Cloning and Characterization of Vitis viniferaUDP-Glucose:Flavonoid 3-O-Glucosyltransferase, a Homologue of the Enzyme Encoded by the Maize Bronze-1Locus That May Primarily Serve to Glucosylate Anthocyanidins in Vivo *

We report here the cloning and optimized expression at 16 °C and the characterization of a Vitis vinifera UDP-glucose:flavonoid 3-O-glucosyltransferase, an enzyme responsible for a late step in grapevine anthocyanin biosynthesis. The properties of this and other UDP-glucose:flavonoid 3-O-glucosyltransferases, homologues of the product encoded by the maize Bronze-1locus, are a matter of conjecture. The availability of a purified recombinant enzyme allowed for the unambiguous determination of the characteristics of a flavonoid 3-O-glucosyltransferase. Kinetic analyses showed that k cat for glucosylation of cyanidin, an anthocyanidin substrate, is 48 times higher than for glucosylation of the flavonol quercetin, whereasK m values are similar for both substrates. Activity toward other classes of substrates is absent. Cu2+ ions strongly inhibit the action of this and other glucosyltransferases; however, we suggest that this phenomenon in large part is due to Cu2+-mediated substrate degradation rather than inhibition of the enzyme. Additional lines of complementary biochemical data also indicated that in the case of V. vinifera, the principal, if not only, role of UDP-glucose:flavonoid 3-O-glucosyltransferases is to glucosylate anthocyanidins in red fruit during ripening. Other glucosyltransferases with a much higher relative activity toward quercetin are suggested to glucosylate flavonols in a distinct spatial and temporal pattern. It should be considered whether gene products homologous to Bronze-1 in some cases more accurately should be referred to as UDP-glucose:anthocyanidin 3-O-glucosyltransferases.

We report here the cloning and optimized expression at 16°C and the characterization of a Vitis vinifera UDPglucose:flavonoid 3-O-glucosyltransferase, an enzyme responsible for a late step in grapevine anthocyanin biosynthesis. The properties of this and other UDP-glucose:flavonoid 3-O-glucosyltransferases, homologues of the product encoded by the maize Bronze-1 locus, are a matter of conjecture. The availability of a purified recombinant enzyme allowed for the unambiguous determination of the characteristics of a flavonoid 3-O-glucosyltransferase. Kinetic analyses showed that k cat for glucosylation of cyanidin, an anthocyanidin substrate, is 48 times higher than for glucosylation of the flavonol quercetin, whereas K m values are similar for both substrates. Activity toward other classes of substrates is absent. Cu 2؉ ions strongly inhibit the action of this and other glucosyltransferases; however, we suggest that this phenomenon in large part is due to Cu 2؉ -mediated substrate degradation rather than inhibition of the enzyme. Additional lines of complementary biochemical data also indicated that in the case of V. vinifera, the principal, if not only, role of UDP-glucose:flavonoid 3-Oglucosyltransferases is to glucosylate anthocyanidins in red fruit during ripening. Other glucosyltransferases with a much higher relative activity toward quercetin are suggested to glucosylate flavonols in a distinct spatial and temporal pattern. It should be considered whether gene products homologous to Bronze-1 in some cases more accurately should be referred to as UDPglucose:anthocyanidin 3-O-glucosyltransferases.
Phytochemists have identified some 70,000 plant chemicals, many thousands of which are glycosylated (1). From a chemical point of view, two main properties differ between the glycosylated secondary products and their respective aglycones, as glycosylation invariably results in enhanced water solubility and lower chemical reactivity. Glycosylated compounds are therefore often thought of as transportable storage compounds or indeed waste/detoxification products assumed to lack physiological activity (2).
Despite the widespread occurrence of glycosylated secondary metabolites, including flavonols (3), anthocyanins (4), monoterpenes (5), plant hormones (2), and metabolites of systemic fungicides (6), isolation and characterization of purified enzymes responsible for their metabolism have only been reported in a couple of select instances (7,8). The most widely studied groups of plant glycosyltransferases are those associated with the biosynthesis of flavonoid glucosides, including flavonol glucosides, flavanone glucosides, and anthocyanins (3,4,9,10). Earliest reports included the detection of an anthocyanidin and flavonol glucosylating activity in endosperm extracts of maize (Zea mays), (11)(12)(13). This work formed the basis for identification by transposon tagging of the gene product of the maize Bronze-1 locus (14). cDNAs encoding flavonoid glucosyltransferases have been isolated from a number of plant species utilizing homology to the maize Bronze-1 cDNA. In the grapevine, Sparvoli and co-workers (15) cloned a partial cDNA from the variety Lambrusco based upon homology to a putative flavonoid glucosyltransferase of Antirhinnum majus, itself identified by homology to the maize cDNA (16). Recently, Boss et al. (17)(18)(19) used the "Sparvoli" cDNA to detect the expression of anthocyanidin glucosyltransferase mRNA during the development of berries of Vitis vinifera variety Shiraz and to show that red and white grapes, and color mutants (sports) of several varieties, differed in whether or not they express the UFGT 1 gene. Although several reports on the characteristics of expression levels and transgenesis of the UFGT gene have appeared, a clear picture of the properties of the corresponding gene product is not available. Thus, despite the suspected pivotal role played by UFGT in pigment accumulation, this enzyme has never been subjected to a thorough characterization in a purified and guaranteed homogeneous state. Indeed, a confusing picture of the nature and extent of flavonoid glucosyltransferases has arisen in the very extensive literature on these activities. For example, Jonsson et al. (20), Teusch et al. (21), and Hrazdina (22) conclude a common identity of UDP-glucose: anthocyanidin 3-O-glucosyltransferase and UDP-glucose:flavonol 3-O-glucosyltransferase with the enzyme significantly more active toward the flavonols. In contrast, preliminary analysis of a Gentiana triflora recombinant UFGT in crude bacterial lysates indicated a preference for anthocyanins over flavonols (23). Additionally Do et al. (24), recently concluded that a partially purified UFGT from grapes exhibits activity with anthocyanins but not at all with flavonols.
