Formation of β-glucogallin, the precursor of ellagic acid in strawberry and raspberry

Highlight Biochemical characterizations of glycosyltransferases in green strawberry and raspberry fruits demonstrate formation of β-glucogallin as being an initial step for biosynthesis of the polyphenolic antioxidants, ellagic acid/ellagitannins.


Introduction
Plant phenols are mostly products of the phenylpropanoid and shikimate pathway, and comprise a large variety of compounds (Quideau et al., 2011). They have been associated with miscellaneous in planta functions including defense against bacteria, fungi, viruses (Akiyama et al., 2001;Reddy et al., 2007), and animal herbivores (Barbehenn and Constabel, 2011) as well as with protection against solar radiation (Harborne and Williams, 2000). In addition, polyphenolic components of major dietary constituents have been linked to a number of potential health benefits (Quideau et al., 2011). Among the bioactive polyphenols are tannins, which can be divided into condensed tannins (proanthocyanidins) and hydrolyzable tannins (ellagitannins and gallotannins). The latter are heterogeneous polymers formed from phenolic acids, especially gallic acid, esterified with simple sugar molecules (Niemetz and Gross, 2005). Recent studies revealed that ellagitannins and ellagic acid (dimeric gallic acid) can be effective remedies against various human diseases such as breast cancer, prostate cancer, and cardiovascular and neurodegenerative disorders, and they are thought to play important roles in long-term health protection (Landete, 2011;Giampieri et al., 2012). Although the biological significance of these compounds is widely acknowledged, there is still little known about the genetic background of their biosynthetic pathway.
Ellagitannins and gallotannins probably derive from 1,2,3,4,6-pentagalloylglucose by addition of further galloyl residues or by oxidation, whereas the precursor is formed by successive galloylation of glucose (Niemetz and Gross, 2001). The biosynthesis starts with the formation of β-glucogallin (1-O-galloyl-β-d-glucopyranose), which is generated by esterification of gallic acid and glucose and gives rise to di-, tri-, tetra-, and pentagalloylglucose by transesterification reactions (Niemetz and Gross, 2005). Although formation of β-glucogallin is the essential step in ellagitannin/gallotannin biosynthesis, only four genes coding for gallic acid UDPglucose glucosyltransferases (GTs) have been isolated from Vitis vinifera (Khater et al., 2012) and Quercus robur (Mittasch et al., 2014) to date. The recombinant enzymes showed promiscuous activity towards several (hydroxyl)benzoic and (hydroxyl)cinnamic acids including gallic acid that resulted in the formation of the related 1-O-acyl glucose esters, while UDP-glucose was acting as the activated donor substrate.
Strawberry and raspberry contain high amounts of bioactive polyphenols, in particular ellagic acid/ellagitannins (Aaby et al., 2005(Aaby et al., , 2012Vrhovsek et al., 2012), being about three times higher than in other fruits or nuts (Häkkinen et al., 1999;Williner et al., 2003;Wang et al., 2008). Thus, we employed strawberry (Fragaria×ananassa and Fragaria vesca) and raspberry (Rubus idaeus) as plant systems to study genes that encode GTs which are able to catalyze the formation of β-glucogallin amongst other 1-O-β-d-glucose esters. Comprehensive in vitro enzyme assays confirmed the activity of the recombinant proteins towards gallic acid. Feeding of isotopically labeled gallic acid to fruits of stable transgenic plants verified the in planta function of GT enzymes in F.×ananassa.

Materials and methods
Plant material Fragaria×ananassa cv. Calypso and Elsanta, F. vesca cv. Yellow Wonder, and stable transgenic FaGT2-silenced (Calypso background) strawberry plants (Lunkenbein et al., 2006) were cultivated in the Call Unit for plant research of the TUM School of Life Sciences in Freising, Germany. The transgenic lines were rejuvenated continuously by collecting clones from stolons, which strawberry plants regularly form. Fruits were collected from April until August 2013 and 2014, directly freeze-dried after harvest, and stored at -20 °C until processed further. Rubus idaeus cv. Tulameen and F. vesca (unknown accession) were grown on the campus of Fondazione Edmund Mach in Vigalzano di Pergine, Italy. Fruits were collected in summer 2013 (F. vesca) or 2014 (R. idaeus), directly freeze-dried after harvest, and stored at -80 °C until processed further.

