Oxidative Transformation of Dihydroflavonols and Flavan-3-ols by Anthocyanidin Synthase from Vitis vinifera

Twelve polyphenols from three distinct families (dihydroflavonols, flavan-3-ols, and flavanones) were studied as potential substrates of anthocyanidin synthase from Vitis vinifera (VvANS). Only flavan-3-ols of (2R,3S) configuration having either a catechol or gallol group on ring B are accepted as substrates. Only dihydroflavonols of (2R,3R) configuration are accepted as substrates, but a catechol or gallol group is not mandatory. Flavanones are not substrates of VvANS. HPLC and MS/MS analyses of the enzymatic products showed that the VvANS-catalyzed oxidative transformation of (+)-dihydroflavonols, such as dihydroquercetin, dihydrokaempferol and dihydromyricetin, leads only to the corresponding flavonols. Among the flavan-3-ols recognized as substrates, (+)-gallocatechin was only transformed into delphinidin by VvANS, whereas (+)-catechin was transformed into three products, including two major products that were an ascorbate–cyanidin adduct and a dimer of oxidized catechin, and a minor product that was cyanidin. Data from real-time MS monitoring of the enzymatic transformation of (+)-catechin suggest that its products are all derived from the initial C3-hydroxylation intermediate, i.e., a 3,3-gem-diol, and their most likely formation mechanism is discussed.


Introduction
Anthocyanins, an important subfamily of flavonoids, are among the most widely distributed water-soluble pigments in plants, and are largely responsible for the colors of flowers, fruits, and vegetables that range from orange or red to purple or blue. They have long been reported to fulfill multiple ecological and physiological functions within plants by contributing to their growth and subsistence in many ways. For example, they enable plants to maintain higher growth rates in metal-polluted soils and to postpone plant senescence under nitrogen deficiency or nutrient deficiency [1]. Moreover, they act as visual signals that attract pollinators and seed-dispersers, protect against pathogens and herbivores through phytoalexins and chemical repellents, act as photo-protecting screens against damage from intense light and UV-B radiation, and inhibit the overproduction of reactive oxygen species (ROS) by chelating metals and scavenging or destroying ROS [2][3][4][5][6][7][8][9]. When used as nutraceuticals, additional properties of interest include their
ules 2022, 27, x FOR PEER REVIEW 3 [40] and VvANS [32]. Four other dihydroflavonols were tested, including (+)-DHK DHM, (+)-epiDHQ, and (-)-epiDHQ. (+)-DHK has a single phenolic hydroxyl on ring B (Figure 1), and its transforma by VvANS was studied first. As shown in Figure 2A, reverse-phase HPLC analysis o commercial standard of (+)-DHK shows two peaks on the chromatogram. The m peak 1 (16.1 min) corresponds to (+)-DHK, and the small peak 2 (19.6 min) is a cont nant of the commercial product of (+)-DHK; this was unambiguously identified b verse-phase HPLC and MS/MS as naringenin, using a commercial standard of naringenin ( Figure S1). Upon incubation with VvANS ( Figure 2B, in blue), a single p uct (peak 3, 22.3 min) was observed with the same retention time as that of a comme standard of kaempferol (red peak 3′, 22.3 min). This product was collected during H analysis, and its characterization by MS and MS/MS was subsequently carried out u commercial kaempferol as external standard. As shown in Figure 2C, a single major p at m/z 287.04 is observed on the MS spectrum of peak 3, which should correspond to [M+H] + ion of the product. Therefore, the product visualized as peak 3 has a molec weight of 286.03 Da, which is the same as that of kaempferol ([M+H] + , m/z 287.04) (Fi 2D). Further MS/MS fragmentation of the ion with m/z 287.04 yields several fragm ions ( Figure 2E), which are very similar to those of the [M+H] + ion of kaempferol (Fi 2F). We conclude that (+)-DHK is accepted by VvANS as a substrate to be exclusi transformed into kaempferol. (+)-DHK has a single phenolic hydroxyl on ring B (Figure 1), and its transformation by VvANS was studied first. As shown in Figure 2A, reverse-phase HPLC analysis of the commercial standard of (+)-DHK shows two peaks on the chromatogram. The major peak 1 (16.1 min) corresponds to (+)-DHK, and the small peak 2 (19.6 min) is a contaminant of the commercial product of (+)-DHK; this was unambiguously identified by reversephase HPLC and MS/MS as naringenin, using a commercial standard of (±)-naringenin ( Figure S1). Upon incubation with VvANS ( Figure 2B, in blue), a single product (peak 3, 22.3 min) was observed with the same retention time as that of a commercial standard of kaempferol (red peak 3 , 22.3 min). This product was collected during HPLC analysis, and its characterization by MS and MS/MS was subsequently carried out using commercial kaempferol as external standard. As shown in Figure 2C, a single major peak at m/z 287.04 is observed on the MS spectrum of peak 3, which should correspond to the [M+H] + ion of the product. Therefore, the product visualized as peak 3 has a molecular weight of 286.03 Da, which is the same as that of kaempferol ([M+H] + , m/z 287.04) ( Figure 2D). Further MS/MS fragmentation of the ion with m/z 287.04 yields several fragment ions ( Figure 2E), which are very similar to those of the [M+H] + ion of kaempferol ( Figure 2F). We conclude that (+)-DHK is accepted by VvANS as a substrate to be exclusively transformed into kaempferol.

