• Open Access

Functional analysis of soybean genes involved in flavonoid biosynthesis by virus-induced gene silencing



This article is corrected by:

  1. Errata: Erratum Volume 6, Issue 1, 103, Article first published online: 7 December 2007

* Correspondence (fax +81-11-706-4933; e-mail kanazawa@res.agr.hokudai.ac.jp)


Virus-induced gene silencing (VIGS) is a powerful tool for functional analysis of genes in plants. A wide-host-range VIGS vector, which was developed based on the Cucumber mosaic virus (CMV), was tested for its ability to silence endogenous genes involved in flavonoid biosynthesis in soybean. Symptomless infection was established using a pseudorecombinant virus, which enabled detection of specific changes in metabolite content by VIGS. It has been demonstrated that the yellow seed coat phenotype of various cultivated soybean lines that lack anthocyanin pigmentation is induced by natural degradation of chalcone synthase (CHS) mRNA. When soybean plants with brown seed coats were infected with a virus that contains the CHS gene sequence, the colour of the seed coats changed to yellow, which indicates that the naturally occurring RNA silencing is reproduced by VIGS. In addition, CHS VIGS consequently led to a decrease in isoflavone content in seeds. VIGS was also tested on the putative flavonoid 3′-hydroxylase (F3′H) gene in the pathway. This experiment resulted in a decrease in the content of quercetin relative to kaempferol in the upper leaves after viral infection, which suggests that the putative gene actually encodes the F3′H protein. In both experiments, a marked decrease in the target mRNA and accumulation of short interfering RNAs were detected, indicating that sequence-specific mRNA degradation was induced. The present report is a successful demonstration of the application of VIGS for genes involved in flavonoid biosynthesis in plants; the CMV-based VIGS system provides an efficient tool for functional analysis of soybean genes.


Functional genomics constitutes a general approach to understanding the functions of genes, and is one of the most important research fields in the postgenomic era. For the elucidation of gene function in plants, tools have been developed for reverse genetic engineering, such as gene targeting (Hanin et al., 2001; Terada et al., 2002; Endo et al., 2006), T-DNA (Krysan et al., 1999) and transposon tagging (Parinov et al., 1999; Speulman et al., 1999). However, these approaches have limitations with respect to practical use. Homologous recombination occurs at very low frequency in flowering plants (Puchta and Hohn, 1996; Reski, 1998; Mengiste and Paszkowski, 1999), which results in very low throughput of gene targeting. Although T-DNA and transposon tagging are powerful tools for generating novel mutants, they are based on random or non-targeted mutagenesis and, thus, the desired mutant may never be found. Another powerful approach is that of RNA silencing, which is a high-throughput tool for suppressing gene expression in a sequence-specific manner, either through RNA degradation or transcriptional repression (reviewed in Baulcombe, 2004). RNA silencing has been modified extensively and improved to facilitate investigations of gene functions (reviewed in Lu et al., 2003b; Waterhouse and Helliwell, 2003; Watson et al., 2005).

The induction of RNA silencing using a viral vector, which is termed virus-induced gene silencing (VIGS), has proved to be a very useful tool for the analysis of gene function (Lu et al., 2003b). VIGS has many advantages over the other methods, in that it is rapid and does not require plant transformation (Burch-Smith et al., 2004). Over the past decade, several RNA virus vectors have been developed for VIGS. These include Tobacco mosaic virus (Kumagai et al., 1995), Potato virus X (Ruiz et al., 1998), Tobacco rattle virus (TRV; Ratcliff et al., 1999; Liu et al., 2002a), Barley stripe mosaic virus (BSMV; Holzberg et al., 2002), Satellite tobacco mosaic virus (Gosseléet al., 2002), Pea early browning virus (PEBV; Constantin et al., 2004), Tomato mosaic tobamovirus (Hori et al., 2004), Bean pod mottle virus (BPMV; Zhang and Ghabrial, 2006), and Cucumber mosaic virus (CMV; Otagaki et al., 2006). VIGS has also been achieved using a vector based on a DNA virus, Geminivirus (Kjemtrup et al., 1998; Peele et al., 2001; Turnage et al., 2002). Recently, we have developed a vector by manipulating RNA2 of CMV, which infects a wide range of host plants. This vector can efficiently induce sequence-specific silencing of a transgene by targeting either a transcribed sequence or a promoter sequence, which results in mRNA degradation or transcriptional repression, respectively (Otagaki et al., 2006). A decrease in the mRNA level of a target gene within 6 days of viral infection indicates that this vector is suitable for rapid induction of gene silencing (Otagaki et al., 2006).

Despite the conserved nature of the biochemical pathways of secondary metabolites in plants, assignment of genes to each step of the pathway is sometimes hampered by the fact that some target genes show limited sequence homology to their counterparts in model plants, and the high sequence similarity among cognate genes within each plant species. Among these genes are those involved in the synthesis of flavonoids, which represent a large variety of related compounds. Flavonoids comprise the most common group of plant polyphenols, and have a multitude of biological functions, including the protection of plant cells against ultraviolet light and phytopathogens, signalling during nodulation, male fertility, and auxin transport (Koes et al., 2005; Treutter, 2005). Flavonoids provide much of the flavor and colour of fruits, vegetables, and flowers and, thus, they have horticultural value for humans. Flavonoids also benefit human health owing to their antioxidant and free-radical scavenging abilities when included as ingredients in foodstuffs (Ross and Kasum, 2002). A typical example of this is the presence of isoflavones in legumes, especially in soybean (Glycine max L.), the consumption of which is associated with human health benefits, such as decreases in the risk of heart disease, menopausal symptoms, and some hormone-related cancers (Yu et al., 2003).

