Identification of a Bifunctional Maize C- and O-Glucosyltransferase*

Background: Plant UDP-glycosyltransferases add sugars to acceptors like flavonoids either via hydroxyls (O-linkage) or carbons (C-linkage). Results: A maize glycosyltransferase produces both flavonoid C-glycosides and O-glycosides. Conclusion: This is the first description of a bifunctional C-/O-glycosyltransferase with a dual role in nature. Significance: This enzyme might be involved both in the biosynthesis of the natural insecticide maysin and in the formation of O-glycosides. Flavonoids accumulate in plant vacuoles usually as O-glycosylated derivatives, but several species can also synthesize flavonoid C-glycosides. Recently, we demonstrated that a flavanone 2-hydroxylase (ZmF2H1, CYP93G5) converts flavanones to the corresponding 2-hydroxy derivatives, which are expected to serve as substrates for C-glycosylation. Here, we isolated a cDNA encoding a UDP-dependent glycosyltransferase (UGT708A6), and its activity was characterized by in vitro and in vivo bioconversion assays. In vitro assays using 2-hydroxyflavanones as substrates and in vivo activity assays in yeast co-expressing ZmF2H1 and UGT708A6 show the formation of the flavones C-glycosides. UGT708A6 can also O-glycosylate flavanones in bioconversion assays in Escherichia coli as well as by in vitro assays with the purified recombinant protein. Thus, UGT708A6 is a bifunctional glycosyltransferase that can produce both C- and O-glycosidated flavonoids, a property not previously described for any other glycosyltransferase.


Flavonoids accumulate in plant vacuoles usually as O-glycosylated
derivatives, but several species can also synthesize flavonoid C-glycosides. Recently, we demonstrated that a flavanone 2-hydroxylase (ZmF2H1, CYP93G5) converts flavanones to the corresponding 2-hydroxy derivatives, which are expected to serve as substrates for C-glycosylation. Here, we isolated a cDNA encoding a UDP-dependent glycosyltransferase (UGT708A6), and its activity was characterized by in vitro and in vivo bioconversion assays. In vitro assays using 2-hydroxyflavanones as substrates and in vivo activity assays in yeast co-expressing ZmF2H1 and UGT708A6 show the formation of the flavones C-glycosides. UGT708A6 can also O-glycosylate flavanones in bioconversion assays in Escherichia coli as well as by in vitro assays with the purified recombinant protein. Thus, UGT708A6 is a bifunctional glycosyltransferase that can produce both C-and O-glycosidated flavonoids, a property not previously described for any other glycosyltransferase.
Glycosyltransferases are enzymes that catalyze the transfer of a sugar moiety to an acceptor molecule. The glycosyltransferases that use uridine diphosphate (UDP) sugar molecules as donors are referred to as UDP-dependent glycosyltransferases (UGTs), 4 and they are members of glycosyltransferase family 1 (1,2). This family contains most plant UGTs, which utilize different small molecules derived from specialized metabolisms as acceptors, such as terpenoids, flavonoids, saponins, plant hormones, and xenobiotics (2). Thus, plant UGTs are involved in different cellular processes that include specialized metabolism, modification of plant hormones, detoxification of xenobiotics, and plant-pathogen interactions. The glycosylation of specialized metabolites, such as flavonoids, affect their properties, enhancing their stability and solubility, and are believed to be important for the compartmentalization, storage, and biological activity of many specialized metabolites (3)(4)(5)(6)(7)(8). Flavonoids are classified in six major subgroups, chalcones, flavones, flavonols, flavandiols, anthocyanins, and proanthocyanidins or condensed tannins, and a few species also produce aurones, isoflavonoids, 3-deoxyanthocyanins, and phlobaphenes (9). In general, plants accumulate flavonoids in vacuoles as O-glycoside derivates; however, bryophytes, ferns, gymnosperms, and several angiosperms also produce flavonoid C-glycosides (10,11). In particular, cereals like wheat, rice, and maize mainly accumulate C-glycosyl flavones that are involved in protection against UV-B radiation and defense against pathogens (12)(13)(14). For example, maysin, the C-glycosyl flavone predominant in silk tissues of some maize varieties, is a natural insecticide against the corn earworm Helicoverpa zea (15,16), whereas C-glycosyl flavonoids identified in cucumber leaves act as phytoalexins in defense against powdery mildew fungi (17,18). From another perspective, there is an increasing interest for C-glycosyl flavones because of their benefits for human health and their possible applications in the prevention of diverse diseases (19,20). For example, C-glycosyl flavones inhibit pancreatic lipases, allowing their applications as chemopreventive compounds against obesity (21). In addition, because of their potential antioxidant properties, they are commonly used as nutraceutical components in the human diet (22,23).
