Transgenic Research

, Volume 21, Issue 4, pp 757–771

Coloring genetically modified soybean grains with anthocyanins by suppression of the proanthocyanidin genes ANR1 and ANR2

Authors

    • Bioproducts and Bioprocesses, Research BranchAgriculture and Agri-Food Canada
    • Department of Biology, Ottawa–Carleton Institute of BiologyCarleton University
  • Ammar Saleem
    • Department of Biology and Center for Research in Biopharmaceuticals and BiotechnologyUniversity of Ottawa
  • Tara L. Rintoul
    • Bioproducts and Bioprocesses, Research BranchAgriculture and Agri-Food Canada
  • Daniel C. W. Brown
    • Southern Crop Protection and Food Research CentreAgriculture and Agri-Food Canada
  • John T. Arnason
    • Department of Biology and Center for Research in Biopharmaceuticals and BiotechnologyUniversity of Ottawa
  • Brian Miki
    • Bioproducts and Bioprocesses, Research BranchAgriculture and Agri-Food Canada
Original Paper

DOI: 10.1007/s11248-011-9566-y

Cite this article as:
Kovinich, N., Saleem, A., Rintoul, T.L. et al. Transgenic Res (2012) 21: 757. doi:10.1007/s11248-011-9566-y

Abstract

Detection and quantification of the levels of adventitious presence of genetically modified (GM) soybeans in non-GM grain shipments currently requires sophisticated tests that can have issues with their reproducibility. We show here that pigment biosynthesis in the soybean seed coat can be manipulated to provide a distinct color that would enable the simple visible detection of the GM soybean grain. We observed that a distinct red-brown grain color could be engineered by the simultaneous suppression of two proanthocyanidin (PA) genes, ANTHOCYANIDINREDUCTASE1 (ANR1) and ANR2. Multiple reaction monitoring by liquid chromatography tandem mass spectrometry was used to quantify differentially accumulated seed coat metabolites, and revealed the redirection of metabolic flux into the anthocyanin pigment pathway and unexpectedly the flavonol-3-O-glucoside pathway. The upregulations of anthocyanin isogenes (DFR1 and GST26) and the anthocyanin/flavonol-3-O-glycosyltransferase (UGT78K2) were identified by quantitative RT-PCR to be endogenous feedback and feedforward responses to overaccumulation of upstream flavonoid intermediates resulting from ANR1 and ANR2 suppressions. These results suggested the transcription of flavonoid genes to be a key component of the mechanism responsible for the redirection of metabolite flux. This report identifies the suppression of PA genes to be a novel approach for engineering pigmentation in soybean grains.

Keywords

Metabolic engineeringVisual markerSoybean seed coat colorAnthocyaninProanthocyanidinANTHOCYANIDIN REDUCTASE

Introduction

The increase in cultivation (Lemaux 2008) and the diversification of commercial genetically modified (GM) crops for pharmaceutical (Karg and Kallio 2009) and industrial (Basaran and Rodriguez-Cerezo 2008) uses has brought added pressure for effective detection and monitoring of the adventitious presence of GM grains in non-GM materials (Demeke et al. 2006). Cases of adventitious presence are the basis for the opposition to GM/non-GM coexistence (Ramessar et al. 2010) and typically result in severe economic penalties for exporters (Demeke et al. 2006; Karg and Kallio 2009), international trade barriers (Smyth and McHughen 2008), non-science-based regulations (Krueger and Le Buanec 2008; Ramessar et al. 2010) and elevated public concern over the capability of regulatory bodies to effectively detect and monitor GM grains. Presently, the detection and quantification of the levels of adventitious presence are based on sophisticated tests (e.g. PCR, qPCR, and ELISA) (Demeke et al. 2006) that require a knowledge of molecular biology for proper interpretation of results and are not always reproducible (Morisset et al. 2009). Coloring transgenic plants, plant organs, and other transgenic materials intended for commercial markets to enable their visual identification could simplify the detection process and facilitate more effective monitoring and quantitation (Shen and Petolino 2006; Kovinich et al. 2011a).

Anthocyanins are flavonoid pigments found in plants that provide red-to-blue color to flowers, leaves, and grains. Proanthocyanidins (PAs) represent another class of flavonoid pigments that if present in plant grains provide a brown color when oxidized during desiccation. The biosynthesis of anthocyanins and PAs has been well-characterized over the past several decades, with numerous structural, regulatory, and transport genes identified (reviewed by Mol et al. 1998; Winkel-Shirley 2001; Lepiniec et al. 2006; Zhao and Dixon 2010). Although often studied separately, the biosynthesis of anthocyanin and PA pigments are intimately linked by their requirement for supply of common flavonoid precursors (Fig. 1). The anthocyanin branchpoint enzyme UDP-glycose:flavonoid-3-O-glycosyltransferase (UF3GT; EC 2.4.1.115) and the PA branchpoint enzyme anthocyanidin reductase (ANR; EC 1.3.1.77) both utilize the unstable flavonoid precursor, cyanidin, as a substrate (Fig. 1). However, engineered suppression of ANR or UF3GT has not been reported, and it remains to be determined whether the suppression of one flavonoid pathway branch would impact the other.
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig1_HTML.gif
Fig. 1

Simplified diagram of flavonoid biosynthesis in the seed coat of soybean variety Clark (irTO)

Numerous studies have reported the successful manipulation of flower color as a result of engineering anthocyanin biosynthesis (reviewed by Holton 1995; Tanaka et al. 1998; Nishihara and Nakatsuka 2011). Further, flower color has been manipulated by redirecting metabolic flux into the anthocyanin pathway by suppression of the branchpoint gene for flavonol biosynthesis (FLS) (Holton et al. 1993; Tsuda et al. 2004; Nakatsuka et al. 2007). However, successful reports of engineering novel grain color may be limited to studies that have overexpressed anthocyanin transcription factors in Maize (Shen and Petolino 2006), Arabidopsis, and Tobacco (Geekiyanage 2009). Engineering grain color by manipulating PA biosynthesis has not been reported, and PA gene knock-outs produce a transparent seed coat (testa) in the model plant Arabidopsis (Lepiniec et al. 2006) which may suggest little potential for engineering grain color by manipulating PA genes.