Construction of expression plasmids
The full-length ORFs of FaGT2* and FaGT5 were obtained from cDNA of ripe fruit of F.×ananassa cv. Elsanta by amplification with primers introducing BamHI/NotI (FaGT2*) and EcoRI/XhoI (FaGT5) restriction sites (Supplementary Table S1 at JXB online). The genes were subcloned into the pGEM-T Easy vector (Promega, Madison, WI, USA). The sequence of FaGT2 was synthesized by Eurofins Genomics (Ebersberg, Germany), already flanked by BamHI and NotI restriction sites, and subcloned into the pEX-K4 vector. Subsequently, the genes were cloned into the pGEX-4T-1 vector (Amersham Bioscience, Freiburg, Germany) in-frame with the N-terminal tag. Sequencing of the complete insert (Eurofins Genomics) confirmed the identity of the cloned sequences. The fulllength ORF of FvGT2 was amplified from cDNA of turning fruits (stage 3) of F. vesca and cloned into pGEX4T-1 as a BamHI/NotI fragment. The full-length ORF of RiGT2 was amplified from cDNA of turning fruits (stage 3) of R. idaeus 'Tulameen' and cloned into pGEX4T-1 as a BamHI/EcoRI fragment. Primer sequences to amplify RiGT2 were obtained from Judson Ward (personal communication). All constructs were verified by DNA sequencing.

Site-directed mutagenesis
Site-directed mutagenesis of FaGT2* was carried out according to the QuikChange ® protocol (Agilent Technology, Santa Clara, USA). The primers were designed as described in the manual, and the success of the mutation was confirmed by sequencing (Supplementary  Table S2).

Heterologous protein expression
The recombinant proteins were expressed in Escherichia coli BL21 (DE3) pLysS (Novagen, Darmstadt, Germany). Preparatory cultures were grown overnight at 37 °C in LB medium containing 100 µg ml −1 ampicillin and 23 µg ml −1 chloramphenicol. The next day, 1 liter of LB was inoculated with 10 ml of the pre-culture and grown at 37 °C at 160 rpm until the OD 600 reached 0.5-0.7. To induce protein expression, 1 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG) was added and the cultures were kept at 16-18 °C at 160 rpm overnight before they were harvested by centrifugation (10 min, 5000 g) and stored at -80 °C.

Cell lysis and purification
Recombinant glutathione S-transferase (GST) fusion proteins were purified by GST Bind resin (Novagen) following the manufacturer's instructions. Cells were disrupted by sonication (3 min, 5 s intervals, 50% intensity; Bandelin Sonoplus, Berlin, Germany) after re-dissolving them in 10 ml of binding buffer containing 10 µM of the proteinase inhibitor phenylmethylsulfonyl fluoride. The crude protein extract was incubated for 2 h with the resin in order to bind the GST fusion proteins. The recombinant proteins were eluted with GST elution buffer containing 100 mM reduced glutathione and quantified (Bradford, 1976). The identity of the heterologously expressed proteins was verified by SDS-PAGE and western blot (anti-GST antibody).

Radiolabeled enzyme activity assay and kinetics
Substrate screens were performed at optimized conditions (100 mM Tris-HCl pH 7.0, 30 °C), containing 2 µg of recombinant protein, 1 mM substrate, and 10 010 pmol UDP-glucose (9980 pmol unlabeled and 30 pmol labeled UDP-[ 14 C]glucose). Empty vector protein extract was processed under identical conditions and employed as the negative control. All reactions were stopped by addition of 1 ml of water-saturated 1-butanol. The mixture was vortexed and the phases were separated by centrifugation (2 min at 12 000 g). An aliquot of the organic phase (800 µl) was mixed with 2 ml of Pro Flow P+ cocktail (Meridian Biotechnologies Ltd, Epsom, UK) and radioactivity was determined by liquid scintillation counting (LSC; Tri-Carb 2800TR Perkin Elmer, Waltham MA, USA). The kinetic data were determined with increasing concentrations of substrate (50 µM to 1 mM) and the same fixed mixture of labeled and unlabeled UDP-glucose (10 010 pmol) which was used for the screens. Assays were stopped after 30 min. To determine the kinetic data of UDP-glucose, a fixed concentration of 4-hydroxybenzoic acid (200 µM) was used and varying mixtures of unlabeled UDP-glucose and UDP-[ 14 C]glucose (BIOTREND Chemikalien GmbH, Koeln, Germany) ranging from 50 µM to 1 mM. The kinetic data were calculated from Hanes-Woolf plots. To evaluate the obtained constants, non-linear regression of the Michaelis-Menten equation was used. Only data that matched both calculations were included in the Results section. The GST tag was not removed prior to determination of the kinetic data. Therefore, we used the molecular weight of the fusion protein for calculation of k cat .

