Sweet orange trees (Citrus sinensis ‘Valencia’) were grown in the greenhouse at the National Citrus Germplasm Repository, the Citrus Research Institute (CRI) of the Chinese Academy of Agricultural Sciences (CAAS), Chongqing, China. A total of seven developmental stages were collected from the fruit-setting period, which consisted of 10 DAB (days after full blooming), 30, 60, 90, 120, 150, and 180 DAB. All fruit samples were separated into two parts: peel (also called exocarp) and pulp (called endocarp). All samples were immediately frozen in liquid nitrogen and stored at -80°C. Tobacco plants (Nicotiana tabacum) used in the transgenic experiments were grown in the growth chamber at a constant temperature (28 ± 3°C), and were exposed to a 12/12 h cycle (light/dark).

Flavonoids in sweet orange were extracted using the procedure described by Yang et al. (2016). After drying at 40°C for 48 h, the peels and pulp were powdered and filtered through a 60-mesh screen. About 0.5 g of the powdered sample was ultrasonically extracted with 7 mL methanol at 200 W for 30 min. After centrifugation at 5,000 rpm for 10 min, the resulting supernatants were collected in a 15-mL tube. The process was repeated three times. The final volume was adjusted to make a consistent volume of 25 mL of methanol. About 0.4 mL of the extract solution was diluted with 0.6 mL deionized water and filtered through a membrane with 0.2-μm pore diameter and temporarily stored at 4°C in the refrigerator.

The tobacco flavonoid compounds were extracted by the following method: about 0.2 g of the powdered samples were extracted in solution (80% methanol, 19% water, 1% hydrochloric acid) by vortexing for 20 s followed by water-bath sonication for 30 min. After centrifugation at 6,000 rpm for 10 min, the resulting supernatants were collected and placed in a fresh 5-mL tube, then extracted twice with chloroform to remove chlorophyll. The extraction process was replicated three times, and the final volume was adjusted to 5 mL. Finally, the supernatants were filtered through a membrane with 0.2-μm pore diameter and temporarily placed at 4°C in the refrigerator.

Ultraperformance liquid chromatography analyses were implemented as described by Yang et al. (2016). After separation on an ACQUITY UPLC BEH C18 column (2.1 mm × 100 mm, 1.7 mm, United Kingdom), samples were scanned by a photodiode array detector with the absorption spectrum set from 240 to 400 nm. Xevo G2-S Q-TOF (Waters MS Technologies, Manchester, United Kingdom), a quadrupole, orthogonal acceleration, time-of-flight tandem mass spectrometer was used with an electron spray ionization source. Both positive and negative ion modes were employed to ionize the chemical compounds. The detection conditions were as follows: capillary voltage at 0.45 kV, cone voltage 40 V, source temperature 100°C, desolvation temperature 400°C, cone gas flow 50 L h^-1, desolvation gas flow 600 L h^-1, low energy 6 V, and high energy ramp 20–40 V. TOF-MS was set from 100 to 1000 m/z. The scan time was 0.2 s. Data was obtained using real-time collection (scan time 0.5 s, interval 15 s). Sample UPLC-Q-TOF-MS data were collected and processed using the Waters UNIFI 1.7 software.

Total RNA was extracted with the TaKaRa MiniBEST Universal RNA Extraction Kit (Takara Bio, China). cDNA synthesis was performed using the PrimeScript RT Master Mix Perfect Real-Time Kit (Takara Bio, China). According to the genomic data from Phytozome v12.1, primers were designed to obtain the full-length cDNA and to detect the relative expression levels. All PCR primer sequences are shown in Supplementary Table S1. The relative expressions of the genes were determined according to the 2^-ΔΔCt method (Livak and Schmittgen, 2001). Based on the analysis by geNorm (Vandesompele et al., 2002), three reference genes, citrus β-actin, SDH1-1, and GAPDH, were used to normalize the expressions of the candidate genes. Three replicates of the experiments were performed for each gene.

The coding sequence of CsUGT76F1 was amplified by PCR with forward and reverse primers (Supplementary Table S1), following which the PCR product was sub-cloned into the pMAL-c2X expression vector with a maltose-tag (New England Biolabs, Ipswich, MA, United States). The recombinant plasmid was introduced into E. coli NovaBlue (DE3) competent cells (Novagen, Schwalbach am Taunus, Germany). The positive clones were identified in 5 mL of lysogeny broth with 80 mg/L ampicillin for 8–12 h at 37°C. Two milliliters of E. coli culture were transferred to 300 mL of lysogeny broth containing 80 mg/L ampicillin and shaken at 200–250 rpm until an optical density (O.D.) of 0.6 at a wavelength of A₆₀₀ was reached. Isopropyl-β-D-thiogalactopyranoside (IPTG) was employed to induce the expression of CsUGT76F1. After induction at 28°C for 48 h, the cells were precipitated by centrifugation and then disrupted by sonication for 20–25 min. The recombinant protein was purified via an amylose resin affinity chromatography system (New England Biolabs, E8201S).

