Redirection of flavonoid biosynthesis through the down-regulation of an anthocyanidin glucosyltransferase in ripening strawberry fruit.

Strawberry (Fragaria × ananassa) fruit contains several anthocyanins that give the ripe fruits their attractive red color. The enzyme that catalyzes the formation of the first stable intermediate in the anthocyanin pathway is anthocyanidin-3-O-glucosyltransferase. A putative glycosyltransferase sequence (FaGT1) was cloned from a strawberry fruit cDNA library and the recombinant FaGT1 transferred UDP-glucose to anthocyanidins and, to a lesser extent, flavonols, generating the respective 3-O-glucosides. Quantitative polymerase chain reaction revealed that transcripts of FaGT1 were almost undetectable in green fruits, but gene expression increased dramatically in both turning and ripe red fruit, corresponding closely to the accumulation of anthocyanins during fruit ripening. The expression of FaGT1 is fruit associated and negatively regulated by auxin. To elucidate the in planta function of FaGT1, Agrobacterium tumefaciens cells harboring an intron-hairpin construct of a partial FaGT1 sequence were injected into midsized ripening fruits. In about one-third of the injected fruits, this led to significant down-regulation of FaGT1 transcript levels that corresponded to reduced concentrations of anthocyanin pigments in ripe strawberry fruits. In contrast, significant levels of epiafzelechin—formed by anthocyanidin reductase (ANR) from pelargonidin—were identified in FaGT1-silenced fruits, indicating competition of FaGT1 and FaANR for the common anthocyanidin substrate. Thus, FaGT1 represents an important branching-point enzyme because it is channeling the flavonoid pathway to anthocyanins. These results demonstrate a method to redirect the anthocyanin biosynthesis into flavan-3-ol production to increase the levels of bioactive natural products or modify pigments in plant tissues.

Strawberry (Fragaria 3 ananassa) fruit contains several anthocyanins that give the ripe fruits their attractive red color. The enzyme that catalyzes the formation of the first stable intermediate in the anthocyanin pathway is anthocyanidin-3-O-glucosyltransferase. A putative glycosyltransferase sequence (FaGT1) was cloned from a strawberry fruit cDNA library and the recombinant FaGT1 transferred UDP-glucose to anthocyanidins and, to a lesser extent, flavonols, generating the respective 3-O-glucosides. Quantitative polymerase chain reaction revealed that transcripts of FaGT1 were almost undetectable in green fruits, but gene expression increased dramatically in both turning and ripe red fruit, corresponding closely to the accumulation of anthocyanins during fruit ripening. The expression of FaGT1 is fruit associated and negatively regulated by auxin. To elucidate the in planta function of FaGT1, Agrobacterium tumefaciens cells harboring an intron-hairpin construct of a partial FaGT1 sequence were injected into midsized ripening fruits. In about one-third of the injected fruits, this led to significant down-regulation of FaGT1 transcript levels that corresponded to reduced concentrations of anthocyanin pigments in ripe strawberry fruits. In contrast, significant levels of epiafzelechin-formed by anthocyanidin reductase (ANR) from pelargonidin-were identified in FaGT1-silenced fruits, indicating competition of FaGT1 and FaANR for the common anthocyanidin substrate. Thus, FaGT1 represents an important branching-point enzyme because it is channeling the flavonoid pathway to anthocyanins. These results demonstrate a method to redirect the anthocyanin biosynthesis into flavan-3-ol production to increase the levels of bioactive natural products or modify pigments in plant tissues. Strawberry (Fragaria 3 ananassa) is one of the most popular fruit crops worldwide and has emerged as a model for nonclimacteric fruit ripening (Giovannoni, 2001). The strawberry fruit is an aggregate composed of a swollen receptacle and the numerous achenes that are the true fruit (Hancock, 1999). In addition to its flavor, much of the popularity of strawberry fruit is due to the attractive, deep red color caused by anthocyanin pigments. Fruit ripening is triggered by a decline in the levels of auxins released by the achenes in early developmental stages (Perkins-Veazie, 1995). The major pigments of the strawberry cultivar 'Elsanta' are pelargonidin 3-O-glucoside (92%), followed by cyanidin 3-O-glucoside (4%) and minor amounts of other pelargonidin derivatives ( Fig. 1; Bakker et al., 1994). More recently, metabolite analyses revealed that pelargonidin-and cyanidin-3-Oglucoside-malonate as well as anthocyanin-flavanol conjugates such as (epi)afzelechin-pelargonidin 3-O-glucoside also occur in strawberries (Fossen et al., 2004;Yoshida and Tamura, 2005). Interestingly, the achenes and the receptacle differ significantly in their anthocyanin composition (Aaby et al., 2005;Yoshida and Tamura, 2005). Achenes have higher concentrations of malonylated pigments and contain almost equal amounts of cyanidin-and pelargonidin-derived anthocyanins. Quercetin and kaempferol are the major flavonols in strawberry fruits and occur as 3-O-glucosides and 3-O-glucuronides (Ryan, 1971;Häkkinen et al., 1999). It has been reported that anthocyanins are undetectable in green and white fruits and that their levels increase rapidly in turning and red stages (Given et al., 1988;Cheng and Breen, 1991). In contrast, the concentrations of polyphenols and non-tannin flavonoids reach their maximum in green fruits and decrease afterward (Cheng and Breen, 1991;Wang and Lin, 2000).
Whereas the chemical composition of the anthocyanins has been studied in detail, genetic and biochemical information about the last steps in anthocyanin biosynthesis in strawberry fruit is still limited (Almeida et al., 2007). The first stable product of the anthocyanin pathway is formed when a glycosyltransferase attaches a sugar molecule to the hydroxyl group at position 3 on the anthocyanidin aglycone. It has been reported that anthocyanin concentration and flavonoid-3-Oglucosyltransferase activity increase in parallel during fruit ripening (Given et al., 1988). Cheng et al. (1994) isolated a glucosyltransferase from strawberry fruit that preferentially glucosylates biochanin A and several flavonols, but does not accept any anthocyanidins. Recently, an anthocyanidin glucosyltransferase gene has been isolated from strawberry fruits and the recombinant protein preliminarily characterized (Almeida et al., 2007).
From grape (Vitis vinifera), another commercially important fruit crop, a UDP-Glc:flavonoid 3-O-glycosyltransferase (VvGT1) has been cloned that preferentially accepts anthocyanidins in vitro, whereas flavonols are less favored (Ford et al., 1998). Comparing glycosylation activities of recombinant VvGT1 and total enzyme extracts from different tissues, the authors concluded that VvGT1 is mainly responsible for the glucosylation of anthocyanidins, whereas additional glycosyltransferases would account for the formation of flavonol glycosides. This is in agreement with earlier findings that VvGT1 mRNA, unlike transcripts of other flavonoid pathway genes, can only be detected in the skin of red grape varieties and that gene expression of VvGT1 is induced after véraison (Boss et al., 1996a(Boss et al., , 1996b. Recently, the three-dimensional structure of VvGT1 has been reported, which enabled the identification of key amino acids involved in substrate binding and catalytic activity (Offen et al., 2006).
Anthocyanidin 3-O-glucosyltransferases have been isolated from flowers of many ornamental plants in which anthocyanins are the major determinants of flower color, including Gentiana triflora, Petunia hybrida, and Iris hollandica (Tanaka et al., 1996;Yamazaki et al., 2002;Yoshihara et al., 2005). The genome of the model plant Arabidopsis (Arabidopsis thaliana) contains over 100 putative glycosyltransferase sequences, and most of the respective proteins have been expressed heterologously Ross et al., 2001). Of 91 glycosyltransferases tested, 29 have been reported to accept the flavonol quercetin (Lim et al., 2004). However, there is only one report on an enzyme from Arabidopsis that glycosylates anthocyanidins (Tohge et al., 2005).
Here, we report the cloning and biochemical characterization of a glucosyltransferase involved in anthocyanin biosynthesis in strawberry fruit. We provide data on the ripening-related and auxin-controlled expression of this gene. Whereas other techniques have been used to verify the function of glycosyltransferases in planta, this article reports on the RNA interference (RNAi)-mediated down-regulation of an anthocyanidin-3-O-glycosyltransferase gene in a commercially important fruit crop, thus confirming its function in planta.

