐type controls (Figure 4c). Finally, the EnT‐1376 mutant gene was confirmed to be allelic to sg‐5 because the sg‐5 phenotype was not restored in the F₁ seeds derived from a cross between EnT‐1376 and ‘Tohoku 152’, while it was clearly restored when ‘Enrei’ and ‘Tohoku 152’ were crossed (Figure 4d). These results indicate that a disabled Glyma15g39090 is responsible for the sg‐5 phenotype.

Figure 4

Glyma15g39090 is responsible for group A saponin biosynthesis.

(a) Mutation sites in EnT‐1376 (R44*) and EnT‐1339 (S348P). Both mutants were isolated from the induced mutant library developed for Sg‐5 cultivar ‘Enrei’ (Tsuda et al., 2015).

(b) DNA sequence electropherograms indicating nucleotide changes in EnT‐1376 and EnT‐1339. A ‘C’ to ‘T’ change resulted in the introduction of a premature stop codon (R44*) in EnT‐1376. A ‘T’ to ‘C’ change caused an amino acid substitution (S348P) in EnT‐1339.

(c) Changes in endogenous saponin levels in EnT‐1376 and EnT‐1339. Endogenous levels of group A and DDMP saponins in the hypocotyls of mature seed were analyzed by liquid chromatography‐mass spectrometry. Saponin levels in ‘Enrei’ that express the wild‐type Sg‐5 allele are provided for comparison. Asterisks indicate the significance of the changes in saponin levels in the mutants relative to those in ‘Enrei’ plants (Welch's t‐test; **: P ≤ 0.01; *: 0.01 < P ≤ 0.05). Details of sugar compositions in group A saponins (Ab, Ac, Af), deacetylated group A saponins (dAb + 1Ac, dAb + 2Ac, dAb + 3Ac), and DDMP saponins (αg, βg) are shown in Figure S1. Error bars represent the standard deviation of three biological replicates. n.d.: not detected.

(d) EnT‐1376 is allelic to sg‐5. Wild‐type ‘Enrei’ or EnT‐1376 lines were crossed with ‘Tohoku 152’, and the resulting F₁ seeds were subjected to saponin analysis. The endogenous levels of group A (Ab) and DDMP saponins (saponins αg and βg) in the hypocotyls of mature seeds were analyzed by liquid chromatography‐mass spectrometry. Saponin levels in ‘Tohoku 152’ (natural sg‐5 allele) and EnT‐1376 (R44*) are provided for comparison. Error bars represent the standard deviation of four biological replicates. n.d.: not detected.

Spatial and hormonal regulation of Sg‐5 gene expression

To determine tissue‐specific Sg‐5 expression levels, we analyzed two Sg‐5 cultivars (‘Williams 82’ and ‘Enrei’) using quantitative reverse transcription PCR (qRT‐PCR). The highest Sg‐5 expression levels were observed in the hypocotyls of developing seeds (Figure 5), which was also the tissue with the highest group A saponin levels (Figure S2). Sg‐5 expression was also observed in the roots of 2‐week‐old plants and the cotyledons of developing seeds, although the expression levels were more than 10‐fold lower than those of seed hypocotyls. It was almost undetectable in the hypocotyls, cotyledons, leaves, and stems of 2‐week‐old plants. We also compared the expression patterns of CYP72A61 and CYP93E1 that are responsible for C‐22 or C‐24 hydroxylation of basic β‐amyrin structure, respectively (Shibuya et al., 2006; Ebizuka et al., 2011), and are essential for the biosynthesis of both SA and SB. The expression levels of these genes were also the highest in the hypocotyls of developing seeds expressing Sg‐5; however, unlike Sg‐5 plants, these genes were also highly expressed in the cotyledons (Figure S2). These results indicate that Sg‐5 expression is coordinately regulated with the accumulation of group A saponins.

Figure 5

Tissue‐specific expression of g‐5.

Sg‐5 gene expression in 2‐week‐old plants (R: root, HC: hypocotyl + cotyledon, L: leaf, YLS: young leaf + stem) or developing seeds (SC: seed cotyledon, SH: seed‐hypocotyl) in ‘Enrei’ and ‘Williams 82’ was analyzed by qRT‐PCR. Expression levels of Glyma15g39098, saponin metabolic genes (i.e., Sg‐1, CYP72A61, CYP93E1, GmSGT2, GmSGT3, BAS1 [β‐amyrin synthase]), and UBC (ubiquitin conjugating enzyme) are presented for comparison. Error bars represent the standard deviation of three biological replicates.

