De novo biosynthesis of myricetin, kaempferol and quercetin in Streptomyces albus and Streptomyces coelicolor

Flavonols are a flavonoid subfamily widely distributed in plants, including several ones of great importance in human and animal diet (apple, tomato, broccoli, onion, beans, tea). These polyphenolic nutraceuticals exert potent antimicrobial (membrane potential disruptors), antioxidant (free-radical scavengers), pharmacokinetic (CYP450 modulators), anti-inflammatory (lipoxygenase inhibitors), antiangiogenic (VEGF inhibitors) and antitumor (cyclin inhibitors) activities. Biotechnological production of these nutraceuticals, for example via heterologous biosynthesis in industrial actinomycetes, is favored since in plants these polyphenols appear as inactive glycosylated derivatives, in low concentrations or as part of complex mixtures with other polyphenolic compounds. In this work, we describe the de novo biosynthesis of three important flavonols, myricetin, kaempferol and quercetin, in the industrially relevant actinomycetes Streptomyces coelicolor and S. albus. De novo biosynthesis of kaempferol, myricetin and quercetin in actinomycetes has not been described before.


Introduction
Flavonoids (from Latin flavus, yellow) are a family of about 6000 nutraceuticals widely distributed in plant cells, many of them found in dietary plants [1][2][3][4][5]. All flavonoids have a generic chemical structure consisting of 15 carbon atoms (C6-C3-C6): two aromatic rings (rings A and B) connected by a heterocyclic pyran C which contains one oxygen (ring C, Fig 1) [6][7][8][9][10][11]. This basic skeleton can have multiple substituents, such as hydroxyl or methyl groups, as well as sugars [12]; indeed, chemical modifications in ring C lead to the formation of more than 9,000 flavonoid derivates [13].
Depending on the pattern of hydroxylation and the substituents on the heterocyclic ring C, flavonoids can be classified into several sub-groups, but in this paper we will focus on flavonols. They are an important subfamily, as some of their members, like myricetin, kaempferol and quercetin, represent the major intake of dietary flavonoids in most societies [2,3,9]. PLOS  generating dihydroquercetin (taxifolin) (Fig 1), which finally is transformed in quercetin by the action of flavonol synthase 1 (FLS1) [12]. Myricetin, kaempferol, quercetin and their dihydro precursors show interesting nutraceutical activities, as it has been described above. This makes these bioactive compounds attractive targets for genetic and metabolic engineering experiments, like the heterologous expression of their plant biosynthetic gene pathways in suitable microbial factories such as actinomycetes. In this work, we have carried out this by using combinatorial biosynthesis, where genes from different organisms are grouped in an artificial gene cluster directing the production of the natural bioactive compound [34,38]. Previous studies have reported the biosynthesis of flavonols in microorganisms; for instance, myricetin has been already heterologously produced, after feeding with naringenin precursor, in a strain of E. coli containing an incomplete flavonol biosynthetic gene cluster [39]. Also, kaempferol has been already produced in E. coli [40] and Saccharomyces cerevisiae [41], needing in some cases feeding with coumaric acid or naringenin [39]. In the case of its heterologous production in Streptomyces venezuelae, feeding with naringenin was also necessary [42]. Finally, quercetin has been also produced in E. coli and in S. cerevisiae after feeding with coumaric acid or naringenin [39,41]. In this work, we have achieved de novo production of these three flavonols by means of biosynthetic pathways heterologously expressed in S. albus and S. coelicolor, without feeding with precursors.

Material and methods
Bacterial strains, plasmids and culture conditions E. coli TOP10 (Invitrogen) and pUC57 (Fermentas) were used for routine sub-cloning while E. coli ET12567 [43] was required to obtain non-methylated DNA for later protoplasts transformation in Streptomyces coelicolor M1154 [44]. The high-copy number E. coli-Streptomyces shuttle vector pIAGO, a derivative of pWHM3 which contains the strong constitutive promoter for ermE � (P ermE � ) [45], was used as expression plasmid. The strain Streptomyces albus J1074 [46] was also used for the production of flavonols.
E. coli strains were grown in TSB liquid broth or TSB agar, supplemented with the corresponding antibiotics (ampicillin 100 μg/ml, Sigma Aldrich) for plasmid selection. S. coelicolor M1154 and S. albus J1074 were grown at 30˚C in YEME 34% and 17% sucrose respectively, for protoplasts preparation. Both species were sporulated on SFM and Bennet medium respectively [47], supplemented with the corresponding antibiotics when necessary (thiostrepton 50 μg/mL).
For flavonols production, S. albus and S. coelicolor clones were grown on 3 ml of solid R5A medium [48], supplemented with the corresponding antibiotic, during 5 days at 30˚C. Spores were previously quantified and an inoculum of 10 7 spores/mL was used for each culture.