Phylogenetic Analysis
FaGT1 and three other putative glycosyltransferase genes (FaGT2-FaGT4) were identified among a set of 1,100 strawberry ESTs (Aharoni and O'Connell, 2002) and the full-length sequences were successfully cloned (Lunkenbein et al., 2006a). Further screening of other strawberry cDNA libraries and subsequent application of RACE led to the identification of three additional strawberry glycosyltransferase genes (FaGT5-FaGT7). Phylogenetic analysis (Fig. 2) revealed the relationship between FaGT1 and other strawberry as well as biochemically characterized plant glycosyltransferases. FaGT1 was most similar to the UDP-Glc: flavonoid 3-O-glycosyltransferase (VvGT1) from grape, which preferentially catalyzes the transfer of UDP-Glc to anthocyanidins (Ford et al., 1998). Similarities to glycosyltransferases from Dianthus caryophyllus (DicGT1, DicGT3; Ogata et al., 2004), P. hybrida (PhF3GalT; Miller et al., 1999), Vigna mungo (VmUFGT1; Mato et al., 1998), and G. triflora (GtF3GT;Tanaka et al., 1996) were also found, all of which transfer either UDP-Glc or UDP-Gal to anthocyanidins or flavonols. FaGT1 also shows close homology to UGT78D2 and UGT78D1 from Arabidopsis and can be allocated to group F of Arabidopsis glycosyltransferases. All genes from this group contain intron 2, which is the most widespread intron among the Arabidopsis glycosyltransferases . A 166-bp intron inserted after position 493 of the FaGT1 open reading frame is consistent with the Arabidopsis intron. UGT78D2 has been expressed heterologously and has shown activity with anthocyanidins and flavonols using UDP-Glc as the sugar donor (Tohge et al., 2005). Whereas UGT78D1 was first reported to catalyze the transfer of UDP-Rha to position 3 of kaempferol and quercetin (Jones et al., 2003), it was later shown that UGT78D1 also accepts UDP-Glc (Lim et al., 2004). UGT78D2 has been identified as UDP-Glc:flavonoid 3-O-glucosyltransferase (Lee et al., 2005). Strawberries contain only pelargonidin-and cyanidin-derived pigments, mostly 3-O-glucosides (Bakker et al., 1994). Glycosides of quercetin and kaempferol are the main flavonols in strawberry fruit (Ryan, 1971;Häkkinen et al., 1999).
FaGT1 displayed a broad pH optimum in Tris-HCl buffer with maximal activity occurring between pH 7.0 and pH 8.0. After storage at 4°C for 1 week, the enzyme still showed 70% of its initial activity. The temperature optimum was located at 30°C and product formation was linear for 60 min under these conditions. Under optimal conditions, the maximal enzyme activity was with the substrates pelargonidin and cyanidin, both of which are endogenous strawberry anthocyanidins (Fig. 3;Bakker et al., 1994). Peonidin was also among the preferred substrates, whereas flavonols were poor substrates. It is noteworthy that anthocyanidins lacking substitution at position 5# (pelargonidin, cyanidin, and peonidin) were better substrates than those with a hydroxyl or methoxy group at this position. The substitution of a hydroxyl group by a methoxy group seems to further reduce product formation (Fig. 3). Kinetic constants were determined for UDP-Glc and pelargonidin because it is the most common strawberry anthocyanidin. The K m for pelargonidin was 30 mM and V max was 21 nkat/mg, whereas UDP-Glc had a K m of 1.1 mM and a V max of 73 nkat/mg.

