Biosynthesis of phloretin and its C-glycosides through stepwise culture of Escherichia coli

Abstract

Phloretin (PT) belongs to the dihydrochalcones (DHCs) family and is found in apple and rooibos tea. Its glycosides, including phlorizin (PT 2′-O-glucoside), trilobatin (PT 4′-O-glucoside), and nothofagin (NF, PT 3′-C-glucoside), are present in various plants. Phloretin and its related glycosides possess health benefits, including antioxidant, anti-inflammatory, and antibacterial activities. To biosynthesize PT and its glycosides, the relevant pathways in plants were studied and introduced into Escherichia coli. We reconstructed the biosynthetic pathways pertaining to PT and three PT C-glycosides (NF, PT 3′, 5′-di-C-glucoside [PDG], and PT 3′-C-arabinoside [PARA]) in E. coli. To prevent the undesirable synthesis of flavonoids instead of PT, we strategically divided the entire pathway into two parts: the first involved the synthesis of tyrosine to phloretic acid (PA), while the second involved the synthesis of PA to PT and its glycosides. The gene set pertaining to each part was incorporated into a different engineered microbe. We optimized phloretin microbial biosynthesis by improving enzyme affinity, identifying the gene that increased the output, refining the production design to a stepwise culture approach, and analyzing the culture conditions (substrate and yeast extract concentrations and pH) conducive to maximum output and the prevention of product degradation. Using the stepwise culture approach, 12.8 mg/L of PT, 26.1 mg/L of NF, 30.0 mg/L of PDG, and 18.1 mg/L of PARA were synthesized. This study provides valuable information for future approaches in the microbe-based synthesis of PT derivatives.

Introduction

Dihydrochalcones (DHCs) are a class of flavonoids having a C6-C3-C6 carbon skeleton; however, unlike flavanones, they possess an unsaturated C-ring structure [1]. Phloretin (PT) is a type of DHC containing a few hydroxyl groups, and it exists as glycosides such as phlorizin (PT 2′-O-glucoside), trilobatin (PT 4′-O-glucoside), nothofagin (NF, PT 3′-C-glucoside) [2, 3], and PT 3′, 5′-di-C-glucoside (PDG) [2,3,4,5]. PT and its glycosides exert health benefits such as anticancer properties, antimicrobial effects, anti-inflammatory activity, hepatoprotective effect, and anticholinesterase activity [6,7,8,9,10,11,12].

The PT synthesis pathway has been elucidated in plants [13]. The synthesis begins with p-coumaroyl-CoA, which is a common building block for various phenolic compounds found in plants. Dehydrogenase (DH) reduces p-coumaroyl-CoA to phloretoyl-CoA (4-hydroxydihydrocinnamoyl-CoA) in the presence of NADPH [14]. Then, chalcone synthase (CHS) synthesizes PT from phloretoyl-CoA using three malonyl-CoA molecules. Although the CHS family enzymes show a wide substrate selectivity range, the CHSs from apple and Psilotum nudum display higher substrate specificity for phloretoyl-CoA compared to other enzymes such as p-coumaroyl-CoA and cinnamoyl-CoA [15, 16]. Hence, some CHSs have been postulated to participate in PT synthesis [15, 17].

A system for converting phloretic acid (PA) to PT and synthesis of PT from tyrosine was designed in Saccharomyces cerevisiae by introducing the corresponding genes. Using the engineered yeast system, 619.5 mg/L of PT from PA was synthesized [18]. From phenylalanine, 42.7 mg/L of PT was synthesized in another S. cerevisiae strain [19]. S. cerevisiae was more effective than Escherichia coli in synthesizing PT. Although 4CL (4-coumarate-CoA ligase) specific for PA was selected, E. coli produced only 1.85 mg/L of PT from PA, and CHS was mutated to increase the specificity for phloretoyl-CoA. However, when E. coli synthesized PT from tyrosine, only 20 µg/L of PT was produced [20].

Phloretin glycosides are synthesized by glycosyltransferases (GTs). The PT O-glycosides, phlorizin and trilobatin, were synthesized from PT using E. coli expressing specific UGT to obtain titers of 287.29 mg/L [20] and 107.64 mg/L [21], respectively. The GTs responsible for NF synthesis have been identified in Oryza sativa Indica (OsCGT) [22] and Fagopyrum esculentum (UGT708C1, FeCGT) [23]. S. cerevisiae expressing OsCGT converted 79.3% of 205.5 mg of PT to NF [24]. Additionally, E. coli expressing FeCGT has demonstrated the conversion of 102 mg PT to NF with an efficiency of > 90% [25]. S. cerevisiae, using CGT (C-glycosyltransferase) from O. sativa, synthesizes NF from phenylalanine to produce 59 mg/L of NF, converting almost all PT molecules produced during the process [19]. Furthermore, FcCGT (UGT708G1) from Fortunella crassifolia and CuCGT (UGT708G2) from Citrus unshiu [26] are known to synthesize PDG from NF. E. coli expressing either FcCGT or CuCGT converted 100 µM of PT to NF and PDG with > 50% efficiency [26]. GgCGT from Glycyrrhiza glabra and DcaCGT from Dendrobium catenatum also synthesized PDG [27–28]. Recently, bifunctional C-arabinosyl and C-glucosyltransferase were identified [29]. UGT708 (uridine diphosphoglucuronosyltransferases) uses UDP-arabinose (UDP-Ara) or UDP-glucose (UDP-Glc) to glycosylate PT and 2-hydroxynaringenin, respectively. OsUGT708A39 from O. sativa predominantly uses UDP-Ara to produce PT 3′-C-arabinoside (PARA) [30].

