MYB5 and MYB14 play pivotal roles in seed coat polymer biosynthesis in Medicago truncatula

In Arabidopsis, the major MYB protein regulating proanthocyanidin (PA) biosynthesis is TT2, named for the transparent testa phenotype of tt2 mutant seeds that lack PAs in their coats. In contrast, the MYB5 transcription factor mainly regulates seed mucilage biosynthesis and trichome branching, with only a minor role in PA biosynthesis. We here characterize MYB5 and MYB14 (a TT2 homolog) in the model legume Medicago truncatula . Over-expression of Mt MYB5 or Mt MYB14 strongly induces PA accumulation in M. truncatula hairy roots and both myb5 and myb14 mutants of Medicago exhibit darker seed coat color than wild-type plants, with myb5 also showing deficiency in mucilage biosynthesis. myb5 mutant seeds have a much stronger seed color phenotype than myb14 . The myb5 and myb14 mutants accumulate, respectively, about 30% and 50% of the PA content of wild-type plants, and PA levels are reduced further in myb5 myb14 double mutants. Transcriptome analyses of over-expressing hairy roots and knockout mutants of Mt MYB5 and Mt MYB14 indicate that Mt MYB5 regulates a broader set of genes than Mt MYB14. Moreover, we demonstrate that Mt MYB5 and Mt MYB14 physically interact and synergistically activate the promoters of anthocyanidin reductase ( ANR ) and leucoanthocyanidin reductase ( LAR ), the key structural genes leading to PA biosynthesis, in the presence of Mt TT8 and Mt WD40-1. Our results provide new insights into the complex regulation of PA and mucilage biosynthesis in Medicago . vector (internal control for transfection efficiency) were used to transform 100 μ L batches of protoplasts (~1 x 10 6 cells/mL). The Dual-Luciferase® Reporter Assay kit (Promega) was used to quantify the luciferase activities according to the manufacturer’s instructions. A GloMax® 96 Microplate Luminometer (Promega) was used to read the luciferase activities.


INTRODUCTION
myb5 mutants have a stronger phenotype than myb14 mutants. Medicago MYB5 and MYB14 physically interact, and synergistically activate transcription from the ANR and LAR promoters in the presence of Mt TT8 and Mt WD40-1, providing a revised model for regulation of the PA biosynthesis pathway.

Identification of Mt MYB5 and Mt MYB14
A WD-40 protein is a key component of the transcriptional complex that activates PAbiosynthetic genes (Lepiniec et al., 2006;Pang et al., 2009), and we recently reported that the MYB transcription factor Mt PAR (Proanthocyanidin Regulator) can activate the promoter of Medicago WD40-1 in yeast (Verdier et al., 2012). To search for new positive regulators of WD40-1, which may therefore also play a role in PA biosynthesis, we performed a yeast onehybrid (Y1H) screen against a cDNA library of young seeds harvested at 10 days after pollination (DAP). We found five MYB transcription factors able to activate the WD40-1 promoter in this assay. The activation activities of these MYB proteins were confirmed by using the full-length cDNAs to perform the Y1H assay a second time. Among these MYB transcription factors, two show similarity to Arabidopsis mucilage regulators, two show similarity to Arabidopsis MYB4 and one (Medtr3g083540) shows high similarity to known transcription factors activating PA biosynthesis. The fact that most of the genes identified by Y1H screening are related to known functions of WD40 proteins in plants suggests that these transcription factors are probably true regulators of Medicago WD40-1. Here, we report the detailed characterization of the PA biosynthesis regulator encoded by Medtr3g083540. which may require additional factors that are not present in the yeast assay (see below) ( Figure   1A). Arabidopsis TT2 is the best studied MYB transcription factor regulating PA biosynthesis (Nesi et al., 2001). No Medicago homolog of TT2 could be found in earlier versions of the M. truncatula genome database. However, a homology search using the At TT2 protein sequence to blast the Medicago genome version Mt3.5v4 (Young et al., 2011) resulted in the positive identification of contig_238935_1 as the closest homolog of At TT2 in Medicago (Supplemental Fig. S3; Supplemental file 2; in the most recent Medicago genome Mt4.0v1, contig_238935 is assembled as Medtr4g125520.1). Contig_238935_1 has only one amino acid difference from Mt MYB14-2 (AFJ53058.1), which is the Medicago ortholog of Trifolium arvense MYB14, a protein that induces PA accumulation in red clover and alfalfa (Hancock et al., 2012). We conclude that contig_238935_1 is a new allele of Mt MYB14-2, the single amino acid difference probably being due to the fact that the two genes were cloned from different ecotypes of M.
truncatula. Y1H assay indicated that, similar to Mt PAR and Mt MYB5, Mt MYB14 is also able to activate the promoter of Mt WD40-1, but not ANR, in yeast (Fig. 1A).
As a first step to analyze the functions of Mt MYB5 and Mt MYB14, we performed phylogenetic analysis of MYBs known to regulate PA biosynthesis. In general, these MYBs form two clades in the phylogenetic tree. One includes Mt MYB5-related proteins, the other includes At TT2, Mt PAR, Vv MYBPA2, Lj TT2s and MYB14-like proteins (Fig. 1B, Supplemental file 3). Among these TT2-related MYBs, MYB14s and Lj TT2s are closely clustered. Inspection of the multiple alignment of these proteins reveals that besides the high homology at their N termini, these MYBs also share considerable amino acid identity at their Cterminal regions (Supplemental Fig. S3). We next performed qRT-PCR to check the transcript levels of MYB5 and MYB14 in different Medicago tissues. Both MYB5 and MYB14 are highly expressed in seeds (Fig. 1C).
MYB5 is also highly expressed in flowers and moderately expressed in roots. There is moderate MYB14 expression level in flowers. These different expression patterns suggest that MYB5 might have broader roles than MYB14 in Medicago.