(+)-Dihydromyricetin (DHM)
The enzymatic transformation of (+)-DHM, a dihydroflavonol containing a gallol group on ring B (Figure 1), was studied next. As shown in Figure 3A, reverse-phase HPLC analysis of the commercial standard of (+)-DHM showed only a major peak (peak 1, 11.2 min) and no contaminant was observed. Upon incubation with VvANS, two small new peaks were observed (peaks 2 and 3, Figure 3B). The blue peak 3 observed at 18.4 min corresponded to the single enzymatic product of (+)-DHM, and it was assigned to myricetin based on the HPLC retention time of the corresponding commercial standard (red peak 3′, 18.4 min). Further MS and MS/MS analyses of the collected peak 3 revealed that its MS and MS/MS spectra were very similar to those of commercial myricetin (Figure S2). The blue peak 2 observed at 12.3 min was considered as a non-enzymatic product of (+)-DHM or a complex of (+)-DHM with iron(II), because this peak was again observed upon incubation of (+)-DHM in the reaction mixture containing iron(II) salt but no enzyme ( Figure 3C). We conclude that (+)-DHM is a substrate of VvANS that gives myricetin as a single enzymatic product.

(+)-Dihydromyricetin (DHM)
The enzymatic transformation of (+)-DHM, a dihydroflavonol containing a gallol group on ring B (Figure 1), was studied next. As shown in Figure 3A, reverse-phase HPLC analysis of the commercial standard of (+)-DHM showed only a major peak (peak 1, 11.2 min) and no contaminant was observed. Upon incubation with VvANS, two small new peaks were observed (peaks 2 and 3, Figure 3B). The blue peak 3 observed at 18.4 min corresponded to the single enzymatic product of (+)-DHM, and it was assigned to myricetin based on the HPLC retention time of the corresponding commercial standard (red peak 3 , 18.4 min). Further MS and MS/MS analyses of the collected peak 3 revealed that its MS and MS/MS spectra were very similar to those of commercial myricetin ( Figure S2). The blue peak 2 observed at 12.3 min was considered as a non-enzymatic product of (+)-DHM or a complex of (+)-DHM with iron(II), because this peak was again observed upon incubation of (+)-DHM in the reaction mixture containing iron(II) salt but no enzyme ( Figure 3C). We conclude that (+)-DHM is a substrate of VvANS that gives myricetin as a single enzymatic product.

(+)-Catechin
The transformation of flavan-3-ols by VvANS was first carried out using (+)-catechin as substrate, and a control experiment was performed in the absence of VvANS ( Figure 5A). As shown in Figure 5B (in red), four peaks (peaks 1-4) are observed upon incubation with VvANS, among which peak 2 (9 min) corresponds to residual (+)-catechin, whereas the other three peaks correspond to new enzymatic products. As shown in Figure 5B (in red), the small peak 4 (18.6 min) could be assigned to cyanidin based on the HPLC retention time of the commercial standard of cyanidin chloride (blue peak 4′, 18.5 min), and further MS analyses of the collected peak 4 confirmed that its MS and MS/MS spectra were very similar to those of commercial cyanidin chloride ( Figure S3).