We used two engineered E. coli strains and a stepwise culture strategy to synthesize PT (from tyrosine) and PT glycosides. The first microbe synthesized PA using the shikimate pathway gene module along with TAL (tyrosine ammonia lyase) and ER (enoate reductase) from Clostridium acetobutylicum, which converts p-coumaric acid to PA). The second microbe was incorporated with 4CL, CHS, matBC (Malonyl-CoA synthase (matB) and malonate carrier protein (matC)) or accABCD (acetyl-CoA carboxylase) to synthesize PT (Fig. 1). We introduced the GT and nucleotide pathways to synthesize the glycosides NF, PT 3′,5′-di-C-glucoside, and PT 3′-C-arabinoside from PT.

Fig. 1

(A) Biosynthesis scheme for PA, PT, NF, PDG, and PARA in Escherichia coli. Introduced genes (SeTAL [tyrosine ammonium lyase from Saccharothrix espanaensis], ER [2-enoate reductase from Clostridium acetobutylicum], Os4CL [4-coumarate CoA ligase from Oryza sativa], PeCHS [chalcone synthase from Populus euramericana], accABCD [acetyl-CoA carboxylase from Photorhabdus luminescens], matBC [malonyl-CoA synthase (matB) and malonate carrier protein (matC) from Rhizobium trifolii], FeCGT [phloretin C-glucosyltransferase from Fagopyrum esculentum], ZmCGT [phloretin di-C-glucosyltransferase from Zea mays], OsCGT [phloretin C-glycosyltransferase from O. sativa], Ecugd [UDP-glucose dehydrogenase from Escherichia coli], AtUXS [UDP-xylose synthase from Arabidopsis thaliana] and OsUGE [UDP-xylose epimerase from O. sativa]) are indicated in the rectangular box. The best strain used for the synthesis of each compound was indicated as green ellipses. ppsA, phosphoenolpyruvate synthase; tktA, transketolase; tyrR, phenylalanine DNA-binding transcription repressor; aroG, deoxyphosphoheptonate aldolase; aroB, dehydroquinate synthase; aroD, dehydroquinate dehydratase; aroE, shikimate dehydrogenase, aroL, shikimate kinase 2; aroA, 3-phosphoshikimate 1-carboxyvinyltransferase; aroC, chorismate synthase; tyrA, prephenate dehydrogenase; pheA prephenate, tyrB, aromatic-amino-acid aminotransferase. (B) Constructs used for the synthesis of PA, PT, NF, PDG, and PARA. To synthesize PT, NF, PDG, and PARA, the E. coli strain harboring constructs of PA synthesis was used as a first cell

Materials and methods

Plasmid constructs and E. coli strains

2-ER from Clostridium acetobutylicum (ER [AEI32805.1]) was cloned using polymerase chain reaction (PCR). The forward primer aaggatccaatgaacaaatacaagaaattatttga, containing the BamHI restriction enzyme recognition site (represented in lower case letters), and the reverse primer aagtcgacTTATATATGGTTTGCAACTTCAA, containing the SalI site (represented in lower case letters), were used. The PCR products were sequenced and cloned into the BglII/XhoI sites of pCDFDuet and pETDuet (Novagen) (pC-ER and pE-ER, respectively). SeTAL, which was previously cloned [31], was subcloned into the EcoRI and HindIII sites of pC-ER (pC-SeTAL-ER).

Os4CL, PeCHS, and acc were previously cloned [32,33,34]. Malonyl-CoA synthase (matB) and malonate carrier protein (matC) from Rhizobium trifolii were cloned using the genomic DNA of R. trifolii. For cloning matB, the forward primer aaggatccaATGAGCAACCATCTTTTCGAC, containing the BamHI site, and the reverse primer aaaagcttTTACGTCCTGGTATAAAGATCG, containing the HindIII site, were used. The PCR products were sequenced and cloned into pACYCDuet (Novagen) to generate pA-matB. The forward primer aaacatATGGGTATTGAATTACTGTCCA, containing NdeI, and the reverse primer aaggtaccTCAAACCAGCCCGGGC, containing KpnI, were used to clone matC. The PCR product was sequenced and subcloned into pA-matB to generate pA-matBC.

