Introduction

Flavonoids are plant secondary metabolites derived from the amino acids, phenylalanine, and tyrosine via the phenylpropanoid pathway (Winkel-Shirely 2001). Most flavonoids exist as O-glycosylated forms, but some plants produce C-glycosylated flavonoids (Bowles et al. 2006). O-Glycosylation reactions occur at hydroxyl groups during the final stage of flavonoid biosynthesis (Vogt and Jones 2001) while C-glycosylation occurs during flavonoid biosynthesis (Brazier-Hicks et al. 2009). Flavonoids can be glycosylated with various sugars. Sugars commonly linked to flavonoids include arabinose, galactose, glucose, rhamnose, glucuronic acid, and xylose (Bowles et al. 2006).

The biosynthesis of flavonoid-O-glycosides is mediated by nucleotide diphosphate-dependent glycosyltransferases (UGTs) (Vogt and Jones 2001), most of which use uridine diphosphate sugars as sugar donors, although thymidine derivatives can also be used (Lim et al. 2006). Thus, the biosynthesis of nucleotide sugars is a prerequisite to the formation of flavonoid-O-glycosides. Nucleotide sugar biosynthesis has been extensively studied in plants due to its role in cell wall biosynthesis (Seifert 2004). UDP-glucose is eventually converted into UDP-galactose, UDP-glucuronic acid, or UDP-rhamnose. UDP-glucuronic acid serves as a precursor to UDP-xylose, UDP-adipose, and UDP-arabinose. GDP-sugars such as GDP-mannose, GDP-galactose, GDP-fucose, and GDP-glucose are also found in plants. Most of the genes involved in nucleotide sugar biosynthesis in plants have been characterized (Reiter 2008). Furthermore, genes regulating nucleotide sugar biosynthesis in plants and microorganisms have been successfully expressed in heterologous systems such as Escherichia coli and Saccharomyces cerevisiae (Oka and Jigami 2006; Yang et al. 2011). These systems serve as host that provide sugar donors for the synthesis of secondary metabolite-glycosides, including flavonoid glycosides.

The biosynthesis of phytochemicals in E. coli has been widely studied. The flavonoid biosynthesis pathway has been a good model for biosynthesis of phytochemicals in E. coli (Horinouchi 2008; Flower and Koffas 2009; Du et al. 2010). Simple modification reactions, semi-synthesis, and total synthesis have been performed by introducing flavonoid biosynthetic genes into E. coli. Although flavonoids have been successfully synthesized from glucose in E. coli (Santos et al. 2011; Kim et al. 2013a, b), there is an advantage to introducing simple modifications into flavonoids through biotransformation (Kim et al. 2005; 2006). Biotransformation of flavonoids using E. coli harboring UGT has been used successfully to produce flavonoid glucosides (Lim et al. 2006; Ko et al. 2006). E. coli provides the sugar donors (nucleotide sugars) necessary to synthesize flavonoid glycosides using exogenous flavonoids. E. coli synthesizes diverse nucleotide sugars for polysaccharide biosynthesis, which serve as sugar donors. To date, several flavonoid glycosides have been synthesized through biotransformation in E. coli. These include flavonoid-O-glucoside (Ko et al. 2006: Lim et al. 2004), flavonoid-O-xyloside (Pandey et al. 2013; Han et al. 2014), flavonoid-O-rhamnoside (Kim et al. 2012a; 2013a, b), flavonoid-O-(N-acetylglucosamine) (Kim et al. 2012b), flavonoid-O-(6-deoxytalose) (Yoon et al. 2012), and flavonoid-3-O-(4-deoxy-4-formamido-L-arabinose) (flavonoid-3-O-Ara4FN) (Kim et al. 2010). Although naringenin 7-O-glucuronide was synthesized previously, the final yields were not reported (Simkhada et al. 2010).

