Domestication and breeding of citrus species/varieties for flavor and other characteristics, based on the ancestral species pummelo, mandarin and citron, has been an ongoing process for thousands of years. Bitterness, a desirable flavor characteristic in the fruit of some citrus species (pummelo and grapefruit) and undesirable in others (oranges and mandarins), has been under positive or negative selection during the breeding process of new species/varieties. Bitterness in citrus fruit is determined by the composition of branched-chain flavanone glycosides, the predominant flavonoids in citrus. The flavor-determining biosynthetic step is catalyzed by two branch-forming rhamnosyltransferases that utilize flavanone-7-O-glucose as substrate. The 1,2-rhamnosytransferase (encoded by Cm1,2RhaT) leads to the bitter flavanone-7-O-neohesperidosides whereas the 1,6-rhamnosytransferase leads to the tastelessflavanone-7-O-rutinosides. Here, we describe the functional characterization of Cs1,6RhaT, a 1,6-rhamnosyltransferase-encoding gene directing biosynthesis of the tasteless flavanone rutinosides common to the non-bitter citrus species. Cs1,6RhaT was found to be a substrate-promiscuous enzyme catalyzing branched-chain rhamnosylation of flavonoids glucosylated at positions 3 or 7. In vivo substrates include flavanones, flavones, flavonols and anthocyanins. Cs1,6RhaT enzyme levels were shown to peak in young fruit and leaves, and gradually subside during development. Phylogenetic analysis of Cm1,2RhaT and Cs1,6RhaT demonstrated that they both belong to the branch-forming glycosyltransferase cluster, but are distantly related and probably originated separately before speciation of the citrus genome. Genomic data from citrus, supported by a study of Cs1,6RhaT protein levels in various citrus species, suggest that inheritance, expression levels and mutations of branch-forming rhamnosyltransferases underlie the development of bitter or non-bitter species/varieties under domestication.
The suggested origin of true citrus fruit trees is South east Asia, and it is widely accepted that the ancestral citrus species, which gave rise through hybridization to all the edible citrus species/varieties known today, are pummelo (Citrus maxima), mandarin (Citrus reticulata) and citron (Citrus medica) (Spiegel-Roy and Goldschmidt, 1996). Domestication and breeding of citrus species/varieties for flavor aroma and other qualities, based on the ancestral species, has been an ongoing process during the last several thousand years, initially in South east Asia and China, and much more recently in other countries around the globe. New varieties with novel flavors and aromas are continuously generated, and are mostly based on variability in the combinations and concentrations of secondary metabolites. Thus, the composition of secondary metabolites in the citrus varieties known today is the result of selection under domestication, and often reflects human preferences for fruit flavors and aromas, and not necessarily physiological advantages for the tree.
Bitterness is a common flavor characteristic in the fruit of some species of the citrus genus, and is determined by the quantity (concentration) and composition of branched-chain flavanone glycosides, the prevailing flavonoids in citrus (Harborne, 1967; Berhow et al., 1998; Gattuso et al., 2007). The bitter flavanone 7-O-neohesperidosides (e.g. neohesperidin and naringin) are the dominant and, in some cases, the only flavanone glycosides in bitter citrus species (i.e. pummelo, grapefruit and bitter orange), and comprise the branched-chain disaccharide neohesperidose (rhamnose-2-O-glucose) O-linked to position 7 of the flavanone (Horowitz and Gentili, 1961; Jourdan et al., 1985; Peterson et al., 2006a). The tasteless 7-O-rutinosides (e.g. hesperidin and narirutin) are the only flavanone glycosides in non-bitter citrus species (i.e. sweet oranges, mandarins, clementine, citron and lemon), and comprise the branched-chain disaccharide rutinose (rhamnose-6-O-glucose) O-linked to position 7 of the flavanone (Hall, 1925; Rousseff et al., 1987; Peterson et al., 2006b). Beyond the effect on fruit flavor, it is assumed that flavanone glycosides have a role in protecting young citrus tissue against herbivory or disease (Del Rio et al., 2004), because they accumulate to very high concentrations in young tissue (mainly leaves and fruit) and are gradually diluted during continued development (Jourdan et al., 1985; Castillo et al., 1992; Bar-Peled et al., 1993; Ortuno et al., 1995).
Citrus flavanone glycosides are biosynthesized via the phenylpropanoid pathway, similar to other flavonoids (Figure 1) (Winkel-Shirley, 2001). The flavanone naringenin, which constitutes a major junction in flavonoid biosynthesis, is the most abundant flavonoid skeleton in some citrus species, while in others it is methylated or hydroxylated to form other flavanones, such as hesperetin, eriodictyol and isosakuranetin. Citrus flavanones accumulate as glycosides and undergo two glycosylation steps (Figure 1) involving enzymes of the glycosyltransferase 1 family (Vogt and Jones, 2000; Li et al., 2001; Cantarel et al., 2009). The first step involves an O-linked glucosylation at position 7 of the flavanone, catalyzed by a 7-O-glucosyltransferase (7GlcT) (McIntosh et al., 1990; Berhow and Smolensky, 1995). The second step involves one of two rhamnosyltransferases. Flavanone-7-O-glucosides are either metabolized into bitter neohesperidosides (e.g. neohesperidin and naringin) by a 1,2-rhamnosyltransferase (1,2RhaT) (Bar-Peled et al., 1991; Frydman et al., 2004), or into tasteless rutinosides (e.g. hesperidin and narirutin) by a 1,6-rhamnosyltransferase (1,6RhaT) (Lewinsohn et al., 1989). The citrus flavones neodiosmin and diosmin) are similarly products of 1,2- or 1,6- branched-chain rhamnosylation.