The phytohormone jasmonate has been known to stimulate saponin biosynthesis in some plants (Suzuki et al., 2002; Mangas et al., 2006; Yendo et al., 2010); therefore, the effects of exogenously applied abscisic acid, gibberellin A₃, indole‐3‐acetic acid, methyl jasmonate (MeJA), and salicylic acid were analyzed in the roots of 7‐day‐old ‘Enrei’ seedlings to investigate the hormonal regulation of Sg‐5 expression. Sg‐5 expression levels were more than two‐fold higher in MeJA‐treated roots than in untreated roots (Figure S5). Methyl jasmonate also up‐regulated the expression of BAS1 (β‐amyrin synthase), CYP72A61 (C‐22 hydroxylase), CYP93E1 (C‐24 hydroxylase), and GmSGT2 and GmSGT3, both of which encode glycosyltransferases for soybean saponins (Shibuya et al., 2010). These results suggest that the jasmonate signaling pathway stimulates expression of saponin biosynthesis genes in soybean seedlings.

Diversity of CYP72A genes in soybean, common bean, and barrel medic

Despite the single gene nature of Sg‐5 (Figure 4; Takada et al., 2013), the soybean genome carries at least 11 copies of putative CYP72A genes that encode proteins of >400 amino acids (Glyma1.1 annotation). Of these genes, five were determined to be closely related homologs of Sg‐5 with 69–76% similarity in the amino acid sequences of the gene products (Figure S6). This implies that soybean CYP72A proteins have diverse functions despite the similarities in protein sequence. To investigate the evolution of CYP72A genes, we compared the genomes of soybean, common bean (P. vulgaris) (Schmutz et al., 2014), and barrel medic (M. truncatula) (Young et al., 2011). Unlike common bean and barrel medic, the soybean genome has undergone at least two rounds of whole genome duplication during its evolution, thereby becoming enriched in inter‐chromosomal paralogous genomic regions. We detected 29 genes that encoded canonical CYP72A proteins (>400 amino acids). When neighbour‐joining clustering was carried out using amino acid sequences for the CYP72A genes, they could be classified into three groups notated as Clades I to III (Figure 6a). Clade III CYP72A genes are conserved (85–93% sequence similarity) and each plant had only one copy of a Clade III gene. For example, the soybean CYP72A61 gene encoding a β‐amyrin C‐22 hydroxylase (Glyma08g25950) (Ebizuka et al., 2011) is in this clade. In contrast, the genes in Clades I and II are highly diverse, consisting of 26 genes and interestingly, Sg‐5 was classified into Clade I (Figure 6a). Additionally, a genome‐wide search for syntenic regions among chromosomes identified two soybean Clade I genes as the Sg‐5 paralogs, Glyma13g33690 and Glyma13g33700 (Figures 6b and S7). Chromosome 5 of common bean also consists of a genomic region that is highly paralogous to the soybean Sg‐5 region. Within this region, three Sg‐5 paralogs were classified into Clade I (Figures 6b and S7). However, unlike soybean and common bean, barrel medic has only one copy of a Clade I gene, with no paralogous relationship between this gene and Sg‐5. These results suggest that Clade I CYP72A multiplication event(s) have occurred in separate species.

Figure 6

Diversity of CYP72A genes in soybean, common bean, and barrel medic.

(a) Phylogenetic analysis using neighbor‐joining clustering of CYP72A genes in soybean (Glycine max), common bean (Phaseolus vulgaris), and barrel medic (Medicago truncatula). CYP72A genes were classified into three groups (Clades I to III). Soybean, common bean, and barrel medic CYP72A genes are indicated by orange, green, and blue squares, respectively. Asterisks indicate the Sg‐5 paralogs presented in (b).

