Early steps in proanthocyanidin biosynthesis in the model legume Medicago truncatula.

Oligomeric proanthocyanidins (PAs) composed primarily of epicatechin units accumulate in the seed coats of the model legume Medicago truncatula, reaching maximal levels at around 20 d after pollination. Genes encoding the single Medicago anthocyanidin synthase (ANS; EC 1.14.11.19) and leucoanthocyanidin reductase (LAR; EC 1.17.1.3) were cloned and the corresponding enzymes functionally identified. Recombinant MtANS converted leucocyanidin to cyanidin, and, more efficiently, dihydroquercetin to the flavonol quercetin. Levels of transcripts encoding dihydroflavonol reductase, ANS, and anthocyanidin reductase (ANR), the enzyme responsible for conversion of anthocyanidin to (-)-epicatechin, paralleled the accumulation of PAs in developing seeds, whereas LAR transcripts appeared to be more transiently expressed. LAR, ANS, and ANR proteins were localized to the cytosol in transfected tobacco (Nicotiana tabacum) leaves. Antisense down-regulation of ANS in M. truncatula resulted in reduced anthocyanin and PA levels, but had no impact on flavonol levels. Transgenic tobacco plants constitutively overexpressing MtLAR showed reduced anthocyanin content, but no catechin or increased levels of PAs were detected either in leaves or in flowers. Our results confirm previously ascribed in vivo functions for ANS and ANR. However, the apparent lack of catechin in M. truncatula PAs, the poor correlation between LAR expression and PA accumulation, and the lack of production of catechin monomers or oligomers in transgenic plants overexpressing MtLAR question the role of MtLAR in PA biosynthesis in Medicago.


al, 2006).
Most genetic studies on PA biosynthesis to date have been performed in Arabidopsis thaliana (Lepiniec et al., 2006) or barley (Jende-Strid, 1993). Because of the importance of PAs for the quality of forage legumes, it is important to develop a model system for studies on the molecular genetics of PA biosynthesis in legumes. Medicago truncatula has been selected as a model legume for genetic and genomic studies (Oldroyd and Geurts, 2001), and a wide range of genomic and genetic resources are now available for this species, including a whole genome sequence that should be complete in 2008. We here describe the distribution, nature and deposition kinetics of PAs in M. truncatula, and the functional characterization and tissue-specific expression of MtLAR and MtANS in relation to PA biosynthesis. Our results provide the background for further development of Medicago as a model system for studying PA biosynthesis and deposition in forage legumes. Furthermore, they raise questions as to the role of LAR in the PA pathway in Medicago seeds.

Tissue-specific accumulation of PAs in M. truncatula
Tissues of M. truncatula cultivar Jemalong A17 were extracted with 70% acetone/0.5% acetic acid to extract soluble PAs, which were then quantified by reaction with DMACA (dimethylaminocinnamaldehyde) reagent. Highest PA levels (about 0.7 mg catechin equivalents per g fresh weight [FW]) were found in the seed coat, with very low levels (1.9 µ g catechin equivalents per g FW) in the remainder of the seed after the coat had been removed ( Fig. 2A). Because of the high sensitivity of the detection method, very low levels of soluble PAs were detectable in flower, leaf, root and stem. The size heterogeneity of the soluble PA oligomers from seed coats was subsequently determined by HPLC analysis with post-column derivatization with DMACA reagent (Peel and Dixon, 2007). This revealed the presence of low levels of free (-)-epicatechin monomers, along with a range of oligomers with an estimated highest degree of polymerization (DP) of greater than 12 (Fig. 3A).
Insoluble (protein-or cell wall-bound) PAs were extracted from the residues left after extraction of soluble PAs by repeated sonication in butanol-HCl, followed by heating to generate colored anthocyanin from degradation of the PAs. Using this method, the highest levels of insoluble PAs were detected in the seed coat (about 5.8 mg procyanidin B1 equivalents per g FW) (Fig. 2B) and in whole seed and immature pods, in both cases presumably the result of the presence of the seed coats. Much lower levels were detected in flower, leaf, stem and root. Due to the possible interference of cell wall components with the quantification of insoluble PAs (Marles et al., 2003), the insoluble PA content could be an underestimate. Furthermore, since different methods are used for quantification of soluble and insoluble PAs, it is difficult to compare the relative amounts of the two fractions within the different tissues; nevertheless, both fractions were clearly highest in seed coats, and insoluble PAs predominated.
PAs were extracted from developing seeds at various days after pollination (dap).
Levels of soluble PAs were already close to maximal (around 2.5 mg catechin equivalents per g FW) at the earliest time point sampled (10 dap), whereas insoluble PAs associated with the cell wall fraction reached maximum levels (around 16 mg per g FW) at 20 dap (Fig. 2C, D).