MYB5 Promotes PA Accumulation in Medicago Hairy Roots
To demonstrate that Mt MYB5 can promote PA accumulation in planta, we generated MYB5 over-expressing transgenic hairy roots in M. truncatula ecotype A17. Control roots were transformed with a β-glucuronidase (GUS) construct. Some lines of MYB5 transformed hairy roots showed retarded growth (Figure 2A), whereas growth was unaffected in other lines.
Staining with dimethylaminocinnamaldehyde (DMACA) suggested that both the retarded and normal growing hairy roots accumulated large amounts of PAs ( Fig. 2A, bottom panels). To avoid negative growth effects, we selected normally growing roots (derived from line #7) for further analysis. There was no significant difference in anthocyanin accumulation between GUScontrol and MYB5 over-expressing roots (Fig. 2B). In contrast, both soluble and insoluble PA levels were strongly induced in MYB5 over-expressing roots, reaching 250 and 500 μg per g fresh weight (FW), respectively ( Fig. 2C and 2D).
PA profiling by normal phase HPLC coupled with post-column DMACA derivatization indicated that most of the soluble PA-like DMACA-positive material accumulating in MYB5 over-expressing lines was of relatively low molecular weight (Fig. 2E); epicatechin monomer elutes at a retention time of approximately 15 min in this HPLC system (Pang et al., 2013). Since DMACA can react with other unknown metabolites, we then performed the more diagnostic phloroglucinolysis analysis to determine the nature and composition of putative PAs in MYB5 over-expressing lines after purification of the soluble PAs on Sephadex LH20 resin. Our results indicated that epicatechin is the main component of the PAs accumulating in MYB5 overexpressing hairy roots, both as starter and extension unit (Supplemental Fig. S5). The average degree of polymerization of the soluble PAs calculated by this method (ratio of extension unit to starter unit peak areas) was determined to be approximately 31. Taken together, our data indicate that the high molecular weight DMACA-positive metabolites accumulating in MYB5 over-expressing lines are authentic PAs and similar to those occurring naturally in the Medicago seed coat (Pang et al., 2007).
We utilized ultra-performance liquid chromatography-mass spectrometry (UPLC-MS) to determine whether the monomer flavan-3-ols also accumulate in MYB5-expressing hairy roots.
This analysis revealed significant levels of epicatechin and epicatechin-3′-O-glucoside in MYB5 over-expressing lines; these compounds were barely detectable in GUS controls (Supplemental Fig. S6).

MYB14 Promotes PA Accumulation in Medicago Hairy Roots
To confirm the function of Mt MYB14 as a PA regulator, we generated MYB14 over-expressing hairy roots in M. truncatula A17. Unlike MYB5, over-expression of MYB14 did not result in growth retardation of hairy roots (

Phenotypes of Medicago myb5 Mutants
To better understand the functions of Mt MYB5, we screened the Tnt1 retrotranspon mutagenized M. truncatula R108 population (Tadege et al., 2008) to isolate myb5 mutants. In total, we obtained three independent lines of MYB5 Tnt1 insertion mutants. All three Tnt1 insertions were located in the first exon of the MYB5 gene (Fig. 4A). RT-PCR could not detect the full length MYB5 transcript in any of the three lines, indicating that all three were null mutants (Supplemental Fig. S8A). There were no visible phenotypic differences in the vegetative organs between mutant and wild-type plants during the whole developmental period under our growth conditions. However, striking differences were observed in mature seeds, which exhibited a conspicuous dark-red color, whereas wild-type seeds are light yellow (Fig. 4B,  acetone solutions, suggesting that these pigments are cross-linked in the seed coat (see discussion). DMACA-stained mutant seeds displayed a much lighter blue color than wild-type seeds, suggesting that the PA content in mutant seeds was reduced (Fig. 4B, bottom panels).
Quantification of soluble PAs by the DMACA method indicated that mutant seeds accumulated approximately 30% of the PA content of wild-type seeds (Fig. 4C). However, quantification of insoluble PAs by the butanol-HCl method did not detect differences between mutant and wildtype (Supplemental Fig. S9). The butanol-HCl method relies on hydrolysis of PAs into anthocyanidin monomers, and cannot therefore distinguish anthocyanidin released from true PAs and anthocyanidin released from non-PA complexes, the presence of which is suggested by the dark-red color of the myb5 mutant seeds. Profiling of the soluble PA fraction by normal phase HPLC coupled with DMACA staining indicated that the myb5 mutant seeds produce, besides a lower amount of total soluble PAs, a more diverse range of small PA oligomers (Fig. 4D).
Phloroglucinolysis analysis did not detect any difference in PA monomer composition or mean degree of polymerization between mutant and wild-type seeds (Supplemental Fig. S10).
At MYB5 has been demonstrated to positively regulate mucilage biosynthesis in the Arabidopsis seed. To test whether Mt MYB5 is also involved in mucilage biosynthesis in Medicago, we stained the myb5 mutant seeds with ruthenium red, a dye used to detect mucilage (Western et al., 2001). Similar to the AT myb5 mutant, Mt myb5 mutant seeds exhibited much lighter ruthenium red staining that wild-type seeds, indicating that the mutant seeds accumulate much less mucilage in the seed coat ( Fig. 4E, top panel).
To determine if there are any differences in seed coat morphology between the myb5 mutant and wild-type, we examined the seed coat epidermal cells by SEM. In wild-type seeds, the epidermal cells are tightly arranged to form a smooth dome-like surface, with the cell walls between neighboring cells tightly sealed. In contrast, in mutant seeds, there were many cracks between the epidermal cells, and many epidermal cells were wrinkled with irregular shape (Fig.   4E, bottom panel). The cells were also generally somewhat smaller than in the wild type, reflecting the overall somewhat smaller size of the seeds themselves (Fig. 4B,E).