(b) Identification of Sg‐5 paralogs. Genomic regions paralogous to the Sg‐5 region were identified on Chr.5 and Chr.13 of the common bean (upper) and soybean (lower) genomes, respectively, based on the results of a genome‐wide BLAST analysis (Figure S7). Solid lines indicate the paralogous relationship between two genes with >60% sequence similarity in coding sequences, while dashed lines indicate possible paralogous relationships according to sequence similarities and gene positions. Sg‐5 is indicated with a red box and CYP72A genes paralogous to Sg‐5 are presented in blue filled boxes. Putative CYP72A genes (i.e., pseudo‐genes) encoding truncated proteins with <400 amino acids are indicated with striped boxes.

(c) Biochemical analysis of soyasapogenols in soybean (‘Williams 82’ and ‘Enrei’), common bean (‘G19833’ and ‘BAT93’), and barrel medic (‘A17’). Endogenous soyasapogenol A and B aglycone quantities in the roots or shoots of 2‐week‐old soybean and common bean plants, or in the roots or shoots of 4‐week‐old barrel medic plants were determined following acid hydrolysis of extracted saponins. Aglycone abundance was analyzed by liquid chromatography‐mass spectrometry. Error bars represent the standard deviation of three biological replicates. n.d.: not detected.

Soyasapogenol A biosynthesis in common bean and barrel medic

To obtain insight into the evolutionary relationship between gene multiplication event(s) and the acquisition of SB C‐21 hydroxylase activity, we compared endogenous SA accumulation in two soybean cultivars (‘Williams 82’ and ‘Enrei’), two common bean accessions (‘G19833’ and ‘BAT93’), and one barrel medic accession (‘A17’). Liquid chromatography‐photodiode array tandem mass spectrometry analyses of roots or aerial parts of 2‐week‐old soybean and common bean plants, or 4‐week‐old barrel medic plants revealed the accumulation of similar amounts of SB in all plant species (Figure 6c). The SA levels were below the detection limit for common bean and barrel medic, but were detectable for soybean. These results suggest that common bean and barrel medic do not contain functional C‐21 hydroxylase genes, despite the presence of Clade I CYP72A genes in their genomes.

Discussion

Sg‐5 is responsible for the metabolic switching of undesirable group a saponins with beneficial DDMP saponins in soybean

In this study, we provide evidence that Sg‐5 (Glyma15g39090) encodes CYP72A69, which is a C‐21 hydroxylase involved in the biosynthesis of SA, and that the aglycone moiety is essential for the synthesis of all group A saponin compounds. Genetic linkage analyses combined with molecular biological analyses revealed that the natural sg‐5 mutant carries a premature stop codon (L164*) in Glyma15g39090. The sg‐5 allele might arise spontaneously from non‐homologous recombination between the functional Glyma15g39090 and disabled Glyma15g39098 (Figure 2d). Further characterization of the EnT‐1376 (R44*) and EnT‐1339 (S348P) mutants and F₁ progenies derived from a cross between EnT‐1376 and ‘Tohoku 152’ demonstrated that a loss‐of‐function mutation in Glyma15g39090 is necessary and sufficient to prevent the production of undesirable group A saponins in soybean (Figure 4). With an enzyme activity assay, we also demonstrated that Glyma15g39090 acts enzymatically on SB (Figure 3). The result of gas chromatography‐mass spectrometry of the recombinant yeast expressing Glyma15g39090 was identical to the SA standard, suggesting that this protein had an SB C‐21 hydroxylase activity. However, this result itself cannot exclude the possibility that hydroxylation could occur at other position(s) on the β‐amyrin skeleton (e.g. C‐30). Such activity may result in the production of compound(s) whose gas chromatography‐mass spectrometry profile is similar to that of the SA standard. However, this hypothesis is unlikely because it cannot thoroughly explain the sg‐5 mutant phenotype of SA deficiency and SB overaccumulation (Takada et al., 2013) that was also present in the heterologous system. Therefore, together with soybean mutant phenotype, the recombinant enzyme activity assay suggests that it is most likely that Glyma15g39090 encodes an SB C‐21 hydroxylase essential for SA biosynthesis. Although there remains the possibility that Sg‐5 also hydroxylates the C‐21 position of SB intermediates such as β‐amyrin, sophoradiol, and 24‐hydroxy‐β‐amyrin (Figure 7), our results nevertheless indicate that Sg‐5 is a predominant factor required for SA biosynthesis.