Nature of M. truncatula PAs
The butanol: HCl-hydrolyzate of the insoluble PA fraction from the seed coat was subjected to HPLC analysis and shown to contain primarily cyanidin, with lower amounts of delphinidin and pelargonidin (Fig. 3B,C). By these criteria, the M. truncatula PAs are comprised primarily of epicatechin and/or catechin units (which yield cyanidin on hydrolysis), with much lower levels of (epi)/gallocatechin (yielding delphinidin) and (epi)/afzelechin (yielding pelargonidin) units (Fig. 3C).
To further determine the structure of the M. truncatula seed coat PAs, the soluble PA fraction was isolated from seeds at various stages of development and subjected to cleavage with phloroglucinol. The released phloroglucinol adducts (extension units) and free starter units were then separated by HPLC and quantified (Fig. 3D, E). By these criteria, the soluble seed coat PAs were demonstrated to contain primarily epicatechin as starter unit, and exclusively epicatechin extension units, consistent with the butanol: HCl hydrolysis data. Only very small peaks were present at retention times corresponding to catechin starter or extension (phloroglucinol adduct) units, and these may reflect (-)-catechin units that occur via non-enzymatic epimerization of (-)-epicatechin (Xie et al., 2004b). Phloroglucinolysis analysis revealed that the mean degree of polymerization of the PAs increased from a value of 6 in seeds at 10 dap to a value of 17 in mature seeds ( Fig. 3F), consistent with the normal phase HPLC analysis.

Identification and properties of MtLAR
Informatic analysis revealed one candidate LAR homolog from M. truncatula, represented by GenBank accession #BN000703, and DNA gel blot analysis confirmed the presence of a single LAR gene in the M. truncatula genome (Supplemental Fig. 1). conserved regions of 13, 11 and 10 amino acids containing, respectively, the LARspecific motifs RFLP, ICCN and THD as described previously (Tanner et al., 2003;Bogs et al., 2005). As shown in Figure 4, the conserved sequences including and immediately following these motifs were identical in MtLAR and DuLAR, with only two amino acid substitutions from the VvLAR and LcLAR sequences. The calculated isoelectric point and molecular weight of MtLAR were 5.88 and 38 KDa, respectively.
To determine the in vitro activity of recombinant MtLAR protein, the ORF was subcloned into the E. coli expression vector pQE30 fused with a 6 × histidine tag at the N-terminus. Recombinant MtLAR protein had a molecular weight of about 45 kDa by analysis on a 12.5% SDS-PAGE protein gel (Fig. 5A). The discrepancy between the calculated and the apparent molecular weights of the MtLAR protein could be due to its particular amino acid composition, which can affect electrophoretic migration (Klenova et al., 1997). The protein was purified and assayed using 3 H-labeled leucocyanidin as substrate in the presence of NADPH, followed by analysis of products by HPLC-UV and radioactivity detection. The vector control extract showed no product formation ( Fig.  5C), whereas a peak with the same retention time as that of (+)-catechin was detected when MtLAR protein was incubated with 3 H-labeled leucocyanidin (Fig. 5D, E).