Flavonoid-O-galactoside and flavonoid-O-glucuronide, whose synthesis has not been extensively attempted in E. coli, have diverse biological activities. For example, quercetin-3-O-galactoside (hyperoside) has antidepressant effects (Zheng et al. 2012) and anti-inflammatory activity (Kim et al. 2011). Quercetin-3-O-glucuronide (miquelianin) also has antidepressant effects (Juergenliemk et al. 2003) and can prevent Alzheimer’s disease and hypertension (Ho et al. 2013; Balasuriya and Rupasinghe 2012). Luteolin-7-O-glucuronide has anti-genotoxic and choleretic effects (Orhan et al. 2013; Benedek et al. 2006). In the present study, we synthesized flavonoid-O-glucuronide (luteolin-7-O-glucuronide and quercetin-3-O-glucuronide) and flavonoid-O-galactoside (quercetin-3-O-galactoside) in E. coli. We expressed specific UGTs for either UDP-glucuronic acid or UDP-galactose in E. coli and engineered its nucleotide sugar metabolism. Using this approach, we could achieve yield of more than 280 mg/L for flavonoid-O-galactoside and flavonoid-O-glucuronides.

Materials and methods

Nucleic acid manipulation

UGT from Antirrhinum majus (AmUGT10; Gene ID AB362988; Noguchi et al. 2009), which was used for the production of luteolin-7-O-glucuronide, was cloned using reverse transcription polymerase chain reaction (RT-PCR). Total RNA was isolated from the flower of A. majus using the Qiagen Plant Total RNA Isolation kit (Qiagen, Hilden, Germany). cDNA was synthesized using Omniscript reverse transcriptase with oligo dT as a primer. The primers for AmUGT10 are listed in Table 1. AmUGT10 was subcloned into the EcoRI/NotI site of pETDuet-1 (Novagen), and the resulting construct was named pE-AmUGT (Table 1). To synthesize quercetin-3-O-galactoside, UGT from Petunia hybrid (PhUGT, GenBank ID; AF165148; Miller et al. 1999) was cloned by polymerase chain reaction (PCR) using cDNA as template. The resulting PCR product was digested with EcoRI/NotI and subcloned into the corresponding sites of pGEX 5X-3 (pG-PhUGT in Table 1). OsUGE (UDP-glucose epimerase from Oryza sativa; GenBank ID; NP_001062869.1) (Kim et al. 2009) was subcloned into the EcoRI/NotI site of pCDFDuet (pC-OsUGE in Table 1). UDP-glucose 4-epimerase (galE, GenBank ID; 945354) from E. coli DH5α was subcloned into the EcoRI/NotI site of pCDFDuet (pC-galE in Table 1). UGT from Vitis vinifera (VvUGT; GenBank ID XM_002280887; Ono et al., 2010), which was used for the production of quercetin-3-O-glucuronide, was cloned using RT-PCR. Total RNA was isolated from the leaves for V. vinifera, and cDNA was synthesized as described above. The PCR product was subcloned into the EcoRI/NotI site of pGEX 5X-3 (pG-VvUGT in Table 1). UDP-glucose dehydrogenase (ugd; Gene ID, 946571) from E. coli was cloned using PCR using E. coli genomic DNA as a template. The resulting PCR product was sequenced and subcloned into the NdeI/XhoI site of pE-AmUGT (pE-AmUGT-Ecugd in Table 1). Ugds from Arabidopsis thaliana (Gene ID, AT3G29360) and Glycine max (Gene ID, U53418) were cloned using RT-PCR. The resulting PCR products were subcloned into the NdeI/KpnI site of pE-AmUGT for AtUgd (pE-AmUGT-AtUgd) and into the NdeI/XhoI site of pE-AmUGT for GmUgd (pE-AmUGT-GmUgd). Primers used are listed in Table 1.

Table 1 Plasmids and strains used in the present study
Full size table

Gene deletion in E. coli BL21 (DE3) was carried out using the Quick & Easy Conditional Knock Out Kit (Gene Bridges, Heidelberg, Germany). The arnA deletion mutant was developed as described in Kim et al. (2010).

Biotransformation of flavonoid using E. coli

An overnight culture of E. coli harboring specific genes was inoculated into 50 mL fresh of LB medium containing appropriate antibiotics. The cells were grown until the OD600 reached 0.8. Genes in the E. coli expression vector were induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 1 mM. The transformants were grown for an additional 20 h at 18 °C. Cells were harvested by centrifugation and resuspended in fresh M9 medium containing 2 % glucose and antibiotics and 1 mM IPTG. To compare the production of luteolin-7-O-glucuronide among different E. coli strains (B500-B507, Table 1), the cell density was adjusted to an OD600 value of 2. The total reaction volume was 2 mL in a test tube (14 mm × 145 mm). One hundred micromoles of luteolin was added, and the reaction mixture was incubated at 25 °C for 1.5 h. Then, an additional 100 μM luteolin was added, and the mixture was incubated for an additional 1.5 h. To determine the optimum cell concentration, 200 μM of luteolin was added. After 2 h, an additional 200 μM luteolin was added, and the reaction mixture was incubated for 2 h. To compare the production of quercetin-3-O-glucuronide among different E. coli strains, the cell densities were adjusted to an OD600 value of 3, and the reactions were carried out for 2 h after addition of 200 μM quercetin.