In soybean, several mutants with altered flavonoid content have been reported (Palmer et al., 2004). However, as observed for many plants other than Arabidopsis thaliana and rice, links between most of the mutant phenotypes and gene sequences have not been established owing to a lack of efficient tools to knock down a particular gene and difficulties associated with map-based cloning of the mutated gene in the plant species. On the other hand, a comprehensive expressed sequence tag database has been accumulated for soybean (Stacey et al., 2004). Therefore, there is an increasing need for sequence-specific silencing of genes that are involved in various biological processes, including flavonoid biosynthesis in soybean.

Given these circumstances, in the present study, we tested whether our CMV-based vector induced down-regulation of the genes involved in flavonoid biosynthesis in soybean. This is the first report on engineering by VIGS of the flavonoid biosynthesis pathway in plants, and provides an efficient tool for functional analysis of genes in soybean.


Infection with a virus that contains the CHS7 insert causes loss of pigmentation in seed coat tissues

We have previously developed an RNA virus vector (CMV2-A1) by manipulating the RNA2 of CMV-Y (Otagaki et al., 2006) (Figure 1a). For CMV infection, three genomic RNA (RNA1–3) are necessary. In order to determine whether the CMV2-A1 vector mediates VIGS of the genes involved in flavonoid biosynthesis in soybean, we first targeted the chalcone synthase (CHS) genes (Figure 1b), since the silencing of CHS genes is manifested as altered seed coat colour as a result of the inhibition of anthocyanin and proanthocyanidin synthesis. In seed coats, the CHS7/CHS8 genes, which share 98% nucleotide sequence identity in the coding sequences, are predominantly expressed among the eight members of the CHS gene family in soybean (Kasai et al., 2004; Tuteja et al., 2004). Taking this into account, a 244-nucleotide fragment of exon 2 of the CHS7 gene was cloned into the CMV2-A1 vector in the antisense orientation, and the resulting construct (designated A1:CHS7) was used for infection of soybeans. For viral infection, a pseudorecombinant virus that consists of RNA components derived from different CMV strains was made because CMV-Y, the origin of all three RNA components of our original vector system, did not infect soybean. Previous analysis demonstrated that a pseudorecombinant virus containing RNA3 derived from the soybean strain of CMV (CMV-Sj) could infect soybeans (Senda et al., 2004; Hong et al., 2007; Matsuo et al., 2007). Accordingly, CMV-Y RNA1 (Y1) and CMV-Sj RNA3 (S3) were mixed with A1:CHS7, to create a pseudorecombinant virus. We found that the pseudorecombinant virus that consisted of Y1/CMV2-A1/S3 systemically infected soybeans without exhibiting any prominent phenotypic change (Figure 2a).

Figure 1.

Structure of the CMV2-A1 vector and flavonoid biosynthesis pathways targeted by virus-induced gene silencing (VIGS) in the present study. (a) Structure of the CMV2-A1 vector. Fragments of exon 2 of the CHS7 gene or exon 3 of the sf3′h1 gene were cloned into the StuI and MluI sites of the CMV2-A1 vector (Otagaki et al., 2006), and the resulting constructs were used for induction of VIGS. The open reading frames (ORF) of the 2a and 2b proteins are indicated. The hatched and grey boxes indicate the T7 RNA polymerase promoter and the cloning site, respectively. The dotted rectangle indicates the segment of the 2b ORF that was deleted from the original CMV-Y RNA2. (b) Schematic diagram of the stages of flavonoid biosynthesis. The steps catalysed by CHS or F3′H are highlighted. Enzymes involved in each step are indicated as follows: CHS, chalcone synthase; CHI, chalcone isomerase; F3H, flavanone 3-hydroxylase; F3′H, flavonoid 3′-hydroxylase; FLS, flavonol synthase.

Figure 2.

Virus-induced silencing of the CHS genes in soybean. (a) Symptomless viral infection in the vegetative stage of soybean growth. Left, a mock-inoculated plant of cultivar Fusanari; centre, a Fusanari plant infected with a control virus that lacks an insert; right, a Fusanari plant infected with the virus that carries the CHS7 insert. (b) Changes in seed coat colour resulting from infection with virus that contains the CHS7 insert. Left, seeds produced by Fusanari plants infected with a control virus that lacks an insert. Right, seeds produced by Fusanari plants infected with the virus that carries the CHS7 insert.