We previously identified the anthocyanin branchpoint gene UGT78K1 encoding UDP-glycose:flavonoid-3-O-glycosyltransferase (UF3GT) from the black soybean seed coat and demonstrated its encoded activity in vitro, and by complementation of the Arabidopsis ugt78d2 knock-out mutation (Kovinich et al. 2010). Subsequently, a second UF3GT gene UGT78K2 was identified by comparing metabolite and transcriptome data between isogenic black and brown soybean seed coats differing in alleles of the R locus (Kovinich et al. 2011b). The PA branchpoint genes ANTHOCYANIDINREDUCTASE1 (ANR1) and ANR2 were identified from the brown soybean seed coat using homology-based methods, and the activity of the encoded proteins for (−)-epicatechin synthesis were confirmed in vitro (Kovinich et al. Submitted). The same study demonstrated that soybean lines with a distinct red-brown seed coat color had reduced ANR1 gene expressions and/or activity, coupled with reduced amounts of soluble PAs and elevated amounts of anthocyanins (Kovinich et al. Submitted). The elevated anthocyanin amounts did not coincide with elevated anthocyanin branchpoint gene expressions, suggesting that a red-brown soybean grain color could be engineered simply by the suppression of the ANR1 gene.

In the present study, we found that a distinct red-brown soybean grain color could be engineered by the simultaneous suppression of the PA branchpoint genes ANR1 and ANR2. Gene suppression resulted in the increased accumulation of metabolites with expected and unexpected identities. Further genetic analysis identified the upregulation of flavonoid genes, suggesting positive feedback and feedforward mechanisms to be responsible for the increased accumulation of metabolites with unexpected identities in the soybean seed coat.

Materials and methods

Chemicals

Cyanidin was purchased from Indofine (Somerville, NJ, USA), NADPH from Fisher Scientific (Ottawa, ON, CAN), (−)-epicatechin, procyanidin B2, cyanidin 3-O-galactoside, and the 3-O-glucosides of cyanidin, quercetin and isorhamnetin were purchased from Extrasynthese (Lyon, FR). All chemicals were of the highest available purity. Solvents for metabolite extraction and LC–MS/MS were LC–MS grade (Fisher Scientific; Ottawa, ON, CAN).

Plant materials and growth conditions

Glycinemax (L.) Merr. variety Clark (PI547438) was obtained from the US Department of Agriculture Soybean Germplasm Collection (Agricultural Research Service, University of Illinois at Champaign-Urbana). Seeds were surface-sterilized as described previously (Kovinich et al. 2011b), dried, and germinated in vermiculite in a Conviron E15 cabinet with a photoperiod of 16 h light (590 μE m2 s) at 25°C, and 8 h dark at 20°C. After 12 days, seedlings were transplanted to soil and 30 days later the photoperiod was changed to 12 h light/12 h dark to encourage reproductive development. Pods were routinely harvested on ice and seed coats were dissected from pod and embryo material, immediately frozen in liquid nitrogen, lyophilized, and stored at −80°C for future analysis.

Nucleic acid isolation, vector construction, soybean transformation and PCR genotyping

RNA was isolated from soybean seed coats using a previously described method (Wang and Vodkin 1994) then cleaned using the RNeasy Plant Mini Kit (QIAGEN; Toronto, ON, CAN) to ensure high purity. The quantity and purity of RNA samples were determined by spectrophotometry using a NANODROP 2000 (Thermo Scientific; Nepean, ON, CAN). RNA samples (3 μg) were treated with amplification grade DNase I (Invitrogen; Toronto, ON, CAN) at 37°C for 15 min prior to cDNA synthesis using SuperScript III Reverse Transcriptase (Invitrogen; Toronto, ON, CAN) according to the manufacturer’s instructions.

For assembly of the ANR-RNAi construct, 254 bp of the 3′-CDS of ANR1 was amplified from seed coat cDNA isolated from the 25–50 mg fresh seed weight (FSW) stage of seed development using primers ASF/ASR (ESM Table S1). The amplicon was cloned into the pENTR/D-TOPO entry vector (Invitrogen; Toronto, ON, CAN), recombined into the pANDA35HK vector (Miki and Shimamoto 2004) as described by the suppliers using the LR clonase enzyme mix (Invitrogen; Toronto, ON, CAN), and the resultant ANR-RNAi construct was sequenced to confirm the integrity of the ANR1 hairpin. Soybean variety Clark was transformed with the ANR-RNAi construct by bombardment of embryonic suspension culture tissues as described by Finer and McMullen (1991). For genotyping, DNA was isolated and amplified using the REDExtract-N-Amp Plant PCR Kit (Sigma-Aldrich; Oakville, ON, CA) and the primers GUSF/GUSR (ESM Table S1).

qRT-PCR

qRT-PCR was performed on seed coat cDNA isolated from the 25–50 mg FSW stage using 250 nM of forward and reverse primers and the iQ SYBR Green Supermix (BioRad; Mississauga, ON, CAN) as described previously (Kovinich et al. 2011b). Parallel reactions were performed in the absence of Superscript III to test for contamination of genomic DNA. Gene expressions from each sample were normalized to the endogenous reference PEPC16. PCR cycling and qRT-PCR analysis of gene expressions and PCR efficiencies were analyzed using the Opticon Monitor 3 software (BioRad; Mississauga, ON, CAN) as described previously (Kovinich et al. 2011b). To verify the specificity of the qRT-PCR reactions, melting curves were performed subsequent to each reaction in addition to fractionation of RT-PCR products on agarose gels. Primer sequences used in this study are listed in ESM Table S1.

