Integrated metabolite and transcript profiling identify a biosynthetic mechanism for hispidol in Medicago truncatula cell cultures.

Metabolic profiling of elicited barrel medic (Medicago truncatula) cell cultures using high-performance liquid chromatography coupled to photodiode and mass spectrometry detection revealed the accumulation of the aurone hispidol (6-hydroxy-2-[(4-hydroxyphenyl)methylidene]-1-benzofuran-3-one) as a major response to yeast elicitor. Parallel, large-scale transcriptome profiling indicated that three peroxidases, MtPRX1, MtPRX2, and MtPRX3, were coordinately induced with the accumulation of hispidol. MtPRX1 and MtPRX2 exhibited aurone synthase activity based upon in vitro substrate specificity and product profiles of recombinant proteins expressed in Escherichia coli. Hispidol possessed significant antifungal activity relative to other M. truncatula phenylpropanoids tested but has not been reported in this species before and was not found in differentiated roots in which high levels of the peroxidase transcripts accumulated. We propose that hispidol is formed in cell cultures by metabolic spillover when the pool of its precursor, isoliquiritigenin, builds up as a result of an imbalance between the upstream and downstream segments of the phenylpropanoid pathway, reflecting the plasticity of plant secondary metabolism. The results illustrate that integration of metabolomics and transcriptomics in genetically reprogrammed plant cell cultures is a powerful approach for the discovery of novel bioactive secondary metabolites and the mechanisms underlying their generation.


INTRODUCTION
Phenylpropanoid metabolism encompasses a complex network of branching biochemical pathways that collectively provide plants with thousands of compounds that have diverse functions in plants, most notably in defense, such as cell wall strengthening and repair (e.g.. lignin and suberin), antimicrobial activity (e.g. furanocoumarin, pterocarpan and isoflavonoid phytoalexins), and as signaling compounds such as luteolin (Peters et al., 1986) and apigenin (Peters and Long, 1988). In the core pathway, phenylalanine is converted into p-coumaroyl-CoA, and the condensation of this molecule with 3 molecules of malonyl-CoA, via chalcone synthase (CHS) activity, yields 4,2',4',6'-tetrahydroxychalcone (naringenin chalcone), which is the primary C15 flavanone skeleton from which a large number of flavonoid subclasses diverge (Fig. 1). In legumes, the 6'-deoxy chalcone isoliquiritigenin is produced through the concerted activity of CHS and chalcone reductase (CHR; (Ralston et al., 2005). This trihydroxychalcone can be further metabolized by chalcone isomerase (CHI) to form liquiritigenin, a precursor for a range of 5-deoxy flavones, flavonols and isoflavones (Fig. 1). Exceptions to this biosynthetic route are the aurones, a flavonoid subclass which can be directly synthesized from isoliquiritigenin (Strack, 1997). Aurones (Fig. 1) serve significant roles in flower pigmentation (Nakayama et al., 2001) and defense responses (Paré et al., 1991). In humans, aurones have recently drawn much attention for their therapeutic potential as anticancer, antidiabetic, antibacterial, antiparasitic and antihormonal agents (Boumendjel, 2003). While flavone, flavonol and isoflavone biosynthetic pathways have been extensively studied using different molecular and biochemical approaches, only a few studies on aurone biosynthesis have been reported.
