Reaction mechanism from leucoanthocyanidin to anthocyanidin 3-glucoside, a key reaction for coloring in anthocyanin biosynthesis.

In the conversion from colorless leucoanthocyanidin to colored anthocyanidin 3-glucoside, at least two enzymes, anthocyanidin synthase (ANS) and UDP-glucose:flavonoid 3-O-glucosyltransferase (3-GT), are postulated to be involved. Despite the importance of this reaction sequence for coloring in anthocyanin biosynthesis, the biochemical reaction mechanism has not been clarified, and the possible involvement of a dehydratase has not been excluded. Here we show that recombinant ANSs from several model plant species, snapdragon, petunia, torenia, and maize, catalyze the formation of anthocyanidin in vitro through a 2-oxoglutarate-dependent oxidation of leucoanthocyanidin. Crude extracts of Escherichia coli, expressing recombinant ANSs from these plant species, and purified recombinant enzymes of petunia and maize catalyzed the formation of anthocyanidin in the presence of ferrous ion, 2-oxoglutarate, and ascorbate. The in vitro formation of colored cyanidin 3-glucoside from leucocyanidin, via a cyanidin intermediate, was demonstrated using petunia ANS and 3-GT. The entire reaction sequence did not require any additional dehydratase but was dependent on moderate acidic pH conditions following the enzymatic steps. The present study indicated that the in vivo cytosolic reaction sequence involves an ANS-catalyzed 2-oxoglutarate-dependent conversion of leucoanthocyanidin (flavan-3,4-cis-diol) to 3-flaven-2,3-diol (pseudobase), most probably through 2,3-desaturation and isomerization, followed by glucosylation at the C-3 position by 3-GT.

The reaction leading from colorless leucoanthocyanidin to anthocyanidin and its 3-O-glucoside is the critical step in the formation of colored metabolites in anthocyanin biosynthesis (1-3) (Fig. 1). Although anthocyanidin is the first colored metabolite in the biosynthetic pathway, it is hardly detected in plant tissues because of its instability. Anthocyanidin 3-glucoside is the first stable colored metabolite detected in plants. The entire reaction formally involves dehydrogenation at C-2, dehydration at C-3 and C-4, and glucosylation at the hydroxyl group of C-3.
Anthocyanidin synthase (ANS) 1 is presumed to catalyze the first half of the reaction involving dehydrogenation at C-2 and dehydration at C-3 and C-4. cDNAs and genes that presumably encode ANS have been isolated from several plant species (4 -12). The involvement of ANS in this step is supported by the ability of a cDNA encoding maize ANS to restore anthocyanin formation in an ANS Ϫ mutant of maize, through transient expression by particle bombardment (4). Sequence comparison with deduced amino acid sequences of these ANSs suggested that they belong to a family of soluble oxygenases depending upon 2-oxoglutarate (13,14).
The precise biochemical reaction mechanism from leucoanthocyanidin to anthocyanidin has not been clarified to date. As shown in Fig. 2, three possible mechanisms have been postulated for the reaction sequence from leucoanthocyanidin to 3-flaven-2,3-diol (pseudobase), which is chemically converted into anthocyanidin (flavylium ion) by removal of the C-2 hydroxyl anion. In the first mechanism (2) (Fig. 2, pathway (A)), ANS catalyzes a desaturation at the C-2 and C-3 positions of (2R,3S,4S)-leucoanthocyanidin (15,16) in the presence of 2-oxoglutarate and molecular oxygen, yielding 2-flaven-3,4diol, with concomitant release of CO 2 , succinate, and H 2 O. After isomerization of the 2,3-double bond to the 3,4-position of 2-flaven-3,4-diol, by removal of the hydroxyl group at C-4, double bond migration, and insertion of a hydroxyl group at C-2, 3-flaven-2,3-diol is formed. Alternatively, ANS may catalyze a hydroxylation at the C-2 position of leucoanthocyanidin, followed by 3,4-dehydration (1) (Fig. 2, pathway (B)). In the third postulation, 3,4-dehydration occurs as the first step, followed by a hydroxylation at C-2 (1) (Fig. 2, pathway (C)). The last two assumptions (pathways B and C) imply that the formation of 3-flaven-2,3-diol may require not only ANS but also a specific dehydratase catalyzing 3,4-dehydration. Alternatively, ANS may catalyze the dehydration of the reaction. So far there is no evidence to suggest the involvement of an external dehydratase in the whole reaction sequence, except for a preliminary report describing a genetic locus that possibly encodes a dehydratase in snapdragon (Antirrhinum majus) (17).