Identification of Reaction Products
FaGT1 formed only a single monoglucoside from each substrate (Fig. 4, A-C). Authentic reference compounds were used to identify the product derived from pelargonidin and the flavonols quercetin and kaempferol. The products and the reference compounds were analyzed by HPLC-diode array detector (DAD) and LC-ESI-MS n . Anthocyanins exhibit characteristic UV/  Vis spectra and the recorded spectrum of the FaGT1 reaction product was identical to that of authentic pelargonidin 3-O-glucoside (Fig. 4E). In LC-ESI-MS n analyses, both the retention times and mass spectra of pelargonidin 3-O-glucoside, quercetin 3-O-glucoside, and kaempferol 3-O-glucoside corresponded closely to those of the respective assay products. Furthermore, the formation of glycosylated products could be confirmed in product ion experiments because the mass spectra of all FaGT1 reaction products were characterized by the loss of one Glc moiety to yield the respective aglycons (Fig. 4D). Together, the findings confirm that FaGT1 forms anthocyanidin and flavonol 3-Oglucosides in vitro.

Spatial and Developmental Expression
The spatial and temporal expression pattern of FaGT1 was studied with quantitative PCR (qPCR). RNA was extracted from vegetative tissue (flowers, roots, runners, leaves, and crowns) as well as from receptacles and achenes of small-sized green (G1), full-sized green (G3), white (W), and red (R) fruits. Following reverse transcription (RT), the respective mRNA levels were quantified using sequence-specific primers. The generated amplicon was sequenced and melting point determination, as well as gel electrophoresis, was applied to every sample to ensure the specificity of the qPCR analyses. FaGT1 clearly showed ripening-related expression in both achenes and receptacles, with the highest transcript levels being detected in fully ripe red receptacles (Fig. 5A). The expression in receptacles was more than an order of magnitude higher than in achenes. FaGT1 transcripts could also be detected in runners, leaves, flowers, and crowns, but only at very low levels similar to the level in green receptacles (data not shown). These results clearly indicate that the expression of FaGT1 is fruit associated and highly ripening related.

Hormonal Control of FaGT1 and Developmental Expression
To investigate whether FaGT1 is under the control of auxin, midsized green fruits were carefully de-achened and gene expression was studied after 5 d. Additionally, de-achened fruits were treated with a lanolin paste containing naphthalene acetic acid (NAA), a synthetic auxin. The expression of FaGT1 increased substantially after 5 d in de-achened fruits and could be largely reversed by the auxin treatment (Fig. 5B). These results show that the expression of FaGT1 is ripening related and negatively regulated by auxin.

Metabolite Studies in Ripening Strawberry Fruit
To study the content of individual flavonoids during strawberry fruit ripening, small-sized green (G1), midsized green (G2), white (W), turning (T), and ripe red (R) fruits were extracted with methanol and analyzed by LC-ESI-MS n (Fig. 6). Only small amounts of pelar-gonidin and cyanidin 3-O-glucoside and pelargonidin 3-O-glucoside-malonate were detected in fruits of the early developmental stages. These compounds showed the highest concentration in ripe red fruits. Unlike its potential precursors, cyanidin 3-O-glucoside-malonate displays two concentration maxima during strawberry fruit ripening. High levels of both kaempferol 3-Oglucoside and quercetin 3-O-glucoside were found in small green fruits and their levels decreased in further developmental stages. However, both flavonol glucosides exhibit a second maximum at late-ripening stages.

Transient Gene Silencing in Ripening Strawberry Fruit
To elucidate the function of FaGT1 in planta, the expression of FaGT1 was down-regulated in strawberry fruit by RNAi using Agrobacterium tumefaciens cells harboring an intron-hairpin construct consisting of two inverted repeats of the FaGT1 sequence (Hoffmann et al., 2006). About one-third (21 of 65) of the fruits injected with pBI-FaGT1i were phenotypically different from fruits injected with the control construct pBI-Intron. Their color was generally less intense and of a different hue compared to the bright red fruits of the controls (Fig. 7, A and B). In all fruits that showed a phenotypic change, a significant down-regulation of FaGT1 expression was determined by quantitative PCR in comparison to the transcript levels in control fruits (Fig. 7C).
The effect of FaGT1 silencing on the metabolite concentrations was analyzed by LC-ESI-MS n . At first, levels of potential metabolites downstream of anthocyanidins and flavonols and the concentrations of phenylpropanoid Glc esters were determined. Due to biological variation and heterogeneous silencing effects, the variance of each compound within a sample group was relatively high. Therefore, statistical methods were applied to identify metabolites with significantly altered concentrations. Because the levels of the individual compounds were not normally distributed, we used the Wilcoxon-Mann-Whitney U test (Hart, 2001;Hoffmann et al., 2006). Only the levels of pelargonidin 3-O-glucoside malonate and pelargonidin 3-O-glucoside decreased significantly (P , 0.05) in fruits injected with pBI-FaGT1i, whereas changes in the levels of other anthocyanin derivatives, flavonol glucosides, and phenylpropanoid Glc esters were not significant, except for cinnamoyl Glc, whose level was higher in the controls (Fig. 8, A and B). To explore also unexpected effects, we applied the software packages MZmine and XCMS for differential analyses of the LC-ESI-MS n data (Katajamaa et al., 2006;Smith et al., 2006). Both computer programs performed untargeted metabolite profiling and uncovered that the levels of epiafzelechin, catechin, and an unknown compound with a pseudomolecular ion m/z 495 were also significantly different in fruits injected with pBI-FaGT1i when compared with control fruits (Supplemental Figs. S1 and S2). The Wilcoxon-Mann-Whitney U test confirmed the findings (Fig. 8, A and B). These results show that pelargonidin is a substrate of FaGT1 in planta, whereas quercetin and kaempferol are probably glucosylated by different enzymes in strawberry fruit. The effect of FaGT1 downregulation on the levels of upstream metabolites is much stronger than that on downstream natural products.