Plant peroxidases (EC 1.11.17,PRXs) are ubiquitous, heme-containing glycoproteins that catalyze the oxidation of diverse organic and inorganic substances at the expense of hydrogen peroxide (H 2 O 2 ). Higher plants possess a number of PRX isoenzymes that are usually classified as anionic, neutral or cationic based upon their isoelectric point (Barz et al., 1990). Anionic and neutral PRXs are usually cell wall bound, and cationic forms are typically confined to the vacuole (Kawalleck et al., 1995). Because of their location, anionic and neutral PRXs are believed to be mainly involved in plant defense. PRXs can oxidize vacuolar phenolic pools, and also play key roles in the polymerization steps involved in lignification and suberization of plant cell walls (Chittoor et al., 1997). Functional characterizations of PRXs have been based primarily upon genomic sequence mining and in vitro biochemical assays. However, the extremely wide substrate specificities of PRXs, the high number of PRX genes, and the diversity of PRX structures raise important questions about their true in vivo substrates and functions.
Large-scale metabolite profiling (metabolomics) is a powerful tool for analyzing metabolism and gene function (Sumner et al., 2003;Bino et al., 2004). Many early metabolomics studies were focused upon primary metabolites for plant genotyping (Taylor et al., 2003), detection of silent phenotypes in transgenic potato (Weckwerth et al., 2004), and the examination of stress responses (Broeckling et al., 2005). Parallel profiling of transcript and metabolites has also been applied to study the effect of cold acclimation (Kaplan et al., 2004), phosphorus stress (Hernandez et al., 2007) and arbuscular mycorrhizal interactions (Schaarschmidt et al., 2007) in plants. More recently, the profiling of natural products has also been incorporated into the metabolomics approach and integrated with transcriptome analysis to identify novel gene functions associated with flavonoid biosynthesis (Tohge et al., 2005;Naoumkina et al., 2007;Farag et al., 2008;Yonekura-Sakakibara et al., 2008). Such an integrated approach represents a powerful platform for the clarification of gene function in plant secondary metabolism (Fridman and Pichersky, 2005).
Barrel medic (Medicago truncatula) is a rapidly developing model for legume biology (Cook, 1999;Young et al., 2005) and an excellent species for studying the rich and unique 7 secondary metabolism of legumes (Dixon and Sumner, 2003). We have recently utilized elicited M. truncatula liquid suspension cell cultures to study biotic stress responses using an integrated functional genomics approach that included transcriptomics (Suzuki et al., 2005), proteomics (Lei et al., 2005), and metabolomics (Broeckling et al., 2005;Farag et al., 2008). These studies involved independent applications of the phytohormone methyl jasmonate (MeJA) and a yeast cell wall preparation (yeast elicitor, YE, a fungal pathogen mimic) to M. truncatula root suspension cell cultures. We report here that the antimicrobial aurone, hispidol, is a major YEinduced secondary metabolite in these cultures, and correlation analyses between metabolite and transcript profiles implicated specific PRXs in aurone biosynthesis. The functionalities of these PRXs were subsequently confirmed by biochemical analysis of recombinant proteins. The integration of metabolomics and transcriptomics data has thus led to the discovery of both a novel bioactive secondary metabolite and a novel mechanism for its biosynthesis. These results are discussed in terms of biochemical and cellular responses to biotic stress.

