Biocatalytic synthesis of flavones and hydroxyl-small molecules by recombinant Escherichia coli cells expressing the cyanobacterial CYP110E1 gene

Background Cyanobacteria possess several cytochrome P450s, but very little is known about their catalytic functions. CYP110 genes unique to cyanaobacteria are widely distributed in heterocyst-forming cyanobacteria including nitrogen-fixing genera Nostoc and Anabaena. We screened the biocatalytic functions of all P450s from three cyanobacterial strains of genus Nostoc or Anabaena using a series of small molecules that contain flavonoids, sesquiterpenes, low-molecular-weight drugs, and other aromatic compounds. Results Escherichia coli cells carrying each P450 gene that was inserted into the pRED vector, containing the RhFRed reductase domain sequence from Rhodococcus sp. NCIMB 9784 P450RhF (CYP116B2), were co-cultured with substrates and products were identified when bioconversion reactions proceeded. Consequently, CYP110E1 of Nostoc sp. strain PCC 7120, located in close proximity to the first branch point in the phylogenetic tree of the CYP110 family, was found to be promiscuous for the substrate range mediating the biotransformation of various small molecules. Naringenin and (hydroxyl) flavanones were respectively converted to apigenin and (hydroxyl) flavones, by functioning as a flavone synthase. Such an activity is reported for the first time in prokaryotic P450s. Additionally, CYP110E1 biotransformed the notable sesquiterpene zerumbone, anti-inflammatory drugs ibuprofen and flurbiprofen (methylester forms), and some aryl compounds such as 1-methoxy and 1-ethoxy naphthalene to produce hydroxylated compounds that are difficult to synthesize chemically, including novel compounds. Conclusion We elucidated that the CYP110E1 gene, C-terminally fused to the P450RhF RhFRed reductase domain sequence, is functionally expressed in E. coli to synthesize a robust monooxygenase, which shows promiscuous substrate specificity (affinity) for various small molecules, allowing the biosynthesis of not only flavones (from flavanones) but also a variety of hydroxyl-small molecules that may span pharmaceutical and nutraceutical industries.

In order to function as terminal monooxygenases, P450s must be associated with one or two additional proteins (or protein domains) to transfer two electrons from NAD(P)H to the heme domain of the P450 protein [6,7]. The vast majority of bacterial P450s need a FADcontaining ferredoxin reductase to receive electrons from NAD(P)H, and a ferredoxin (small iron-sulfur protein) to receive them from ferredoxin reductase, which subsequently reduces P450 itself [class I system; Additional file 1: Figure S1a] [7]. P450RhF (CYP116B2) derived from Rhodococcus sp. NCIMB 9784 was discovered to be a self-sufficient P450 protein, in which the P450 domain is Figure 1 Phylogenetic positions of cyanobacterial CYP110 family using 27 CYP110 proteins derived from 11 cyanobacteria, whose amino acid sequences are shown in CyanoBase. The accession numbers in parentheses show those of respective P450 proteins. CYP110E1 is shown in boldface. The phylogenetic tree was constructed using the neighbor-joining method. The number shown next to each node indicates the percentage bootstrap value of 1,000 replicates (only 50% and higher are cited). The scale bar indicates a genetic distance of 0.02 (Knuc). C-terminally fused to a reductase domain (here called RhFRed) [8]. RhFRed contained an FMN-binding ferredoxin reductase subdomain to receive electrons from NAD(P)H and a [2Fe-2S] ferredoxin subdomain [9]. A short linker region of 16 amino acids existed between P450 and RhFRed [8]. The redox chain of P450RhF resembles that of the class I system [Additional file 1: Figure S1c]. Thus, vector pRED was constructed for the functional expression of bacterial P450 (class I) genes in Escherichia coli, using the linker sequence and RhFRed domain sequence [(Additional file 1: Figure S2] [10]. This vector has been demonstrated to be useful for functional expression of the P450cam gene (CYP101A1) [10,11], P450bzo gene (CYP203A) [10], P450balk gene (CYP153A13a) [12,13], other CYP153A genes [14], and P450 PikC gene [15], constituting corresponding selfsufficient P450 monooxygenation enzymes. We elucidate here that the CYP110E1 gene is functionally expressed on pRED in E. coli to synthesize a robust monooxygenase, which shows promiscuous substrate specificity (affinity) for various small molecules.

