Expression of the R2R3-MYB transcription factor TaMYB14 from Trifolium arvense activates proanthocyanidin biosynthesis in the legumes Trifolium repens and Medicago sativa.

Proanthocyanidins (PAs) are oligomeric flavonoids and one group of end products of the phenylpropanoid pathway. PAs have been reported to be beneficial for human and animal health and are particularly important in pastoral agricultural systems for improved animal production and reduced greenhouse gas emissions. However, the main forage legumes grown in these systems, such as Trifolium repens and Medicago sativa, do not contain any substantial amounts of PAs in leaves. We have identified from the foliar PA-accumulating legume Trifolium arvense an R2R3-MYB transcription factor, TaMYB14, and provide evidence that this transcription factor is involved in the regulation of PA biosynthesis in legumes. TaMYB14 expression is necessary and sufficient to up-regulate late steps of the phenylpropanoid pathway and to induce PA biosynthesis. RNA interference silencing of TaMYB14 resulted in almost complete cessation of PA biosynthesis in T. arvense, whereas Nicotiana tabacum, M. sativa, and T. repens plants constitutively expressing TaMYB14 synthesized and accumulated PAs in leaves up to 1.8% dry matter. Targeted liquid chromatography-multistage tandem mass spectrometry analysis identified foliar PAs up to degree of polymerization 6 in leaf extracts. Hence, genetically modified M. sativa and T. repens plants expressing TaMYB14 provide a viable option for improving animal health and mitigating the negative environmental impacts of pastoral animal production systems.

Proanthocyanidins (PAs), also known as condensed tannins, are polyphenolic plant secondary metabolites that accumulate in a wide range of plant tissues . Their function in plants is mainly associated with the ability to bind proteins and metal ions, thereby providing an effective defense against pathogens and herbivores (Gonzalez de Colmenares et al., 1998;Aziz et al., 2005). The beneficial effects of PAs found in green tea (Camellia sinensis), grape seed (Vitis vinifera), or cranberry (Vaccinium oxycoccos) on human health have been widely reported and have resulted in considerable interest in the biosynthesis and chemistry of this compound class (Ahmad et al., 2000;Bagchi et al., 2000;Dufresne and Farnworth, 2001).
PAs also play a crucial role in forage-based animal production systems with high forage legume content. The effects of PAs on grazing mammals can be detrimental when concentrations are very high (above 6%), as found in leaves of some Lotus species, leading to reduced forage palatability and digestibility (Aerts et al., 1999). On the other hand, the predominant pasture legume species Medicago sativa (lucerne, alfalfa) and Trifolium repens (white clover), which contain no or very low levels of PAs in foliar tissues (Marshall et al., 1980;Li et al., 1996), can also have negative effects on animal production. These legumes are very rich in proteins, which are rapidly fermented in the rumen, resulting in gas and foam formation, the cause of a serious condition called pasture bloat that can be lethal (McMahon et al., 2000). In the presence of PAs, excess dietary proteins as well as bacterial enzymes are complexed, and the level of protein degradation in the rumen is significantly reduced. This leads to an increase of protein bypass to the ruminant's gut and improved absorption of essential amino acids, resulting in increased milk and meat production (Aerts et al., 1999;Douglas et al., 1999;McMahon et al., 2000). Moreover, decreased protein degradation in the rumen also decreases methane production and ammonium excretion in urine, which could contribute to significant reductions in emissions of the potent greenhouse gases methane and nitrous oxide from pastures (Woodward et al., 2001;Kingston-Smith and Thomas, 2003;Beauchemin et al., 2007;Smith et al., 2008). Legumes containing moderate amounts of PAs have also been associated with reduced intestinal parasite load in ruminants (Aerts et al., 1999). The production of M. sativa and T. repens cultivars with moderate amounts of foliar PAs (2%-4%), therefore, is of considerable interest to the pastoral agricultural industry (Reed, 1995;Aerts et al., 1999;Smith et al., 2008).
PAs are derived from the phenylpropanoid/flavonoid pathway (for review, see Holton and Cornish, 1995;Winkel-Shirley, 2001;Dixon et al., 2005), sharing most precursors with flavonols and anthocyanins. Molecular genetic studies of Arabidopsis (Arabidopsis thaliana) transparent testa (tt) and tannin-deficient seed mutants have allowed the identification of a number of structural and regulatory genes (Shirley et al., 1995;Winkel-Shirley, 2001;Abrahams et al., 2003;Lepiniec et al., 2006;Fig. 1). Briefly, starting from the common precursor p-coumaric acid, the chalcone naringenin is formed by chalcone synthase (CHS). Naringenin is subsequently hydroxylated by flavanone 3-hydroxylase (F3H) to dihydrokaempferol, which can be hydroxylated at the B-ring by F39H and F3959H to dihydroquercetin and dihydromyricetin. These dihydroflavonols are either converted by flavonol synthase (FLS) to the flavonols kaempferol, quercetin, and myricetin or reduced by dihydroflavonol 4-reductase (DFR) to the corresponding leucoanthocyanidins (flavan-3,4-diols). A further reduction by leucoanthocyanidin reductase (LAR) yields the monomeric PA units, transflavan-3-ols (afzelechin, catechin, and gallocatechin). This step seems not to be involved in PA biosynthesis in Arabidopsis, as its PAs do not contain any transflavan-3-ols but appear to be made up exclusively of the cis-flavan-3-ol epicatechin. LAR was first isolated from the legume Desmodium uncinatum, which accumulates PAs based on both the trans-and cisflavan-3-ols (Tanner et al., 2003). Leucoanthocyanidins can also be converted by anthocyanidin synthase (ANS; syn. leucoanthocyanidin deoxygenase) to anthocyanidins (pelargonidin, cyanidin, and delphinidin), which are either glycosylated to the stable, colored anthocyanins or reduced by anthocyanidin reductase (ANR) to form the second set of monomeric PA units, cis-flavan-3-ols (epiafzelechin, epicatechin, and epigallocatechin). Most of these genes have also been identified in legumes (for review, see Rasmussen, 2009). Many questions about the transport into vacuoles and polymerization of these monomeric PA units remain open (Zhao et al., 2010), but recently, a UDP-Glc transferase that glycosylates epicatechin to the corresponding glucoside was identified in the model legume Medicago truncatula (Pang et al., 2008). Epicatechin 3-O-glucoside acts as a substrate for the vacuolar multidrug and toxic compound extrusion (MATE) transporters, which also transport glycosylated anthocyanidins (Marinova et al., 2007;Zhao and Dixon, 2009).