Metabolite analysis

Lyophilized seed coats (400–500 mg FSW) were ground to a fine powder in liquid nitrogen using an electric drill. Samples were aliquotted into 10–12 mg, and extracted with MeOH:H2O:HCO2H (80:15:5, v/v; 20 μL mg−1) for 16 h on ice with shaking (250 rpm). Subsequent to centrifugation the supernatant was filtered through 0.2 μm Teflon (Chromatographic Specialties; Brockville, ON, CAN). For compound identification, 5 μL aliquots were analyzed by LC–MS/MS in scanning mode as described previously (Kovinich et al. 2011b). For compound quantification, 2 μL aliquots were analyzed by LC–MS/MS in multiple reaction monitoring (MRM) mode in positive polarity at unit resolution. The system consisted of a 1200 series Agilent liquid chromatograph with high performance autosampler (model G1376B), binary pump (model G1312A), column thermostat (model G1316A), and PEEK 0.12 mm ID lines from column inlet onwards till tri-way splitter of a 3200 QTRAP with PEEK 0.17 mm ID × 60 cm lines on splitter outlet.

For MRM, two Q1 (parent) and Q3 (product) transitions were developed per authentic standard. The most sensitive transition was used for quantification and the second for authentication of identity. For quantitative optimization, authentic standards (10–20 μg mL−1 dissolved in methanol) were infused into the turbo ion source using an electrospary probe at a flow rate of 10 μg mL−1. Flow Injection Analysis (FIA) was applied to optimize source conditions using the same LC conditions as for sample analysis; a linear gradient of 32–70% MeOH in 5% HCO2H over 22 min at a flow rate of 1 mL min−1. The column (Luna C18 (2), 150 × 4.6 mm, 5 μm particle size, (Phenomenex Inc; Torrence, CA, USA) was washed by ramping up the MeOH % from 70 to 100% for 8 min. The initial conditions were brought back in 0.1 min and the column was equilibrated for 6 min before next injection. Samples were injected at a draw/inject speed of 200 μL min−1. The injection needle was washed in a flush port for 3 s after every injection to minimize carry over. The column thermostat was 20°C.

Detailed QTRAP settings for MRM analysis of the 6 investigated compounds are listed in ESM Table S2. Optimal signal-to-noise ratio was achieved with vertical and horizontal probe settings of 7 and 5 mm, respectively. Quantitation of the 6 target compounds was performed using Analyst software 1.5.1. The integration of peaks was performed after applying a bunching factor/smoothing of 5. The calibration curves were generated by diluting authentic standard mixes serially in methanol. Calibration curves were linear fit (R2 ≥ 0.998).

Results

Engineering red-brown soybean seed coats using the RNAi construct ANR-RNAi

An RNA interference (RNAi) construct (ANR-RNAi) designed to silence ANTHOCYANIDINREDUCTASE1 (ANR1) was transferred into brown (irTO) soybean variety Clark using a biolistics protocol (Finer and McMullen 1991). The ANR-RNAi construct consisted of 254-bp inverted repeats of the 3′-coding region of the ANR1 gene (Kovinich et al. Submitted) driven by the 35S constitutive promoter and separated by a non-functional (non-translated) fragment of GUS linker DNA sequence derived from the pANDA35HK vector (Miki and Shimamoto 2004) (Fig. 2a). Regenerated T0 plants were screened by PCR of the GUS linker sequence, and by the visual inspection for a color alteration of immature seeds dissected from pod material at various stages of development.
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig2_HTML.gif
Fig. 2

Structure of the ANR-RNAi DNA construct and phenotypes associated with its genomic integration in soybean. a Structure of the ANR-RNAi DNA construct: 35S promoter (35S); 3′-CDS of the G. maxANTHOCYANIDINREDUCTASE1 (ANR1); GUS linker (GUSLinker); transcription terminator (T); herbicide resistance gene hygromycinphosphotransferase (HPT). b, c Seed coat phenotypes at the 100–200 mg fresh seed weight stage (FSW) and the 200–300 mg FSW stage, respectively. Untransformed variety Clark (Control); T0 line A1A1 (A1A1). d PCR detection of GUS linker sequence from genomic DNA and dried grain phenotype of selected T1 lines

The three regenerated T0 lines (A1A1, A1A2, and A1A3) tested positive for the presence of the GUS linker sequence (GUS+) by PCR of genomic DNA, however only line A1A1 survived to reproductive maturity. Line A1A1 exhibited a novel red pigmentation in the hilum of immature seeds in the <25–200 mg fresh seed weight (FSW) stage of development (Fig. 2b). This phenotype was not observed in the control (untransformed Clark) (Fig. 2b), or in spontaneous or EMS-mutagenized red-brown varieties previously shown to have reduced ANR1 gene expression or enzyme activity (Kovinich et al. Submitted). The red pigment became visible over the entire seed coat surface by 200–300 FSW stage of seed development (Fig. 2c). Flowers, stems and leaves of the transgenic line did not exhibit any difference in pigmentation in comparison to the untransformed control (not shown).