To date, in vitro biochemical analysis regarding this crucial reaction for anthocyanin biosynthesis has only been carried out with recombinant ANS from Perilla frutescens (10). Recombinant perilla ANS was shown to catalyze the formation of anthocyanidin from leucoanthocyanidin according to a 2-oxoglutarate-dependent manner when followed by an acidification with HCl. It is necessary to verify this important reaction mechanism with ANSs from various model plant species for anthocyanin biosynthesis under in vivo conditions, where ANS is combined with a UDP-glucose:flavonoid 3-O-glucosyltransferase (3-GT) at cytosolic and vacuolar pH.
The second half of the reaction (Fig. 1), the formation of anthocyanidin 3-glucoside from anthocyanidin, has also not been completely clarified. From intertissue complementation assays in maize (18) it has been generally assumed that leucoanthocyanidin first is converted to 3-flaven-2,3-diol (pseudobase) by ANS and is then glucosylated by 3-GT (1). No in vitro biochemical experimental evidence has been provided to confirm the reaction sequence from leucoanthocyanidin to anthocyanidin 3-glucoside, although 3-GT activity has been detected in crude enzyme preparations of various plant species (19 -22) and in preparations containing recombinant 3-GT (23,24). The reaction sequence catalyzed by ANS and 3-GT is postulated to occur in the cytosol, followed by transport into vacuoles, where the colored anthocyanin (flavylium ion) is formed as a consequence of the acidic pH (25). However, no in vitro experimental evidence for this hypothesis has been provided yet. Therefore, it is also necessary to examine whether anthocyanidin 3-glucoside can be formed from leucoanthocyanidin using both ANS and 3-GT recombinant proteins under physiological conditions, where enzyme reactions take place under cytosolic conditions (pH 7) followed by a shift to vacuolar conditions (pH 5) (26).
In the present study, we expressed and functionally verified, for the first time, four recombinant ANSs from snapdragon (A. majus), petunia (Petunia hybrida), torenia (Torenia fournieri), and maize (Zea mays). The reaction catalyzed by these ANSs was shown to proceed via a 2-oxoglutarate-dependent oxygenation of leucoanthocyanidin in the formation of anthocyanidin. Evidence, which implicates that ANS catalyzes the reaction from leucoanthocyanidin to anthocyanidin in the absence of an external dehydratase or strong acid conditions, is provided. We show that the combination of recombinant ANS and 3-GT was sufficient and essential to yield cyanidin 3-glucoside from leucocyanidin under physiological conditions, by mimicking cytosolic enzyme reactions and transport into vacuoles.

EXPERIMENTAL PROCEDURES
cDNA Clones from Various Model Plant Species-The cDNA clones Candi and A2, encoding ANS of A. majus (5) and Z. mays (4), were kind gifts from Dr. Cathie Martin and Dr. Alfons Gierl, respectively. The cDNA clone TAN1, encoding ANS from T. fournieri cv. Summer Wave Blue (Suntory Ltd.) was isolated as described previously (12). The cDNA encoding petunia ANS was isolated by PCR screening of a cDNA library constructed from P. hybrida cv. Surfinia (Suntory Ltd.) using the forward primer 5Ј-CCG GAA TTC ATG GTG AAT GCA GTA GTT ACA AC-3Ј and the reverse primer 5Ј-GCT CTA GAA CTC GTG ATT CCA ACA CAT CAT C-3Ј. Amplification was carried out for 25 cycles at the following conditions: 94°C/0.5 min, 55°C/1.5 min, 72°C/2.8 min.