Although the early metabolic steps resulting in flavanone formation and the branching point for the formation of different classes of flavonoids are well characterized in plants (24), the genes involved in the biosynthesis of glycosyl flavones in maize have not yet been fully identified (16). We have previously demonstrated that a flavanone 2-hydroxylase (ZmF2H1), CYP93G5, converts flavanones into the corresponding 2-hydroxyflavanones (25), which are proposed to serve as substrates for C-glycosylation, followed by dehydration as has been described in other grasses (9,26,27). However, the specific enzyme responsible for C-glycosylating 2-hydroxyflavanones in maize remains unknown. Thus, the aim of this study was to identify a C-glycosyltransferase involved in the formation of C-glycosyl flavones in maize. Here, we show that UGT708A6 is a C-glycosyltransferase that can catalyze the addition of a glucose molecule to 2-hydroxyflavanones, generating C-glycosyl flavones. Surprisingly, UGT708A6 can also accept flavanones as substrates to form O-glycosidated products. These dual activities were confirmed by both in vivo bioconversion assays and in vitro assays with the recombinant protein, revealing that UGT708A6 is a bifunctional enzyme with the ability to form both C-glycoside and O-glycoside derivatives using as acceptors 2-hydroxyflavanones and flavanones, respectively.

EXPERIMENTAL PROCEDURES
Plant Material, Growth Conditions, and Chemicals-B73 seeds were obtained from the Instituto Nacional de Tecnología Agropecuaria (Pergamino, Buenos Aires, Argentina). Maize plants were grown in greenhouse conditions with supplemental visible lighting to 1000 microeinstein m Ϫ2 s Ϫ1 with 15 h of light and 9 h of dark. Samples were collected from hypocotyls, radicles (3-day-old plants), anthers, roots (21-day-old plants), seedlings (7-day-old plants), and juvenile leaves (21-day-old plants). Flavonoid standards and UDP-glucose were purchased from Sigma-Aldrich and Indofine Chemical Co. (New Orleans, LA).
Cloning and Expression of ZmUGTs and Purification of UGT7-08A6s-A full-length cDNA corresponding to GRMZM2-G162783 (UGT708A6) was amplified by PCR using the primers UGT708A6-NdeI-forward and UGT708A6-Not-reverse harboring the NdeI and NotI restriction sites, respectively, for further cloning. PCRs were performed with GoTaq (Promega) and Pfu polymerases (Invitrogen) (10:1) using 1ϫ buffer, 2 mM MgCl 2 , 0.5 M each primer, 0.5 mM each dNTP, 0.5 unit of enzyme, and cDNA from B73 leaves in a 25-l final volume under the following cycling conditions: 2-min denaturation at 94°C and 35 cycles at 94°C for 20 s, 60°C for 30 s, and 72°C for 120 s followed by 7 min at 72°C. Primers for cDNA were designed based on the sequence provided by the maize genome sequence (MaizeSequence, release 5b.60, GRMZM2G162783). The PCR product was purified from the gel, cloned in pGEMTeasy vector (Promega), and sequenced. The pGEMT-UGT708A6 construct was digested with the corresponding restriction enzymes NdeI and NotI, and the insert was purified and cloned in pET28a vector generating the construct pET28-UGT708A6. Full-length cDNAs corresponding to GRMZM2-G162755 (UGT708A5), GRMZM2G063550 (UGT707A8), and GRMZM2G180283 (UGT91L1) were obtained from Arizona Genomics Institute (Tucson, AZ). ZmUGTs were amplified from the bacterial artificial chromosome clones by PCR using the primers described in supplemental Table 1 for further cloning in pET28 vector. PCRs were performed as described above for UGT708A6. The PCR products were purified from the gels, digested with the corresponding restriction enzymes, purified, cloned into pET28 vector, and sequenced. BL21(DE3) cells with the chaperone expression plasmid pGRO (28) were transformed with the construct pET28-ZmUGTs and the empty vector pET28. Cell cultures (200 ml of LB medium containing 30 mg liter Ϫ1 kanamycin and 35 mg liter Ϫ1 chloramphenicol) were grown at 37°C until A 600 reached 0.4, and L-arabinose (2 mg ml Ϫ1 ) was added to induce chaperone proteins. The cultures were grown at 37°C to midlog phase (A 600 0.5-0.6), and recombinant N-terminal His 6 -ZmUGTs expression was achieved by induction with 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside for 20 h at 22°C.