The confusing picture concerning the multiplicity and properties of anthocyanidin/flavonol 3-O-glucosyltransferases undoubtedly stems from the repeated inability of investigators to purify any such enzyme to homogeneity. Additionally, the highly labile nature of anthocyanidins but not of flavonol substrates at the basic pH optimum of most glucosyltransferases has inevitably contributed further to the confusion regarding the relative activities and observed specificities of these enzymes.
We report here the cloning of a cDNA encoding a full-length UFGT from grapes of V. vinifera, the world's largest fruit crop with annual production of 60 million tons (25). Following development of a protocol which employs very slow growth of Escherichia coli at 16°C, expression of the cDNA and subsequent purification of a histidine-tagged rUFGT in an active form has permitted, for the first time, the unambiguous determination of the properties of a member of this ubiquitous enzyme family and allowed us to highlight some of the pitfalls that could contribute to erroneous conclusions regarding anthocyanidin glucosyltransferases. It has been demonstrated that UFGT exhibits much higher catalytic efficiency against anthocyanidins than against flavonols such as quercetin and kaempferol. Additional data also strongly indicate that in the case of grapes, the principal role of UFGT is to glucosylate anthocyanidins in red fruit, whereas products of other genes serve to glucosylate flavonols in a distinct spatial and temporal pattern.

EXPERIMENTAL PROCEDURES
Biochemicals and Reagents-All biochemicals were of analytical grade or higher. Flavonoid substrates and authentic glucosides were obtained from Fluka, Castle Hill, New South Wales, Australia; Apin Chemicals, Oxford, UK; Carl Roth GmbH, Karlsrü he Germany; Extrasynthèse, Genay, France; and Indofine, Somerville, NJ.
Cloning of Full-length V. vinifera UFGT cDNA-Employing mRNA isolated from whole V. vinifera cv. Shiraz grape berries 10 weeks after flowering (26), a Superscript Choice cDNA synthesis kit (Life Technologies, Inc.), ZAPII arms pre-digested with EcoRI (Stratagene) and EcoRI adaptors, a cDNA library was prepared according to the manufacturer's instructions by C. Davies (CSIRO Plant Industry, Adelaide, Australia). An aliquot of the amplified library, consisting of approximately 150,000 plaques, was screened using a 532-base pair partial cDNA clone for UFGT from grape (15) as a probe. Hybridization was carried out overnight at 65°C in Church-Gilbert buffer (27); washings, also at 65°C, were in 2ϫ SSC. This would allow the detection of nucleic acid species with up to approximately 30% mismatched sequences (28). Two hybridizing clones were isolated, and their nucleotide sequences determined using standard protocols for chain termination sequencing (29) of both encoding and non-encoding strands.
Synthesis and Purification of Recombinant UFGT-PCR amplification reactions were carried out using 1ϫ Vent DNA polymerase buffer, 0.2 mM dNTPs, 1 M each forward (UFGTF1 [5Ј-GCGGATCCGCATA-TGTCTCAAACCACCACCAACCCCCATGTGGCCGTC-3Ј]) and reverse (UFGTR1 [5Ј-GTGTCAAAACCAAAGGATGTCTAGGGATCCAAGCTT-GAATTCCG-3Ј]) DNA primer, 2 mM MgSO 4 , and 0.8 units of Vent DNA polymerase (New England Biolabs) in a 20-l final volume with approximately 1 ng of plasmid pGUT3 DNA template. Thermal cycling parameters were 94°C, 5 min, 30 ϫ (94°C for 1 min, 58°C for 1 min, and 72°C for 2 min) and a final 5 min at 72°C. After digestion of the 1403-base pair amplification product with restriction enzymes NdeI and BamHI, the DNA was ligated into pET14b (Novagen, Madison, WI) and used to transform competent cells of E. coli strain JM109. Plasmid was prepared from recombinant clones and used to transform the E. coli BL21 pLysS strain (30) to yield the clone pET14-VVUFGT used in the subsequent production of recombinant UFGT.
Soluble, active UFGT was expressed from pET14-VVUFGT following growth and induction at 16°C. Briefly, L broth (100 ml) held at 16°C and containing ampicillin and chloramphenicol was inoculated with 100 l of overnight culture (grown at 37°C) and grown until an A 600 between 0.4 and 0.6 was reached (typically 24 -40 h). Induction of recombinant protein expression was initiated by the addition of isopropyl-1thio-␤-D-galactopyranoside to 0.4 mM, and cultures were grown for a further 24 h. Cells were harvested by centrifugation at 1400 ϫ g for 15 min, and the drained pellets stored at Ϫ20°C. Bulk growth (1-liter scale) was carried out in 5-liter Ehrlenmeyer flasks using 5 ml of an overnight starter culture.