(+)-Catechin
The transformation of flavan-3-ols by VvANS was first carried out using (+)-catechin as substrate, and a control experiment was performed in the absence of VvANS ( Figure 5A). As shown in Figure 5B (in red), four peaks (peaks 1-4) are observed upon incubation with VvANS, among which peak 2 (9 min) corresponds to residual (+)-catechin, whereas the other three peaks correspond to new enzymatic products. As shown in Figure 5B (in red), the small peak 4 (18.6 min) could be assigned to cyanidin based on the HPLC retention time of the commercial standard of cyanidin chloride (blue peak 4 , 18.5 min), and further MS analyses of the collected peak 4 confirmed that its MS and MS/MS spectra were very similar to those of commercial cyanidin chloride ( Figure S3).
The other two major products eluted at 5.5 and 10.5 min (peaks 1 and 3) were also collected and analyzed by MS and MS/MS. As shown in Figure 5C, six major ions, with m/z 115.04, 445.12, 463.09, 519.15, 536.17, and 593.17, were observed on the MS spectrum of collected peak 1, among which only the peak at m/z 463.09 corresponds to the [M+H] + ion of the product; the five others are common contaminant ions encountered in positive ionization mass spectrometry [42]. Further MS/MS analysis of this [M+H] + ion with m/z 463.09 yielded several fragment ions ( Figure 5D), of which the two major ions with m/z 177.02 and 287.07 correspond respectively to the [M+H] + ion derived from ascorbic acid (theoretical exact mass 176.03 Da) and M + ion from cyanidin (theoretical exact mass 287.06 Da). We conclude that m/z 463.09 corresponds to the [M+H] + ion of a covalent adduct of ascorbate with cyanidin (176.03 + 287.06 = 463.09 Da), which means that the product visualized as peak 1 should be an ascorbate-cyanidin adduct with a molecular weight of 462.08 Da.  Figure 5D), of which the two major ions with m/z 177.02 and 287.07 correspond respectively to the [M+H] + ion derived from ascorbic acid (theoretical exact mass 176.03 Da) and M + ion from cyanidin (theoretical exact mass 287.06 Da). We conclude that m/z 463.09 corresponds to the [M+H] + ion of a covalent adduct of ascorbate with cyanidin (176.03 + 287.06 = 463.09 Da), which means that the product visualized as peak 1 should be an ascorbate-cyanidin adduct with a molecular weight of 462.08 Da.
As shown in Figure 5E, only the major ion with m/z 575.11 is observed on the MS spectrum of collected peak 3, which should be the [M+H] + ion of the product. Further MS/MS analysis of this [M+H] + ion with m/z 575.11 yielded several fragment ions ( Figure  5F); among them, the most important fragment ion is that with m/z 287.07, which corresponds to the flavylium cation of cyanidin or an equivalent tautomeric form of the cation. Therefore, we conclude that m/z 575.11 most likely corresponds to the [M+H] + ion of a dimer of cyanidin or oxidized catechin (theoretical exact mass 287.06 + 287.06 = 574.12 Da), which means that the product visualized as peak 3 should be a dimer of cyanidin As shown in Figure 5E, only the major ion with m/z 575.11 is observed on the MS spectrum of collected peak 3, which should be the [M+H] + ion of the product. Further MS/MS analysis of this [M+H] + ion with m/z 575.11 yielded several fragment ions ( Figure 5F); among them, the most important fragment ion is that with m/z 287.07, which corresponds to the flavylium cation of cyanidin or an equivalent tautomeric form of the cation. Therefore, we conclude that m/z 575.11 most likely corresponds to the [M+H] + ion of a dimer of cyanidin or oxidized catechin (theoretical exact mass 287.06 + 287.06 = 574.12 Da), which means that the product visualized as peak 3 should be a dimer of cyanidin or oxidized catechin with a molecular weight of 573.10 Da. Given the similarity of our results and those of Wellman et al. [41] with ANS from Gerbera hybrida, we believe that this dimer is the same as theirs, i.e., a C 4 -C 4 symmetrical dimer of oxidized catechin, which they identified by NMR.

(+)-Gallocatechin
(+)-Gallocatechin, which possesses a gallol group on ring B and therefore one more phenolic hydroxyl group than catechin (Figure 4), was tested next. A control experiment was also performed in the absence of VvANS ( Figure 6A). A single enzymatic product (peak 2, 17.2 min) was observed upon incubation with VvANS ( Figure 6B, in blue) with no residual substrate, indicating that (+)-gallocatechin is a substrate of VvANS that is more efficiently transformed by the enzyme than (+)-catechin. As shown in Figure 6B (in red), the single enzymatic product of (+)-gallocatechin could be assigned to delphinidin based on the HPLC retention time of the commercial standard of delphinidin chloride (peak 2 , 17.6 min); this was further confirmed by MS and MS/MS analyses using commercial delphinidin chloride as an external standard ( Figure S4). which they identified by NMR.

(+)-Gallocatechin
(+)-Gallocatechin, which possesses a gallol group on ring B and therefore one more phenolic hydroxyl group than catechin (Figure 4), was tested next. A control experiment was also performed in the absence of VvANS ( Figure 6A). A single enzymatic product (peak 2, 17.2 min) was observed upon incubation with VvANS ( Figure 6B, in blue) with no residual substrate, indicating that (+)-gallocatechin is a substrate of VvANS that is more efficiently transformed by the enzyme than (+)-catechin. As shown in Figure 6B (in red), the single enzymatic product of (+)-gallocatechin could be assigned to delphinidin based on the HPLC retention time of the commercial standard of delphinidin chloride (peak 2′, 17.6 min); this was further confirmed by MS and MS/MS analyses using commercial delphinidin chloride as an external standard ( Figure S4).