For the purification of UGT708A6, cells were harvested by centrifugation at 3000 ϫ g for 20 min at 4°C. The pellet was resuspended in binding buffer (50 mM sodium phosphate, pH 7.5, 500 mM NaCl, 20 mM imidazole, 5% glycerol) containing 0.1% Tween 20, 1 mM phenylmethylsulfonyl fluoride, and Complete EDTA-free protease inhibitor mixture (Roche Applied Science). Cells were disrupted by sonication and then centrifuged at 12,000 ϫ g for 20 min at 4°C to obtain soluble cell extracts. The protein was bound to a nickel-nitrilotriacetic acid resin (Invitrogen) by rocking at 4°C for 1 h, and then the resin was loaded onto a column, washed three times with 15 volumes of binding buffer followed by three washes with 7 volumes of washing buffer (50 mM sodium phosphate, pH 7.5, 500 mM NaCl, 5% glycerol, 40 mM imidazole). Elution was carried out by five sequential additions of 1 ml of elution buffer (50 mM sodium phosphate, pH 7.5, 500 mM NaCl, 5% glycerol, 200 mM imidazole). Finally, the recombinant protein was desalted in desalting buffer (25 mM Hepes-NaOH, pH 7.5, 10 mM 2-mercaptoethanol, 5% glycerol) by four cycles of concentration and dilution using Amicon Ultra-15 30,000 (Millipore) and stored at Ϫ80°C. The protein level was estimated both by comparison with dilution series of bovine serum albumin on a Coomassie Blue-stained SDS-polyacrylamide gel and by using the Bradford reagent (Bio-Rad; Ref. 29). The yield of 90% pure recombinant protein obtained in these conditions was 6 mg liter Ϫ1 of culture.
To express each ZmUGT in yeast, the full-length cDNAs were amplified by PCR using primers harboring restriction sites (supplemental Table 1) and each pET28-ZmUGT construct as templates. The PCR product was purified, digested with the corresponding enzymes, and cloned in p5AX43 vector generating the plasmids p5AX43-ZmUGTs: p5AX43-UGT708A5, p5AX43-UGT707A8, p5AX43-UGT91L1, and p5AX43-UGT708A6. The p5AX43 vector corresponds to a modified version of plasmid YEplac181 (30) in which the glyceraldehyde-3-phosphate dehydrogenase promoter was inserted at the HindIII site. The p5AX43-ZmUGT plasmids and p5AX43 empty vector were transformed into competent WAT11 (31) yeast cells harboring pGZ25-ZmF2H1 or pGZ25 empty vector (25), respectively, following the Trafo protocol (32). Yeast colonies harboring the plasmids were selected by growth on synthetic complete medium (SC) agar plates lacking uracil, tryptophan, and leucine (SC UraϪ TrpϪ LeuϪ).
Bioconversion Experiments-For in vivo yeast activity assays, an individual recombinant yeast colony was grown for 40 h at 30°C in 5 ml of liquid SC UraϪ TrpϪ LeuϪ medium containing 2% (w/v) glucose. Then an aliquot of this culture corresponding to an A 600 of 1.0 was collected by centrifugation, washed in sterile water, and used to seed 5 ml of induction medium, SC UraϪ TrpϪ LeuϪ containing 2% (w/v) galactose and 3% (v/v) glycerol. The flavonoid substrates were then added to a final concentration of 40 g ml Ϫ1 . After incubation for 48 h at 30°C, flavonoids were extracted with ethyl acetate from 1-ml culture aliquots by adding 500 l of ethyl acetate and vortexing for 1 min. Solvent layers were separated by centrifugation at 13,000 rpm for 1 min, and flavonoids (both the aglycones and the glycosides) were recuperated in the organic layer. The organic layer was then twice re-extracted with 500 l of ethyl acetate, and the organic layers were combined. The organic phase was dried in a SpeedVac and resuspended in methanol for subsequent liquid chromatography-mass spectrometry (LC-MS) analysis.
For in vivo E. coli activity assays, BL21(DE3) cells harboring pGRO (for expression of GroEL-GroES chaperone complex) and pET28-ZmUGTs or empty pET28a plasmids were grown at 37°C in LB with appropriate antibiotics. Expression of chaperones and UGT proteins was induced by the addition of L-arabinose and 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside, respectively, as described above, and cultures were simultaneously supplemented with 40 g ml Ϫ1 flavonoids. Cultures were grown at 22°C for 24 -48 h and then centrifuged at 15,000 ϫ g for 5 min. One-milliliter medium aliquots were extracted with ethyl acetate as described above, vacuum-dried, and resuspended in methanol for subsequent LC-MS analysis.