Screening experiments
Nostoc sp. strain PCC 7120, Nostoc punctiforme PCC 73102, and Anabaena variabilis ATCC 29413 possess six, ten, and four P450s, respectively. We screened the biocatalytic functions of these P450s using 47 small molecules that contain flavonoids, sesquiterpenes, lowmolecular-weight drugs, naphthalene derivatives, and other chemicals with benzene rings [Additional file 1: Figure S3)]. E. coli BL21 (DE3) cells carrying each P450 gene inserted into the pRED vector were co-cultured with the substrates and possible bioconversion products were analyzed by HPLC. Consequently, CYP110E1 of Nostoc PCC 7120 was found to be promiscuous for the substrate range mediating the biotransformation of various small molecules. The CYP110E1 enzyme that is C-terminally fused to RhFRed was confirmed to constitute the active P450 form by CO difference spectral analysis [Additional file 1: Figure S4]. Thus, cells of E. coli BL21 (DE3) carrying plasmid pCYP110E1-Red were used for the following experiments.
The molecular formula of S-2 was determined to be C 15 H 24 O 2 by HREI-MS. Consistent with its molecular formula and 1 H-NMR spectrum, the introduction of an alcoholic OH group in S-1 was proposed. The position of the alcoholic OH group was clarified to be C-13 by the observation of an oxymethylene signal (δ H 3.92 and δ H 4.02, H-13) and the 1 H-13 C long range coupling from this oxymethylene to C-5 (δ C 35.0), C-6 (δ C 140.1), and C-7 (δ C 126.0). The identity of S-2 was thus determined as (6Z,10E)-6-hydroxymethyl-2,9,9-trimethylcycloundeca-2-ene-1-one ( Figure 4). This product (S-2) was a novel compound according to the CAS database.
The molecular formula of S-3 was determined to be C 15 H 24 O 2 by HREI-MS. Consistent with its molecular formula and 1 H-NMR spectrum, the introduction of an alcoholic OH group in S-1 was proposed. The position of the alcoholic OH group was clarified to be C-8 by the observation of an oxymethylene signal (δ H 4.24, H-8) and the 1 H-13 C long range coupling from H-14 (δ H 1.12) and H-15 (δ H 1.26) to C-8 (δ C 75.5). The identity of S-3 was thus determined as (6E,10E)-8-hydroxy-2,6,9,9tetramethylcycloundeca-2,6-dien-1-one ( Figure 4). This product (S-3) was a novel compound according to the CAS database.

Bioconversion of aryl compounds by E. coli (pCYP110E1-Red)
A variety of aryl compounds, which include naphthalene derivatives and low-molecular-weight drugs, were biotransformed through co-cultivation with cells of E. coli (pCYP110E1-Red). Converted compounds were identified by chromatographic and spectroscopic analyses. Spectroscopic data are shown in Additional file 2.
The molecular formula of A-3 was determined to be C 11 H 10 O 2 by HRAPCI-MS. Consistent with its molecular formula and 1 H-NMR spectrum, the introduction of one phenolic OH group in the aromatic ring was proposed. The position of this phenolic OH group was determined to be C-2 by the observation of 1 H-1 H vicinal spin coupling of H-3 (δ 7.23, d, J = 8.5 Hz) and H-4 (δ 7.57, d, J = 8.5 Hz) and 1 H-13 C long range couplings from H-4 to C-2 (δ 145.4) and C-5 (δ 128.3). The identity of A-3 was thus determined to be 1-methoxynaphthalen-2-ol ( Figure 5) [19].
The molecular formula of A-4 was determined to be C 12 H 12 O 2 by HRAPCI-MS. Consistent with its molecular formula and 1 H-NMR spectrum, the introduction of one phenolic OH group in the aromatic ring was proposed. The position of this phenolic OH group was determined to be C-4 by the observation of 1 H-1 H vicinal spin coupling of H-2 (δ 7.55, d, J = 8.7 Hz) and H-3 (δ 7.23, d, J = 8.7 Hz) and the 1 H-13 C long range couplings from H-5 (δ 7.78) to C-4 (δ 145.8). The identity of A-4 was thus determined to be 4-ethoxynaphthalen-1-ol ( Figure 5) [20].
A compound converted from 7-ethoxycoumarine The EtOAc extract from the bioconversion mixture (0.5 ml) with 7-ethoxycoumarine and E. coli (pCYP110E1-Red) was subjected to HPLC to yield a product (A-12). A-12 was identified as 6-hydroxy-2 H-chromen-2-one ( Figure 6) with HPLC by its comparison with an authentic sample.
The molecular formula of A-16 was determined to be C 16 H 15 FO 3 by HRAPCI-MS. Consistent with its molecular formula and 1 H-NMR spectrum, the introduction of one phenolic OH group in the aromatic ring was proposed. The position of this phenolic OH group was determined to be C-13 because the signals of H-12 and H-14 were observed to be doublet (J = 8.6 Hz) and at high field (δ 6.89). The identity of A-16 was thus determined to be methyl 2-(2-fluoro-4'-hydroxy-[1,1'-biphenyl]-4-yl) propanoate (Figure 7).