Frozen cells (about 0.8 -1.0 g per 50 ml of culture) were thawed and resuspended in lysis buffer (20 mM Tris-Cl, pH 8.0, 100 mM NaCl, 10 mM 2-mercaptoethanol, 10% v/v glycerol; 3 ml per g wet cell weight) in 10-ml tubes followed by three freeze/thaw cycles using liquid nitrogen. Phenylmethylsulfonyl fluoride was added to 1.5 mM, DNase I to 690 units ml Ϫ1 , and the suspension mixed thoroughly by passage through an 18-gauge syringe needle 10 times before incubation on a mixing wheel for 20 min at room temperature. The suspension was transferred to 2-ml tubes and centrifuged for 20 min at full speed in a bench top microcentrifuge. The resulting supernatant and pellet contained, respectively, soluble and insoluble protein fractions.
Recombinant UFGT was purified from the soluble fraction by IMAC using the Talon system (CLONTECH, Palo Alto, CA), utilizing the 6-histidine tag placed at the N terminus of the recombinant protein.
Protein in the soluble fraction was bound to the affinity resin and eluted using a buffer containing 50 mM imidazole as described in the manufacturer's protocol. All buffers were modified to contain 10 mM 2-mercaptoethanol and 10% v/v glycerol. Using 0.5 ml of Talon resin, recombinant protein from approximately 50 ml of culture could be purified. Incorporation of 15 mM imidazole in the wash buffers resulted in a homogeneous preparation of the enzyme, albeit at the expense of yield.
Enzyme Assays-Glucosylation was routinely assayed in a modification of the buffer system reported by Do et al. (24). In a final assay volume of 200 l, the reaction conditions were 100 mM buffer (Tris-Cl, pH 8.0, or HEPES-OH, pH 8.0), 10 mM poly(ethylene glycol) 3400, 14 mM 2-mercaptoethanol, 2 mM dithiothreitol, 9 mM UDP-glucose, 100 M flavonol or anthocyanidin acceptor substrate and enzyme. The amount of enzyme used was varied according to the acceptor substrate under study. Except in the case of anthocyanidin substrates, assays were routinely started with the addition of the enzyme and were incubated at 30°C for up to 6 min (anthocyanidin substrates) and 10 min (flavonol substrates). Samples were taken at two time points to ensure linearity of all data points and hence to measure initial rates of activity. Flavonols and anthocyanidins were prepared fresh as required from stock solutions stored at Ϫ20°C. Unless stated otherwise, the solvent used for substrate dilution was 2-methoxyethanol (ethylene glycolmonomethyl ether (2-methoxyethanol)).
Reactions were stopped by acidification. For flavonol acceptor substrates, 50 l of glacial acetic acid was used, and for anthocyanidins, 150 l of 5% HCl was added to the reaction mixture.
The radioactive donor substrate UDP-[U 14 C]glucose was used for the determination of the rates of synthesis of glucosides for which extinction coefficient data were not available. UDP-[U 14 C]glucose (10.6 GBq mmol Ϫ1 , DuPont) was added to a final concentration of 2.5 M and supplemented with "cold" UDP-glucose to 100 M. For analysis by liquid scintillation counting or thin layer chromatography, flavonol glucosides were extracted into 500 l of ethyl acetate to remove unincorporated donor substrate. Liquid scintillation counting of ethyl acetate extracts (200 l, duplicates) was performed using Ultima Gold XR mixture (Canberra Packard) in a LS6000TD counter (Beckman). TLC analysis was performed upon ethyl acetate extracts following concentration at 70°C in a Centrivap Concentrator (Labconco) under reduced pressure. Residues were dissolved into 20 l of ethyl acetate. TLCs were run on Silica Gel 60 F254 plates (Merck) in a solvent comprising 35 parts ethyl acetate, 2 parts formic acid, 2 parts water (32). Developed plates were dried at 80°C for 1 h and exposed to storage phosphorimaging plates (Molecular Dynamics) prior to scanning on a Storm 860 PhosphorImager (Molecular Dynamics). Glucoside formation from anthocyanidin, phenol, and monoterpene acceptor substrates was assessed after separation using SepPak disposable reverse phase cartridges (Millipore) to remove the unincorporated UDP-[U 14 C]glucose prior to liquid scintillation counting or TLC assay. Briefly, acidified assay mixtures were loaded onto pre-charged SepPak cartridges, washed with 5 ml of MilliQ water, and eluted in 2-3 ml of ethanol (33). After lyophilization under reduced pressure, residues were dissolved into 20 l of ethyl acetate. For assays with anthocyanidin substrates, ethanol used for elution was acidified to 0.1% v/v HCl.