Acid hydrolysis was performed to differentiate between Oand C-glycosylated products as an acidic treatment hydrolyzes O-glycosidic linkages, whereas C-linked conjugates are stable to this treatment. After extraction with ethyl acetate, an equal volume of 2 N HCl was added to the samples followed by incubation at 90°C for 1 h. One volume of 100% methanol was added to prevent the precipitation of aglycones.
In Vitro UGT708A6 Activity Assays-The reaction mixture contained 50 mM Hepes-NaOH, pH 7.5, 10 mM 2-mercaptoethanol, 100 g ml Ϫ1 flavonoid substrates, 2 mM UDP-glucose, and 5 g of recombinant purified protein in a final volume of 100 l. Reactions were initiated by the addition of the enzyme and terminated by extraction with ethyl acetate. Activity assays were performed at 30°C for up to 60 min.
Glycoside Product Analyses by LC-MS-Reaction products were analyzed by LC-MS using a system consisting of an Agilent 1100 high-performance liquid chromatography pump, and a Bruker micrOTOF-Q II mass spectrometer in a positive-ion mode configured with a Turbo-ion spray source setting collision energy 25 eV. Samples (10 l) were chromatographed on a Phenomenex Hypersil GOLD C18 (3 m; 2.0 by 150 mm) at 200 l/min with a linear gradient from 20% MeCN to 100% in 0.1% formic acid over 10 min. The eluate was delivered unsplit into the mass spectrometer source. Compounds were identified by comparison of mass spectra to those of authentic commercial standards (Sigma-Aldrich and Indofine Chemical Company). Absorbance units were detected at 295 and 360 nm.
Gene Expression Analyzes by RT-(Quantitative (q)) PCR-Tissues from three independent biological replicates were frozen in liquid nitrogen and stored at Ϫ80°C. Total RNA was extracted following the Trizol protocol (Invitrogen) followed by DNase treatment (Promega). cDNAs were synthesized from 4 g of total RNA using Superscript Reverse Transcription Enzyme II (Invitrogen) with oligo(dT) as a primer. The resulting cDNAs were used as templates for qPCR in a iCycler iQ detection system with the Optical System Software version 3.0a (Bio-Rad) using the intercalation dye SYBR Green I (Invitrogen) as a fluorescent reporter and Platinum Taq polymerase (Invitrogen). Primers were designed to generate unique 150 -250-bp fragments using PRIMER3 software (33). Three biological replicates were used for each sample plus a negative control (reaction without reverse transcriptase). To normalize the data, primers for actin1 (J01238) were used (supplemental Table 1). Amplification conditions were as follows: 2-min denaturation at 94°C and 40 -45 cycles at 94°C for 10 s, 57°C for 15 s, and 72°C for 20 s followed by 5 min at 72°C. Melting curves for each PCR product were determined by measuring the decrease of fluorescence with increasing temperature (from 65 to 95°C). To confirm the size of the PCR products and to check that they corresponded to a unique and expected PCR product, the final PCR products were separated on a 2% (w/v) agarose gel, stained with SYBR Green (Invitrogen), and sequenced. Primers used for UGT708A6 are listed in supplemental Table 1 (UGT708A6-RT-forward and UGT708A6-RT-reverse).
Extraction of Total Flavonoid from Maize Silks-Flavonoid extraction was performed as described previously (12). Fresh silks and 25-day-after pollination pericarps were rinsed with water and lyophilized for 1 day. Dry weight was measured, and the sample was ground to a powder with a mortar and pestle. The powder was extracted for 8 h with 12 volumes of acidic methanol (1% (v/v) HCl in methanol) followed by a second extraction with 12 volumes of chloroform and 6 volumes of distilled water. The extracts were vortexed and centrifuged for 2 min at 3000 ϫ g, and organic phases were collected. Flavonoid extracts were analyzed by LC-MS/MS.
Phylogenetic Analysis-The tree was constructed using MEGA 4.0 software with the neighbor joining method based on ClustalW multiple alignments (34).
Computational Analyses from High Throughput Available Data-The heat map was generated with all the gene models with the glycosyltransferase domain (IPR002213) present on the maize genome (version 5b.60) using bronze1 (GRMZM2G165390) as a model. These gene models were further used to generate a list to cross-reference to data publicly available from Morohashi et al. (25) (P1-rr and P1-ww pericarps and silks) and from publicly available data sets (root, shoot, and leaf from the B73 inbred line) RNA sequencing results (35). These data were further used to generate a heat map on the MeV Multiple Array Viewer (36).