Sample extraction
For quantification of secondary metabolites in strawberry fruit, samples were extracted according to Ring et al. (2013). Each tissue sample was ground individually using a mortar and pestle and extracted with methanol. The extracts were evaporated to dryness in a vacuum concentrator and re-dissolved in 50 µl of water. After 1 min of vortexing, 10 min of sonication, and 10 min of centrifugation at 16 000 g, the clear supernatant was used for LC-MS analysis.

LC-MS analysis
The levels of gallic acid, β-glucogallin, and ellagic acid were determined in two tissue types (achenes and receptacles) and five ripening stages (small green, big green, white, turning, and ripe) of F.×ananassa cv. Calypso fruits and three ripening stages (small green, white, and ripe) of F. vesca cv. Yellow Wonder. It was not possible to distinguish the big green stage from the small green stage and turning from ripe fruit of the Yellow Wonder genotype, because this genotype produces small fruits and turns from green to white ripe fruits very quickly. The ripe fruits do not turn red in color. On account of this, we harvested three ripening stages and separated achenes from the pulp. The three metabolites were identified and quantified by LC-MS analysis in the positive and negative MS mode by the internal standard method (Ring et al., 2013). The values were expressed as per mil (‰) equivalent of the dry weight. The identity of the metabolites was confirmed by reference compounds run under identical conditions. Levels of labeled and unlabeled compounds were determined with an Agilent 1100 HPLC/UV system (Agilent Technologies) equipped with a reverse-phase column [Luna 3u C18(2) 100A, 150 × 2 mm; Phenomenex] and connected to a Bruker esquire3000plus ion-trap mass spectrometer (Bruker Daltonics). The system was adjusted, analysis was performed, and metabolite levels were measured according to Ring et al. (2013). As reference compounds, chemically pure gallic acid, β-glucogallin (LC Scientific Inc., Concord, Ontario, Canada), and ellagic acid were purchased and run under identical conditions in order to identify their retention times and mass spectra (MS and tandem MS).

Transcriptome data set of F. vesca cv. Yellow Wonder tissues
The expression of FaGT2 (AY663785; gene26265) and FaGT5 (DQ289586; gene26249) corresponding transcripts in three stages of green fruit development of F. vesca cv. Yellow Wonder was evaluated in a transcriptome data set publicly available at the SGR database Kang et al., 2013; http://bioinformatics.towson.edu/strawberry/newpage/Compare_Samples. aspx, accessed August 2015).

Results
Selection of putative gallic acid UDP-glucose glucosyltransferases from F. vesca, F.×ananassa, and R. idaeus Sequence alignment and a phylogenetic comparison of three UDP-glucose:gallic acid GTs from Vitis vinifera (VvgGT1-VvgGT3; Khater et al., 2012) with putative GTs encoded by the F. vesca genome (Shulaev et al., 2011) yielded two sequences whose closest homologs in F.×ananassa have been named FaGT2 and FaGT5 ( Supplementary Fig. S2). Recently, FaGT2 has been shown to be involved in the formation of cinnamoyl and 4-coumaroyl glucose ester in F. × ananassa (Lunkenbein et al., 2006;Landmann et al., 2007). Re-amplification of the ORF of FaGT2 from cDNA obtained from F.×ananassa cv. Elsanta produced a sequence FaGT2* that differed from the published version in 18 single nucleotide polymorphisms (SNPs). These SNPs resulted in eight amino acid changes and were located before and after the Plant Secondary Product Glycosyltransferase (PSPG) box, a conserved region assumed to be responsible for the interaction with the sugar donor ( Fig. 1). In addition to both FaGT2 sequences and FaGT5, we selected GT2 orthologs from F. vesca (FvGT2) and R. ideaus (RiGT2) for further analysis. Two mutants of FaGT2* were also generated by site-directed mutagenesis to test the effect of the amino acid changes between FaGT2* and FaGT2 on the catalytic activity and substrate specificity of the enzymes. VvgGT1 was used as a positive control (Khater et al., 2012). The GT2 protein sequences and VvgGT1 show a pairwise identity of >80% (Fig. 1), while GT2 homologs from strawberry and raspberry even have an identity of >93%. The genes were cloned into the expression vector pGEX-4T-1. The recombinant proteins were expressed with an N-terminal GST tag and enriched by affinity purification. The presence of the enzymes was verified by SDS-PAGE and western blot.