Figure 7

Sg‐5 is responsible for the metabolic switching of group A and DDMP saponins in soybean.

Disabled Sg‐5 function impairs the soyasapogenol A (SA) biosynthesis pathway, leading to increased metabolic flux in the soyasapogenol B (SB) biosynthesis pathway. The altered SA/SB balance results in a switch from the undesirable group A saponins to beneficial DDMP saponins in the sg‐5 mutants. Solid arrows indicate hydroxylation and glycosylation steps that have been experimentally substantiated. Dashed arrows indicate possible soyasapogenol biosynthetic pathways that have not been confirmed by experiments using each substrate. There still remains the possibility that Sg‐5 could catalyze C‐21 hydroxylation of β‐amyrin, sophoradiol, and 24‐hydroxy‐β‐amyrin in addition to soyasapogenol B.

We previously identified Sg‐1 UDP‐glycosyltransferase (Glyma07g38460) as a key factor responsible for the biosynthesis of acetylated group A saponins (Sayama et al., 2012). However, because Sg‐1 is not involved in metabolizing aglycones and only attaches the second sugar moiety to the C‐22 position of SA, endogenous levels of DDMP saponins are unaffected in these mutants. In contrast, we observed that sg‐5 mutations increased the endogenous DDMP saponin levels (Figure 4c), which is consistent with findings reported in Takada et al. (2013). It is likely that the inhibition of the SA biosynthesis pathway by the sg‐5 mutation leads to increased metabolic flux in the SB biosynthesis pathway, which ultimately increases the abundance of DDMP saponins (Figure 7). Interestingly, sg‐5 mutations did not affect the accumulation of saponin βg but rather, only affected that of saponin αg (Figures 1b and 4c). Clearly, increased DDMP saponin levels are attributed to increased saponin αg levels in the sg‐5 mutants. This implies that endogenous saponin βg levels are stable and saponin αg levels can fluctuate. Because saponin βg stimulates root growth in lettuce and Arabidopsis thaliana (Tsurumi et al., 2000), it is possible that saponin βg concentrations are precisely regulated during plant development. Taken together, our results demonstrate that the metabolic switching from undesirable group A saponins to beneficial DDMP saponins can be achieved with sg‐5 null mutations. This may have implications for the breeding of new soybean cultivars that can be used to increase the quality and consumer acceptance of soy‐based food products.

Evolution and functional diversification of CYP72A genes

In this study, we provide several pieces of evidence that show that the SB C‐21 hydroxylase gene evolved through neofunctionalization, during which a duplicated gene acquired a novel function. First, we showed that the soybean genome has multiple copies of CYP72A genes (Figure 6a). Considering that sg‐5 was identified as a single recessive allele (Takada et al., 2013) and that all group A saponins are eliminated by disablement of only Sg‐5 (Figures 4 and S2), there are probably differences in the activities of Sg‐5 and its closely related homologs. This notion is also supported by the soybean gene expression atlas (Severin et al., 2010), which shows that closely related Sg‐5 paralogs (i.e., Glyma13g33690, Glyma13g33700) are transcribed at levels comparable to Sg‐5 in developing seeds. Furthermore, CYP72A61, which belongs to Clade III (Figure 6a), encodes a β‐amyrin C‐22 hydroxylase (Ebizuka et al., 2011). Thus, it is likely that CYP72A gene functions have diversified at least in soybean. Second, we show that the evolution of SB C‐21 hydroxylase activity is associated with Clade I gene duplication event(s) (Figure 6a,b). Because common bean and barrel medic are able to accumulate substantial amounts of SB similar to soybean, these plants undoubtedly have functional C‐22 hydroxylases. Likewise, the genomes of common bean and barrel medic have one copy of a closely related CYP72A61 homolog (Figure 6a). Given the considerable sequence similarities (85–93%) among the Clade III genes of common bean, barrel medic, and soybean, it is possible that C‐22 hydroxylase has a fundamental role in saponin biosynthesis and plant development. However, Clade I contains several very diverse genes including soybean Sg‐5. At least one gene multiplication event probably occurred in a shared ancestor of soybean and common bean because the common bean genome contains Sg‐5 paralogs within a genomic region syntenic to the soybean Sg‐5 region (Figures 6b and S7). Despite this close relationship, the common bean Sg‐5 paralogs are unlikely to encode a functional C‐21 hydroxylase because common bean does not accumulate SA relative to SB (Figure 6c). Additionally, the barrel medic genome has only one copy of a Clade I gene, which is consistent with the fact that SA is not actively produced in this plant species (Figure 6c; Confalonieri et al., 2009). Therefore, at least one duplication event is likely to be required for the functional diversification of Clade I genes and the evolution of the SB C‐21 hydroxylase activity. As mentioned above, the soybean genome has closely related Sg‐5 paralogs on chromosome 13 (Glyma13g33690 and Glyma13g33700; Figures 6b and S7) because the soybean genome underwent at least two rounds of whole genome multiplication during its evolution (Schmutz et al., 2010; Cannon and Shoemaker, 2012). Both of these multiplication events involved these two Clade I genes (Figure 6a), but these genes are unlikely to encode a functional SB C‐21 hydroxylase because Sg‐5 is the unique gene responsible for this enzymatic activity. Rather, these Clade I genes may have a closer relationship with common bean Sg‐5 paralogs than the functional soybean Sg‐5. Future investigation into the phenotypic consequences of mutations in these genes may involve the use of the soybean TILLING mutant library. In addition, to identify the specific amino acid residue(s) that determine the reaction specificity of CYP72A members, it is also required to conduct the detailed analysis of reaction products as well as an enzyme assay using a series of amino acid substituted recombinant proteins. These studies will help characterize the diversity of CYP72A genes and clarify the evolution of the soybean cytochrome P450 gene.