RESULTS
Expression and Phylogenetic Analysis of ZmUGTs-To determine a putative candidate for C-glycosylation reaction of flavonoids in maize, we followed two criteria. First, we evaluated how genes of candidates were expressed in different maize tissues and whether they are regulated by the P1 transcription factor, extensively known to be involved in the regulation of C-glycosyl flavone biosynthesis (16,(37)(38)(39). Therefore, we built a list of 157 putative UGTs in maize using bronze1 (GRMZM2G165390), one of the best studied maize UGTs and one involved in anthocyanin biosynthesis (40,41), as a starting point. We next intersected this list with RNA sequencing data publicly available from maize leaves, shoots, and roots from the B73 inbred line and RNA sequencing data from silks and pericarps with contrasting P1 alleles in the common A619 genetic background and referred to here as P1-rr and P1-ww (supplemental Fig. 1) (25,42). From these results, we selected four genes that were highly up-regulated in P1-rr compared with P1-ww pericarps: UGT708A5, UGT91L1, UGT707A8, and UGT708A6. These candidate ZmUGTs contain the characteristic plant secondary product glycosyltransferase motif characteristic of plant UGTs with 10 conserved amino acids proposed to be involved in the interaction with the UDP-sugar molecule (Fig. 1).
Our second criterion was that any gene model taken into consideration would have sequence similarity with previously characterized UGTs capable of performing C-glycosyl bond formation, such as the rice C-glycosyltransferase (10) (Fig. 1). With this, we generated a phylogenetic tree with selected UGTs that use mainly flavonoids as substrate acceptors. The tree shows five well defined clusters characterized by the regioselectivity of some of these enzymes (Fig. 2). Enzymes in cluster 1 transfer UDP-sugars onto the 7-hydroxyl group of their sub-strates; cluster 2 includes UGTs that utilize flavonoid glycosides as acceptors and catalyze the formation of sugar-O-sugar links. Clusters 3 and 4 are constituted by UGTs that transfer sugars onto the 3-and 5-hydroxyl groups of the acceptors, respectively. Finally, cluster 5 includes members characterized by having a broad plasticity in the position of glycosylation (3Ј-, 3-, and 7-hydroxyl groups) and by the formation of more than one glycoside product. From this analysis, we placed UGT91L1 in cluster 2, which includes UGTs that utilize flavonoid glycosides as acceptors and catalyze the formation of sugar-O-sugar links like Ph1-6RhaT from Petunia hybrida that adds rhamnose to the 6-O-glucose of anthocyanidin (43). UGT708A5, UGT707A8, and UGT708A6 were included in cluster 5 as well. It is important to take into consideration that phylogenetically distant UGTs can have similar substrate specificity, whereas evolutionary close UGTs may accept different substrates and that the selectivity for acceptors cannot be inferred only by the similarity in their primary sequences (1,2). Interestingly, UGT708A6, included in cluster 5 together with the C-glycosyltransferase from Oryza sativa (OsCGT) and the bifunctional N-and O-glycosyltransferase from Arabidopsis thaliana, UGT72B1 (10,44), shows the highest identity (67%) to OsCGT, a rice UDP-glucosyltransferase that uses 2-hydroxyflavanones as flavonoid acceptors (10). Thus, we predict that UGT708A6 is among the best candidates to catalyze the C-glycosylation reaction in the C-glycosyl flavone biosynthetic pathway because it is up-regulated in P1-rr tissues and has the highest identity to a previously described CGT.
Bioconversion Assays in E. coli and Yeast-To evaluate whether UGT708A6 or any of the other of the selected ZmUGTs are involved in the C-glycosyl flavone pathway catalyzing the reaction that follows that of ZmF2H1 as it was described in rice, the full open reading frames of each UGT were cloned in the pET28a vector, and the proteins were expressed in E. coli as N-terminal fusion proteins with a His 6 tag as described under "Experimental Procedures." Glycosyltransferase activity was assayed in vivo by feeding 2-hydroxynaringenin as a flavonoid acceptor to E. coli cultures expressing each of the ZmUGTs. After a 2-day fermentation assay, flavonoids were extracted with ethyl acetate, and products were analyzed by LC-MS. Of all the glycosyltransferases tested (UGT708A5, UGT707A8, UGT91L1, and UGT708A6), only UGT708A6 was able to produce a compound (1) that was identified as apigenin 6-C-glucoside (isovitexin) by comparison with an isovitexin standard using LC-MS/MS (Fig. 3, A and C). The negative control, E. coli containing the empty vector, did not show production of this compound (Fig. 3A).