Kinetic Measurements-K m values of the recombinant glucosyltransferase were determined for the donor substrate UDP-glucose and the acceptors quercetin, cyanidin, malvidin, and delphinidin. Additionally, V max was determined for quercetin and cyanidin. For the measurement of the K m for UDP-glucose, 200 M quercetin was used as the acceptor substrate, and the concentration of UDP-glucose was varied from 0.25 to 9 mM. Acceptor substrate K m values were determined with 9 mM UDP-glucose as the donor substrate, and the concentration of the acceptor varied from 1 to 200 M as applicable. Assays were as described above. Additionally, for the determination of the K m for cyanidin, reducing agents were omitted from the reaction mixture, and cyanidin, rather than enzyme, was added to start the reaction. Experimental data from HPLC analyses were integrated using Gold Nouveau software and for cyanidin and quercetin glucosylation, converted to specific activities by reference to standard curves obtained using the authentic glucosides. Data were transformed and plotted as Lineweaver-Burk graphs to allow calculation of K m and V max values.
Metal Inhibition Studies-The effect of divalent metal ions upon the glucosylation of quercetin and cyanidin was studied using radiolabeled UDP-[U 14 C]glucose and liquid scintillation counting. Standard 200-l assays performed in the absence of added reducing agents were supplemented with CuCl 2 , MgCl 2 , CaCl 2 , MnCl 2 , and ZnCl 2 , to final concentrations of 0.01, 0.1, 1, and 10 mM as specified.
Product Inhibition Assays-The inhibition of glucosyltransferase activity was tested for quercetin and cyanidin glucosylation. In both cases the cognate and alternative glucoside were added individually to glucosylation assays at concentrations between 1 and 100 M. Product formation was determined by HPLC analysis.
Production of Antibodies, Western Transfer, and Immunoblotting-Rabbit antibodies were produced by the School of Biochemistry, La Trobe University, Melbourne. rUFGT (approximately 500 g) was mixed with an equal volume of Freund's complete adjuvant and injected intramuscularly. The rabbit was given two more boosts at 4-week intervals before blood samples were obtained. Serum was separated from red cells by centrifugation at 3000 ϫ g for 15 min at 4°C and stored in 0.02% sodium azide at 4°C. Total soluble protein extracts of leaves and berries of grapevine varieties as specified were prepared by a modification of the method of Hawker (34,35); full details of this protocol will appear elsewhere. 2 Following preparation, extracts were desalted on a Bio-Gel P6 column and concentrated in a stirred cell unit (Amicon) using a YM30 membrane. Proteins were resolved by SDS in 12% T, 2.63% C (w/v) Tris-Gly gels (36) and blotted onto nitrocellulose membranes (MSI Laboratories, Westbro, MA) using a semi-dry transfer unit (LKB, Broma, Sweden) as described by Harlow and Lane (37). A dilution of antibody of 1:10,000 was used for immunoblotting experiments. Immunoblots were probed with anti-rUFGT antibodies and horseradish peroxidase-coupled goat anti-rabbit antibodies (Promega) before detection using ECL TM reagents (Amersham Pharmacia Biotech) and exposure to Hyperfilm-MP TM film (Amersham Pharmacia Biotech).

RESULTS
Cloning and Identification of Full-length cDNAs Encoding UFGT from V. vinifera cv. Shiraz, Overexpression and Purification of Soluble Active Recombinant UFGT-By using the partial 532-base pair grape cDNA previously isolated by Sparvoli et al. (15), two clones homologous to putative plant UFGT sequences were isolated from a screen of approximately 150,000 plaque forming units of a V. vinifera cv. Shiraz postveraison (coloring) berry cDNA library. Clone GUT8 constitutes a full-length UFGT cDNA sequence but surprisingly contains a 77-base pair intron located at nucleotide 492, while the second clone, GUT3, was intronless and almost full length. Determination of the nucleotide sequence of both clones (Gen-Bank TM accession numbers AF000371 and AF000372 for GUT3 and GUT8, respectively) allowed us to predict the amino acid sequence for the corresponding V. vinifera UFGT (VVUFGT; Fig. 1). Comparison of the predicted V. vinifera UFGT amino acid sequence with that of other glucosyltransferases (Fig. 1) reveals overall positional identities ranging from 42% (V. vinifera UFGT versus Bronze-1 encoded UFGT) to 25% (V. vinifera UFGT versus solanidine glucosyltransferase). Positional identity is not surprisingly greater in the C-terminal third of the protein where the suspected UDP-glucose binding domain, common to all glucosyltransferases, is located (46% V. vinifera UFGT versus Bronze-1 encoded UFGT, 35% V. vinifera UFGT versus solanidine glucosyltransferase) (38). Quite clearly, sequence comparisons of the N-terminal parts of these various glucosyltransferases do not allow solid conclusions to be drawn regarding the enzymic properties of the encoded enzyme, particularly because there exists very little correlation between protein/gene sequences and biochemical characterization of the encoded products.
The recombinant UFGT (rUFGT) was therefore expressed as a soluble, active protein in E. coli using the pET14b 6-His fusion vector system. Proteins expressed from sequences cloned into pET14b contain an additional 20 NH 2 -terminal amino acid residues arising from the vector, including the 6-His "tag" used for metal-chelate affinity purification and a recognition site for the protease thrombin. Preliminary growth and induction trials at 30 and 25°C yielded protein that was expressed mainly as insoluble inclusion bodies (data not shown). Further reduction of the growth and induction temperature to 16°C resulted in the expression of significant amounts of a protein, the size expected for recombinant UFGT, in the soluble fraction. This is illustrated in Fig. 2, which shows an SDS-polyacrylamide gel analysis of total proteins extracted from both uninduced cells and those harvested after isopropyl-1-thio-␤-D-galactopyranoside induction. The detailed analysis of this expression protocol and the general applicability of low temperature growth of E. coli for production of "difficult" recombinant proteins will be described elsewhere. 2 When soluble protein extract prepared from cells containing putative rUFGT was assayed for enzymic activity, significant quercetin glucosylation was observed, suggesting that an ac-tive enzyme had been successfully expressed. Uninduced cell lysate and lysate prepared from E. coli BL21 pLysS cells that were not transformed with pET14-V. vinifera UFGT did not contain any glucosyltransferase activity against quercetin (data not shown).