The nucleotide sequence was verified by sequence analysis using the dideoxy chain termination method with Thermo Sequenase (Amersham Pharmacia Biotech) using a DNA sequencer (model DSQ-2000L; Shimadzu, Kyoto, Japan) and was shown to be identical with the published sequence (6). The cDNA clone PGT8, encoding petunia 3-GT (GenBank TM accession number AB027454), was isolated from P. hybrida cv. Surfinia. 2 Heterologous Overexpression in Escherichia coli-An EcoRI site and 2 Yamazaki et al., submitted for publication.
XbaI site was created at the 5Ј-and 3Ј-terminal ends, respectively, of the coding region of ANS by PCR amplification with each primer set as follows: for snapdragon ANS, 5Ј-CCG GAA TTC ATG GCA CCG GCA ATA GCT CCA CCC TCA CGG G-3Ј and 5Ј-GCT CTA GAT TAC TCA  ACA ACC TTC TCA TCT TGT TTC CGA T-3Ј; for torenia ANS, 5Ј-CCG  GAA TTC ATG GTT TCT CCA GCA TCT CCG AGC CCG GCC C-3Ј and  5Ј-GCT CTA GAT CAC TCA ACA CTC TTA TCA TCA TGC TCA  ACA-3Ј; for maize ANS, 5Ј-CCG GAA TTC ATG GAG TCG TCG CCG  CTG CTG-3Ј and 5Ј-GCT CTA GAT CAG TTG GTC TGC GGC GGC  GGC-3Ј. For petunia ANS, an EcoRI site and XbaI site were created by PCR screening using the primer set as above. The amplified fragments were ligated into the EcoRI-XbaI site of pMAL-c2 (New England Biolabs, Beverly, MA) to afford pMSDA1 (snapdragon), pMPTA9 (petunia), pMTOA6 (torenia), and pMMZA3 (maize), respectively. An XbaI site and PstI site were created at the 5Ј-and 3Ј-terminal ends of the coding region of petunia 3-GT by PCR amplification with the primer set 5Ј-GCT CTA GAA TGA CTA CTT CTC AAC TTC ACA-3Ј and 5Ј-AAC TGC AGT CAA GTA AGC TTG TGA CAT TTA-3Ј. The amplified fragment was ligated into the XbaI-PstI site of pMAL-c2 to afford p3GTM7. The ANSs and 3-GT were expressed as fusion proteins with maltose-binding protein (MBP, 42.7 kDa) at the N terminus. Each vector was introduced into E. coli Nova Blue for overexpression.
For purification, the MBP-tagged ANS proteins from petunia and maize were then subjected to amylose resin column chromatography (New England Biolabs) in Buffer B (20 mM potassium P i , pH 7.0, 200 mM NaCl, 10 mM maltose, 5 mM dithiothreitol, 10% glycerol) according to the manufacturer's protocol. All purification procedures were performed at 4°C.

SDS-Polyacrylamide Gel Electrophoresis (PAGE) and
Western Blotting-SDS-PAGE was carried out using 8% polyacrylamide gels as described previously (27). Western blotting and immunostaining were performed on an Immobilon P membrane (Millipore) using phosphatase-labeled goat anti-rabbit IgG (Kirkegaard & Perry Laboratories) and 5-bromo-4-chloro-3-indolylphosphate p-toluidine/nitro blue tetrazolium chloride (Life Technologies, Inc.) as a substrate. The rabbit primary antibodies were prepared against the His-tagged recombinant ANS from P. frutescens (10) and applied at a 1:200 dilution.