Transformation of Flavanones by VvANS
Finally, flavanones, which are characterized by the absence of a substituent at C3, could be envisaged as potential substrates of VvANS, since naringenin, a member of this family, has been shown to act as a substrate of AtANS [35,43]. Therefore, naringenin was also tested as substrate of VvANS, and two different commercial standards of naringenin were used, including a racemic mixture of naringenin [(±)-naringenin] and a pure natural stereoisomer of naringenin (2S-naringenin). No enzymatic product could be visualized in either case ( Figure S6), indicating that, unlike AtANS, VvANS does not recognize naringenin as substrate.

Transformation of Flavanones by VvANS
Finally, flavanones, which are characterized by the absence of a substituent at C 3 , could be envisaged as potential substrates of VvANS, since naringenin, a member of this family, has been shown to act as a substrate of AtANS [35,43]. Therefore, naringenin was also tested as substrate of VvANS, and two different commercial standards of naringenin were used, including a racemic mixture of naringenin [(±)-naringenin] and a pure natural stereoisomer of naringenin (2S-naringenin). No enzymatic product could be visualized in either case ( Figure S6), indicating that, unlike AtANS, VvANS does not recognize naringenin as substrate.
In summary, only five of the twelve tested polyphenols are substrates of VvANS, including three polyphenols of the family of dihydroflavonols and two polyphenols of the family of flavan-3-ols. The main information concerning the enzymatic transformation of these five substrates by VvANS is summarized in Table 1. The other six polyphenols, including (+)-epiDHQ, (−)-epiDHQ, (+)-afzelechin, (−)-catechin, (+)-epicatechin, and (−)-epicatechin, are not accepted by VvANS as substrates.

