(Iso)flavonoids are commonly accumulated as malonylated or acetylated glycoconjugates in legumes. Sequence analysis on EST database of the model legume Medicago truncatula enabled us to identify nine cDNA sequences encoding BAHD super-family enzymes that are distinct from the most of the characterized anthocyanin/flavonol acyltransferase genes in other species. Functional characterization revealed that three of these corresponding enzymes, MtMaT1, 2 and 3, specifically recognize malonyl CoA as an acyl donor and catalyze the malonylation of a range of isoflavone 7-O-glucosides in vitro. These malonyltransferase genes displayed distinct tissue-specific expression patterns and responded differentially to biotic and abiotic stresses. Consistent with gene expression, the level of the accumulated malonyl isoflavone glucoside was altered in the roots of M. truncatula grown under normal and drought-stressed conditions. Overexpression of the MtMaT1 gene in a previously engineered Arabidopsis line that accumulates genistein glycosides (Proc. Natl Acad. Sci. USA, 99, 2002:14578) led to a malonylated product. Confocal microscopy of the transiently expressed MtMaT1–GFP fusion revealed strong fluorescence in both the cytoplasm and nucleus of M. truncatula and tobacco leaf cells. A truncated MtMaT1 lacking the C-terminal polypeptide of 110 amino acid residues that include the DFGWG motif, the single conserved sequence signature of BAHD super-family members, retained considerable catalytic efficiency, but showed an altered optimum pH preference for maximum activity. Such C-terminal polypeptide deletion or deletion of the DFGWG motif alone led to improper folding of the transiently expressed GFP fusion protein in living cells, and impaired nuclear localization of the enzyme.
Isoflavonoids are large subfamily of phenylpropanoid metabolites that confer resistance to fungal pathogens in legume plants and serve as signaling molecules modulating legume–Rhizobium symbiosis (Dewick, 1988, 1993; Dixon, 1999). In the Leguminosae, such as alfalfa (Medicago sativa L.), barrel medic (M. truncatula), white sweet clover (Melilotus alba), chick pea (Cicer arietinum) and soybean (Glycine max), the constitutively accumulated (iso)flavonoids often exist as glucosides and malonylated or acetylated glucoconjugates (Baggett et al., 2002; Graham et al., 1990; Kachlicki et al., 2005; Kudou et al., 1991; Mackenbrock et al., 1993; Park et al., 1992; Williams and Harborne, 1989). These malonylated or acetylated glycoconjugates are generally regarded as ‘storage forms’ that accumulate in the central vacuole or other specific cell compartments, serving as a chemical pool of biosynthetic precursors or inactive forms of phytoalexins (Dewick, 1993; Dixon, 1999; Figure 1). The malonyl residues substituted on the sugar moiety of (iso)flavonoids potentially prevent enzymatic degradation of the glucoconjugates, change their lipophilicity and act as a signal moiety promoting the sequestration of the conjugates into specific compartments, such as vacuoles or the cell wall (Harborne, 2000; Markham et al., 2000).
The enzymatic acylation of small molecules in plants requires at least two types of high-energy acyl donors, the acyl CoA thioesters and 1-O-acylglucosides. Correspondingly, two families of acyltransferases, acyl CoA-dependent acyltransferases and serine carboxypeptidase-like proteins (SCPL), are involved (Steffens, 2000). Although an enzyme of the SCPL family demonstrated sinapoylation (aromatic acylation) activity on anthocyanin conjugates (Fraser et al., 2007), enzymatic malonylation or acetylation of (iso)flavonoid glycoconjugates requires acyl CoA-dependent acyltransferases in most cases (Dewick, 1993; Harborne, 2000). In plants, acyl CoA-dependent acyltransferases are involved in forming and modifying a variety of special metabolites, including the alcoholic esters for floral scents and fruit aromas (D’Auria et al., 2002; Dudareva and Pichersky, 2000; Dudareva et al., 1998), and the flower pigment anthocyanins and flavonols (Fujiwara et al., 1998; Luo et al., 2007; Suzuki et al., 2001, 2004b). Consistent with their versatile biological functions in plant metabolism, acyl CoA-dependent O- or N-acyltransferases comprise a large functionally divergent protein family, known as the BAHD super-family (D’Auria, 2006; St-Pierre and De Luca, 2000). Members of the BAHD family characteristically contain two highly conserved motifs, HXXXD and DFGWG, in their primary sequences. The enzymes characterized so far are exclusively cytosolic proteins, with a molecular mass ranging from 48 to 55 kDa; none appears to have signal peptides or other transit sequences that lead to specific localizations or secretions (D’Auria, 2006). Within the BAHD super-family, several characterized anthocyanin/flavonol acyltransferases (AATs) constitute a phylogenetically specific group that separates them from other BAHD members, and these generally share an additional conserved sequence motif, NYFGNC, which is used as their diagnostic signature (Nakayama et al., 2003).