Selfing line A1A1 produced T1 progeny with 3 combinations of traits: (1) GUS+ progeny with red-brown grains, (2) GUS+ with brown grains identical to the control, and (3) GUS− (tested negative for the presence of GUS by PCR) progeny with brown grains. The red-brown and brown grain color phenotypes observed in the T1 generation segregated stably with GUS linker sequences in the T2 and T3 generations, suggesting that multiple transgene insertions had been present in the parental T0 line. T2 lines representing all combinations of traits and containing single or multiple transgene insertions (as determined by inheritance of GUS and seed coat color in the T3 generation) were selected for in-depth analysis. Line A1A1-3-4 had a single transgene insertion (χ0.05, 12 (3:1, GUS+:GUS−) = 0.27) and red-brown grain color, line A1A1-1-2 had multiple transgene insertions (χ0.05, 12 (3:1, GUS+:GUS−) = 7.8) and a red-brown phenotype, line A1A1-2-2 was GUS− with a brown grain and was selected to represent the null segregating control, and line A1A1-4-2 had a multiple transgene insertions (χ0.05, 12 (3:1, GUS+:GUS−) = 5.6) and a brown grain color. The grain phenotype and the detection of GUS linker sequence from genomic DNA by PCR are shown for each selected line in Fig. 2c. GUS PCR products were sequenced to confirm their identity.

High-level suppression of ANR1 and ANR2 genes by the ANR-RNAi construct confers a red-brown grain phenotype while low-level suppression fails to modify seed coat color

To measure the level of gene suppression in transgenic lines carrying the ANR-RNAi construct, quantitative RT-PCR (qRT-PCR) of ANR genes was performed using seed coat cDNA isolated at the 25–50 mg FSW stage of seed development. The 25–50 mg FSW stage was selected as the developmental time point for this analysis because ANR1 transcripts have been shown to be at maximum levels at this stage in the seed coat of brown (irT) soybean variety Clark (Kovinich et al. 2011b). To ensure the seeds of all lines were at the same stage of development, the days post anthesis, pod length, embryo morphology, and transcript levels of the developmental marker gene Gm-r1083-1191 were ensured to be equivalent among all materials prior to testing (ESM Table S3). Comparisons of the ANR-RNAi sequence with other reductase-epimerase-dehydrogenase (RED) superfamily genes previously shown to be expressed in the seed coats of pigmented soybeans (Kovinich et al. 2011b; Submitted) found ANR-RNAi to have high identity only with ANR1 (100%) and ANR2 (86%), and moderate identity (55–47%) to other RED superfamily genes (ESM Table S4). To determine the specificity of gene silencing, other RED superfamily genes with highest identity to ANR-RNAi (i.e. LAR1 and DFR1) were included in the qRT-PCR analysis. Phosphenol pyruvate carboxylase 16 (PEPC16) was used as the endogenous reference for normalization of qRT-PCR measurements of target genes across seed coat samples. PEPC16 is expressed at similar levels in a wide range of soybean tissues and is commonly used as a reference for gene expressions during soybean seed coat development (Sugimoto et al. 1992; Tuteja et al. 2004; Kovinich et al. 2011b; Submitted).

Quantitative RT-PCR (qRT-PCR) revealed a 35.2–38.8 fold (97.1–97.4%) reduction in the levels of ANR1 transcripts, and a 3.8–7.0 fold (73.5–85.7%) reduction in ANR2 transcripts in transgenic lines with a red-brown seed coat (A1A1-1-2 and A1A1-3-4), compared to the untransformed control or the segregating line (A1A1-2-2) that lacked the ANR-RNAi construct (Fig. 3). By contrast, only a 2.1 and 1.7 fold reduction in the levels of ANR1 and ANR2 transcripts, respectively, was observed in line A1A1-4-2 (Fig. 3), that tested GUS+ but contained the control phenotype (brown seed coat color) (Fig. 2d). The magnitude of GUS linker sequence expression was inversely proportional to the level of ANR gene expression (Fig. 3), supporting that ANR-RNAi expression was the causal agent of reduced ANR1 and ANR2 transcript levels in transgenic lines.
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig3_HTML.gif
Fig. 3

qRT-PCR measurement of transcript profiles from the seed coat of transgenic soybean carrying the ANR-RNAi construct. ANTHOCYANIDINREDUCTASE1 (ANR1), ANR2, the ANR-RNAi construct GUS linker sequence (GUS), LEUCOANTHOCYANIDINREDUCTASE1 (LAR1), and DIHYDROFLAVONOL4-REDUCTASE1 (DFR1) profiles relative to the PEPC16 gene. Transgenic lines with red-brown seed coat color (greybars), lines with control seed color (whitebars). Values are mean and SD from three biological replicates. Student’s t test compared to the control significant at P < 0.01 (**), P < 0.05 (*)

LAR1 and DFR1 are RED superfamily genes expressed in the soybean seed coat that have moderate sequence similarity to ANR-RNAi (ESM Table S4). qRT-PCR demonstrated no significant decrease in the levels of LAR1 or DFR1 transcripts in transgenic lines relative to the control or the segregating GUS− line (Fig. 3), suggesting that ANR-RNAi-mediated gene suppression was highly specific to ANR1 and ANR2 genes. Interestingly, DFR1 expressions were upregulated 1.2–1.6 fold in transgenic lines with a red-brown seed coat, suggesting that an endogenous positive feedback mechanism may have been stimulated in response to engineered ANR gene suppressions.