Identification and properties of MtANS
A partial fragment of a putative MtANS sequence was selected from several M. truncatula expressed sequence tags (ESTs) by querying the TIGR EST database with Arabidopsis ANS (Wilmouth et al., 2002 (Wilmouth et al., 2002). Multiple sequence alignment (Supplemental Fig.   2) confirmed the presence of three conserved residues (His232, His288 and Asp234 for MtANS) required to coordinate ferrous-iron at the catalytic center of iron-containing soluble oxygenases, and an arginine residue (Arg298 for MtANS) that is assumed to contribute to the specific binding of 2-oxoglutarate (Britsch et al., 1993;Prescott and John, 1996;Lukacin and Britsch, 1997). The theoretical isoelectric point and molecular weight of the deduced MtANS protein were predicted to be 5.9 and 40 KDa, respectively.
MtANS shares around 40% identity at the amino acid level with other plant 2-ODDs (flavonol synthase [FLS] and flavanone 3-β-hydroxylase [F3H]) involved in flavonoid biosynthesis (Prescott and John, 1996). The deduced amino acid sequence of MtANS, together with those of FLS and F3H from several plant species, were aligned using CLUSTALW (Thompson et al., 1994) and subjected to phylogenetic analysis using the Neighbor-Joining method (Saitou and Nei, 1987). As viewed with Phylodraw (Choi et al., 2000), the three groups of functionally distinct plant 2-ODDs, FLS, F3H and ANS, are clearly divided into three distinct clusters, and MtANS clustered within the ANS group (Supplemental Fig. 3).
The MtANS ORF was expressed as a 6 × histine tagged N-terminal fusion in E. coli.
Recombinant MtANS protein had a molecular weight of around 40 kDa (Fig. 6A), as predicted by bioinformatics tools. The recombinant protein was purified and its activity assayed using radiolabeled 3 H-leucocyanidin as substrate in the present of ferrous iron and 2-oxoglutaric acid. The product was detected by HPLC monitored at 530 nm, and its identity confirmed as cyanidin by comparison of retention time and UV/Vis spectrum with those of authentic standards (Fig. 6B, D). The cyanidin peak was collected and confirmed to contain tritium label by scintillation counting (Fig. 6E), although the overall enzymatic activity was low (50 pmol/min/mg protein), comparable to that of recombinant LAR. No 3 H-leucocyanidin was converted to cyanidin in incubations with protein from the vector control (Fig. 6C).
Leucocyanidin is quite unstable, and converts to its precursor dihydroquercetin during storage. Some 3 H-quercetin was detected in incubations of MtANS with 3 Hleucocyanidin, raising the question of whether it was derived from 3 H-dihydroquercetin.
To test this hypothesis, the same batch of purified recombinant MtANS used for the assays with labeled leucocyanidin was incubated with unlabeled dihydroquercetin in the presence of ferrous iron and 2-oxoglutaric acid. Significant production of quercetin was observed by HPLC analysis (Supplemental Fig. 4C), whereas no quercetin was observed when dihydroquercetin was incubated with protein from an empty vector control (Supplemental Fig. 4B). Thus, MtANS is a bifunctional enzyme within the anthocyanidin/PA and flavonol biosynthetic pathways, at least in vitro. Similar observations have been made with ANS from Arabidopsis and rice (Turnbull et al., 2000;Tanner et al., 2003;Reddy et al., 2007). The rate of reaction with dihydroquercetin was almost 10-fold higher (460 pmol/mim/mg protein) than with leucocyanidin.

Tissue-specific expression of PA biosynthetic genes in M. truncatula
To assess the potential involvement of MtLAR and MtANS in PA biosynthesis, their To further define the site of expression in seeds, RNA from pods, whole seeds, seed coats and seeds without seed coats was analyzed by quantitative real time PCR (Fig. 7B).
In spite of high variation at the lowest transcript levels, the data suggest that the expression of ANS, ANR and LAR in seeds is almost entirely the result of expression in the seed coat, thereby ruling out the possibility of significant contamination of seed coats from underlying tissue.