Uridine diphosphate-rhamnose was synthesized using the procedure described by Shibuya et al. (2010) and Hsu et al. (2017). Full-length cDNA of the RHM2/MUM4 gene (At1g53500) from A. thaliana was cloned into the expression vector pET21a and then introduced into the BL21-CodonPlus (DE3)-RIPL. The recombinant protein was prepared according to the method reported by Shibuya et al. (2010). The 80-μL of cell extract prepared from the RHM2/MUM4-expressing cells was incubated with 5 mM UDP-glucose, and 10 mM NADPH at 37°C for 2 h. The presence of UDP-rhamnose in the reaction mixture was identified via comparisons with the data in available literature (Hsu et al., 2017), and the concentrations were calculated using UDP-glucose as a reference.

The standard reaction mixture for enzyme assay consisted 100 mM Tris-HCl (pH 7.5), 0.1% (v/v) β-mercaptoethanol, 2.5 mM sugar donors, 0.2 mM substrates, and 20 μg recombinant protein in a total volume of 50 μL. Firstly, flavonoid substrates were dissolved in dimethylsulfoxide (DMSO) to obtain a concentration of 15 mM. For substrate specificity analysis and to obtain optimal reaction conditions, flavonoid substrates dissolved in DMSO were then added to the reaction mixture with a final concentration of 0.2 mM. Although there was the possibility of slight precipitation of flavonoid aglycones occurring, such minor precipitation was believed to have no significant effect on the specificity analysis and the optimal reaction conditions that were derived from the presence or the maximum accumulation of the glycosylated flavonoids. A series of reaction mixtures containing different concentrations of the substrate (0.02, 0.04, 0.06, 0.08, 0.10, and 0.12 mM) were used to test the kinetic parameters of UGT. This selected range of substrate concentrations was far below the solubility of the flavonoid aglycones. The reaction was conducted at 30°C for 1 h and terminated by the addition of 60 μL pure methanol.

Optimal reactions were performed at varying pH and temperatures. Enzyme assays were conducted following the methods described above using different buffers. Three buffers were used to control the pH levels, including acid-sodium citrate buffer (pH 4.0–7.5), Tris-HCl buffer (pH 7.0–10.0), and NaHCO₃–Na₂CO₃ buffer (pH 9.0–11.0). Reaction temperatures ranged from 10 to 70°C, with 5°C increments. Products from the enzymatic activity assay were analyzed using the UPLC-Q-TOF-MS system described above.

The coding region of CsUGT76F1 was amplified by PCR with forward and reverse primers (Supplementary Table S1). The PCR product was introduced into the vector pDONR207 using the Gateway BP Clonase Enzyme mix (Invitrogen, United States). Subsequently, CsUGT76F1 was transferred into the expression vector pCB2004 using the Gateway LR Clonase system (Invitrogen, United States). The recombinant pCB2004-CsUGT76F1 plasmid was transferred into the Agrobacterium tumefaciens EHA105-competent cells through electroporation. The positive cells were screened on the agar-solidified medium consisting 50 mg/L spectinomycin and 50 mg/L kanamycin, and genetic transformation was implemented via the leaf disk transformation method (Maiti et al., 1988). The transgenic plants were screened on Murashige and Skoog medium containing 25 mg/L phosphinothricin, and then identified using both genomic PCR and RT-PCR. Three lines with high transcript levels were used in the flavonoid profiling analyses.