Biochemical Characterization
FaGT1 exhibited a rather broad substrate preference, accepting all tested anthocyanidins and flavonols in vitro (Fig. 3). However, aglycones, belonging to chemically similar subgroups, such as flavones, flavanones, and dihydroflavonols, were not converted. VvGT1, the enzyme from grape that shows greatest similarity to FaGT1, exhibited a comparable substrate spectrum and was also unable to convert morin at a reasonable rate (Ford et al., 1998). This might be due to morin's unusual hydroxyl group at position 2#, which may cause a sterical hindrance for the deprotonation of hydroxyl group 3 by His-20 of VvGT1 (Offen et al., 2006). In accordance with published results, FaGT1 showed highest activity with pelargonidin and cyanidin, the endogenous substrates in strawberries (Almeida et al., 2007). FaGT1 activity decreased along with the substitution pattern (H . OH . OCH 3 ) at the 5# position of the anthocyanidin molecule. Replacement of a B-ring hydroxyl group with one methoxy group (e.g. cyanidin to peonidin) decreased the reactivity slightly, but dimethoxylated malvidin was the poorest substrate. These substitutions seem to hinder the efficient binding of hydroxyl group 3 near the His-22 residue of FaGT1. It is worth noting that there are no anthocyanins with a 5# substitution in strawberries (Bakker et al., 1994). Contrary to its broad substrate tolerance for the flavonoid aglycones, FaGT1 only accepted UDP-Glc as a sugar donor.
In contrast to FaGT1, the VvGT1 from grape showed weak activity with pelargonidin (Ford et al., 1998), which is not an endogenous substrate in grape (Mazza and Miniati, 1993). However, the obtained K m values for anthocyanidins and UDP-Glc are within the same order of magnitude for both enzymes.
The attachment of the sugar was highly regiospecific because FaGT1 formed only anthocyanidin and flavonol 3-O-glucosides. This is consistent with the fact that 3,7-dihydroxyflavone, but not chrysin (5,7-hydroxyflavone), could serve as substrate for FaGT1. In accordance, plant secondary product glycosyltransferases have been reported to exhibit strict regioselectivity  A, Ripening-related gene expression of FaGT1 in receptacle tissue (white bars) and achenes (black bars). RNA was extracted from pooled fruits at small-size green (G1), full-size green (G3), white (W), and red (R) stages. Transcript levels were analyzed by qRT-PCR as described in ''Materials and Methods.'' Gene expression is calibrated against expression in receptacle tissue from full-size green fruit (G3). B, Hormonal control of FaGT1 expression. Achenes were carefully removed at the midsize green stage (G2) and the fruits were harvested after 5 d. One set of de-achened green fruits was then treated with a lanolin paste containing the synthetic auxin NAA. FaGT1 expression was analyzed by qRT-PCR as described in ''Materials and Methods'' and normalized against the expression in untreated strawberries with attached achenes.
toward the position of the attached sugar (Vogt and Jones, 2000).

Gene Expression Studies
Key enzymes of the flavonoid pathway exhibit two maxima in enzymatic activity during strawberry fruit development (Halbwirth et al., 2006). This includes flavonoid 3-O-glycosyltransferase activity when measured with quercetin as the substrate, and Phe ammonia lyase (PAL) activity (Cheng and Breen, 1991). At the transcript level, it has been reported that the dihydroflavonol 4-reductase (DFR), putative chalcone synthase (CHS), and flavonoid 3-hydoxylase (F3H) genes show two peaks in gene expression at early and late developmental stages (Manning, 1998;Moyano et al., 1998), whereas glycosyltransferase genes highly homologous to FaGT1 were only expressed at late developmental stages (Manning, 1998;Almeida et al., 2007).
In this article, we have confirmed the fruit-associated and ripening-related expression of FaGT1. The substantial increases in FaGT1 transcript levels in turning and ripe red fruit agree with our findings (Fig. 6, B and C) and earlier reports that both anthocyanin concentration and flavonoid 3-O-glycosyltransferase activity with malvidin increase during fruit development (Given et al., 1988). In contrast, the levels of polyphenols, non-tannin flavonoids, and cyanidin 3-O-glucosidemalonate are at a maximum shortly after flower opening, peaking again in the late ripening stages (Fig. 6, A and B; Cheng and Breen, 1991;Wang and Lin, 2000). Thus, it can be concluded that at least two different enzymes are involved in the glucosylation of anthocyanidins and flavonols in strawberry: a transferase, which forms the glucosides found at the early ripening stages, and FaGT1, which catalyzes the Glc transfer in ripe strawberry fruits.
In grape berries, most anthocyanin pathway genes are expressed up to 4 weeks after anthesis and after véraison mainly in the skin, whereas VvGT1 transcripts are only detected in the skin of red grapes after véraison (Boss et al., 1996a(Boss et al., , 1996b. Thus, VvGT1 seems to be regulated differently than other anthocyanin pathway genes and appears to be the major control point in anthocyanin biosynthesis. Interestingly, the expression of VvGT1 can be delayed by application of a synthetic auxin (Davies et al., 1997). Therefore, FaGT1 and VvGT1 share negative regulation by auxin, which plays an important role in nonclimacteric fruit ripening.