To verify the ability of UGT708A6 to convert 2-hydroxynaringenin to isovitexin, we took advantage of a yeast strain that we had previously generated that expresses the A. thaliana cytochrome P450 reductase and ZmF2H1, accumulating small amounts of 2-hydroxynaringenin when fed with naringenin (25). Thus, yeast cultures expressing both ZmF2H1 and one of the ZmUGTs or harboring the corresponding combination of empty vectors were supplied with the flavanones naringenin or eriodictyol as substrates, and the glycoside products were analyzed by LC-MS. In these combinatorial assays, only when UGT708A6 was expressed along with ZmF2H1 were the 6-C- WAPQIQVLSHDAVGVVITHGGWNSVVESIAAGVPVICRPFFGDH * * :* : : . : * *: * * . : * :: glucosyl derivatives of the respective flavones, isovitexin and isoorientin, identified as products (1 and 2) as compared with the respective standards by LC-MS/MS (Fig. 3, B-F). These compounds show the characteristic fragment ions of the C-glycoside moiety, [M ϩ H Ϫ 90] and [M ϩ H Ϫ 120] (Fig. 3, D and  H). Furthermore, the formation of the C-glucoside products was verified due to the stability of these compounds under acid hydrolysis (10, 26) (not shown). In addition to isoorientin (luteolin 6-C-glucoside), another reaction product with an m/z of 449.1 and different retention time was observed (3). Further analysis of the relative intensity of the product ion found by positive electrospray ionization (LC-MS/MS) allowed validation of reaction product 3 as orientin (luteolin 8-C-glucoside) (Fig. 3H) (45).
Previous experiments showed that a yeast dehydratase activity was responsible for converting 2-hydroxyflavanones into the corresponding flavones (46). To verify that the flavones generated by dehydration from the 2-hydroxyflavanones are not the actual substrate acceptors for the UGT708A6 C-glycosyltransferase activity, flavones (apigenin and luteolin) were fed to yeast cultures expressing only UGT708A6; however, no glycosylation products were detected. In addition, to verify the specificity of UGT708A6, different flavonoids were fed to E. coli cultures expressing this enzyme. No glycoside product was detected when flavonols (quercetin and kaempferol), flavones (apigenin, luteolin, and chrysin), and anthocyanidins (cyanidin) were used as substrates. However, when E. coli cultures were fed with the flavanones naringenin and eriodictyol as substrates, production   (Fig. 4, A and D). Interestingly, the fragmentation patterns of these new glycoside derivatives showed the typical neutral loss of 162 (transition 435.1 3 273.1 for naringenin and 451.1 3 289.1 for eriodictyol, respectively) corresponding to a hexose residue in a flavonoid O-glycoside (Fig. 4, C, F, and  G). These results were confirmed by acid hydrolysis (not shown). Finally, the O-glycoside flavonoid products were identified as naringenin 7-O-glucoside (4) and eriodictyol 7-Oglucoside (5) as compared with the respective standards by LC-MS/MS (Fig. 4). Hence, the results described for the bioconversion assays in E. coli and yeast show that UGT708A6 is a novel enzyme able not only to C-glucosylate 2-hydroxyflavanones but also to O-glucosylate flavanones.
In Vitro Activities of the Recombinant Purified UGT708A6 Protein-To verify that UGT708A6 is a glucosyltransferase able to produce both O-and C-glucosyl products as shown in the bioconversion experiments in E. coli and yeast, we purified the recombinant protein expressed in E. coli to perform in vitro activity assays (Fig. 5A). When the recombinant UGT708A6 was assayed using the flavanones naringenin and eriodictyol as acceptors and UDP-glucose as a donor, products corresponding to the flavanone O-glucosides were detected (not shown). Similarly, as observed by in vivo assays in E. coli, when naringenin was assayed as a substrate, the formation of one naringenin O-glycoside compound was detected, whereas eriodictyol generated two O-glycosides derivatives, which could correspond to the glucose molecule bound to different -OH groups. Furthermore, the sensitivity of these compounds to acid hydrolysis confirmed that they correspond to O-glycosides.