The effect of EcgalE and OsUGE on the production of quercetin-3-O-galactoside in E. coli BL21 (DE3) was tested using strains B510 and B511 (Table 1). Each strain was cultured as described above. One hundred micromoles of quercetin was added, and the mixture was incubated at 30 °C for 1.5 h with shaking.

To analyze quercetin-3-O-galactoside, the culture supernatant was harvested, extracted twice with an equal volume of ethyl acetate and dried in a vacuum. The dried sample was dissolved in dimethyl sulfoxide (DMSO) and analyzed by high-performance liquid chromatography (HPLC) using a Varian HPLC equipped with a photodiode array (PDA) detector and a Varian C18 reverse-phase column (Varian, 4.60 × 250 mm, 3.5 μm particle size). Cell growth was monitored by determining the absorbance at 600 nm. For analysis of flavonoid-O-glucuronic acid, the reaction supernatant was boiled for 3 min and centrifuged for 10 min at 13,000g. The supernatant was analyzed with HPLC.

The mean and standard error of the mean were calculated from three experiments. Analysis of variance (ANOVA) was carried out using Tukey’s method with a significance level of P = 0.01. Statistical analyses were performed using Microsoft Office Excel 2010.

The molecular masses of the reaction products were determined by mass spectrometry as described previously (Lee et al. 2007). The structures of the reaction products were determined using nuclear magnetic resonance (NMR) spectroscopy as described previously (Kim et al. 2004). The following data were obtained. Quercetin-3-O-galactoside: 1H NMR (400 MHz, DMSO-d 6 ) δ (ppm) 7.67 (dd, J = 8.5, 2.2 Hz, 1H, H-6′), 7.54 (d, J = 2.2 Hz, 1H, H-2′), 6.83 (d, J = 8.5 Hz, 1H, H-5′), 6.41 (d, J = 2.0 Hz, 1H, H-8), 6.21 (d, J = 2.0 Hz, 1H, H-6), 5.38 (d, J = 7.7 Hz, 1H, H-1″), 3.66 (d, J = 3.1 Hz, 1H, H-4″), 3.58 (m, 1H, H-2″), 3.47 (dd, J = 9.6, 5.2 Hz, 1H, H-6″), 3.39 (dd, J = 9.5, 3.3 Hz, 1H, H-3″), 3.33 (d, J = 2.9 Hz, 1H, H-5″), 3.31 (m, 1H, H-6″); 13C NMR (100 MHz, DMSO-d 6 ) δ (ppm) 177.4 (C-4), 164.2 (C-7), 161.2 (C-5), 156.3 (C-2), 156.2 (C-9), 148.4 (C-4′), 144.8 (C-3′), 133.4 (C-3), 121.1 (C-1′), 122.0 (C-6′), 115.9 (C-2′), 115.2 (C-5′), 103.9 (C-10), 101.8 (C-1″), 98.7 (C-6), 93.5 (C-8), 75.8 (C-5″), 73.2 (C-3″), 71.2 (C-2″), 67.9 (C-4″), 60.1 (C-6″). Quercetin-3-O-glucuronic acid: 1H NMR (CD3OD, 400 MHz): δ 7.61–7.65 (m, 2H), 6.85 (d, J = 8.4 Hz, 1H), 6.39 (d, J = 2.0 Hz, 1H), 6.20 (d, J = 2.1 Hz, 1H), 5.34 (d, J = 7.6 Hz, 1H), 3.74 (d, J = 9.7 Hz, 1H), 3.60 (d, J = 8.8 Hz, 1H), 3.56 (d, J = 3.1 Hz, 1H), 3.45–3.53 (m, 3H); 13C NMR (CD3OD, 100 MHz): δ 178.7, 165.4, 162.4, 158.4, 157.9, 149.3, 148.3, 145.4, 134.8, 122.8, 122.2, 116.6, 115.4, 105.0, 103.6, 99.3, 94.1, 77.0, 76.6 74.8, 72.3. Luteolin-7-O-glucuronic acid: 1H NMR (400 MHz, DMSO-d 6 ) δ (ppm) 7.44 (d, J = 8.0 Hz, 1H), 7.43 (s, 1H), 6.90 (d, J = 8.0 Hz, 1H), 6.81 (d, J = 2.0 Hz, 1H), 6.75 (s, 1H), 6.45 (d, J = 2.0 Hz, 1H), 5.26 (d, J = 6.8 Hz, 1H), 4.00 (d, J = 9.2 Hz, 1H), 3.38 (t, J = 9.2 Hz, 1H), 3.33 (t, J = 9.0 Hz, 1H), 3.29 (t, J = 7.2 Hz, 1H); 13C NMR (100 MHz, DMSO-d 6 ) δ (ppm) 181.85, 170.38, 164.45, 162.49, 161.10, 156.91, 149.92, 145.75, 121.27, 119.12, 115.96, 113.49, 105.37, 103.10, 99.35, 99.13, 94.48, 75.65, 75.15, 72.74, 71.29.