PT C-glucosyltransferase from Fagopyrum esculentum (AB909375.1) was cloned using reverse transcription (RT)-PCR. Total RNA was isolated from the 2-week-old F. esculentum seedlings, and cDNA was synthesized using OmniScript reverse transcriptase. We performed PCR using aagaattcaATGATGGGAGATTTAACAACTTCTTTT as the forward primer and aagcggccggTCAACGTTTAAGACTTCCGATGAT as the reverse primer, which contained the EcoRI and NotI sites, respectively (indicated in lower-case letters). The resulting PCR product was sequenced and subcloned into the EcoRI and NotI sites of the pGEX 5X-3 vector (GE Healthcare) to generate plasmid.

PT di-C-glucosyltransferase from Zea mays (ZmCGT; MK894455) was cloned using RT-PCR. The cDNA was synthesized as previously described. Two primers containing the EcoRI site (underlined) aagaattcaATGGCCCCGCCGGCAAC or the Hind III site (underlined) aaaagcttTCAAGACGCCACGGTTGCTTTG were used for the PCR. The PCR product was subcloned into the EcoRI or Hind III site of pETDuet to produce pE-ZmCGT.

PT C-glycosyltransferase from Oryza sativa Indica (OsCGT; OsUGT708A39) was cloned using RT-PCR. The cDNA was synthesized as previously described [35]. The sequences of the two primers were aagaattcaATGCCACCGCCCACGGT for EcoRI and aactcgagTTAAGCAGCCTTGAGCTTTGCAACAA for XhoI. The PCR product was subcloned into the EcoRI and XhoI sites of pGEX 5X-3. The genes required for the biosynthesis of UDP-Ara (AtUXS, Ecugd, and OsUGE) were previously cloned [35, 36]. OsCGT containing a glutathione S-transferase (GST) tag sequence was subcloned into Bgl II and XhoI of pC-OsUGE to produce pC-OsUGE-GST-OsCGT. The forward primer aaagatctaATGTCCCCTATACTAGGTTATT, containing the Bgl II site, was used and the reverse primer was aactcgagTTAAGCAGCCTTGAGCTTTGCAACAA. AtUXS-Ecugd from pA-AtUXS-Ecugd was digested using BamHI and XhoI sites and subcloned into the Bgl II and XhoI sites in pETDuet to form pE-AtUXS-Ecugd. OsUGE-GST-OsCGT was digested with NcoI and XhoI sites and subcloned into the NcoI and SalI sites of pE-AtUXS-Ecugd to produce pE-OsUGE-GST-OsCGT-AtUXS-Ecugd. The plasmid constructs and E. coli strains used in this study are listed in Table 1.

Table 1 Plasmids and Escherichia coli strains used in this study

Synthesis of PA, PT, NF, PDG, and PARA

E. coli BL21 (DE3) cells were used as the transformation host, and the transformants are listed in Table 1. The transformants were precultured overnight in 2 mL of Luria–Bertani (LB) broth with 0.1% appropriate antibiotics at 37 °C. Each culture was inoculated into 3 mL of fresh LB medium containing antibiotics, and the main culture was performed at 37 °C until OD₆₀₀ reached 1.0. Isopropyl β-D-1-thiogalactopyranoside (IPTG) was then added to the culture at a final concentration of 1 mM and incubated for induction at 18 °C for 16 h. Following this, the cells were harvested, and the concentration of the culture was adjusted until OD₆₀₀ was 3 using 1 mL of M9 broth containing 2% glucose, 1% yeast extract, 0.1% antibiotics, and 1 mM IPTG (YM9) for the bioconversion study. Then, p-coumaric acid, PA with malonate, or PT were added to the medium at 100 µM concentration when required. Malonate was added at a final concentration of 2 g/L. The cultures were incubated at 30 °C for 24 h with shaking at 180 rpm. The reaction products, except PDG, were extracted using ethyl acetate and dried using a speed vacuum. The sample was then dissolved in dimethyl sulfoxide (DMSO) and analyzed using high-performance liquid chromatography (HPLC). The PDG was extracted by heating the reaction product in boiling water and collecting the supernatant after centrifugation. The collected sample was analyzed using HPLC.

The stepwise culture was carried out as follows; The cells were prepared as described above. The cell concentration of the first E. coli strain was adjusted to OD₆₀₀ = 1 and the following strains were adjusted to OD₆₀₀ = 3. The culture filtrate of the first reaction cell was mixed with the cells for the second reaction. The resulting cells were incubated at 30 °C for 24 h with shaking at 180 rpm.

PA and PT were quantified using commercially purchased compounds as standards, whereas NF, PDG, and PARA were quantified using phlorizin as a standard. PA, PT, and phlorizin were purchased from Tokyo Chemical Industry. The mean and standard error of the mean were calculated for triplicate experiments.