DNA manipulation
Restriction enzymes were purchased from Takara Biochemicals, T4 DNA ligase from Thermo Scientific, and Dream Taq DNA Polymerase from Thermo Scientific. Synthetic genes for the following ORFs were generated by Genscript after codon optimization:

Construction of plasmids for flavonoids production
All constructed plasmids described below were verified by restriction enzymes digestions and also by sequencing of the cloned regions. Streptomyces producing clones were confirmed by PCR. Primers used amplify the first two common genes: 5'-GTGATCGAGCTGGACATGAA-3' as the forward primer and 5'-GGCGTCCACGAGGTGC-3' as the reverse primer.

Construction of pKF
The plasmid pKF contains the ermE � promoter (P ermE � ) and the 6 genes responsible for kaempferol biosynthesis. All synthetic gene cassettes were independently cloned in pUC57 and plasmids were named pLMF1 (pUC19 containing TAL gene), pLMF2 (4CL), pLMF3 (CHS), pLMF-FLS (FLS1), pLMF5 (CHI) and pLMF-N3DOX (N3DOX) ( Table 1). Additionally, TAL gene was subcloned into vector pSL1180 as HindIII-BamHI (pLMF7) to start with the cloning strategy. 4CL gene (from pLMF2) was cloned into pLMF7 as PstI-BamHI gene cassette, generating pLMF8. Next step was subcloning FLS1 gene cassette from pLMF-FLS1 into pLMF3 as an EcoRI DNA fragment, giving rise to pKF11. The correct orientation of each DNA fragment was always confirmed by restriction enzymes digestions and sequencing. The two gene cassettes from pKF11 (CHS and FLS1) were subcloned together into pLMF8 as SacI-BamHI DNA band, in order to get the first 4 genes together in a plasmid (pKF14). Finally, N3DOX gene was subcloned into pLMF5 (opened EcoRV-BamHI) as an EcoRI (blunt ended)-BamHI gen cassette and the two genes together (CHI and N3DOX) were subcloned into pKF14 as XbaI-BamHI resulting in the generation of pKF17, which contains the 6 genes required for kaempferol biosynthesis. As the expression host was Streptomyces, a further subcloning was required, and the BglII-BamHI DNA fragment carrying the 6 genes was finally subcloned into pIAGO plasmid, a derivative of the bifunctional replicative vector pWHM3, which contains the ermE � promoter, giving rise to the final plasmid pKF.

Construction of pQR
The plasmid pQR contains the ermE � promoter (P ermE � ) and the 7 genes required for the biosynthesis of quercetin. To obtain this plasmid, it was required to add one more gene to the previous plasmid pKF17, synthetized to produce kaempferol. This gene, F3'H, was cloned into pKF17 as DraI-BamHI gene cassette giving rise to pQR2. The 7 genes contained in this plasmid were subcloned in the vector pIAGO to be further expressed in Streptomyces. The gene cassette was cloned as BglII-BamHI DNA fragment to obtain the final plasmid pQR ( Table 1).

Construction of pMYR
The plasmid pMYR contains the ermE � promoter (P ermE � ) and the 7 genes required for the biosynthesis of myricetin. To have the 7 genes together in the same vector, it was necessary to clone the F3'5'H gene into the previously constructed plasmid pKF17. The gene was cloned as DraI-BamHI gene cassette giving rise to pMYR2. The 7 genes contained in this plasmid were subcloned in the vector pIAGO to be further expressed in Streptomyces. The gene cassette was cloned as BglII-BamHI DNA fragment to obtain the final plasmid pMYR ( Table 1).

