Functional Characterization of MtrGSTF7, a Glutathione S-Transferase Essential for Anthocyanin Accumulation in Medicago truncatula

Flavonoids are essential compounds widespread in plants and exert many functions such as defence, definition of organ colour and protection against stresses. In Medicago truncatula, flavonoid biosynthesis and accumulation is finely regulated in terms of tissue specificity and induction by external factors, such as cold and other stresses. Among flavonoids, anthocyanin precursors are synthesised in the cytoplasm, transported to the tonoplast, then imported into the vacuole for further modifications and storage. In the present work, we functionally characterised MtrGSTF7, a phi-class glutathione S-transferase involved in anthocyanin transport to the tonoplast. The mtrgstf7 mutant completely lost the ability to accumulate anthocyanins in leaves both under control and anthocyanin inductive conditions. On the contrary, this mutant showed an increase in the levels of soluble proanthocyanidins (Pas) in their seeds with respect to the wild type. By complementation and expression data analysis, we showed that, differently from A. thaliana and similarly to V. vinifera, transport of anthocyanin and proanthocyanidins is likely carried out by different GSTs belonging to the phi-class. Such functional diversification likely results from the plant need to finely tune the accumulation of diverse classes of flavonoids according to the target organs and developmental stages.


Introduction
Flavonoids comprise a large group of secondary metabolites, including anthocyanins, proanthocyanidins (PAs), flavonols and isoflavones, which contribute in various ways to the growth and survival of plants [1][2][3]. Anthocyanidins are attractants for pollinators [4], protect plants against UV light damage and serve as defense molecules toward pathogens. Anthocyanins and simple monomeric and/or oligomeric PAs have been reported to protect against free radical injury and cardiovascular disease in humans [4]. PAs also act as antinutrients affecting both dietary protein and micronutrient availability in animals [5].
As other flavonoids, anthocyanin and the flavan 3-ols, catechins and epicatechins, the building blocks of PAs, are synthesised in the cytosol but accumulate within the vacuole, and therefore the efficiency of their transportation across the tonoplast affects their accumulation [6]. At least two different routes of flavonoid transport have been proposed: vesicle trafficking and transporter-mediated [7]. GSTs have been recognised as flavonoid transporters in a number of different taxa [6]. Moreover, they have been Tnt1 flanking regions from both lines were cloned by thermal asymmetric inte PCR (TAIL-PCR), and their genomic positions were identified within the R108 refe sequence [24]. It is noteworthy that one insertion from R45 and one from line R39 mapped to close genomic positions, both within the MtrunR108HiC_012760, whi codes for the MtrGSTF7 phi class glutathione S-transferase protein [13]. Sequence t site (STS) markers were developed for each cloned insertion (see Figure 2A) and tes sibs of the two mutant lines. Both STS markers tagging the MtrGSTF7 insertions cogated with the phenotype according to mono-factorial recessive mutation models. findings were also confirmed on progenies produced by selfed heterozygous sibs o and R39 lines ( Figure 2B). Tnt1 flanking regions from both lines were cloned by thermal asymmetric interlaced PCR (TAIL-PCR), and their genomic positions were identified within the R108 reference sequence [24]. It is noteworthy that one insertion from R45 and one from line R39 were mapped to close genomic positions, both within the MtrunR108HiC_012760, which encodes for the MtrGSTF7 phi class glutathione S-transferase protein [13]. Sequence tagged site (STS) markers were developed for each cloned insertion (see Figure 2A) and tested on sibs of the two mutant lines. Both STS markers tagging the MtrGSTF7 insertions co-segregated with the phenotype according to mono-factorial recessive mutation models. These findings were also confirmed on progenies produced by selfed heterozygous sibs of R45 and R39 lines ( Figure 2B).