Purification of rUFGT by IMAC from 5 g of cell paste proved straightforward, yielding large quantities of the recombinant enzyme. The inclusion of 15 mM imidazole in the wash buffer resulted in recovery of an essentially homogeneous preparation of rUFGT (Fig. 2, lane 5). Purified recombinant UFGT prepared in this way was stored at Ϫ80°C, where it was stable for many months. At least three freeze-thaw cycles could be tolerated without apparent loss of activity. Additionally, the recombinant enzyme was stable when stored for at least 120 h both at Ϫ20°C and 4°C and showed only a 30% decrease in quercetin glucosylation after the same period of storage at room temperature (data not shown).
Pitfalls in the Assay of Glucosyltransferase Activities-Assaying for enzyme activities with flavonol and anthocyanidin substrates in particular has inherent pitfalls that appear to be not always fully recognized, and this can in some cases lead to erroneous conclusions concerning the properties of the enzymes under consideration. For example, in our initial assays of the recombinant enzyme, it proved very difficult to achieve linearity between time points when assaying for glucosylation of cyanidin at pH values where activity could be observed (data not shown). The possibility that product inhibition was responsible for this phenomenon was effectively discounted by supplementing assays with the product. No effect upon the extent of cyanidin nor quercetin glucosylation was observed by the prior addition to the assay of either quercetin glucoside or cyanidin glucoside to a final concentration of 100 M. Likewise the use of very low enzyme concentrations failed to achieve linearity between assays typically performed for 5, 10, and 20 min (data not shown). Given the relatively stable nature of rUFGT described above, these observations indicate that loss of linearity is due either to substrate or product instability during the assay procedure.
In detailed chemical studies Brouillard (39) demonstrated that both very fast and slower interconversions of tautomeric anthocyanidin forms occur in aqueous solutions, particularly at pH values above 4. The red flavylium form AH ϩ occurs as the predominant species at pH values below approximately 2.5. If the pH is raised above 4, the blue quinoidal anhydrobase A, a kinetic product, forms instantaneously. This unstable intermediate rapidly decolorizes in aqueous solution (seconds) through the formation of a carbinol pseudobase B arising from nucleophilic attack by water at the 2-position of the C 15 flavonoid backbone. Re-acidification at any time during this process to a pH below 2 will fully restore the red flavylium form of the anthocyanidin. However, if the anthocyanidin solution is left for extended times (minutes) at pH 4 and above, it undergoes conversion to an open chain cis-chalcone and then the final thermodynamic product the trans-chalcone. At this point, recovery of the flavylium form cannot be achieved by simple re-acidification (39). The observation that very little of the red flavylium form of the anthocyanidin substrate cyanidin was reformed upon acidification to pH Ͻ1 after incubation under assay conditions suggested that irreversible non-enzymic conversion of substrate was occurring, and that this itself may be responsible for the lack of linearity seen in many assays.
To explore this possibility further, enzyme assays were performed following preincubation of substrate in reaction buffer for a variety of times. For cyanidin, incubation in assay buffer for 90 min with no added reducing agents resulted in a substrate unable to support the formation of glucoside by recombinant UFGT. In parallel, full absorbance spectra were monitored at regular intervals, and absorbance data were recorded continuously at 585 nm, the absorbance maximum for the quinoidal (blue) tautomer of cyanidin which forms instanta-neously at pH 8.0. Such prolonged incubation of the substrate led to loss of its characteristic spectrum (data not shown). This suggests a conversion of the unstable substrate into a form (a chalcone as suggested by Brouillard (39)) that is not glucosylated by the enzyme, a most likely contributor to the nonlinearity of the assay at extended incubation times. This was further underscored by spectral analysis of quercetin, which showed that the ease with which the assays for glucosylation of this substrate could be achieved over extended time intervals was paralleled by its stability in the assay buffer. To make valid comparisons between the efficiency of the two substrates with UFGT, we therefore concluded that the length of assays for cyanidin glucosylation should be restricted to 6 min or less where linearity was achieved.
The assays described in subsequent sections were mostly performed at pH 8, close to the pH optimum of the rUFGT, with the extent of glucosylation of anthocyanidin substrates determined by absorbance following HPLC separation of reaction products. To enable accurate determinations to be made of the amounts of anthocyanins synthesized, assays were stopped by the addition of HCl to a final pH of less than 1. Under these conditions, the colored flavylium form of both substrate and product is the predominant equilibrium product and is stable for extended periods (39). Duplicate HPLC determinations of individual assays analyzed at the beginning and end of batch HPLC runs (up to 48 h apart) gave essentially equal data sets, suggesting that no decomposition of anthocyanidin glucosides was occurring at this low pH (data not shown).