Temporal and spatial regulation of these various parts of the flavonoid pathway is complex, and a large number of transcription factors (TFs) that regulate the expression and tissue specificity of anthocyanin and PA synthesis were identified in Arabidopsis and a number of other plants. A ternary TF complex of TT2-TT8-TTG1 involving an R2R3-MYB, a basic helix-loophelix (bHLH), and a WD40 repeat protein regulates the expression of AtTT3 (DFR), AtTT18 (ANS), AtBAN (ANR), and AtTT12 (MATE) and is essential for PA biosynthesis and accumulation in Arabidopsis seed coats (Nesi et al., 2001;Baudry et al., 2004;Lepiniec et al., 2006). Several R2R3-MYB TFs have been isolated recently from grapevine (VvMYBPA1, VvMYBPA2, and VvMYB5) and shown to be essential for the expression of PA biosynthetic genes and the accumulation of PAs in grapevine (Bogs et al., 2007;Deluc et al., 2008;Terrier et al., 2009). Other species in which MYB TFs have been reported to be crucial for PA biosynthesis include Brassica napus, Diospyrios kaki (persimmon), and Populus tremuloides (Wei et al., 2007;Akagi et al., 2009;Mellway et al., 2009), indicating that MYB TFs are common regulators of PA biosynthesis. The first legume R2R3-MYB TFs reported to regulate PA biosynthetic genes are three Lotus japonicus TT2 orthologs, which, if coexpressed with AtTT8, AtTTG1, and LjANR in transient Arabidopsis leaf cell assays, activate the promoter of LjANR (Yoshida et al., 2008). Constitutive expression of LjTT2a results in the production and accumulation of PAs and anthocyanins in Arabidopsis seedlings and induces the expression of CHS, DFR, ANS, ANR, and a glutathione S-transferase predicted to be involved in PA monomer transport (Yoshida et al., 2010).
In common with the forage legume M. sativa, T. repens does not accumulate significant levels of PAs in foliar tissues; relatively high levels of PAs, however, are found in M. sativa and T. repens seed coats and clover flower petals (Goplen et al., 1980;Koupai-Abyazani et al., 1993;Foo et al., 2000). Despite the importance of foliar PAs in these forage species, no convincing strategies of engineering sufficient amounts of PAs into the leaves of these species have been reported. Attempts to induce PA biosynthesis into M. sativa leaves using ectopic coexpression of PAP1, a MYB TF responsible for regulating most of the steps in the phenylpropanoid pathway except the biosynthesis of PAs from Arabidopsis (Borevitz et al., 2000) and MtANR (Xie et al., 2006), or ectopic expression of LcMYC, a bHLH TF from Zea mays (Ray et al., 2003), have had only limited success so far. It has been suggested that the activation potential of flavonoidregulating TFs might be species specific, as induction of anthocyanin biosynthesis was not seen in M. sativa when overexpressing PAP1 from Arabidopsis but required the expression of Legume Anthocyanin Pro-duction1 (LAP1), a PAP1 ortholog from M. sativa . The recently identified MYB TF MtPAR (Verdier et al., 2012) has been shown to regulate PA biosynthesis in seed coats of wild-type M. truncatula plants and, if constitutively expressed, also induced PA accumulation in M. truncatula hairy roots and to a minor degree in M. sativa shoots (Verdier et al., 2012).
Our major goal for this study was to identify and isolate a TF regulating PA biosynthesis in foliar tissues of Trifolium species and to test if this TF would induce PA accumulation in leaves of our species of interest, T. repens. The clover genus Trifolium is one of the largest genera in the family Leguminosae, with approximately 255 species (Ellison et al., 2006). Only two Trifolium species, namely Trifolium arvense (rabbit's foot clover) and Trifolium affine (Trifolium preslianum), are known to accumulate significant levels of PAs in leaves (Fay and Dale, 1993). We used T. arvense complementary DNA (cDNA) libraries to identify R2R3-MYB TFs specific for the regulation of PA biosynthesis in legume leaves and isolated and functionally characterized the R2R3-MYB TF TaMYB14 as well as homologs from several Trifolium and Medicago species. We silenced TaMYB14 in T. arvense and determined that PA accumulation in leaves of transformed plants was almost completely blocked. Heterologous constitutive expression under the control of the cauliflower mosaic virus (CaMV) 35S promoter of TaMYB14 was used to test its effects on PA biosynthesis in Nicotiana tabacum (tobacco), M. sativa, and T. repens; PA monomers and oligomers were identified using liquid chromatography-multistage tandem mass spectrometry (LC-MS n ) analysis and compared with known grape seed PA standards. The activation of structural and transporter PA genes in T. repens transformants was analyzed using reverse transcriptase (RT)-quantitative PCR (qPCR).

RESULTS
Isolation and Identification of TaMYB14, an R2R3-MYB TF from T. arvense To isolate R2R3-MYB TFs potentially involved in PA biosynthesis in T. arvense, one of the two Trifolium species known to accumulate PAs in foliar tissues (Jones et al., 1976), we initially searched in-house (T. repens EST CS35, Pastoral Genomics NR) and public legume databases (http://compbio.dfci.harvard.edu/tgi/plant. html and http://www.kazusa.or.jp/e/resources/index. html; GenBank accession nos. CO511688-CO517334) for sequences containing the highly conserved R2R3-MYB domain (Stracke et al., 2001;Miyake et al., 2003). We designed degenerate primers (all primer sequences used in this study are listed in Supplemental Table S1) to amplify putative R2R3-MYB TFs by RT-PCR of mRNA isolated from immature and mature T. arvense leaf tissues, cloned the PCR products, and randomly selected 200 clones for sequencing. Those amplification products likely to be coding for R2R3-MYB TFs were translated and analyzed using internal National Center for Biotechnology Information TBLASTX analysis against the Database of Arabidopsis Transcription Factors (http://datf.cbi.pku.edu.cn/index.php); more than 50% of the sequences did not return any significant hits to known R2R3-MYB protein sequences. The remaining were identified as sequences containing significant similarity to R2R3-MYB TFs related to abiotic stress responses, phytohormone stimuli, stomatal movement, epidermal cell growth, meristem control, and the down-regulation of caffeic acid O-methyltransferase (data not shown).

Identification of TaMYB14 Homologs from Trifolium Species
Primers designed to the start and stop region of TaMYB14 were used to amplify full-length homologs from genomic DNA isolated from the Trifolium species T. arvense, T. affine, T. repens, and T. occidentale and the Medicago species M. sativa and M. truncatula. All of the corresponding full-length genomic MYB14 sequences had three exons and two introns of varying sizes (data not shown), except for MtMYB14-1 (JN157820) and MsMYB14-3 (JN225514), which contained the first intron only. Two PCR products were amplified from T. arvense genomic DNA, one of which corresponded to the expressed sequence TaMYB14; the other corresponded to an unidentified isoform/allelic variant of TaMYB14, designated TaMYB14_2 (JN052773); expression of this variant was not detected in any tissue of the T. arvense accession AZ2925 used in this study. Alignment of the two open reading frames and protein sequences revealed the presence of several deletions and insertions in TaMYB14_2 compared with TaMYB14 (Supplemental Fig. S1). We also identified two MYB14 isoforms/allelic variants in T. affine (JN117921, JN117922), T. occidentale (JN117926, JN117927), and M. truncatula (JN157820, JN157821), whereas M. sativa (JN120176, JN120177, JN225514) and T. repens (JN117923, JN117924, JN117925) appeared to contain three isoforms. The translated sequences of the identified R2R3-MYB14 TFs shared 95% similarity with TaMYB14, with the majority of amino acid differences being located in the 39 region downstream of the conserved MYB domain (Supplemental Fig. S1). All of the sequences contained the bHLH-binding and nineresidue consensus sequences as described above.
