Mechanistic studies on three 2-oxoglutarate-dependent oxygenases of flavonoid biosynthesis: anthocyanidin synthase, flavonol synthase, and flavanone 3beta-hydroxylase.

Anthocyanidin synthase (ANS), flavonol synthase (FLS), and flavanone 3beta-hydroxylase (FHT) are involved in the biosynthesis of flavonoids in plants and are all members of the family of 2-oxoglutarate- and ferrous iron-dependent oxygenases. ANS, FLS, and FHT are closely related by sequence and catalyze oxidation of the flavonoid "C ring"; they have been shown to have overlapping substrate and product selectivities. In the initial steps of catalysis, 2-oxoglutarate and dioxygen are thought to react at the ferrous iron center producing succinate, carbon dioxide, and a reactive ferryl intermediate, the latter of which can then affect oxidation of the flavonoid substrate. Here we describe work on ANS, FLS, and FHT utilizing several different substrates carried out in 18O2/16OH2, 16O2/18OH2, and 18O2/18OH2 atmospheres. In the 18O2/16OH2 atmosphere close to complete incorporation of a single 18O label was observed in the dihydroflavonol products (e.g. (2R,3R)-trans-dihydrokaempferol) from incubations of flavanones (e.g. (2S)naringenin) with FHT, ANS, and FLS. This and other evidence supports the intermediacy of a reactive oxidizing species, the oxygen of which does not exchange with that of water. In the case of products formed by oxidation of flavonoid substrates with a C-3 hydroxyl group (e.g. (2R,3R)-trans-dihydroquercetin), the results imply that oxygen exchange can occur at a stage subsequent to initial oxidation of the C-ring, probably via an enzyme-bound C-3 ketone/3,3-gem-diol intermediate.

The flavonoids are a large class of plant secondary metabolites. They contain a 15-carbon phenylpropanoid core, which is extensively modified by rearrangement, alkylation, oxidation, and glycosylation (1). In plants the flavonoids fulfill a diverse array of roles including pigmentation and protection against UV photodamage and can act as signaling molecules (1,2). They have been reported to possess a variety of biomedicinal properties including antimalarial, antioxidant, and antitumor activities. Furthermore flavonoids have been shown to modulate the hypoxic response in endothelial cells, an effect that might be mediated by their ability to chelate iron or directly inhibit hydroxylase enzymes involved in the response (3,4). Due to their historic importance in genetic studies and their biomedicinal properties, flavonoid biosynthesis is of interest to both biologists and chemists (1,5,6).
The flavonoid biosynthetic pathway has been extensively studied in plants by a combination of genetic and labeling studies (7)(8)(9). Advances in genetic techniques and recombinant expression technologies have enabled proposals regarding the substrate and product selectivities of individual enzymes to be studied in vitro. The steps leading to the flavonoid core have now been established (see Fig. 1). Formation of chalcone 1 is catalyzed by a type III polyketide synthase known as chalcone synthase via condensation of para-coumaryl-CoA 2 and three malonyl-CoA thioesters 3 (10,11). Chalcone isomerase then catalyzes the intramolecular cyclization of chalcone 1 to give (2S)-naringenin 4, a precursor of many flavonoids 1 including the anthocyanin (e.g. cyanin 5) subfamily (12).
The 2OG 7-dependent oxygenases (Fig. 2) are involved in a range of important pathways, including those leading to collagen, the ␤-lactam antibiotics, and modified amino acids and peptides (36). Roles for 2OG oxygenases have also been identified in the hypoxic signaling pathway and in DNA repair (37)(38)(39)(40)(41). The available evidence suggests that catalysis by 2OG oxygenases proceeds via bidentate binding of 2OG 7 to the active site iron (42,43). Substrate binding is thought to enable dioxygen to displace a ligating water from the catalytic iron center (44,45). Oxidative decarboxylation of 2OG 7 then occurs, producing succinate 10, CO 2 , and a ferryl [Fe(IV)ϭO 7 Fe(III)-O ⅐ ] 11 intermediate, which can subsequently affect hydroxylation or desaturation of the substrate (Fig. 2) (46,47).