Function of FaGT1 in Planta
To demonstrate the role of FaGT1 in planta, we used a recently developed transient gene-silencing approach based on RNAi to down-regulate FaGT1 expression in strawberry fruit (Hoffmann et al., 2006). A large number of glycosyltransferases involved in plant secondary metabolism have already been cloned and heterologously expressed (Bowles et al., 2005). However, only in a few cases have their biological products been analyzed in planta (Chong et al., 2002;Jones et al., 2003;Morita et al., 2005;Sepú lveda-Jiménez et al., 2005).
In about one-third of the fruits, the injection of pBI-FaGT1i caused a different phenotype with less intense color (Fig. 7A). In contrast, all fruits injected with A. tumefaciens carrying the pBI-FaCHSi construct showed reduced levels of anthocyanins and FaCHS transcripts (Hoffmann et al., 2006). The reason for the lower efficiency of FaGT1 down-regulation remains unknown, but we suspect that it depends on the part of the genespecific sequence chosen for the cloning of the construct. The average FaGT1 transcript levels in pBI-FaGT1i fruits were about 15% of the levels in fruits injected with a control construct (Fig. 7C). When stable strawberry antisense lines of CHS were analyzed, only fruits with ,25% of the original transcript level showed a detectable change in pigmentation (Lunkenbein et al., 2006b). Although the content of pelargonidin 3-Oglucoside in these fruits was reduced to 7.5% of control levels, the fruits were still colored orange and pink.
A principal problem in silencing glycosyltransferases in plants is the occurrence of enzymes with redundant functions. In our study, the presence of other glycosyl-transferases active on anthocyanidins and not silenced by pBI-FaGT1i can account for the partial phenotype observed. The large number of glycosyltransferase sequences in the genomes of Arabidopsis , Oryza sativa (Ko et al., 2006), and Medicago truncatula (Achnine et al., 2005), and the presence of at least seven glycosyltransferase genes (Fig. 2) in strawberry, suggest Figure 7. Down-regulation of FaGT1 in fruits injected with Agrobacterium cells harboring a FaGT1 intronhairpin construct (pBI-FaGT1i). A and B, Phenotype of control fruits and fruits infiltrated with pBI-FaGT1i. Control fruits (A) infiltrated with Agrobacterium containing the control construct (pBI-Intron) are deeply red colored, whereas FaGT1-silenced fruits (B) show a less intense red pigmentation with a different hue. Photographs were taken 14 d after injection. C, qRT-PCR analysis of FaGT1 expression in strawberry fruits. cDNA was generated from total RNA isolated from single fruits infiltrated with Agrobacterium containing either the silencing pBI-FaGT1i (white bars) or the control pBI-Intron vector (gray bars). qRT-PCR was performed using FaGT1 and FaActin gene-specific primers, the latter used as internal control for normalization. that strawberries should also contain a large multigene family of glycosyltransferases. This consideration is further supported by the fact that the octoploid background of the cultivated strawberry genotype used in this study leads to a theoretical 4-fold increase of genes. One of these enzymes could be compensating for downregulated FaGT1. In a similar case in Arabidopsis, the ugt78d2 mutant, lacking UGT78D2 function closely related to FaGT1 activity, still accumulated small amounts of anthocyanins (21% of the wild type), leading the authors to suggest the activity of other 3-Oglucosyltransferases (Tohge et al., 2005). Thus, the simultaneous silencing of several glycosyltransferase genes is likely to be necessary to obtain a clear white phenotype.
Down-regulation of FaGT1 is accompanied by significantly (P , 0.05) reduced levels of the strawberry pigments pelargonidin 3-O-glucoside malonate and pelargonidin 3-O-glucoside, but the concentrations of other anthocyanins did not change significantly (Fig. 8,  A and B). A second glucosyltransferase (e.g. the one that glucosylates cyanidin at early developmental stages and provides the precursor for the malonated derivative) could compensate for FaGT1 silencing in the case of cyanidin 3-O-glucoside (Fig. 6B). The strongly reduced concentrations of pelargonidin 3-O-glucoside-malonate in pBI-FaGT1i fruits are consistent with the idea that the first modification of the anthocyanidin structure is the attachment of a Glc molecule followed by esterification with malonic acid, and shows that pelargonidin is a substrate of FaGT1 in the late ripening stages in vivo. Weak enzymatic activity of the malonyl transferase at low pelargonidin 3-O-glucoside levels would lead to strongly reduced amounts of the malonylated derivative, whereas the level of pelargonidin 3-Oglucoside remains almost constant. This scenario would explain-at least for the pelargonidin derivativeswhy the level of the 3-O-glucoside-malonate is stronger affected than the level of the glucoside.
Surprisingly, the level of cinnamoyl Glc was also significantly reduced in fruits injected with pBI-FaGT1i, even though FaGT1 showed no activity toward cinnamic acid in vitro. Because we observed a similar effect when FaCHS was transiently silenced, we reason that manipulation of the flavonoid pathway modulates PAL activity through transcriptional and posttranscriptional mechanisms as has been shown by antisense expression of an alfalfa (Medicago sativa) cinnamic acid 4-hydroxylase (Supplemental Fig. S3; Blount et al., 2000;Hoffmann et al., 2006). The modulation of PAL activity by the metabolites of the phenylpropanoid pathway has already been reported and could cause the reduced level of the cinnamic acid derivative (Shirsat and Nair, 1986).
Based on in vitro substrate preference and additional biochemical data, it has been proposed that VvGT1 is primarily responsible for the glucosylation of anthocyanidins in vivo (Ford et al., 1998). In contrast, in the homozygous Arabidopsis ugt78d2 mutant, the total level of anthocyanins (consisting exclusively of cya-nidin glucoside derivatives) and the levels of four flavonol 3-O-glycosides were reduced (Tohge et al., 2005). Thus, because recombinant UGT78D2 glucosylates anthocyanidins (cyanidin, pelargonidin, and delphindin) and flavonols (kaempferol, quercetin, and myricetin) in vitro, it was concluded that the native enzyme catalyzes the glucosylation of both cyanidin and flavonols at the 3 position as UDP-Glc:flavonoid 3-O-glucosyltransferase in planta. Our data provide experimental evidence that FaGT1 is involved only in the formation of anthocyanins in planta, whereas flavonols can be excluded as in vivo substrates. This conclusion is based on the significantly reduced levels of pelargonidin 3-O-glucoside malonate and pelargonidin 3-O-glucoside and the strongly increased concentrations of flavan-3-ols (epiafzelechin, catechin, and epicatechin) in fruits in which FaGT1 transcript level had been genetically downregulated. Because epiafzelechin is produced from pelargonidin by the action of anthocyanidin reductase (ANR), increased amounts of the ANR product can only be explained by enhanced availability of the pelargonidin substrate because average FaANR transcript levels were unaffected in pBI-FaGT1i fruits ( Fig.  9; Supplemental Fig. S4). Thus, the increased level of epiafzelechin confirms the efficient and specific downregulation of FaGT1 activity and provides further evidence for pelargonidin as in planta substrate of FaGT1. The successful redirection of the flavonoid biosynthesis also suggests that the reciprocal regulation of ANR and FaGT1 is important in directing the metabolic flux to either anthocyanins or flavan-3-ols (Lee et al., 2005). But FaGT1 silencing also provides enhanced levels of cyanidin and its precursor leucocyanidin, which are converted by FaANR and leucoanthocyanidin 4-reductase (LAR) to epicatechin and catechin, respectively (Fig. 9). The amounts of both flavan-3-ols increased as a consequence of FaGT1 downregulation (Fig. 8). Because LAR from Fragaria shows strict substrate specificity toward leucocyanidin, the lack of afzelechin in strawberry is conceivable (Almeida et al., 2007). The unknown metabolite with a pseudomolecular ion of m/z 495 might be an anthocyanin derivative (downstream metabolite) according to its mass spectral data (data not shown) and reduced level in pBI-FaGT1i fruits.
In addition to the findings in Arabidopsis, this article reports on gene silencing of an anthocyanidin 3-O-glucosyltransferase from a commercial fruit crop. We could clearly demonstrate that FaGT1 acts on anthocyanidins (pelargonidin and cyanidin), whereas other glycosyltransferases should be responsible for the glucosylation of flavonols (kaempferol and quercetin) in the receptacle. The redirection of the metabolic flux toward flavan-3-ols through down-regulation of FaGT1 offers a new method to increase the levels of these bioactive metabolites in fruit crops. The findings emphasize the necessity of in planta experiments for the elucidation of biological functions and enable novel insights into the flux control of the flavonoid pathway.