Assay of ANS Enzymatic Activity-(2R,3S,4R)-leucocyanidin and (2R,3S,4R)-leucopelargonidin were purchased from Plant Chemistry Industrial Research (Lower Hutt, New Zealand). These are stereoisomers of natural leucoanthocyanidins having 2R,3S,4S stereochemistry and are easily converted to natural isomers in the reaction mixture through a mechanism via 4-carbocation formed by removal of the hydroxyl anion from the C-4 position (28,29). The reaction mixture (100 l), containing 20 mM potassium P i (pH 7.0), 200 mM NaCl, 5 mM dithiothreitol, 10% glycerol, 4 mM sodium ascorbate, 0.4 mM FeSO 4 , 1 mM 2-oxoglutaric acid, 1 mM leucoanthocyanidin, and 300 -500 g of ANS protein, was incubated at 30°C for an appropriate period. The reaction was terminated by addition of 1 l of concentrated HCl and/or by extraction with 100 l of isoamyl alcohol, and anthocyanidins were analyzed by high performance liquid chromatography (HPLC). HPLC was carried out using a YMC-ODS-A312 column (diameter, 6 mm ϫ 150 mm) with CH 3 OH-CH 3 COOH-H 2 O (20:15:65) as an eluent at a flow rate of 1.0 ml/min at 40°C. The quantity of anthocyanidin was determined by monitoring the absorbance at 520 nm. The quantitative calibration curves were obtained with standard anthocyanidins. Protein was determined by the method of Bradford (30) using a protein determination kit (Bio-Rad).
Combined Reaction of ANS and 3-GT-The standard reaction mixture (200 l), containing 20 mM potassium P i (pH 7.0), 200 mM NaCl, 5 mM dithiothreitol, 10% glycerol, 4 mM sodium ascorbate, 0.4 mM FeSO 4 , 1 mM 2-oxoglutaric acid, 1 mM leucoanthocyanidin, 1 mM UDP-glucose, 300 -500 g of ANS and 3-GT proteins each, was incubated at 30°C for 30 min for the coexistence reaction. The reaction was terminated by addition of 2 l of concentrated HCl. The products were extracted with 200 l of isoamyl alcohol and subjected to HPLC analysis as described above. For sequential reactions, the enzyme protein, either ANS or 3-GT, was first added and incubated for 30 min; then the first enzyme protein was removed by ultrafiltration using Ultrafree-MC (Millipore); and then the second enzyme protein was added and incubated for 30 min. After addition of concentrated HCl and extraction with isoamyl alcohol, the products were analyzed by HPLC.
Cyanidin 3-Glucoside Formation by a Shift of pH Conditions-The reaction containing both ANS and 3-GT was carried out in the mixture (600 l) at pH 7 as described above (the ratio of ANS to 3-GT protein was about 1:9) for 60 min, followed by removal of proteins by ultrafiltration. The product composition was confirmed to be ϳ90% cyanidin 3-glucoside and ϳ10% cyanidin by HPLC analysis. The filtrate was then divided into three portions (200 l each) and subjected to changed pH conditions: pH 2 by adding 2 l of concentrated HCl; pH 5 by adding 1 l of 20% H 3 PO 4 ; pH 7 by no addition. The cyanidin 3-glucoside was determined from ultraviolet and visible absorption spectra obtained by spectrometry using a spectrophotometer U-2001 (Hitachi).

Heterologous Expression of ANS and 3-GT in E. coli-
The open reading frames of cDNAs encoding ANS from snapdragon, torenia, and maize and the cDNA encoding 3-GT from petunia were amplified by PCR and subcloned into pMAL-c2 expression vectors under the control of the tac promoter. The constructs carrying ANS-encoding cDNA from snapdragon, torenia, and maize and 3-GT-encoding cDNA from petunia were named pMSDA1, pMTOA6, pMMZA3, and p3GTM7, respectively. The construct, pMPTA9, was generated by subcloning the coding region of the petunia ANS-encoding cDNA, including the 78base pair nucleotide sequence of the 3Ј-flanking region, into pMAL-c2. The vectors were used to transform E. coli Nova Blue. The empty vector, pMAL-c2, was applied as a negative control, and pLMK1, carrying the ANS-encoding cDNA from perilla (10) Blue to confirm expression (Fig. 3B, lanes 2-6 and 9). Western blot analysis indicated that each recombinant ANS preparation contained antigens of the expected molecular mass when an antibody raised against recombinant perilla ANS (10) was employed (Fig. 3A). This cross-reactivity can be attributed to the relatively high identities (47.6 -72.8%) shared between the deduced amino acid sequences of the four ANS-encoding cDNAs and the perilla ANS-encoding cDNA. Recombinant ANSs from petunia and maize were purified by affinity chromatography using an amylose resin column (Fig. 3, lanes 7 and 8). The expression of recombinant petunia 3-GT was shown by the presence of an extra band of the expected molecular mass (92 kDa) (Fig. 3, lane 9).