In summary, our results show that transgenic lines with a novel red-brown seed coat color have high-level suppression of ANR1 and ANR2 genes, whereas the GUS+ transgenic line that has the control (brown) seed coat color has only low levels of ANR1 and ANR2 suppression.

Red-brown transgenic grains undergoing high-level suppression of ANR1 and ANR2 genes have major reductions in the amount of PAs along with elevated levels of cyanic pigments in the seed coat

We previously demonstrated that proanthocyanidins (PAs) from brown or black seed coats of variety Clark are exclusively (−)-epicatechin based (Kovinich et al. 2011b). As ANTHOCYANIDIN REDUCTASE1 (ANR1) and ANR2 enzymes from G. max have been shown to convert cyanidin to (−)-epicatechin (Kovinich et al. Submitted), high-level suppression of ANR1 and ANR2 expressions by the ANR-RNAi construct would be expected to cause significant reductions in the amounts of epicatechin-based PAs in the soybean seed coat.

Reaction of seed coat extracts with DMACA reagent revealed a 71.5–71.7% reduction (37.3–37.4 mg g−1 LSC) in amounts of soluble PAs in the transgenic lines that had red-brown grain color and high-level ANR gene suppressions (A1A1-1-2 and A1A1-3-4) in comparison to the untransformed or segregating null controls (Fig. 4a). By contrast, no significant change in soluble PA amounts relative to the controls was observed in the transgenic brown-colored line (A1A1-4-3) (Fig. 4a) that had only low levels of ANR gene suppression (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig4_HTML.gif
Fig. 4

Total PA and cyanic pigment amounts from the seed coat of transgenic soybean carrying the ANR-RNAi construct expressed in units mg per gram lyophilized seed coat (LSC). a Soluble PAs measured by reaction of seed coat extracts with DMACA reagent, using the PA dimer procyanidin B2 (B2) as the quantitative standard. b Solvent insoluble PAs measured by acid catalyzed hydrolysis of previously extracted seed coat residues, quantitatively compared to the authentic standard cyanidin chloride. c Cyanic pigments (anthocyanins or anthocyanidins) measured by photospectrometry in comparison to cyanidin chloride. Transgenic lines with red-brown seed coat color (greybars), lines with control seed color (whitebars). Values are mean and SD from three biological replicates. Student’s t test compared to the control significant at P < 0.01 (**)

Measurement of solvent-insoluble PAs by subjecting pre-extracted seed coat residues to acid-catalyzed hydrolysis revealed a 73.9–82.9% reduction (1.3–1.4 mg g−1 LSC) in the amounts of solvent insoluble PAs in the red-brown transgenic lines (A1A1-1-2 and A1A1-3-4) and no significant change in amounts in the transgenic line that had the brown grain phenotype (A1A1-4-3) compared to the untransformed or segregating null controls (Fig. 4b).

The red component of the red-brown grain color suggested elevated levels of cyanic pigment molecules (likely anthocyanidins or their glycosides, the anthocyanins) in the seed coat. Measurement of the total amount of cyanic pigments from seed coat extracts was performed by photospectroscopy in comparison to the authentic standard cyanidin. Red-brown transgenic lines (A1A1-1-2 and A1A1-3-4) had a 7.9–8.3 fold increase (3.2–3.4 mg g−1 LSC) in the total amounts of cyanic pigments compared to the control (Fig. 4c). By contrast, the transgenic line with brown grain color phenotypically identical to the control (A1A1-4-3) exhibited no significant change in the amount of cyanic pigments.

Overall, these results demonstrated that transgenic lines with engineered red-brown grain color undergoing high-level suppression of ANR1 and ANR2 genes had major reductions in the amounts of soluble and solvent-insoluble PAs, along with significant increases in the amounts of cyanic pigments (either anthocyanidins or anthocyanins).

High-level suppression of soybean ANR1 and ANR2 genes redirects metabolic flux from PA biosynthesis into the anthocyanin and flavonol-3-O-glucoside pathways

The enzymes anthocyanidin reductase (ANR) and UDP-glycose:flavonoid-3-O-glycosyltransferase (UF3GT) utilize cyanidin as a substrate for proanthocyanidin (PA) and anthocyanin biosynthesis, respectively (Fig. 1). As the ANR gene ANR1 and the UF3GT gene UGT78K2 were shown to be expressed at high and low levels in the brown soybean seed coat, respectively (Kovinich et al. 2011b), we hypothesized that suppression of ANR genes would redirect a moderate amount of metabolic flux from the PA pathway into the anthocyanin pathway, and would result in a phenotypic change from brown to a cyanic-colored soybean grain.