To identify the flavonoids and their derivatives, we employed UPLC-Q-TOF-MS in both positive and negative ion modes to analyze the peel and pulp extracts. The representative chromatograms of BPI corresponding to the negative and positive signals of sweet orange peels at 180 DAB are shown in Figure 2. A total of 13 flavonoids and their derivatives were detected and unambiguously characterized via comparisons with the reference standards and literature data (Abad-García et al., 2012; Yang et al., 2016) (MS data presented in Supplementary Table S2). Four flavonoid-O-glycosides in the fruit of sweet orange were identified, consisting eriocitrin (also called eriodictyol-7-O-rutinoside, peak 4), narirutin (naringenin-7-O-rutinoside, peak 5), hesperidin (hesperetin 7-O-rutinoside, peak 6), and didymin (isosakuranetin-7-O-rutinoside, peak 7). These flavonoid-O-glycosides often contain a disaccharide (rhamnosyl-α-1,6-glucose) linked to an aglycone through the C-7 hydroxyl group. In addition, two flavonoid-C-glycosides- apigenin-6,8-di-C-glucoside (peak 1) and chrysoeriol-6,8-C-glucoside (peak 3), were also identified. The polymethoxylated flavones are found almost exclusively in the Citrus species. Our results revealed the presence of seven polymethoxylated flavones in the fruit of sweet orange, consisting of isosinensetin (peak 9), sinensetin (peak 10), dihydroxy-tetramethoxyflavone nobiletin (peak 11), 5, 7, 8, 4′-tetramethoxyflavone (peak 12), 3, 5, 6, 7, 8, 3′, 4′-heptamethoxyflavone (peak 13), 5-hydroxy-6, 7, 8, 3′, 4′-pentamethoxyflavone (peak 14), and tangeretin (peak 15). Overall, these four flavonoid O-glycosides described above, viz., eriocitrin, narirutin, hesperidin, and didymin, appear to be ubiquitously present, and consequently detected, in all sweet orange fruit peels and pulp (with the exception of didymin in pulp).

One hundred and seven Arabidopsis UGTs were employed as a query for BLASTP searches against the C. sinensis genome¹, which identified 123 putative UGTs sorted by e value (1e^-5). Based on previous transcriptome data sets (Wu et al., 2014; Huang et al., 2016; Wang J. et al., 2016), 10 putative UGT genes were found to be highly expressed in sweet orange.

To predict whether the putative Citrus UGTs are involved in flavonoid glycosylation, an unrooted NJ phylogenetic tree (Figure 3) was constructed based on the full-length proteins of 10 UGTs and 12 functionally characterized flavonoid UGTs. From the phylogenetic trees, three putative UGTs (orange1.1g046033m, orange1.1g010093m, and orange1.1g012735m) were clustered with C12RT1, BpUGAT, UGT78G1, and Cs1,6Rhat (encoded by orange1.1g037721m). These UGTs are known to function as glucosyltransferase, glucuronosyltransferase, or rhamnosyltransferase of flavonoids. Alignment analysis revealed the presence of a conserved PSPG box in the C-terminus of all UGTs in this cluster (Supplementary Figure S1).

According to the genomic data obtained from Phytozome v12.1, PCR amplification was performed to produce the full-length cDNA of three genes from the fruit of sweet oranges. Only orange1.1g012735m could be obtained, however; this gene is 1,495 bp long and contains a 3′ untranslated region (UTR) of 41 bp, a 5′ UTR sequence of 80 bp, and an ORF of 1,374 bp that encodes a 475-amino acid protein. The base sequence information of orange1.1g012735m could be retrieved from Phytozome v12.1 and the gene, CsUGT76F1, was named according to the UGT committee nomenclature protocols. qRT-PCR was employed to investigate the expression abundance of CsUGT76F1 in fruit pulp and peels during fruit development (Figure 4). Transcripts of CsUGT76F1 had higher levels of expression in peels than in pulp. In peels, CsUGT76F1 was downregulated during fruit development and ripening, whereas in pulp, the transcript was upregulated initially and then gradually decreased with fruit ripening.

The full-length coding region of CsUGT76F1 was transferred into a pMAL-C2X expression vector with a maltose-binding protein (MBP) tag and introduced into E. coli. The recombinant CsUGT76F1 was purified by a MBP tag purification system. Initial in vitro assays using UDP-glucose as the sugar donor indicated that the protein could catalyze the glucosyl transfer to the 7-hydrogen sites of hesperetin, naringenin, and diosmetin (Figures 5A–C). The biotransformation products were verified using UPLC-Q-TOF-MS by comparing them with reference standards, the mass of the molecular ion, and the resulting fragments (Supplementary Table S3). Subsequently, substrate specificity studies for the recombinant protein with flavonols were also conducted, including quercetin and kaempferol, the glycosylated products of which had been previously identified in other Citrus plants (Abad-García et al., 2012; Wang S. et al., 2016). The results demonstrated that the recombinant UGT was able to glucosylate quercetin at the 3- or 7-hydrogen position (Figure 5D), but using kaempferol as substrate, three products— kaempferol 3,7-O-diglucoside, kaempferol 3-O-glucoside, and kaempferol 7-O-glucoside, were detected (Figure 5E).