Hispidol and hispidol-4'-O-β-D-glucoside are novel phenylpropanoid compounds induced in the response of Medicago truncatula to YE
A large-scale elicitation experiment was conducted using liquid suspension cell cultures and two elicitors (YE and MeJA) to generate an integrated global data set (transcriptome, proteome and metabolome) and to facilitate gene discovery and novel insight into biotic and abiotic stress responses associated with natural product pathways in M. truncatula. Triplicate biological samples from control and elicited cell cultures were harvested at 21 different time points between 0 and 48 hours post-elicitation for each elicitor and each replicate from independent culture flasks (Broeckling et al., 2005). Global metabolite, protein, and transcript profiles were obtained for the sampled cells and specifically queried as related to the nature and extent of effect on phenylpropanoid biosynthesis. The expression levels of approximately 16,000 tentative consensus sequences (TC) were monitored using a custom M. truncatula cDNA oligonucleotide microarray. The accumulation patterns of approximately 1000 proteins were 8 monitored with two-dimensional gel electrophoresis (Lei et al., 2005). Approximately 180 secondary metabolites were profiled using HPLC-PDA-MS (Farag et al., 2007;Farag et al., 2008) and approximately 500 primary metabolites were profiled by GC-MS (Broeckling et al., 2005).  2B). P2 (R t = 33.5 min) also had an ion peak at m/z = 253 suggesting that P2 might be the aglycone moiety observed as part of the hexose conjugate in P1 (Fig. 2C). Enzymatic hydrolysis of P1 with β-glucosidase followed by HPLC-PDA-MS analysis yielded a single peak with the same retention time and spectral characteristics as P2. GC-MS analysis of the sugar hydrolysate confirmed D-glucose as the hexose moiety in P1. The ultraviolet spectra for P1 (inset, Fig. 2B) and P2 (inset, Fig. 2C) were similar with a λ max at 390 nm implying the presence of a highly conjugated chromophore typical of aurones. Aurones absorb near the visible region due to an extended conjugated double bond/pi system found in chalcones, flavones or isoflavones (Mabry et al., 1970).
The basic aurone nucleus has a molecular weight of 222, and the 32 Da mass difference between the aurone nucleus and P2 suggests the presence of two additional hydroxyl groups on either the A or B rings of P2. Several authentic aurone standards having a molecular weight of 254, the same as P2, were analyzed using the same HPLC-MS conditions, of which only hispidol (6,4'-dihydroxyaurone) matched P2 in retention time and spectral characteristics. Thus, P2 was identified as hispidol (H), a compound previously detected in soybean seedlings (Wong, 1967). P1 was identified as hispidol-4'-O-β-D-glucoside (HG) through HPLC-PDA-MS and co-9 characterization with an authentic standard. This is the first report of the presence of these aurones in M. truncatula.
Higher intracellular levels of hispidol glucoside rather than hispidol were observed in response to YE ( Fig. 2A). This might indicate a rapid in vivo conversion of hipsidol to hispidol glucoside, or a limiting availability of isoliquiritigenin. However, metabolite profiling of the extracellular medium from YE-treated cell cultures showed that hispidol accumulation was induced up to 50-fold (Supplementarl Fig. 1); whereas, hispidol glucoside was below the detection limit of the HPLC-PDA-MS. The large accumulation of hispidol in the media suggests that the lower intracellular levels of hispidol compared to hispidol glucoside are likely due to secretion of this compound into the media.