Recent crystallographic work has demonstrated that ANS contains the double-stranded ␤-helix common to other 2OG oxygenases and identified the residues involved in substrate binding (38,48).
ANS and FLS are closely related (50 -60% sequence similarity at the polypeptide level) but display a lower level of similarity to FHT and FNS I (Ͻ35%) (26,29). This has led to the proposal that there are two distinct subgroups of flavonoid 2OG oxygenases, one containing FNS I and FHT and another containing FLS and ANS. Substrate analogue work supports this proposal; FNS I and FHT appear to possess relatively narrow substrate selectivities compared with ANS and FLS (21,26,28,29,49).
Here we report work that enhances mechanistic understanding of the non-heme dioxygenases of flavonoid biosynthesis by carrying out assays of ANS, FHT, and FLS in 18 O-labeled dioxygen and water with a variety of substrates.
Cell Growth, Lysis, and Protein Purification-A. thaliana ANS was prepared and purified as reported previously (51). For FLS the pET-3a A. thaliana FLS plasmid was transformed into Escherichia coli BL21(DE3) Gold cells, and an overnight starter culture was used as a 1% inoculum in 600 ml of 2TY medium in 2-liter unbaffled flasks containing ampicillin (100 g ml Ϫ1 ). After inoculation, growth was continued at 37°C and 250 rpm until induction with 0.2 mM isopropyl-   pre-equilibrated at 50 mM Tricine, pH 7.3 before elution using a linear gradient of NaCl from 0 to 0.5 M NaCl over ϳ5 column volumes. FLS eluted at ϳ230 mM NaCl, and the fractions containing FLS at ϳ45-55% purity (by SDS-PAGE analysis) were pooled and further purified. The FLS protein was concentrated and buffer-exchanged into 50 mM MES, pH 6.15, 10% (w/v) glycerol before loading onto the reactive green C-19 column equilibrated using the same buffer. The protein was eluted with a linear gradient of NaCl from 0 to 2 M NaCl. Fractions of ϳ90 -95% purity were pooled, concentrated, buffer-exchanged, and then stored at Ϫ80°C. Petunia hybrida ANS was prepared as reported previously (22). P. hybrida FHT and FLS were prepared from corresponding cDNAs amplified by PCR as for P. hybrida ANS.
A racemic mixture of cis/trans-DHQ was produced by C-2 epimerization of commercial DHQ as reported previously (54). cis-DHQ was purified from the cis/trans-DHQ mixture as follows. Racemic cis/trans-DHQ was dissolved in MeOH (10 mg, 1 ml) and loaded onto a Phenomenex Luna C-18 250-mm ϫ 10-mm (bead size, 4 m) column pre-equilibrated with 50 mM NaH 2 PO 4 , pH 6.8, 20% MeOH (v/v) and run at 3.6 ml min Ϫ1 . The column was run isocratically for 15 min in 50 mM NaH 2 PO 4 , pH 6.8, 20% MeOH (v/v) before a gradient was run to 50 mM NaH 2 PO 4 , pH 6.8, 50% MeOH (v/v) over 25 min followed by 15 min of isocratic elution. The peaks corresponding to the racemic DHQ diastereomers were collected, diluted to Ͻ10% MeOH by addition of Milli-Q H 2 O, and freeze-dried. The purified DHQ (trans-and cis-) diastereomers were dissolved in the minimum volume of 50 mM NaH 2 PO 4 , pH 6.8, 20% MeOH (v/v) and then loaded onto Strata C18E solid phase extraction columns pre-equilibrated as for the purification of the naringinase  Tables I and II). While it is known that prime substrate oxidation is not always fully coupled to that of 2OG 7 (4,(55)(56)(57)(58), the results support previous work implying that both ANS and FLS possess a broad substrate selectivity in vitro (21,26,28,29,49) and provide some new insights.