Determination of pH Optima and Substrate Specificity of the Recombinant Enzyme-Incubation of rUFGT with either cyanidin or quercetin aglycones in the presence of 9 mM UDPglucose donor substrate resulted in the formation of the respective 3-O-glucosides as determined by co-chromatography and diode array spectroscopy using authentic glucosides assayed by both TLC and HPLC analyses. Assays in the absence of the UDP-glucose donor substrate yielded no product (data not shown).
The pH optimum of the recombinant UFGT was analyzed for quercetin and cyanidin acceptor substrates and in both cases was observed to fall around pH 8.0 (data not shown). Quercetin glucosylation exhibited a marked buffer effect, with values in Tris buffer approximately 85% of those at the same pH in HEPES buffer. This effect was not seen with cyanidin glucosylation, but the instability of anthocyanidin substrates over much of the pH range studied inevitably impacted upon the final activity obtained.
The activity of the recombinant UFGT was tested against a wide range of donor and acceptor substrates. Of the donor substrates, only UDP-glucose was able to support the glucosylation of quercetin. No activity was observed with UDP-galactose, UDP-xylose, UDP-glucuronic acid, UDP-mannose, ADPglucose, CDP-glucose, GDP-glucose, or TDP-glucose (data not shown). This agrees with data obtained for non-homogeneous preparations of other flavonoid 3-O-glucosyltransferases showing exclusive acceptor-substrate specificity for anthocyanidins and flavonols (reviewed by Heller and Forkmann (40)).
The enzyme was specific for the glucosylation of flavonol and anthocyanidin substrates, with no activity observed for any other substrate tested. The range of substrates for which glucosides could be formed by enzymes in a total soluble protein extract prepared from V. vinifera variety Shiraz berries of equivalent developmental status to those used for the isolation of UFGT cDNAs was much greater than that catalyzed by rUFGT alone. Glucosylation of flavonols, a flavone, a flavanone, an isoflavone, and monoterpenes was detected when total berry proteins were used in assays for glucosyltransferase ac- tivity. 2 No glucosylation was seen with dihydroflavonol or flavan-3-ol substrates (data not shown).
Kinetic Parameters of the Recombinant Glucosyltransferase-The kinetics of the recombinant glucosyltransferase were studied both for donor and acceptor substrates (Fig. 3). By using quercetin (200 M) as the acceptor substrate, the dependence of activity upon the concentration of UDP-glucose was analyzed between 0.25 and 8 mM. A hyperbolic saturation curve was obtained, from which Lineweaver-Burk transformation gave a K m for UDP-glucose of approximately 1.9 mM and V max for quercetin glucoside formation of 18.9 nanokatals mg Ϫ1 . When the concentration of UDP-glucose was maintained constant at 9 mM and the concentration of quercetin varied between 1 and 40 M, a K m for quercetin of approximately 15 M was obtained. The V max estimated for quercetin glucoside formation under these conditions was also 18.9 nanokatals mg Ϫ1 .
When the concentration of cyanidin was varied between 1 and 80 M with UDP-glucose maintained at 9 mM, a K m for cyanidin of 30 M and a V max for cyanidin glucoside formation of 905 nanokatals mg Ϫ1 (48 times that observed for quercetin glucosylation) were obtained. All data were determined and averaged from three independent assays. However, due to the instabilities described above associated with the use of anthocyanidin substrates, greater variability was observed in the apparent K m data for this substrate than for either quercetin or UDP-glucose. By using the same reaction conditions, the K m for delphinidin was determined to be approximately 16 M and for malvidin approximately 35.7 M. V max data were not obtained for these acceptors. Additionally, thrombin cleavage of the 6-His N-terminal tag, removing 17 non-UFGT amino acids, had no effect on the ratio of cyanidin to quercetin glucosylation (data not shown).
Relative Activity Against Flavonol and Anthocyanidin Substrates-The relative activity of rUFGT with a range of flavonol and anthocyanidin substrates was tested. As shown (Fig. 4), cyanidin and quercetin gave the highest rates of anthocyanidin and flavonol glucosylation, respectively. Anthocyanidin glucosylation rates were compared by HPLC analyses, with the total absorbance for each glucoside measured at the absorption maximum determined from the chromatogram of each product. These products were identified by comparison with authentic glucoside standards (cyanidin, malvidin, and peonidin) and by the appearance, when anthocyanidin substrate was incubated with rUFGT, of a peak exhibiting an approximately 10 -20 nm lower absorbance maximum and a retention time shorter than the authentic anthocyanidin substrate (delphinidin and pelargonidin), as expected for a glucosylated product. Relative rates of flavonol glucoside formation were determined by liquid scintillation counting of ethyl acetate extracts from assays using radiolabeled UDP-[U 14 C]glucose donor substrate. This allowed the accurate determination of total glucoside formation, otherwise not possible for those flavonol glucosides for which molar extinction coefficients or authentic samples were not available.