Anthocyanins were undetectable by spectrophotometric analysis of leaves of both R45 and R39 homozygous individuals. A segregating progeny of an R45 heterozygous sib was cultivated in vitro on anthocyanin inducing medium (6% sucrose). Even under these conditions, we could not detect anthocyanins in leaves of R45 homozygous individuals, whereas individuals heterozygous for the insertions accumulated anthocyanins in leaves, though at levels significantly lower than wild-type sibs ( Figure 3). R45 mutant sibs did not show qualitative differences for PAs when glandular trichomes and seed coats were stained with 4-(dimethylamino)-cinnamaldehyde (DMACA) (data not shown). Anthocyanins were undetectable by spectrophotometric analysis of leaves of both R45 and R39 homozygous individuals. A segregating progeny of an R45 heterozygous sib was cultivated in vitro on anthocyanin inducing medium (6% sucrose). Even under these conditions, we could not detect anthocyanins in leaves of R45 homozygous individuals, whereas individuals heterozygous for the insertions accumulated anthocyanins in leaves, though at levels significantly lower than wild-type sibs ( Figure 3). R45 mutant sibs did not show qualitative differences for PAs when glandular trichomes and seed coats were stained with 4-(dimethylamino)-cinnamaldehyde (DMACA) (data not shown).  Notably, spectrophotometric analysis of extracts from seeds harvested from R45 homozygotes and heterozygotes showed a significant increase in soluble PAs compared to the wild-type seeds (t-test, p < 0.05), whereas the levels of insoluble PAs did not change ( Figure 4). Notably, spectrophotometric analysis of extracts from seeds harvested from R45 homozygotes and heterozygotes showed a significant increase in soluble PAs compared to the wild-type seeds (t-test, p < 0.05), whereas the levels of insoluble PAs did not change ( Figure 4).
Previous studies have identified 9 phi GSTs in M. truncatula [13]. Based on sequence homology, we analysed the genetic distances among these proteins and phi GSTs characterised in other species ( Figure 5). containing medium. WT is wild-type R108. HET and MUT stands for heterozygous and homozygous referred to Tnt1 insertion in MtrGSTF7. Anthocyanins could not be detected in MUT. Data indicate mean values ±SD of four replicates (*** p < 0.001).
Notably, spectrophotometric analysis of extracts from seeds harvested from R45 homozygotes and heterozygotes showed a significant increase in soluble PAs compared to the wild-type seeds (t-test, p < 0.05), whereas the levels of insoluble PAs did not change ( Figure 4). Previous studies have identified 9 phi GSTs in M. truncatula [13]. Based on sequence homology, we analysed the genetic distances among these proteins and phi GSTs characterised in other species ( Figure 5).  M. truncatula phi GSTs were assigned to four clusters. MtrGSTF7 grouped in cluster I with GSTs essential for anthocyanin transport from the cytoplasm to the vacuole. The Malus domestica MdGST6 and the V. vinifera VVGST4 which complement both the anthocyanin and PA deposition in Arabidopsis tt19 mutant belong to this cluster [19].
MtrGSTF6, MtrGSTF8 and MtrGSTF9 were in cluster 2 along with V. vinifera VvGST3, which complements PAs but not anthocyanin accumulation defect in tt19 mutant [18]. The remaining phi GSTs from M. truncatula were positioned in two additional clusters (3 and 4), which showed the least distance from VvGST1 [18] and bz2 [14] tau GSTs.

Expression Analysis
Several studies have demonstrated that phi GSTs transcript levels are differentiated among organs and developmental stages. The gene expression atlas of Medicago organs [25] was interrogated for all available phi GSTs probes. MtrGSTF9 and 4 showed the highest levels of tissue specificity being readily detectable only in roots and almost not detectable in other organs. MtrGSTF8 was expressed at high levels almost exclusively in seeds M. truncatula phi GSTs were assigned to four clusters. MtrGSTF7 grouped in cluster I with GSTs essential for anthocyanin transport from the cytoplasm to the vacuole. The Malus domestica MdGST6 and the V. vinifera VVGST4 which complement both the anthocyanin and PA deposition in Arabidopsis tt19 mutant belong to this cluster [19].
MtrGSTF6, MtrGSTF8 and MtrGSTF9 were in cluster 2 along with V. vinifera VvGST3, which complements PAs but not anthocyanin accumulation defect in tt19 mutant [18]. The remaining phi GSTs from M. truncatula were positioned in two additional clusters (3 and 4), which showed the least distance from VvGST1 [18] and bz2 [14] tau GSTs.