Enzymatic activity of GT2 homologs
A set of (hydroxy)benzoic acid and (hydroxy)cinnamic acid derivatives was selected to elucidate the substrate specificity of the GT2 homologous enzymes in vitro. The substrates gallic acid, 4-hydroxybenzoic acid, and protocatechuic acid were chosen because they are known precursors of hydrolyzable tannins (Haslam and Cai, 1994), while others show structural similarity (salicylic acid and gentisic acid) or their O-acyl glucoside has been reported in strawberry, raspberry, and grape (e.g. cinnamic acid and p-coumaric acid). All recombinant proteins converted gallic acid to the corresponding 1-O-acyl glucose ester (Table 1). FaGT2* preferred 4-and, 3-hydroxybenzoic acid and vanillic acid, but also efficiently (>90% relative activity) converted protocatechuic acid and m-coumaric acid. FaGT2 showed a preference for the benzoic acid derivatives as well, and esterified veratric acid with the highest efficiency. The enzyme activity of FvGT2 and RiGT2 indicated a higher affinity for cinnamic acid derivatives, and both enzymes glucosylated benzoic acid derivatives likewise with a comparatively high activity. With the exception of veratric acid, FaGT5 clearly favored hydroxycinnamic acid derivatives (sinapic acid 100%). VvgGT1 preferred m-coumaric acid but also converted veratric acid and sinapic acid with a high efficiency. All enzymes were able to use sorbic acid as the acceptor molecule and, except for FaGT2, showed a low activity towards salicylic acid and gentisic acid. The formation of ester bonds was confirmed by alkaline hydrolysis and visualized by LC-MS, exemplarily shown for sinapoyl glucose ester ( Supplementary Fig. S3).

Biochemical characterization
To determine the kinetic values, assays were performed with optimized conditions (100 mM Tris-HCl buffer, pH 7, 2 µg of purified enzyme, 30 °C). For biochemical characterization of UDP-glucose, a fixed p-hydroxybenzoic acid concentration (200 µM) was used. The kinetic parameters were determined from Hanes-Woolf plots and non-linear regression of the Michaelis-Menten equation. The results confirm the high activity of the tested enzymes towards gallic acid, except for FaGT5 (Table 2). FaGT2*, FaGT2, FvGT2, RiGT2, and VvgGT1 show a higher substrate affinity (lower K M ) and enzyme efficiency (k cat /K M ) towards almost all of the tested substrates including gallic acid than FaGT5. FaGT5 preferred sinapic acid, as the k cat /K M value of 15555 s −1 M −1 for this acid exceeded the corresponding values of the other substrates tested.
Site-directed mutagenesis of FaGT2* FaGT2* and FaGT2 show a pairwise identity of >98%. Eighteen SNPs result in eight amino acid changes. To test the effects of the amino exchange on the substrate specificity, three amino acids were reversed in FaGT2* by site-directed mutagenesis. Because no amino acid change was observed in the PSPG box, the presumed region interacting with the UDP-sugar, we decided to mutate amino acids in the vicinity of the PSPG box (E420 and I422) and an amino acid (R230) close to a region which was identified by Bönisch et al. (2014a) to be in the vicinity of the active site. In mutant FaGT2*_R230S, an arginine at position 230 was mutated into a serine while in mutant FaGT2*_E420D_I422V two amino acids, glutamate and isoleucine, were changed to aspartate and valine, respectively. A substrate screening was performed with 18 acids, and the difference (Δ) in the relative activities of the mutants towards the substrates was calculated against the relative activities of FaGT2* (Fig. 2). Both mutants showed an enhanced enzymatic activity towards gallic acid (35% and 38% for FaGT2*_R230S and FaGT2*_ E420D_I422V, respectively), o-coumaric (24% and 39%), and caffeic acid (25% and 31%), whereas formation of sorbic acid glucose ester was decreased (35% and 24%).