Results

Production of flavonoid-O-glucuronides

We planned to use E. coli to synthesize flavonoid glucuronide. E. coli synthesizes UDP-glucuronic acid but does not contain a UGT specific for luteolin and UDP-glucuronic acid. AmUGT10 uses UDP-glucuronic acid as a sugar donor and luteolin as the optimal sugar acceptor to make luteolin-7-O-glucuronide (Noguchi et al. 2009). E. coli harboring pE-AmUGT (B500 in Table 1) was used for biotransformation of luteolin, and analysis of the reaction product with HPLC showed a new peak at 462 Da (Fig. 1a), whose molecular mass was consistent with that of a glucuronic acid attached to luteolin (Fig. 1d). In addition, the reaction product differed in retention time from luteolin-7-O-glucoside (Fig. 1c). MS/MS analysis of the reaction product revealed a compound with a molecular mass of 286 Da (Fig. 1d), which corresponded to the molecular weight of luteolin. Thus, we concluded that E. coli harboring pE-AmUGT synthesized luteolin-7-O-glucuronic acid from luteolin. Furthermore, NMR revealed the structure of the biotransformation product to be luteolin-7-O-glucuronide (see “Materials and methods”).

Fig. 1
figure 1

a HPLC profile of the reaction product (P1) obtained from E. coli strain B500 supplemented with luteolin (S1); b luteolin-7-O-glucose standard (S2); c mixture of the reaction product and luteolin-7-O-glucose standard; d MS/MS analysis of the reaction product (P1). The mass spectrometer was operated in negative mode

Full size image

To increase the production of luteolin-7-O-glucuronide, it was necessary to increase the supply of UDP-glucuronic acid in E. coli because luteolin was supplied exogenously, but UDP-glucuronic acid was synthesized by the E. coli itself. UDP-glucuronic acid is synthesized from UDP-glucose by UDP-glucose dehydrogenase (ugd) and is then converted into UDP-4″-ketose by UDP-L-Ara4N formyltransferase/UDP-GlcA C-4″-decarboxylase (arnA), which is involved in the biosynthesis of outer membrane lipopolysaccharides (Breazeale et al. 2005). Thus, in order to increase the supply of UDP-glucuronic acid, arnA was first deleted in E. coli (B-araA in Table 1), and B-arnA was transformed with pE-AmUGT. The resulting strain B504 produced 31.1 mg/L of luteolin-7-O-glucuronide, which is approximately 121 % more than B500 (14.1 mg/L) (Fig. 2).