Cloning of FaGT1 Genomic Sequence
DNA used as a template for PCR amplification was isolated from strawberry (Fragaria 3 ananassa 'Elsanta') leaves using a commercial kit (Qiagen). Products obtained with primers located at the 5#-and the 3#-end of the fulllength sequences were cloned into the pGEM-T vector (Promega) and sequenced by a commercial sequencing service (MWG Biotech).

Construction of FaGT1 Expression Plasmid
The cDNA library construction and sequence analysis of the ESTs have been described (Aharoni and O'Connell, 2002). The FaGT1 full-length open reading frame was subcloned in the pGEM-T vector (Promega). For the FaGT1 construct, 5#-TGAATTCATGGCACCAGTATCAAACC-3# was used as the forward primer and 5#-TCTCGAGATTGGTTGTAGTCATTTCC-3# as the reverse primer, introducing EcoRI and XhoI restriction sites that were used for subsequent cloning into the pET-29a(1) expression vector (Novagen). The insert was cloned in frame with a coding region for a C-terminal His-tag. The identity of the cloned gene was confirmed by sequencing the complete insert (MWG Biotech).

Heterologous Protein Expression
Escherichia coli BL21 (DE3) pLysS cells (Novagen) were transformed with the pET-29a(1) expression vector containing the FaGT1 open reading frame. Cultures were grown overnight at 37°C in Luria-Bertani medium containing 25 mg/mL kanamycin and 34 mg/mL chloramphenicol. The next day, the cultures were diluted to an OD 600 of 0.06 with Luria-Bertani medium containing the appropriate antibiotics in a final volume of 800 mL. This culture was grown at 37°C to an OD 600 of 0.4 to 0.6, cooled to 16°C, and 1 mM isopropylthio-b-galactoside was added to induce protein expression. After overnight incubation at 16°C to 18°C, cells were harvested by centrifugation and stored at 220°C.