Real-Time MS Monitoring of the Enzymatic Transformation of (+)-Catechin
As above described, an ascorbate-cyanidin adduct and a dimer of cyanidin (or oxidized catechin) were observed as VvANS products of (+)-catechin. To try to clarify their formation mechanism, the enzymatic transformation of (+)-catechin was monitored by means of real-time mass spectrometry. However, the continuous injection of a reaction mixture containing free iron salt had to be avoided. Therefore, a VvANS-iron(II)-oxoglutarate complex was prepared extemporaneously ( Figure S7) and used for the enzymatic transformation of (+)-catechin in the absence of free iron(II) salts. We had first checked that this VvANS-iron(II)-oxoglutarate complex was fully active, as shown by its ability to produce the same enzymatic product as the enzyme incubated in the presence of free iron(II) salt ( Figure S8).
As shown in Figure 7, a rapid decrease of (+)-catechin is observed once the enzymatic reaction is initiated. Another molecule with m/z 290.07 decreases at about the same rate. It gives a much smaller signal, and it is not observed in the absence of enzyme, which means that it is most likely the one-electron oxidation product of catechin that is immediately formed by enzyme-catalyzed hydrogen abstraction, a generic step in iron/oxoglutarate oxygenases [44]. The next major event is the slower formation of four products (m/z 288.07, 307.09, 463.09, and 287.06). The first peak (m/z 288.07) has an extremely weak signal intensity and it corresponds to a three-electron oxidation product of catechin. The next peak (m/z 307.09) has a much higher signal intensity, and it is again clearly an intermediate rather than a stable end product. It can be assigned to a monohydroxylated product of (+)-catechin, based on its molecular mass (theoretical exact mass of the [M+H] + ion: 290.08 + 17.00 = 307.08 Da), and it is most likely C 3 -hydroxylated (+)-catechin, i.e., a 3,3-gem-diol according to our previous study [32]. This intermediate is produced at a much higher initial rate than the other products, reaching a maximal concentration by approximately 5 min and then disappearing quite rapidly and becoming almost entirely consumed by 30 min, when the other products have reached their maximal values. This strongly suggests that such products are derived from this monohydroxylated intermediate.
The first two products immediately observed are the ascorbate-cyanidin adduct ([M+H] + , m/z 463.09) and cyanidin itself (M + , m/z 287.06), which is the four-electron oxidation product. A fourth product is the dimer of oxidized catechin ([M+H] + , m/z 575.12), but it is only observed after a lag time of at least 8 min. This suggests that the dimer is not directly produced from monohydroxylated catechin.
Based on signal intensities, it is likely that the ascorbate-cyanidin adduct is a major product in this continuous flow experiment, whereas cyanidin and the dimer of oxidized catechin are two minor products. Molecules 2022, 27, x FOR PEER REVIEW 10 of 19 Figure 7. Time-course of (+)-catechin consumption and product formation by real-time mass spectrometry.
The first two products immediately observed are the ascorbate-cyanidin adduct ([M+H] + , m/z 463.09) and cyanidin itself (M + , m/z 287.06), which is the four-electron oxidation product. A fourth product is the dimer of oxidized catechin ([M+H] + , m/z 575.12), but it is only observed after a lag time of at least 8 min. This suggests that the dimer is not directly produced from monohydroxylated catechin.
Based on signal intensities, it is likely that the ascorbate-cyanidin adduct is a major product in this continuous flow experiment, whereas cyanidin and the dimer of oxidized catechin are two minor products.
The two radical species (m/z 290.07 and 288.07) were not detected after 30 min in the reactional medium coupled to HPLC, which is expected for species with a rather short half-life. More importantly, the hydroxylation product of (+)-catechin had not been observed either in the reactional medium coupled with HPLC analysis (Figure 5B), which means that after a 30 min incubation period, the residual level of this intermediate was too low to be detected.
By contrast, the dimer of oxidized catechin, which is observed as a minor product in these experiments of continuous flow and real-time mass spectrometry, was a major product (together with the ascorbate-cyanidin adduct) in the reactional medium The two radical species (m/z 290.07 and 288.07) were not detected after 30 min in the reactional medium coupled to HPLC, which is expected for species with a rather short halflife. More importantly, the hydroxylation product of (+)-catechin had not been observed either in the reactional medium coupled with HPLC analysis (Figure 5B), which means that after a 30 min incubation period, the residual level of this intermediate was too low to be detected.
By contrast, the dimer of oxidized catechin, which is observed as a minor product in these experiments of continuous flow and real-time mass spectrometry, was a major product (together with the ascorbate-cyanidin adduct) in the reactional medium coupled with HPLC analysis ( Figure 6B). There are at least two distinct experimental features that may be responsible for this. The first is that the reaction temperature was 35 • C in the reactional medium coupled with HPLC analysis, whereas it had to be fixed at 22 • C for continuous flow MS experiments. This suggests therefore that the formation of the dimer of oxidized catechin could be temperature-dependent. Perhaps a free radical precursor with a longer half-life at 22 • C would favor the dimerization path. The second difference relates to the availability of dissolved oxygen. In the experiments coupled with HPLC analysis, the reaction medium is in a vessel open to ambient air and under continuous stirring for 30 min, which means that the initial concentration of dissolved oxygen (230-260 µM under 1 Atm) is not limiting. By contrast, in the real-time MS experiment, the reaction medium is introduced at t 0 from a tight syringe with no headspace, meaning that the initial concentration of dissolved O 2 is limiting, which may result in a faster decrease in oxygen that might differently affect the yield of each product. For example, a free-radical mediated carbon-carbon dimerization would be more efficient at lower oxygen concentration where the formation of peroxyl radicals is less favorable. More generally, this oxygen-limiting experimental setup is probably the reason why low yet significant amounts of free radical intermediates (m/z 290.07 and m/z 288.07) are detected by MS.

Structure-Activity Relationships
VvANS possesses a wide range of substrates, as expected from what had been observed with ANS from Arabidopsis thaliana and Gerbera hybrida [35,40,41,43], although some differences were observed with VvANS, such as the absence of transformation of naringenin, and the transformation of (+)-catechin into an ascorbate-cyanidin adduct. Based on our study, the following structure-activity relationships can be formulated for VvANS: Flavonols only exist as glucosides in grape berries, and the most abundant flavonol glucosides observed in Cabernet sauvignon grape berries are derived from quercetin and myricetin [45]. VvANS does not more efficiently to produce flavonols that are di-or trihydroxylated on ring B, such as quercetin and myricetin, but it does produce them, which may have some physiological significance.

Lack of Reliable Kinetic Data
Although multiple kinetic experiments have been performed over the years in our laboratory with dihydroflavonol reductases, leucoanthocyanidin reductases, and anthocyanidin reductases, we were not able to measure steady-state initial rates with VvANS to estimate distinct k cat and K M values for flavan-3-ols and dihydroflavonols. There are several reasons for this. The first reason is that our products cannot be specifically monitored by spectrophotometry, except for cyanidin, which is not a major and single product. Moreover, the instability of cyanidin (half-life close to 30-40 min in our conditions) is incompatible with a steady-state initial rate measured over more than a few minutes, which would also be the case with delphinidin. An alternative is to monitor oxygen consumption, but we could not prepare stock solutions of active oxoglutarate-enzyme complex at concentrations higher than 2 µM. We had therefore to initiate the reaction by adding the polyphenolic substrate in an ethanolic or methanolic stock solution that invariably resulted in a transient increase in apparent dissolved oxygen, which perturbed the signal for several minutes with either an oxygen Clark electrode or an optical fiber-based oximeter (see Figure S9). We therefore could not measure initial rates at steady-state by oximetry. To the best of our knowledge, no steady-state kinetics have ever been reported with purified anthocyanidin synthases.