The enzyme activity of malonyl CoA:isoflavone 7-O-glucoside-6′′-O-malonyltransferase was first detected in chick pea (Cicer arietinum) hypocotyls (Koester et al., 1984). However, molecular characterization of the corresponding genes and enzymes was not performed until recently, while this paper was in preparation, when an isoflavone 7-O-malonyltransferase gene was characterized from soybean by a sequence homology-based PCR cloning strategy and in vitro functional analysis (Suzuki et al., 2007). In our study, we adopted a biochemical genomics approach in order to systemically identify the enzymes responsible for acylation of (iso)flavonoid glycoconjugates. Using the conserved sequence motifs of BAHD family members and the sequences of several functionally characterized acyltransferases, we searched the EST database of M. truncatula and identified approximate 77 tentative consensus sequences (TCs) and singletons that encode the putative BAHD super-family enzymes. Seventeen of them are clustered within the AAT gene family and further separated into three clades. Nine comprise a specific clade that is distant from most characterized genes encoding AATs. Our functional characterization demonstrated that the enzymes encoded by at least three members of this clade, namely MtMaT1, 2 and 3, specifically utilize malonyl CoA as an acyl donor and catalyze malonylation of a range of isoflavone 7-O-glucosides, including genistin (genistein 7-O-glucoside), daidzin (daidzein 7-O-glucoside), formononetin 7-O-glucoside and 6, 4′-dihydroxyisoflavone 7-O-glucoside. These three enzymes are phylogenetically distinct from the newly identified soybean isoflavone 7-O-malonyltransferase. The three malonyltransferase genes displayed distinct tissue-specific expression patterns and responded differentially to stresses caused by pathogen infection, insect herbivory and various abiotics. Drought stress, in particular, greatly altered expression of the malonyltransferase genes in M. truncatula roots. The level of accumulated formononetin 7-O-glucoside malonate was consistently altered, suggesting biological function of the identified malonyltransferases in isoflavonoid biosynthesis. Subsequent overexpression of the MtMaT1 gene in a previously engineered Arabidopsis line that accumulates genistein glycosides in leaf tissue (Liu et al., 2002) produced genistein malonyl glucoside. An MtMaT1–GFP fusion protein expressed in the leaf cells of tobacco and M. truncatula resulted in strong fluorescence in both the cytoplasm and nucleus. A truncated MtMAT1 lacking the C-terminal 110 amino acid residues, including the conserved DFGWG motif, displayed reduced activity at the neutral pH conditions that are optimal for the intact enzyme but retained considerable catalytic efficiency for malonylation of (iso)flavone glucosides under acidic conditions. Deleting the C-terminal polypeptide or the DFGWG motif caused partial aggregation of the highly expressed GFP fusion in living cells and impaired its nuclear localization.
Identification of putative acyl CoA-dependent acyltransferase genes
We used the conserved sequence motifs of the BAHD family members, namely DFGWG and NYFGNC (specific for AATs), and the full-length sequences of several acyltransferases with known function to search The Institute for Genomic Research (TIGR)M. truncatula EST databases (see Experimental procedures). We identified approximately 77 putative acyl CoA-dependent acyltransferase genes, including 51 TCs and 26 singletons, from among 36 878 annotated cDNAs. Multiple gene-sequence alignments demonstrated that, of these 77 candidates, 17 cDNA sequences including 12 TCs and five singletons, constituted a separated group that clustered with several anthocyanin/flavonol acyltransferase (AAT) genes of known function (data not shown). The encoded polypeptides of the 17 cDNAs together with all the known AATs were further phylogenetically classified into four clades, and the 17 M. truncatula sequences were separated into three clades that are distinct from the most of the known anthocyanin/flavonol acyltransferases in other species (Figure 2). Clade I contains two TCs (TC97945 and 109329) and one singleton, which are clustered with the recently identified two anthocyanin/flavonol hydroxycinnamoyltransferases from Arabidopsis (Luo et al., 2007). Clade II comprises seven M. truncatula TCs and two singletons. There were no characterized enzymes classified within this clade at the time of initiating this study; however, recently, two malonyltransferases, Arabidopsis anthocyanin 5-O-glucoside and soybean (iso)flavone 7-O-glucoside malonyltransferases (Luo et al., 2007; Suzuki et al., 2007) were characterized and are distantly clustered within this clade (Figure 2). The M. truncatula sequences in this clade are further separated into two sub-groups, implying potential function disparity of these putative enzymes in the acylation of phenolic glycosides. The remaining three M. truncatula TCs and two singletons are distantly classified within clade III, which contains one unique anthocyanin 5-O-glucoside 4′′′-O-malonyltransferase identified in scarlet sage (Salvia splendens; Suzuki et al., 2004b).
Isolation and sequence analysis of acyl CoA-dependent acyltransferase genes
As a functional genomics-based approach to systemically identify M. truncatula (iso)flavonoid acyltransferases, we isolated five full-length cDNAs of clade II (TC102229, TC109903, TC107840, TC100057 and TC98548; Figure 2) from the corresponding M. truncatula EST library collection of the Samuel Roberts Noble Foundation. The EST clone of TC102229 from the root library contains 1519 nucleotides but encodes a polypeptide of only 362 amino acid residues. This deduced polypeptide is 110 amino acids shorter at its C-terminus, including the DFGWG motif, than the annotated TC sequence deposited in the TIGR gene index database (Figure 3). The truncation is caused by insertion of an extra thymidine in a thymidine-rich region of the nucleotide sequence (data not shown). To verify whether this insertion mutation occurs naturally in planta or was generated artificially during construction of the cDNA library, cDNA fragments spanning the thymidine-rich region were amplified using mRNAs prepared from M. truncatula root, leaf and suspension-cultured cells. All the amplified fragments contain only seven consecutive thymidines, as does the annotated TC in the database, suggesting that the extra thymidine was artificially inserted in the obtained EST clone. Therefore, we deleted the thymidine by mutagenesis to produce a full-length cDNA that encodes a polypeptide containing 472 amino acid residues with a deduced molecular mass of 53.064 kDa. This full-length cDNA was designated MtMaT1 (accession number EU272030), and the truncated cDNA was designated MtMaT1-Δ110. The EST clone of TC109903, designated MtMaT2 (accession number EU272032), has 1436 nucleotides and encodes a polypeptide of 476 amino acids with a deduced molecular mass of 53.605 kDa, while that of TC107840, designated MtMaT3 (accession no. EU272031), contains 1606 nucleotides and encodes 483 amino acid residues with a deduced molecular mass of 54.510 kDa. Aligning the amino acid sequences revealed that MtMaT1 and MtMaT2 were more closely clustered, sharing approximately 73% similarity at the amino acid level, while MtMaT3 is further from MtMaT1 and 2, with only 59% and 55% similarity, respectively. The sequences of the TC100057 and TC98548 clones were separated from those of the three isolated MtMaTs (Figure 2), and the encoded enzymes exhibited distinct functions, which will be reported elsewhere.