Fig. 2
figure 2

Comparison of luteolin-7-O-glucuronide production from luteolin using different combinations of ugd and E. coli mutant strains. Data are mean values ± SD from three independent experiments

Full size image

In order to increase the supply of UDP-glucuronic acid, ugd, which encodes a protein converting UDP-glucose to UDP-glucuronic acid, was overexpressed in E. coli. Three ugd genes from E. coli, A. thaliana, and G. max were tested in two different E. coli strains (wild type and B-araA). The product of biotransformation of luteolin for each strain (B500∼B507 in Table 1) was analyzed using HPLC. E. coli transformants containing ugd produced more luteolin-7-O-glucuronide than E. coli transformants not containing ugd. The ugd of E. coli (Ecugd) was the most efficient followed by those of A. thaliana and G. max (Fig. 2). Strain B505 (B-araA harboring pE-AmUGT-Ecugd) produced 197 % more luteolin-7-O-glucuronide (92 mg/L) than strain B504 (B-araA harboring only pE-AmUGT, 31.1 mg/L). Thus, an increased supply of UDP-glucuronic acid by ugd overexpression is required to increase the production of luteolin-7-O-glucuronide. As shown above, the B-araA strain produced more luteolin-7-O-glucuronide than BL21 for all constructs. B500, in which only AmUGT was overexpressed in wild type E. coli BL21 (DE3), synthesized approximately 14 mg/L luteolin-7-O-glucuronic acid while B505, in which both AmUGT and Ecugd were overexpressed in B-araA, produced approximately 92 mg/L of luteolin-7-O-glucuronide.

We employed another UGT (VvUGT) to synthesize quercetin-3-O-glucuronide from quercetin in E. coli. Strain B508 (Table 1) was used for biotransformation of quercetin, and the biotransformation product was analyzed. A new peak was observed at 478 Da (Fig. 3a), the molecular mass corresponding to that of quercetin-3-O-glucuronide. The structure of the reaction product was determined to be quercetin-3-O-glucuronide by NMR (see “Materials and methods”). Similar to the synthesis of luteolin-7-O-glucuronide, the production of quercetin-3-O-glucuronide increased in the araA deletion strain (B-araA) and overexpressing Ecugd. Strain B508 (E. coli BL21 [DE3] harboring pG-VvUGT) produced approximately 38 mg/L quercetin-3-O-glucuronide, while strain B509 (B-ara harboring pG-VvUGT and pA-Ecugd) produced approximately 62 mg/L quercetin-3-O-glucuronide. Therefore, the engineered E. coli strain produced approximately 1.6-fold as much quercetin-3-O-glucuronide as wild type E. coli strain.

Fig. 3
figure 3

HPLC analysis of the reaction product obtained with strain B508 supplemented with quercetin as a substrate. a The reaction product obtained with E. coli BL21 harboring empty vector, 5X-3; b HPLC profile and MS analysis of the reaction product obtained with strain B508. The mass spectrometer was operated in negative mode

Full size image

The production of luteolin-7-O-glucuronide was optimized with strain B505. First, we determined the optimal cell density. After induction, the cell density was adjusted to OD600 values of 1, 1.5, 2.0, 2.5, 3.0, 3.5, 4, 5, and 7. Initially, 200 μM (57.2 mg/L) luteolin was added. After incubation of the reaction mixture at 25 °C for 2 h, an additional 200 μM of luteolin was added, and the reaction mixture was incubated 25 °C for 6 h. In the reaction mixture containing cells at an OD600 value of 3.5, 320 μM (148 mg/L) luteolin-7-O-glucuronide was produced, which was the highest yield among the cell densities tested. The production of luteolin-7-O-glucuronide was monitored. The cell density of strain B505 was 3.5 at 600 nm. One hundred micromolar of luteolin (28.6 mg/L) was added at 0, 2, 4, 6, 8, 10, and 12 h (the final concentration was 700 μM [200.4 mg/L]) because the solubility of luteolin is low and a high concentration might cause cell lysis. The mixture was incubated at 25 °C for 48 h. The reaction mixture was collected periodically and analyzed by HPLC. The production of luteolin-7-O-glucuronide continued to increase until 24 h. At this point, 300.5 mg/L of luteolin-7-O-glucuronide (approximately 650 μM) was produced (Fig. 4a). The conversion yield from luteolin to luteolin-7-O-glucuronide was 92.8 %.