Phenotypes of Medicago myb14 Mutants
We isolated two independent myb14 mutants from the Medicago Tnt1 insertion mutant population (Fig. 5A). myb14-1 and myb14-2 harbor Tnt1 insertions in the second exon and the  whereas very low levels of MYB14 transcripts were detected in line myb14-2, indicating that myb14-1 is a null mutant and myb14-2 is a hypomorphic mutant (Supplemental Fig. S8B and   8C). Both myb14-1 and myb14-2 seeds display slightly but noticeably darker red-brown color than wild-type seeds, suggesting that the reduction of MYB14 transcripts in hypomorphic myb14-2 is sufficient to cause a similar phenotype as in the null mutant (Fig. 5B). Compared with the myb5 mutant, the seed color of myb14 is closer to that of the wild type. To exclude the possibility that the different phenotype of the Medicago myb14 is a result of gene redundancy, we used qPCR with two pairs of primers, one primer pair located in the conserved R2R3 domain and the other located in the divergent C-terminal region, to measure the Medicago MYB14 gene copy number. Results from both primer pairs indicated that MYB14 is a single copy gene in the Medicago genome (Supplemental Fig. S11).
myb14 mutants accumulate approximately 50% the PA level of wild-type seeds ( Figure   5C). Similar to myb5 mutants, we could not detect differences in levels of insoluble PAs between myb14 mutant and wild-type seeds by the butanol-HCl method (Supplemental Fig. S7). Based on the same rationale, we conclude that insoluble PA levels in myb14 mutant seed cannot be determined by the butanol-HCl method due to the presence of insoluble anthocyanin moieties.
No differences in mucilage staining were observed in myb14 mutant seeds compared to wild type (Supplemental Fig. S12).
To dissect the genetic interaction between MYB5 and MYB14 in Medicago, we generated the myb5 myb14 double-knockout mutant by crossing myb5-1 and myb14-1. The seed coat color of the double mutant exhibited a more uniform dark red-brown color than the single mutant of myb5, which typically exhibits a lighter red-brown color at one end of the seed (Fig. 5B, arrow).
These results suggest that MtMYB5 and MtMYB14 act redundantly to regulate PA biosynthesis.
The soluble PA content in the double mutant was reduced to approximately 20% of that of the wild type (Fig. 5D).