In all sequences, the putative catalytic motif of the BAHD family enzymes, HXXXD, was highly conserved; by contrast, the DFGWG motif varied and was DFGFG in MtMaT3, while the other conserved motif, NYFGNC, among the characterized AATs, was NYLGNC in MtMAT1 and NYCGNC in MtMaT3 (Figure 3).
Characterization of MtMaT functions
The recombinant proteins MtMaT1, 2, and 3, as well as the truncated MtMaT1-Δ110, were purified from Escherichia coli and assayed using the aliphatic thioester donors malonyl CoA and acetyl CoA, the aromatic acyl donors p-coumaroyl CoA, feruloyl CoA and caffeoyl CoA, and several glycosidic phenolic conjugates (Table 1 and Figure 4). LC-MS profiling of the reaction products revealed that all three full-length enzymes specifically utilized malonyl CoA as the thioester donor, and converted all the tested 7-O-glycosidic isoflavones and flavanones into a corresponding product (Figure 5). Each reaction product resolved by LC-MS displayed a UV spectrum similar to that of its parental substrate, but exhibited a mass increment of 86 m/z, indicating the occurrence of malonylation (Figure 5). Collision-induced dissociation (CID) fragmentation of the product suggested that malonylation occurred at the 7-O-glycoside, presumably at the 6′′-OH of the sugar moiety (data not shown). None of the enzymes displayed significant activities on flavonoid 3-O-glycosides, or cyanidin 3- or 5-O-mono- or di-glycosides (Table 1). No activity was detected for any of the three enzymes when using aromatic acyl CoAs as donors.
Table 1. Substrate preferences of MtMaTs
aThe structures of the substrates are shown in Figure 4.
bActivity was measured with 1.5 μg protein incubated with 100 μm malonyl CoA and phenolic substrate at 30°C for 30 min. That for genistin was set at 100, and other activities are calculated relative to that for genistin. The measured activities for genistin were 50, 31.8, 1.7 and 5.8 nmol mg−1 min−1 for MtMaT1, MtMaT1-Δ110, MtMaT2 and MtMaT3, respectively. ND, not detectable.
6,4′-dihydroxy-7-O-glucosyl isoflavone (2)
Formononetin 7-O-glucoside (4)
Naringenin 7-O-glucoside (5)
The optimum pH for the activities of MtMaT2 and MtMaT3 was around 8.0, while MtMaT1 was optimally active at the neutral pH of 7.5 (Figure S1). The kinetics of the recombinant MtMaT1, 2 and 3 for various (iso)flavonoid glycosides were determined at their optimum pH. MtMaT1 demonstrated a highest catalytic ratio for the isoflavone 7-O-glucoside genistin, although the variations in its kcat/Km values for various substrates are within an order of magnitude (Table 2). In contrast, MtMaT2 and 3 showed an in vitro preference for the flavanone 7-O-neohesperidoside naringin (Table 2), a compound that mainly exists in grapefruit and is not found in Medicago species (Bisby et al., 1994). At the saturated concentration of acceptor substrate, the apparent Km of MtMaT1 and MtMaT2 for the malonyl CoA donor were approximately 42.2 and 11.1 μm, respectively (Table 2), values that are within the same range as those of other anthocyanin/flavonol malonyltransferases (Suzuki et al., 2001, 2002), but MtMaT3 had a Km more than 10 times higher than that of MtMaT1 (Table 2).
Table 2. Catalytic properties of Medicago truncatula (iso)flavone-7-O-glucoside malonyltransferases
The truncated MtMaT1-Δ110 showed an obvious shift of the apparent optimum pH for maximum activity from 7.5 to 6.0 (Figure S1). The truncation caused reduction of the overall enzyme activity, particularly at the neutral and higher pH conditions suitable for the intact enzyme (Figure S1). However, at the optimum pH of 6.0, the truncated enzyme retained considerable catalytic efficiency and high binding affinity for all the tested 7-O-glycosidic (iso)flavonoids and the malonyl CoA donor (Table 2).
Tissue-specific and stress-inducible expression of MtMaT genes and phenolic metabolite profiling
The (iso)flavonoid malonyltransferase genes that we identified had distinct tissue-specific and stress-responsive expression patterns (Figure 6a–d). MtMaT1 was predominantly expressed in M. truncatula root and leaf, where its expression was clearly downregulated by drought stress; in contrast, challenge of M. truncatula seedlings with copper chloride remarkably induced its expression (Figure 6a). Further, while MtMaT1 exhibited a high level of constitutive expression in leaves, it showed little response to infection by the leaf-rot pathogen Phoma medicaginis (Figure 6b) or insect damage to young leaves (Figure 6c), and elicitation of a root cell suspension culture of M. truncatula by yeast elicitor (Figure 6d) slightly induced its expression. The overall expression pattern of MtMaT2 resembled that of MtMaT1 in the various tissues and for various treatments, except it had an apparently lower transcription level. In contrast, MtMaT3 exhibited constitutive expression in root, leaf and stem, but was invariably absent from cell suspension cultures. In contrast to MtMaT1 and 2, drought stress and fungal infection increased the expression of MtMaT3, and copper treatment decreased its expression.