To identify metabolites that had accumulated differentially between transgenic and control seed coats, we analyzed seed coat extracts by liquid chromatography-tandem mass spectrometry (LC–MS/MS) in scanning mode. Six compounds were identified by LC–MS/MS to be present in different amounts in the transgenic red-brown seed coats compared to the untransformed control: (−)-epicatechin, procyanidin B2, cyanidin-3-O-glucoside (C3G), cyanidin-3-O-galactoside (C3Ga), quercetin-3-O-glucoside (Q3G), and isorhamnetin-3-O-glucoside (I3G) (Fig. 5a).
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig5_HTML.gif
Fig. 5

Identification and quantification of seed coat metabolites that accumulate to different amounts in transgenic lines with red-brown grain color compared to the untransformed control. a Mass spectra of (–)-epicatechin (Rt: 4.3 min), cyanidin-3-O-galactoside (Rt: 3.8 min), quercetin-3-O-glucoside (Rt: 9.5 min), procyanidin B2 (Rt: 3.0 min), cyanidin-3-O-glucoside (Rt: 4.2 min), isorhamnetin-3-O-glucoside (Rt: 12.0 min). b Multiple reaction monitoring (MRM) of the compounds listed in (a) expressed in units mg per gram lyophilized seed coat (LSC). Transgenic lines with red-brown seed coat color (greybars), lines with control seed color (whitebars). Values are means ± SD, n = 3 biological replicates. Student’s t test compared to the control significant at P < 0.01 (**)

To compare the amounts of each of these compounds between transgenic and control lines, LC–MS/MS was performed in multiple reaction monitoring (MRM) mode optimized for the detection of each of the six compounds. MRM revealed significant reductions in the amounts of (−)-epicatechin (38.8–41.5%; 9.0–9.6 mg g−1 LSC) and procyanidin B2 (52.1–62.2%; 1.9–2.3 mg g−1 LSC), along with increases in the amounts of the 3-O-glycosylated anthocyanins C3Ga (13.0–14.3 fold; 0.76–0.84 mg g−1 LSC) and C3G (2.5–2.7 fold; 0.27–0.32 mg g−1 LSC), and the flavonol-3-O-glucosides Q3G (166.0–178.6 fold; 25.9–27.9 mg g−1 LSC) and I3G (8.0–9.6 fold; 1.9–2.3 mg g−1 LSC) from red-brown transgenic seed coat extracts (Fig. 4b).

These results demonstrate that high-level suppression of soybean ANR1 and ANR2 genes redirects metabolic flux from PA biosynthesis into the anthocyanin pathway to produce a red-brown soybean seed coat. The major increase in flavonol-3-O-glucoside amounts in the transgenic seed coat was an unexpected result of engineering the suppression of ANR genes, as the branch into the flavonol-3-O-glucoside pathway precedes the biochemical step catalyzed by ANR (Fig. 1). These results suggest that metabolite differences resulting from the suppression of ANR genes may not be attributed solely to the increased accumulation of the ANR substrate, cyanidin.

High-level suppression of soybean ANR1 and ANR2 genes causes the upregulation of anthocyanin and flavonol-3-O-glucoside isogene transcripts

Metabolomes have a tendency to rebalance themselves in response to metabolic engineering events by mechanisms that are poorly understood. The overaccumulation of flavonol-3-O-glucosides in red-brown transgenic seed coats undergoing ANR gene suppression (Fig. 5b) represented one such example. However, the unexpected upregulation of DFR1 transcripts (Fig. 3) suggested that transcriptional responses are involved in the mechanism responsible for re-routing metabolic flux. To gain further insight into this mechanism, we compared the expressions of several anthocyanin and flavonol-3-O-glucoside isogenes between transgenic red-brown seed coats and the control. As the suppression of ANR may be predicted to result in the accumulation of cyanidin, (the substrate of ANR and the product of the anthocyanidin synthase (ANS) enzyme) (Fig. 1), we investigated whether ANR suppression may cause alterations in the expressions of ANS genes (ANS2/ANS3). Furthermore, cyanidin is also the substrate for UDP-glucose:flavonoid-3-O-glucosyltransferase (UF3GT), which produces anthocyanins from anthocyanidins (Fig. 1). For this reason, we also investigated whether ANR suppression may cause alterations in the expressions of UF3GT genes (UGT78K1 and UGT78K2).

qRT-PCR demonstrated 2.0–2.1 and 1.7–1.8 fold increases in the expressions of UGT78K2 and GST26, respectively, in red-brown transgenic seed coats undergoing ANR gene suppression compared to the control (Fig. 6). By contrast, no significant differences in expressions were observed for the UF3GT isogene UGT78K1, or the ANS genes ANS2/ANS3 (Fig. 6).
https://static-content.springer.com/image/art%3A10.1007%2Fs11248-011-9566-y/MediaObjects/11248_2011_9566_Fig6_HTML.gif
Fig. 6

qRT-PCR measurement of transcript profiles identified positive feedforward/feedback induction of flavonoid gene expressions in the seed coat of transgenic soybean carrying the ANR-RNAi construct. ANTHOCYANIDINSYNTHASE1/ANTHOCYANIDINSYNTHASE2 (ANS1/ANS2), UDP-GLYCOSE:FLAVONOID-3-O-GLYCSYLTRANSFERASE isogene 1 (UGT78K1), UDP-GLYCOSE:FLAVONOID-3-O-GLYCSYLTRANSFERASE isogene 2 (UGT78K2), and putative anthocyanin GLUTATHIONES-TRANSFERASE26 (GST26) profiles relative to the PEPC16 gene. Transgenic lines with red-brown seed coat color (greybars), lines with control seed color (whitebars). Values are mean and SD from three biological replicates. Student’s t test compared to the control significant at P < 0.01 (**), P < 0.05 (*)

Overall, the results from Figs. 3, 6 demonstrate the upregulation of 3 flavonoid isogenes (DFR1, UGT78K2, and GST26) from transgenic red-brown soybean seed coats undergoing ANR gene suppressions. These results identify the upregulation of a network of gene transcripts as a component of positive feedback and feedforward mechanisms responsible for the redistribution of metabolic flux in response to ANR gene suppressions in the soybean seed coat.