Antimicrobial activity of hispidol
The induction of hispidol and hispidol glucoside in response to YE suggested that these compounds might be defense related or antimicrobial compounds. Thus, antifungal activity assays were conducted with the fungal pathogen Phoma medicaginis which causes spring black stem and leaf spot disease in alfalfa to assess potential biological activities of hispidol and its glucoside. Other endogenous Medicago metabolites induced in response to YE were also tested and included formononetin, ononin, afrormosin, irisolidone and isoliquiritgenin. All phenolics were tested at a concentration of 100 μ M. Coumestrol, a known antimicrobial compound in alfalfa, was included in the assay as a positive control and showed the strongest inhibition of fungal growth followed by hispidol, afrormosin, ononin and hispidol 4'-O-β-D-glucoside ( Fig.   3). Thus, hispidol can be classified as a potential and relatively potent antifungal compound in M. truncatula cell cultures. The direct precursor of hispidol, isoliquiritigenin showed very weak antifungal activity against P. medicaginis.

Selection of putative hispidol synthases
Two primary mechanisms for aurone biosynthesis have been proposed based upon polyphenol oxidase (Nakayama et al., 2000;Nakayama et al., 2001) or peroxidase activity (Wong, 1967;Rathmell and Bendall, 1972;Wilson and Wong, 1976 Putative PRXs were then evaluated as potential hispidol synthases. The custom M. truncatula oligonucleotide microarray chip included 88 putative PRX TC sequences. Ten of the PRXs showed altered expression in response to YE (Fig. 4A) and seven in response to MeJA respectively). In general, the qRT-PCR and Affymetrix microarray expression levels were consistent with the 16k custom oligonucleotide microarray results. For example, the data showed that YE elicitor induced a significant transient increase from 0.5 -4 h in TC106558, whereas TC106484 and TC110836 showed a sustained increase from 2 -24 h (Fig. 4). At the metabolite levels, both hispidol and HG showed similar induction kinetics, with a sustained increase from 8 to 36 h post-elicitation (Fig. 4G). Both qRT-PCR and microarray (Fig. 4) analyses showed no change or a mild decrease in the same three TCs in response to MeJA that correlated with no change or moderate decline in both hispidol and hispidol glucoside levels ( Fig.   4E). In summary, the transcript profiles and qRT-PCR data correlated well with that of hispidol and hispidol glucoside accumulation, supporting the hypothesis that TC110836, TC106484, and TC106558 might be involved in hispidol formation. A C-terminal propeptide that targets peroxidases for vacuolar import (Welinder et al., 2002) was present in Mt PRX2 and Mt PRX3, but not in Mt PRX1 (Fig. 5A). However, the possibility still exists that Mt PRX1 is targeted to the vacuole due to its strong cationic nature (pI 8.9) as most cationic PRXs are vacuolar localized due to the acidic nature of the vacuole (Ros 12 Barcelo et al., 2003). In addition, many vacuolar targeted proteins contain still undefined sorting signals (Carter et al., 2004).