Expression Analysis
Several studies have demonstrated that phi GSTs transcript levels are differentiated among organs and developmental stages. The gene expression atlas of Medicago organs [25] was interrogated for all available phi GSTs probes. MtrGSTF9 and 4 showed the highest levels of tissue specificity being readily detectable only in roots and almost not detectable in other organs. MtrGSTF8 was expressed at high levels almost exclusively in seeds at the first developmental stages. MtrGSTF7 and MtrGSTF1 transcripts were readily detectable in aerial organs: MtrGSTF7 was almost undetectable in seeds while a decreasing trend was observed for MtrGSTF1.
In M. truncatula, high amounts of PAs accumulate in the seed coat. Phi GST expression data in seed coat indicate a high expression for MtrGSTF6, MtrGSTF8 and MtrGSTF1, suggesting that these three GSTs are possibly involved in PA accumulation in this tissue ( Figure 6). at the first developmental stages. MtrGSTF7 and MtrGSTF1 transcripts were readily detectable in aerial organs: MtrGSTF7 was almost undetectable in seeds while a decreasing trend was observed for MtrGSTF1.
In M. truncatula, high amounts of PAs accumulate in the seed coat. Phi GST expression data in seed coat indicate a high expression for MtrGSTF6, MtrGSTF8 and MtrGSTF1, suggesting that these three GSTs are possibly involved in PA accumulation in this tissue ( Figure 6). Next, we analysed in silico the expression of phi GSTs in leaves of M. truncatula overexpressing the transcription factor LAP1 and showing hyperaccumulation of anthocyanins in leaves [26]. As shown in Figure 7, only MtrGSTF7 was upregulated in these conditions, while the expression of other GSTs was nearly unchanged.  Next, we analysed in silico the expression of phi GSTs in leaves of M. truncatula overexpressing the transcription factor LAP1 and showing hyperaccumulation of anthocyanins in leaves [26]. As shown in Figure 7, only MtrGSTF7 was upregulated in these conditions, while the expression of other GSTs was nearly unchanged. Since glandular trichomes are known sites of PA accumulation, we interrogated the Trichome expression database [27] for phi GST accumulation. Figure 8 reports the expression levels of M. truncatula phi GSTs in glandular and non-glandular trichomes. Higher expression was detected for MtrGSTF1 and MtrGSTF6 with log2FC value below 2 between the two types of trichomes. All other MtrGSTFs showed lower expression values. This was expected for MtrGSTFs with organ-specific expression in root and seeds. Interestingly, MtrGSTF7, which is highly expressed in leaves, showed low expression in trichomes. A redirection of metabolic flux from anthocyanin to flavonol biosynthesis has been demonstrated in the Arabidopsis tt19 mutant background [15]. It is noteworthy that the metabolic shift is not associated with a reduced expression of anthocyanin biosynthetic genes which are instead expressed at higher levels in tt19 compared to wt plants. A similar picture was observed in MtrGSTF7 mutant lines, in which the dramatic reduction in the MtrGSTF7 transcripts levels was paralleled neither from that of early (C4H, 4CL, CHS, CHI and F3′H) nor from that of late (DFR, ANS and ANR) key genes of the phenylpropanoid Since glandular trichomes are known sites of PA accumulation, we interrogated the Trichome expression database [27] for phi GST accumulation. Figure 8 reports the expression levels of M. truncatula phi GSTs in glandular and non-glandular trichomes. Higher expression was detected for MtrGSTF1 and MtrGSTF6 with log 2 FC value below 2 between the two types of trichomes. All other MtrGSTFs showed lower expression values. This was expected for MtrGSTFs with organ-specific expression in root and seeds. Interestingly, MtrGSTF7, which is highly expressed in leaves, showed low expression in trichomes. Since glandular trichomes are known sites of PA accumulation, we interrogated the Trichome expression database [27] for phi GST accumulation. Figure 8 reports the expression levels of M. truncatula phi GSTs in glandular and non-glandular trichomes. Higher expression was detected for MtrGSTF1 and MtrGSTF6 with log2FC value below 2 between the two types of trichomes. All other MtrGSTFs showed lower expression values. This was expected for MtrGSTFs with organ-specific expression in root and seeds. Interestingly, MtrGSTF7, which