In addition, the sugar-donor specificities for the recombinant protein were also investigated, including UDP-rhamnose, UDP-galactose, and UDP-xylose. The results demonstrated that the recombinant UGT can only accept UDP-rhamnose to rhamnosylate the 7-hydrogen position of quercetin and kaempferol (Figures 5F,G), but not to catalyze the rhamnosylation of hesperetin, naringenin, and diosmetin. Since flavonoid-O-glycosides in the Citrus plants often contain a disaccharide (rhamnosyl-α-1,6-glucose or rhamnosyl-α-1,2-glucose), which is sequentially added to the flavonoid aglycones, we further tested whether the recombinant UGT was able to rhamnosylate flavonoid-O-monoglycosides, including hesperetin-7-O-glucoside, naringenin-7-O-glucoside, diosmetin-7-O-glucoside, quercetin-3 and 7-O-glucoside, and kaempferol-3 and 7-O-glucoside. No flavonoid O-rutinoside (rhamnosyl-α-1,6-glucose) or O-neohesperidose (rhamnosyl-α-1,2-glucose) products were detected. These results indicated that CsUGT76F1 in vitro could function as flavonoid 3-O-, 7-O-, 3,7-O-glucosyltransferase, and 7-O-rhamnosyltransferase.

The optimal reaction conditions for CsUGT76F1 were determined using naringenin and kaempferol as substrates, and UDP-glucose and UDP-rhamnose as the sugar donors (Figure 6). The pH and temperature of the reaction buffer ranged from 4.0–11.0 to 10–70°C, respectively. The maximum levels of naringenin 7-O-glucoside product accumulation occurred at pH 8.0 and a temperature of 40°C; for kaempferol 7-O- and 3-O-glucoside, the maximum accumulations occurred at pH 8.0 and 40–45°C temperature, and pH 8.0–8.5 and 40°C temperature, respectively; and for kaempferol 3,7-O-diglucoside, the maximum accumulation occurred at pH 7.5–8.5 and temperatures of 35–45°C. The maximum accumulation of the kaempferol 7-O-rhamnoside product could be observed at pH 7.5 and a temperature of 35–40°C.

The kinetic parameters of CsUGT76F1 were further investigated on the flavonoid substrates at pH 8.0 and 40°C (Table 1). The enzyme showed the highest affinity toward the substrates hesperetin (K_m = 15.16 μM) and naringenin (K_m = 20.41 μM) compared to that toward diosmetin (K_m = 43.27 μM), quercetin (K_m = 36.78 μM), and kaempferol (K_m = 28.09 μM). However, the enzyme used diosmetin (k_cat = 1.60 s^-1) most efficiently, followed by naringenin (k_cat = 0.71 s^-1) and hesperetin (k_cat = 0.77 s^-1). The kcat values for the quercetin and kaempferol were 0.58 and 0.46 s^-1, respectively. The k_cat/K_m ratio was the highest for hesperetin (kcat/Km = 50.39 M^-1 s^-1), followed by that for naringenin (kcat/Km = 34.79 M^-1 s^-1), diosmetin (kcat/Km = 36.98 M^-1 s^-1), quercetin (kcat/Km = 15.77 M^-1 s^-1), and kaempferol (kcat/Km = 16.38 M^-1 s^-1).

In order to clarify whether CsUGT76F1 functions as a rhamnosyltransferase, the coding region of CsUGT76F1 under the 35S promoter was introduced into tobacco plants, with the positive transgenic plants identified using antibiotic selection and flavonoid compounds extracted from the leaves of transgenic and control tobacco. Compared with control tobacco, one new flavonol glycoside (Figure 7, peak 11) was produced in transgenic tobacco, and two other flavonol glucosides were significantly higher (peak 12 and peak 13). Based on the results of both the UPLC-Q-TOF-MS analysis and previous studies, the peak 11 product was identified as quercetin 7-O-rhamnoside, peak 12 was quercetin 7-O-glucoside, and peak 13 was kaempferol 7-O-glucoside (Supplementary Table S3). Higher accumulation of quercetin 7-O-rhamnoside, quercetin 7-O-glucoside and kaempferol 7-O-glucoside in the transgenic plants demonstrated that CsUGT76F1 had flavonoid 7-O-glucosyltransferase and 7-O-rhamnosyltransferase activities in vivo.