Fig. 4
figure 4

Production of luteolin-7-O-glucuronide using strain B505 (a) and quercetin-3-O-glucuronide using B509 (b). Data are mean values ± SD from three independent experiments. For the synthesis of luteolin-7-O-glucuronide, 100 μM luteolin was added at 0, 2, 4, 6, 8, 10, and 12 h (total 700 μM luteolin was added). For the synthesis of quercetin-3-O-galactoside, 200 μM quercetin was added every 2 h up to 12 h and then at 24 h (total 1.4 mM quercetin was added). Sample was taken at each time point, and the product was analyzed using HPLC

Full size image

The production of quercetin-3-O-glucuronide was also monitored for 48 h in strain B509. Quercetin (100 μM = 30.2 mg/L) was added every hour up to 12 h and then at 24 h. Therefore, the final amount of quercetin added was 1.4 mM (422.8 mg/L). The initial cell density was set to an OD600 value of 3. The production of quercetin-3-O-glucuronide increased rapidly for 12 h, and then the rate of increase slowed. After 48 h, 659.6 mg/L of quercetin-3-O-glucuronide (1.38 mM) was produced (Fig. 4b). The conversion yield was 98.5 %.

Production of quercetin-3-O-galactoside

PhUGT, flavonol-3-O-galactosyltransferase, was used to produce quercetin-3-O-galactoside (Miller et al. 1999). E. coli BL21 (DE3), which is a generally used host for expressing foreign genes that was transformed with pG-PhUGT (Table 1). However, galE, which is a UDP-glucose 4-epimerase and catalyzes the interconversion of UDP-glucose and UDP-galactose, is absent in E. coli BL21 (DE3), and this strain no longer synthesizes UDP-galactose. Therefore, biotransformation of quercetin using E. coli BL21 (DE3) harboring pG-PhUGT did not produce any reaction product (Fig. 5a). It clearly showed that UDP-galactose is absent from E. coli BL21 (DE3) due to the absence of galE. On the other hand, E. coli DH5α harboring pG-PhUGT produced a new reaction product from quercetin, which was predicted to be quercetin-3-O-galactoside, because E. coli DH5α contains an intact copy of galE. Thus, both galE and PhUGT are required to synthesize quercetin-3-O-galactoside in E. coli BL21 (DE3). We decided to use E. coli BL21 (DE3) instead of DH5α because this strain is commonly used to express foreign genes.

Fig. 5
figure 5

HPLC profile of the reaction product obtained with strain B511 supplemented with quercetin as a substrate. a Reaction product obtained with E. coli BL21 harboring PhUGT; b HPLC profile and MS analysis of the reaction product obtained from strain B510 supplemented with quercetin as a substrate. The mass spectrometer was operated in negative mode

Full size image

In order to provide UDP-galactose to E. coli BL21, two genes, EcgalE and OsUGE, were tested. Both EcgalE and OsUGE interconvert UDP-glucose and UDP-galactose. E. coli BL21 (DE3) was transformed with pC-galE or pC-OsUGE along with pG-PhUGT (Table 1). The production of quercetin-3-O-galactoside in strains, B510 and B511 (Table 1) were compared. Strain B511 produced 38 mg/L quercetin-3-O-galactoside, which was 1.9-fold as much as that produced by strain B510 (20 mg/L). This indicated that OsUGE more effectively converted UDP-glucose to UDP-galactose. The structure of the biotransformation product was determined to be quercetin-3-O-galactoside by NMR (see “Materials and methods”).

Using strain B511, we monitored the production of quercetin-3-O-galactoside. The initial cell density was set to OD600 value of 3. Quercetin (100 μM = 30.2 mg/L) was added at 0, 3, 6, 9, 13, 24, and 36 h (total 700 μM [211.5 mg/L] quercetin was added). Quercetin-3-O-galactoside was rapidly produced until 24 h. After 36 h, its production did not increase substantially. Approximately 280.4 mg/L quercetin-3-O-galatoside (approximately 603 μM) was produced after 60 h of biotransformation (Fig. 6). The conversion yield was 86.1 %.