is highly expressed in leaves, showed low expression in trichomes. A redirection of metabolic flux from anthocyanin to flavonol biosynthesis has been demonstrated in the Arabidopsis tt19 mutant background [15]. It is noteworthy that the metabolic shift is not associated with a reduced expression of anthocyanin biosynthetic genes which are instead expressed at higher levels in tt19 compared to wt plants. A similar picture was observed in MtrGSTF7 mutant lines, in which the dramatic reduction in the MtrGSTF7 transcripts levels was paralleled neither from that of early (C4H, 4CL, CHS, CHI and F3′H) nor from that of late (DFR, ANS and ANR) key genes of the phenylpropanoid  A redirection of metabolic flux from anthocyanin to flavonol biosynthesis has been demonstrated in the Arabidopsis tt19 mutant background [15]. It is noteworthy that the metabolic shift is not associated with a reduced expression of anthocyanin biosynthetic genes which are instead expressed at higher levels in tt19 compared to wt plants. A similar picture was observed in MtrGSTF7 mutant lines, in which the dramatic reduction in the MtrGSTF7 transcripts levels was paralleled neither from that of early (C4H, 4CL, CHS, CHI and F3 H) nor from that of late (DFR, ANS and ANR) key genes of the phenylpropanoid pathway when compared to wt plants grown under controlled conditions (Figure 9). However, when the same plants were grown under cold condition (cold greenhouse with average temperatures of 10 • C, see Figure 1A,B,D,E), an increase in the expression levels of DFR and a decrease in those of MtrGSTF6 was detected in R45 leaves ( Figure 10). Since R45 showed an increment in the levels of soluble PAs in the seeds, we decided to investigate the expression patterns of key genes as above along with that of LAR and UGT72L1 in wt and R45 pods collected at an early stage (6 days after pollination) of their development when immature seeds therein stained positively with DMACA. As shown in Figure 11, the steady state mRNA levels were not statistically different for all the genes investigated except for LAR and UGT72L1. The former codes for the enzyme are not only responsible for the synthesis of catechin [28,29] but also for the control of the degree of PA polymerisation in that it promotes the accumulation of soluble rather than insoluble PAs [30]. The latter codes for an epicatechin-specific glucosyltransferase, which is expressed in the seed coat of M. truncatula [31]. pathway when compared to wt plants grown under controlled conditions ( Figure 9). However, when the same plants were grown under cold condition (cold greenhouse with average temperatures of 10 °C, see Figure 1A,B,D,E), an increase in the expression levels of DFR and a decrease in those of MtrGSTF6 was detected in R45 leaves ( Figure 10). Since R45 showed an increment in the levels of soluble PAs in the seeds, we decided to investigate the expression patterns of key genes as above along with that of LAR and UGT72L1 in wt and R45 pods collected at an early stage (6 days after pollination) of their development when immature seeds therein stained positively with DMACA. As shown in Figure  11, the steady state mRNA levels were not statistically different for all the genes investigated except for LAR and UGT72L1. The former codes for the enzyme are not only responsible for the synthesis of catechin [28,29] but also for the control of the degree of PA polymerisation in that it promotes the accumulation of soluble rather than insoluble PAs [30]. The latter codes for an epicatechin-specific glucosyltransferase, which is expressed in the seed coat of M. truncatula [31].

Restoration of Anthocyanin Transport Deficiency of the Arabidopsis tt19 Mutant
Arabidopsis tt19 mutant is defective in anthocyanin accumulation in vegetative tissues and shows pale testa colour in the seeds [15]. To better elucidate MtrGSTF7 function in vivo, we adopted a functional complementation approach on tt19-1 by generating transgenic tt19-1 plants expressing MtrGSTF7 under the constitutive promoter 35S from CaMV. Seedlings of transgenic and control plants were grown on high sucrose substrate to induce anthocyanin accumulation. As shown in Figure 12, the 35S::MtrGSTF7 construct in tt19 fully restored the anthocyanin accumulation in vegetative organs but had no effect on the levels of soluble PAs in seeds ( Figure 12).