Discussion
The phylogenetic analysis (Figure 1) showed that CYP110E1 of Nostoc sp. strain PCC 7120 was located in close proximity to the first branch point in the phylogenetic tree of the CYP110 family, i.e., among 27 CYP110 proteins derived from 11 cyanobacteria, only four P450s, CYP110K1, CYP110D2, CYP110E6, and CYP110D3, were located closer to the first branch point than CYP110E1. In this study, CYP110E1, whose function had been unknown, was found to function as a substratepromiscuous monooxygenase when it was C-terminally fused to the RhFRed reductase domain of P450RhF (CYP116B2) by the use of the pRED vector. Naringenin was converted directly to apigenin with a significant conversion ratio (31.5%) (Figure 2). Naringenin and apigenin belong to typical flavonoids that can be biosynthesized in higher plants. Artificial flavonoids, flavanone, 6-hydroxyflavanone and 7-hydroxyflavanone, were also converted to the respective flavones, even if with low conversion ratios, i.e., major parts of the substrates remained without being biotransformed even after 48-h co-culture. These results revealed that CYP110E1 functions as a flavone synthase. Such an activity is reported for the first time in prokaryotic P450s. When using flavanone as the substrate, 3-hydroxyflavanone was additionally generated (Figure 3). It may be possible that 3-hydroxyflavanone is the intermediate from flavanone to flavone. However, E. coli (pCYP110E1-Red) did not biotransform 3-hydroxyflavanone when it was added as the substrate (data not shown). It is therefore likely that flavones and 3-hydroxyflavanone were generated independently from flavanone. The higher-plant CYP93B was characterized as P450-type flavone synthase (FSII) and was proposed to convert flavanones to flavones by way of 2hydroxyflavanones [23,24]. Such a catalytic route may be the case with CYP110E1.
P450BM3 (also described as P450BM-3 or P450 BM3 ; CYP102A1) derived from Bacillus megaterium, one of the best-characterized prokaryotic P450s, is a natural fusion enzyme composed of a P450 part and a eukaryotetype NADPH-P450 reductase domain [Additional file 1: Figure 1d] [25][26][27]. This P450 part is closely related to the CYP110 family [1] and exhibited 24.6% amino acid sequence identity to CYP110E1. P450BM3, whose native substrates are thought to be long-chain fatty acids, has been shown to possess substrate and catalytic promiscuities [28][29][30][31]. Specifically, P450BM3 variants incorporating active site mutations that include F87V (or F87A) were found to acquire broader substrate affinity not only for a variety of aryl compounds including substituted naphthalenes [17,32,33], but also for the monoterpene α-pinene and the sesquiterpene amorpha-4,11diene [30,34]. The P450BM3 variant F87V that was Nterminally fused to an archaeal peptidyl-prolyl cis-trans isomerase (PPIase), which was synthesized by the plasmid named pFusionF87V in E. coli cells, was shown to elevate the stability of the P450 protein [17].
Zerumbone, a sesquiterpene contained in the shampoo ginger (Zingiber zerumbet Smith), is a promising chemopreventive agent, since its anti-inflammatory and antitumor activities have been investigated [35][36][37][38]. E. coli BL21 (DE3) carrying plasmid pCYP110E1-Red was shown to hydroxylate zerumbone to produce two novel compounds (S-2 and S-3) via the endogenous metabolite in E. coli (S-1; Figure 4). On the other hand, E. coli BL21 (DE3) carrying plasmid pFusionF87V was not able to synthesize compound S-3 from zerumbone, although it bioconverted zerumbone to compound S-2 with a higher conversion ratio than that of E. coli (pCYP110E1-Red) (Figure 4). E. coli (pFusionF87V) was able to hydroxylate β-eudesmol (a sesquiterpene contained in edible plants of the Zingiberaceae family) at its C-5 position [17], while E. coli (pCYP110E1-Red) was not able to biotransform β-eudesmol (data not shown). The two prokaryotic P450s, CYP110E1 and P450BM3 (F87V), are likely to have the ability to biotransform some sesquiterpenes of higher-plant origin complementally. CYP109B1 from Bacillus subtilis was also found to possess a wide substrate range for saturated fatty acids, n-alcohol, and some isoprenoids, and convert the sesquiterpene (+)-valencene to (+)-nootkatone, a high addedvalue compound found in grapefruit juice [39,40].
The present study also showed that E. coli (pCYP110E1-Red) biotransformed the methylester forms of nonsteroidal anti-inflammatory drugs ibuprofen and flurbiprofen to produce the respective hydroxyl derivatives that are difficult to synthesize chemically (Figure 7). E. coli (pUCRED-Balk) converted ibuprofen methylester to another hydroxyl substituent [13], while it was not able to biotransform flurbiprofen methylester (data not shown). The four drug metabolites produced by recombinant E. coli cells (Figure 7) or their free carboxylate forms, awaiting the determination of the absolute configuration, could be used as standards in studies on the metabolisms of ibuprofen and flurbiprofen with human P450s [41].