Subcellular localization of MtANS, MtLAR and MtANR
Because of the suggestion that the PA pathway may exist as a metabolic channel associated with cellular membranes (Winkel, 2004), it was of interest to determine the subcellular localizations of LAR, ANS and ANR. This was achieved by making gene constructs encoding C-terminal EGFP/YFP fusions of the respective ORFs for transient expression by particle bombardment and localization of fluorescence by confocal microscopy. As a control for endomembrane localization, we utilized a construct in which EGFP was fused to the membrane anchor of the ER-localized cytochrome P450 enzyme cinnamate 4-hydroxylase, as described previously (Achnine et al., 2004) The  (Xie et al., 2003(Xie et al., , 2006. Flower tissues from the five lines were first subjected to RT-PCR, to confirm high MtLAR expression (Supplemental Fig. 6B), and then extracted and analyzed for anthocyanin and soluble and insoluble PA levels. Several of the transgenic lines exhibited a small reduction in anthocyanin levels (although in one line, #38, they were increased). However, PA levels were reduced rather than increased in all transgenic lines (Fig. 8A, B), and no increase in material eluting from HPLC at the retention time of free catechin was observed (Fig. 8C). Similarly, no production of catechin or PAs was reported in tobacco or white clover expressing Desmodium LAR, even though the recombinant enzyme was shown to be catalytically active after expression in planta (Tanner et al., 2003). We were unable to detect catechin in transgenic tobacco leaves or flowers expressing Desmodium LAR (data not shown).
The potential in vivo function (s)  To investigate the potential additional function for ANS in formation of flavonols from dihydroflavonols, as suggested from the in vitro activity of the enzyme, total flavonoid profiles of vector control and ANS antisense lines were compared by HPLC (Supplemental Fig. 8). We could not detect any significant difference in the levels of individual or total flavonols in lines in which ANS down-regulation led to a reduction in foliar anthocyanin levels ( Fig. 9E).

M. truncatula as a model for the study of PA biosynthesis
Arabidopsis has become the major model for studies on PA biosynthesis, largely as a result of the generation and analysis of a number of transparent testa (tt) and tannindeficient seed (tds) mutants that lack PAs in the seed coat ( (Xie et al., 2004b), little is known about PAs, the enzymes specific for their biosynthesis, or the transcriptional regulation of the PA pathway, in this species.