Restoration of Anthocyanin Transport Deficiency of the Arabidopsis tt19 Mutant
Arabidopsis tt19 mutant is defective in anthocyanin accumulation in vegetative tissues and shows pale testa colour in the seeds [15]. To better elucidate MtrGSTF7 function in vivo, we adopted a functional complementation approach on tt19-1 by generating transgenic tt19-1 plants expressing MtrGSTF7 under the constitutive promoter 35S from CaMV. Seedlings of transgenic and control plants were grown on high sucrose substrate to induce anthocyanin accumulation. As shown in Figure 12, the 35S::MtrGSTF7 construct in tt19 fully restored the anthocyanin accumulation in vegetative organs but had no effect on the levels of soluble PAs in seeds ( Figure 12).

Discussion
Anthocyanins are part of the flavonoid family, and they are the final products of a specific branch leading to the formation of flavonols, phlobaphenes and proanthocyanidins as well [32]. This class of metabolites confers resistance to several stresses in plant species via ROS signalling [33]. The biosynthesis of anthocyanins is extensively characterised, and both biosynthetic and regulatory genes have been identified in several plant species [34] showing a high degree of conservation. Concerning the transport of anthocyanins to the vacuole, two main mechanisms have been proposed, one using membrane transporters, and the second one involving the formation of vesicles which will merge with the tonoplast. In both models, the transport is assisted by GST. In particular, GST of the phi group are those involved in anthocyanin transport, with GSTs also having a plethora of different roles [35]. Biochemical assay has proved the binding of GST to anthocyanin in a few species, including Arabidopsis tt19; the conservation of the anthocyanin binding function was proved for GST from several other species by complementation analysis of the Arabidopsis tt19 mutant. Among these are FL3 from carnation [36], PfGST1 from Perilla frutescens [37], VvGST1 and VvGST4 from grapevine [18,38], CkmGST3 from Cyclamen [39], CMGSTF12 from Camelina sativa [40], LcGST4 from Litchi chinensis [16], PpRiant1 from peach [41], FvRAP from strawberry [42], IbGSTF4 from sweet potato [43], MdGSTF6 from apple [19] and AcGST1 from kiwifruit [44].
Here, we report on the cloning and functional characterization of MtrGSTF7. M. truncatula functional genomics via mutant analysis has proved very powerful in understanding several aspects of plant biology of legumes, with particular focus on nodulation [45] and biosynthesis of secondary metabolites [46]. In this context, the presence of a leaf red blot is a useful marker to detect mutations in the anthocyanin pathway. In our work, the visual screening of a Tnt1 mutagenized M. truncatula collection [23] led to the original identification of two mutants lacking the anthocyanin pigmentation of the leaf typical of the R108 genotype. The anthocyanin content remained negligible even after sucrose induction in in vitro grown plants of the mutant for MtrGSTF7 compared to the wild-type R108. The cloning of the gene responsible for the mutation undertaken by TAIL-PCR analysis of the transposon flanking regions led to the identification of MtrGSTF7. The two independent knock-out mutants, though lacking the ability to accumulate anthocyanins, are able to store PAs in both seed coat and glandular trichomes. Expression analysis by qRT-PCR of key biosynthetic genes of the phenylpropanoid pathway in leaves of plants grown under controlled conditions, while confirming a dramatic decrease in the levels of MtrGSTF7 in the mutant line, did not show any significant difference between this and wt for all the other genes considered. Conversely, under cold conditions, differences emerged between R45 and wt plants, namely the upregulation of DFR in R45 and that of MtrGSTF6 in wt leaves. Although both these results call for an in depth characterisation of both transcriptional profile and function of these genes, they let us argue that the block of the anthocyanin pathway due to the knocking out of MtrGSTF7 causes a reprogramming of the metabolic pathway in leaves only under environmental conditions that promote flavonoid biosynthesis. Accurate targeted metabolomic analyses by HPLC-MS would help us to address this issue.