Incubation of LCDs with ANS and FLS-ANS (A. thaliana) catalyzes the in vitro oxidative decarboxylation of 2OG at a greater rate in the presence of (2R,3S,4R)-trans-LCD 13 than with its natural substrate, (2R,3S,4S)-cis-LCD 6 (Table I) (50). As described above, it has been suggested that in vitro the initial product of oxidation of (2R,3S,4S)-cis-  Table I).
Incubation of DHQ Stereoisomers with ANS and FLS-With DHQ as a substrate, both ANS and FLS (both A. thaliana) showed a preference for the "natural" (2R,3R)-trans 8 stereochemistry (Tables I and II). Both 2OG 7 turnover and HPLC assays imply that racemic cis-DHQ is a poor substrate for ANS

TABLE V The percentage of incorporation of 18 O label into the products of ANS reactions with leucoanthocyanidin substrates
The absolute chirality of the cis-and trans-DHQ and the exact site of 18   leads to products including (2S,3S)-trans-DHK 21, a compound with "unnatural" C-2 and C-3 stereochemistries (28, 29) (Fig.  3E). In comparison, FNS I (P. crispum) and FHT (P. hybrida) demonstrate a high C-2 selectivity with (2R)-naringenin 20 not being accepted as a substrate (24,29). This suggests that the ANS/FLS subfamily may select for substrates with a ␤-face C-3 hydroxyl group. In comparison, the presence of the C-2 ␣-face B ring is relatively more important for FHT/FNS I. The higher C-2 stereoselectivity of FHT/FNS I is perhaps unsurprising as unlike ANS/FLS their natural substrates (2S)-flavanones (e.g.

4) do not possess a C-3 hydroxyl group.
Differential Substrate Selectivities of ANS/FLS and FHT/ FNS I-The C-3 hydroxylation of naringenin 4/20 substrates by ANS, FLS, and FHT suggests that oxidation of both natural and unnatural substrates by these enzymes occurs via initial C-3 oxidation. The high sequence similarity between FNS I and FHT (both P. crispum) (94% sequence homology) also supports the hypothesis that FNS I is a C-3 oxygenase. However, recent work showed that an enzyme-free C-3-hydroxylated intermediate/product was not detected during FNS I-mediated oxidation of its natural substrate (2S)-naringenin 4 (26). Further incubation of a C-3-hydroxylated flavanone (i.e. a dihydroflavonol, e.g. 18) with FNS I did not result in enzyme-mediated dehydration. However, this evidence does not rule out the possibility of an ordered sequential mechanism in which dehydration can only occur subsequent to formation of a dihydroflavonol intermediate in the active site. Consistent with this, we propose two possible mechanisms for FNS I action (Fig. 4), both proceeding by initial abstraction of the ␤-face C-3 hydrogen. There are then two possibilities, either a non-concerted divincial C-2,C-3 desaturation (path A) or a sequential ␤-face hydroxylation and dehydration (path B).
Substrate Assays under 18 18 O labeling studies with prokaryotic 2OG oxygenases have demonstrated that during hydroxylation reactions less than stoichiometric incorporation of oxygen into the hydroxyl group can occur (59,60). This is thought to be due to solvent exchange of an iron-oxygen intermediate, mammalian type I prolyl-4-hydroxylase (62), and hypoxic inducible factor hydroxylases (63,64), show that Ͼ90% input of oxygen from dioxygen occurs on hydroxylation of their substrates. This contrasts with results for eukaryotic lysyl hydroxylase where Ͻ10% of 18 O was incorporated into the product (65) (data reviewed in Mehn et al. (66)).

O Environments and Mechanistic Insights into the 2OG-dependent Oxygenases of Flavonoid Biosynthesis-
Assays with ANS, FLS, and FHT were carried out in labeled oxygen environments of 18 (Tables III, IV, and V) revealed several trends, implying certain common mechanistic processes.