We previously reported the isolation and functional characterization of the gene Cm1,2RhaT from pummelo (C. maxima). This gene encodes a flavanone-7-O-glucose-1,2-rhamnosyltransferase, the key branch-forming enzyme directing biosynthesis of the bitter flavanone neohesperidoside compounds in bitter citrus species (Frydman et al., 2004). This enzyme was found to specifically catalyze rhamnosylation of flavanone-7-O-glucose and flavone-7-O-glucose substrates at position 2 of the glucose moiety. Here we describe the isolation and functional characterization of a parallel gene, Cs1,6RhaT, from oranges (Citrus sinensis). This gene encodes a branch-forming flavonoid glucose 1,6-rhamnosyltransferase, the key enzyme directing biosynthesis of the tasteless flavanone rutinosides common to non-bitter citrus species. Cs1,6RhaT was found to catalyze branched-chain rhamnosylation of flavanones, flavones, flavonols and anthocyanins, and is a more ‘promiscuous’ enzyme than Cm1,2RhaT, catalyzing rhamnosylation of both flavonoid-7-O-glucose and 3-O-glucose substrates at position 6 of the glucose moiety. Expression of the enzyme Cs1,6RhaT was found to be prominent in young tissue, and gradually diminished during tissue development. Levels of the enzyme correlate well with the composition of rutinosides in the various citrus species. Phylogenetic analysis of the two citrus branch-forming rhamnosyltransferases demonstrates that they both belong to the branch-forming glycosyltransferase (GT) cluster, but are distantly related and appear to have diverged early in evolution prior to speciation of the citrus genus. The present research, combined with data mining of recently available citrus genome sequences, sheds new light on the molecular basis of evolution of bitter versus non-bitter species under domestication.
Cloning and characterization of Cs1,6RhaT, which encodes a branch-forming flavanone-7-O-glucose 1,6-rhamnosyltransferase
Cloning of a gene encoding flavanone-7-O-glucose 1,6-rhamnosyltransferase was based on a candidate gene approach using the orange (C. sinensis), clementine (Citrus clementina) and mandarin (C. reticulata) EST databases, as these species contain high levels of the rutinoside hesperidin (Hall, 1925; Harborne, 1967; Berhow et al., 1998), and are therefore expected to express a 1,6-rhamnosyltransferase as the major branch-forming glycosyltransferase. Criteria for the candidate EST screen included: (i) relative abundance of the EST in non-bitter citrus species (C. sinensis, C. clementina and C. reticulata), (ii) absent or rare occurrence of the EST in bitter citrus species (C. maxima, Citrus aurantium and Citrus paradisi), and (iii) phylogenetic mapping of deduced amino acid sequence to the branch-forming glycosyltransferase gene cluster. C. sinensis EST sequences meeting the above criteria (#EY700040 and #EY700418) were used to design specific primers for cloning complete cDNA of the candidate gene using 5′ and 3′-RACE (Eyal et al., 1999). The activity of the 1,6RhaT candidate gene-encoded protein was analyzed in vivo by stable expression under the control of the CaMV 35S promoter in transgenic BY-2 cells, as previously described for Cm1,2RhaT (1,2RhaT) (Frydman et al., 2004). Wild-type BY-2 and cell lines transgenic for 1,2RhaT and the 1,6RhaT candidate gene were fed with hesperetin or naringenin substrate for 48 h, and the resulting biotransformation products were extracted and identified by accurate-mass LC/MS based on retention time, the mass of the molecular ion and the resulting fragments (Figure 2). We note that primary glycosylation of hesperitin and naringenin to 7-O-glucosides occurs naturally within the BY-2 cells (Taguchi et al., 2000; Frydman et al., 2004), which in turn serve as substrates for catalysis by transgenically expressed branch-forming GTs. The results demonstrate that branched-chain rhamnoglucoside modifications of hesperetin or naringenin were produced only in the transgenic (1,2RhaT and 1,6RhaT candidate) cell lines but not the BY-2 control (Figure 2; a detailed analysis of all the major modifications occurring during biotransformation of hesperetin and naringenin in the wild-type and transgenic BY-2 cells is shown in Figures S1 and S2). Although the accurate masses of the branched-chain products obtained equally fit both hesperidin/neohesperidin (m/z 611.198) and both narirutin/naringin (m/z 581.187), the retention times clearly distinguish between the products of 1,2RhaT and the 1,6RhaT candidate (Figure 2). The biotransformation products of 1,2RhaT cell lines correspond to the neohesperidosides neohesperidin and naringin, as previously shown (Frydman et al., 2004), while products of the 1,6RhaT candidate cell lines were identified, by comparison with standards, as the rutinosides hesperidin and narirutin (Figure 2). Thus, the 1,6RhaT candidate gene was shown to encode a flavanone-7-O-glucose-6-O-rhamnosyltransferase (named Cs1,6RhaT) that directs biosynthesis of the tasteless flavanones hesperidin and narirutin that are common to non-bitter citrus species.