Experimental procedures

Plant materials and growth conditions

Seeds of soybean cultivars ‘Williams 82’, ‘Enrei’, and ‘Ohsuzu’ were obtained from the National Institute of Agrobiological Sciences Genebank and National Agricultural Research Organization Tohoku Region Agricultural Research Center in Japan. ‘Tohoku 152’, which has the natural sg‐5 allele, was derived from the wild soybean line ‘B01082’ as previously reported (Sasama et al., 2010; Takada et al., 2013). Seeds of the common bean accessions ‘BAT93’ and ‘G19833’ were obtained from the CIAT International Center for Tropical Agriculture (http://isa.ciat.cgiar.org/urg/showstaff.do). The barrel medic accession ‘Jemalong A‐17’ seeds were provided by the USDA Agricultural Research Service (http://www.ars-grin.gov/npgs/). Plants were grown in a soil mixture composed of ‘Nippi’ (Japan Agricultural Cooperatives, Tokyo, Japan) and ‘SuperMix’ (Sakata Seed Corp., Yokohama, Japan) (v/v: 2:1) at 28°C under 16‐h light/8‐h dark conditions in air‐conditioned greenhouses. The soil was inoculated with Rhizobium and Azospirillum species (Tokachi Nokyoren, Hokkaido, Japan) to facilitate plant growth. For biochemical analysis of saponins and isolation of total RNA, harvested plant tissues were frozen in liquid nitrogen and stored at −80°C.

Genetic mapping of Sg‐5

For genetic mapping of the Sg‐5 locus, 148 BC₄F₂ and 342 BC₄F₃ individuals derived from an ‘Ohsuzu’ (Sg‐5) × ‘Tohoku 152’ (sg‐5) cross were used. Saponin phenotypes (i.e., presence or absence of group A saponins) were analyzed in the descendants of each individual. Because the Sg‐5 locus is positioned between the two SSR markers Satt117 and GMES0332 (Takada et al., 2013), new SSR markers were designed based on the genomic sequence between the two existing markers. The soybean genomic sequence was obtained from the Glyma1.1 gene set (http://www.phytozome.net/soybean). Primer pairs were designed with the SSR Candidate Marker Search Tool in the Comprehensive Phytopathogen Genomics Resource (http://cpgr.plantbiology.msu.edu/). Initially, markers were chosen based on the following criteria: total length of 100–500‐bp, SSR repeat region >20‐bp, and amplifying primers were approximately 20‐bp with melting temperatures around 60°C (using the default setting of the SSR marker search tool). The PCR was completed using the GoTaq Green Master Mix (Promega, Madison, WI, USA). The amplicons were analyzed by 10% non‐denaturing polyacrylamide gel electrophoresis and ethidium bromide staining (Hwang et al., 2009). For genotyping, multiplex PCR and fluorescence‐based sequencing were used (Oetting et al., 1995; Sayama et al., 2011). An M13 tail (5′‐CACGACGTTGTAAAACGAC‐3′) was attached to the 5′‐end of each forward primer. Additionally, an M13 primer with a PET fluorophore (Applied Biosystems, Foster City, CA, USA) attached to the 5′‐end was also used. Details regarding the primers used for Sg‐5 mapping are provided in Data S1.