Consistent with the constitutive expression of MtMaT1 and 2 in normally growing M. truncatula roots, a high level of malonylated formononetin 7-O-glucoside was detected in the same tissues, but the unmethylated isoflavone genistein and daidzein conjugates and the flavanone conjugates naringenin 7-O-glucoside or naringin were not present in detectable amounts (Figure 6e). The malonylated formononetin glucoconjugate was resolved with the same UV spectrum as the in vitro enzymatic malonylated product from formononetin 7-O-glucoside, and its structure was confirmed by tandem MS analysis (Figure 6g). The level of this malonyl glucoside accumulated in roots was approximately 30.3 nmol g−1 FW. However, when plants were exposed to drought, the amount of formononetin conjugates present in roots decreased; the level of the malonyl glucoside of formononetin decreased to approximately 4.7 nmol g−1 FW, more than a six-fold reduction, which is consistent with downregulation of MtMaT1 and 2 gene expression under the same conditions (Figure 6a). These results imply that MtMaT1 and/or MtMaT2 may function as formononetin 7-O-glucoside malonyltransferases in M. truncatula.
Overexpression of MaT genes in an engineered Arabidopsis tt6/tt3 mutant
Arabidopsis normally does not accumulate isoflavonoids. Expression of soybean isoflavone synthase (IFS) in a tt6/tt3 double mutant ecotype Landsbergerecta (Ler) with both the flavanone 3-β-hydroxylase (F3H) and dihydroflavonol reductase (DFR) genes of the flavonol/anthocyanin synthetic pathway knocked-out (Winkel-Shirley, 2001)], resulted in accumulation of isoflavone genistein glycoconjugates in transgenic leaf tissues (Liu et al., 2002). This tt6/tt3 mutant line containing the IFS transgene was used as the host for co-expression of the identified MtMaT1 gene. HPLC profiling phenolic extracts from leaves of the T2 transgenic lines MtMaT1 7t-2–6, 7t-2–7 and 7t-2–8, revealed a novel genistein 7-O-glucoside malonate peak in the profile (Figure 7a) that exhibited the same retention time and UV spectrum as that produced from the in vitro reaction of MtMaT1 with genistein 7-O-glucoside and malonyl CoA (Figure 7c). This peak was absent from the extract of control lines that only carry the IFS transgene (Figure 7b and Figure S2). Additional peaks were also observed in the profile of tt3/tt6-IFS-MtMaT1 transgenic lines (Figure 7a). The peak at 41.7 min showed a UV spectrum with absorption at 260 nm (Figure S2), which is characteristic of isoflavone genistein, suggesting that another potential malonylated derivative from the distinct genistein glycoconjugates accumulated in the IFS transgenic line (Liu et al., 2002). The peak at 50.4 min showed UV absorption at 235 nm, and was not found in other transgenic lines (Figure S2). In all tt3/tt6 transgenic lines, no specific malonylated naringenin 7-O-glucoside or naringin was resolved.
When the IFS gene is expressed in Arabidopsis Ler wild-type, the plants do not produce significant or detectable amounts of isoflavone but accumulate predominantly flavonoid conjugates, as does the Ler wild-type (Liu et al., 2002). Overexpressing the MtMaT1 gene in Ler-IFS transgenic plants did not result in the detection of any novel malonylated flavonoid conjugates or obvious changes in the endogenous flavonoid conjugates (Figure 7d), compared to the Ler wild-type (Figure 7e) or Ler-IFS plants (data not shown). Taking together, these results suggest that MtMaT1 functions biologically as an isoflavone 7-O-glycoside malonyltransferase in planta.
Subcellular localization of MtMaT1
MtMaT1, MtMaT1-Δ110, MtMaT1 without the DFGWG motif, and the C-terminal polypeptide of 110 amino acids from MtMaT1 were fused in-frame to the N-terminus of enhanced GFP (Figure 8). The chimeric genes, driven by a double 35S promoter, were bombarded into epidermal cells from the leaves of Nicotiana benthamiana and M. truncatula. As shown in Figure 8(a,b), the fluorescence of the MtMaT1–GFP fusion was distributed uniformly in both the cytoplasm and nuclei of the cells of both plants, as was GFP itself (Figure 8c). As controls, a Arabidopsis BAHD family enzyme (At5g41040; C.-J.L. and X.-H.Y., unpublished data) fused with GFP and expressed in tobacco leaf cells did not exhibit any nuclear localization signal (Figure 8d), and an isoflavone 7-O-methyltransferase (I7OMT) fused with GFP in a similar construct to the MtMaT1 fusion (Liu and Dixon, 2001) showed only cytoplasmic distribution in normal-growth leaf cells (Figure S3). Fusion of GFP to MtMaT1 without the C-terminal 110 amino acid residues (MaT1-Δ110) or without the DFGWG motif resulted in the appearance of irregularly aggregated fluorescence granules that clustered in the cytoplasm and on peri-nuclei (Figure 8e–g). Fluorescence was largely absent from the nuclei (Figure 8f). Interestingly, fusion of GFP with the C-terminal polypeptide of 110 amino acid residues that contains the DFGWG motif (Δ110-GFP) produced even smaller but much more uniform fluorescent granules throughout the cytoplasm (Figure 8h). Time-lapse imaging and staining with a mitochondria marker excluded the possibility of mitochondrial localization of this short peptide–GFP fusion. Under high magnification, these evenly distributed fluorescent spots in the cytosol were observed to occur predominantly on the cytoskeleton (data not shown).