Plants of the soybean cultivar Fusanari, which normally produces seeds with pigmented coats, were infected with the pseudorecombinant virus that contained A1:CHS7 (CMV-A1:CHS7). No prominent outward changes appeared as a consequence of viral infection during the vegetative stage of plant growth (Figure 2a), except for the appearance of a weak temporary mosaic on the leaves at an early stage of viral infection. As a result of infection with CMV-A1:CHS7, the pigmentation of the seed coat was clearly inhibited (Figure 2b). On the other hand, the pigmented seed coat colour of plants infected with the virus that lacked the CHS7 insert (empty vector) was unchanged (Figure 2b), as was that of plants not infected with the virus (not shown). Viral infection of both leaf and seed coat tissues was confirmed by an enzyme-linked immunosorbent assay (ELISA; Figure S1 in Supplementary material). Similar results have been obtained for identical experiments performed with the cultivar Chakaori. The silenced seeds showed mottling phenotypes, to various extents (from brown to nearly complete yellow), in both plant lines, and the degree of silencing varied between these lines (Table 1). The partial pigmentation seen in the seed coats of the CMV-A1:CHS7-infected plants indicates incomplete induction of silencing in limited cells. This could be due to a spontaneous deletion of the insert from the virus vector or a low level of viral accumulation in these cells.

Table 1.  Effects of VIGS by A1:CHS7 on pigmentation in soybean seed coats
Soybean cultivarsExperimentsNo. of seeds (No. of plants)Phenotypes (%)
  • *

    Including seeds with minor stain (< 10% in area) as shown in Figure 2b.

  • Control experiments in which the virus carrying no CHS7 insert was infected.

Fusanari1128 (5) 06931
2 78 (3) 26335
3 64 (9) 03034
Control 1 25 (3)100 0 0
Control 2 42 (3)100 0 0
Chakaori1 28 (2) 1479 7
Control 1 32 (3)100 0 0

VIGS involves both reductions in the CHS mRNA levels and the production of small RNAs, and reproduces naturally occurring gene silencing

To confirm that the observed changes in seed coat colour involve VIGS of the CHS genes, changes in the mRNA levels of CHS genes after infection of the recombinant virus were analysed by gel-blot analysis of RNA extracted from the seed coats of premature developing seeds and also from leaves. A PCR-amplified DNA fragment that hybridizes to all eight CHS genes (Senda et al., 2002) was used as a probe. The mRNA levels of the CHS genes in both the seed coats and leaves were markedly decreased by infection with CMV-A1:CHS7. Quantification of the signal intensity of hybridization using 18S rRNA as an internal standard indicated that the CHS mRNA levels in the seed coats and leaf tissues of plants infected with CMV-A1:CHS7 were reduced to 12.4% and 47.0% (as an average of two repeated experiments) of the levels in control plants infected with the empty vector, respectively (Figure 3).

Figure 3.

Changes in the levels of CHS mRNA caused by virus-induced gene silencing. Total RNA samples were isolated from the leaves or seed coats of Fusanari plants infected with the virus that contains the CHS7 insert or the control virus that lacks an insert. The leaves were harvested from three plants. The seed coats were harvested from two seeds from each of three plants. CHS mRNA was detected by hybridization with the labelled CHS gene-specific probe. Ethidium bromide staining of 18S rRNA is shown as a loading control. The relative levels of CHS mRNA are calculated by dividing the CHS-specific radioactivity by the intensity of the 18S rRNA bands. The value of the empty vector-infected plants is set at 1. The results of two repeated experiments are shown.

The production of short interfering RNA (siRNA) is a hallmark of the occurrence of RNA silencing that involves sequence-specific degradation of a target mRNA (Hamilton and Baulcombe, 1999). We investigated whether siRNAs homologous to the CHS genes accumulated in the seed coats and leaves of plants infected with CMV-A1:CHS7. A previous analysis has demonstrated that in yellow seed coats of soybean, CHS gene expression is repressed by naturally occurring silencing of the CHS genes (Senda et al., 2004). In the present study, we isolated low-molecular-weight RNA fractions from the soybean cultivar Toyohomare (TH), which has yellow seed coats, and a spontaneous mutant line of Toyohomare (THM), which has pigmented seed coats. These RNAs were used as a control for siRNA detection. In the seed coats, siRNAs that hybridized to the CHS gene probe accumulated in TH but not in THM, as reported previously (Senda et al., 2004) (Figure 4). No siRNA was detected in the leaf tissues, which indicates that the occurrence of silencing is specific to the seed coats in TH.

Figure 4.

Virus-induced silencing of the CHS genes causes accumulation of CHS siRNAs in seed coat and leaf tissues. Low-molecular-weight RNAs were isolated from seed coat or leaf tissues of Fusanari plants infected with the virus that contains the CHS7 insert (A1:CHS7) or the control virus that lacks an insert (Empty vector), and hybridized with labelled CHS gene-specific sense [CHS(-) siRNA] or antisense [CHS(+) siRNA] probes. For comparison with naturally occurring CHS RNA silencing, low-molecular-weight RNA was also isolated from seed coat or leaf tissues of TH or THM lines and used for the gel-blot analysis. For the isolation of RNA, leaves and seed coats were harvested as described in the legend to Figure 3. A DNA oligonucleotide was loaded as a size control; its size is indicated to the left. The same membrane was re-hybridized with a probe specific for the 5S rRNA gene, as an internal control.