Cell Lysis and Protein Purification
His-tagged protein was affinity purified using the His Bind Quick 900 cartridges (Novagen) as recommended by the manufacturer. Briefly, cells were resuspended in binding buffer and sonicated three times on ice. Viscosity was reduced through incubation with benzonase nuclease (Novagen) and the protein raw extract was applied to the cartridges. Recombinant protein was eluted with a buffer containing 1 M imidazol and kept on ice. Protein concentration was determined (Bradford, 1976) and the presence of recombinant protein was confirmed by SDS-PAGE and western blot using anti-His antibodies (Novagen). The amount of FaGT1 present in the partially purified eluate was quantified using the Alpha Imager 2200 documentation and analysis system and Alpha Ease software (Alpha Innotech). Aliquots of purified protein were stored at 220°C.

Enzyme Assays
Enzyme activity was assayed in a buffer containing 100 mM Tris-HCl (pH 7.0), 10% glycerol, and 10 mM b-mercaptoethanol. Standard assays contained 5 mM UDP-Glc, 200 mM substrate, and protein raw extract with approximately 0.25 mg of recombinant protein in a total volume of 250 mL. Enzyme assays were incubated for 30 min at 30°C and stopped by the addition of 200 mL of 5% HCl (anthocyanidin substrates) or 50 mL of acetic acid (all other substrates). As a control, BL21 (DE3) pLysS cells were transformed with an empty pET29-a(1) vector and the resulting protein extract was assayed under the same conditions. As an additional control, assays were conducted with heat-inactivated enzyme solution (5 min at 95°C). Initial enzyme activity was monitored by detecting the reaction products with LC-ESI-MS n . The identity of the FaGT1 glycosylation product was also confirmed by HPLC-DAD.

Enzyme Kinetics
The biochemical characterization was carried out with radioactively labeled UDP-Glc allowing for detection of the reaction product by liquid Figure 9. Biosynthetic pathway leading to anthocyanins and flavan-3-ols. FHT, Flavanone 3-hydroxylase; ANS, anthocyanidin synthase. Because of the strict specificity of Fragaria LAR for leucocyanidin, afzelechin is not found in strawberry fruit (Almeida et al., 2007). scintillation counting. Assay conditions were essentially the same, except only 0.1 mg of affinity-purified recombinant protein was used in a total volume of 200 mL. The UDP-Glc was a mixture of 1 mL of 0.016 mM [6-3 H]UDP-Glc (1 mCi/mL) and 99 mL of unlabeled 101 mM UDP-Glc. Assays were incubated at 30°C for 30 min and extracted with 1 mL of water-saturated n-butanol. The radioactivity of the products was determined by liquid scintillation counting (LKB Rackbeta 1219) after the addition of 4 mL of Ultima Gold XR LSC cocktail (Perkin-Elmer). The kinetic constants were calculated with SigmaPlot 8.0 software (Systat Software) assuming single-site saturation binding.

HPLC-DAD
The instrument used was a LaChrom HPLC (Merck-Hitachi) equipped with a DAD. HPLC separation was performed with a Phenomenex Luna C-8 column (150 mm long 3 4.6 mm i.d., particle size 3 mm) applying a gradient that went from 100% A (0.05% formic acid in water) to 100% B (acetonitrile) in 30 min at a flow rate of 1 mL/min. Spectra and chromatograms were acquired with the Chromatography Data Station software (Merck-Hitachi).

LC-ESI-MS n
A Bruker Daltonics esquire 3000 plus ion trap mass spectrometer (Bruker Daltonics) connected to an Agilent 1100 HPLC system (Agilent Technologies) equipped with a quaternary pump and a variable wavelength detector was utilized for all experiments. Components were separated with a Phenomenex Luna C-18 column (150 mm long 3 2.0 mm i.d., particle size 5 mm) that was held at 25°C. Enzyme assays were analyzed using a linear gradient that went from 100% A (0.1% formic acid in water) to 100% B (acetonitrile) in 30 min with a flow rate of 0.2 mL/min. For metabolite analyses in strawberry fruit extracts, the gradient went from 100% A to 40% B in 40 min, then to 100% B in 5 min. The detection wavelength was either 520 (anthocyanidins) or 280 nm (other substrates and metabolite analyses). The ESI voltage of the capillary was set to 24,000 V and the end plate to 2500 V. Nitrogen was used as dry gas at a temperature of 300°C and a flow rate of 10 L/min. The full-scan mass spectra were measured in a scan range from 50 to 800 m/z with a scan resolution of 13,000 m/z/s until the ICC target reached 20,000 or 200 ms, whichever was achieved first. Tandem MS was carried out using helium as the collision gas (3.56 3 10 26 mbar) with 1-V collision voltage. Spectra were acquired in the positive and negative ionization mode. Data analysis was performed using the DataAnalysis 3.1 software (Bruker Daltonics).

Auxin Treatment
The hormone treatment was performed as described previously (Medina-Escobar et al., 1997). Briefly, achenes from midsized green fruits were removed carefully with a scalpel blade. One set of fruits was treated with a lanolin paste containing 1 mM NAA in 1% (v/v) dimethyl sulfoxide. The other set of deachened fruits was treated with the same paste without the synthetic auxin NAA. Fruits were harvested after 5 d and immediately frozen in liquid nitrogen.