Malonyltransferases responsible for (iso)flavonoid biosynthesis
Our genome-wide analysis of the M. truncatula EST database enabled us to identify nine cDNA sequences that phylogenetically differ from most non-legume malonyl CoA:anthocyanin/flavonol malonyltransferases (Figure 2). Three of the encoded enzymes, MtMaT1, MtMaT2 and MtMaT3, specifically recognize malonyl CoA but not acetyl or aromatic acyl CoAs for modification of a range of isoflavone-7-O-glucosides. Like the enzyme purified from chickpea (Koester et al., 1984), these characterized recombinant enzymes also accept other flavonoid 7-O-glucosides as substrates in vitro. However, none of them show any significant activity on flavonoid 3-O-glycosides and anthocyanin mono- or di-O-glycosidic conjugates, indicating a strict substrate specificity and regio-selectivity for sugar moieties linked at various positions on the polyphenolic core structure. Expression of MtMaT1 and 2 in M. truncatula root tissues correlates well with accumulation of the isoflavone formononetin 7-O-glucoside malonate. The MtMaT1 and 2 genes were also constitutively expressed in M. truncatula leaves, which is consistent with the previous observation of constitutive gene expression of isoflavone synthase and isoflavone O-methyltransferases in M. truncatula leaf tissues (Deavours et al., 2006), although we were unable to resolve the particular isoflavone conjugates from the leaf phenolic extracts by UV-HPLC profiling (data not shown) due to overlapping with the many other leaf phenolics that are accumulated. Moreover, overexpression of the MtMaT1 gene in engineered Arabidospsis that accumulates genistein glycosides resulted in the production of malonylated isoflavone glucoconjugates, but expression of the same gene in Arabidopsis that only produces flavonol glycosides (Liu et al., 2002; Kerhoas et al., 2006) did not lead to the production of novel malonylated flavonoid conjugates or changes in metabolite profiles (Figure 7d,e). Together, these data suggest that MtMaT1 and/or 2 function as malonyl CoA:isoflavone 7-O-malonyltransferases in vivo. The type of malonylated isoflavone conjugates formed in a given species, e.g. M. truncatula or Arabidopsis, depends more on the available internal concentrations of the isoflavone substrates than on the Km and kcat values that the enzymes display in vitro.
A soybean isoflavone 7-O-malonyltransferase was recently characterized (Suzuki et al., 2007). It shares only 36–42% sequence similarity with the characterized MtMaTs at the amino acid level, and hence is separated into different sub-group, where it is phylogenetically closer to M. truncatula TC110716 and the singleton NF082E07EC1F1053 (56% and 52% similarity at the amino acid level, respectively), and moderately close to other two enzymes responsible for acetylation of (iso)flavonoids in vitro (X.-H. Yu and C.-J.L., unpublished data; Figure 2). Although the enzymes from both soybean and M. truncatula species share similar substrate specificity with respect to thioester donors and 7-O-glucosides of isoflavone conjugates, the soybean enzyme has not been tested with any flavonoid 7-O-glucoside as potential substrate (Suzuki et al., 2007). The enzymes from neither species show significant activity towards 3-O-glucosides of flavonol and 3-O- or 5-O-glucosides of anthocyanidins.
Nuclear localization of MtMaT1
Most of the BAHD family members characterized so far are thought to be localized in the cytosol (D’Auria, 2006). An immunolocalization study consistently verified that a anthocyanin 5-O-glucoside aromatic acyltransferase from Gentiana triflora is predominantly localized in the cytoplasm of epidermal cells (Fujiwara et al., 1998). However, the MtMaT1–GFP fusion protein expressed in the epidermis of M. truncatula and tobacco leaves is clearly localized not only in the cytoplasm but also in nuclei (Figure 8a,b). The nucleocytoplasmic localization of the MtMaT1–GFP is not due to aberrance of the GFP fusion itself as the fused protein was properly expressed in living cells and other enzymes fused with the same fluorescence protein in the same construct did not exhibit the same localization behavior as MtMaT1 (Figure 8 and Figure S3). The nuclear localization of MtMaT1 was not affected by stress or elicitation of M. truncatula cells (data not shown). Typically, small molecules (<50 kDa) diffuse freely in and out of the nucleus through the nuclear pores. However, the size of the MtMaT1–GFP fusion is approximately 80 kDa, far larger than the reported permeability of the nuclear pore complex by passive diffusion (Keminer and Peters, 1999). To access the nucleus, larger proteins require a specific monopartite or bipartite nuclear-localization sequence to mediate binding to a cargo protein called importin for their transport through the nuclear pore (Keminer and Peters, 1999; Rout and Aitchison, 2001). Analysis of the MtMaT1 sequence, however, did not identify any apparent nuclear-localization signals or transit peptides. So far, only one BAHD-like member, CER2 from Arabidopsis thaliana, which is phylogenetically clustered with BAHD enzymes but lacks the typical BAHD family sequence signatures, has been demonstrated to be a nuclear enzyme, but the CER2 protein sequence does not contain any of the commonly recognized nuclear-localization signals (Xia et al., 1997). Expression analysis indicated that CER2 might act as regulatory enzyme rather than a biochemical catalyst in the extension of epicuticular wax (Costaglioli et al., 2005). Both in vitro and in vivo analyses demonstrated that MtMaT1 functions as biosynthetic enzyme for modifying (iso)flavonoids. The constitutively accumulated (iso)flavonoid conjugates were generally thought to be stored in the central vacuole (Harborne, 2000; Mackenbrock et al., 1992). However, recent observations have indicated accumulation of flavonoids in the nucleus of various species, such as Arabidopsis thaliana (Buer and Muday, 2004; Peer et al., 2001), the tea plant (Camellia sinensis; Feucht et al., 2004b), Picea abies (Hutzler et al., 1998), Tsuga canadensis and Taxus baccata (Feucht et al., 2004a). Consistently, Saslowsky et al. (2005) observed that two flavonoid biosynthetic enzymes, chalcone synthase (CHS) and chalcone isomerase (CHI), occur in the nucleus in several cell types of Arabidopsis; these specially localized enzymes might be responsible for in situ synthesis of flavonoids in the nucleus. So far there is no clear biochemical evidence showing whether isoflavonoid glycoconjugates, as well as flavonoids, accumulate in the nucleus. Interestingly, an M. truncatula (iso)flavonoid β-glucosidase, which digests pre-stored (iso)flavonoid glucoconjugates, was found to localize to the nucleus in the cells under stress conditions (Naoumkina et al., 2007). Whether the unexpected nuclear localization of isoflavone malonyltransferase is due to its involvement in the formation of (iso)flavonoid conjugates in specific sites of the cell, or whether the nucleus of the cell functions as a specific storage site for particular enzymes requires further investigation.