Construction of pREC4 for malonyl-CoA metabolic engineering in S. albus
The plasmid pREC4 is a derivative of the E. coli-Streptomyces bifunctional vector pSEVA98c1 (colE1 and pIJ101 origins of replication for E. coli and Streptomyces respectively, both high copy number; apramycin resistance gene aac(3)IV). Vector pSEVA98c1 was digested with PacI-SacI in order to introduce the P ermE � promoter for Streptomyces, giving rise to pREC3. Finally, pREC3 was digested with BamHI-HindIII, in order to introduce the 5.1 kb DNA fragment containing the S. coelicolor chromosomal genes birA (biotin ligase, SCO4927 gene), accA2 (alpha subunit of acetyl-CoA carboxylase, SCO4921 gene) and accBE (beta and epsilon subunits of acetyl-CoA carboxylase, SCO5535 and SCO5536 genes respectively) [49][50][51]. These plasmids were amplified by PCR from S. coelicolor chromosomal DNA, using the following primers, which contain sequences for restriction enzymes at their ends (marked in cap-

Extraction and analysis of flavonols
S. albus J1074 and S. coelicolor clones harboring pKF, pQR, pMYR and pIAGO (negative control), or pREC4 (for malonyl-CoA metabolic engineering) were cultivated (three replicas for each strain were extracted and quantified separately  1), filtrated (0.2 μm PVDF) and analyzed by liquid chromatography-electrospray ionization mass spectrometry (LC-E-SI-MS/MS, Agilent technologies 1290 Infinity, Triple Quadrupole), which was carried out using a Zorbax Eclipse Plus C18 column (50 mm x 2.1 mm, 1.8 μm) in the negative ion mode. The analytes were eluted at a flow rate of 0.3 mL/min using a gradient of 0.1% (v/v) formic acid in water (A) and 0.1% (v/v) formic acid in acetonitrile (B) at 0-10% of B for 1 min, which was increased to 35% for 3 min and maintained at 35% for 1 min, then increased to 80% for 3 min and maintained at 80% for 2 min and finally decreased to 10% for 1 min. Flavonoids quantification was carried out in multiple reaction monitoring (MRM) mode in MS/MS. To accomplish this, the following ion sets were selected to detect the transitions of the parent ions to the product ions specific to the analytes:

Heterologous expression of kaempferol
In microorganisms, the activity of four enzymes, TAL, 4CL, CHS and CHI, is required to produce naringenin, which is the main flavonols precursor (Fig 1). To obtain the flavonol kaempferol, the activity of other two enzymes is also needed. The first enzyme, N3DOX, hydroxylates naringenin in the position C3 to form the immediate kaempferol precursor, dihydrokaempferol. Later, FLS1 catalyzes de formation of a double bond between C2 and C3 to finally obtain kaempferol (Fig 1). In this work, all the six synthetic genes encoding for the enzymes required for the biosynthesis of kaempferol (with codon usage adapted to the transcription characteristics of Streptomyces) were cloned in a replicative high-copy number shuttle vector for E. coli-Streptomyces, under the control of P ermE � (see Materials and Methods section). The final plasmid, pKF, was transformed and expressed in two different species of Streptomyces: S. albus and S. coelicolor.
Cultures of S. albus-pKF and S. coelicolor-pKF in R5A solid medium were analyzed by HPLC-MS chromatography in multiple reaction monitoring (MRM) in MS/MS mode, in order to identify and quantify the final product, kaempferol, as well as its intermediate precursors naringenin and dihydrokaempferol. The presence of the hydroxylated form of kaempferol, quercetin, was also analyzed. In S. coelicolor-pKF no kaempferol was detected. However, low production levels (below 0.1 μM) of the precursor, dihydrokaempferol, and the flavonol quercetin were observed (Fig 2A, Table 2).
In S. albus-pKF, kaempferol was detected at 0.212 μM, but also dihydrokaempferol was detected below 0.1 μM and traces of quercetin (0.2 μM) (Fig 2B, Table 2). The presence of kaempferol, quercetin, myricetin and their precursors, naringenin and dihydrokaempferol was analyzed in parallel in the negative controls (S. albus-pIAGO and S. coelicolor-pIAGO). No flavonoids were detected in any case in these negative control strains.