Metabolite analysis
In F.×ananassa cv. Calypso, the precursor molecules gallic acid and β-glucogallin and the final product of ellagitannin biosynthesis, ellagic acid, are strongly enriched in achenes of the green stage (Fig. 3). The levels decreased towards the late stages of ripening. Similar trends were observed for the levels of the ellagitannin precursors in receptacles, as well as for the F. vesca cv. Yellow Wonder tissues. Overall, the level of analyzed metabolites was higher in achenes than in receptacles and higher in F.×ananassa cv. Calypso than in fruits of F. vesca cv. Yellow Wonder (Supplementary  Table S3). Thus, in the Calypso and Yellow Wonder genotypes, the investigated metabolites of ellagitannin biosynthesis are more abundant in the early stages of fruit development than in ripe fruits.

Confirmation of FaGT2*/FaGT2 activity in vivo
To confirm that FaGT2*/FaGT2 act as a gallic acid GT in vivo, aqueous solutions of deuterium-labeled gallic acid were injected into both F.×ananassa cv. Calypso control fruits and FaGT2-silenced fruits of stable transgenic strawberry plants. The generation of the stable antisense transgenic FaGT2i line has been described by Lunkenbein et al. (2006). After 24 h,  the levels of d 2 -gallic acid, and the downstream metabolites d 2 -β-glucogallin, d 2 -gallic acid glucoside, d 2 -di-galloyl glucose, and d 2 -ellagic acid were quantified by LC-MS (Fig. 4). A higher level of d 2 -gallic acid and d 2 -gallic acid 4-O-glucoside and a lower level of d 2 -ellagic acid were detected in the transgenic fruits compared with control fruits. The 4-O-glucoside was unambiguously identified by LC-MS in comparison with authentic reference material ( Supplementary Fig. S4). Since the pathway to β-glucogallin is severely blocked, the surplus gallic acid probably results in a shift of the downstream metabolism in favor of formation of gallic acid 4-O-glucoside. In contrast, the amount of d 2 -β-glucogallin and d 2 -di-galloyl glucose remained unchanged.

Discussion
Putative gallic acid UDP-glucose GTs in F. vesca, F.×ananassa, and R. idaeus GTs comprise a diverse class of enzymes that transfer activated sugar compounds onto a large number of acceptor molecules, including secondary metabolites, proteins, lipids, and other sugars (Yonekura-Sakakibara and Hanada, 2011). The classification of this enzyme class is based on the catalytic mechanism and on sequence and structure homologies (Campbell et al., 1997;Coutinho et al., 2003). FaGT2*, FaGT2, FvGT2, RiGT2, and FaGT5 were selected as candidates as they are close homologs to VvgGT1-VvgGT3  Supplementary Table S1. Achenes were removed from receptacles post-harvest and analyzed separately. n.a., not analyzed; n.d., not detectable, n=3-5 biological replicates. ( Fig. 1), three gallic acid GTs from V. vinifera. They belong to the family 1 GTs, also known as UDP-sugar-dependent GTs, that often show specificity towards the donor molecule, mostly UDP-glucose, but they also glycosylate diverse small molecules (http://www.cazy.org/,accessed August 2015; Jones et al., 2003;Sawada et al., 2005). GTs cluster into phylogenetic groups, and it has been shown that within these groups proteins can have comparable substrate specificity (Radominska-Pandya et al., 1999;Ross et al., 2001;Gachon et al., 2005). The phylogenetically related GTs investigated in this study produce 1-O-acyl-esters of (hydroxy)benzoic and (hydroxy)cinnamic acids and share a pairwise identity of >80% with VvgGT1 from V. vinifera. The GT2 homologous proteins show >90% similarity. FaGT2* and FaGT2 differ in only eight amino acids. These two sequences are most probably alleles with similar activity and arise from the octaploid genome of the garden strawberry (Pauli and Kutchan, 1998;Bönisch et al., 2014a, b). Unlike FaGT5 and VvgGT1, the GT2 proteins from strawberry and raspberry have a C-terminal overhang of ~80 amino acids (Fig. 1). Similarly, UGT84A13, a hydroxybenzoic acid GT from Q. robur, shows 510 amino acids in total (Mittasch et al., 2014). The function of the additional sequence remains unknown.