Cs1,6RhaT is a substrate-promiscuous flavonoid branch-forming rhamnnosyltransferase
The substrate specificity of the enzyme encoded by Cs1,6RhaT was further tested in transgenic BY-2 cells to examine the possibility that it is involved in the biosynthesis of additional flavonoid rutinosides in citrus. The flavone diosmetin was biotransformed into diosmetin glucoside (m/z 460.6) in both wild-type BY-2 and 1,6RhaT transgenic cells (Table 1). However, the rutinoside product diosmin was obtained only in the 1,6RhaT cells, showing that flavone rutinoside biosynthesis in citrus is catalyzed by Cs1,6RhaT. Flavonol rutinoside biosynthesis was also shown to be catalyzed by Cs1,6RhaT; the quercetin substrate was biotransformed to a rutinoside by transgenic Cs1,6RhaT transgenic cell cultures (but not by wild-type cells) (Table 1), but this experiment could not distinguish between two theoretically possible products: a 7-O-rutinoside versus a 3-O-rutinoside, as quercetin fed to the cells may be subject to both 7-O and 3-O primary glucosylation activity endogenous to the tobacco BY-2 cells (Taguchi et al., 2000, 2001).
Table 1. Cs1,6RhaT activity on flavone (diosmetin glucoside) and flavonol (quercetin glucoside) substrates
Primary glycoside (glucoside)
Branched-chain glycoside (rutinoside)
Wild-type (BY-2) and transgenic Cs1,6RhaT-expressing cells (1,6RhaT) were fed with diosmetin or quercetin during the logarithmic phase of growth, as described in Experimental procedures. Cells were harvested after 48 h, and the flavonoid biotransformation products were extracted and identified by comparison of retention time and mass (m/z) to known standards using ion-trap LC/MS.
Diosmetin glucoside (m/z 460.6)
Diosmetin glucoside (m/z 460.6)
diosmin (m/z 607.2; retention time 24.2 min)
m/z 607.2; retention time 24.1 min
m/z 607.2; retention time 24.6 min
Quercetin glucoside (m/z 462.6)
Quercetin glucoside (m/z 462.6)
rutin (m/z 609.0; retention time 20.7 min)
m/z 609.2; retention time 20.6 min
In order to clarify the specificity of the Cs1,6RhaT enzyme with regard to the primary glucosylation site of the flavonoid substrate, we employed the BY-2 biotransformation system and used 7-hydroxyflavone or 3-hydroxyflavone artificial substrates that contain only one hydroxyl group, and therefore can be glycosylated at only one position on the flavonoid skeleton. Primary glucosylation of 7-hydroxyflavone or 3-hydroxyflavone (m/z 401.126) catalyzed by BY-2 endogenous enzymes occurred in all cell lines examined (i.e. wild-type BY-2, 1,6RhaT and 1,2RhaT) (Figures S3 and S4). Both of the transgenic lines (1,6RhaT and 1,2RhaT) further biotransformed the 7-hydroxyflavone substrate into a branched-chain rhamnoglucoside product (Figure 3a; detailed analysis in Figure S3). However, only the 1,6RhaT cell line further biotransformed the 3-hydroxyflavone substrate into a branched-chain rhamnoglucoside product (Figure 3b; detailed analysis in Figure S4), demonstrating the promiscuity of the Cs1,6RhaT enzyme with regard to the primary glucosylation site of the substrate. These results were further confirmed by studying biotransformation of flavonols previously glycosylated at either the 3-O or 7-O positions (kaempferol 3-O-glucoside or kaempferol 7-O-glucoside); kaempferol rutinoside biotransformation products were obtained using both substrates (Figure S5). Thus, the enzyme encoded by Cs1,6RhaT is substrate-promiscuous and catalyzes 1,6-rhamnosylation of both flavonoid 3-O and 7-O mono-glucosylated substrates.
The activity of Cs1,6RhaT on flavonoid 3-O-glucoside substrates raised the question of its ability to catalyze rhamnosylation of anthocyanidin-3-glucosides. Biotransformation of anthocyanidins using the BY-2 system was not an option due to poor stability of anthocyanidins at the pH of the BY-2 cell culture. We therefore studied the activity of a transgenically expressed Cs1,6RhaT on anthocyanidin-3-glucose naturally produced in a red grape (Vitis vinifera) cell-culture system (Gollop et al., 2002). Red grape (RG) cells produce a variety of anthocyanins (see Table S1), including abundant peonidin-3-glcoside and cyanidin-3-glucoside, but no anthocyanidin rutinosides. RG cells were transformed with the plant binary plasmid pME16RT (see Experimental procedures), designed for independent expression of Cs1,6RhaT and the green fluorescent protein (GFP). Transgenic cell lines were confirmed by visualization of GFP fluorescence (Figure 4a) and by immunodetection of the heterologously expressed Cs1,6RhaT protein (Figure 4b). Anthocyanins were extracted from wild-type RG and from the transgenic RG 1,6RhaT cell line, and analyzed by accurate-mass LC/MS (Figure 4c). The data demonstrate that peonidin rutinoside (m/z 609.182; retention time 1.99 min) and cyanidin rutinoside (m/z 595.166; retention time 1.80 min) were produced in the transgenic cell line, but were absent from wild-type RG cells. It was concluded that Cs1,6RhaT can catalyze branched-chain rhamnosylation of anthocyanidin glucosides.