Mutant Seeds
To gain insights into MYB5-and MYB14-regulated genes in Medicago, we performed transcriptome analysis of hairy roots by Affymetrix microarray analysis. We considered the genes for which the transcript level changed by more than two fold, and with a statistical significance of p<0.05, as differentially expressed genes. In total, 628 and 460 probesets were induced in MYB5 and MYB14 over-expressing hairy roots, respectively (Supplemental Datasets 1 and 2). qRT-PCR analysis was performed on a set of selected genes to confirm the microarray results ( Fig. 6B and 6C). Genes involved in the PA and anthocyanin biosynthesis pathways were highly induced ( Fig. 6A and 6B). Noticeably, ANR, the gene responsible for epicatechin synthesis, was induced by 255-fold (second strongest induced) and 985-fold (strongest induced) in MYB5 and MYB14 overexpressing hairy roots, respectively, consistent with the observation that these hairy roots accumulate large amounts of PAs. One hundred and forty nine probesets were induced more than 2-fold by both MYB5 and MYB14 (Supplemental Dataset 3). Known PA and flavonoid biosynthesis genes, such as ANR, LAR, UGT72L1, LDOX, CHS and F3′H, were among the common induced genes. Interesting, an aldehyde dehydrogenase and a vacuolar acid invertase gene were also highly induced by both MYB5 and MYB14, possibly indicating that sugar metabolism is altered to accommodate the high level of PA biosynthesis in these roots.
Transcripts encoding the transcription factors Mt TT8 (Medtr1g072320.1, the putative homolog of Arabidopsis TT8) and Mt WD40-1 (Pang et al. 2009), which are the integral components of the MBW complex and essential for the activation of PA pathway structural genes, were also up-regulated in transgenic hairy roots. Mt TT8 (probeset Mtr.22479.1.S1_at) was induced 2.6-and 2.1-fold by Mt MYB5 and Mt MYB14, respectively. Mt WD40-1 (Mtr.39774.1.S1_at) was induced 2.6-fold by Mt MYB14, but less than 2-fold by Mt MYB5. Because of the relatively low level of induction of these genes compared to that of the PA pathway structural genes, qRT-PCR was performed to quantify the expression levels of Mt TT8 and Mt WD40-1 in hairy roots. The results confirmed that both Mt MYB5 and Mt MYB14 can induce Mt TT8 and MtWD40-1 (Fig. 6B, C).
One MYB transcription factor gene, Medtr5g079670.1, represented by two probesets (Mtr.16432.1.S1_at and Mtr.34401.1.S1_s_at), was induced 53-to 80-and 10-fold by MYB5 To identify transcript changes underlying the phenotypes of the myb5 and myb14 mutant seeds, we performed Affymetrix microarray analysis. Seeds of mutant and wild-type plants at 12 DAP were dissected from pods for isolation of RNA. Two biological replicates were collected, and transcript levels compared to those of wild-type plants. Nine hundred and ninety one genes (500 down-regulated and 491 up-regulated >2 fold) were differentially expressed in myb5 seeds (Supplemental Dataset 4), whereas only 151 genes were differentially expressed (100 downregulated and 51 up-regulated >2 fold) in myb14 seeds (Supplemental Dataset 5).
We analyzed the 500 down-regulated genes in myb5 seeds by PathExpress (Goffard and Weiller, 2007) to identify the potential metabolic pathways associated with these genes. This indicated that flavonoid biosynthesis and sugar metabolism are the main pathways downregulated in the myb5 mutant seeds (p value <0.01) (Supplemental Table 1 ANR, which is highly induced by MYB5 over-expression in hairy roots, was only slightly downregulated in the myb5 seeds. However, LAR, another key structural gene for PA biosynthesis, was down-regulated more than 6-fold. The down-regulation of the above genes was subsequently confirmed by qRT-PCR from independent biological samples ( Fig. 6D). It is noteworthy that most of the above genes belong to multigene families in the Medicago genome, but only a subset of each gene family was down-regulated in myb5 seeds, suggesting that different gene family members might be involved in different flavonoid biosynthetic pathways. Medtr5g079670.1, the putative repressor highly induced in hairy roots, was the most strongly down-regulated gene in myb5 seeds (Fig. 6D, Supplemental Dataset 4). The strong down-regulation of an anthocyanin repressor was associated with up-regulation of certain anthocyanin biosynthesis genes; two genes encoding anthocyanin acyltransferase and three genes encoding DFR were up-regulated more than two-fold in myb5 seeds (Supplemental Dataset 4).
Genes encoding glycosyl hydrolase and α /β fold hydrolase are down-regulated in Arabidopsis myb5 mutant seeds and were assumed to be associated with mucilage biosynthesis (Gonzalez et al., 2009;Li et al., 2009). Similarly, many genes encoding glycoside hydrolases and other cell wall carbohydrate modifying enzymes are down-regulated in Mt myb5 mutant seeds (Supplemental Table 2). The homolog of one gene in Arabidopsis that has been directly demonstrated to be involved in mucilage biosynthesis (mum2, Supplemental regulated in myb5 seeds. Interestingly, genes encoding the aldehyde dehydrogenase and vacuolar acid invertase identified in MYB5 over-expressing hairy roots were also down-regulated in myb5 seeds, suggesting they are probable targets of MYB5 (Fig. 6D). These enzymes are typically involved in carbohydrate and/or lipid metabolism, suggesting they are probably associated with mucilage biosynthesis in the Medicago seed coat.
Analysis of 100 down-regulated genes in myb14 seeds by PathExpress indicated that flavonoid biosynthesis was likely the only pathway significantly affected (Supplemental Table   3), with ANR, UGT72L1 and CHS being down-regulated more than two-fold in myb14 seed.
Subsequent qRT-PCR analysis confirmed the microarray data (Fig. 6D). One gene encoding flavonoid 3', 5'-hydroxylase (F35H1), which was down-regulated slightly less than 2-fold in myb5 and myb14 seeds (54% of wild-type) on the basis of microarray analysis, was found to be down-regulated by more than 2-fold by qRT-PCR analysis. The results show that loss of function of MYB14 only affects a small set of genes in the PA biosynthesis pathway, consistent with its less severe phenotypic effect than loss of function of MYB5. Unlike myb5 mutant seeds, no gene implicated in mucilage biosynthesis was down-regulated in myb14 seeds.