In silico expression analysis of Mt PRX1, PRX2, and PRX3
In silico expression analysis was performed for Mt PRX1, Mt PRX2, and Mt PRX3 to evaluate the spatial expression of these PRXs (Fig. 6

Biochemical Characterization of Recombinant PRX Enzymes
To determine the biochemical activities of each of the three candidate PRXs, their cDNAs  Table 1). Further, QVE was an observed product of horseradish peroxidase and not detected in elicited M. truncatula cell cultures.
Iisoliquiritgenin 4-O-glucoside (Zhu et al.) was also converted to hispidol glucoside by the recombinant PRXs (Table 1, Supplemental Fig. 2). A considerable proportion of the isoliquiritigenin glucoside was isomerized in all in vitro assays to liquiritigenin glucoside as flavanones are the thermodynamically more stable isomers of chalcones (Boland and Wong, 1975). In the case of Mt PRX3, a quinol vinyl ether glucoside was the major product (Table 1).
To further assess Mt PRXs 1-3 as putative vacuolar peroxidases, enzymatic assays were performed that monitored the formation of hispidol from isoliquiritigenin by Mt PRXs 1-3 as a function of pH over the range of pH 3 to 9. Vacuolar enzymes tend to have lower optimum pH consistent with the acidic environment of the vacuole, which is typically pH of 5.5 (Taiz, 1992; Ros Barcelo et al., 2003) and is the case for snapdragon aureusidin synthase (Nakayama et al., 2000;Nakayama et al., 2001). Both Mt PRX 2 and 3 exhibited an activity optimum at pH 5; whereas, Mt PRX 1 showed a slightly higher activity optimum at pH 6 (supplemental Fig. 3).
The pH profile of Mt PRX 2 and 3 were bimodal with a small peak at pH 7.

Substrate specificities and mechanisms of Mt PRXs
The ability of the recombinant Mt PRXs to synthesize aurones from a variety of chalcone substrates was evaluated to determine the specificity of these enzymatic reactions and the underlying mechanism(s) ( Table 1). Chalcones tested either lacked one of the three (4, 2', or 4') free hydroxyl groups present in isoliquiritigenin or contained an additional hydroxyl group on either the A or B ring. Chalcones lacking the 2' or 4-hydroxyl functions (i.e. compounds 6, 7, and 20) generated no aurones, whereas the absence of a 4' hydroxyl group (3) did not impede enzymatic activity and the respective aurone was produced as monitored by HPLC-PDA-MS.
Interestingly, chalcone analogs having a glucose at the 4-position (2) were also acceptable substrates for aurone production, whereas those with a methoxy group at this position (8) were not, highlighting the crucial role of ring substituents at this position on enzymatic activity.
Interestingly, naringenin chalcone that serves as a substrate for aureusidin formation by PPO in snapdragon (Nakayama et al., 2000) generated no aurone when incubated with expressed Mt PRXs 1-3 (Table 1). In cases where aurone formation was absent (i.e. substrates 6, 7 and 8), the corresponding flavanone dimers were produced with Mt PRX2 and Mt PRX3, but not with Mt PRX1 (Table 1). The presence of extra hydroxyl groups at the 3, or 3 and 5' positions in chalcones (4 and 5) did not prevent formation of the corresponding aurones, although this occurred at lower amounts than with substrates 1, 2 and 3.