The ability of Cs1,6RhaT to catalyze branched-chain rhamnosylation of anthocyanidin glucosides in transgenic red grape cells raises the question of whether anthocyanidin rutinosides occur in vivo in citrus (none have been described as yet, to the best of our knowledge). Lemon (Citrus limon) accumulates clearly visible levels of anthocyanins in newly emerging leaves, which have been used for analysis of anthocyanin content, and specifically for the occurrence of the rutinoside of cyanidin, the dominant anthocyanidin in citrus (Hillebrand et al., 2004). A peak of m/z 595.166 at retention time 1.80 min was detected in lemon anthocyanin extracts (Figure 4d), corresponding to cyanidin rutinoside. The associated mass spectrum showing the molecular ion (595.166) and the expected fragments (cyanidin glucoside, m/z 449.110; cyanidin, m/z 287.057) further supports this identity. We concluded that Cs1,6RhaT is involved in the biosynthesis of anthocyanidin rutinosides in citrus leaves.
Cs1,6RhaT enzyme levels peak in young fruit and young leaf tissue
The bulk of citrus flavanone accumulation is known to occur in very young leaves and fruit (Jourdan et al., 1985; Castillo et al., 1992; Bar-Peled et al., 1993; Moriguchi et al., 2001), reaching concentrations as high as 14 g/100 g fresh weight in grapefruits of approximately 1 cm diameter (Ortuno et al., 1995). Therefore, it was of interest to determine the levels of Cs1,6RhaT enzyme throughout leaf and fruit development. Cs1,6RhaT-specific antibodies were generated and used for immunological detection of enzyme levels in C. sinensis fruit and leaves. The results demonstrate that the highest levels of the enzyme occur in very young fruit (0.5 cm diameter) or very young leaves (1 cm length), and that enzyme levels decrease gradually during leaf and fruit development (Figure 5). Cs1,6RhaT enzyme was virtually undetectable in mature fruit, and low, but still detectable, levels were present in fully expanded leaves (10 cm).
Cs1,6RhaT enzyme levels are correlated with rutinoside production in citrus species
In previous work (Frydman et al., 2004), we showed that the enzyme 1,2RhaT was present in bitter citrus species (i.e. pummelo, grapefruit and bitter orange), but was not detected in non-bitter species (i.e. orange, mandarin and citron). To complement this work, Cs1,6RhaT-specific antibodies were used to determine enzyme levels in young leaves of bitter and non-bitter citrus species and their correlation with flavanone rutinoside production (Figure 6). Levels of Cs1,6RhaT were high in oranges, mandarins and lemons, in close correlation with the high flavanone rutinoside content. Levels of Cs1,6RhaT were relatively low in citron, which has been shown to contain overall low levels of flavanones (Berhow et al., 1998). Low levels of Cs1,6RhaT were detected in grapefruit and bitter orange, which contain high levels of flavanone neohesperidosides, but low levels of flavanone rutinosides. Cs1,6RhaT was not detected in pummelo, which accumulates abundant flavanone neohesperidosides, but no rutinosides.
Evolution of branch-forming rhamnosyltransferases under domestication – selection underlies the development of bitter and non-bitter citrus species/varieties
The molecular basis for the development of bitter versus non-bitter citrus species/varieties during domestication remains an open question. While data at the protein level (this work, and Frydman et al., 2004) show a strong correlation between fruit bitterness and the levels of the enzymes 1,2RhaT versus 1,6RhaT in various citrus species, the underlying explanation (at the gene level) for the complete lack of 1,2RhaT products in mandarin and orange or the complete lack of 1,6RhaT products in pummelo has not been determined. Two possible explanations may be envisaged: (i) absence of the 1,2RhaT-encoding gene in mandarin and orange and absence of the 1,6RhaT-encoding gene in pummelo, or (ii) extremely low expression levels of the 1,2RhaT-encoding gene in mandarin and orange and of the 1,6RhaT-encoding gene in pummelo. The recent availability of drafts for two citrus genomes, C. sinensis (orange) and C. clementina (clementine, closely related to mandarin: C. reticulata; Saunt, 2000) (http://www.phytozome.net), provides tools to shed some light on this issue. We therefore performed a blast analysis of the clementine genome using Cs1,6RhaT and Cm1,2RhaT sequences. The clementine genome was found to contain a single Cs1,6RhaT ortholog candidate (Cc1,6RhaT), which contains only three amino acid substitutions compared with Cs1,6RhaT (Figure S6). In contrast, we did not detect a Cm1,2RhaT ortholog in the clementine genome. Further support for the absence of a Cm1,2RhaT ortholog in mandarin and its close relatives comes from recent Illumina-based genome sequencing performed on the well-known tangerine variety Murcott (closely related to mandarin). The genome of Murcott was found to contain a single Cs1,6RhaT ortholog candidate containing only one amino acid substitution relative to Cs1,6RhaT (Figure S6), but lacks an ortholog of Cm1,2RhaT.