Promoters in the Presence of Mt TT8 and Mt WD40-1
MYB transcription factors interact with bHLH and WD40 transcription factors to form an MBW complex (Lepiniec et al., 2006). To test the hypothesis that MYB5 and MYB14 form complexes with Medicago bHLH and WD40 transcription factors to regulate the PA biosynthesis pathway, we conducted transient promoter activation assays in Arabidopsis protoplasts. The promoters of Medicago ANR and LAR were cloned in front of the firefly luciferase reporter gene (Fig. 7A).
The reporter constructs were co-transfected into the protoplasts with various effectors driven by the constitutive CaMV 35S promoter (Fig. 7A). In addition to MYB5 and MYB14, the effectors included one bHLH transcription factor Mt TT8 and Mt WD40-1 (Pang et al., 2009).
Both MYB5 and MYB14 alone were able to weakly activate the ANR promoter (Fig.   7B); this contrasts with their inability to interact with the promoter in the yeast one-hybrid assay ( Fig. 1). MYB5 alone had higher activation activity than MYB14 alone (23-fold versus 3.8-fold).
When MYB5 and MYB14 were co-transfected with TT8 and WD40-1, the ANR and LAR promoter activities were strongly activated. We observed a 1,965-fold induction when MYB5 was co-transfected with TT8 and WD40-1 and a 794-fold induction when MYB14 was cotransfected with TT8 and WD40-1. These results support the hypothesis that MYB5 and MYB14 are independently able to form a complex with TT8 and WD40-1 to activate the ANR promoter.
We also observed low transactivation, 2.8-and 1.8-fold, for TT8 and WD40-1 alone, respectively, which, as with the low level of transactivation by MYB5 or MYB14 alone, likely reflects the non-specific formation of complexes with endogenous Arabidopsis activators. All these data are consistent with the operation of a classical MBW ternary complex.
To test whether MYB5 and MYB14 act redundantly in separate ternary complexes, we co-transfected the protoplasts with all four transcription factors, resulting in a 12,400-fold activation of the ANR promoter (Fig. 7B). Because the induction activity of the combination of all four transcription factors is more than four times higher than the sum of the activation by MYB5 and MYB14 ternary complexes with TT8 and WD40-1, we conclude that MYB5 and MYB14 are able to activate the ANR promoter synergistically.
We next tested the activation activities of these transcription factors on the promoter of LAR (Fig. 7C). Similar results were obtained to those with the ANR promoter. A weak activation activity was observed for MYB5 alone (20-fold) and MYB14 alone (2.8-fold). When TT8 and WD40-1 were co-transfected with MYB5 or MYB14, we observed 51-fold and 127-fold induction activities, respectively. Similar to the results with the ANR promoter, the highest induction activity (448-fold) was observed when all four transcription factors were cotransfected. This activation is 2.5 times higher than the sum of the transactivation activities of the MYB5 and MYB14 complexes with TT8 and WD40-1. We therefore conclude that MYB5 and MYB14 also synergistically activate the LAR promoter.
To determine whether the apparent synergistic action between MYB5 and MYB14 reflects the existence of a MYB5/MYB14 complex on the same promoter, we performed bimolecular fluorescence complementation (BiFC) assays between MYB5 and MYB14 in the presence of WD40-1, TT8 and the ANR promoter. We transfected the Arabidopsis protoplasts with MYB5 fused with the C terminal half of EYFP (YFPc:MYB5) and MYB14 fused with the N terminal half (YFPn:MYB14) along with TT8, WD40-1 and the ANR promoter. The complementary YFP signals between MYB5 and MYB14 were readily observed when TT8 and WD40-1 were co-transfected with the BiFC plasmids ( Figure 7D, Top left). In contrast, when TT8 or WD40-1 were omitted, no complementary YFP signal could be observed ( Figure 7D bottom left and middle, respectively), indicating that MYB5 and MYB14 form the complex in a TT8 and WD40-1 dependent manner. It is noteworthy that the complementary YFP signals between MYB5 and MYB14 are multiple spotted signals, suggesting that MYB5 and MYB14 are parts of multiple transcription foci. The complementary YFP signals between MYB5 and TT8 were more diffuse (Fig. 7D, top middle), similar to those of full length EYFP fused with MYB5 (YFPf:MYB5, Fig. 7D, top right), suggesting that the interaction between MYB5 and TT8 is not dependent on transcription foci. Taken together, our results strongly suggest that MYB5 indeed forms a complex with MYB14, and that this is part of a quaternary complex with TT8 and WD40-1 for the high level trans-activation of PA biosynthesis.

Both MYB5 and MYB14 are PA Regulators in Medicago
Medicago MYB5 belongs to a small family of MYB transcription factors characterized by their highly conserved R2R3 domains at the N terminus and two other highly conserved motifs, C1 and C3, at the C terminus. Although members of this family have been shown to be involved in PA biosynthesis, loss of function mutants of these transcription factors result in quite different phenotypes in different species, suggesting that the mechanisms underlying transcriptional control of PA biosynthesis may be different in different species. In Arabidopsis, loss of function of At myb5 does not affect the expression level of genes related to PA biosynthesis. The involvement of At MYB5 in PA biosynthesis is only apparent when At myb5 and At tt2 are both knocked out, indicating that At MYB5 plays only a minor role in PA biosynthesis (Gonzalez et al., 2009). In persimmon, complete loss of PA in non-astringent type (NA-type) fruit is associated with mutation of Dk myb4, and most of the structural genes in PA biosynthesis are turned off (Akagi et al., 2009), suggesting that Dk MYB4 is the major regulator, if not the sole regulator, of PA biosynthesis in persimmon. Interestingly, in petunia, the MYB TF PH4 regulates vacuolar acidification but has no effect on anthocyanin content (Quattrocchio et al., 2006). No PA-related phenotype was reported in ph4 mutants, but mutation of the gene encoding PH5, a proton pump which is the direct target of PH4, drastically reduces the PA content in seeds and produces the typical lighter seed color of PA-deficient mutants (Verweij et al. 2008). In contrast to these observations, Mt myb5 mutants accumulate significantly reduced levels of soluble PAs in the seed coats, but exhibit a dark-red seed coat color (presumably arising from an anthocyanin-related compound, see discussion below), suggesting that the regulatory architecture at the anthocyanin/PA branch in Medicago is distinctly different from that in the other systems described above. In many plants, the color of the seed coat is mainly determined by its anthocyanin and PA contents. Seed coats with high PA and low anthocyanin content exhibit bright brown colors due to PA oxidization. In contrast, seed coats with high anthocyanin content exhibit a dark red color.
In soybean, the recessive i allele, which releases CHS from silencing, results in high anthocyanin accumulation in the seed coat which can be dark brown or even black depending on other loci such as R and T (Tuteja et al., 2004;Yang et al., 2010). In contrast, the dominant I allele, which silences CHS expression and hence blocks both anthocyanin and PA biosynthesis, results in a completely colorless seed coat. In Arabidopsis, most of the mutants isolated from screening for seed coat color display various degrees of reduced color to colorless phenotypes (transparent testa, tt mutants), indicating that these mutations abolish or reduce both anthocyanin and PA biosynthesis. So far in Arabidopsis, only the ban mutant, which harbors a mutation in the ANR gene, contains high anthocyanin content in the seed coat and hence exhibits a dark-red seed coat phenotype (Devic et al., 1999;Xie et al., 2003). The ban phenotype supports the exclusive role of ANR in PA biosynthesis. In soybean, three varieties with red-brown seed coats are associated with reduced ANR mRNA levels (Kovinich et al., 2012).
The tt2 mutant of Arabidopsis exhibits a strong transparent testa phenotype, suggesting that TT2 regulates both the anthocyanin and PA pathways (Nesi et al., 2001). Mutation of At Unlike mutants of their homologs in Arabidopsis, both Mt myb5 and Mt myb14 mutant seeds exhibit darker color than wild type Medicago seeds, with Mt myb5 exhibiting the stronger phenotype in terms of seed color. Due to the fact that these pigments are resistant to acid methanol and aqueous acetone extraction, their identity cannot be determined at present.
However, given the fact that anthocyanin and PA biosynthesis share most of the biosynthetic steps, it is reasonable to assume these pigments are some kind of cross-linked anthocyanin related compound. One plausible hypothesis to interpret the seed color phenotypes is that both MYB5 and MYB14 mainly regulate PA biosynthesis with little or no effect on anthocyanin biosynthesis, allowing for the accumulation of anthocyanins (which are then further processed, for example by oxidation) in the seed coat and hence the darker seed color.