Metal Inhibition Studies-Many reports of the properties of partially purified glucosyltransferases include the effect of added divalent metal ions (10, 11, 20, 21, 24, 32,  cause over 80% reduction in quercetin glucosylating activity. Although our preliminary experiments in the absence of added reducing agents suggested Cu 2ϩ inhibition of the recombinant glucosyltransferase itself, further investigation proved this to be incorrect. The addition of even 0.1 mM Cu 2ϩ to either quercetin or cyanidin glucosylation assays caused rapid and irreversible bleaching of the substrates, thus rendering the detection of product formation by HPLC and absorbance measurement impossible. Spectroscopic analyses of both quercetin and cyanidin before and after the addition of 1 mM Cu 2ϩ to the assay buffer showed the effect more clearly (Fig. 5). In each case, the colored component of the absorption spectrum was lost, leaving only the absorbance at around 280 nm arising from the phenolic ring. It seems likely that Cu 2ϩ is acting to catalytically convert the C 15 flavonoid into a compound not recognized by rUFGT. These data meant that it was not possible to attribute the observed lack of glucoside formation in the presence of Cu 2ϩ to specific inhibition of the glucosyltransferase itself.
Comparison of Glucosyltransferase Activities of Cloned V. vinifera UFGT with Total Activities in Grapevine Leaves and Berries-The specificity studies reported above clearly showed that the activity of rUFGT was likely to be restricted to glucosylation of flavonols and anthocyanidins, as originally suggested for the products of the Bronze-1 allele (41, 42). However, it does not follow that glucosylation of both these types of substrates in grapes is the consequence of rUFGT action in vivo, as has been assumed to be the case for certain lines of maize (43). Indeed, Dooner (44) and Styles and Ceska (45) reported accumulation of flavonol-3-glucosides in the endosperm, and coleoptiles, of bz-1/bz-1 lines of maize in which there was no detectable anthocyanin accumulation. In both cases the authors predicted the existence of a Bz-1-independent glucosyltransferase, which may be both tissue-specific (44) and with differing positional specificity to that encoded at the Bronze-1 locus (45). To investigate this in grapevine tissues, a comparison was made between the cyanidin-3-glucoside/quercetin-3-glucoside ratio made on the one hand by rUFGT and on the other hand by enzyme extracts from leaves and berries of the grapevine variety Muscat of Alexandria. The expression of the UFGT gene in berries of this white grape variety, in which anthocyanins are present only in trace amounts (46), cannot be detected by Northern blotting (19). The results of HPLC assays for glucoside formation from each acceptor substrate are given in Table I. This indicates very strongly that there must exist in leaves and berries additional enzyme activities capable of glucosylating quercetin at a greater rate than observed for rUFGT, since the ratio of cyanidin to quercetin glucosyltransferase activity is much smaller in both tissues (ratios of 1.1 and 2.4, respectively) than is seen for rUFGT (ratio of 48), an enzyme which may serve predominantly to glucosylate anthocyanidins at and after veraison in red berries. This became even clearer when extracts of berries from the red variety Shiraz (also the source of the rUFGT cDNA) and the white variety Muscat of Alexandria were compared for their ability to glucosylate anthocyanidins and flavonols (Table I). We found that while the total units and specific activity of quercetin glucosylating activity were practically identical in extracts made from these grapes, the extract from the red Shiraz berries had a specific activity for cyanidin glucosylation which was almost 25-fold higher than that seen in extracts from berries of Muscat of Alexandria. This strongly suggests a very minor role, if any, for UFGT in the production of flavonol glucosides in berries. That this indeed appears to be the case was further demonstrated by Western blotting of the extracts on which the enzymic assays had been performed (Fig. 6). Only red grapes are seen to contain significant quantities of immunoreactive UFGT, the abundance of which correlates with UDP-glucose:cyanidin 3-O-glucosyltransferase activity and not at all with UDP-glucose: quercetin 3-O-glucosyltransferase activity (cf. lanes 1-4 in Fig.  6). The lack of immunoreactive material in the extracts from the white variety cannot be explained by changes in protein sequence of UFGT within varieties of the same species. Our previous work on other grape proteins indicates a very strong retention of immunological reactivity across V. vinifera varieties (47). Indeed, given the different electrophoretic mobilities of immunoreactive material from the two varieties, it is possible that UFGT is practically absent from the white fruit and that this constitutes the real reason anthocyanins do not accumulate in these white varieties since Northern analyses previously have shown most other genes in the anthocyanin biosynthesis pathway to be expressed (19). DISCUSSION Despite the great importance of glucosylation to many of the metabolic processes underlying plant growth, development, and response to environmental stress (2) glucosyltransferases, the final state of purity achieved has been some way from homogeneity. This presents many problems for the assignment of specificities and is further complicated by the structural similarity of many distinct substrates, which may permit cross-reactivity under experimental conditions.
An alternative approach for the study of glucosyltransferases, capable both of yielding large amounts of protein and offering a simple way to achieve homogeneity, is to clone the cDNAs of the desired enzyme into a suitable expression system. At least four reports of the production of recombinant plant glucosyltransferases are known. However, in one of these (48), the enzyme was inactive, and in the other three no effort was made to purify the expressed enzyme from the background of host proteins (23,49,50).