Integrated metabolite and transcript profiling identify hispidol and putative PRXs
PRX multigene families have been found in all species thus far; however, the assignment of individual PRXs to a specific in vivo function still remains difficult (Quiroga et al., 2000).
The genetic approach used to assess the function of PRXs involves the production of transgenic plants either over-expressing or under-expressing a specific PRX gene (Kajita et al., 1994;McIntyre et al., 1996;Lagrimini et al., 1997), but this approach has often failed to provide definitive information, and the in vivo roles of many PRXs still remain elusive (Quiroga et al., 2000). Here, we describe an integrated functional genomics approach involving correlations between transcriptomics and metabolomics for the discovery of unique natural products and their biosynthetic mechanisms. This approach is similar to studies that coupled cDNA-AFLP or DNA microarray to metabolite profiling for gene discovery in alkaloid biosynthesis in tobacco (Goossens et al., 2003) and glucosinolate biosynthesis in Arabidopsis (Hirai et al., 2004;Hirai et al., 2005). Global metabolic profiling of M. truncatula cell cultures first led to the identification of a novel phytoalexin hispidol, which was induced in response to YE, but not to MeJA.
Hispidol has been previously identified in soybean, and is derived from isoliquiritigenin chalcone (Wong, 1967). However, the enzymology of this reaction was uncertain, but likely to contain a peroxidase-mediated step (Strack, 1997). Through correlation analysis of hispidol accumulation and high density oligonucleotide microarray analyses of YE and MeJA induced cell cultures, we were able to identify three PRXs that could be associated with hispidol biosynthesis. The potential roles of Mt PRX1, Mt PRX2 and Mt PRX3 in aurone biosynthesis were each tested by heterologous expression in E. coli and in vitro assays using isoliquiritigenin and isoliquiritigenin glucoside as substrates.

Evidence implicating Mt PRX1 and Mt PRX2 as hispidol synthases
All cloned Mt PRXs catalyzed the oxidation of isoliquiritigenin and its glucoside to the corresponding aurone as a single product or one of multiple products. However, only Mt PRX1 and Mt PRX2 are likely to be functional hispidol synthases. Although Mt PRX3 catalyzes hispidol biosynthesis, it also catalyzes a greater quantitative production of a quinol vinyl ether (QVE) which is believed to originate from the rearrangement of the 2-(αperoxbenzyl)coumaranone (Wong, 1967). Less specific PRXs such as commercial horseradish peroxidase were also tested and found to catalyze greater levels of QVE than hispidol from In silico expression analyses were also performed for Mt PRX1, Mt PRX2, and Mt PRX3 to evaluate the spatial expression of these PRXs (Fig 6). Mt PRX1 was most highly expressed in flower, pods and roots. Mt PRX2 expression levels were approximately 200 fold higher in roots and nodules than any other tissues. These data illustrate the high tissue specificity of Mt PRX2 which would translate to the present cell cultures that were initiated from roots (Broeckling et al., 2005). Finally, Mt PRX3 expression levels were more evenly distributed throughout all tissues suggesting a more general peroxidase role, and therefore, a lower likelihood of a primary role in hispidol biosynthesis.
To better assess the substrate specificities of each of the peroxidases by kinetic analysis, multiple approaches were exhaustively pursued to express and purify the recombinant Mt PRX proteins. The proteins were heterologously expressed in E. coli, yeast, and insect cell systems.
In all systems, only low levels of the peroxidases were observed to accumulate, suggesting a potent cell toxicity of these enzymes in prokaryotic and eukaryotic systems. Enrichment attempts using immobilized metal affinity chromatography (IMAC) to purify the His-tagged proteins were also unsuccessful, and it is assumed that the His-tag was non-accessible for affinity binding and purification. Thus, quantitative kinetic data for these peroxidases could not be obtained.