Analysis of synthesized compounds

The synthesized compounds were analyzed using HPLC [37]. The mobile phases for PA and PARA were 80% water (containing 0.1% formic acid) and 20% acetonitrile (containing 0.1% formic acid) at 0 min. The linear gradient of acetonitrile from 20 to 90% was used between 12 and 15 min. The mobile phases for PT, NF, and PDG started and ended with 10% acetonitrile instead of 20%. To determine the structure of the NF using NMR [38], the reaction product was purified using thin layer chromatography (TLC silica gel 60 F254; Millipore, Burlington, MA, USA). The mobile phase was a mixture of ethyl acetate, acetone, formic acid, and H₂O (10:1:1:2, v/v). The spot corresponding to ND was scraped off from the plate, and the synthesized ND was eluted with methanol. Silica was separated by centrifugation. Methanol was evaporated, and the dried compound was dissolved in DMSO. Nuclear magnetic resonance spectra were recorded with a JNM-ECZ500R (Tokyo, Japan) at 500 MHz; ¹H NMR (500 MHz, DMSO-d6): δ 2.77 (t, J = 7.6 Hz, 2 H), 3.13–3.18 (m, 3 H), 3.23 (t, J = 7.9 Hz, 2 H), 3.43 (d, J = 11.4 Hz, 1H), 3.65 (d, J = 11.8 Hz, 1H), 3.88 (t, J = 9.0 Hz, 1H), 4.53 (d, J = 9.2 Hz, 1H), 5.95 (s, 1H), 6.67 (d, J = 8.4 Hz, 2 H), 7.02 (d, J = 8.3 Hz, 2 H). The proton NMR spectroscopy results for NF matched those in previously published literature [39]. The molecular masses of the synthesized PDG and PARA were determined using ultra-performance liquid chromatography-high-resolution mass spectrometry [40]. The MS/MS fragmentation patterns were compared with those of published data [27, 30].

Results

Synthesis of PA

We synthesized PT from PA by using a process that differs from the flavonoid synthesis process, as flavonoids are typically synthesized from p-coumaroyl-CoA and PT is synthesized from phloretoyl-CoA. For the synthesis of PT, the conversion of p-coumaric acid to PA was required. Enoate reductase (ER) catalyzes this reaction; ER from C. acetobutylicum reduces p-coumaric acid to PA [41]. The ER gene was cloned and introduced into E. coli (B-PA1), and the B-PA1 synthesized PA when p-coumaric acid was provided. The molecular weight of the synthesized PA molecule was 165.0560 Da (Supplementary Fig. 2A), which corresponded to that of standard PA. The next step was to synthesize PA without providing p-coumaric acid. p-Coumaric acid was previously synthesized from tyrosine in E. coli by using TAL [31]. The TAL gene (SeTAL) was transformed into E. coli along with ER, and the resulting transformant (B-PA2) was tested for PA production. B-PA2 strain synthesized PA; however, no p-coumaric acid was detected (Fig. 2). These results indicate that PA was successfully synthesized in the B-PA2 strain without the need for p-coumaric acid, and the final titer of PA might increase with an increased supply of tyrosine. For this, we used the shikimate pathway module constructs known to increase the final titer of various phenolic compounds derived from chorismate or tyrosine [37, 42, 43]. Six different constructs containing various combinations of genes in the shikimate pathway and a control plasmid were introduced into the E. coli strain B-PA2. The PA titers in the resulting transformants (B-PA3–B-PA9) were measured, and all E. coli strains showed an improvement in PA. Among them, strain B-PA7, into which the five genes aroL, aroG^fbr, ppsA, tktA, and tyrA^fbr had been incorporated, synthesized the highest titer of PA (72.8 mg/L), followed by B-PA6 (65.2 mg/L), B-PA5 (55.1 mg/L), B-PA8 (52.3 mg/L), B-PA9 (46.6 mg/L), B-PA4 (23.2 mg/L), and B-PA3 (21.2 mg/L) (Fig. 3A). No p-coumaric acid was detected in the filtrate from any of the strains, indicating that the PA titer could be enhanced by increasing the tyrosine concentration. We used E. coli mutants, BT and BTP, which produce more tyrosine compared to that produced by the wild-type [31]. The BT strain had tyrR deleted, while the BTP strain had both tyrR and pheA deleted. These mutant strains were transformed with the plasmids pC-SeTAL-ER and pA-aroL-aroG^f-ppsA-tktA-tyrA^f, respectively. The resulting transformants (BT-PA7 and BTP-PA7) showed a significant increase in PA titer compared with that of the parent strain (B-PA7). The E. coli strain BTP-PA7 synthesized 150.1 mg/L PA, which was 1-fold higher than that of the parent strain (72.8 mg/L), whereas BT-PA7 synthesized 125.3 mg/L (Fig. 3B). We monitored the production of PA by BTP-PA7 for 72 h and found that 310.9 mg/L (1871.0 µM) of PA was synthesized. In a recent study, an E. coli mutant containing ptsG, which encodes membrane-integral permease EIICB consuming phosphoenolpyruvate, tyrR, and pheA in addition to TAL from Flavobacterium johnsoniaeu and ER from C. acetobutylicum, synthesized 693.87 mg/L PA in 36 h [20].