The existence/absence of a Cm1,2RhaT ortholog in non-bitter species was further studied using the orange genome (C. sinensis). Oranges were apparently derived from a cross between pummelo and mandarin, and may therefore have inherited a Cm1,2RhaT gene from the pummelo parent, but orange fruit is not bitter and does not accumulate flavanone neohesperidosides (Berhow et al., 1998). The orange genome was indeed found to harbor a Cm1,2RhaT ortholog corrupted by frameshift mutations within the coding region (scaffold00124; nucleotides 80 553–79 197 at http://www.phytozome.net/citrus.php). As the available C. sinensis genomic sequence is still a draft, it is not possible to conclude that the Cm1,2RhaT ortholog in orange is in fact corrupted by frameshift mutations without further confirmation. We therefore amplified and sequenced the orange Cm1,2RhaT ortholog from C. sinensis cv. valencia genomic DNA. Comparison of this sequence (named Cs1,2rhaT*) with the sequence encoding Cm1,2RhaT from pummelo (Figure S7) shows 92% identity at the nucleotide level over the entire sequence length. Two frameshift mutations disrupt the Cs1,2RhaT* reading frame at nucleotide positions 278 and 1288 (highlighted in red; Figure S7). RT-PCR analysis using Cs1,2RhaT*-specific primers did not amplify a product, suggesting that there is no expression of this corrupted gene at the mRNA level.
Phylogenetic analysis of Cs1,6RhaT, Cm1,2RhaT and other branch-forming glycosyltransferases
The primary glycosylation position specificity of GTs involved in flavonoid biosynthesis apparently developed early in plant evolution and was maintained during plant speciation. Thus, genes encoding GTs for each specific flavonoid position (i.e. 3-O, 5-O and 7-O) share a common evolutionary origin (Figure 6) (Li et al., 2001). It was therefore of interest to determine whether the position specificity of branch-forming GTs (1,2 and 1,6) was similarly an early specialization in plant evolution that was conserved during plant speciation, or whether branch-forming GTs of different specificity within each genus are related. The amino acid sequences of functionally characterized branch-forming GTs were aligned (together with primary GTs with various position specificities), and used to generate a phylogenetic tree (Figure 7). The results show that all known 1,2 and 1,6 branch-forming GTs that catalyze branched-chain glycosylation of flavonoids/polyphenols constitute a separate branch, and apparently underwent early duplication during plant evolution (i.e. before speciation of the citrus genus). However, there is no support for early divergence of genes encoding 1,6 branch-forming versus 1,2 branch-forming GT activity.
We describe the cloning and functional characterization of Cs1,6RhaT, which encodes a branch-forming rhamnosyltransferase that directs biosynthesis of the tasteless flavanone rutinosides that are common to non-bitter citrus species. We demonstrate that the Cs1,6RhaT enzyme is promiscuous with regard to its substrate specificity, unlike Cm1,2RhaT. Cs1,6RhaT catalyzes branched-chain rhamnosylation of both 7-O-glucoside and 3-O-glucoside substrates, whereas Cm1,2RhaT is specific for flavonoid-7-O-glucosides. This finding correlates well with the composition of flavonoid glycosides found in citrus fruit (Harborne, 1967; Berhow et al., 1998; Gattuso et al., 2007). Flavonoid neohesperidosides (putative products of Cm1,2RhaT catalysis) are found in citrus only with a 7-O linkage (e.g. neohesperidin), while flavonoid rutinosides (putative products of Cs1,6RhaT catalysis) are found with both 7-O and 3-O linkages (e.g. hesperidin and rutin, respectively). Thus, Cs1,6RhaT is apparently responsible for the biosynthesis of flavanone-7-O-rutinosides (e.g. hesperidin and narirutin) flavone-7-O-rutinosides (e.g. diosmin) and flavonol-3-O-rutinosides (e.g. rutin and kaempferol-3-O-rutinoside), which are all present in the non-bitter citrus species (Berhow et al., 1998). Based on our analysis of Cs1,6RhaT substrate specificity, it is also more than likely that flavonol-7-O-rutinosides (e.g. quercetin-7-O-rutinoside and kaempferol-7-O-rutinoside) are produced in citrus, but they are difficult to distinguish from the corresponding flavonol-3-rutinosides as they have the same mass and the same retention time.