Gel-blot analyses of the low-molecular-weight RNA fraction from CMV-A1:CHS7-infected Fusanari plants showed that siRNAs that hybridized to the CHS gene probe accumulated in both the seed coats and leaf tissues (Figure 4). There were no siRNA accumulations in the empty vector-infected plants. The level of CHS siRNA in the seed coats was higher in the CMV-A1:CHS7-infected plants than in the TH line (Figure 4). These observations suggest that the loss of pigmentation in the seed coats of CMV-A1:CHS7-infected plants is the result of VIGS of the CHS genes.

Changes in isoflavone content in seeds as a consequence of VIGS of the CHS genes

We also examined whether VIGS of the CHS genes has an effect on the levels of isoflavones, which are produced in the downstream of CHS in flavonoid biosynthesis pathways and are major flavonoid compounds accumulated in soybean seeds. Changes in the levels of glucosides (daidzein and genestein) and malonylglucosides (malonyldaidzein and malonylgenistein) of isoflavones in the cotyledons of the seeds produced on soybean cultivar Fusanari after infection with CMV-A1:CHS7 were analysed by high-performance liquid chromatography (HPLC). The levels of daidzin, malonyldaidzin, genistin, and malonylgenistin in the seeds produced on the CMV-A1:CHS7-infected plants were reduced to 77.2%, 68.2%, 85.5%, and 80.8%, respectively, of those on the empty vector-infected plants (Table 2). These results indicate that VIGS of the CHS genes affected accumulation of flavonoids in embryonic tissues in addition to seed coat and leaf tissues.

Table 2.  Effects of VIGS by A1:CHS7 on isoflavone content in soybean seeds
TreatmentsIsoflavone content (mg/100 g dry weight)
  1. The data shown represent the mean ± standard deviations obtained from three replicates of the analysis. *Statistical analysis of difference between treatments was performed using the t-test.

A1:CHS727.19 ± 0.9034.99 ± 3.5136.48 ± 3.8463.01 ± 14.16161.66 ± 22.21
Empty vector35.21 ± 4.1251.27 ± 7.2542.64 ± 2.7078.03 ± 4.58207.16 ± 15.84
P* 0.030 0.025 0.085 0.155  0.045

Down-regulation of sf3′h1 gene expression

We also silenced the soybean gene sf3′h1 (DDBJ/GenBank/EMBL accession number AB061212; Toda et al., 2002). A genetic study has strongly suggested that the sf3′h1 gene encodes the flavonoid 3′-hydroxylase (F3′H) protein (Toda et al., 2002; Zabala and Vodkin, 2003), although the enzymatic activity of the sf3′h1 gene product has not been analysed. F3′H catalyses the production of quercetin from kaempferol through hydroxylation at the 3′ position of the B-ring of flavonoids (Figure 1b). Since F3′H is reported to be closely linked to the T locus, which controls pubescence colour, we used soybean near-isogenic lines for the T gene, i.e. To7B (genotype TT), in which both kaempferol and quercetin are present, and To7G (genotype tt), in which kaempferol is present but quercetin is absent (Takahashi and Asanuma, 1996). We cloned a 300-nucleotide portion of exon 3 of the gene into the CMV2-A1 vector, and the resulting construct (A1:F3′H) was used to infect To7B plants.

Infection with this virus did not produce prominent viral symptoms in the plants, as observed for the CMV-A1:CHS7-infected Fusanari plants (Figure 2a). Changes in the level of sf3′h1 mRNA in the upper leaves were analysed 3 weeks after viral infection. The mRNA level was analysed by quantitative RT-PCR, since the level of putative F3′H mRNA in soybean leaves is too low to be detected by RNA gel-blot analysis (Zabala and Vodkin, 2003). The experiment was performed using three replicates of individual plants, in which viral infection without deletion of the insert was confirmed by RT-PCR (not shown). The quantitative RT-PCR analyses indicated that sf3′h1 mRNA levels decreased in To7B plants infected with the virus that contained A1:F3′H (CMV-A1:F3′H), as compared to both mock-inoculated plants and the plants infected with the virus that contained an empty vector (Figure 5a). The average mRNA level of the sf3′h1 gene in the CMV-A1:F3′H-infected plants was 31.3% of that in the empty vector-infected plants, and was as low as the level in To7G plants. The lower mRNA level of the sf3′h1 gene in To7G plants than in To7B plants has been attributed to destabilization of the mRNA caused by a base substitution in the 3′-untranslated region and/or a base substitution that led to premature termination of translation (Toda et al., 2002), the latter of which is known as nonsense-mediated mRNA decay (reviewed by Maquat, 2004). Gel-blot analyses of the low-molecular-weight RNA fraction showed that siRNA that hybridized to the sf3′h1 gene probe accumulated in the CMV-A1:F3′H-infected To7B plants (Figure 5b), which confirms the occurrence of VIGS of the sf3′h1 gene.

Figure 5.