A phylogenetic tree was constructed with the amino acid sequences of other plant R2R3-MYB TFs involved in the regulation of PAs. Related MYB proteins from Gossypium hirsutum (Cedroni et al., 2003), whose functions have not yet been confirmed, were also included. The sequences were first aligned using the multiple sequence alignment program MAFFT (Katoh et al., 2009), and a maximum-likelihood tree was constructed using PhyML from the program SeaView version 4 (Gouy et al., 2010). The tree identifies two broad phylogenetic groups for R2R3-MYBs involved in PA biosynthesis (Fig. 3). Group 1 consists of MYB14 TFs from Trifolium and Medicago genera isolated in this study, MtPAR recently reported by Verdier et al. (2012), AtTT2 (Nesi et al., 2001), LjTT2a, -b, and-c (Yoshida et al., 2008), VvMYBPA2 (Terrier  (Mellway et al., 2009), and GhMYB10, -36, and -38 (GenBank accession nos. AAK19615, AAK19617, and AAK19618, respectively). MYB14 homologs from the Trifolium genera have the least number of substitutions between them (as represented by the very short branch lengths) and exhibit greater homology across species than what is observed for the LjTT2 MYB family within the same species. Interestingly, MtPAR, which was shown to activate PA biosynthesis in Medicago, is conserved to MtMYB14 sequences only in the MYB domain. Its C-terminal region exhibits considerable variation, and this is reflected in its long branch length in the phylogenetic tree (Fig. 3).
As a preliminary test for the putative relationship of MYB14 with PA biosynthesis and accumulation, we analyzed various tissues of wild-type T. arvense and T. repens for the presence of PAs by DMACA staining and the co-occurrence of ANR, LAR, and MYB14 transcripts in cDNA libraries isolated from the same tissues (Table I; Supplemental Fig. S3). TaMYB14 transcripts were present in immature and mature T. arvense leaves but not in callus tissues, which do not accumulate PAs (Table I). Transcripts of MYB14 orthologs were also detected in cDNA libraries from meristematic leaf tissues, trichomes, and floral tissues from T. repens and T. occidentale but not in young and mature leaf tissues, stolons, internodes, roots, and petioles (Table I). No MYB14 transcripts were detected in a red-leafed T. repens line (cv Isabelle), which accumulates high levels of anthocyanins in foliar tissues but does not accumulate foliar PAs except in trichomes. Expression of MYB14 was paralleled by the expression of the structural genes ANR and LAR (Table  I; Supplemental Fig. S3), which code for enzymes catalyzing the last steps of PA monomer synthesis ( Fig.   1). This indicates that expression of MYB14 coincides with PA biosynthesis and that PA biosynthesis is activated in T. arvense in both immature and mature tissues, whereas it is only activated in meristematic leaf and floral tissues of T. repens.

Functional Analysis of TaMYB14 in T. arvense by Gene Silencing
To test the function of TaMYB14, the gene was effectively silenced by transformation of T. arvense with an RNA interference (RNAi) construct (Helliwell and Waterhouse, 2003); silencing of TaMYB14 was confirmed by RT-PCR (data not shown). Initially, transformed callus tissues and regenerated plants were analyzed for PAs using DMACA staining. Neither wild-type nor silenced T. arvense callus tissues contained PAs; leaves from regenerated wild-type plants stained positive (Fig. 4A), whereas leaves from regenerated TaMYB14-silenced plants showed only a light blue staining in the leaves (Fig. 4E), indicating that PA biosynthesis was severely reduced in TaMYB14-silenced plants. RT-qPCR analysis of MYB14 and ANR expression in silenced plants from nine independent transformation events revealed that transcript levels in most lines were under the detection limit (Table II); MYB14 expression was detectable in one silenced line only, whereas ANR transcripts were detectable in four of the nine tested silenced lines, albeit at a more than 1,000-fold lower level than in wildtype plants (Table II).
Extracts of regenerated wild-type T. arvense plantlets were analyzed by LC-MS n to confirm the presence of PAs. For comparison purposes, a commercial grape seed extract was analyzed under the same conditions, as the PA composition of such extracts has been extensively examined by LC-mass spectrometry (MS; Hayasaka et al., 2003;Monagas et al., 2006;Weber et al., 2007). A targeted ion-trap MS n approach (Fraser et al., 2012) was used to detect and provide characterizing data on PA monomers, catechin and epicatechin (procyanidins [PC]), and gallocatechin and epigallocatechin (prodelphinidins [PD]), as well as B-type PA dimers and trimers. A comprehensive study of the fragmentation of PAs reported by Li and Deinzer (2007) detailed many of the fragmentation mechanisms observed in an ion-trap mass spectrometer for PAs up to trimers, aiding in reliable detection and classification of the PA structures. For example a PC:PC dimer MS 2 (daughter ion) spectrum ( Fig. 5D) gives the retro-diels alder fragments 409 and 427 mass-to-charge ratio (m/z) and a monomer fragment 291 m/z due to interflavanic bond cleavage. This approach of comparing MS 2 spectra from samples of known PA composition with unknown samples provides the qualitative structural evidence for confirming that coeluting peaks in different samples are in fact the target compound of interest (Fraser et al., 2012). The MS 2 spectra of known PAs from the grape seed extract were compared with coeluting MS 2 spectra in T. arvense extracts for qualitative confirmation of the PAs. Extracts of wild-type T. arvense leaves contained high levels of PA monomers (Supplemental Fig. S4), mainly catechin and, to a lesser extent, gallocatechin, as well as PC:PC (Fig. 5B), PC:PD and PD:PD dimers (data not shown), and trimers of the various PC:PD combinations (Supplemental Fig. S5). Figure 5, D and E, shows MS 2 spectral evidence that the highlighted PC:PC dimer peak in the grape seed control and the T. arvense wild-type extracts are identical. Our method also detects multiple dimers and trimers in the grape seed and experimental samples by LC-MS n ; however, without authentic standards, we are unable to assign their identities, so all polymers are referred to by the PA subunits they are constructed of (e.g. PC:PC dimers being those containing catechin and/or epicatechin). No higher degree of polymerization (DP) polymers were seen in MS 1 (parent ion) scans, but further work on concentrated acetone fractions revealed the presence of up to DP6 PAs in wild-type T. arvense leaves (data not shown). In contrast, only traces of PA monomers, but no dimers (Fig. 5C), trimers (Supplemental Fig.  S5), or larger PA polymers were detected in leaf extracts from silenced plants, and no PAs were detected using butanol-HCl extraction. These results clearly indicate that PA biosynthesis had been severely reduced by TaMYB14 silencing and that the expression of TaMYB14 is required for PA synthesis in T. arvense.