Anthocyanidin Production and Hypothetical Reactional Mechanism Involved in (+)-Catechin Transformation
Overall, the real-time MS analysis, which was summarized in Figure 7, enables us to visualize an initial hydroxylated intermediate ([M+H] + , m/z 307.09) from which plausible sequences of product formation are summarized in Figure 8. An initial C 3 -hydrogen abstraction is catalyzed by VvANS on (+)-catechin 1 to produce the transient C 3 -free radical 2 from which hydroxylated catechin 3 (3,3-gem-diol) is the first metastable intermediate produced by the enzyme. The corresponding 3,3-gem-diol is expected to dehydrate into an enediol with a double bond at C 2 -C 3 (compound 4) or at C 3 -C 4 . However, the subsequent formation of cyanidin 5 (drawn in uncharged form here) and of the symmetrical C-C dimer 7 could be achieved only with a double bond at C 2 -C 3 (compound 4). Cyanidin should predominate in its two uncharged tautomeric forms (quinone methide) in the reaction mixture (pH 6.3). They should have the same reactivity (electrophilic C 4 ) and therefore only one of them has been drawn on Figure 8. The formation of a symmetrical C-C dimer 7 can only be conceived from the recombination of C 4 free radicals (compound 4a). The latter should be produced again by hydrogen abstraction, a major catalytic step in iron/oxoglutarate oxygenases, which is generally associated with the transition Fe IV =O to Fe III -OH at the active site [44]. oxygenases is to reduce Fe(III) to Fe(II), thereby preventing oxidative inactivation of the active site during catalysis. We do not yet know the structure of the ascorbate-cyanidin adduct, and it should be underlined that it is not observed in the absence of enzyme. Moreover, no ascorbatedelphinidin adduct was observed when (+)-gallocatechin was used as a substrate. Among the five substrates that were identified in this work, (+)-dihydroflavonols were only transformed into the corresponding flavonols, and (+)-gallocatechin was converted into a single product, delphinidin.
The VvANS-catalyzed oxidative transformation of (+)-catechin is more complex, Figure 8. Hypothetical reactional scheme for VvANS transformation of (+)-catechin. 2-oxoglu = 2-oxoglutarate; AH 2 = ascorbic acid; All m/z refer to (M+1) cationic species, and m/z which were observed by MS are in bold blue. The expected iron oxidation states of the enzyme active site are indicated only for hydrogen abstraction steps. Note that the flavylium cation of cyanidin would have the same m/z 287.06 as that observed for the mono-protonated form of the uncharged quinone methide since they are tautomers.
We therefore conclude that structures 1, 2, 3, 4, 4a, and 7 are all mandatory, but there are several possibilities for the formation of both cyanidin 5 and the ascorbate-cyanidin adduct 6 from such structures. The direct transformation of 4 to 5 by VvANS cannot be ruled out, but it should be considered as highly speculative in the absence of realistic mechanistic rationale. We think that the transformation of structure 4a into cyanidin is a plausible alternative, and two possibilities can be envisaged. The first is that of step (I), a VvANS-catalyzed C 4 -hydroxylation, which despite a change in regiospecificity from C 3 to C 4 , would be classical for an iron/oxoglutarate oxygenase. The second is step (II), which relies on dioxygen addition to the C 4 -centered free radical to produce a peroxyl radical intermediate. The addition of dioxygen to carbon-centered free radicals is indeed known to be very fast [46]. With such a peroxyl radical intermediate, a fast irreversible extrusion of superoxide O 2 − would be expected, because at pH 6.3, the non-enzymatic dismutation of superoxide is also fast [47]. Presumably, the short half-life of this putative peroxyl radical would imply that the corresponding M+1 ion with m/z 320 would not be visualized by mass spectrometry.
As for the formation of the ascorbate-cyanidin adduct 6, we tentatively suggest in Figure 8 that it could be formed by nucleophilic addition of the ascorbate anion AH − to the electrophilic C 4 center of cyanidin 5, although a free radical mechanism involving compound 4a cannot be ruled out. Ascorbate is a good one-electron donor that behaves as an essential cofactor of VvANS. Its generic function in iron/oxoglutarate/ascorbate oxygenases is to reduce Fe(III) to Fe(II), thereby preventing oxidative inactivation of the active site during catalysis.
We do not yet know the structure of the ascorbate-cyanidin adduct, and it should be underlined that it is not observed in the absence of enzyme. Moreover, no ascorbate-delphinidin adduct was observed when (+)-gallocatechin was used as a substrate.
Among the five substrates that were identified in this work, (+)-dihydroflavonols were only transformed into the corresponding flavonols, and (+)-gallocatechin was converted into a single product, delphinidin.
The VvANS-catalyzed oxidative transformation of (+)-catechin is more complex, since three products were observed, including an ascorbate-cyanidin adduct ([M+H] + , m/z 463.09), a dimer of oxidized catechin ([M+H] + , m/z 575.12), and cyanidin. However, the data obtained from real-time MS monitoring of the enzymatic reaction imply that these three products are all derived from an initial common intermediate that is a monohydroxylated (+)-catechin, and most likely a 3,3-gem-diol ([M+H] + , m/z 307.09), based on our recent work [32].
We cannot explain why the ascorbate-cyanidin adduct and the dimer of oxidized catechin are observed in these in vitro experiments with our recombinant VvANS even though they were never described in biological extracts. This might result from in vitro conditions that do not exactly reproduce the in vivo conditions (microenvironment, enzyme/substrate ratio, cofactors), including the possibility that in vivo, ANS is part of a multi-enzyme channeling complex from which the formation of such cyanidin by-products is negligible.
However, the fact that VvANS can more efficiently produce pure delphinidin from gallocatechin than cyanidin from catechin may have some physiological significance, since malvidin and its glucoside, which are derived from delphinidin, are considerably more abundant in Vitis vinifera than cyanidin and its glucoside [48,49].
At a minimum, this work demonstrates that flavan-3-ols are substrates of VvANS, and that significant amounts of anthocyanidins can be produced in vitro by the enzyme from such substrates, which is not the case with leucoanthocyanidins [32]. The formation of significant amounts of cyanidin from (+)-catechin has also been confirmed with ANS from Gerbera hybrida [41] and from Arabidopsis thaliana [43]. In addition, with metabolically engineered bacteria, the optimal production of anthocyanins also required catechin [50][51][52][53].
For years, ANS and LDOX (leucoanthocyanidin dioxygenase) have been implicitly considered as two distinct names of a single enzyme. More recently, a thorough inves-tigation of functional genomics and purified enzymes in Medicago truncatula led to the conclusion that two distinct enzymes were able to produce anthocyanins in two parallel pathways with distinct regulations [54]: one that the authors called MtLDOX (access nb: XP_003601080.1) accepted (+)-catechin as a substrate to produce transient cyanidin immediately reduced by ANR to (+)-epicatechin, and the other that the authors called MtANS (access nb: XP_003611189.1) accepted leucocyanidin as a substrate to produce transient cyanidin immediately glycosylated to anthocyanins or reduced by ANR to (−)-epicatechin. In their interpretation, (−)-epicatechin produced from leucocyanidin and MtANS would serve as a precursor unit of polymeric proanthocyanidins whereas (−)-epicatechin produced from (+)-catechin would serve as an extension unit.
We do not know which role VvANS might play in the production of starter and extension units of proanthocyanidins in grapes, but our results support the idea that leucoanthocyanidin reductase (LAR) of Vitis vinifera could play a major role in the biosynthetic pathway of anthocyanidins by transforming leucoanthocyanidins into flavan-3-ols, which would then be the immediate precursors of anthocyanidins in vivo as ANS substrates. Just as for ANS/LDOX, there are two homologs of LAR enzymes (LAR1 and LAR2) in grape, but their structures and activities are almost identical [54]. They are however not only producing flavan-3-ols since they also convert the C 4 -thioethers 4β-(S-cysteinyl)-catechin and 4β-(S-cysteinyl)-epicatechin into (+)-catechin and (−)-epicatechin, respectively [55].
Data obtained on purified enzymes are very useful to know which transformations should be taken into consideration, especially for bioproduction optimization. However, given the enzymatic promiscuity of major actors, such as LAR, ANS/LDOX, and even ANR [56], unraveling the complexity of the biosynthetic pathway of anthocyanins in plants remains a considerable challenge.