Fig. 2

HPLC analysis of the reaction product from the E. coli strain B-PA2. (A) HPLC profile of standard PA (B) HPLC profile of reaction product in B-PA2 strain

Fig. 3

Selection of shikimate pathway module constructs (A) and E. coli mutant (B) for PA production. The mean and the standard error of the mean were calculated for triplicate experiments

Synthesis of PT

Phloretin is synthesized by the condensation of phloretoyl-CoA with three molecules of malonyl-CoA, which is mediated by two enzymes, 4CL and CHS. To examine the synthesis of PT from PA, two genes, Os4CL and PeCHS, were introduced into the E. coli strain B-PT1. These strains were used to test the synthesis of PT using 100 µM of PA. Analysis of the culture filtrate using HPLC revealed a peak corresponding to PT but no peak corresponding to PA (Supplementary Fig. 1B). Its molecular mass identified using MS was 273.0782 Da (Supplementary Fig. 2B), which corresponded with the predicted molecular mass of the standard PT. In addition, an unidentified peak with a molecular mass of 231.0662 Da was detected as a major product. The molecular mass corresponds to the reaction intermediate dihydro-bisnoryangon, formed when two molecules of malonyl-CoA attach to PA. This intermediate has been previously reported during PT biosynthesis [18, 20]. This result indicates that the supply of malonyl-CoA was insufficient, leading to the formation of a reaction intermediate having only two malonyl-CoA molecules as a major product. To increase the supply of malonyl-CoA, accABCD was introduced into E. coli through pC-Os4CL-PeCHS. This enzyme increases malonyl-CoA concentration by converting acetyl-CoA to malonyl-CoA and increases flavonoid synthesis [34, 37, 44]. The resulting E. coli strain B-PT2 synthesized 6.9 µM of PT from 100 µM PA (Fig. 4C), which was 8.9-fold higher than that produced by B-PT1 (0.7 µM). However, an intermediate major peak was still observed. We observed that the conversion of PA to PT was approximately 5%, whereas the conversion to the intermediate was approximately 25%, irrespective of the initial PA concentration (100, 200, 300, 400, 500, or 600 µM). As PT is synthesized from PA and malonyl-CoA, the supplementation with malonyl-CoA was insufficient to convert all PA into PT. Alternatively, accABCD could have been inhibited by PA, PT, and/or the reaction intermediate(s). To increase the supply of malonyl-CoA, we used other genes (matB and matC) that affect flavonoid biosynthesis in E. coli [18]. The strain B-PT3, which overexpresses matBC along with the other genes in B-PT1, synthesized 45.1 µM of PT from 100 µM PA when 2 g/L malonate (19.2 mM) was supplied, and no PA peak was detected. Additionally, the peak for the intermediates decreased notably, as observed using HPLC (Supplementary Fig. 4D). Figure 4 showed the amount of intermediate dihydro-bisnoryagon and PT synthesized from 100 µM PA. Some of the synthesized PT seemed to be degraded (See below). Therefore, the strain B-PT3 was supplied with various concentrations (100, 200, 300, 400, 500, and 600 µM) of PA, and 112.5 µM of PT was obtained from 300 µM PA. No unreacted PA was observed. However, when 400 µM PA was supplied, 115.6 µM of PT was synthesized, and 29.1 µM of unreacted PA was detected.

Fig. 4

Effect of different malonyl-CoA suppliers on the production of PT from PA. The strain B-PT2 harbors accABCD and the strain B-PT3 contains matBC. 100 µM of PA was supplied in E. coli. The synthesized dihydro-bisnoryangon and PT were indicated as gray bars and white bars, respectively

We found that PT degraded in the culture medium, which explains the low yield of 112.5 µM PT from 300 µM PA. We tested PT degradation by incubating 100 µM of PT overnight in various culture media (YM9 [pH 7.0], M9 [pH 7.0], YM9 [pH 5.7], and distilled water). The results showed that 90% of PT degraded in YM9 [pH 7.0] medium, 80% in YM9 [pH 5.7] medium, and 40% in distilled water, whereas only 20% degraded in M9. This result indicates that the pH and yeast extract caused the degradation of PT. Next, various concentrations of yeast extracts (0%, 0.2%, 0.4%, 0.6%, 0.8%, and 1%) were used for the synthesis of PT from 400 µM of PA using the strain B-PT3 to determine the optimal concentration for balancing PT synthesis and degradation. The E. coli used only one-fourth of the supplied PA without yeast extract, which was much lower than that used when 1% yeast extract was present (approximately three-fourths of PA was used). We observed that PT synthesis was reduced to 2.2 µM when M9 containing 0% yeast extract was used and that it gradually increased up to 137.4 µM when 1% yeast extract was used. This indicates that the amount of PT synthesis at higher yeast extract concentrations exceeded the amount of PT degradation. Furthermore, adjusting the pH of YM9 from 7.0 to 5.7 with 50 mM 2-(N-morpholino) ethanesulfonic acid (MES) buffer slightly decreased the degradation rate (from 90 to 80%). Thus, PT stability increased as the pH decreased (Zhang et al., 2018). Therefore, the use of yeast extract was shown to be more advantageous for PT synthesis because it grows at a lower pH than E. coli. Two E. coli strains, B-PT2 (housing accABCD) and B-PT3 (housing matBC), were used to synthesize PT at different pH levels in the media supplemented with 100 µM PA. B-PT2 synthesized 4.4 µM of PT at pH 7.0, which was the pH of M9 medium, whereas 11.7 µM of PT was synthesized at pH 5.7, suggesting that lower pH increases the PT titer. However, B-PT3 synthesized 45.1 µM of PT at pH 7.0 and 26.4 µM at pH 5.7 with observed cell death. This might be attributed to the excessively low pH, as B-PT3 was supplied with additional malonate. In conclusion, we used the B-PT3 cells under normal media conditions for further experiments because this yielded the maximum PT titer.