Comparison of LC-UV chromatograms (280 and 350 nm) revealed substantial differences in flavonoid and isoflavonoid levels between different genotypes of wild-type as well as silenced plants (data not shown), but there was no consistent effect observed on the levels of flavonoids between the silenced and wildtype plants, suggesting that only PA biosynthesis was reduced by TaMYB14 silencing.
Functional Analysis of TaMYB14 in Transgenic N. tabacum Plants N. tabacum is a plant species that, like Arabidopsis, accumulates PAs in seed coats only and has been used previously to test the function of PA-regulating R2R3-MYB TFs (Deluc et al., 2006(Deluc et al., , 2008. It has also been shown that constitutive expression of PA-regulating TFs in Arabidopsis can result in dwarfed phenotypes and death at early stages of development (Bogs et al., 2007). Therefore, we tested the function of TaMYB14 in the heterologous plant system tobacco (rather than Arabidopsis) by transforming plants with TaMYB14   Table I. PA accumulation (indicated by DMACA staining) and the presence of MYB14, ANR, and LAR transcripts in tissue-specific cDNA libraries from T. arvense, T. repens, and T. occidentale Transcripts were detected using RT-PCR (Supplemental Fig. S3). under the control of the CaMV 35S promoter. Integration into the genome and expression of TaMYB14 were confirmed by RT-PCR (data not shown). The effects of introducing TaMYB14 into wild-type ("green" tobacco) plants as well as into tobacco plants expressing the Solanum lycopersicum (tomato) LeANT1 TF under the control of the constitutive cassava vein mosaic promoter (Verdaguer et al., 1996;Mathews et al., 2003) on PA accumulation were tested by DMACA and LC-MS n analysis. LeANT1-expressing tobacco plants accumulate high levels of anthocyanins ("red" tobacco), and we used these plants for transformation with the CaMV 35S::TaMYB14 construct to ensure the presence of sufficient precursor flavonoids for the production of PAs. Leaves from both the green and red tobacco plants transformed with the CaMV 35S::TaMYB14 construct stained positive for PAs with DMACA, whereas untransformed tobacco plants did not (Fig. 4, B and F (Fig. 6), and trace levels of trimers (data not shown) were detected by LC-MS n analysis of flavonoid extracts from TaMYB14-expressing tobacco plants, whereas PAs were undetectable in control plants. The red TaMYB14-expressing tobacco lines did not accumulate more PA monomers and oligomers compared with the green lines; however, a loss of red color and corresponding anthocyanins in the former was noticeable (data not shown). Comparison of LC-UV chromatograms (280 and 350 nm) showed that there was no consistent effect of TaMYB14 expression on tobacco leaf flavonol levels (data not shown). These results provide strong evidence that TaMYB14 is able to up-regulate the biosynthesis of PAs, even in plants phylogenetically distant from T. arvense, and indicate that the up-regulation of PA biosynthesis competes with substrates for anthocyanin biosynthesis or accumulation.  confirmed by LC-MS n analysis, whereas PAs were undetectable in control plants. A glycosylated monomer previously identified in M. truncatula seed coats, epicatechin glycoside (Pang et al., 2008), was also detected by LC-MS n (MS 1 m/z 453, MS 2 m/z 291, MS 3 m/z 123, 139, 151, 165) in TaMYB14-transformed plants only, with levels 10-fold lower relative to free epicatechin (data not shown). There was no consistent effect of TaMYB14 expression on leaf flavonoid levels (data not shown).
Quantification of soluble PAs using the butanol-HCl method (Terrill et al., 1992) in leaves of eight independent M. sativa lines constitutively expressing TaMYB14 showed an accumulation of PAs in six of the tested lines, ranging from 0.1% to 1.0% dry mass (Supplemental Table S2).

Functional Analysis of TaMYB14 in Transgenic T. repens Plants
To test the function of TaMYB14 in T. repens, a close relative of T. arvense and widely used in pastoral mixed grazing systems, plants were transformed with TaMYB14 under the control of the CaMV 35S promoter; the presence and expression of the transgene were confirmed by RT-PCR (data not shown). Leaves from regenerated plantlets were screened for PA accumulation using DMACA staining, and a number of plants transformed with TaMYB14 tested positive. Leaves from nontransformed wild-type plants showed, as expected, positive DMACA staining in the trichomes on the abaxial leaf layers only (Fig. 4D), whereas plants transformed with TaMYB14 accumulated PAs in epidermal leaf cells as well as in trichomes (Fig. 4H). Microscopic examination under higher magnification revealed that stomatal guard cells as well as mesophyll cells stained positive with DMACA (Fig. 4, H and L). Staining in trichome tier and apical cells was much stronger compared with wild-type trichomes (Fig. 4, J and K), and staining of spongy mesophyll cells indicated PA accumulation in multiple vacuole-like organelles (Fig. 4L). In comparison, wild-type T. arvense plants accumulate PAs in epidermal leaf cells only as shown by DMACA staining of a young leaf crosssection (Supplemental Fig. S9). In some transgenic T. repens plants, PAs were also detected in root and petiole cells (data not shown). A number of plants appeared to accumulate very high levels of foliar PAs, indicated by almost black staining with DMACA (Fig.  4I); however, all of these plants died before sufficient material for LC-MS 2 analyses could be harvested. These results indicate that the constitutive expression of TaMYB14 results in the ectopic production and accumulation of PAs in foliar (and other) tissues and cell layers of T. repens that normally do not accumulate PAs.
To confirm the presence of PA units in T. repens leaves expressing TaMYB14, leaf extracts were analyzed by LC-MS n . No PA units were detected in wild-type T. repens leaves, whereas PA monomers (Supplemental Fig. S10), dimers (Fig. 8), as well as trace levels of trimers (Supplemental Fig. S11) were detected in T. repens plants transformed with the CaMV 35S::  TaMYB14 construct. The major monomer detected was epigallocatechin, with 10-fold lower levels of epicatechin and gallocatechin and only traces of catechin detected (Supplemental Fig. S10). The dimers and trimers detected were PC and PD or all PD based, consistent with the observation of both PC and PD monomers in transformed T. repens leaves. As seen for T. arvense, there were considerable differences in flavonol and isoflavonoid levels in different T. repens genotypes; however, there was no consistent effect of TaMYB14 expression on leaf flavonol and isoflavonoid levels ( Supplemental Fig. S12). These results provide evidence for TaMYB14 being able to specifically activate the biosynthesis and accumulation of PAs in white clover foliar tissues.
Quantification of soluble PAs in leaves of 16 independent T. repens lines constitutively expressing TaMYB14 showed an accumulation of PAs ranging from 0.5% to 1.8% dry mass. We selected five of these lines for further analysis of transcript levels of phenylpropanoid biosynthetic and transporter genes (Fig. 9).