Conclusion
In conclusion, among twelve tested polyphenols, only three dihydroflavonols and two flavan-3-ols were accepted as substrates of VvANS, with structure-activity relationships, mechanistic information, and one technical innovation that may be summarized as follows: (1) Only 2R,3S-flavan-3-ols having either a catechol group (two adjacent phenolic hydroxyl groups) or a gallol group (three adjacent phenolic hydroxyl groups) as ring B are accepted as substrates. A simple phenol is not accepted. (2) Only dihydroflavonols of (2R,3R) configuration are accepted as substrates.
(3) VvANS is able to produce anthocyanidins from flavan-3-ols such as (+)-gallocatechin or (+)-catechin, but not from leucocyanidin. (4) When (+)-catechin is used as a substrate, cyanidin is not the major product observed; instead, the main products include an ascorbate covalent adduct of cyanidin and a dimer. For this complex multiple transformation, an experimental setup has been developed to monitor substrate, intermediates, and products by real-time mass spectrometry. The results imply that all products are derived from an initial 3,3-gem-diol intermediate that undergoes 2,3-dehydration to a flav-2-en-3-ol. Further transformation of this structure into C-C-dimer, ascorbate adduct, or cyanidin most likely involves a C 4 -free radical intermediate whose fate may artificially favor the dimer and the adduct because of the time-dependent limitation on dissolved oxygen. Further investigations will be required to understand why such a reactional complexity is not observed with (+)-gallocatechin, but it is noteworthy that its single product, namely delphinidin, is the precursor of malvidin and its glucoside, which are considerably more abundant in grapes of Vitis vinifera than cyanidin and its glucoside. (5) This work also shows for the first time that an active iron/oxoglutarate dioxygenase can be extemporaneously prepared by gel filtration in the form of an iron/oxoglutarate complex, which can then be used without introduction of free iron salt in the reactional medium.