Enzymatic activity and biochemical characterization
VvgGT1 from V. vinifera was employed as a positive control in the substrate screens and the kinetic assays because among the three published VvgGT enzymes, it exhibited the highest efficiency towards gallic acid (Khater et al., 2012). All tested enzymes showed a broad substrate tolerance in vitro. FaGT2 has already been reported to glucosylate a multitude of natural and xenobiotic compounds (Landmann et al., 2007). Similarly, a bi-functional resveratrol/hydroxycinnamic acid GT of Concord grape (Vitis labrusca) showed 99% sequence identity with VvgGT2 (corresponds to an exchange of one amino acid) and glycosylated (hydroxyl)cinnamic acids and the stilbene resveratrol (Hall and Luca, 2007). The overall acceptor substrate preference of GT2 proteins and VvgGT1 was similar, except for salicylic acid and gentisic acid, while FaGT5 showed a clear bias for (hydroxyl)cinnamic acids. Kinetic data revealed gallic acid, sinapic acid, and syringic acid as the favored substrates for the GT2 enzymes, suggesting an overall preference for tri-functionalized acceptors with functional groups in the o-, m-, and p-position. The K M and k cat values for gallic acid of all GTs studied showed a narrow range from 72 µM to 264 µM and from 0.8 s −1 to 1.8 s −1 , respectively. The highest specificity constants k cat /K M of 20 000 s −1 M −1 and 15 277 s −1 M −1 for gallic acid were calculated for RiGT2 and VvgGT1, respectively. These numbers slightly exceed the values published for VvgGT1-VvgGT3 (4231, 3194, and 3873 s −1 M −1 , respectively; Khater et al., 2012) but are clearly lower than the value of the gallic acid GT UGT84A13 from Q. robur (25 952 s −1 M −1 ; Mittasch et al., 2014). In contrast, biochemical characterization confirmed that FaGT5 favored sinapic acid as the phenolic acid substrate. Strawberry fruit (F.×ananassa) has been identified as a rich source of sinapic acid (450 µg g −1 DW; Russell et al., 2009) and its hexose derivatives were found to accumulate during the mid and late stage of receptacle development (Fait et al., 2008) where they might contribute to monolignol biosynthesis and lignin formation in the fruit vasculature (Humphreys and Chapple, 2002).
Site-directed mutagenesis of three FaGT2* positions showed that R230S and E420D_I422V exchanges enhanced the activity of the GT towards gallic acid, o-coumaric acid, and caffeic acid (Fig. 2). The amino acids are not directly part of the PSPG box, the proposed active site of GTs (Wang, 2009). Nevertheless they have a notable effect on the activity and selectivity of FaGT2*. This demonstrates that amino acid residues outside of the active cleft might also cause a change in substrate specificity, as has been shown for a monoterpenol GT from V. vinifera (Bönisch et al., 2014b).

Metabolite analysis
The strawberry fruit is a rich source of ellagitannins and consists of two principle tissue types, pulp (receptacle) and achenes (seeds), in which these compounds are variably distributed. During the ripening process of the fruit, there seems to be an organ-and development-specific, dynamic fluctuation of the ellagitannin level (Fait et al., 2008;Zhang et al., 2011). In one study, β-glucogallin was mainly detected in the receptacle while higher galloylated forms accumulate in achenes (Fait et al., 2008). In contrast, in a second study, similar to our result, higher levels of ellagic acid, galloyl-glucoses, and ellagitannins were found in achenes and not in the receptacle (Aaby et al., 2005). The concentration of β-glucogallin peaked in the early ripening stages in achenes and the receptacle, and decreased towards the ripe stage of the fruit (Fig. 3). Gallic acid and ellagic acid levels followed a similar trend. In general, the white-fruited F. vesca cv. Yellow Wonder contained less tannins than red-fruited F.×ananassa cv. Calypso.
Gallic acid is the precursor compound of the biosynthesis of hydrolyzable tannins whereas ellagic acid is released from ellagitannins by hydrolysis at the end of the biosynthetic pathway (Niehaus and Gross, 1997). Tannins are part of the plant defense mechanism and are released from plant cells upon attack by fungi, bacteria, and insects (Silva et al., 1997). They inhibit the growth of microorganisms and even offer protection against ruminants through formation of complexes with animal proteins that create an unpleasant sensation (Edelmann and Lendl, 2002). For strawberry plants, it might be beneficial to accumulate astringent and antimicrobial compounds during the generative phase to prevent the regenerative organ from infestation. In the late phase of fruit ripening, the fruit firmness decreases concomitant with an increase of the sugar content. Similarly, the reduction of the tannin level and thus the astringency contributes to the appeal of the fruit.