Nature and distribution of PAs in M. truncatula
PAs have been described in alfalfa (Medicago sativa), a species very closely related to M.
truncatula. Essentially, PAs are either absent, or present in very small amounts, in the aerial organs of alfalfa. They are present at relatively high levels in seed coats, where they have been reported to exist as oligomers of DP from 4.4-6.5, with a mean DP of 5, containing epicatechin as the major extension unit and catechin as the major starter unit To date, the genetic evidence for LAR function is the observation that Arabidopsis, does not possess an obvious LAR ortholog and has a PA comprised of only epicatechin (Abrahams et al., 2003;Tanner et al., 2003;Routaboul et al., 2006). Furthermore, LAR expression is found in non-seed tissues of some plants, such as Lotus corniculatus (Paolocci et al., 2007) and grape (Bogs et al., 2005), that accumulate high levels of PAs in these tissues, and LAR is co-regulated with ANR via MYB family transcription factors (Bogs et al, 2007;Paolocci et al, 2007).
The Medicago gene that we selected as encoding a potential MtLAR had been identified as such informatically by previous workers (Bogs et al., 2005;Paolocci et al., 2007). The recombinant enzyme exhibits the same in vitro catalytic activity as the LAR from Desmodium uncinatum (Tanner et al., 2003), namely the conversion of 3 Hleucocyanidin to a compound with the same retention time and UV absorption characteristics as (+)-catechin, although, even with scaled-up reactions, we were unable to recover sufficient product for a rigorous identification with determination of stereochemistry. This has yet to be achieved for the product of any of the recombinant LARs described to date (Bogs et al., 2005;Pfeiffer et al., 2006;Takos et al., 2006;Paolocci et al., 2007). The in vitro activity of the purified recombinant Mt LAR was very low (only 40 pmol/min/mg protein) compared to the activity purified from Desmodium leaves (Tanner et al, 2003). However, a recent report failed to detect any activity for recombinant Lotus LAR when assayed as a single protein with leucocyanidin as substrate; weak activity was detected by a TLC assay only if the leucocyanidin was Medicago (this work) and ANR in tobacco (Xie et al., 2006). In the latter case, ectopic expression of MtANR in tobacco flowers resulted in production of considerable levels of both PAs and free epicatechin. Furthermore, a previous study in which Desmodium LAR was ectopically expressed in tobacco failed to report accumulation of catechin (Tanner et al., 2003), and we were unable to show catechin production in tobacco expressing Desmodium LAR.
In the absence of evidence that proteins encoded by LAR-like genes actually catalyze a different reaction from the conversion of leucocyanidin to (+)-catechin, we remain cautious as to the true in vivo function of MtLAR. We have yet to generate transgenic Medicago plants in which LAR is down-regulated, although it is not clear what the PA phenotype of these plants would be since the Medicago PAs appear to contain insignificant levels of catechin. We can not at present rule out the possibility that a subset of insoluble PAs containing catechin units are made in Medicago, or that some type of metabolic channeling exists in vivo such that LAR does not properly couple with endogenous pathway enzymes (such as DFR, see below) in transgenic plants expressing a "foreign" LAR, thereby accounting for the lack of catechin production in transgenic plants over-expressing the enzyme, although this is not a problem with transgenic expression of ANR for epicatechin production (Xie et al., 2006). The reduction in PA levels in tobacco flowers expressing MtLAR could likewise be explained by incorrect insertion of a heterologous functional LAR into a metabolic channel for PA formation, or by LAR having alternate or additional functions that divert flux from PA biosynthesis.
The D. uncinatum LAR protein was shown to be inhibited by naringenin, dihydroflavonols and flavonols (Tanner et al., 2003), so it is possible that the enzyme activity may be blocked in vivo in transgenic plants in which pools of flavonoid intermediates are higher than in the natural situation.

Metabolic channeling at the branch point for PA biosynthesis?
Because of the instability and expense of leucoanthocyanidin, the first detection of LAR activity in crude plant cell extracts utilized coupled assays in which dihydroflavonol was converted by endogenous DFR to leucoanthocyanidin and then on to catechin (Stafford and Lester, 1985;Kristiansen, 1986). A similar coupled assay with recombinant enzymes was shown to be essential for detection of LAR activity from Lotus corniculatus (Paolocci et al, 2007). A protein complex from Onobrychis viciifolia could produce catechin from dihydroflavonol (Singh et al., 1997), although the exact nature of the protein components was not reported. These observations are consistent with a recently proposed model for metabolic channeling in flavonoid biosynthesis, whereby PA-specific enzymes such as LAR, ANS and ANR, possibly along with DFR, might form a complex through which intermediates are channeled directly into the formation of PAs. A separate channel would be envisaged to produce anthocyanins (Winkel, 2004).
In such a model, the flavonoid biosynthetic pathway is seen as localized as one or two endoplasmic reticulum (ER)-associated multi-enzyme complexes (Stafford, 1974).
However, our results suggest that LAR, ANS and ANR are all localized to the cytosol. This is consistent with the fact that none of these enzymes has clear signal peptides or membrane-targeting domains. Nevertheless, these experiments were performed by particle bombardment of single enzyme constructs into tobacco leaf epidermal cells, not into the "natural" milieu of the endothelial layer of the Medicago seed coat. Our data suggest that none of the enzymes alone has a membrane localization, but can not rule out possible complex formation when all are present together in the same cell. Furthermore, we can not rule out the possibility of associations, between DFR and LAR, which specifically channel common precursors away from anthocyanin synthesis and into PA formation. In this respect, it is interesting to note the presence of two distinct DFR genes in M. truncatula. Although the specificities of the corresponding enzymes have been determined, and exhibit differences as regards the A-and B-ring substitution patterns of the dihydroflavonol substrates (Xie et al., 2004a), the exact stereochemistries of the products have yet to be determined. This may be important for understanding LAR function.
In summary, our results provide the information on PA content and composition required to allow for the development of M. truncatula as a model for functional genomic approaches to understanding PA biosynthesis, with translational opportunities for quality improvement in alfalfa. Future studies will address the function of LAR in Medicago through direct forward and reverse genetic approaches.
Individual tissues were harvested from greenhouse material, separated, and frozen in liquid nitrogen until further processing. Transgenic tobacco and M. truncatula (cv. R108) were grown under the same conditions, and leaf tissue for anthocyanin analysis was collected one month after the Medicago seedlings were transferred from media to soil. To collect immature seed, the pollination date of each flower was recorded, the flower labeled, and individual pods then collected at either 10, 12, 16, 20, 24 or 36 dap. Seeds from the same dap were then separated, pooled, and frozen in liquid nitrogen.