Analysis of the Expression of Phenylpropanoid Pathway Genes in T. repens Plants Expressing TaMYB14
To analyze the effects of TaMYB14 on the expression of genes coding for enzymes of the phenylpropanoid pathway and a putative MATE-like PA transporter, RNA was isolated from leaves of T. repens plants constitutively expressing TaMYB14. Nucleotide sequences coding for putative TrCHS, TrF39H, TrF3959H, TrFLS, TrDFR, TrANS, TrANR, TrLAR, TrMATE1, and TrMATE2 were identified from our in-house Trifolium gene databases, and mRNA levels were quantified in five independent lines transformed with TaMYB14 using RT-qPCR with gene-specific primers ( Fig. 9; for primer sequences, see Supplemental Table S1).
No consistent differences in the expression of TrCHS, TrF3H, TrFLS, and TrMATE2 in transgenic TaMYB14-expressing compared with wild-type T. repens leaves were detected, although CHS transcript levels were higher (approximately 4-fold) in three of the five tested lines. Expression of TrF3959H, TrDFR, and TrANS was consistently up-regulated in transformed plants compared with wild-type plants, with a very strong induction of TrF3959H (more than 600fold). Expression of the putative PA-specific genes TrANR and TrLAR could only be detected in leaves from plants constitutively expressing TaMYB14 but not in wild-type plants. We also analyzed the expression of two genes putatively coding for MATE transporters and show here that TrMATE1 was not expressed in wild-type plants but was highly expressed in T. repens plants transformed with CaMV 35S::TaMYB14, whereas TrMATE2 was expressed in wild-type plants and not up-regulated in the transgenic lines. TaMYB14 expression varied considerably in the transgenic lines and was highest in the line with the highest PA levels (Fig. 9). Our results demonstrate that expression of the R2R3-MYB TF TaMYB14 is necessary and sufficient to upregulate genes of the PA pathway and to activate the synthesis and accumulation of PAs in leaves of T. repens. All transgenic plants were grown in a greenhouse under the same conditions. However, one previously greenleafed line that accumulated PAs to levels comparable with the other transgenic lines (indicated by DMACA staining; data not shown) changed its phenotype after 6 months at the onset of winter and subsequently exhibited a red-leafed phenotype. At this stage, high levels of anthocyanins but only trace amounts of PAs were detectable. The three anthocyanins detected were shown by LC-MS n to be pelargonidin, cyanidin, and delphinidin glycosides (Supplemental Fig. S13). Expression analysis (Supplemental Table S3) revealed that all of the putative phenylpropanoid pathway genes related to the production of anthocyanins (i.e. CHS, F3H, DFR, ANS, and F3959H) as well as both MATE transporters were up-regulated, with expression levels for DFR, ANS, and F3959H in this line being much higher (more than 200-fold) compared with the PA-accumulating greenleafed T. repens lines. No expression of TrANR and TaMYB14 could be detected in this line, whereas TrLAR expression was much lower compared with the greenleafed TaMYB14 transformants.

TaMYB14 Is an R2R3-MYB TF Regulating PA Biosynthesis in Legumes
Only a few out of the 255 known Trifolium species accumulate PAs in leaves, one of which is T. arvense (Jones et al., 1976). Therefore, we used leaf tissues of T. arvense to isolate the R2R3-MYB14 TF regulating foliar PA biosynthesis. Several lines of evidence are presented here that indicate that TaMYB14 is in fact regulating PA biosynthesis: (1) expression of TaMYB14 coincides with ANR and LAR expression and PA accumulation in T. arvense tissues; (2) silencing of TaMYB14 in T. arvense almost completely blocked foliar PA biosynthesis; and (3) constitutive expression of TaMYB14 in N. tabacum, M. sativa, and T. repens resulted in ectopic accumulation of PAs in leaf tissues of these plants as well as up-regulation of PA biosynthetic pathway genes in T. repens transformants.
Silenced TaMYB14 T. arvense plants still showed very weak DMACA staining and traces of PA monomers, whereas PA dimers and trimers were completely absent. The synthesis of traces of PA monomers might have been the result of very low residual expression of TaMYB14, although we were not able to detect any TaMYB14 transcripts in these plants. It is also possible that T. arvense possesses one or more additional PA-regulating R2R3-MYB TFs, as has been described for grapevine and poplar (Bogs et al., 2007;Deluc et al., 2008;Terrier et al., 2009;Wilkins et al., 2009). The TaMYB14 ortholog TaMYB14_2 identified in this study was not expressed in any of the tissues tested, and it is not clear if this gene is in fact coding for a functional variant of TaMYB14 in T. arvense.
R2R3-MYB TFs have been described to regulate PA biosynthesis, often in a complex with bHLH and WD40 proteins (Baudry et al., 2004;Yoshida et al., 2008Yoshida et al., , 2010Hichri et al., 2011), in a wide range of plants. In our study, ectopic expression of TaMYB14 under the control of the CaMV 35S promoter resulted in PA accumulation in leaves of N. tabacum, M. sativa, and T. repens without the need for coexpression of additional factors. Transcript analysis of T. repens transformants showed expression of the putative PAspecific genes TrANR, TrLAR, and TrMATE1, which were completely silent in wild-type T. repens leaves, and a more than 600-fold up-regulation of TrF3959H. The expression profile of TrMATE1 would be consistent with it being the ortholog of the PA-specific MATE transporter TT12 Marinova et al., 2007), whereas TrMATE2, which was expressed in both wild-type and transformed plants, seems to be an ortholog of the anthocyanin/flavonol MtMATE2 transporter from M. truncatula (Zhao et al., 2011). Genes required for precursors of both the anthocyanin and PA pathways (i.e. DFR and ANS) were also up-regulated, Figure 9. Analysis of phenylpropanoid and transporter gene expression and PA accumulation in T. repens leaf tissue. Transcript and extractable PA levels in wild-type (WT) and TaMYB14expressing T. repens lines are shown. Transcript levels were normalized relative to the transcript levels of T. repens actin. PAs were extracted and quantified using the butanol-HCl method. Three technical replicates were analyzed for each line. Error bars represent SD. DW, Dry weight. but to a lesser extent, whereas early flavonoid pathway genes (i.e. CHS, F3H, and FLS) were not consistently upregulated. In comparison, overexpression of the recently identified TF MtPAR in M. truncatula hairy roots did not induce the expression of CHS, FLS, DFR, and LAR but did result in higher transcript levels of F3H, ANS, and ANR (Verdier et al., 2012). In contrast, overexpression of VvMYBPA2 or VvMYBPA1 in grapevine hairy roots and PtrMYB134 in poplar induced the expression of both early and late flavonoid pathway genes as well as the PA-specific genes ANR and LAR1 (Mellway et al., 2009;Terrier et al., 2009). Constitutive expression of these R2R3-MYB TFs was sufficient to induce PA biosynthesis without the need for additional factors such as bHLH or WD40 repeat proteins, as was seen in our study for TaMYB14 as well. This is in contrast to the situation in grapevine cell suspensions, where cotransformation with a bHLH factor was required for VvMYBPA1 and VvMYB5b to activate the promoters of PA-specific genes in a transient luciferase assay (Bogs et al., 2007;Deluc et al., 2008). This cotransformation was also required for activating AtDFR, AtANS, and AtANR promoters with LjTT2s in transient Arabidopsis leaf cell assays, whereas constitutive overexpression of LjTT2a was sufficient to result in the expression of AtCHS, AtDFR, and AtANR and the accumulation of PAs in stably transformed Arabidopsis seedlings (Yoshida et al., 2010). Constitutive expression of R2R3-MYB TFs generally did not result in PA accumulation in all cells of transformed plants in the above studies but rather in particular cell types, as seen in our study as well. These results could indicate that the required cofactors, such as the AtTT8 and AtTTG1 orthologs, are constitutively expressed in at least some plant tissues and hairy root cells, but not in cell cultures. Constitutive expression of a WD40 repeat protein (MtWD40-1) has been demonstrated in M. truncatula .