Detection of Enzymatic Activity of ANSs-Crude protein extracts of E. coli transformed with pMSDA1, pMPTA9, pM-TOA6, and pMMZA3 were assayed for ANS activity in the presence of cofactors and leucocyanidin as a substrate. The identity of cyanidin formed by the in vitro reaction was confirmed by co-chromatography of the reaction product with a standard using HPLC. Representative results using petunia ANS, indicating the enzymatic formation of cyanidin from leucocyanidin, are shown in Fig. 4. In the standard procedure of the assay, HCl was added to the reaction mixture to terminate the reaction prior to isoamyl alcohol extraction and subsequent HPLC analysis. However, even without addition of HCl, the same amount of cyanidin was formed. These results indicate that HCl acidification is not absolutely necessary, but isoamyl alcohol extraction and the acidic conditions found during HPLC analysis (pH 2.0 -2.2) were sufficient for the formation of anthocyanidin. Furthermore, petunia ANS could also catalyze the formation of pelargonidin from leucopelargonidin (Fig. 4F), despite an absence of pelargonidin-type pigments in petunia.
Cyanidin was formed by extracts of E. coli expressing recombinant ANS from all of the different plant species (Fig. 5), although cyanidin formation did not show linearity with time after 10 min of incubation, most likely because synthesized cyanidin decomposed due to an inherent instability or because ANSs were inactivated during the incubation.
Purified recombinant MBP-ANSs also exhibited ANS activity and were able to catalyze the formation of cyanidin (Table  I). The specific activities for the formation of cyanidin were 0.972 and 0.215 microkatal/kg of protein for petunia MBP-ANS and maize MBP-ANS, respectively, obtained by initial velocities of 5 min of incubation. These results indicated that, for the conversion of leucoanthocyanidin to anthocyanidin, only ANS is required, and no additional dehydratase is needed.
Enzyme Properties of ANS as a 2-Oxoglutarate-dependent Oxygenase-The cofactor requirements of the purified recombinant ANSs from petunia and maize were investigated, and the results are summarized in Table I. The formation of cyanidin was absolutely dependent on the presence of ferrous ion, ascorbate, and 2-oxoglutarate. When O 2 was substituted with N 2 , cyanidin production was not observed. These results clearly indicate that ANS belongs to a family of 2-oxoglutarate-dependent oxygenases requiring O 2 and ferrous ion. In addition, no dehydrated compound was produced upon incubation of leucocyanidin with purified petunia ANS in the absence of cofactors (data not shown), suggesting that ANS itself has no dehydration activity toward leucoanthocyanidin.
The Entire Reaction Sequence Catalyzed by ANS and 3-GT-To clarify the entire reaction sequence between leucoanthocyanidin and anthocyanidin 3-glucoside, we examined combinatory and sequential reactions involving both recombinant ANS and 3-GT in vitro. Fig. 6A shows the reaction products obtained from leucocyanidin by the combined presence of ANS and 3-GT together with the appropriate cofactors for both enzymes (2-oxoglutaric acid, sodium ascorbate, FeSO 4 , and UDP-glucose). Both cyanidin and cyanidin 3-glucoside were formed. Upon the sequential reaction of ANS followed by 3-GT, where the intermediates were only exposed to one enzyme at a time, the main product was cyanidin 3-glucoside, despite the presence of cyanidin (Fig. 6B). However, upon the opposite sequential reaction, i.e. 3-GT followed by ANS, only cyanidin was formed by the action of ANS; no cyanidin 3-glucoside was produced (Fig. 6C). Furthermore, no glucosylated compound could be detected when only 3-GT was incubated with leucocyanidin (data not shown), indicating that petunia 3-GT does not accept leucoanthocyanidin as a substrate. These results indicate that the reaction sequence from leucoanthocyanidin to anthocyanidin 3-glucoside involves anthocyanidin formation from leucoanthocyanidin as the first step, followed by 3-glu-cosylation of anthocyanidin as the second step. More importantly, only two enzymes, ANS and 3-GT, are necessary and sufficient for the catalysis of the sequential reactions leading from leucoanthocyanidin to anthocyanidin 3-glucoside. Therefore, no additional enzyme(s) such as dehydratase are required to complete the entire reaction sequence.