Next, we attempted PT synthesis from p-coumaric acid. Four genes (ER, 4CL, CHS, and matBC) were introduced into E. coli (B-PT4), and PT synthesis was examined. Unexpectedly, chalcone was primarily synthesized, and only 3.4 µM of PT was obtained from 100 µM p-coumaric acid. The higher production of naringenin compared to that of PT was likely due to the difference in the conversion rate between Os4CL and ER; the conversion of p-coumaric acid into p-coumaroyl-CoA by Os4CL was much faster than that of PA by ER, and the higher production of p-coumaroyl-CoA resulted in increased synthesis of naringenin. To eliminate the possibility of converting p-coumaric acid into p-coumaroyl-CoA and prevent naringenin synthesis, we bifurcated the pathway into two portions: the first, from tyrosine to PA, and the second, from PA to PT. Genes associated with each portion of the pathway were introduced into different E. coli strains. Therefore, to synthesize PT from tyrosine, the BTP-PA7 strain was cultured in YM9 to synthesize PA. The PA concentration in the culture filtrate was adjusted to approximately 300 µM with the fresh YM9, which was the maximum PA concentration used in the previous experiments. Next, the B-PT3 strain was resuspended in the filtrate to synthesize PT. This method of synthesis yielded 9.9 mg/L of PT (36.0 µM) with 29.1 µM of unreacted PA remaining. However, the quantity of PT thus obtained was much lower than that obtained when PA was used. Therefore, the strain harboring accABCD would be better suited for PA synthesis in a stepwise culture because this strain used a higher initial PA concentration (> 600 µM). Therefore, the PA concentration in the culture filtrate was adjusted to 476.9 µM when the strain BTP-PA6 was used. Using YM9 containing B-PT2, 12.8 mg/L PT (46.8 µM) was synthesized (Fig. 6), demonstrating that the strain harboring accABCD was better than that harboring matBC in the stepwise synthesis of PT.

Synthesis of PT C-glycosides

We synthesized three PT C-glycosides: NF, PDG, and PARA. NF can be synthesized from PT via glucosylation. C-glucosyltransferase from F. esculentum was transformed into E. coli to produce strain B-NF1 and test NF synthesis. The culture filtrate from strain B-NF1 was analyzed using HPLC, and a new peak was detected (Fig. 5A). This compound was purified using thin-layer chromatography (TLC), and its structure was determined to be NF using proton nuclear magnetic resonance (NMR) spectroscopy (see the Methods section). We then investigated the ability of FeCGT to convert PT to NF in E. coli. The B-NF1 strain was treated with various concentrations of PT (100, 200, 300, 400, 500, and 600 µM). The results showed that B-NF1 converted 600 µM of PT into 408.2 µM of NF. Another study using FeCGT and sequential addition of PT (200 µM PT in every 1.5 h) showed a six-fold higher final yield with > 90% efficiency [25]. To synthesize NF from tyrosine, the BTP-PA7 strain was used to synthesize PA, and fresh YM9 medium was added to the reaction medium to adjust the PA concentration to approximately 300 or 476.9 µM (for matBC or accABCD, respectively). Similar to the synthesis of PT, the strain housing accABCD produced a higher PT titer than the strain housing matBC. The strain B-NF2, which housed matBC in addition to the other genes, synthesized 3.1 mg/L of NF (7.1 µM) and 4.1 mg/L of PT (15.1 µM); 151.7 µM of PA remained unreacted. The strain B-NF3 housing accABCD synthesized 26.1 mg/L of NF (59.9 µM), the titer of which was 7.4-fold higher than that of the strain housing matBC. No unreacted PA or PT were observed (Fig. 6).