(i) C-3 hydroxylation of naringenin 4/20 by ANS and FLS (both A. thaliana) and FHT (P. hybrida) revealed high levels of 18 O incorporation from incubation in an 18 O 2 atmosphere (Table III). Only very low levels of 18 O label incorporation were observed in incubations in a 16 O 2 / 18 OH 2 environment. This indicates that the oxygen of the introduced C-3 hydroxyl group is derived from dioxygen (Fig. 5), and at least in the case of naringenin 4/20 oxidation, the ferryl oxygen 11 (or other intermediate) does not readily exchange with water during catalysis.
(ii) On incubation of ANS/FLS (A. thaliana or P. hybrida) with trans-DHQ 8/15 or LCD 6/13 substrates under an 18 O 2 / 16 OH 2 atmosphere no 18 O label was incorporated into the products (Tables IV and V). If, as previously proposed, C-3 oxidation is the initial step in catalysis (48,49), together with (i), this is consistent with a mechanism in which the initially introduced hydroxyl group is subsequently lost.
(iii) Incubation of ANS/FLS (A. thaliana or P. hybrida) with trans-DHQ 8/15 or LCD 6/13 substrates in 18 OH 2 environments led to some 18 O incorporation into enzymatic products. Control experiments implied that the incorporation was not into the phenolic OH groups of the products and that nonenzymatic oxygen exchange does not occur at either the oxygen of the C-3 hydroxyl group or C-4 ketone of the quercetin 9 product on the time scale of the incubation. Together with (i) and (ii), these observations imply that introduction of the 18 O label from 18 OH 2 occurs at an intermediate stage following initial substrate oxidation. Since 2OG oxygenases commonly catalyze hydroxylations and both ANS and FLS can catalyze the C-3 hydroxylation of naringenin 4/20, it seems likely that ANS and FLS react with leucoanthocyanidin (e.g. 6) and dihydroflavonol (e.g. 8) substrates by initial C-3 oxidation. Two pieces of evidence suggest that, as with the eukaryotic oxygenases, the ferryl 11 (or other reactive oxygen species) does not exchange during catalysis by the flavonoid oxygenases ANS, FLS, and FHT: first, the high level of incorporation of 18 O from 18 O 2 gas into the products of incubations with naringenin 4/20 (Table III), and second, the observation that where exchange does occur, different levels of incorporation are seen in 16 (Tables IV and V). Equilibration between 3-one 22 or 3,3-gem-diol 23 intermediates provides a mechanism of oxygen exchange after initial oxidation, rationalizing the incorporation of some 18 O label from 18 OH 2 into products derived from substrates with a C-3 hydroxyl group (Fig. 6).
For oxidation of trans-DHQ 8/15 and leucoanthocyanidin (e.g. 6) substrates in a 18 O 2 / 16 OH 2 atmosphere by ANS/FLS, the lack of 18 O label in the products obtained is consistent with initial C-3 hydroxylation, providing the introduced C-3 hydroxyl is stereospecifically removed in a subsequent step. Since both (2R,3S,4S)-cis-LCD 6 and (2R,3R)-trans-DHQ 8 possess a ␤-face hydroxyl group, the ␤-face C-3 hydroxyl of the 3,3-gem- diol 23 intermediate will be derived from the substrate, while the ␣-face hydroxyl will be derived from dioxygen (assuming no inversion at C-3) (Fig. 6). Four pieces of evidence suggest that the ␣-face hydroxyl group might be coordinated, or in close proximity, to the active site iron. (i) Structural work implies that the C-3 ␣-face hydroxyl of a 3,3-gem-diol 23 is likely to lie in close proximity to the active site iron, which may function as a Lewis acid promoting its stereospecific removal (50). (ii) The ␣-face hydroxyl will have been delivered from the ferryl 11, therefore likely to be proximate to the ferryl. (iii) Incubation of ANS and FLS with (2S)-naringenin 4 produces predominantly (2R,3S)-cis-DHK 18 (i.e. ␣-face C-3 hydroxyl) (26,28,49). (iv) The formation of the thermodynamically less stable (2R,3S)cis-DHQ 14 diastereoisomer in incubations of ANS with (2R,3S,4R)-trans-LCD 13 substrates suggests that the active site is orientated so that complexation of an ␣-face hydroxyl group to the active site Fe(II) is favored (21,49).