Extraction and quantification of PAs, anthocyanins and flavonols
For extraction of PAs, tissues were ground in liquid nitrogen, and 0.25-1.0 g batches were extracted with 5 ml of extraction solution (70% acetone/0.5% acetic) acid by vortexing followed by sonication at 30°C for 30 min. Following centrifugation at 2,500g for 10 min, the residues were re-extracted twice as above. The pooled supernatants were then extracted with 30 ml of chloroform, and the aqueous supernatant re-extracted twice with chloroform and three times with hexane. Samples were freeze-dried and re-suspended in extraction solution to a final concentration of 3g original sample/ml. The samples were spun briefly, transferred to another tube, and soluble PA content determined using DMACA reagent with catechin standards. In brief, aliquots of samples or standards (2.5µl) were mixed with 197.5µl of DMACA reagent (0.2% w/v DMACA in methanol-3N HCl (1: 1)) in microplate wells; for blanks, the same samples were replaced with 2.5 µl methanol-3N HCl. Samples, blanks and standards were read within 15 min on a Wallac Victor 2 plate reader equipped with a 640 nm emission filter. Blanks were subtracted from samples and PA content calculated as catechin equivalents.
For measurement of insoluble PAs, the residues from the above tissue extractions were dried in air for two days, and 1 ml butanol-HCl reagent was then added and the mixture sonicated at room temperature for 60 min, followed by centrifugation at 2,500 g for 10 min. The supernatants were transferred to cuvettes for determination of absorption at 550 nm, and the samples were then boiled for 1 h. After cooling to room temperature, the absorbance at 550 nm was recorded again and the first value subtracted from the second. Absorbance values were converted into PA equivalents using a standard curve (2.5, 5, 10, 20 and 40 µ g) of procyanidin B1 (Indofine, NJ, USA).
Normal phase HPLC analysis of soluble PAs was performed using an HP 1100 system equipped with a diode-array-detector. Samples were separated on a 250 × 4.6 mm Luna 5µ silica column, and post-column derivatization was accomplished using a separate HPLC pump (Alltech model 426) which was used to deliver the DMACA reagent (1% DMACA in 1.5M H 2 SO 4 in methanol) to a mixing tee where effluent from the column and the reagent combined and passed through an 8 m coil of 0.2mm i.d.
PEEK tubing prior to detection at 640nm (Peel and Dixon, 2007).
The soluble PAs from seeds at different development stages were purified on a Sephadex LH-20 column (Pharmacia, Uppsala, Sweden) using 50% (v/v) methanol, followed by 70% acetone to elute the PAs. Acetone was removed by rotary evaporation, and the aqueous phase was freeze-dried. One hundred µg of the purified PA samples were then analyzed by the phloroglucinolysis procedure, with detection of the PA degradation products by HPLC, as described previously (Kennedy and Jones, 2001).
For analysis of anthocyanin levels, 5 ml methanol: 0.1% HCl was added to 0.5 g ground tissue and sonicated for 1 h, followed by shaking overnight at 120 rpm. After centrifugation at 2,500g for 10 min, 1 ml water was added to 1 ml extract followed by  µ l methanol, 20 µl of which was used for reverse phase HPLC analysis as described below.