Conclusion
The present study revealed that cyanobacterial cytochrome P450 CYP110E1, C-terminally fused to the P450RhF (CYP116B2) RhFRed reductase domain, is promiscuous for substrate and catalytic ranges and is useful for biosynthesizing not only flavones (from flavanones), but also a variety of hydroxyl-small molecules that are difficult to synthesize chemically, which may span pharmaceutical and nutraceutical industries.

Bacterial strains and genetic manipulation
Three cyanobacterial strains, Nostoc sp. strain PCC 7120, Nostoc punctiforme PCC 73102 (=ATCC 29133), and Anabaena variabilis ATCC 29413 were obtained from Pasteur Culture collection, Paris, and grown autotrophically in BG 11 medium. For the isolation of genomic DNA, cyanobacteria were harvested from the log phase and were immediately treated with lysozyme (10 mg/ml for 1 h). Genomic DNA was then isolated with the GenElute plant genomic DNA kit (Sigma-Aldrich, St. Louis, MO).
E. coli DH5α (ECOS Competent E. coli DH5α; Nippon Gene, Tokyo, Japan) was utilized as the host for DNA manipulations. E. coli BL21 (DE3) (Nippon Gene) was used for the functional expression of each P450 gene, which was inserted into the pRED vector [10]. PCR amplifications were performed using Prime STAR Max Premix DNA polymerase (Takara Bio, Ohtsu, Japan) and a thermal cycler (Applied Biosystems, Foster City, CA). Restriction enzymes and DNA-modifying enzymes were purchased from New England BioLabs (Beverly, CA) or Takara Bio. A Ligation-Convenience Kit (Nippon Gene) was also used. Plasmid DNA was prepared with a QIAprep Miniprep Kit (Qiagen, Hilden, Germany). All recombinant DNA experiments were carried out according to the suppliers' manuals or Sambrook and Russell (2001) [42].

Nucleotide sequencing and computer analysis
Nucleotide sequences were confirmed with Bigdye terminator cycle sequencing ready reaction kit version 3.1 (Applied Biosystems) and a model 3730 DNA analyzer (Applied Biosystems). Homologous protein sequences in the protein sequence database were retrieved from Cyano-Base of Kazusa DNA Research Institute (http://genome. kazusa.or.jp/cyanobase) with the BLAST program [43], and aligned by Clustal W program in Molecular Evolutionary Genetics Analysis (MEGA) software version 5.0 (http://www.megasoftware.net/). A phylogenetic tree was also constructed according to MEGA 5.0.