Fig. 5

Synthesis of three PT C-glycosides from PT using engineered E. coli. HPLC condition of PARA starts with 10% ACN in this result. (A) HPLC profile of reaction product with 600 µM PT in B-NF1, (B) reaction product with 300 µM PT in B-PDG1, and (C) reaction product with 100 µM PT in B-PARA1

Fig. 6

The production of PT, NF, PDG, and PARA from tyrosine in E. coli using stepwise culture. Two different malonyl-CoA suppliers (matBC in A and accABCD in B) were used. The first cell (E. coli strain BTP-PA7) was used to synthesize PA, and the second cells, which indicated at the X-axis of figures, were used to synthesize PT, NF, PDG, and PARA. The optimal amount of PA was fed to the cell for the synthesis of each compound. Not only the final product but also the reaction intermediate(s) were shown. Triplicate experiments were carried out to calculate the mean and the standard error of the mean

Previous studies [26] have reported that PDG can be synthesized from PT through di-glucosylation. The E. coli strain B-PDG1, transformed with C-glucosyltransferase from Zea mays (ZmCGT), was reacted with PT to synthesize PDG. The reaction product was analyzed, and the new peak was purified using HPLC (Fig. 5B). The molecular mass of the reaction product was 597.1904 Da, which was the predicted molecular mass of PDG. Additionally, the tandem mass spectrometry (MS/MS) fragmentation pattern (Supplementary Fig. 3) of the reaction product matched that of the published data (Ren et al., 2020), indicating the reaction product to be PDG. ZmCGT was tested to determine the maximum amount of PT that could be converted into PDG by the E. coli strain B-PDG1. The B-PDG1 strain was treated with various concentrations of PT (100, 200, 300, 400, 500, and 600 µM), and the reaction filtrate was analyzed. We found that 300 µM of PT yielded 91.5 µM of PDG and 10.1 µM of NF with no residual PT remaining. However, at a higher PT concentration of 400 µM, 99.6 µM PDG and 18.8 µM NF were synthesized with 1.0 µM of the unreacted PT remaining. Therefore, we concluded that approximately 300 µM of PT was optimal for the synthesis of PDG. We used stepwise synthesis to synthesize PDG from tyrosine, and PA was synthesized using the BTP-PA7 strain. Then, the B-PDG2 strain (housing matBC) was used as the second whole-cell biocatalysis for PDG synthesis from 300 µM PA. In total, 3.2 mg/L of PDG (5.2 µM), 2.0 mg/L of NF (4.6 µM), and 1.4 mg/L of PT (5.1 µM) were synthesized, leaving 83.3 µM of unreacted PA. When the B-PDG3 strain (housing accABCD) was used as the second whole-cell biocatalysis, 30.0 mg/L of PDG (50.2 µM) was synthesized; moreover, unreacted PA, PT, and NF were not observed when 476.9 µM PA was supplied (Fig. 6). Thus, PDG was successfully synthesized from tyrosine using stepwise synthesis, and the strain housing accABCD was more efficient at PDG synthesis compared to the matBC-housing strain.

Furthermore, PARA can be synthesized from PT through arabinosylation. C-glycosyltransferase from O. sativa was transformed into E. coli housing AtUXS, Ecugd, and OsUGE, which participate in UDP-Ara synthesis. The resulting B-PARA1 strain reacts with PT. We performed HPLC analysis of the reaction filtrate to confirm the presence of PARA (Fig. 5C), which has a molecular mass of 405.1249 Da. The MS/MS fragmentation pattern (Supplementary Fig. 3) matched that of the published PARA [30], indicating that the synthesized product was PARA. The strain B-PARA2, housing AtUXS, Ecugd, OsUGE, and OsCGT in one plasmid, was supplied with various concentrations of PT (100, 200, 300, 400, 500, and 600 µM) to determine the maximum amount of PT that could be converted into PARA. The strain B-PARA2 synthesized 74.2 µM of PARA when 100 µM PT was provided. However, it produced 145.4 µM of PARA with 4.0 µM of residual PT remaining when 200 µM of PT was supplied. PARA was also synthesized from tyrosine using a two-cell system. We synthesized PA as described above. Strain B-PARA3, housing Os4CL, PeCHS, matBC, AtUXS, Ecugd, OsUGE, and OsCGT, was used to synthesize PARA from PA. The culture filtrate was analyzed, which showed that it contained 1.7 mg/L of PARA (4.1 µM), 7.7 mg/L of PT (28.0 µM), and approximately 184.7 µM of unreacted PA. Conversely, B-PARA4 housing accABCD instead of matBC synthesized 18.1 mg/L PARA (44.4 µM) and 4.8 mg/L PT (17.6 µM) with 13.5 µM of unreacted PA remaining (Fig. 6). This indicates that PARA was synthesized from tyrosine. For PT-glycoside synthesis, accABCD, rather than matBC, appears to be a better malonyl-CoA supplier.