Changes in the levels of sf3′h1 mRNA and accumulation of siRNA as a consequence of infection with virus that contains the sf3′h1 insert. (a) Changes in the levels of sf3′h1 mRNA. The sf3′h1 mRNA level relative to the β-tubulin mRNA level was assessed by quantitative RT-PCR for the following plants: mock-inoculated To7B plants; empty vector-infected To7B plants; To7B plants infected with the virus that contains the sf3′h1 insert; and To7G plants. RNA was isolated from the leaves of virus-infected plants 3 weeks after inoculation. The sf3′h1 mRNA level relative to the β-tubulin mRNA level in one of three mock-inoculated To7B plants was assigned a value of 1. The data shown represent the mean and standard errors obtained from three replicates of the analysis. The data for three individuals for each treatment are shown. (b) Accumulation of the sf3′h1 siRNAs in leaves. Northern blot analysis was performed on low-molecular-weight RNA from the leaves of mock-inoculated To7B plants, empty vector-infected To7B plants, and To7B plants infected with the virus that contains the sf3′h1 insert. A DNA oligonucleotide was loaded as a size control; its size is indicated to the left. Ethidium bromide-stained 5S RNA and tRNA bands are shown below the panel, to demonstrate that an equal amount of the small RNA fraction was loaded. (c) VIGS-mediated changes in the content of quercetin relative to that of kaempferol. The levels of quercetin relative to those of kaempferol were quantified by LC-MS-MS using leaves of mock-inoculated To7B plants, empty vector-infected To7B plants, To7B plants infected with the virus that contains the sf3′h1 insert, and To7G plants. The data shown represent the mean and standard deviations obtained from three replicates of the analysis. The data for three individuals, for each treatment, are shown.

VIGS of the sf3′h1 gene results in changes in flavonol content

In the flavonoid synthesis pathway, F3′H catalyses three steps, i.e. naringenin to eriodictyol, dihydrokaempferol to dihydroquercetin, and kaempferol to quercetin (Figure 1b). In soybean leaves, flavonol glycosides (kaempferol and quercetin glycosides) are the most abundant flavonoids (Buttery and Buzzell, 1973; Romani et al., 2003). We compared the changes in the relative amounts of quercetin and kaempferol in leaves following VIGS of the sf3′h1 gene. The relative amounts of these compounds were quantified by liquid chromatography-tandem mass spectrometry (LC-MS-MS) analysis (Figures 5c and 6). The relative amounts of quercetin vs. kaempferol were lower in the A1:F3′H-infected To7B plants than in the empty vector-infected control plants, which indicates that down-regulation of the sf3′h1 gene affects the production of quercetin from kaempferol (Figure 5c). These results support the notion that the sf3′h1 gene encodes the F3′H protein. The level of quercetin in the To7G plants was below the detection limit of the LC-MS-MS analysis (Figure 6). This is consistent with a previous report that To7G has a single base deletion in the coding sequence of sf3′h1, which results in a nonsense mutation (Toda et al., 2002).

Figure 6.

Representative results from the LC-MS-MS analysis. Chromatograms of kaempferol (left) and quercetin (right) from the leaves of mock-inoculated To7B plants, empty vector-infected To7B plants, To7B plants infected with the virus that contains the sf3′h1 insert, and To7G plants are shown. RT, retention time (min); PAI, peak area index.


VIGS as an efficient tool for functional analysis of genes involved in flavonoid biosynthesis pathways in soybean and for genetic engineering

In the present study, we demonstrate that our VIGS system suppresses the genes involved in flavonoid synthesis in soybean, and we show that it is possible to modify flavonoid content by VIGS in plants. We first targeted the CHS genes that are highly expressed in seed coat tissues. In pigmented seed coats of soybean, the CHS7 and/or CHS8 genes are prominently expressed, and other CHS genes, such as CHS2 and CHS3, are expressed at a very low level (Kasai et al., 2004; Tuteja et al., 2004). The mRNA levels of all these genes are low in yellow seed coats (Kasai et al., 2004; Tuteja et al., 2004). The very low mRNA levels of CHS genes in seed coats after the CHS VIGS treatment (Figure 3) indicate that at least the CHS7/CHS8 genes are markedly down-regulated, just like in the yellow seed coats, in which CHS silencing is naturally induced. To date, there have been four reports of natural post-transcriptional gene silencing (PTGS) phenomena (Kusaba et al., 2003; Della Vedova et al., 2005; Koseki et al., 2005), including the yellow seed coats in soybean (Senda et al., 2004). Thus, we have directly demonstrated the natural PTGS in the soybean by reproducing the phenomenon using VIGS of CHS.

VIGS of CHS also induced a decrease in the mRNA levels of the CHS genes in leaves, although the extent of the decreases was slightly lower than those in the seed coats. There was no significant difference in the level of accumulation of CMV-A1:CHS7 in these tissues (Figure S1). One plausible explanation for the differential effects of VIGS on these tissues may be that the intrinsically higher accumulation of CHS transcripts in seed coats enhances the degradation of CHS mRNA. Alternatively, it is also possible that the limited sequence homology (79%–80%) between the A1:CHS7 and the CHS1-CHS3 genes, the transcripts of which make up approximately 40% of the total CHS transcript content of leaf tissues (Tuteja et al., 2004), results in the degradation of the CHS1-CHS3 transcripts at a much lower efficiency than the degradation of CHS7/CHS8 transcripts.