A thorough M. truncatula GST gene family characterisation has revealed the presence of nine phi GSTs with partially overlapping expression patterns. It is noteworthy that while the expression of MtrGSTF2-3 is mainly root-specific and that of MtrGSTF8 is concentrated in the last stages of seed development, the other M. truncatula GSTFs show overlapping patterns of expression in vegetative tissues. However, the observation that MtrGSTF7 knockout totally abolishes the anthocyanin accumulation in vegetative tissues demonstrates that other M. truncatula phi GSTs, although being co-expressed with MtrGSTF7, are unable to functionally complement it. Several hypotheses could explain such a conclusion. First, other phi GSTs have a non-redundant function, for example they are not able to bind anthocyanin. Second, other GSTs are engaged in anthocyanin transport downstream of MtrGSTF7. One possibility would be that MtrGSTF7 could transform anthocyanin to a form recognisable by other GSTs. Enzymatic studies in Petunia and Arabidopsis have excluded that enzymatic modification of anthocyanin mediated by GSTs is needed for their binding and transport. Other studies have demonstrated that anthocyanin can also be transported by other transporters, such as MATE2 [47], and that, differently from MtrGSTF7, their knockout does not impair completely anthocyanin accumulation. These data lend support to the hypothesis of anthocyanin transport as a sequence of events, with MtrGSTF7 working at the early stages of this sequence. At the same time, MtrGSTF7 mutant seeds do not show loss in PA content, highlighting the specific involvement of MtrGSTF7 in the transport of one subclass of flavonoid, the anthocyanin. Interestingly, only MtrGSTF7 expression notably increases upon overexpression of the MYB transcription factor LAP1, with LAP1 being an inducer of anthocyanin synthesis and accumulation in M. truncatula and M. sativa [26]. The other phi GSTs expressed in leaves do not show significant expression changes in LAP1 overexpressing leaves. At the same time, the expression of MtrGSTF7 is low in glandular trichomes, which accumulate PAs compared to whole leaf. Other GSTs, such as MtrGSTF1 and MtrGSTF6, showed higher expression in trichomes. These GSTs are also expressed in other sites of preferential PA accumulation (e.g., seeds), in which MtrGSTF7 expression is barely detectable. All these data suggest that MtrGSTF7 expression is associated with anthocyanin but not with PA accumulation. Better still, the increase in the levels of soluble PAs in the seeds of R45, coupled to the increase in the steady state levels of LAR and UGT72L1 genes in the pods at early stage of R45 with respect to those from wt, are all evidence suggesting that knocking out of MtrGSTF7 causes a flux diversion from anthocyanin towards PAs in M. truncatula reproductive organs. In turn, these results pave the way to future time course analyses to compare the metabolic and molecular profiles between R45 and wt developing seeds.
The complementation experiment of Arabidopsis tt19 mutant showed that MtrGSTF7 is able to rescue the accumulation of anthocyanins in plant tissues but does not restore the presence of soluble PAs in the seeds. Thus, the transport mechanism in which MtrGSTF7 is involved is similar to what happens in Arabidopsis but limited to the anthocyanins. The tt19 complementation assay is able to discriminate between GST transporters of anthocyanins and GST transporters of both anthocyanins and PAs. For instance, AN9, PpGST, LcGST4, Bract1, RAP, ScGST3, MdGSTF6, GhGSTF12, LhGST and PpGST1 only complement anthocyanin accumulation in vegetative tissues [16,19,25,41,42,[48][49][50][51][52]. Differently, heterologous expression of Camelina sativa CMGSTF12, grape VviGST4, radish RsGST1, tea CsGSTa, CsGSTb and CsGSTc in Arabidopsis tt19 mutant not only complements the anthocyanin accumulation in vegetative tissues but also complements the brown pigmentation in the seed coat [17,18,40,44]. Therefore, in M. truncatula, either a fine subfunctionalisation between GSTs involved in the transport of terminal branches of the flavonoid pathway occurs, or the GST responsible for the transport of both compounds has not been identified yet. Embracing the latter hypothesis, we note that the M. truncatula genome harbours several phi type GSTs, some of which, namely MtrGSTF1, 6 and 8, have high expression in seeds. Further analysis, based on mutant characterisation, needs to be undertaken to unravel the function and organ specificity of the above mentioned M. truncatula genes in flavonoid transport under different environmental conditions. Finally, we note that while the writing of the current manuscript was underway, the paper by Wang et al. [17] concerning the functional characterisation of MtrGSTF7 as an anthocyanins specific transporter has been published. The present study not only reinforces but also provides additional evidence with respect to the above-mentioned study about the functional role of MtrGSTF7.  [53] with and without 50 g/L of additional sucrose (final concentrations 1% and 6% respectively). A. thaliana plants, Columbia and tt19-1 mutant [15], were grown in a growth chamber under the same conditions used for M. truncatula. Samples were collected from plants and immediately frozen in liquid nitrogen. Frozen samples were stored at −80 • C until analysis.