Mimicking the in Vivo Sequential Reactions in Vitro-The in vivo conditions involved in the transport of anthocyanidin 3-glucoside into vacuoles were mimicked in vitro, whereby the sequential reactions of ANS and 3-GT yielding cyanidin 3-glucoside as ϳ90% of the products were followed by a shift in pH. After the incubation with both enzymes at pH 7 (cytosolic pH), the enzymes were removed by ultrafiltration, and a part of the reaction mixture was acidified to pH 2 or 5 (vacuolar pH) or maintained at pH 7 (cytosolic pH). The ultraviolet and visible spectra of all of the reaction products under these pH conditions were identical with those of standard cyanidin 3-glucoside (Fig. 7). These results indicate that the initial reaction product of ANS and 3-GT under neutral cytosolic conditions most likely is the pseudobase of anthocyanidin 3-glucoside and that the flavylium ion of anthocyanidin 3-glucoside is formed after the pseudobase has been transported into the acidic environment of vacuoles (pH 5).

DISCUSSION
Most reactions involved in anthocyanin biosynthesis have been biochemically characterized with purified enzyme preparations and recombinant proteins (31,32). However, the reactions from leucoanthocyanidin to anthocyanidin 3-glucoside ( Fig. 1) have not been well characterized. In particular, the reaction catalyzed by ANS has not been well documented, because the detection of ANS activity in vitro has not been possible with crude enzyme preparations or recombinant proteins. Our recent study on P. frutescens ANS was the only evidence provided for the formation of anthocyanidin by ANS (10), although this study could not completely exclude the possible involvement of an external dehydratase in a variety of model plant species. The present study indicated that all tested putative ANSs (not only perilla ANS) were capable of catalyzing the 2-oxoglutarate-dependent reaction that converts leucoanthocyanidin to an oxidized metabolite, in the absence of additional enzyme(s) such as a dehydratase, which can be readily converted to anthocyanidin by acidification. Furthermore, the inclusion of an in vitro 3-GT reaction and the effect of  The product composition was confirmed to be ϳ90% cyanidin 3-glucoside and ϳ10% cyanidin by HPLC analysis. After removal of ANS and 3-GT by ultrafiltration, the pH of the incubation mixture was changed from 7 to 2 or 5. The reaction and determination of absorption spectra were carried out as described under "Experimental Procedures." ABS, absorbance.
a pH shift that mimics transport of the glucoside into vacuoles have provided evidence for a plausible in vivo reaction sequence (as shown in Fig. 8). Namely, the 3-GT reaction takes place after the ANS reaction, and the actual substrate of the 3-GT is therefore most likely the pseudobase form of anthocyanidin present at cytosolic pH (pH 7), not the leucoanthocyanidin. The pseudobase form of anthocyanidin 3-glucoside is then transported into vacuoles, where the colored flavylium ion of anthocyanidin 3-glucoside is formed as a result of the acidic pH conditions of vacuoles.