Isolation of LAR and ANS cDNAs from M. truncatula
Total RNA from developing seeds of M. truncatula (cv. Jemalong A17) was isolated using modified CTAB extraction buffer (2% CTAB, 2% polyvinylpyrrolidone, 100 mM The isoelectric points and molecular weights of the MtLAR and MtANS proteins were calculated using the pI/MW calculation tools at www.expasy.org.

In vitro expression of recombinant MtLAR and MtANS proteins
The open reading frame of the MtLAR gene was PCR amplified from pGEMT-MtLAR

Transformation of tobacco with MtLAR
The ORF of MtLAR was amplified with the primers MtLAR-BF and MtLAR-SR, and then cloned into the BamHI and SacI sites of the binary vector pBI121. The resulting vector pBI121-MtLAR, with MtLAR driven by the cauliflower mosaic virus 35S promoter, was transferred into Agrobacterium tumefaciens strain LBA4404 by electroporation. Transgenic tobacco plants (Nicotiana tabacum L. cv Xanthi) were generated by Agrobacterium-mediated transformation of leaf discs on MS media supplied with 100 mg/l kanamycin using standard protocols (Horsch et al., 1988). Only one bud was picked from each explant to ensure independent transformants.

Transformation of M. truncatula with antisense MtANS
The MtANS open reading frame in the antisense orientation was amplified with the primers MtANS-SF (5'-GAGCTCATGGGAACGGTGGCTCAAAGAG-3', Sac I site is double underlined and the start codon is single underlined) and MtANS-BR (5'-GGATCCTCATTTTTTGGGATCATCTTTCTTC-3', Bam HI site is double underlined and the stop codon is single underlined). The amplified PCR product was cloned into the BamHI and SacI sites of the binary vector pBI121, and the resulting vector pBI121-MtANSanti, with MtANS in the antisense orientation driven by the 35S promoter, was transferred into Agrobacterium tumefaciens strain LBA4404 by electroporation.
Transgenic M. truncatula (cv R108) plants were generated by Agrobacterium-mediated transformation of cotyledons as described previously . Each transgenic plant was from a separate cotyledonary explant.

Analysis of LAR, ANS and ANR transcript levels by RT-PCR in transgenic plants
Total Inc., Eugene, OR).

RT-PCR
Total RNA from, pod, seed, seed coat and seed without coat was isolated using the CTAB method, and all RNAs were purified and concentrated, and first strand cDNA synthesized, as described above. The primers for quantitative real time RT-PCR were

Microarray analysis
Affymetrix microarray analysis of various Medicago tissues and six seed development stages has been described recently (Benedito et al., 2007) obtained using dCHIP. Gene selections based on student t-test and associative t-test were made using Matlab (MathWorks, Natick, MA). A Bonferroni-correction P value threshold of 1 x 10 -5 was used. Affymetrix microarray analysis of total RNA from Medicago seed coats was conducted as above with three biological replicates. Sequences of MtLAR and MtANS are available in the GenBank database as accession #s BN000703 and EF544389, respectively.

SUPPLEMENTAL MATERIAL
The following supplemental material is available for this article online:         seed coats at various stages of seed development (days after pollination, dap) as determined by Affymetrix microarray analysis Letters reflect differences that were statistically significant at p = 0.05 by t-test.