Downstream Targets of MYB5 and MYB14 in Medicago
Gene expression profiling of MYB5 or MYB14 over-expressing hairy roots revealed that both genes strongly induce the expression of structural genes specific to the PA biosynthesis pathway.
ANR was induced more than 250-fold in MYB5 over-expressing hairy root, and 985-fold in MYB14 over-expressing hairy roots. These extremely high values mirror the effects of high over-expression of AtTT2 in Medicago hairy roots (Pang et al., 2008), and are in part explicable in terms of extremely low ANR expression in control tissues. As in At TT2 expressing Medicago hairy roots, UGT72L1, a glycosyltransferase specific for epicatechin (Pang et al., 2008), was also highly induced in MYB5 and MYB14 expressing hairy roots. Besides PA pathway specific genes, genes encoding enzymes functioning upstream of ANR or LAR, such as CHS, F3'H, and DFR, which could potentially be involved in both anthocyanin and PA biosynthesis, were also induced by MYB5 and MYB14. However, compared with ANR, these genes are induced with much lower fold increase, suggesting that they might be induced through a secondary, feedback/forward mechanism. Mt TT8 and Mt WD40-1, the two essential components of the MBW complex, are up-regulated around 2-3-fold, suggesting that there is a self-activating feedback regulation for the complex. In petunia, two MYB anthocyanin activators (DPL and PHZ) also  Striking observations from the current study are that myb5 mutant seeds exhibit a more conspicuous dark-red seed color than myb14 seed, and less soluble PA can be extracted from myb5 seeds. In contrast to the minor role of MYB5 in PA biosynthesis in Arabidopsis, these results suggest that Mt MYB5 plays at least as important a role as Mt MYB14 (orthologous to At TT2) in PA biosynthesis in Medicago. However, examination of the expression levels of major PA pathway structural genes did not reveal significant difference between myb5 and myb14 seeds. Instead, a putative anthocyanin repressor, Medtr5g079670, appeared to be exclusively down-regulated in myb5 seeds. One hypothesis to explain the more conspicuous seed coat color of the myb5 mutant is that the anthocyanin repressor is essential to direct the metabolic flux to PA biosynthesis. In this way, knock-out of myb5 will cause metabolic flux to be redirected to anthocyanin biosynthesis, as well as via the down-regulation of the PA pathway structural genes.
The combination of these two effects results in the stronger phenotype seen in the myb5 seeds.
Another possible hypothesis is that myb5 predominantly regulates some unknown steps, such as epicatechin transport and condensation. The petunia homolog of Mt MYB5 directly regulates a vacuolar proton pump (PH5) which is essential for PA transport (Verweij et al. 2008). It will be interesting to investigate whether a similar proton pump exists in Medicago. An acid vacuolar invertase is strongly down-regulated in myb5 seeds; this might possibly affect the vacuolar osmotic pressure and consequently affect PA accumulation in the vacuoles.