Functions of the C-terminal polypeptide of MtMaT1
The BAHD family enzymes share a conserved tertiary architecture and consist of two α/β core domains linked by flexible loops. Between these two structural domains, a solvent channel forms that hosts the putative active site (Ma et al., 2005; Unno et al., 2007). The thioester donor and acceptor molecule bind to the solvent channel from each side to allow the transacylation reaction. Homology models for MtMaT1 and the truncated MtMaT-Δ110, based on the crystal structure of anthocyanidin 3-O-glucoside-6′′-O-malonyltransferase (Protein Data Bank code, 2E1T), demonstrated that the C-terminal amino acid residues deleted from MtMaT1-Δ110 constitute a large proportion of the α-helices (α17–19), β-sheets (β13–16) and adjunct loops involved in constitution of the hydrophobic channel and binding of the acyl CoA donor (Figure 9a). It is apparent that deletion of the C-terminal amino acid residues will severely damage the substrate binding site (Figure 9b), thus probably abolishing enzyme activity. However, the truncated MtMaT1-Δ110 exhibited considerable enzyme activity and high binding efficiency for the tested substrates at its optimum pH (pH 6.0; Tables 1 and 2). These results imply that the truncated malonyltransferase probably undergoes a great degree of conformational rearrangement in the solution at the optimum pH to maintain its activity. On the other hand, the truncated enzyme exhibited lower activity under the neutral or high pH conditions that are optimal for the intact enzyme in vitro (Figure S1), and fusion proteins of GFP with MtMaT1-Δ110 in living cells under physiological pH showed irregular aggregations, suggesting improper folding. Consistently, overexpression of the truncated MtMaT1-Δ110 in the Arabidopsis tt3/tt6 mutant containing an IFS transgene did not lead to accumulation of malonylated isoflavones (data not shown). Hence, the C-terminal residues of MtMaT1, including the DFGWG motif, might be involved in stabilizing the protein conformation at a certain pH range and conferring the characteristic pKa of the enzyme to ensure that the intact enzyme functions properly under physiological conditions (pH 7.5). This is in agreement with the previously predicted role of the DFGWG motif of anthocyanindin malonyltransferase from scarlet sage (Suzuki et al., 2003).
The DFGWG motif and adjacent amino acid residues form a flexible loop on the protein’s surface (Figure 9a). When the C-terminal polypeptide containing 110 amino acid residues, including the DFGWG motif, was fused with GFP, the fusion proteins produced uniquely aggregated fluorescence and interacted with the cytoskeleton in the cytoplasm (Figure 8h), implying that the C-terminal polypeptide and perhaps the DFGWG loop region may contain specific subcellular localization information, in addition to its role in protein stabilization.
Biotechnological implications for (iso)flavonoid engineering
Flavonoids and isoflavonoids constitute a promising group of health-promoting nutraceuticals. Previous studies overexpressing isoflavone synthase (IFS) in the Arabidopsis tt3/tt6 mutant demonstrated the production of isoflavone glycoconjugates (Liu et al., 2002). Co-expression of the characterized malonyltransferase resulted in generation of the more lipophilic malonyl glucoconjugate. Generally, it is believed that free genistein aglycone is highly bioavailable and exerts biological activity (Sfakianos et al., 1997). Nevertheless, studies on the cytostatic and cytotoxic activities of synthetic genistein conjugates have suggested that free genistein aglycone often undergoes rapid biodegradation in human cancer cells, whereas the acylated glucosides of genistein with high lipophilic properties display cytostatic and cytotoxic activities at levels that are several times greater than those of aglycone genistein (Polkowski et al., 2004). In addition, the accumulation of malonyl conjugates may be a favorable trait for engineering isoflavonoid nutraceuticals in planta because it results in storage of the compound in the vacuoles or other specific compartments (Mackenbrock et al., 1992), protected from further potential metabolism.
Medicago trunctula seeds were germinated as described previously (Fedorova et al., 2002). The seedlings were transferred to soil and cultured at 25°C under 14 h light/10 h dark. For drought treatment, 11-week-old plants were deprived of water for 4 days and the leaves and roots were collected separately. The methods utilized for infection of leaves by the fungus P. medicaginis, yeast elicitation of M. truncatula root suspension cells and insect herbivory damage on young seedlings were as described previously (Liu et al., 2003).
All flavonoid and isoflavonoid chemicals were purchased from Indofine Chemical Co. (http://www.indofinechemical.com). Cyanidin 3-O-glucoside (kuromanin), cyanidin-3-O-rutinoside (keracyanin), cyanidin 3,5-O-diglucoside and quercetin 3-O-galactoside (hyperoside) were purchased from ChromaDex (http://www.chromadex.com). All other chemicals and solvents were obtained from Sigma-Aldrich (http://www.sigmaaldrich.com/) unless otherwise specified.