Sequencing of Sg‐5 and Glyma15g39098

To obtain Sg‐5 and Glyma15g39098 cDNA sequences, 1.5‐kb DNA fragments were amplified from a cDNA library using high‐fidelity KOD DNA polymerase (Toyobo, Osaka, Japan) and the following primers: ‘common‐F1’ (5′‐ GAGAGAAATGGAAGCAGCATGG‐3′) and ‘Sg‐5‐gR1’ (5′‐CTAATGTTTTTGGATAAAGGACAGTTTTAATCT‐3′) or ‘common‐F1’ and ‘Glyma15g39098‐gR1’ (5′‐CTTGCTCTTGGTCATGTCCTCATCA‐3′). After eliminating excess dNTPs and primers with ExoSap‐IT (GE Healthcare, Little Chalfont, UK), samples were sequenced using the ABI PRISM 3500xl and 3730xl Genetic Analyzers (Applied Biosystems) according to the manufacturer's protocols. The primers used for DNA sequencing are listed in Data S2. ‘Enrei’, ‘Ohsuzu’, and ‘Tohoku 152’ cDNA sequences were obtained by aligning sequences with the ‘Williams 82’ reference sequence using CLUSTALW 2.0 (Larkin et al., 2007). To obtain ‘Tohoku 152’ Sg‐5 and Glyma15g39098 genomic sequences, the corresponding DNA fragments were amplified with PrimeSTAR GXL DNA polymerase (TaKaRa, Kyoto, Japan) and the following primers: 5′‐ATTTTTGTCGCCATGCAGCATGTATTTCAC‐3′ and 5′‐GGATAAAGGACAATTTTAATCCCTTATTGA‐3′. Amplicons were cloned into the pCR4Blunt‐TOPO vector (Life Technologies, Carlsbad, CA, USA). Sequencing reactions were completed as described above using the primers listed in Data S2.

Enzyme activity assay of glyma15g39090

To investigate enzymatic activities of the Sg‐5 candidates (Glyma15g39090, Glyma15g39098, and Glyma15g39160), cDNA fragments were amplified by PCR using the primers listed in Data S3 and cloned via the GATEWAY donor vector pDONR221 (Life Technologies, Carlsbad, CA, USA) into the pELC‐GW yeast expression vector (Seki et al., 2011). A detailed experimental procedure is also provided in Methods S1.

Isolation of induced sg‐5 mutants

The EnT‐1376 (R44*) and EnT‐1339 (S348P) soybean lines were isolated from the TILLING mutant library, which was developed using the Japanese cultivar ‘Enrei’ as previously described (Tsuda et al., 2015). For high‐resolution melting (HRM) analysis of sg‐5 mutants, 1.4‐kb or 2.4‐kb DNA fragments containing the Sg‐5 genomic region were amplified with ExTaq (TaKarRa, Kyoto, Japan) in 1536 M₂ soybean lines using the following gene‐specific primer pairs: 5′‐CCCTTTATAACCAAAACTAAAGGAGGG‐3′ and 5′‐GTAAAACAACTTGTAAGAGGATGAAGTG‐3′; or 5′‐CATATTCTAATTGTTGCTGATGGCATG‐3′ and 5′‐GCGTTGATTAATGATATAACAGCTATGC‐3′. The primary candidates were further screened by HRM analysis using the MeltDoctor HRM master mix (Life Technologies) and ViiA7 QPCR system (Applied Biosystems) with the primers listed in Data S4. Genomic DNA fragments that were targeted in the HRM analysis were amplified with ExTaq (Takara) and sequenced as described in Method S1. The mutant alleles of EnT‐1376 and EnT‐1339 consisted of a premature stop codon (R44*) or encoded an amino acid substitution (S348P), respectively. For genotyping, three to five thin slices of seed cotyledon were obtained from individual seeds, crushed with TissueLyzer (Qiagen, Venlo, Netherlands), and then genomic DNA was extracted using the BioSprint 96 DNA Plant Kit (Qiagen). Genomic DNA fragments were amplified by PCR using the following primers: (EnT‐1376) 5′‐CCCTTTATAACCAAAACTAAAGGAGGG‐3′ and 5′‐CATTTGCATCTTTAATGTCTCCTTCGTG‐3′; or (EnT‐1339) 5′‐CCTTCATTTCATGATTATCACAGGACTG‐3′ and 5′‐GCGTTGATTAATGATATAACAGCTATGC‐3′. The zygosity of the mutations was analyzed by sequencing with the following primers: (EnT‐1376) 5′‐GTCTTTTTCTTATTTCTTTTAAACCAAGTATTTCTC‐3′ or (EnT‐1339) 5′‐AGGAACACCAACTCCTGGAG‐3′.