The second target in this study was the sf3′h1 gene (Toda et al., 2002), which is believed to encode the F3′H protein. The fact that silencing of the sf3′h1 gene inhibited the synthesis of quercetin from kaempferol provides strong evidence that the hypothesized gene actually encodes a protein with the F3′H activity. Although it is known that the T locus controls pubescence colour in soybean (Woodworth, 1921; Palmer et al., 2004), no change in pubescence colour was observed when the sf3′h1 gene was down-regulated by VIGS in To7B plants grown under the natural greenhouse conditions. However, we have found that loss of pigmentation in pubescence by sf3′h1 VIGS depends on growth conditions of plants, which affect steady-state mRNA level of the sf3′h1 gene (A. Nagamatsu et al., unpublished).

The allocation of a gene sequence to an enzyme involved in a metabolic pathway facilitates genetic engineering of the process, with a view to a practical purpose. For example, identification of the F3′H gene in soybean enables an increase in quercetin content via transformation of plants with the F3′H gene driven by a strong promoter. Quercetin has antioxidant activity and, thus, an increase in quercetin content increases the nutritional value for humans. An increase in quercetin content may also cause an increase in the tolerance of plants to environmental stresses, which is beneficial for crop production. In fact, To7B plants are more tolerant to chilling than To7G plants that lack quercetin (Takahashi and Asanuma, 1996), and this trait may be enhanced by genetic engineering.

In addition to analysis of gene function, VIGS can also be used to predict the consequences of transgene-mediated silencing of a particular gene over a short period of time (e.g. 2 weeks after viral inoculation) for the purpose of genetic engineering. This is particularly useful in plants, such as soybean, for which a long period of time is required to construct transgenic plants. There are several reports of transgene-mediated gene silencing in soybean (Kinney et al., 2001; Herman et al., 2003; Yu et al., 2003; Subramanian et al., 2005; Li et al., 2006; Nunes et al., 2006; Tougou et al., 2006; Lozovaya et al., 2007). Two of these reports describe the modification of flavonoid synthesis, which reflects the importance of engineering the flavonoid pathways, especially in seeds. We here demonstrate that VIGS of CHS genes resulted in reduction in isoflavone content in soybean seeds (Table 2). This could be due to indirect effects of CHS VIGS in tissues other than seeds such as leaves, since the developing seed is a sink for many products including isoflavones synthesized in the other organs (Dhaubhadel et al., 2003). Alternatively, it is also conceivable that CHS VIGS was actually induced in the embryonic tissues by the presence of CMV-A1:CHS7 and/or CHS siRNAs transmitted to the tissues, which subsequently influenced isoflavone synthesis in developing embryos. In any case, our VIGS system is useful for modifying even seed quality of soybean.

Use of the pseudorecombinant CMV vector system for VIGS in metabolic pathways of plants

VIGS experiments have been performed, in the main, using Nicotiana benthamiana, which is a model plant for studies of viruses and which exhibits a strong and persistent response to VIGS (Lu et al., 2003a). For practical applications, several viral vectors have been applied to VIGS in other plant species, including vectors based on BSMV for barley (Holzberg et al., 2002), PEBV for pea (Constantin et al., 2004), TRV for solanaceous plants (Liu et al., 2002b; Brigneti et al., 2004; Chung et al., 2004; Ryu et al., 2004; Valentine et al., 2004), BPMV for soybean (Zhang and Ghabrial, 2006), TRV for opium poppy (Hileman et al., 2005), and TRV (Valentine et al., 2004; Cai et al., 2006) or Geminivirus (Turnage et al., 2002) for A. thaliana. However, endogenous genes silenced by VIGS systems, especially those involved in metabolic pathways, have so far been confined to genes whose silencing results in visible phenotypic changes, such as photo-bleaching and/or morphological abnormalities (Kumagai et al., 1995; Ruiz et al., 1998; Gosseléet al., 2002; Holzberg et al., 2002; Liu et al., 2002a; Turnage et al., 2002; Constantin et al., 2004; Hori et al., 2004). Silencing of the sf3′h1 gene in the present study is a clear demonstration of modification of metabolic product content by VIGS without a visible phenotype of the plants being affected.

With regard to the virus vectors available for use in soybean, Zhang and Ghabrial (2006) have developed a BPMV-based vector for producing foreign proteins in soybean. They have also performed VIGS using a portion of the phytoene desaturase (PDS) gene and have reported a visible phenotypic change, although no analysis of RNA was described (Zhang and Ghabrial, 2006). Their vector induced clear symptoms on both inoculated and systemic leaves of soybean plants (Zhang and Ghabrial, 2006), which is an obstacle that needs to be overcome when evaluating the effects of VIGS of metabolic pathways. To minimize this type of viral effect, which distorts cellular gene expression, we can use pseudorecombinant viruses, as demonstrated in the present study. Different strains of CMV preferentially infect different plant species. For example, the soybean strain of CMV can infect soybean (Senda et al., 2004; Hong et al., 2007; Matsuo et al., 2007; present study), while ordinary strains, such as CMV-Y, CMV-L, or CMV-O, can infect various plant species, e.g. Petunia hybrida (Koseki et al., 2005) and A. thaliana (Matsuo et al., 2007). Viable pseudorecombinant CMV strains can be created by mixing the three viral RNAs from different virus strains. Accordingly, the use of pseudorecombinant CMV that consists of the CMV2-A1 vector and RNA1 and/or RNA3 of different strains of CMV allows optimization of its infectivity and modification of the viral host range. We have previously shown that the use of a pseudorecombinant virus is effective in reducing viral symptoms in N. benthamiana plants (Otagaki et al., 2006). We consider that this flexible feature, as well as its wide host range (> 800 species; Palukaitis et al., 1992), is advantageous to VIGS especially for the analysis of plant metabolites.