On the other hand, when 2-hydroxynaringenin was assayed as a substrate, two reaction products with an m/z of 433.1 [M ϩ H ϩ ] were observed, one corresponding to isovitexin (apigenin 6-C-glucoside, 1) in comparison with the available standard (Figs. 5B and 3C). Analysis of the relative intensity of the product ion found by positive electrospray ionization allowed the identification of selective ions for the C8 isomer ([ 0.3 X Ϫ H 2 O Ϫ CO] ϩ and [ 0.2 X Ϫ CHO Ϫ CO] ϩ with m/z values of 297.3 and 256.4, respectively), indicating that the second reaction product (7) corresponds to vitexin (apigenin 8-C-glucoside) (45) (Fig. 5D). Together, both in vitro and in vivo bioconversion activity assays demonstrate that UGT708A6 is a bifunctional enzyme able to   catalyze both the C-glucosylation of 2-hydroxyflavanones and the O-glucosylation of flavanones. Flavonoid Glycosides in Maize Pericarps and Silks-Supplemental Fig. 1 shows that UGT708A6 is expressed in pericarps and silks, and its expression is positively regulated by P1, showing significantly higher mRNA levels in P1-rr than in P1-ww pericarps and silks (25). Thus, to correlate UGT708A6 activities with the flavonoid glycosides present in these organs, methanolic extracts of maize P1-rr pericarps and silks were analyzed by LC-MS/MS. As shown in Table 1, both C-glycosyl flavones derived from apigenin and luteolin (isoorientin and isovitexin) with the glycosylated substitutions at the C6 position were identified as was reported previously (47,48) (Table 1). Interestingly, we could identify flavanone O-glycosides (both for naringenin and eriodictyol) in accordance with the detected expression of UGT708A6 in these tissues (25). In addition, successive losses of hexoyl units were observed for naringenin O-glycosides, indicating the presence of di-O, O-hexosides. Isomers with different retention times were detected for naringenin O-glycosides that likely represent the different glycosylation positions of these compounds. Overall, metabolic profiling analysis demonstrates that this enzyme could catalyze the biosynthesis of both C-and O-glycoside products in planta.

DISCUSSION
Glycosylation is an important step in flavonoid biosynthesis that contributes to flavonoid stability, solubility, storage, and biological activity changes (3). Although flavonoid glycosides have been described in maize, for example the characterization of a glycosyltransferase involved in anthocyanin biosynthesis (bronze1), information about other glycosyltransferases implicated in flavonoid metabolism have not been reported (40,41).
Here we have characterized a maize glycosyltransferase, UGT708A6, involved in the biosynthesis of C-glycosyl flavones by in vitro and in vivo bioconversion activity assays. Previously, we have demonstrated that the first step in the formation of the C-glycosyl flavone involves the conversion of flavanones into 2-hydroxyflavanones by ZmF2H1 (CYP93G5) (25). Here, through bioconversion assays in yeast expressing ZmF2H1 with UGT708A6, we have demonstrated the formation of isovitexin and isoorientin, the 6-C-glucosyl derivatives of the flavones apigenin and luteolin, respectively. Furthermore, both in vitro activity assays with the recombinant purified UGT708A6 pro- tein and bioconversion assays in yeast showed the formation of both isomers of apigenin C-glucosides (vitexin and isovitexin) and luteolin C-glucosides (orientin and isoorientin), respec-tively. These results indicate that UGT708A6 is a C-glycosyltransferase that uses 2-hydroxyflavanones as substrates to generate C-glycosyl flavones similarly to a flavonoid C-glycosyltransferase from Fagopyrum esculentum and the rice CGT (10,26,27,49). In addition, both bioconversion assays in E. coli expressing UGT708A6 and in vitro experiments showed that UGT708A6 can also O-glucosylate the flavanones naringenin and eriodictyol, generating one and two different glucoside products, respectively. Consequently, these results show that UGT708A6 is a bifunctional enzyme that has the ability to form both C-glycoside and O-glycoside links with the flavonoid acceptors 2-hydroxyflavanones and flavanones, respectively, a property that has been only described for a modified glycosyltransferase from Streptomyces fradiae using an unnatural substrate (UrdGT2; Ref. 50). Interestingly, carbon-carbon-based and carbon-oxygen-based prenylation of a diverse collection of hydroxyl-containing aromatic acceptors like naringenin was described for bacterial prenyltransferases (51). These enzymes have a bimolecular nucleophilic substitution (S N 2)-like reaction mechanism similar to that of plant UGTs (52). The reaction involves a carbon-mediated nucleophilic attack on C1 of geranyl diphosphate with the diphosphate moiety stabilized by Mg 2ϩ coordination and the basic character of the diphosphate binding site serving as a leaving group. On the other hand, based on crystal structures and genetic evidence, plant O-glycosyltransferases contain a highly conserved histidine residue in the active site that acts as a general base to abstract a proton from the acceptor substrate. A nearby aspartate residue interacts with the histidine, forming a triad substrate-His-Asp that helps to stabilize the histidine charge after deprotonating the flavonoid substrate (53). It was proposed that the deprotonated acceptor displaces the UDP by attacking the C1 carbon center of the UDP-sugar to form the ␤-glucoside product (53). Protein sequence alignments showed that UGT708A6 has the conserved His-Asp residues corresponding to the active site of O-glycosyltransferases (Fig. 1). On the substrate site, the distribution of charges in the deprotonated phenolic structure of ring A of flavonoids can permutate between the carbon and the adjacent oxygen substituent. Thus, in a way similar to that suggested for prenyltransferases (52) and described for isopentenyl pyrophosphate transferases involved in terpene biosynthesis (54), the dual function of UGT708A6 may be explained by the phenolic character of the substrate that alternatively can mediate either the carbon or oxygen nucleophilic attack on C1 of UDP.