Fig. 6
figure 6

Production of quercetin-3-O-galactoside from strain B511. Data are mean values ± SD from three independent experiments. Quercetin (100 μM) was added at 0, 3, 6, 9, 12, 24, and 36 h (total 700 μM was added)

Full size image

Discussion

UDP-glucose is a substrate for the synthesis of UDP-galactose, UDP-glucuronic acid, and UDP-Ara4FN in E. coli. UDP-galactose is not synthesized in E. coli BL21 (DE3) because it lacks galE, which encodes the enzyme that converts UDP-glucose into UDP-galactose. Introducing galE into E. coli BL21 (DE3) resulted in the production of UDP-galactose, which served as a substrate for the production of quercetin 3-O-galactoside. For the production of flavonoid-O-glucuronide, we also overexpressed ugd to enhance the conversion of UDP-glucose to UDP-glucuronic acid. This increased in luteolin-7-O-glucuronic acid production from 14 to 65 mg/L. Furthermore, an E. coli strain in which a gene (arnA) for a protein converting UDP-glucuronic acid to UDP-Ara4FN produced more luteolin-7-O-glucuronic acid (30 mg/L) than wild type (14 mg/L). Finally, E. coli strain, in which ugd was overexpressed and arnA was deleted, was the most effective for the production of luteolin-7-O-glucuronic acid (92 mg/L).

Two flavonoid-O-glucuronides (luteolin-7-O-glucuronide and quercetin-3-O-glucuronide) and quercetin-3-O-galactoside were synthesized in E. coli via biotransformation of exogenously supplied luteolin and quercetin, respectively. The UGTs that used (namely, AmUGT10, VvUGT, and PhUGT) have both sugar donor specificity and acceptor specificity (Miller et al. 1999; Noguchi et al. 2009; Ono et al., 2010). In regard to sugar acceptors, AmUGT10 could use flavonoids having a 7-hydroxy group such as naringenin, apigenin, and quercetin. Therefore, it is possible to make naringenin-7-O-glucuronide, apigenin-7-O-glucuronide, and quercetin-7-O-glucuronide. Kaempferol, which also has a 3-hydroxy group, serves as a substrate for PhUGT and VvUGT for the synthesis of kaempferol-3-O-galactoside and kaempferol-3-O-glucuronide, respectively.

Flavonoid-O-glucuronides and flavonoid-O-galactose have not been synthesized previously in E. coli. Most research on flavonoid glycosylation has been about the regioselective glucosylation of flavonoids, although glucuronylation and galactosylation of flavonoids in plants are also common (Xiao et al. 2014). The reason that glucosylation of flavonoids is common is because most UGTs use UDP-glucose as a sugar donor. In this study, we chose UGTs that use either UDP-glucuronic acid or UDP-galatose as a sugar donor. In addition, we engineered E. coli to provide higher levels of the appropriate sugar donors to enhance synthesis of flavonoid-O-glucuronides and flavonoid-O-galactosides. The E. coli strains which nucleotide biosynthesis pathways have been engineered could be used to attach glucuronic acid and galactose to other secondary metabolites.

Flavonoid glucosides were the most prevalent flavonoid glycosides synthesized using biotransformation. The final yield of quercetin- O-glucoside ranged from 1 to 10 mg/L (Lim et al. 2004). Using a controlled fermentation system, yields of approximately 100 mg/L were obtained (Lim et al. 2004). E. coli harboring isoflavone glucosyltransferase gave yields ranging from 10 to 20 mg/L genistein-7-O-glucoside or biochanin A-7-O-glucoside (He et al. 2008). These studies were performed using endogenous UDP-glucose as a substrate donor. In this study, 300 mg/L of luteolin-7-O-glucuronide, 687 mg/L of quercetin-3-O-glucuronide, and 280 mg/L of quercetin-3-O-galactoside were obtained This was achieved by engineering the nucleotide pathway of E. coli. Quercetin-3-O-rhamnoside and kaempferol-3-O-rhamnoside were also synthesized (150 to 200 mg/L) in E. coli strain with an engineered nucleotide synthesis pathway (Kim et al. 2012a, b). Thus, engineering the nucleotide pathway in E. coli is critical to increasing the yield of flavonoid glycosides.

We tested functionally identical genes from different organisms in order to identify the one with the highest productivity. For the synthesis of luteolin-3-O-glucuronide, ugd from E. coli was more effective than those from plants such as A. thaliana and G. max. On the other hand, OsUGE from rice was more effective than that from E. coli in the synthesis of quercetin-3-O-galactoside. Although we did not measure in vitro catalytic efficiency of the individual enzymes encoded by each gene, the most efficient enzyme was selected by comparing the final yield. Furthermore, in vitro catalytic efficiency is always not correlated with in vivo biotransformation efficiency (Lim et al. 2011). Thus, the overall balance of several reaction intermediates and the concentration of enzymes must be modulated in order to maximize the productivity.