Experimental procedures

Plant materials

Cultivated soybean lines Fusanari and Chakaori were used for the analysis of CHS gene silencing. These plants produce pigmented seed coats. The cultivated soybean line Toyohomare (TH; genotype II) and its spontaneous mutant line THM (genotype ii) (Senda et al., 2004) were used as controls in the experiments. Cultivated soybean near-isogenic lines for the T locus, To7B (TT), and To7G (tt) (Takahashi and Asanuma, 1996) were used for the analysis of sf3′h1 gene silencing. The plants for CHS gene silencing and sf3′h1 gene silencing were grown in a plant growth room and in a greenhouse, respectively.

Cloning of portions of the CHS7 and sf3′h1 genes into the CMV2-A1 vector

The 3′ portion of exon 2 of the CHS7 gene was amplified by PCR using the genomic DNA of the TH line as template. The following PCR primers were used: forward 5′-AGGCCTCAAGTCCTTCACCTGTGGTT-3′ and reverse 5′-ACGCGTGAGGCTTTCAACCCATTGAAC-3′. Similarly, the 3′ portion of exon 3 of the sf3′h1 gene was amplified by PCR using the genomic DNA of To7B as template. The following PCR primers were used: forward 5′-AGGCCTGTGCTACACTCTTGGTGAAC-3′ and reverse 5′-ACGCGTGTTAGCCCATACGCTTCATC-3′. The first six nucleotides of the forward primers provide a StuI site, and those of the reverse primer provide an MluI site, which were used in the subsequent plasmid construction. After cloning of the PCR products into the pGEM-T Easy Vector (Promega, Madison, WI, USA), the StuI-MluI fragment of the plasmid that contained the CHS7 or sf3′h1 gene fragments were cloned into the StuI and MluI sites of the CMV2-A1 vector (Otagaki et al., 2006). Using these procedures, the fragments of the CHS7 and sf3′h1 genes were cloned into the vector in the antisense and sense orientations, respectively.

In vitro transcription of viral RNA

The plasmid pCY1, which contains the full-length cDNA of RNA1 of CMV-Y (Suzuki et al., 1991), and the CMV2-A1 vector (derived from RNA2 of CMV-Y; Otagaki et al., 2006) were linearized with NotI, and plasmid pCSD3, which contains the full-length cDNA of RNA3 of CMV-Sj (Hong et al., 2007), was linearized with EcoRI prior to in vitro transcription. The in vitro transcription reaction was performed using 1 µg of linearized vector DNA, 25 U T7 RNA polymerase (TaKaRa Bio, Otsu, Japan), 20 U ribonuclease inhibitor, 5 mm DTT, 1 mm ATP, 1 mm CTP, 1 mm UTP, 0.1 mm GTP, and 1 mm m7G(5′)PPP(5′)G (Invitrogen, Carlsbad, CA, USA) as a cap analogue in 40 mm Tris-HCl (pH 8.0), 8 mm MgCl2, and 2 mm spermidine-HCl, at 37 °C for 60 min.

Viral inoculation of plants

For virus propagation, the leaves of 4-week-old plants of N. benthamiana were dusted with carborundum and rub-inoculated with the in vitro-generated transcripts. For inoculation of soybean plants, the first leaves of the plants were inoculated with the sap from an infected leaf of an N. benthamiana plant. Successful infection of the N. benthamiana and soybean plants without deletion of the inserted sequences was confirmed by the conventional ELISA (Masuta et al., 1995) and RT-PCR of the viral RNA.

Isolation of total RNA and quantification of mRNA

Isolation of total RNA from seed coat and leaf tissues for the analysis of CHS silencing was performed as described previously (Senda et al., 2004). RNA gel-blot analysis and the preparation of the soybean CHS probe were performed as described previously (Senda et al., 2002). Quantification of the hybridization signals was performed as described by Senda et al. (2004). The relative levels of CHS mRNA were calculated by dividing the CHS-specific radioactivity, quantified using the Bio-Imaging Analyser BAS 1000 (Fuji Film, Tokyo, Japan), by the intensity of the 18S rRNA bands, quantified using the NIH Image version 1.63 software. Isolation of total RNA samples from leaf tissues for the analysis of sf3′h1 silencing was performed according to the method of Napoli et al. (1990), with the modification that we removed the genomic DNA from the RNA fraction by treating with DNase I (TaKaRa Bio). Poly(A) + RNA samples were purified from the total RNA using Oligotex super 30 (TaKaRa Bio). Quantitative RT-PCR was performed using poly(A) + RNA as the template, essentially as described by Koseki et al. (2005). The cDNA synthesis reaction mixture was prepared by mixing 4 µL of 5× reaction buffer [250 mm Tris-HCl (pH 8.3), 375 mm KCl, 15 mm MgCl2], 2 µL of 0.1 m DTT, 0.5 µL of RNaseOUT inhibitor (Invitrogen), 1 µL of 100 µm oligo(dT)20 primer, 4 µL of 2.5 mm dNTPs, the RNA solution, and water, to a final volume of 19 µL. After the addition of 1 µL of M-MLV reverse transcriptase (Invitrogen), cDNA synthesis was performed at 42 °C for 1 h. The reverse transcriptase was inactivated by heating the sample at 99 °C for 1 min. Quantitative RT-PCR was carried out using a 1-µL aliquot of the reaction mixture and SYBR Premix ExTaq Perfect Real Time (TaKaRa Bio) with the DNA Engine Opticon 2 System (MJ Research, Waltham, MA, USA). The PCR conditions were: 40 cycles of 94 °C for 30 s, 61 °C for 30 s, 72 °C for 1 min, and 78 °C for 2 s. Fluorescence quantification was carried out before and after the incubation at 78 °C, to monitor the formation of primer-dimers. A reaction mixture without reverse transcriptase was used as a control, to confirm that no amplification occurred from genomic DNA contamination of the RNA sample. The following primers were used for the PCR: for sf3′h1, 5′-GCAGGAACTGACACATCATC-3′ and 5′-GCCAAGTCCTCTTCTTTGAC-3′; and for β-tubulin, 5′-GAGAAGAGTATCCGGATAGG-3′ and 5′-GAGCTTGAGTGTTCGGAAAC-3′.