Developmental Analyses of FaGT1 Expression
Total RNA was isolated from pools of six to seven strawberry fruits from each ripening stage as described (Asif et al., 2000). RNA was treated with DNase I (Amersham Bioscience) to remove any DNA contamination prior to cDNA synthesis. RT was carried out using 1 mg of total RNA and the iScript cDNA synthesis kit (Bio-Rad) as recommended by the manufacturer. qRT-PCR analysis was performed using the iCycler system (Bio-Rad) as described previously (Benítez-Burraco et al., 2003;Raab et al., 2006). FaGT1-specific primers were designed (5#-TGCGGTTGGAACTCGTGCGGTTGGAACTCGGTGCTTGCCT-GTTGTGCGAGTTGTTTTGTGCTT-3# and 5#-GCCTGTTGTGCGAGTTGTTTT-3#) and tested with cDNA from different ripening stages. The obtained amplification products were cloned and subsequently sequenced. The qRT-PCR data for FaGT1 were normalized against the expression levels of an interspacer 26S-18S RNA housekeeping gene (5#-ACCGTTGATTCGCACAATTGGTCA-TACTGCGGGTCGGCAATCGGACGTCG-3# and 5#-TACTGCGGGTCGGC-AATCGGACG-3#). PCR reactions contained 2 mM MgCl 2 , 0.2 mM dNTPs, 0.2 mM of the respective primers, 2.5 mL SYBR-Green I (diluted 1:15,000), 2 mL cDNA, and 0.5 units Taq polymerase (Biotools) in a total volume of 25 mL. The thermal cycling conditions were as follows: 2 min at 94°C, followed by 40 cycles of 95°C for 15 s, 55°C for 30 s, and 72°C for 30 s. Melting-point analysis and agarose gel electrophoresis were performed on every reaction to ensure that only the desired amplicon had been generated. Each reaction was performed at least in triplicate and relative expression levels were calculated by calibrating the results with receptacle expression levels from full-sized green fruits (G3) or untreated controls with attached achenes (Livak and Schmittgen, 2001).

Construction of pBI-FaGT1i
A blunt-end PCR product containing the full-length sequence of FaGT1 was generated using a high-fidelity polymerase (Finnzymes) and the same primers used to subclone the expression vector. This amplicon was then digested with SpeI yielding a fragment of approximately 500 bp, which was used for ligation into the binary vector pBI121 that contained XbaI/NheI and SpeI/SacI (Ecl136II) restriction sites separated by an intron from strawberry (Hoffmann et al., 2006). To do this, the vector was cut with XbaI and a highfidelity polymerase was used to create a blunt end. After cutting the vector with NheI, the 500-bp fragment was ligated in the sense direction. Second, the vector was digested with SpeI and Ecl136II and the 500-bp fragment was inserted in the antisense direction, yielding the intron-hairpin construct pBI-FaGT1i.

Transfection of Strawberries by Agroinfiltration
Strawberry cultivar 'Elsanta' plants were grown under standard conditions at 25°C and a 16-h photoperiod. Transfection was carried out as described previously (Hoffmann et al., 2006). Briefly, Agrobacterium tumefaciens strain AGL0 cells containing the pBI-FaGT1i construct were grown at 28°C until the OD 600 reached approximately 0.8. Cells were harvested and resuspended in modified MacConkey agar medium and injected evenly in fruits still attached to the plant about 14 d after pollination. Ripe fruits with dark achenes were harvested approximately 14 d after injection and stored at 280°C. Control experiments were carried out by injecting strawberry fruits with A. tumefaciens AGL0 cells harboring a pBI-Intron control construct (Hoffmann et al., 2006).

RNA Extraction and Expression Analyses Used for Transfected Fruits
Agroinfiltrated fruits were individually freeze dried in a lyophilizer (Christ ALPHA 1-4) and ground to a fine powder. Fifty milligrams of the powder from each fruit was used for RNA extraction, followed by DNase I treatment and RT, as described before. Five nanograms of cDNA was used for qRT-PCR experiments, carried out with FaGT1, FaANR, and FaActin genespecific primers (Almeida et al., 2007) using an ABI 7900 thermocycler (Applied Biosystems) and Platinum SYBR-Green kit (Invitrogen) according to the manufacturer's instructions. The FaActin gene, having constant expression levels, was used to normalize raw data and calculate relative FaGT1 and FaANR transcript levels.

Metabolite Analysis
Fifty milligrams of freeze-dried strawberry powder was extracted with 250 mL of methanol containing 0.2 mg/mL 4-methylumbelliferyl b-D-glucuronide as an internal standard. Methanol was removed in a rotary vacuum concentrator (Christ RVC 2-18) and the extract was redissolved in 35 mL of water for analysis by LC-ESI-MS n . For metabolite analyses during fruit development, 500-mg samples were extracted with 500 mL of methanol, 0.05 mg of internal standard was added, centrifuged (5,000g, 10 min), and analyzed by LC-ESI-MS n . Metabolite quantification was performed using QuantAnalysis 1.5 software (Bruker Daltonics) normalizing all results against the internal standard. Each analysis was performed in triplicate. Levels of metabolites determined in the RNAi experiment were displayed as box and whisker plots using the software SigmaPlot 8.0 (Systat Software). Statistical significance levels were calculated with the Wilcoxon-Mann-Whitney U Test (Hart, 2001) using the software package R (www.r-project.org).
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession number AAU09442 (FaGT1).

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Comparative HPLC-ESI-MS analysis of the metabolites visualized by MZmine.
Supplemental Figure S3. Effect on the cinnamoyl glucose level by transient silencing of FaCHS.
Supplemental Figure S4. Quantitative PCR analysis of FaANR expression in strawberry fruits.