Mechanism, substrate specificity, and comparison of PRX and PPO mediated aurone biosynthesis
The mechanism for the PRX-mediated conversion of isoliquiritigenin to hispidol proposed here is based upon that originally described by Wong and colleagues (Wong, 1967;Wilson and Wong, 1976). This mechanism involves the formation of a phenoxy radical (II) that undergoes further self rearrangement leading to formation of a free radical (III) (Fig 9). The oxygenation of the α -carbon of III furnishes the peroxide IV, which after intramolecular Michael-like addition, provides a dioxetane type structure (V) (Fig. 9). Intramolecular opening of the dioxetane (V) results in a hydroperoxide 2-(α-peroxbenzyl)coumaranone (VI), which upon cleavage and dehydration yields hispidol (H) (Fig. 9).
The substrate specificity of the Mt PRXs differed from that of the snapdragon flower PPO. Interestingly, naringenin chalcone that serves as a substrate for aureusidin formation by PPO in snapdragon (Nakayama et al., 2000) generated no aurone when incubated with expressed Mt PRXs 1-3 (Table 1). However, naringenin chalcone does not accumulate in YE elicited M.
truncatula cell cultures (Farag et al., 2007). Thus, naringenin chalcone is less likely to be a primary in vivo substrate of the Mt PRXs and/or is rapidly converted to the flavone via CHI. In contrast, significantly elevated levels of isoliquiritgenin and its glucoside were observed, and are more logical in vivo substrates for the Mt PRXs.
Snapdragon aureusidin synthase also utilizes naringenin chalcone glucoside rather than naringenin chalcone as a substrate (Ono et al., 2006), whereas, Mt PRX utilized both isoliquiritgenin chalcone and its glucoside as substrates for the production of aurones. Flavonoid glycosides co-localize in the vacuole with peroxidase and can serve as its substrate (Ros Barcelo et al., 2003). Mt PRX2 and Mt PRX3 contained a C-terminal propeptide that targets these peroxidases for vacuolar import (Welinder et al., 2002), but this signal peptide was not present in Mt PRX1 (Fig. 5A). However, the possibility still exists that Mt PRX1 is targeted to the vacuole due to its strong cationic nature (pI 8.9) as most cationic PRXs are vacuolar localized due to the acidic nature of the vacuole (Ros Barcelo et al., 2003). In addition the acidic pH optimums for Mt PRX 1-3 are consistent with vacuolar localization. Vacuolar PRXs participate in the turnover and degradation of phenolic glycosides (Ros Barcelo et al., 2003). The oxidative breakdown of phenolics by PRXs begins with the enzymatic hydrolysis of glycosides to release the aglycones which are then direct substrates for vacuolar PRXs. We hypothesize that M.
truncatula PRXs utilize chalcone aglycones as substrate produced via cleavage of its glucoside in the vacuole under stress condition to produce the antifungal agent "hispidol". In onion, a similar deglucosidation of quercetin glucosides to quercetin yields an antifungal agent 3,4dihydroxybenzoic acid as catalyzed by the action of peroxidases (Takahama and Hirota, 2000).
The additional hydroxyl group of chalcones containing the functional motif for aurone conversion increased their reactivity for PPO, but not for PRXs in M. truncatula. Both PPO and PRX mechanisms require the presence of 2' and 4 hydroxyl groups in the precursor chalcone.
Substitution of the 4-hydroxyl with a methoxy group (8) prevented conversion to aurone by any of the three Mt PRXs, whereas glucose attachment at this position (2) did not. Further, the 4-methoxy-2',4'-dihydroxychalcone substrate only yielded a flavanone dimer (16) with Mt PRX2 and Mt PRX3 (see Table 1); trace levels were detected with Mt PRX1. Protection of the 4-hydroxyl group with a methoxy group inhibits the formation of a B ring radical species.
Instead, a phenoxy radical species is initiated in the A ring at the 2' hydroxyl which undergoes further cyclization furnishing the pyrone radical (Fig. 10B) and dimerization to form compound C in Fig. 10. Under our experimental conditions, the chalcones 6 and 7 (Table 1) which both contain 2' and 4' hydroxyl groups also produced the corresponding dimers. This suggests that the absence of a hydroxyl group at C4 or its protection with a methyl group provides the mechanistic diversion towards dimer formation. In contrast, the presence of the sugar hydroxyl groups in isoliquiritgenin-4-O-glucoside may contribute to aurone formation. Homolytic cleavage of the O-H bonds of the sugar may also result in the formation of radical species (Fig.   10E) that then follow a similar reaction mechanism as proposed in Fig 9 and the previous literature (Wong, 1967), yielding hispidol-4'-O-β-D-glucoside (Fig. 10H). Our data suggest that aurone biosynthesis is kinetically favored over the dimerization reaction, as dimers were only produced as alternative oxidation products by Mt PRXs (Table 1) in situations where aurone formation was inhibited. Consistently, no dimers were detected for the in vivo Mt PRXs substrates IL and its respective glucoside. To the best of our knowledge, these structure activity relationships provide the first mechanistic insights into aurone glucoside biosynthesis in legumes.
Hispidol biosynthesis in M. truncatula is unlikely to occur via a polyphenol oxidase mechanism. This conclusion is based upon the non detectable levels of all three putative PPOs present on the M. trunatula microarray, and based upon our current mechanistic understanding.
According to the reported PPO-catalyzed mechanism for aurone synthesis (Nakayama et al., 2001), the substrate must contain hydroxyl groups at the 2' and 4 positions, such that all products will then possess a 3,4-dihydroxy-substituted B-ring (Fig. 1). Hispidol contains a 4monohydroxy B-ring and lacks the 3-hydroxyl B ring group (Fig. 1); thus the production of hispidol via a PPO catalyzed mechanism is unlikely. This argument is further supported by the fact that the predicted substrate required for the PPO production of hispidol would contain an unsubstituted B-ring (i.e. 2',4'-dihydroxy chalcone or a 4-dehydroxy isoliquiritigenin) which has not been identified in legumes to date. Nakayama and colleagues also suggest that hispidol biosynthesis occurs via a PRX as opposed to a PPO mediated mechanism (Nakayama et al.,