In rice, C-glycosyl flavone biosynthesis takes place through a pathway different from that of O-glycosyl flavone formation involving the generation of 2-hydroxyflavanones by CYP93G2 activity followed by the C-glycosylation catalyzed by OsCGT (10,26). It has been proposed that an open form of 2-hydroxyflavanones is the actual substrate for OsCGT, resulting in the formation of 2-hydroxyflavanone C-glycoside products that are further dehydrated by a dehydratase (10,26,27). However, it is important to mention that we could not detect the 2-hydroxyflavanone C-glycoside products either by in vivo or in vitro experiments. The failure to detect these intermediates in C-glycosyl flavone biosynthesis could possibly be due to spontaneous dehydration of these unstable compounds during the reaction    process (10,49,55). In addition, the relative abundances of C-glycosyl flavone isomers derived from naringenin and eriodictyol were different. The main product detected for naringenin was the flavone 6-C-glucoside (isovitexin), whereas both flavone 6-C-glucoside (isoorientin) and flavone 8-C-glucoside (orientin) were detected for eriodictyol in a ratio of 1:8 (Fig. 3E).
A similar result of the in vivo assays in yeast was obtained using 2-hydroxy naringenin in vitro, but formation of flavone 8-Cglucoside (vitexin) could also be detected in minor proportion (Fig. 5). These results could be explained by proposing that the actual substrate for the glycosyltransferase is the closed form of the 2-hydroxyflavanone (Fig. 6, compound B) as it has less structural flexibility than the open form (Fig. 6, compound C). Because the only structural differences between the two substrates are the substitutions on the B ring of the flavanone, these hydroxyl groups should be important for substrate accommodation in the active site of the enzyme, something difficult to obtain with an open-chain flavanone. Overall, the results described in this study indicate that UGT708A6 can generate C-glycosides with the glucose molecule at the C6 and C8 positions; however, only flavone 6-Cglycosides have been described in silks of maize (47,48). Taking into consideration the proposed biosynthesis pathway of the C-glycosyl flavone maysin (16), a possible explanation for this is that C6 isomer (isovitexin and isoorientin) consumption by the following rhamnosyltransferase enzyme involved in this pathway may favor the formation of this isomer. Nevertheless, we cannot rule out that flavone 8-C-glycosides are present in maize tissues not yet studied.
The R2R3-MYB P1 transcription factor regulates maysin production in silk tissues of some maize varieties (37)(38)(39). Our results show that UGT708A6, the expression of which is regulated by P1 in silks (25), generates isovitexin and isoorientin, intermediates involved in biosynthesis of apimaysin and maysin, respectively (16), suggesting that this enzyme could be involved in this biosynthetic pathway. Similarly to rice (27), when ZmF2H1 and UGT708A6 enzymes were co-expressed in yeast, the intermediate 2-hydroxyflavanones were not detected; it is also likely that UGT708A6 is not the limiting activity in the C-glycosyl flavone biosynthesis in maize. However, UGT708A6 shows a relatively constitutive expression pattern in different maize tissues (Table 2), consistent with the microarray database from a genome-wide atlas of transcription (42); consequently, this pattern of expression could allow the generation of flavanone O-glycosides in different maize tissues as well as their storage in vacuoles, preventing toxicity and increasing their stability. Nevertheless, we cannot rule out that other non-characterized glycosyltransferase enzymes are also responsible for the formation of these compounds, and it cannot be excluded that additional transcription factors could be involved in the regulation of UGT708A6 expression in maize tissues. Thus, additional studies are required to reveal the involvement of UGT708A6 in other branches of flavonoid biosynthesis besides the C-glycosyl flavone pathway.  In summary, we have identified and characterized the first occurring C-/O-glycosyltransferase, a dual role that has not yet been described for any glycosyltransferase in nature. This enzyme could be involved in the formation of the insecticidal C-glycosyl flavone maysin but can also catalyze the formation of flavanone O-glycosides. Further studies concerning the catalytic mechanism of UGT708A6 will provide useful information to be applied in genetic engineering of other glycosyltransferases to develop therapeutic compounds more stable than O-glycosides to enzymatic degradation by glycosidases.