Cloning and expression of acyltransferases
The conserved sequence motifs of the BAHD super-family and AAT family members, DFGWG and NYFGNC, as well as the sequences of several known acyl CoA-dependent acyltransferases, were used to blast search the Medicago trunculata gene index database (now hosted at http://compbio.dfci.harvard.edu/tgi/). The acyltransferase sequences used included anthocyanin 5-O-glucoside-6′′′-O-malonyltransferase (Ss5MaT1) and 5-O-glucoside-4′′′-O-malonyltransferase (Ss5MaT2) from Salvia splendens (Suzuki et al., 2001, 2004b), anthocyanin 3-O-glucoside-6′′-O-malonyltransferase from Chrysanthemum × morifolium (Dm3MaT1; Suzuki et al., 2004a), anthocyanin 5-aromatic acyltransferase from Gentiana triflora (Gt5AT; Fujiwara et al., 1998), anthocyanin 3-O-glucoside-6′′-O-hydroxycinnamoyltransferase from Perilla frutescens (Pt3AT; Yonekura-Sakakibara et al., 2000), anthranilate N-benzoyltransferase from Dianthus caryophylus (Yang et al., 1997), hydroxycinnamoyl CoA:shikimate/quinate hydroxycinnamoyltransferase from Nicotiana tabacum (NtHCT; Hoffmann et al., 2003), acetyl CoA:benzylalcohol acetyltransferase (BEAT) and benzoyl CoA:benzyl alcohol benzoyltransferase (BEBT) from Clarkia breweri (D’Auria et al., 2002; Dudareva et al., 1998), and phenylpropanoyltransferase BAPT from Taxus cuspidata (Walker et al., 2002). When using the sequence fragments for the blast search, the blast matrix was set as pam100 and the expectation at 100; for full-length sequence BLAST, the matrix was set as blosum62 and the expectation at 10. For phylogenetic analysis of AAT proteins, in addition to the AAT sequences mentioned above, the full set of functionally characterized AATs was also included (see Figure 2 legend).
The corresponding EST clones of the putative acyltransferases were obtained from the EST collections of the Samuel Roberts Noble Foundation. The full-length genes were sub-cloned into the Gateway-compatible expression vector pET-DES (O’Maille et al., 2004), following the published procedure (Yu and Liu, 2006). The full-length MtMaT1 cDNA was obtained by mutagenesis of the EST clone using the QuickChange site-directed mutagenesis kit (Stratagene, http://www.stratagene.com/) with the primers, forward 5′-GATGGTGTTTTTTTAGTTGCTAAGAGT-3′ and reverse 5′-AGCAACTAAAAAAACACCATCTTCTTT-3′, to delete the thymidine insertion.
The recombinant proteins were produced in E. coli BL21 (DE3). The Ni+ affinity-purified recombinant enzymes were digested with thrombin and passed through a benzamidine Sepharose 6B column (Amersham Bioscience, http://www5.amershambiosciences.com/), and the resultant enzymes were used for the assays.
Enzyme assay, kinetic analysis and product identification
Enzymatic reactions were performed with 1.5 μg enzyme in 100 μl of 0.1 mm Tris/HCl (pH 8.0) containing 1 mm DTT and 0.1 mg ml−1 BSA, with 100 μm malonyl CoA and 100 μm isoflavone, flavonoid or cyanidine glycosides, respectively. The reactions were incubated at 30°C for 30 min, and then stopped by adding two volumes of 50% acetronitrile; the reaction mixtures were directly subjected to LC-MS analysis. The enzymatic product was identified using LC-MS and collision-induced dissociation (CID) fragmentation analysis as described previously (Yu and Liu, 2006).
To measure the steady-state kinetic constants of MtMaT1, 2 and 3 for (iso)flavone and flavanone glycosides, their activities were determined using various concentrations of genistein or formononetin 7-O-glucoside (5–400μm), naringenin 7-O-glucoside (5–250μm), and naringenin 7-O-neohesperidoside (naringin, 2.5–500 μm) at a fixed malonyl CoA concentration of 1 mm with 1.5 μg enzyme in a total volume of 100 μl, as described above. The kinetic constants of MtMaT1 and MtMaT1-Δ110 for malonyl CoA were assayed using various concentrations of malonyl CoA (5–250 μm) at a fixed concentration of 0.3 mm formononetin 7-O-glucoside. The kinetics of MtMaT2 and 3 for malonyl CoA were measured using malonyl CoA (5–500 μm) at a fixed concentration of 0.5 mm of naringenin 7-O-glucoside, instead of formononetin 7-O-glucoside, because of the limited solubility of the isoflavone compound and the low affinity of the two enzymes to this compound. All kinetic reactions were incubated for 3 and 6 min, at 30°C. The products were quantified by HPLC (Yu and Liu, 2006). All the reactions were run in duplicate, and each experiment was repeated at least twice.