A) RNA isolation, cDNA labeling and hybridization
Total RNA was isolated from 0.5 g of M. truncatula suspension cells using 5 ml of Tri-Reagent

D) Microarray data processing and analysis
Arrays were imaged with a ScanArray 4000 scanner (Packard, Palo Alto, CA) at 10 µm resolution and variable photomultiplier tube voltage settings to obtain the maximal signal intensities. The fluorescence intensity for each label and each element on the array was captured using GenePix Pro 4.1 (Axon, Union City, CA). Normalization of Cy3 and Cy5 signal was performed by adjusting the signal intensities of the two images using a Lowess (Sub-grid) method and GeneTraffic software (www.iobion.com). The local background was subtracted from the values of each spot on the array. The statistical analysis of variance (ANOVA) of normalized data was performed using GeneSpring software (www.silicongenetics.com).

Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR) analysis
Microarray results were confirmed by qRT-PCR using TC-specific primers for tissue elsewhere (Czechowski et al., 2005). Data were analyzed using the SDS 2.2.1 software (Applied Biosystems). PCR efficiency (E) was estimated using the LinRegPCR software as described (Ramakers et al., 2003) and the transcript levels were determined by relative quantification (Pfaffl, 2001) using actin (TC# 107326) as reference gene (primers above).

Antifungal agar plate bioassays
Antifungal assays were conducted essentially as described (Blount et al., 1992). Stock plates of Phoma medicaginis Malbr. Et Roum were grown on potato dextrose agar media (Difco laboratories) in sterile 100 x 15 mm Petri dishes until the mycelia covered 1/3 of the plate.
Compounds to be tested were dissolved in dimethylsulfoxide (DMSO, Sigma) at stock concentrations of 20mM. The solvent (as a control) or compounds were then added to the melted agar media at 5 μ l ml -1 to give a final phenolic concentration of 0.1 mM. 5 ml samples of media were aseptically pipetted into sterile 35 x 10 mm Petri dishes and allowed to cool. Using a sterile 4 mm cork borer, agar discs with mycelia were cut from a stock fungal culture plate and placed on the center of treatment and control plates. Radial mycelial growth was measured from the edge of the disc to the outer edge of the mycelia, and the results of treatments were calculated as the relative percentage of their corresponding solvent control. truncatula peroxidases. (I) Aurone-producing chalcone substrates are listed with the amounts of aurones produced in each reaction expressed as a ratio relative to that obtained with an empty vector in the following order (Mt PRX1, Mt PRX2, Mt PRX3).
-indicates absence, + indicates ratio < 10, ++ indicates a ratio 10-20, and +++ indicates a ratio > 20. The relative ratio of each product formed by Mt PRX 1, 2 and 3 does not solely reflect difference in enzymatic activities but also differences in expression levels.         designates that a compound aligns with the denoted peak in the corresponding chromatogram. (modified from (Wong, 1967). The pathway is initiated with the peroxidase catalyzed generation of a phenoxy radical (II) and the subsequent steps most likely occur non-enzymatically. The additional steps include radical rearrangement (III) followed by oxygenation of the α -carbon (IV) and epoxide cyclization with the β -carbon (V). The epoxide radical is then utilized for the propagation of an additional substrate radical (i.e. conversion of I to II) which initiates a new cycle of the reaction, and it is converted to the neutral molecule. The neutral epoxide is then reduced and dehydrated to form VIII or the epoxide VII. These products then readily convert to hispidol or the QVE. The specific end products observed for each of the M. truncatula peroxidases are indicated in the final steps.       truncatula PRXs to other functionally annotated peroxidases. A) Amino acid alignments of Medicago truncatula peroxidases generated using ClustalW. Cys residues forming disulfide bridges are highlighted with an asterisk (*) and the two His residues essential for catalysis indicated in bold are underlined. Solid line boxes represent conserved domains in class III peroxidases; the dashed line box represents the C-terminus propeptide sequence targeting PRXs to the vacuole. Gaps are introduced to maximize the alignment. Conserved similarity shading is based on 100% (black), 60% (dark gray), and 30% (light gray). B) Unrooted tree dendogram comparison of the amino acid sequences of M. truncatula peroxidases (Mt PRX1, Mt PRX2 and Mt PRX3, labeled with asterisks), with other functionally characterized plant peroxidases. The genes are: Nicotiana tabacum peroxidase GBID AB027752, Ipomea batatus anionic peroxidase GBID AY206414, Asparagus officinalis basic peroxidase GBID AJ544516, Quercus suber cationic peroxidase AY443340, Peanut Arachis hypogaea cationic peroxidase GBID M37636, Gossypium hirsutum bacterial-induced peroxidase GBID AF155124, Linum usitatissim basic peroxidase AF049881, Scutellaria baicalensis novel peroxidase GBID AB024439, Pea ascorbate peroxidase GBID X62077, Glycine max cytosolic ascorbate peroxidase GB U56634, Glycine max ascorbate peroxidase GBID AF127804, Arabidopsis thaliana ascorbate peroxidese cytosolic protein GBID X59600, tomato ascorbate peroxidase cytosolic protein GBID Y16773, Arabidopsis thaliana class III peroxidase GBID AF452388, Arabidopsis thaliana class III peroxidase GBID AF452387, Populus trichocarpa xylem anionic peroxidase GBID X97351, Populus nigra anionic peroxidase GBID 83225, Populus trichocarpa xylem anionic peroxidase GBID X97350, Linum usitatissimum anionic peroxidase GBID L07554, Glycine max seed coat peroxidase GBID L78163, Glycine max anionic seed coat peroxidase GBID U41657, Medicago sativa peroxidase GBID X90693, Medicago sativa peroxidase GBID X90694, Lupinus albus extensin peroxidase GBID AF403735, Glycine max peroxidase GBID AF007211. The branch lengths are proportional to the degree of divergence with a scale of "0.1" representing 10% change.