Sequence alignments and phylogenetic tree analysis of CYP72A genes

Putative CYP72A genes (encoding >400 amino acids) were identified in the genomes of soybean (G. max; Glyma1.1) (Schmutz et al., 2010), common bean (P. vulgaris; release v1.0) (Schmutz et al., 2014), and barrel medic (M. truncatula) (Young et al., 2011), which are available in the Phytozome database (http://www.phytozome.net/). Alignments of DNA or amino acid sequences were completed using CLUSTALW version 2.1 (Larkin et al., 2007). Phylogenetic analysis was conducted using MEGA version 6.0 (Tamura et al., 2013).

Quantitative reverse transcription PCR analysis

The quantitative reverse transcription PCR (qRT‐PCR) assay was conducted as described in Methods S1. The primers and dual‐labeled probes used for qRT‐PCR are also listed in Data S5.

Identification of paralogs between chromosomes

To detect interspecies or inter‐chromosomal paralogous relationships, genome‐wide BLASTN analyses were first carried out using BLAST software version 2.2.29 +  (Camacho et al., 2009). We used default parameter settings and full soybean coding sequences as the input. The coding sequences of soybean, common bean, and barrel medic used as the database were the same versions as described above. We searched for paralogous relationships between two distinct chromosomes in a window size of 20 successive genes. Paralogs were identified when more than half of the 20 genes were homologous between two chromosomes with >60% nucleotide sequence similarity. A Perl script used for this analysis is provided in Data S6.

Other methods

Experimental procedures for biochemical analysis of saponins, genomic PCR and southern blot analysis of the Sg‐5 genomic region, enzyme activity assay for C‐21 hydroxylase activity, isolation of total RNA, and quantitative reverse transcription PCR analysis are provided in Methods S1.

Accession numbers

Nucleotide sequence data reported in this study are available in the DDBJ/EMBL/GenBank databases under the following accession numbers in the following Glycine max (L.) Merr.) cultivars: Sg‐5 cDNA in ‘Enrei’, LC143440; Sg‐5 cDNA in ‘Ohsuzu’, LC143441; sg‐5 cDNA in ‘Tohoku 152’, LC143442; Glyma15g39098 in ‘Enrei’, LC143443; Glyma15g39098 in ‘Ohsuzu’, LC143444; and Sg‐5 genomic DNA in ‘Enrei’, LC143445. The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1365-313X) is Masao Ishimoto ([e-mail]).

Acknowledgements

We thank Toshiya Muranaka (Osaka University) for encouragement; Toshio Aoki (Nihon University) for helpful discussions; and Saeko Tochigi, Takehiko Shimizu, and Kayo Machida (NIAS) for technical assistance. This work was supported by the Program for the Promotion of Basic and Applied Research for Innovations in the Bio‐oriented Industry (BRAIN) (to K.O., K.S., and M.I.); the Ministry of Agriculture, Forestry and Fisheries of Japan [the Scientific Technique Research Promotion Program for Agriculture, Forestry, Fisheries and Food Industry] (to M.I.); Genomics‐based Technology for Agricultural Improvement (to A.K., T.A., and M.I.); and the NIAS Strategic Research Fund (to A.K. and M.I). The authors declare no conflict of interest.