A Model for Regulation of PA Biosynthesis in Medicago
The current model for regulation of PA biosynthesis is mainly based on results from Arabidopsis, where TT2, TT8 and TTG1 form a ternary complex to activate PA biosynthesis genes (Baudry et al., 2004). In grapes, Vv MYB5a and Vv MYB5b can activate the Vv ANR promoter in the presence of At EGL3, suggesting that MYB5 proteins are also part of a similar ternary complex (Deluc et al., 2008). In our transactivation assays using Arabidopsis protoplasts, we observed that Mt MYB5 and Mt MYB14 alone can weakly activate both the ANR and LAR promoters, with MYB5 showing somewhat higher activity than MYB14. When Mt TT8 and Mt WD40-1 were included in the assay, the activation by Mt MYB5 and Mt MYB14 increased dramatically, supporting the ternary complex model. However, the highest transactivation activity was observed when all four transcription factors were co-transfected, and the effects of MYB5 and MYB14 were synergistic rather than additive. That this involves a previously unsuspected quaternary complex is suggested by the observation that BiFC analysis revealed a physical interaction between MYB5 and MYB14, but only in the presence of TT8 and WD40-1. It was recently reported that petunia MYB27 functions as a co-suppressor by forming a quaternary complex with a second activator MYB, bHLH and WD40 TFs (Albert et al., 2014). This parallels our model in which MYB5 and MYB14 can form a quaternary complex with Mt TT8 and Mt WD40-1. Previous work has suggested the presence of two MYB molecules in transcriptional complexes for activation of anthocyanin biosynthesis. For example, Kong et al.
(2012) proposed a model to explain how different configurations of bHLH protein (maize R) dimerization can function as a switch that directs the MBW complex to be tethered to different promoters. In either of the two configurations, there are two MYB molecules (maize C1) in the complex. It is tempting to speculate that the two MYB molecules can be present as a homodimer or, as suggested by the present work and that of Albert et al. (2014), as a hetero-dimer.
Such heterodimers may consist of two transcriptional activators with synergistic activities, or an activator and a repressor. The promoters of ANR and LAR have both MYB and bHLH binding sites; it remains to be determined exactly how the transcriptional complex is assembled on these promoters.
Recently, we discovered that another Medicago MYB transcription factor, PAR, also regulates PA biosynthesis in the Medicago seed coat (Verdier et al., 2012). Unlike loss of function of MYB5 and MYB14, the par mutant exhibits a lighter seed coat color than wild type, indicating that PAR either regulates both anthocyanin and PA biosynthesis or does not negatively regulate the anthocyanin pathway. The mechanism by which PAR, MYB5 and MYB14 together orchestrate PA biosynthesis is currently under investigation.

Plant Materials
Medicago truncatula ecotype R108 was used as wild type for comparison with Tnt1 insertion mutants. The tobacco Tnt1 insertion mutagenized Medicago population was screened to identify MYB5 and MYB14 mutants as described by (Tadege et al., 2008). Seeds were scarified with concentrated sulfuric acid for 10 min, then washed with a large amount of water five times to remove sulfuric acid. Scarified seeds were sterilized with 10% bleach for 10 min and then rinsed five times with sterile water. myb5 mutant seeds were sterilized with 10% bleach without sulfuric acid scarification. Sterilized seeds were vernalized at 4°C for 4 days on moist, sterile filter paper. Vernalized seeds were germinated on filter paper for 5 days before transfer to soil in pots. The plants were grown in a greenhouse set at 12h/12h day/night cycle, 24°C.

Yeast One-Hybrid Screening and Assays
The Medicago WD40-1 promoter reporter yeast strain was generated as described in Verdier et al. (2012). A cDNA library was constructed from RNAs of 10 DAP seeds of M. trucatula A17 using the CloneMiner™ II cDNA Library Construction Kit (Invitrogen). The cDNA library was then cloned into pAG425GPD-ccdB (Alberti et al., 2007) to generate the library for yeast screening. The library plasmid DNA was used to transform the WD40-1 promoter reporter strain.
Positive colonies were selected on yeast leucine dropout medium supplemented with 200 ng/mL aureobasidin A. To reconfirm the Y1H results from screening, full length cDNAs of putative effectors were cloned into pAG425GPD-ccdB and Y1H assays were performed as described in (Verdier et al., 2012).

Complementation of Arabidopsis tt2 with Mt MYB5, Mt PAR and Mt MYB14
The Arabidopsis ANR (Banyuls) promoter was amplified by the Ban-pro-F/Ban-pro-R primer pair from genomic DNA and cloned into pMDC32 (Curtis and Grossniklaus, 2003) at the HindIII and KpnI sites to replace the 35S promoter. The resulting plasmid pMDC32-ProBan was used to generate pMDC32-ProBan:MYB5, pMDC32-ProBan:PAR and pMDC32-ProBan:MYB14 by LR reactions. The resulting binary vectors were used to transform the Arabidopsis tt2 mutant (Salk_005260) by floral dip method (Clough and Bent, 1998).

Plasmid Constructions for Generating MYB5 and MYB14 Over-Expressing Plants
To generate binary vectors for MYB5 and MYB14 over-expression, the cDNA sequences were amplified by RT-PCR from seed RNA using the primer pairs MtMYB5-F/MtMYB5-R and MtMYB14-F/MtMYB14-R, respectively (Supplemental Table 4  http://www.noble.org/Global/Medicagohandbook/pdf/AgrobacteriumRhizogenes.pdf.

RNA Isolation, qRT-PCR and DNA Microarray Analysis
RNAs from plant tissues except seeds were isolated using the RNeasy Plant Mini Kit (Qiagen).
Except for root samples, which were collected from 1 week old seedlings grown on half MS supplemented with 1% sucrose, all other tissues were collected from soil-grown plants. RNAs from seeds were isolated from 12 DAP seeds using Plant RNA Reagent (Invitrogen). Isolated RNAs were treated with DNase I and then purified by RNeasy MinElute Cleanup Kit (Qiagen).
Cleaned RNAs were used to perform reverse transcription with SuperScript® III Reverse Transcriptase (Invitrogen). qRT-PCR analysis was performed on an ABI 7900HT qPCR machine according to the manufacturer's instructions. Three technical replicates were used for each gene and statistics were performed with Excel. Three biological replicates for hairy root samples and two biological replicates for 12 DAP seeds were used to perform microarray analyses using the Medicago GeneChip (Affymetrix) according to the manufacturer's instructions. Raw data were normalized by robust multichip averaging (RMA) method. Presence and absence calls for probe sets were obtained by using the dCHIP algorithm (Li and Wong, 2001). Type I family-wise error rate was calculated by using a Bonferroni corrected P value (threshold 0.05). To identify differentially regulated probe sets, we used a P value threshold of 5% and at least a twofold difference between transformant/mutant lines and their respective controls.