Enzyme Source and Design of Reaction Medium
Recombinant anthocyanidin synthase from Vitis vinifera (VvANS, untagged enzyme) was produced and stored as previously described [32].

Reverse-Phase HPLC Analysis
After 30 min of reaction, a 100 µL portion of the reaction mixture was analyzed by reverse-phase HPLC using an Atlantis C18 column (5 µm, 4.6 × 250 mm; Waters, Milford, MA, USA) with the same HPLC parameters as previously described [32].

Tandem Mass Spectrometry (MS/MS) Analysis
HPLC elution fractions of the products were collected and analyzed by positiveion electrospray ionization tandem mass spectrometry (ESI-MS/MS), using a Q-Tof Premier mass spectrometer (Waters, Milford, MA, USA) under conditions described by Zhang et al. [32]. We systematically used MS and MS/MS analysis of HPLC-purified products, as well as freshly prepared methanolic solutions of 2 mM commercial standards including kaempferol, quercetin, myricetin, cyanidin and delphinidin, in agreement with classical MS fragmentation analysis of polyphenols [58,59].

Real-Time MS Monitoring of the Enzymatic Transformation of (+)-Catechin
To avoid continuous-flow injection of iron(II) into the ESI source of the mass spectrometer, the transformation of (+)-catechin was carried out in a reaction mixture containing no free iron(II) salt using the extemporaneously prepared holoenzyme loaded with iron(II), further referred to as VvANS-Fe(II)-oxoglutarate complex.
The reaction mixture (2.5 mL, final volume) contained 20 mM ammonium acetate, 20 mM NaCl, 2 mM ascorbate, 1 mM 2OG and ≈ 10 −6 M VvANS-iron(II)-oxoglutarate complex, pH 6.3 (at 22 • C), and the reaction was initiated by the addition of 100 µM (+)-catechin. Upon vortex homogenization within a few seconds, 300 µL of the reaction mixture was taken with a 500 µL Hamilton syringe and infused into the ESI source under a flow rate of 5 µL/min by means of a syringe pump. The reaction was then monitored by real-time MS in positive ion mode in conditions previously described [60].
Supplementary Materials: The following are available online. Figure S1: Identification of the contaminant of the commercial product of (+)-DHK; Figure S2: MS and MS/MS analysis of the product visualized as peak 3 in Figure 3B; Figure S3: MS and MS/MS analysis of the product visualized as peak 4 in Figure 5B; Figure S4: MS and MS/MS analysis of the product visualized as peak 2 in Figure 6B; Figure S5: Reverse-Phase HPLC analysis of the transformation of (+)-afzelechin (A), (−)-epicatechin (B), (+)-epicatechin (C) and (−)-epicatechin (D) by VvANS; Figure S6: Reverse-Phase HPLC analysis of the enzymatic transformation of naringenin; Figure S7: Protocol of production of the VvANS-iron(II)-oxoglutarate complex; Figure S8: HPLC determination of the activity of the VvANS-iron(II) complex; Figure S9: Oxygen signal perturbation upon injection of methanolic solution of (+)-catechin; Figure S10: Sequence alignments of VvANS with ANS and LDOX from Medicago truncatula.