We have now cloned cDNAs representing a V. vinifera glucosyltransferase. Initial attempts at obtaining an active recombinant protein from E. coli cells grown at 30°C failed due to the exclusive accumulation of the rUFGT in insoluble aggregates. However, following expression at 16°C, the encoded protein was shown to be soluble and after purification was shown to harbor UDP-glucose:flavonoid 3-O-glucosyltransferase activity when assayed in vitro. Importantly, this conclusion could not have been safely arrived at from amino acid sequence similarities alone as these never exceeded 50% positional identity. The availability of purified preparations of rUFGT permitted calculation of the protein concentration used in kinetic assays and thereby, for the first time, the determination of k cat values for flavonoid glucoside formation. The k cat for cyanidin glucoside formation was calculated to be 47.63 s Ϫ1 , while for quercetin   6. The presence of UFGT in grapes correlates strongly with cyanidin glucosylating activity but not at all with quercetin glucosylating activity. Total soluble protein extracts from berries of the red grape variety Shiraz, and from berries and leaves of the white grape variety Muscat Of Alexandria, were separated by SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membrane followed by immunoblot analysis using rabbit antibodies raised against rUFGT. glucoside formation the k cat was 0.99 s Ϫ1 . These data showed the relative catalytic turnover of the two substrates and indicated that cyanidin, everything else being equal, is glucosylated 48 times faster than quercetin. Therefore, it is possible that the inability to observe flavonol glucosylation reported for the cyanidin-3-O-glucosyltransferase from V. vinifera by Do et al. (24) for a partially purified enzyme preparation was due to insufficient enzyme in the assay system. In the other instance of the expression of an active UDP-glucose-flavonoid glucosyltransferase, from G. triflora (23), preliminary analysis revealed the relative specific activities for cyanidin and quercetin glucoside formation were 583.33 and 61.66 nanokatals per mg of protein per min, respectively. These data were obtained from crude extracts of bacteria expressing the G. triflora UFGT cDNA, and it should be noted that unlike the grapevine UFGT expressed here, delphinidin was glucosylated more readily than cyanidin. Direct comparison of the results obtained using V. vinifera UFGT with those of Tanaka et al. (23) is difficult, since no detailed kinetic parameters could be obtained from assays with the crude bacterial lysate employed in their study.
Part of the cDNA from which the V. vinifera rUFGT was expressed was used previously to determine the developmental pattern of gene expression for anthocyanin biosynthetic genes in berries of the red grape variety Shiraz (17) and to determine UFGT expression in white grapes that spontaneously have acquired the ability to accumulate anthocyanins (18). In the former case it was concluded that the appearance of anthocyanins at veraison (the beginning of berry softening and coloring) correlated with the detectability of UFGT mRNA, whereas transcripts of anthocyanidin biosynthetic genes in general were detectable throughout berry development. Similarly, in "sports" of white grape varieties showing accumulation of anthocyanins, there were detectable levels of UFGT transcripts, whereas a white sport of the red grape variety Cabernet Sauvignon contained no detectable UFGT transcript. Thus, unlike the situation found in maize and snapdragon where control is exerted at earlier points in the biosynthetic pathway (reviewed in Ref. 51), it appears that in grapes UFGT may play a controlling role in anthocyanin synthesis. In contrast to the situation found with anthocyanins, the presence of UFGT mRNA appears not to correlate with the accumulation of flavonol glucosides in grape berries. Price et al. (52) reported the accumulation of quercetin glycosides (glucoside, glucuronide, and other unidentified glycosides), in sun-exposed berries of the variety Pinot Noir. Additionally, we have observed that in Shiraz, quercetin glycoside accumulation occurs prior to color development. 3 This raises the possibility that the UFGT-encoded protein may not at all or may only partly be responsible for the glucosylation of flavonols in planta, despite its ability to do so in vitro. A comparison of the rates of cyanidin and quercetin glucosylating activities of recombinant UFGT with those obtained from total protein extracts of leaves and berries of Muscat of Alexandria supported this claim. While the recombinant glucosyltransferase formed cyanidin glucoside 48 times faster than quercetin glucoside, the relative rates for berries and leaves were 1.1 and 2.4, respectively. This suggested the presence in leaves and berries of additional glucosyltransferase activities with specificity toward flavonol substrates. It would not have been surprising if such enzymes would be encoded by genes very similar to UFGT, and yet Sparvoli et al. (15) and our observations 4 suggest that the V. vinifera genome contains only one gene encoding UFGT or closely related homologues thereof. Nevertheless, we have in other experiments now confirmed that extracts from Muscat of Alexandria leaves contain at least six separable quercetin glucosylating activities, all of which exhibit lower activity than rUFGT toward cyanidin. 2 To explore the nature of such activities, we produced polyclonal antibodies that subsequently were used in Western blots (Fig.  6) to test the hypothesis that in grape UFGT probably plays a very minor role for flavonol glucosylation in vivo. The results shown in Fig. 6 and Table I together clearly demonstrate that the presence of UFGT in berries of grapes correlates very strongly with the quantitative ability to glucosylate anthocyanidins but not at all with the ability to glucosylate quercetin. We therefore suggest that the UFGT investigated here, from a physiological point of view, more appropriately be considered a UDP-glucose:anthocyanidin 3-O-glucosyltransferase. It appears that these considerations are applicable when assigning roles to homologues of UFGT in other systems where the role of UFGT for production of flavonol glucosides might vary considerably depending on the genetic background, as seen for different lines of maize bz-1/bz-1 mutant lines (44,45).
The availability of protocols for production of active recombinant glucosyltransferases at low growth temperature, combined with good antibodies and carefully executed enzyme assays, should now facilitate a more rigorous assessment of the properties and roles of "UFGT" homologues, a much studied class of enzymes for which the genes and cDNAs have been obtained from a multitude of species across the plant kingdom.