We also demonstrated that Cs1,6RhaT can catalyze branched-chain rhamnosylation on cyanidin-3-glucose and peonidin-3-glucose. This result conforms with the relatively high sequence similarity of Cs1,6RhaT to the Petunia hybrida enzyme encoded at the Rt locus (Kroon et al., 1994), which catalyzes identical activity on anthocyanidin-3-O-glucoside substrate. While cyanidin is the major anthocyanidin in citrus, its rutinoside was not detected in a detailed and comprehensive analysis of the anthocyanins of blood orange fruit (Hillebrand et al., 2004). As anthocyanins in blood orange fruit accumulate late in fruit maturation, while Cs1,6RhaT is expressed mostly in young fruit and is absent in mature fruit, cyanidin rutinoside is apparently not produced in the fruit due to temporal separation of the enzyme and the anthocyanin substrate. Anthocyanin accumulation in young citrus tissue, which overlaps temporally with Cs1,6RhaT expression, occurs exclusively in young leaves and flower buds of lemon and citron. Thus, the data we present showing accumulation of cyanidin rutinoside in young leaves of lemon are consistent with activity of Cs1,6RhaT on anthocyanins specifically in young tissue.
Most citrus varieties known today are the products of evolution and selection under domestication conditions (Saunt, 2000). In this context, cloning of Cs1,6RhaT provides an important component for understanding the molecular basis of evolution of bitterness during domestication/selection of citrus species. The selection during domestication for edible bitter species corresponds at the metabolite level to selection for flavanone neohesperidoside production and against rutinoside production. This is translated at the molecular level to positive selection for the presence, intactness and expression of the Cm1,2RhaT gene, and negative selection against the presence, intactness and/or expression of the Cs1,6RhaT gene. Indeed, we show that the Cs1,6RhaT enzyme is entirely absent in pummelo and its level in grapefruit is low, but Cm1,2RhaT is well expressed in both (Frydman et al., 2004).
Selection for non-bitter citrus species/varieties corresponds at the metabolite level to selection for flavanone rutinoside production and against neohesperidoside production. This is translated at the molecular level to positive selection for the presence, intactness and expression of the Cs1,6RhaT gene, and negative selection against the presence, intactness and/or expression of Cm1,2RhaT. Indeed, we show that Cs1,6RhaT is well expressed in orange and mandarin, but a Cm1,2RhaT ortholog is absent from mandarin-type genomes (clementine and Murcott genomes) or is inactivated due to frameshift mutations in the orange genome.
The finding of a Cm1,2RhaT ortholog in the orange (C. sinensis) genome is in line with the current hypothesis regarding the origin of orange from a cross between pummelo, which harbors a Cm1,2RhaT gene, and mandarin, which lacks this gene. Thus, the Cm1,2RhaT ortholog in the orange genome apparently originated from the pummelo parent and acquired frameshift mutations during domestication and selection to create a non-bitter fruit. Grapefruit and bitter orange, which are ‘siblings’ of orange (i.e. also derived from a cross between pummelo and mandarin), retained both the 1,2RhaT from the pummelo parent and the 1,6RhaT from the mandarin parent. In these species, the latter is apparently expressed at much lower levels than the former, giving rise to bitter fruit.
The biological role of branched-chain flavanones in citrus has not been experimentally determined, but they have been suggested to be involved in plant defense against herbivory (Del Rio et al., 2004). This notion is consistent with the accumulation pattern of these compounds: levels are very high in young tissue and gradually decrease as the tissue grows and matures. It is noteworthy that the ancestral citrus species (pummelo, mandarin and citron) contain only one type of branched-chain flavanone each (i.e. neohesperidoside or rutinoside). This may suggest that the biological role assumed by the two types of flavanone branched-chain glycoside products is redundant, and therefore the presence of only one of them is biologically required.
While only few branch-forming glycosyltransferase-encoding genes have been functionally characterized to date (Petunia 1,6RhaT, Kroon et al., 1994; Citrus 1,2RhaT, Frydman et al., 2004; Ipomoea 1,2GlcT, Morita et al., 2005; Bellis 1,2GAT, Sawada et al., 2005; Sesamum 1,6GlcT, Noguchi et al., 2008; Catharanthus 1,6GlcT, Masada et al., 2009; Actinidia 1,2XylT, Montefiori et al., 2011; Arabidopsis 1,2XylT, Yonekura-Sakakibara et al., 2012; Citrus 1,6RhaT, this work), phylogenetic analyses suggest that 1,2 and 1,6 branch-forming GTs catalyzing branched-chain glycosylation of flavonoids/polyphenols all share a common evolutionary origin (i.e. an ancestral branch-forming glycosylation gene/enzyme). However, the phylogenetic tree shows no distinction between genes/enzymes displaying 1,2 versus 1,6 specificity, raising the possibility that the shift between 1,2 and 1,6 specificities may have occurred more than once during plant evolution. Nevertheless, in the case of citrus, the Cm1,2RhaT and Cs1,6RhaT genes are clearly not closely related, and appear to have diverged early in plant evolution, before speciation of the citrus genus. Based on sequence similarity, Cs1,6RhaT appears to share a common evolutionary origin with a Petunia hybrida anthocyanidin-3-O-glucose 1,6RhaT, which catalyzes an identical reaction on anthocyanin substrates.