Isolation of low-molecular-weight RNA and detection of siRNA

The isolation of low-molecular-weight RNA and detection of CHS siRNA were performed as described by Senda et al. (2004). The detection of sf3′h1 siRNA was performed as described by Goto et al. (2003) using 60 µg of low-molecular-weight RNA isolated from trifoliate leaves. The sf3′h1 gene-specific antisense RNA probe was prepared using the DIG RNA Labeling Mix (Roche, Basel, Switzerland).

Quantification of isoflavone content by HPLC

Extraction of isoflavones was performed essentially as described by Kitamura et al. (1991). Ten milligram of cotyledons of dried seeds were ground into powder in liquid nitrogen with mortar and pestle. The powder was transferred to a microtube and mixed with 500 µL of 70% ethanol. After sonication and extraction at 4 °C for 15 min, the tube was centrifuged at 15 000 g for 10 min, and then the supernatant was centrifuged again at the same condition. The resulting supernatant was used for HPLC analysis. HPLC analysis was performed as described by Kanamaru et al. (2006) except that the following linear gradient system was applied in this study. The mobile phase consisted of two solvent systems, acetonitrile containing 0.1% acetic acid (phase A) and distilled water containing 0.1% acetic acid (phase B). The phase B was set at 80% at the initial running, then changed to 50% during the first 25 min, and then returned to 80% within the last 5 min. The absorbance at 254 nm was used to detect the separated isoflavones. Peak identification was performed using standard isoflavones (Wako Chemicals, Osaka, Japan). The levels of malonyldaidzin and malonylgenistin were quantified using daidzin and genistin as a standard, respectively. The extraction and quantification were performed in triplicate using independent seeds for each treatment.

Quantification of flavonoid content by LC-MS-MS

The samples for LC-MS-MS analysis were prepared according to the method reported previously (Porter et al., 1986). In the present study, the flavonoid levels were calculated from the total amount of flavonoids, including both free flavonoids and flavonoids conjugated with sugars. The apparatus and analytical conditions for the analysis were as follows. LC-MS-MS analyses were performed using a LCQDO liquid chromatograph-mass spectrometer (Thermo Finnigan, San Jose, CA, USA) equipped with an auto-sampler, a column oven, two pumps, a ultraviolet detector, and a degasser (Nanospace SI-2 series; Shiseido, Tokyo, Japan). The electrospray ionization source was used in the positive mode. The separation column used was a Cadenza CD-18 column, measuring 150 × 2 mm, with 3.0-mm particles (Imtakt Co., Kyoto, Japan). Column chromatography was performed using a mobile phase that consisted of MeOH–H2O–TFA (30 : 70 : 0.1, Solvent A) and MeOH–TFA (100 : 0.1, Solvent B). Gradient elution was performed by changing the combination of Solvent A and Solvent B from 80 : 20 to 10 : 90 in the period from 0 min to 11.5 min, followed by elution with A:B (10 : 90) from 11.5 min to 15 min. The flow rate of the mobile phase was 0.2 mL/min, the column temperature was 40 °C, and the amount of injection sample was 2 µL. In positive electrospray ionization, the sheath gas flow rate, aux gas flow rate, spray voltage, spray current, capillary temperature, capillary voltage, and tube lens offset were set at 70 units, 5 units, 4 kV, 0.63 µA, 200 °C, 3 V, and –10 V, respectively. Calculation of the peak area was carried out in SRM mode. The ions of m/z 287 and 303 were selected for analyses of kaempferol and quercetin, respectively, and each ion was further ionized with normalized collision energy of 25%. The peak areas of endogenous kaempferol and quercetin were calculated from the ions of m/z 287 and 303, respectively.


We are grateful to Kyohei Kanamaru and Shaodong Wang for technical support in HPLC analysis, and Tsuyoshi Inukai and Tetsuya Yamada for helpful advice. This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and a grant from Fuji Foundation for Protein Research.