One white clover line transformed with TaMYB14 exhibited a deeply red-colored phenotype due to anthocyanin production and accumulation. This line initially expressed TaMYB14, contained PAs, and showed no anthocyanin accumulation for the first 6 months until the onset of winter, after which it started to accumulate large amounts of anthocyanins. At that stage, only trace levels of PA and no TaMYB14 and ANR transcripts were detectable in leaves of this line, whereas DFR, ANS, and F3959H transcripts were strongly induced. Interestingly, both MATE1 and MATE2 genes were also more highly expressed in the red line compared with the other TaMYB14 transformants. Recent studies have shown that the Arabidopsis MATE transporter TT12 takes up epicatechin-39-glucoside (E39G), a putative PA precursor in M. sativa (Pang et al., 2008), more efficiently than the anthocyanin cyanidin-3-glucoside (Cy3G), whereas MtMATE2 preferentially takes up Cy3G (Zhao et al., 2011). It was also shown that at high Cy3G concentrations, E39G uptake was inhibited (Zhao and Dixon, 2009). Therefore, the high concentrations of anthocyanins found in the red TaMYB14expressing clover line might have inhibited vacuolar PA monomer uptake and accumulation. This could have prevented PA polymerization in the vacuole, as indicated by the absence of oligomers in this line. However, it is unclear what caused the repression of TaMYB14 expression, and more studies are needed to identify factors leading to the switch from PA to anthocyanin accumulation in this line.

Chemical Composition of PAs in TaMYB14-Expressing Plants
It has been shown previously that PA detection by DMACA staining or HPLC analysis alone can be misleading . Therefore, in this study, we used targeted LC-MS n analysis and comparison with known grapevine PA standards, which allows the unambiguous and sensitive identification of PA monomers and oligomers based on their specific ion fragmentation patterns (Fraser et al., 2012). The majority of PA monomers accumulating in TaMYB14-transformed white clover leaves were epigallocatechin, with 10-fold lower concentrations of epicatechin and gallocatechin, and only traces of catechin. Analysis of the PAs in wild-type T. repens flowers has shown that they consist mostly of gallocatechin-and epigallocatechin-based monomeric units, with equal proportions of gallocatechin and epigallocatechin in the terminator units and a 1:2 ratio of gallocatechin to epigallocatechin in the extender units (Foo et al., 2000). Our findings of preferential accumulation of epigallocatechin in TaMYB14-expressing clover leaves is consistent with this; however, concentrations of epigallocatechin were at least 10-fold higher (rather than 2-fold, as seen in wild-type clover flowers) compared with gallocatechin, which might indicate that TaMYB14 induced TrANR more than TrLAR, although this is not apparent from the RT-qPCR results. Nothing is known about the composition of PAs detected in DMACA-positive trichomes of T. repens leaves, and we were not able to detect any PA-specific ions by LC-MS/ MS even in trichome-enriched extracts of immature wildtype T. repens leaves, which could simply be due to PA concentrations being below the detection limit. However, it is also possible that the compounds responsible for positive DMACA staining of wild-type trichomes are not PAs, as Peel et al. (2009) reported that in LAP1-overexpressing Medicago plants, DMACA-positive compounds accumulated, which, upon more detailed analysis, turned out not to be PAs.
Native PAs in M. sativa seeds are mainly based on epicatechin and epigallocatechin (91%), with much lower concentrations of catechin-and gallocatechinbased PAs (Koupai-Abyazani et al., 1993). In contrast, M. sativa foliar monomers induced by ectopic expression of TaMYB14 in this study consisted almost equally of epicatechin and catechin, whereas (epi)gallocatechin monomers were undetectable. This might indicate that TaMYB14 induced both ANR (resulting in epicatechin production) and LAR (resulting in catechin production) almost equally, as was seen in our T. repens transformants, whereas the native M. sativa PA regulator(s) in seeds might be more active with MsANR rather than MsLAR. This would be consistent with the induction of ANR, but not LAR, in hairy roots of M. truncatula overexpressing MtPAR (Verdier et al., 2012). We could not detect any 59-hydroxylated epigallocatechin or gallocatechin monomers in TaMYB14-transformed M. sativa, and the majority of the foliar PA dimers and trimers were PC based, with only traces of a PC:PC:PD trimer identified. This is consistent with a 14:1 ratio of PC-to PDbased PAs found in seeds (Koupai-Abyazani et al., 1993) and could mean that M. sativa has only minor F3959H activity, which is also not induced by TaMYB14. Transcriptomic analysis of MtPAR-transformed hairy roots also did not show any up-regulation of a F3959H (Verdier et al., 2012). Pang et al. (2008) showed that more than 75% of epicatechin accumulates as E39G in developing M. sativa seed coats, whereas its content declined during seed development and was absent in mature seeds. We also detected E39G in TaMYB14-transformed M. sativa leaves, but only at one-tenth of the concentration of the free PA monomers, which could reflect insufficient up-regulation of the corresponding UDP-glucosyltransferase by TaMYB14. We did not detect any glycosylated PA monomers in T. repens or tobacco plants transformed with TaMYB14 or in wildtype T. arvense, and it is not clear if the production of E39G is specific for Medicago species or if levels of E39G were below the detection limit in the other three species analyzed here.

CONCLUSION
Our study conclusively demonstrates that expression of TaMYB14 specifically induces the biosynthesis and accumulation of PAs in foliar tissue of the forage legume T. repens through the up-regulation of late flavonoid biosynthetic pathway genes and the PAspecific transporter TrMATE1. Constitutive expression of TaMYB14 in M. sativa and T. repens resulted in foliar PA accumulation of up to 1.8% dry mass, with PA complexes as high as DP6 and a monomeric unit composition comparable to that of wild-type flower PAs. It is conceivable, therefore, that the production of genetically modified T. repens cultivars expressing TaMYB14 is a viable option for a bloat-safe, highly productive pastoral agriculture with reduced greenhouse gas emissions.
Seeds were germinated on seed trays and plants grown in a glasshouse. Plant tissues were harvested at various developmental stages and either immediately processed for histochemical staining or frozen in liquid nitrogen and stored at 280°C for subsequent DNA, RNA, and PA isolation.