Discussion

Previously we proposed that soybean grain color could be engineered by the suppression of anthocyanin and/or proanthocyanidin (PA) genes in the seed coat to enable the visual detection of commercial transgenic grains (Kovinich et al. 2011a). Here we have reported the successful modification of soybean grain color (Fig. 2d). This was achieved by the simultaneous and high-level suppression of the PA genes ANTHOCYANIDINREDUCTASE1 (ANR1) and ANR2 (Fig. 3). ANR1 and ANR2 are branchpoint genes into PA biosynthesis (Fig. 1), and their high-level suppressions resulted in the reduction of total PAs (Fig. 4), and the overaccumulation of specific anthocyanins (and unexpectedly flavonol-3-O-glucosides) in the soybean seed coat (Fig. 5).

Since the first report of engineering novel flower color by the expression of the maize DFR gene in petunia (Meyer et al. 1987), there have been numerous publications describing successful flower color manipulations in various plant species (reviewed by Holton 1995; Tanaka et al. 1998; Nishihara and Nakatsuka 2011). By contrast, successful reports of engineering novel grain color remain scarce, and may have been limited to the overexpression of anthocyanin transcription factors in Maize (Shen and Petolino 2006), Arabidopsis, and Tobacco (Geekiyanage 2009). Our work demonstrated the suppression of PA branchpoint genes in soybean seed coat tissue as a novel approach for engineering pigmentation in plants. As the seed coats of other economically significant crop plants also accumulate PAs (e.g. canola (Brassicanapus L.), flax (Linumusitatissimum), and wheat (Triticum spp.) (McCallum and Walker 1990; Naczk et al. 1994; Venglat et al. 2011), it remains a possibility that ANR gene suppression could be used to color the GM grains of other important crop species.

The re-channeling of flux between branches of the flavonoid metabolism has been observed previously. The dfr mutant had increased levels of flavonols in the Arabidopsis seed (Routaboul et al. 2006). The flavonol triple regulatory mutant pfg1pfg2pfg3 and the fls1-2 knock-out mutant each had increased accumulation of anthocyanins in the seedling (Stracke et al. 2007, Stracke et al. 2009). And the M. truncatula anthocyanin transporter mutant mate2 had elevated levels of epicatechin-3′-O-glucoside in the flower (Zhao et al. 2011). Further, previous engineering experiments in soybean, Arabidopsis, and petunia have redirected metabolic flux from anthocyanin biosynthesis into the isoflavone pathway, from lignin biosynthesis into the flavonoid pathway, and from flavonol biosynthesis into the anthocyanin pathway, by suppressing anthocyanin, lignin, and flavonol branchpoint genes, respectively (Yu et al. 2003; Besseau et al. 2007; Nakatsuka et al. 2007). The overexpression of the ANR gene from Medicagotruncatula in tobacco resulted in reduced anthocyanin pigmentation in the flower and elevated PA levels (Xie et al. 2003). These results suggested the potential for ANR to compete with the anthocyanin biosynthesis enzyme UDP-glycose:flavonoid-3-O-glycosyltransferase (UF3GT) for the substrate anthocyanidin. Our results demonstrated the converse scenario; that suppression of ANR genes results in increased anthocyanin accumulations (Figs. 4, 5).

The Arabidopsis ANR (a.k.a. BAN) knockout mutant displayed precocious accumulation of cyanic pigments in the seed coat during early seed development (Albert et al. 1997). The accumulations were only temporary, and resulted in a transparent testa (tt) phenotype with black pigmentation confined to the raphe of the dried grain (Albert et al. 1997; Fig. 3). This contrasts the phenotype in soybean, where high-level suppression of ANR genes gives a red-brown grain (Fig. 2d). There may exist underlying mechanistic and metabolite differences that could explain the differences in grain phenotypes between these species. In Arabidopsis, the UF3GT gene (UGT78D2) and the ANR gene are regulated reciprocally, with UGT78D2 expressed with anthocyanins in the seedling, and ANR expressed with PAs in the seed coat (Lee et al. 2005). By contrast, soybean UF3GT genes (UGT78K1 and UGT78K2) and ANR genes (ANR1 and ANR2) are both expressed in the seed coat (Figs. 3, 6). Thus, it is possible that the difference in phenotype between the soybean grain undergoing high-level ANR gene suppressions and the Arabidopsis ANR knock-out grain, may be attributed to the presence and absence of UF3GT expressions, respectively, that stabilize anthocyanins in soybean allowing their accumulations to provide the red-brown grain phenotype. Consistent with this hypothesis, high-level suppression of ANR genes in the soybean seed coat resulted in increased amounts of the enzymatic products of UF3GT (3-O-glycosylated anthocyanins) (Fig. 5). It should be noted that the cyanic pigments that accumulated temporarily in the seed coat of the Arabidopsis ANR knockout mutant were identified by photospectrometric methods that could not distinguish between anthocyanins and the unstable anthocyanidins (Albert et al. 1997, Devic et al. 1999), and thus may not be 3-O-glycosylated anthocyanins. This hypothesis suggests that the expression of the UF3GT gene(s) may be an important consideration when designing strategies for the manipulation of grain pigmentation by the suppression of ANR genes in other species.