Leaves and fruit at various developmental stages were collected from orange (C. sinensis), mandarin (C. reticulata), citron (C. medica), lemon (C. limon), pummelo (C. maxima), grapefruit (C. paradisi) and bitter orange (C. aurantium) trees and frozen at −80°C for the purpose of protein purification.
Tobacco BY-2 cell suspension cultures (Nagata et al., 1992) were maintained and transformed as previously described (Frydman et al., 2004). Red grape cell suspension cultures were maintained and transformed as previously described (Gollop et al., 2002).
Genomic DNA from C. sinensis cv. Valencia was extracted from young leaves as previously described (Jacob-Wilk et al., 1999).
Cloning of the gene Cs1,6RhaT
Total RNA was extracted from young C. sinensis leaves as described by Logemann et al. (1987). mRNA was purified for RACE and RT-PCR applications using a PolyAtract mRNA isolation system III (Promega; www.promega.com). 5′-RACE based on the candidate ESTs from C. sinensis (#EY700040 and #EY700418) was performed as described previously (Eyal et al., 1999).The coding sequence of Cs1,6RhaT was amplified by RT-PCR; poly(A) RNA (500 ng) was reverse-transcribed using Superscript II reverse transcriptase (BRL/Life Technologies; www.lifetechnologies.com) and 10 pmol tailed oligo(dT) primer (5′-GTTTTCCCAGTCACGACGTTTTTTTTTTTTTTT-3′). Amplification was achieved using the oligonucleotides 5′-TCTAGAATGCACGCCCCTTCGAACCAAC-3′ and 5′-GAGCTCTTAAGCTAAGGCTTTGAGATCC-3′ and Pyrobest DNA polymerase (Takara; www.takara-bio.co.jp). The amplified fragment was cloned into the binary vector pME504 at the XbaI and SacI restriction sites under the control of a CaMV 35S promoter.
Biotransformation of flavonoids using a transgenic BY-2 cell culture
The pME504 construct harboring Cs1,6RhaT was used to stably transform a BY-2 cell culture (Nagata et al., 1992) as described previously (Frydman et al., 2004). To initiate biotransformation studies, transgenic and wild-type cell cultures were diluted (400 μl culture into 20 ml new medium), and grown in Erlenmeyer flasks for 4 days. Flavonoid aglycones or glycosides (obtained from Sigma-Aldrich and Extrasynthese) were dissolved in dimethylsulfoxide to a concentration of 20 mM, and added to the flasks to a final concentration of 0.15 mm. The cultures were grown for an additional 50 h before harvest. The cell cultures and media were extracted using water-saturated n-butanol. The n-butanol phase was transferred to a new tube and dried.
Extraction of anthocyanins from red-grape cell suspension cultures
Cell suspensions were centrifuged at 700 g, 20°C for 5 min, the medium was discarded, and the cells were washed with cold H2O. Cells were collected again by centrifugation at 700 g, 20°C for 5 min, and were resuspended in extraction solution (70% methanol/0.1% formic acid) (2 ml solution for 1 ml cells) and stored overnight. Extraction solution containing the extracted anthocyanins was directly used for analysis by accurate-mass LC/MS.
Analysis of flavanone glycoside products by LC/MS
LC/MS analyses were performed using a Waters (www.waters.com) Acquity ultra-performance liquid chromatograph coupled to a hybrid quadropole time-of-flight mass spectrometer (UPLC Q-TofMS Premier, Waters) or an Agilent (www.chem.agilent.com) 1100 HPLC coupled to a Bruker Daltonics (www.bdal.com) Esquire ion-trap mass spectrometer.
For the Acquity UPLC Q-TofMS Premier, samples were filtered through a Millipore (www.millipore.com) Millex-HV Durapore poly(vinylidene difluoride) membrane (0.22 μm) before injection into the LC/MS instrument. Sample analyses were performed by UPLC Q-TofMS with the UPLC column connected online to a photodiode array detector, followed by a Q-TofMS detector equipped with an electrospray ionization (ESI) source (set in ESI-positive mode). Separation was performed on a UPLC BEH C18 column (Waters) (2.1 × 50 mm internal diameter, 1.7 μm. Chromatographic and MS parameters were as follows. The mobile phase consisted of 0.1% formic acid in water (phase A) and 0.1% formic acid in acetonitrile (phase B). The linear gradient program comprised 100–95% A over 0.1 min, 95–5% A over 9.7 min, holding at 5% A over 3.2 min, then return to the initial conditions (95% A) for 4.2 min. The flow rate was 0.3 ml min−1, and the column temperature was kept at 35°C. Accurate mass measurements were performed with the Q-TofMS operated in the ‘V’ mode and at a resolution of 8000 or higher. Mass spectra were acquired with an m/z range of 70–1000 Da using the following settings: capillary voltage 2.8 kV, cone voltage 30 eV, collision energy 5 eV. Argon was used as the collision gas. The Q-TofMS system was calibrated using sodium formate, and leu-enkephalin was used as the lock mass. MassLynx software version 4.1 (Waters) was used to control the instrument and calculate accurate masses.