Fresh tissue samples (mature leaves, flowers, roots, immature/meristematic leaves, and trichomes) were collected from plants, and PAs were histochemically analyzed using the acidified DMACA (Sigma NZ) method essentially as described by Li et al. (1996). Briefly, tissue samples were decolorized in ethanol:acetic acid (3:1) overnight, stained with DMACA (3 mg mL 21 ; methanol:hydrochloric acid, 1:1) for 20 min at room temperature, and destained with several washes of 70% ethanol. Meristematic leaves and trichomes were dissected from end tips of stolons using a microscope. Histochemical staining for GUS activity was performed as described (Jefferson et al., 1987).

LC-MS n Analysis and Quantitation of PAs in Plant Tissues
To extract flavonoids for LC-MS n analysis, leaf tissue (500 mg fresh weight) was frozen in liquid N 2 , ground to a fine powder, and extracted with acetic acid:methanol (80:20, v/v) for 30 min at 4°C. Plant debris was pelleted in a microcentrifuge at 13,000 rpm for 10 min. The supernatant was removed and placed at 220°C for 30 min. Extracts were analyzed by LC-MS n using a linear ion-trap mass spectrometer (Thermo LTQ) coupled to a Thermo Finnigan Surveyor HPLC system equipped with a Thermo photo diode array detector. Thermo Finnigan Xcalibur software (version 2.0) was used for data acquisition and processing. A 5-mL aliquot of sample was injected onto a 150-3 2.1-mm Luna C18(2) column (Phenomenex) held at a constant 25°C. The HPLC solvents used were as follows: solvent A = 0.1% formic acid in water; solvent B = 0.1% formic acid in acetonitrile. The flow rate was 200 mL min 21 , and the solvent gradient used is shown in Supplemental Table S4. Photo diode array data were collected across the range of 220 to 600 nm for the entire chromatogram.
The mass spectrometer was set for electrospray ionization in positive mode. The spray voltage was 4.5 kV, the capillary temperature was 275°C, and the flow rates of sheath gas, auxiliary gas, and sweep gas were set to 20, 10, and 5 arbitrary units min 21 , respectively. The first 4 min and last 11 min of flow from the HPLC device were diverted to waste. The MS apparatus was programmed to scan from 150 to 2,000 m/z (MS 1 scan) and then to perform data-dependent MS 3 on the most intense MS 1 ion. The isolation window for the data-dependent MS 3 method was 2 m (nominal mass units), and fragmentation (35% relative collision energy) of the most intense ion from the MS 1 spectrum was followed by the isolation (2 m) and fragmentation (35% relative collision energy) of the most intense ion from the MS 2 spectrum. The mass spectrometer then sequentially performed product ion scans for selected masses on the different combinations of PC and PD masses shown in Supplemental Table S5, with an isolation window for each selected m/z value of 2.0 m and a relative collision energy of 35%. Each sample was run twice with two data collection methods to ensure that enough points were collected in the mass spectrometer across the chromatographic peaks. The first method included an MS 1 scan, the data-dependent MS 3 scans, followed by product ion scans based on selected ions (Supplemental Table S5) for the PA and PD monomers, dimers, and trimers. The second method contained an MS 1 scan, the data-dependent MS 3 scans, followed by product ion scans based on selected ions (Supplemental Table S5) for the tetramers; this method was used for M. sativa extracts only. The masses listed cover the different combinations of procyanidin (catechin and/or epicatechin) and prodelphinidin (gallocatechin and/or epigallocatechin) masses up to DP4.
Soluble PAs were quantified using the butanol hydrochloric acid methodology (Terrill et al., 1992). A calibration curve was set up with purified PAs isolated from Lotus pedunculatus (Sivakumaran et al., 2006).
Standard Methods for Isolation, PCR, Cloning, and Sequencing of DNA and RNA Genomic DNA was isolated from fresh or frozen plant tissues using the DNeasy Plant Mini Kit (Qiagen) following the manufacturer's instructions. DNA preparations were treated with RNaseH (Sigma NZ) to remove RNA from the samples. Total RNA was isolated from fresh or frozen tissues using the RNeasy Plant Mini Kit (Qiagen). Isolated total RNA was treated with RNasefree DNaseI (Roche NZ) to remove DNA from the samples following the manufacturer's instructions. The concentration and purity of DNA and RNA samples were assessed by determining the ratio of A 260 and A 280 using a NanoDrop ND-100 spectrophotometer (NanoDrop Technologies; Thermo Scientific).
Standard PCR was carried out in a Thermal Cycler (Applied Biosystems) using 5 ng of DNA or 1 mL of cDNA as template. The thermal cycle conditions were as follows: initial reaction at 94°C for 30 s; 35 cycles at 94°C for 30 s, 50°C to 64°C for 30 s (depending on the melting temperature of the primers), and 72°C for 1 to 2 min (1 min kb 21 ), and a final reaction at 72°C for 10 min. PCR products were separated by agarose gel electrophoresis and visualized by ethidium bromide staining. Bands of interest were cut out, and DNA was subsequently extracted from the gel slice using the QIAquick Gel Extraction Kit (Qiagen) following the manufacturer's instructions. Extracted PCR products were cloned into TOPO 2.1 vectors (Invitrogen NZ) and transformed into OneShot Escherichia coli cells by chemical transformation following the manufacturer's instructions. Bacteria were plated, selected, and grown according to the manufacturer's protocols. Bacterial plasmid DNA was extracted and purified from the bacterial cultures using the Qiagen Prep Plasmid Miniprep Kit (Qiagen). The ZAP XR vector system used the ZAP-cDNA Synthesis and Gigapack III Gold Kits (Stratagene).
Isolated plasmid DNA was sequenced using the dideoxynucleotide chain termination method (Sanger et al., 1977) using Big-Dye (version 3.1) chemistry (Applied Biosystems). Either M13 forward and reverse primers or specific gene primers were used. The products were separated on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems), and sequence data were compared with sequence information published in GenBank (National Center for Biotechnology Information) using the AlignX function of Vector NTI (Invitrogen NZ).
A variety of libraries were constructed from T. arvense and T. repens leaf, root, stolon, and floral tissues using the Uni-ZAP XR vector system with the ZAP-cDNA Synthesis and Gigapack III Gold Kits (Stratagene), amplified using the Clontech BD SMART PCR cDNA Synthesis Kit (Clontech), and cloned using TOPO 2.1 vectors (Invitrogen NZ).

Isolation of Putative R2R3-MYB Orthologs from T. arvense
To predict the conserved domain of legume R2R3-MYB TFs, we used the hidden Markov model method (HMMER 3.0;Eddy, 1996) based on R2R3-MYB sequences published previously (Nesi et al., 2001;Miyake et al., 2003). EST and EST contigs from available legume sequence databases (in house: T. repens EST CS35, T. repens Pastoral Genomics NR; public: T. pratense EST contigs, Lotus japonicus EST contigs, Glycine max contigs, Medicago truncatula predicted proteins, M. sativa glandular trichome ESTs, the Dana Farber Gene Indices databases at http://compbio.dfci.harvard.edu/tgi/plant.html, http:// www.kazusa.or.jp/e/resources/index.html, and GenBank accession nos. CO511688-CO517334) were downloaded onto an in-house database system. Sequences were translated in all six reading frames and searched with the predicted R2R3-MYB domain. The total number of legume sequences predicted to have an R2R3-MYB domain identified with the HMM method was 166. A total of 130 full-length domain sequences were aligned using the HMMER alignment program (http://hmmer.wustl.edu/), and degenerate primers were designed to isolate putative R2R3-MYB TFs from T. arvense (Supplemental Table S1).