Quantification of Anthocyanins and PAs
DMACA staining was used for qualitative estimation of PA accumulation plant organs as well as the analyses of soluble and insoluble PAs were performed according to [54]. Anthocyanin extraction and quantification were carried out as reported in [55].

Identification of the Mutated Gene
Tnt1 insertions in R45 and R39 mutants were recovered using a TAIL-PCR protocol as described in the M. truncatula handbook [56]. TAIL-PCR products were inserted in pGEM-T easy vector and cloned in E. coli XL1-blue cells.  Table 1. An aliquot of 2 µL of 1:10 diluted cDNA was used in the PCR reaction, which was performed using the BlasTaq 2X qPCR Mater Mix (ABM, Richmond, Canada) and carried out in an ABI PRISM 7300 SDS apparatus (Applied Biosystems) according to the following cycling parameters: an initial step at 95 • C for 3 min, followed by 40 cycles each, including a step at 95 • C, 15 s and a step at 60 • C, 1 min. A melting curve was added after each run.
For each gene, four technical replicates were amplified. The PCR efficiency for each primer pair was tested as reported in [58]. Gene expression quantification was calculated using the (2 −∆Ct ) method as reported in [59] with actin as a reference gene. Three biological replicates were analysed, and the significance of the differences between mean expression values was analysed by t-test.

In Silico Expression Analysis
MtrGSTFs expression was investigated by analyzing data available for two microarray experiments on plant tissues and seed coat [25,31]. M. truncatula gene expression data was downloaded from the Medicago Gene Expression Atlas web server [60] currently available at https://medicago.toulouse.inrae.fr/GEA, accessed on 25 February 2022 [61]. Expression data in glandular and non-glandular trichomes was downloaded from the TrichOME database currently available at https://trichome.zhaolab.org/trichomedb/, accessed on 10 December 2021 [27].

Phylogenetic Analysis
Sequences of GST proteins related to flavonoid transport were retrieved from the GenBank database. Accession numbers are shown in Figure 5 caption. M. truncatula GSTs of the phi class in the R108 genotype were retrieved based on previous genome wide analysis [13] by Blastp of A17 (M. truncatula v4.01 genome) proteins on R108 v1.0 genome. Proteins were aligned using ClustalW and a phylogenetic tree was constructed with the neighbor joining method (1000 replications of bootstrap test) and p-distance substitution model, pairwise deletion using the MEGA 6 program.

Arabidopsis tt19 Complementation
The ORF of MtrGSTF7 was PCR amplified from leaves cDNA of M. truncatula cv. R108 obtained by standard protocols. Primer GST_ATG_FW (5 -ATG GTG GTG AAA GTT TAT GGT TCA-3 ) and GST_TAA_REV (5 -TTA CTT AGC CAA TTC CTT CAA C-3 ) were used. The PCR product was cloned into RBCTA cloning vector (RBC Biosciences, Taiwan) according to the manufacturer's instructions. Plasmids were isolated from positive colonies with an Innuprep plasmid mini kit (Analytik Jena, Germany) and sequenced by Sanger methods with an ABI3130 sequencing machine. Sequences were verified by Blast analysis with the available genomic resources of M. truncatula; the correct clone was used for sub-cloning into the plant binary vector pBI121 by substituting the GUS gene (XmaI-SacI fragment) with the GST gene obtaining the plasmid pBI121-GST7, which was verified by sequencing with primers 35sFW (5'-CTA TCC TTC GCA AGA CCC TTC-3') and M13 (-21) forward. The pBI121-GST7 plasmid was transferred into Agrobacterium tumefaciens GV3101 cells by electroporation. Colonies were verified by PCR, and the correct colony was stored with glycerol at −80 • C and used for plant transformation via the floral dip method [62]. The Arabidopsis mutant tt19-1 [15] was used for transformation. Transgenic plants were selected on MS plates added with kanamycin 50 mg/L, two independent homozygous lines for a single T-DNA insertion were obtained in T3 generation and seeds were used for further phenotypic analysis. Data Availability Statement: All references to publicly archived datasets were reported in the Section 4 of the manuscript.