Correlation of expression pattern of putative gallic acid UDP-glucose GTs and metabolite levels
A transcriptome data set of F. vesca cv. Yellow Wonder tissues of early developmental stages has been reported recently Kang et al., 2013). Transcripts corresponding to FvGT2 are abundant in tissues of the receptacle (cortex and pith) and achenes (embryo, ghost, and wall) of small green fruit, whereas the levels decreased towards the big green stage (stage 5; Supplementary Fig. S5) consistent with the reduced levels of β-glucogallin in later stages of development (Fig. 3). Expression of FvGT2 is 20-to 30-fold higher in the wall of achenes compared with receptacle tissue (Supplementary Fig. S5) and thus correlates with our metabolite data, where small green achenes were identified as the main source of gallic acid, β-glucogallin, and ellagic acid. In contrast, the levels of FaGT5 transcripts almost always peaked in big green fruit ( Supplementary Fig. S5). Quantitative realtime PCR analysis also indicated that FaGT2 transcripts accumulate to high levels in red, ripe strawberry fruit (Lunkenbein et al., 2006). Thus, it appears that in F.×ananassa FaGT2 has a dual function due to the spatio-temporal expression pattern. In green fruit, FaGT2 seems to be involved in the formation of the ellagic acid precursor β-glucogallin and, in ripe fruit, concomitant with the reduction of ellagic acid/ellagitannins it might glucosylate cinnamic acid (Lunkenbein et al., 2006). GTs form a vast and diverse enzyme class that comprise proteins with not only specific but also promiscuous in planta functions (Hall and De Luca, 2007). Enzyme promiscuity might derive from evolutionary adaption to changing environmental conditions, and proteins with a broad substrate spectrum are believed to be more evolved than single-substrate enzymes (Ulusu, 2015). Our results provide evidence that GT2 homologous proteins catalyze the formation of the first known precursor of ellagic acid biosynthesis not only in garden strawberry (F.×ananassa), but also in woodland strawberry (F. vesca) and raspberry (R. idaeus).

Injection of labeled gallic acid suggests FaGT2 in vivo activity
Formation of ellagitannin precursors was quantified in F.×ananassa cv. Calypso control fruits and fruits of a stable transgenic FaGT2 antisense line after feeding with deuteriumlabeled gallic acid (Fig. 4). Fruits of the green developmental stage were selected for injection as FaGT2 expression and levels of gallic acid derivatives peaked at this stage. d 2 -gallic acid accumulated in the transgenic fruits probably due to reduced FaGT2 activity, whereas ellagic acid formation seemed to be inhibited but was not completely blocked, presumably due to 34% residual FaGT2 expression in the transgenic fruits (Lunkenbein et al., 2006). Alternatively, FaGT5 might contribute to the formation of β-glucogallin and successive products, but with significantly reduced efficiency (Table 1). In contrast, the levels of d 2 -β-glucogallin and d 2 -di-galloylglucose were not significantly different in control fruits when compared with levels in the transgenics. It appears that the residual enzymatic activity of FaGT2 is sufficient to maintain a constant level of the intermediates but cannot sustain the flux through the ellagic biosynthetic pathway. A new metabolite of gallic acid accumulated to high levels and was subsequently identified as gallic acid 4-O-glucoside (Schuster and Herrmann, 1985). Thus, the surplus gallic acid resulted in a shift of the downstream metabolism in favor of glucoside production because the pathway to ester formation is severely blocked. Clearly, this result confirms one of the in planta functions of the FaGT2 enzyme as gallic acid GT in immature strawberry fruit.

Supplementary data
Supplementary data are available at JXB online. Table S1. Primers used for cloning. Table S2. Primers used for site-directed mutagenesis of FaGT2*. Table S3. Relative concentration of gallic acid, β-glucogallin, and ellagic acid in different strawberry tissues. Figure S1. Relative concentration of d 2 -β-glucogallin after injection of different amounts of labeled gallic acid into small green strawberry fruits. Figure S2. Amino acid sequence alignment and phylogenetic tree of strawberry (F. vesca) proteins. Figure S3. Confirmation of the formation of ester bonds by FaGT2*. Figure S4. LC-MS analysis of β-glucogallin and gallic acid 4-O-glucoside. Figure S5. Expression of FaGT2 and FaGT5 corresponding transcripts in three stages of green fruit development of F. vesca cv. Yellow Wonder.