In the flavonoid biosynthetic pathway, there are four reactions that are catalyzed by 2-oxoglutarate-dependent oxygenases: ANS (10, present study), flavanone 3-hydroxylase (33,34), flavonol synthase (35), and flavone synthase I (36). All these reactions are concerned with oxidation at the C-2 and/or C-3 position of the flavonoid skeleton. In the reaction of flavone synthase I, the direct double bond formation at C-2 and C-3 of flavanone was proposed (36). Because no stable hydroxylated intermediates were detected in the reactions of ANS, flavonol synthase, or flavone synthase I, the mechanism involving direct 2,3-desaturation seems to be the most likely route for the formation of 2-flaven-3,4-diol (Fig. 8). Regarding the reaction mechanism of flavone synthase I, a radical mechanism was proposed for the direct 2,3-dehydrogenation of flavanone (36). In the proposed pathway A (Fig. 2) for the formation of the key compound, 2-flaven-3,4-diol, three mechanisms are possible: (i) direct dehydrogenation from C-2 and C-3, probably by a radical mechanism; (ii) hydroxylation at C-2 of leucoanthocyanidin to give flavan-2,3,4-triol and subsequent 2,3-dehydration to 2-flaven-3,4-diol (a modified pathway B in Fig. 2); (iii) hydroxylation at C-3 to give flavan-3,3,4-triol, spontaneous dehydration into 3-oxo-flavan-4-ol, and subsequent keto-enol isomerization to 2-flaven-3,4-diol. However, the route via flavan-2,3,4-triol (ii) may require a dehydratase, either leading to 2-flaven-3,4-diol or to 3-flaven-2,3-diol. If this dehydration indeed takes place, it is now evident that no external dehydratase is required, although we cannot completely exclude the possibility that ANS itself may catalyze the dehydration of flavan-2,3,4-triol (pathway B in Fig. 2, for example). Because the present study provided evidence that ANS does not catalyze the dehydration of leucoanthocyanidin (pathway C in Fig. 2), it is unlikely that ANS is capable of catalyzing the dehydration of flavan-2,3,4triol, because it has a similar structure to that of leucoanthocyanidin. In addition, currently there is no report that describes an additional dehydratase activity by 2-oxoglutaratedependent oxygenases, although ANS may act to promote the dehydration as a general acid-base catalyst due to the common nature of proteins. The mechanism involving C-3 hydroxylation (iii) may be more plausible in analogy to the reaction catalyzed by flavanone 3-hydroxylase. To elucidate the detailed mechanism of 3-flaven-2,3-diol formation, further investigation may be required. Fig. 8 illustrates the most probable reaction mechanism from leucoanthocyanidin (flavan-3,4-diol) to the colored anthocyanidin 3-glucoside (flavylium ion) obtained in this study. The reaction catalyzed by 2-oxoglutarate-dependent oxygenases has been postulated to proceed in two steps (37)(38)(39)(40). In the first step, ANS binds with ferrous ion, which acts as a catalytic center of the reaction, and composes a complex with molecular oxygen and 2-oxoglutarate, followed by the formation of an oxoferryl enzyme complex, succinate, and CO 2 . This first step is common to all 2-oxoglutarate-dependent oxygenases. In the second step, the oxoferryl species is used for the hydrogen radical abstraction from C-2 and C-3 or hydroxylation at C-2 or C-3, followed by spontaneous or, less likely, ANS-catalyzed dehydration that yields 2-flaven-3,4-diol and H 2 O. Subsequently, isomerization of the hydroxyl group and double bond of 2-flaven-3,4-diol occurs spontaneously to yield the thermodynamically more stable 3-flaven-2,3-diol (pseudobase) (41) at cytosolic pH. The 3-GT catalyzes 3-O-glucosylation of the pseudobase form under cytosolic neutral conditions. The pseudobase form of anthocyanidin 3-glucoside is then transported into vacuoles; and under the moderate acidic condition of vacuoles (pH 5), the colored flavylium ion of anthocyanidin 3-glucoside is formed by removal of the hydroxyl anion from the C-2 position of the pseudobase form. For the net formation of colored anthocyanidin 3-glucoside (flavylium ion form) from leucoanthocyanidin, only two enzymes, ANS and 3-GT, are therefore required, and the change of pH following transport of anthocyanidin 3-glucoside into vacuoles is sufficient and essential for the coloring of anthocyanidin 3-glucoside (or more modified anthocyanins). It has been suggested that ANS and 3-GT might work as a multienzyme complex (2). If an ANS-3-GT complex were formed to catalyze the sequential reactions, the intermediate of the reactions, anthocyanidin, could then be rapidly channeled between these two enzymes. However, no evidence has been provided for the channeling of the intermediate (Fig. 6). In addition, our recent study with yeast two-hybrid experiments suggests that there is no apparent interaction between ANS and 3-GT from P. frutescens. 3 Because a protein-protein interaction between several anthocyanin biosynthetic enzymes of Arabidopsis has been reported recently (42), a more complicated organization of the entire biosynthetic machinery may be involved.