DMACA-and Ruthenium Red-Staining
To visualize PAs, seeds were soaked in water for 2 h, then stained with 1% DMACA solution (1% DMACA in 1:1 methanol: concentrated HCl) for 1 h. Stained seeds were then washed in 70% ethanol for 1 h. To visualize mucilage, seeds were stained with 0.01% aqueous ruthenium red solution for 10 min and then washed in water.

Quantification of PAs
For hairy roots, approximately 0.2g fresh tissue was used for extraction of the soluble PA fraction. For seeds, plant materials were first ground to a powder in liquid nitrogen, and freeze dried for 16 h. Approximately 50 mg of freeze-dried materials were then extracted. Extraction was three times with 1 mL 70% acetone, 0.5% acetic acid, as described by Pang et al. (2008). Assay of soluble PAs (with DMACA reagent) and insoluble PAs (with butanol-HCl was performed as described previously (Pang et al. 2008).

Determination of PA Composition by Normal and Reverse Phase HPLC
All analyses, including size estimation by HPLC followed by post-column derivatization with DMACA, and phloroglucinolysis, were performed as described by Pang et al. (2008).

UPLC Analysis of PA Monomers
Ten mg lyophilized tissue powders were extracted with 80% methanol containing 18 μ g mL −1 umbelliferone (internal standard) for 4 h at 4 °C with constant agitation. Samples were then centrifuged at 3,000 x g for 15 min. The supernatants were transferred to liquid chromatography vials and analyzed by UPLC-ESI-QTOF-MS system (Waters, ACQUITY UPLC) according to the procedure as described in (Pang et al., 2009).

Promoter Transactivation Assays
To construct luciferase reporter constructs, approximate 2 kb genomic sequences upstream of the translation start codons of the ANR and LAR genes were PCR amplified from M. truncatula A17 genomic DNA using the primer pairs ANRP2K-F/ANRP2K-R and LARP2k-F/LARP2K-R, respectively (Supplemental Table 4), and cloned into p2GW7 to replace the 35S promoter in the vector. A firefly luciferase gene was then cloned into this vector to generate the promoter::luciferase reporter construct. The cDNAs of the effector transcription factors were first cloned into pENTRD vector and then cloned into p2GW7 by LR reaction to form the 35S::effector constructs. A Renilla luciferase gene was also cloned into p2GW7 to form the reference gene construct. Arabidopsis protoplasts were isolated and transformed as described by

SUPPLEMENTAL DATA
The following materials are available in the online version of this article.
Supplemental Figure S1. Biosynthesis pathway of proanthocyanidin.  Table 2. Genes putatively involved in mucilage biosynthesis that are downregulated in seeds of the Medicago myb5 mutant. Table 3. PathExpress analysis of genes down-regulated in myb14 mutant seeds (Pathway with P<0.01 is highlighted). Table 4. List of oligonucleotide sequences used in this study.      D, Normal phase HPLC analysis, with post column derivatization with DMACA, of soluble PA fractions from wild-type (R108) and myb5-1 mutant. E, Ruthenium red staining (top panels) and SEM images (bottom panels, all three images are at the same magnification) of seeds of ecotype R108 (WT) and myb5 mutants. C and D, Soluble PA levels in myb14 -1, myb5-1 and myb5-1 myb14-1 (myb5 myb14) mutants as determined by the DMACA method. All quantifications used three independent biological replicates. Error bars indicate standard deviations. **p<0.01, *p<0.05, Student's t test, n=3.  was confirmed by BiFC (Top, middle) and the full length EYFP (YFPf) fused with MYB5 was also included. As a negative control, Arabidopsis At-MYB46 fused with YFPc was used to cotransfect protoplasts with YFPn:MYB14 in the presence of TT8 and WD40-1 (bottom, right).

Supplemental
EYFP signals were rendered as green and chloroplast autofluorescence was rendered as red.   indicates that myb5 mutants accumulate less PA than wild type (WT). C, Soluble PA levels in mature dry seeds of wild-type and myb5 mutant seeds. Three biological replicates were analyzed. Error bars indicate standard deviations. **p<0.01, Student's t test, n=3. D, Normal phase HPLC analysis, with post column derivatization with DMACA, of soluble PA fractions from wild-type (R108) and myb5-1 mutant. E, Ruthenium red staining (top panels) and SEM images (bottom panels, all three images have the same scale bar) of seeds of ecotype R108 (WT) and myb5 mutants. A, Scheme of the flavonoid pathway leading to PA production. Squares denote genes which are up-regulated in MYB5 and MYB14 over-expressing hairy roots based on Affymetrix microarray analysis. Triangles denote genes which are down-regulated in myb5-1 mutant seeds. Dots denote genes which are down-regulated in myb14-1 mutant seeds. B and C qRT-PCR analysis to confirm the up-regulation of genes in MYB5 and MYB14 overexpressing hairy roots, respectively. n.d., not detected. Transcript levels were normalized to Mt Actin2 transcript levels. D, qRT-PCR to measure the transcript levels of down-regulated genes in seeds (12DAP) of the myb5-1 and myb14-1 mutants. ALDH: aldehyde dehydrogenase; F35H: flavonoid 3′,5 ′hydroxylase; AVT: acid vacuolar invertase; TT8: Mt TT8; WD40-1: Mt WD40-1. All qRT-PCR analyses were performed on one independent biological replicate (separate from those used for the microarray analysis), giving technical triplicates; error bars indicate standard deviations.