Gene expression analysis by RT-PCR
Extraction of total RNAs from M. truncatula has been described previously (Liu et al., 2003). These RNA samples were treated with DNase I (New England Bio-Labs; http://www.neb.com), and 1 μg of each was used for cDNA synthesis. A 1 μl aliquot of the reverse transcription products was used for the PCR reaction under the following conditions: 94°C for 2 min, then 28 cycles of 94°C for 30 sec, 60°C for 30 sec and 72°C for 30 sec, followed by a final extension at 72°C for 7 min, using primers for MtMaT1 (forward 5′-CCCATTTGAACCACAACTCGAAG-3′; reverse 5′-AAATCTATGAGACCCCGCCACTC-3′); MtMaT2 (forward 5′-CTTCTACGCCCTCCCAAATTCAC-3′; reverse 5′-TGGCAACAAAGTTGGCAGTTCTT-3′); MtMaT3 (forward 5′-TTGGAGCACAATTCACCACAAGA-3′; reverse 5′-CCGCATCCTTAATGATTGGTTGA-3′). The actin gene was used to check for equal amounts of mRNA in the RT-PCR reaction using the primers ActF (5′-CTAGCATAGTAGGACGTCCACGTC-3′) and ActR (5′-GCTCATAGGTCTTCTCAACAGCTG-3′). The resulting PCR products were analyzed by electrophoresis in a 1.2% agarose gel.
Extraction and analysis of isoflavonoids from M. truncatula
Medicago trunctula roots (approximately 0.5 g) grown under normal conditions or drought stress were ground into powder, extracted with 5 ml 80% methanol, and concentrated under an N2 gas stream. The extracts were re-dissolved in 300 μl 80% methanol, and analyzed by LC-MS. The sample was injected into a Gemini or Eclipse C18 reverse-phase column (http://www.phenomenex.com) and resolved in 0.2% formic acid (A) with an increasing concentration gradient of acetonitrile containing 0.2% formic acid (B): 0–5 min, 5% B; 5–15 min, 15% B; 15–20 min, 17% B; 20–25 min, 23% B; 25–60 min, 50% B, at a constant rate of 1.0 ml min−1. UV absorption was monitored at 254, 260, 280, 310 and 510 nm using a multiple-wavelength photodiode array detector (http://www.agilent.com). The tandem MS analysis was performed using an electron spay ionization trap mass spectrometer (Yu and Liu, 2006).
Expression of MtMaT1 in Arabidopsis
MtMaT1 was amplified using the forward primer 5′-GCCTCGAGATGGCTTCCAACAACAAC-3′ and the reverse primer 5′-GCGGATCCAGGGTCAAAACTTAAATC-3′ to introduce XhoI and BamHI enzyme restriction sites (underlined). The amplified cDNA fragment was subcloned into shuttle vector pRT101. The resulting expression cassette containing the 35S promoter and nos terminator were inserted into binary vector pCAMBIA3301 (http://www.cambia.org). The construct was then transferred into Agrobacterium tumefaciens strain C58C1 (Konz and Schell, 1986) using a thaw and freeze method. Arabidopsis line 973 in the Landsberg erecta (Ler) tt3/tt6 double mutant background, which harbors the soybean isoflavone synthase gene and is known to accumulate isoflavone glycosides, and Ler plants harboring the IFS gene that produce normal flavonoids (Liu et al., 2002), were transformed with C58C1 harboring MtMAT1 by the floral-dip method (Clough and Bent, 1998). The leaves of T2 transgenic lines were collected and ground; approximately 0.2 g of leaf powder was extracted with 2 ml of extraction buffer (80% ethanol, 350 mm Tris/HCl, pH 7.2, augmented with 1 μm formononetin as an internal standard) at 4°C for 2 days. The extracts were concentrated under N2 gas, then re-dissolved in 200 μl methanol and analyzed by HPLC using the method described above. Identification of isoflavone products was based on comparison of the chromatographic behavior and UV spectra with those of the malonylated compound produced from the in vitro enzymatic reaction.
Construction and transient expression of the chimericEGFP gene
Plasmid DNAs harboring MtMAT1-GFP, MtMAT1-Δ110-GFP, MtMAT1-ΔDFGWG-GFP, Δ110-GFP or I7OMT-GFP (Liu and Dixon, 2001) under the control of a double 35S promoter were bombarded into N. benthamiana and/or M. truncatula leaf epidermal cells at a pressure of 180 psi using a portable Helios gene gun system (model PDS-1000/He; Bio-Rad, http://www.bio-rad.com/) as described previously (Tzfira et al., 2001). The cells then were incubated for 18–28 h at 25°C to allow expression of the transformed DNA. The Arabidopsis putative BAHD gene (At5g41040) chimeric with GFP and driven by a double 35S promoter on the binary vector was infiltrated into tobacco leaf. GFP fluorescence was detected using a Zeiss AXIOSKOP2 inverted microscope with a Zeiss LSM 5 Pascal laser-scanning confocal attachment (http://www.zeiss.com/).
Homology modeling of MtMaT1 and MtMaT-Δ110
The MtMaT1 and MtMaT-Δ110 amino acid sequences were aligned to that of chrysanthemum anthocyanidin 3-O-glucoside-6′′-O-malonyltransferase using CLUSTAL W version 1.83 (Thompson et al., 1994). Homology models of MtMaT1 and MtMaT-Δ110 were built based on the structure of chrysanthemum anthocyanidin 3-O-glucoside-6′′-O-malonyltransferase (Protein Data Bank code 2E1T; Unno et al., 2007) using the program MODELLER (Marti-Renom et al., 2000). Malonyl CoA was manually docked into the models by superposition of built models with the 2E1T structural complex and extraction of the bound thioester.
We thank Dr Richard Dixon of the Samuel Roberts Noble Foundation for M. truncatula EST clones, Dr Vitaly Citovsky at the State University of New York, Stony Brook, for sharing the confocal microscope, and Michael Blewitt (Brookhaven National Laboratory, USA) for DNA sequencing. The work was supported by the Plant Feedstock Genomics Program from the Department of Energy Office of Biological Environmental Research (130–135), and by Brookhaven National Laboratory’s Laboratory Directed Research and Development program, under contract with the DOE.