For ion-trap-based MS, HPLC separations were performed using a reversed-phase C18 column (5 μm, 4.6 × 250 mm) (J.T. Baker; www.avantormaterials.com). Samples were eluted using a linear gradient of 95% A:5% B–5% A:95% B over 90 min at a flow rate of 0.8 ml/min. The mobile phases consisted of 0.1% aqueous acetic acid (A) and acetonitrile (B). Negative-ion electrospray ionization (ESI) mass spectra were acquired using a source potential of 3000 V and capillary offset potential of −70.7 V. Nebulization was achieved using nitrogen gas delivered at a pressure of 482633 Pa. Desolvation was aided by a counter-current of nitrogen at a pressure of 82737 Pa and a capillary temperature of 360°C. Mass Spectra were recorded over the m/z range 50–2200. The Bruker ion-trap mass spectrometer was operated using an ion current control pre-set at 20 000, a maximum acquisition time of 100 m sec, and a trap drive setting of 60. Tandem mass spectra were obtained using automated LC/MS/MS following selection of the two most abundant ions above m/z 200 as precursor ions for MS/MS. Tandem spectra were then acquired using an isolation width of 2.0, a fragmentation amplitude of 0.83, and a threshold setting of 5000. The ion charge control was set at 2000 with a maximum acquisition time of 100 m sec.
Cs1,6RhaT antibody production
The complete Cs1,6RhaT coding sequence was cloned into the pET-28a expression vector (Novagen; www.emdmillipore.com/life-science-research/novagen) in-frame with the N-terminal His tag, and transformed into Escherichia coli BL21(DE3)pLysE cells. A starter culture of transformed cells was grown overnight at 30°C, diluted 1:100, and grown for approximately 36 h at 16°C until it reached an OD600 of 0.6–0.8. Expression was induced by addition of isopropyl thio-β-d-galactoside to a concentration of 0.5 mm, and cells were grown for 24 h at 16°C. Cells were collected by centrifugation, resuspended in lysis buffer (50 mM Tris/HCl, pH 8.0, 600 mm NaCl, 0.2% Triton X-100) and ruptured using a French press (Thermo Scientific; www.thermoscientific.com). The soluble Cs1,6RhaT-His tag fusion protein was purified by Ni-NTA affinity chromatography (Qiagen; www.qiagen.com) according to the manufacturer's instructions. Antibodies were generated against the purified recombinant protein by a service unit (Sigma), and were tested for specific identification of 1,6RhaT using immunoblots containing recombinant 1,6RhaT and 1,2RhaT protein.
Protein extraction, separation and immunoblotting
Plant tissues were frozen in liquid nitrogen, ground to a powder, and proteins were precipitated by addition of 5 volumes (w/v) of cold acetone (−20°C), followed by overnight incubation at −20°C. Samples were than centrifuged for 10 min at 15 000 g at 4°C, and the supernatant was discarded. Precipitated proteins were re-suspended in USB buffer (20 mm Tris/HCl pH 7.5, 8 m urea, 4.5% SDS and 1 m β-mercaptoethanol) for 1 h at 25°C. Protein extracts were centrifuged for 15 min at 15 000 g at 25°C to remove particulate matter, and the supernatant was transferred to a new tube. Protein extracts were separated by 12% SDS–PAGE and transferred onto nitrocellulose membranes (Whatman; www.whatman.com). Membranes were blocked with TBST (Tris-buffered saline containing 0.05% Tween 20) containing 1% skim milk, incubated overnight with Cs1,6RhaT-specific antibody, dressed with a secondary antibody (goat anti-rabbit horseradish peroxidase-conjugated, diluted 1:10 000; Jackson Immuno Research Laboratories; www.jacksonimmuno.com) for 1 h, and finally developed using Super-Signal West Dura substrate (Thermo Fisher Scientific/Pierce; www.piercenet.com). Decorated immunoblots were visualized by chemiluminescence using an MF-ChemiBis 3.2 bio-imaging system (DNR; www.dnr-is.com).
The phylogenetic tree was created using phyogeny.fr pipeline (Dereeper et al., 2008). First, sequences were multiply aligned using ProbCons (Do et al., 2005). Then, gaps were removed, and the curated alignment was subjected to the maximum-likelihood method using PhyML (Criscuolo, 2011). Our confidence in the inferred tree topology was determined by calculating Shimodaira-Hasegawa -like confidence limits.
The genome of the mandarin relative Murcott was sequenced using an Illumina (www.illumina.com) Hi-seq 2000 sequencing system (unpublished data), and assembled by mapping to the C. clementina scaffolds (downloaded from http://www.phytozome.net) using Bowtie (Langmead et al., 2009). The Murcott genome sequences covered 85% of C. clementina scaffolds, which indicates that most of the Murcott genome was sequenced. Cs1,6RhaT and Cm1,2RhaT orthologous sequences in the Murcott genome were queried by BLAST (Altschul et al., 1990).
This research was supported by a research grant from the Chief Scientist of the Israeli Ministry of Agriculture (Biotechnology Panel). Salary support for L.W.S. and D.H., as well as mass spectrometry equipment, were provided by the Samuel Roberts Noble Foundation (Ardmore, OK).