Coordination of the ␣-face hydroxyl group to the active site iron, which could act as a Lewis acid, will favor its stereospecific removal (Fig. 6). In an 18 Tables IV and V). However, the 3-one 22 intermediate could be rehydrated from the ␤-face with assistance from the ␣-face-located iron (or ironbound water) re-forming a 3,3-gem-diol 23 (Fig. 6). In a 16 O 2 / 18 OH 2 environment this process would lead to the introduction of 18 O label at the ␤-position of C-3. Subsequent stereospecific loss of ␣-face hydroxyl, similar to that described above, could lead to the partially 18 O-labeled products observed in the case of all DHQ (e.g. 8) and LCD (e.g. 6) isomers tested in the presence of ANS/FLS (Tables IV and V Fate of the 3,3-gem-Diol and 3-One Intermediates Formed by Substrate Oxidation-The presence and stereochemistry of substituents at the C-2 and C-4 positions of the 3,3-gem-diol 23 and 3-one 22 intermediates affects product selectivity. The 3,3-gem-diol 24 and 3-one 25 formed on ANS/FLS-mediated oxidation of (2R,3R)-trans-DHQ 8 can only dehydrate/tautomerize via loss of its C-2 hydrogen to form the observed quercetin 9 product (Fig. 7). However, for LCD substrates dehydration/tautomerization of the equivalent intermediates can involve either the C-2 or the C-4 hydrogen.
It has been proposed that 3,3-gem-diol 16 and 3-one 26 intermediates formed on ANS-mediated oxidation of (2R,3S,4S)-cis-LCD 6 undergo dehydration/tautomerization involving loss of the pseudo-axial C-2 hydrogen, the removal of which is kinetically favored rather than the pseudo-equatorial C-4 hydrogen (50). This can lead to a (4S)-flav-2-en-3,4-diol 17, which is trapped in the active site and undergoes a second cycle of oxidation giving quercetin 9 (Fig. 8). In an 18 OH 2 environment there was complete incorporation of label into the quercetin product 9 (Table V). This suggests that the (4S)-flav-2-en-3,4diol intermediate 17 could have a relatively long lifetime in the active site leading to a full exchange with 18 OH 2 . The second oxidative cycle could occur by either C-3 or C-4 oxidation (Fig.  8). No dual incorporation of 18 O was observed in the quercetin 9 oxidation product of (2R,3S,4S)-cis-LCD 6; this suggests that if oxidation at C-4 does occur, it does so without exchange.
In summary, the available results are consistent with the mechanisms for all four of the enzymes ANS/FLS and FHT/ FNS proceeding via initial substrate oxidation at C-3, although in some cases oxidation at C-4 or concerted C-2,C-3 oxidation cannot be ruled out. The results also support the proposal that ANS/FLS and FHT/FNS form pairs of flavonoid oxygenases distinct both in terms of their substrate and stereoselectivities. This conclusion reflects the observation that ANS/FLS are closely related by sequence but are more distantly related to FHT. The results presented here and elsewhere (26,28,29,48,49) also reveal that the ANS/FLS and FHT/FNS pairs may be classified as ␣and ␤-face-selective oxygenases, respectively. The 18 O labeling experiments imply that this is true even when the products do not possess a chiral center at C-3, e.g. ANSmediated production of quercetin 9 from (2R,3S,4S)-cis-LCD 6.
The results also demonstrate that, even when oxygen exchange between dioxygen and water does occur, this is not via exchange of an iron-bound reactive oxidizing species but most likely via C-3 ketone 22/gem-diol 23 equilibration subsequent to the oxidation of the flavonoid C ring (Fig. 6). The lack of significant exchange of the ferryl intermediate renders ANS, FLS, and FHT similar to most but not all eukaryotic 2OG oxygenases (59 -66).