Discussion

We used E. coli to synthesize PA, PT, NF, PDG, and PARA. Previous study using engineered E. coli was mainly bioconversion, and the final yield of PT when it was synthesized from glucose was very low (20 µg/L) [20]. The current study showed the possibility to synthesize five compounds from glucose with an improved final yield. PA was synthesized from tyrosine using TAL and ER. In order to maximize the synthesis of PA, we screened the shikimate module constructs and E. coli mutants (ΔtyrR and ΔtyrR/ΔpheA) which are known to enhance the synthesis of tyrosine. It turned out that the shikimate module construct containing aroL, aroG^f, ppsA, tktA, and tyrA^f showed the best field yield and E. coli mutant deleting both tyrR and pheA was better than the wild type. The use of shikimate module constructs and E. coli mutant to improve the final yield of target compound has been carried out other studies [35, 43,44,45,46]. For the synthesis of PT, and its glycosides, we did not use the shikimate module construct because the downstream of PA was likely to be a rate-limiting. The higher concentration of PA did not enhance the final yield of PT and its glycosides. Therefore, we determined the best PA concentration for PT an its glucosides synthesis and adopted to synthesize them. PA synthesis might be enhanced by use the another mutant such as deletion of typE, which block the tryptophan synthesis. But, the field goal of this study was to synthesis of PT and its glycosides. The yield of PA from the current study was overwhelmed for the synthesis of PT and its glycosides. The final yield of PA was less than the previous studies (Table 2). But, the final goal of the current study was to synthesize PT and its glycosides. The major hindrance in PT synthesis was the specificity of the enzyme for phloretoyl-CoA. In previous studies, CHS was identified as an enzyme with high affinity for phloretoyl-CoA [15, 16]; however, CHS has an equally high affinity for p-coumaroyl-CoA. Hence, structural modification of CHS would likely enhance the preference of CHS for phloretoyl-CoA. However, a previous study showed that the structural changes resulted only in a marginal increase in preference for phloretoyl-CoA over p-coumaroyl-CoA [20]. Therefore, metabolic channeling and/or tight compartmentalization within the site dedicated to flavonoid and PT biosynthesis may be a valid synthesis strategy in plants. However, we used E. coli as a microbial factory; we used two engineered E. coli cells, each harboring different modules for the synthesis of PT from tyrosine. The first E. coli strain contained constructs for the synthesis of PA from tyrosine, while the other contained constructs for the synthesis of PT from PA. This approach provided artificial compartmentalization for the synthesis of PA and PT, preventing the synthesis of naringenin instead of PT. The final yield of PT in this study was 12.8 mg/L, which was much higher than the previous study using E. coli (20 µg/L) but less than the final yield using S. cerevisiae (42.7 mg/L) (Table 2). It seemed that degradation of PT in S. cerevisiae is less than E. coli. The synthesis of NF was better in S. cerevisiae (59 mg/L) [19] than E. coli (26.1 mg/L). No attempt to synthesize PDG and PARA from glucose has been tried except the current study.

Table 2 Synthesis of PA, PT and PT glycosides using microbial system

We observed PT degradation in the E. coli culture medium (M9 containing 1% yeast extract), which appeared to be influenced by yeast extract and/or pH. We used M9 with various concentrations of yeast extract or adjusted the pH up to 5.7 to prevent the degradation of PT. The results indicated that yeast extract exerted a greater effect on PT synthesis than on degradation, and a low pH reduced the degradation of PT but inhibited E. coli survival. In the attempted stepwise culture, the pH of the culture medium for the synthesis of PA using the BTP-PA7 strain was adjusted to approximately 5.8. At this pH, the degradation of PT was lower than that at the initial medium pH of 7.0.

To synthesize PT glycosides, the optimal PT concentration was first determined using various PT concentrations with respect to E. coli housing GT. A higher initial PT concentration did not result in a higher production of PT glycosides. In fact, a PT concentration that was higher than the optimal concentration for a given amount of E. coli lowered the final PT glycoside titer. This was used to determine the optimal PT concentration for the synthesis of PT glycosides from tyrosine.

We tested two genes (accABCD and matBC) to increase malonyl-CoA concentration in E. coli. matBC, which comprised malonyl-CoA synthase and malonate carrier protein, proved to be a better malonyl-CoA supplier when utilizing commercial PA for the synthesis process. Moreover, E. coli overexpressing matBC demonstrated an approximately 9-fold higher quantity of PT compared to E. coli overexpressing accABCD. However, this dynamic changed when PT and its glycosides were synthesized from tyrosine. In this case, accABCD was a better supplier of malonyl-CoA. This could be attributed to the inhibitory effect of unidentified metabolite(s) generated during the synthesis of PA on matB and/or matC. The degree of inhibition of accABCD was much lower.

By using stepwise synthesis and adjusting culture conditions, we successfully synthesized PT and three PT derivatives. We optimized phloretin microbial biosynthesis by improving enzyme affinity, identifying the gene that increased the output, refining the production design with a stepwise culture approach, and analyzing the culture conditions conducive to maximum output and the prevention of product degradation. This study contributes valuable insights to future research methodologies in the microbe-based synthesis of PT derivatives, representing a step forward in the synthetic production of plant-based compounds that are useful in health and wellness.