Feedback and feedforward mechanisms that operate during flavonoid biosynthesis remain poorly understood, and their elucidation could allow for more precise control over the outcomes of metabolic engineering of flavonoid biosynthesis. Mechanisms of feedback and feedforward regulation of flavonoid and phenylpropanoid metabolism operate at the enzymatic level or the level of gene expression. l-phenylalanine ammonia lyase was shown to be inhibited by its own product, trans-cinnamic acid (Bolwell et al. 1986; Blount et al. 2000). The regulation of the bean CHS promoter by the first two intermediates of the phenylpropanoid pathway, trans-cinnamic acid and trans-p-coumaric acid, was observed in fungal elicitor treated alfalfa protoplasts (Loake et al. 1991). The Arabidopsis flavonol synthase (FLS) knockout mutant fls1-2 demonstrated the induction of DFR and LDOX (a.k.a. ANS) genes (Stracke et al. 2009). Similarly, our results demonstrated the upregulation of DFR1, UGT78K2, and the putative anthocyanin glutathione S-transferase gene GST26 in response to ANR gene suppressions (Figs. 3, 6). These results suggest the upregulation of anthocyanin and flavonol gene transcripts to, at least in part, be responsible for the elevated levels of anthocyanins and flavonol-3-O-glucosides (Fig. 5).

Interestingly, the transgenic line (A1A1-4-2) undergoing moderate levels of ANR gene suppression (Fig. 3) did not exhibit a change in seed coat pigmentation relative to the untransformed or null segregating controls (Fig. 2). This was consistent with the previous finding that a threshold of flavonoid 3′-hydroxylase gene down-regulation by virus-induced gene silencing (VIGS) had to be exceeded to cause a change in pubescence pigmentation in soybean (Nagamatsu et al. 2009). DFR1, UGT78K2 and GST26 were not upregulated in line A1A1-4-2 which suggested that the accumulation of upstream metabolites (possibly cyanidin), rather than suppression of the gene itself, was responsible for the feedback/feedforward gene induction. The induction of DFR1, UGT78K2 and GST26 may suggest the involvement of a regulatory factor in the feedback/feedforward induction mechanism, which is consistent with flavonoid genes being coordinately regulated (Hahlbrock et al. 1976; Pelletier et al. 1997; Hartmann et al. 2005). Further work would be needed to determine how the overaccumulation of a flavonoid intermediate is perceived by the plant cell and the subsequent mechanisms that trigger the induction of flavonoid structural genes.

The adventitious presence of genetically modified (GM) grains in non-GM shipments is the basis for opposition to GM/non-GM coexistence (Ramessar et al. 2010) and has resulted in elevated public concern over the capability of government bodies to effectively detect and monitor the products of GM technologies. The use of fluorescent proteins (FPs) as transgene markers has been proposed to enable the optical detection of commercial genetically modified plant materials (Stewart 2005, 2006), however FPs do not have a history of safe use for human consumption and their detection requires specialized equipment which may not allow for convenient detection by consumers, producers, processors, and importers. An alternative approach would be to color transgenic soybeans by altering their endogenous pigment metabolism to produce distinct colors of natural products that are currently already part of the food supply and that are generally recognized as safe (Kovinich et al. 2011a). Recently discovered branchpoint genes for soybean PA biosynthesis (namely, ANTHOCYANIDINREDUCTASE1 (ANR1) and ANR2) and anthocyanin biosynthesis (UGT78K1 and UGT78K2) have been identified in addition to comparative biochemical analyses of soybean lines differing in their expressions and seed coat color (Kovinich et al. 2010, 2011b; Submitted). These advancements have enabled the design of a rationale approaches for engineering soybean grain color. Here we have demonstrated one such approach; the production of a red-brown grain by the suppression of PA branchpoint genes ANR1 and ANR2 to redirect metabolism into anthocyanin pigment biosynthesis. Coloring transgenic soybean grains intended for commercial markets to enable their visual detection would provide identification features for consumers and increased freedom to producers (Kovinich et al. 2011a). The use of recombinant DNA technology, rather than naturally occurring or EMS pigmenation mutants, offers the advantage of physically linking the DNA element for “grain-coloring” to the transgene(s) that encode the desired commercial trait. Traditionally, pigmented soybeans may have less commercial value than non-pigmented grains due to the presence of PAs that bind proteins and complicate their extraction during processing. For this reason, reducing the amount of PAs in transgenic soybean grains by the suppression of PA biosynthesis represents an advantageous approach to engineering a visible marker for industrial, pharmaceutical, or other commercial grains designed to produce proteins. As the prevalence and diversity of transgenic materials increases, the use of different color markers to distinguish materials engineered for pharmaceutical, industrial, and other non-food/feed commercial uses may become increasingly warranted. In the event that the costs of implementing color markers would be more economical than the costs associated with adventitious presence, we hope that government bodies would provide incentives that would encourage the development of GM products that incorporate the use of color markers.

Acknowledgments

We gratefully thank Dr. Vincenzo De Luca for the fruitful conversations, Dr. Ko Shimamoto for providing the pANDA35HK vector and Carla Schmidt for assistance with soybean transformation. This project was funded by the A-base (project inventory 126) from Agriculture and Agri-Food Canada to BM and DB, and NSERC Discovery Grants to BM and JTA.

Supplementary material

11248_2011_9566_MOESM1_ESM.xlsx (12 kb)
Supplementary material 1 (XLSX 12 kb)
11248_2011_9566_MOESM2_ESM.xlsx (11 kb)
Supplementary material 2 (XLSX 10 kb)
11248_2011_9566_MOESM3_ESM.xls (26 kb)
Supplementary material 3 (XLS 25 kb)
11248_2011_9566_MOESM4_ESM.xls (24 kb)
Supplementary material 4 (XLS 23 kb)

Copyright information

© Her Majesty the Queen in Right of Canada 2011