Construction of plasmids
Cyanobacterial P450 genes were amplified by PCR from genomic DNA of Nostoc sp. PCC 7120, N. punctiforme PCC 73102, or A. variabilis ATCC 29413. All synthetic oligonucleotides used in this work were listed in Table 1. PCR amplification was performed in a 50 μl reaction mixture containing 25 ng of genomic DNA, 25 μl of 2 × the DNA polymerase, 10 μM of each primer, and 5% dimethyl sulfoxide (DMSO). The PCR conditions used were the following: preincubation at 98°C for 2 min; a total of 5 cycles at 98°C for 10 sec, 55°C for 10 sec, and 75°C for 15 sec; a total of 30 cycles at 98°C for 10 sec, 62°C for 5 sec and 75°C for 15 sec. An amplified 1.4 kb fragment was digested with NdeI and EcoRI or HindIII, and ligated into the NdeI-EcoRI or NdeI-HindIII site of pRED to construct the desired plasmids. In these plasmids, the stop codons of the respective P450 genes were removed to fuse the N-terminus of RhFRed.
CO difference spectral analysis CO difference spectral analysis was done as described [12].

Bioconversion experiments
E. coli BL21 (DE3) carrying each plasmid was grown in an LB medium including ampicillin (Ap; 100 μg/ml) at 37°C with rotary shaking for 3-4 h until the absorbance at OD 600 nm reached approximately 0.8. For screening experiments, 1.5 ml of this preculture was inoculated into 125 ml of an LB medium including Ap (100 μg/ml), 5-aminolevulinic acid hydrochloride (5-ALA; 80 μg/ml), ammonium iron (II) sulfate (0.1 mM), and IPTG (0.1 mM) in a baffled Erlenmeyer flask, and cultured at 20°C for 20 h on a rotary shaker (140 rpm; Kuhner Shaker Lab Therm LT-X, Basel, Switzerland). Cells were collected by centrifugation and resuspended in 25 ml of CV-3 buffer [sodium phosphate buffer (50 mM, pH 7.2) containing 5% glycerol] in a baffled Erlenmeyer flask. Five hundred μl of this cell suspension was added into each well of a 96 well sterile plate (PR-Master Block 2ML; Greiner Bio-One, Frickenhausen, Germany), together with 1 mM (final concentration) of substrate dissolved in dimethyl sulfoxide (DMSO). Bioconversion was performed by cultivation at 25°C for 24 h with 300 rpm using the Kuhner Shaker.
For structural determination of products, large scale cultivation was carry out, by inoculation of 5 ml of the preculture into 500 ml of LB medium including Ap (100 μg/ml), 5-ALA (80 μg/ml), ammonium iron (II) sulfate (0.1 mM), and IPTG (0.1 mM) in a baffled Erlenmeyer flask at 20°C for 20 h with 140 rpm on the Kuhner Shaker. Cells were collected by centrifugation, and resuspended in 100 ml of CV-3 buffer in a baffled Erlenmeyer flask. Each substrate dissolved in DMSO was added at a final concentration of 1 mM to the cell suspension and bioconversion was performed by cultivation at 25°C for 48 h with 180 rpm.
Substrates and authentic samples used in this study were purchased from Tokyo Chemical Industry Co. (Tokyo, Japan), Sigma-Aldrich Co. (St. Louis, MO), or Wako Pure Chemical Industries (Osaka, Japan).

Extraction and HPLC analysis of products
Five hundred μl of the reaction mixture liquid was added to 100 μl of saturated sodium chloride solution and 500 μl of EtOAc and shaken for 5 min. After centrifugation, the organic phase was analyzed by high pressure liquid chromatography (HPLC; Waters 2695, Waters Corp., Milford, MA) equipped with an on-line photodiode array detector (Waters 2996). An aliquot of the organic phase (20 μl) was applied to HPLC and separated using an XTerra MS C 18 column (I.D. 4.6 mm × 100 mm; Waters), and a flow rate of 1 ml/min was used, with solvent A [5% acetonitrile (CH 3 CN) in 20 mM phosphoric acid] for 3 min, then a linear gradient from solvent A to solvent B (95% CH 3 CN in 20 mM phosphoric acid) for 25 min, and finally with solvent B for 15 min, with the maximum absorbance being monitored in the range of 200-500 nm (max plot).