Total RNA was isolated from immature and mature T. arvense leaves as described above and reverse-transcribed into two separate cDNA libraries using the SMART cDNA Synthesis Kit (Clontech). Library A was synthesized using the SMART CDS primer IIA, which binds to the poly(A) tail; library B was synthesized using an anchored poly(T) primer and the SMART IIA oligonucleotide, which anchors to the 59 untranslated region of the mRNA. PCR products were amplified using the PCR primer IIA and degenerate MYB primers designed to the 59 region of the MYB domain, and amplicons were cloned into TOPO vectors as described above; 200 clones were randomly selected for sequencing and analyzed using BLAST analysis against the Database of Arabidopsis Transcription Factors (http://datf.cbi.pku.edu.cn/index.php).
Genomic DNA was PCR amplified using the primer pair MYB14ATG and MYB14TGA (Supplemental Table S1) to amplify full-length MYB14 genomic sequences from T. arvense, T. repens, T. affine, and T. occidentale.

Genetic Constructs, Plant Transformation, and Regeneration
For overexpression of TaMYB14 or GUS (Jefferson et al., 1987), the genes were first cloned into the primary cloning vector pART7 between the CaMV 35S promoter and the octopine synthase gene 39-untranslated region as an EcoRI fragment (Gleave, 1992). The entire overexpression cartridge was then linearized and shuttled as a NotI fragment into the binary vectors pART27 (Gleave, 1992) for N. tabacum and pHZBAR for M. sativa and T. repens transformation. Transfer-DNAs of the genetic construct showing the orientation of the cloned gene are represented graphically in Supplemental Figure S14.
To silence TaMYB14, a 299-bp fragment from the 39 end of TaMYB14 cDNA was cloned into the hairpin RNAi vector pHANNIBAL (Helliwell and Waterhouse, 2003) and used to transform T. arvense cotyledons. The fragment for cloning in the sense orientation was amplified with the primer pair MYB14F1/MYB14R (Supplemental Table S1), which added XbaI sites to the 59 and 39 ends. For cloning in the antisense orientation, XhoI sites were added to both ends of the fragment with the primer pair MYB14F/MYB14R1 (Supplemental Table S1). The direction of cloning was determined by DNA sequence analysis. The NotI fragment from MYB14pHANNIBAL containing the hairpin RNA cassette was cloned into the binary vector pHZBAR (designated pHZBARSMYB; Supplemental Fig. S15) and transformed into Agrobacterium tumefaciens for transformation of T. arvense.
RT-qPCR Analysis of Gene Expression in CaMV 35S:: TaMYB14-Expressing T. repens Total RNA from wild-type and CaMV 35S::TaMYB14-expressing T. repens leaves was isolated using TRIzol Reagent (Invitrogen NZ) containing 2% polyvinylpyrrolidone and treated with RNase-free DNaseI (Roche NZ) to remove residual genomic DNA. DNase-treated RNA was subsequently cleaned up using the RNeasy Plant Mini Spin Kit (Qiagen) to remove enzyme, salts, and degraded DNA fragments. RNA integrity and quality were checked by denaturing gel electrophoresis, and the absence of genomic DNA was confirmed by PCR using a primer pair designed for T. repens 5.8S rRNA (Supplemental Table S1) prior to reverse transcription. One mg of total RNA was reverse transcribed into cDNA using the SuperScriptVILO cDNA Synthesis Kit (Invitrogen NZ) following the manufacturer's instructions. The synthesized cDNAs were then diluted 100-fold, of which 10 mL was used for subsequent qPCR analysis with a total PCR volume of 25 mL.
Gene expression profiles were quantified by qPCR amplifying the target cDNA using an iCycler (Bio-Rad Laboratories NZ) following the procedures described by Rasmussen et al. (2007). The qPCR was arrayed using SYBR Green PCR Master Mix (Applied Biosystems) at 60°C annealing temperature. Plasmids containing the target genes, ranging from 2 3 10 1 to 2 3 10 7 copies, were used for calibration of the PCR. The expression of the analyzed genes was normalized and expressed as a ratio of the target gene and actin (TrACT) as the housekeeping gene. Primers used for amplification were specific to T. repens putative CHS, FLS, F3H, F3959H, DFR, ANS, ANR, LAR, MATE1, and MATE2 sequences as identified in our in-house T. repens databases; primer sequences are listed in Supplemental Table S1. Primer specificity was confirmed by direct cloning and sequencing of individual PCR amplification products. Reported are results of three technical replicates.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Amino acid alignment of TaMYB14 homologs from Trifolium and Medicago species.
Supplemental Figure S3. RT-PCR analysis of ANR, LAR, and MYB14 transcript presence in cDNA libraries isolated from various tissues of wild-type Trifolium species and cultivars grown under glasshouse or tissue culture conditions.
Supplemental Figure S4. LC-MS n composite extracted ion chromatograms of ions for catechin and gallocatechin in T. arvense leaf extracts.
Supplemental Figure S5. LC-MS n composite extracted ion chromatograms of ions for PA trimers in T. arvense leaf extracts.
Supplemental Figure S6. LC-MS n composite extracted ion chromatograms of ions for PA monomers in N. tabacum leaf extracts.
Supplemental Figure S7. LC-MS n composite extracted ion chromatograms of ions for PA monomers in M. sativa leaf extracts.
Supplemental Figure S8. LC-MS n composite extracted ion chromatograms of ions for PA trimers in M. sativa leaf extracts.
Supplemental Figure S10. LC-MS n composite extracted ion chromatograms of PA monomers in T. repens leaf extracts.
Supplemental Figure S11. LC-MS n composite extracted ion chromatograms of PA trimers in T. repens leaf extracts.
Supplemental Figure S13. LC-MS chromatograms of the anthocyanins in a red-leafed T. repens line.
Supplemental Figure S14. Graphical presentation of the CaMV 35S:: TaMYB14 gene construct used for transformation of N. tabacum, M. sativa, and T. repens.
Supplemental Figure S15. Graphical presentation of the RNAi TaMYB14 gene-silencing construct used for transformation of T. arvense.
Supplemental Table S1. List of primer sequences used in this study.
Supplemental Table S2. PA accumulation in leaves of M. sativa wild-type or transgenic lines transformed with CaMV 35S::TaMYB14.
Supplemental Table S3. RT-qPCR analysis of gene expression in T. repens.
Supplemental Table S4. Solvent gradient used for the separation of PA oligomers.
Supplemental Table S5. Selected ion masses (m/z) for product ion scans of PA monomers, dimers, trimers, and tetramers in positive electrospray ionization mode.