Heterologous expression of quercetin
Quercetin is a hydroxylated form of kaempferol. For its biosynthesis, it is only necessary the activity of one extra enzyme (F3'H), able to hydroxylate kaempferol in the position B3'. After checking that our S. albus host were able to produce kaempferol, the gene encoding for the F3'H was cloned into the plasmid directing the biosynthesis of kaempferol. The new plasmid, pQR, was transformed in both S. albus and S. coelicolor protoplasts. The cultures from positive recombinant strains (in R5A solid medium) were analyzed by HPLC-MS chromatography.
In S. coelicolor-pQR, quercetin and one of the intermediate precursor, dihydrokaempferol were observed, although at levels below 0.1 μM (Fig 3A, Table 2).

Heterologous expression of myricetin
The flavonol myricetin, is also a hydroxylated form of kaempferol. In this case, there are two extra hydroxylations, while in quercetin there is only one. These two extra hydroxyl groups are in the positions B3' and B5', so the activity of a specific enzyme, F3'5'H, is required (Fig 1). The gene encoding this enzyme was added to the previous construction directing the biosynthesis of kaempferol and the new plasmid was further transformed in S. coelicolor and S. albus.
In S. coelicolor-pMYR, myricetin was detected, but also the presence of dihydrokaempferol and quercetin was observed (Fig 4A). However, the levels of production were very low (under 0.1 μM was detected in MRM analysis).
S. albus-pMYR was transformed also with plasmid pREC4, which contains all the genes coding for malonyl-CoA biosynthesis enzymes (BirA and acetyl-CoA carboxylase subunits). In this strain, however, no higher production levels of myricetin were detected in liquid cultures, but 0.3 μM of the shunt product apigenin, a derivative of naringenin (the precursor of dihydrokaempferol, kaempferol and myricetin) was detected. Apigenin was not detected in S. albus-pMYR-pSEVA98c1 (control strain for metabolic engineering experiments). The same strain, S. albus-pMYR-pREC4 was cultivated in R5 liquid medium as well, but including a feeding with naringenin. In this case, no further increase in apigenin production levels was detected, but interestingly, very low amounts of luteolin (a 3'-hydroxylated derivative of apigenin) were detected.

Discussion
In this work, two different strains of Streptomyces, S. albus and S. coelicolor, have been able to biosynthesize de novo the flavonols kaempferol, quercetin (a 3'-hydroxylated kaempferol), myricetin (a double hydroxylated form of kaempferol) and the main precursor of the flavonols pathway, dihydrokaempferol. Previous works reported the biosynthesis of kaempferol in E. coli [39,40], Saccharomyces cerevisiae [41] and S. venezuelae [42]. Two strategies were followed to produce kaempferol in E. coli. The first one consisted of adding L-Tyrosine (the first precursor of flavonoids in prokaryotes) to the culture medium, so the complete biosynthetic pathway was cloned. The levels of production obtained were up to 15 mg/L [40]. The group of Leonard et al. did not clone the gene encoding for the first enzyme of the biosynthetic pathway (tyrosine ammonia lyase, TAL), which means that feeding the cultures with precursors was needed. When p-coumaric acid (second metabolite of the biosynthetic pathway in microorganisms) was added, the yields of kaempferol reached 0.3 mg/L. The productions achieved were better (0.8 mg/L) when naringenin was used for feeding the culture [39].
Kaempferol was also produced in yeast in a similar way that it was synthetized in E. coli. The complete biosynthetic pathway was cloned in S. cerevisiae and three different precursors were added to the cultures: L-phenylalanine (first precursor of flavonoids in eukaryotes), pcoumaric acid and naringenin. The production levels achieved were 1.3 mg/L, 0.9 mg/L and 4.6 mg/L respectively [41].
Regarding the biosynthesis of kaempferol in Streptomyces, only one group succeeded to do it previously. However, the complete biosynthetic pathway was not cloned, only the genes encoding for the N3DOX and FLS1 were cloned in S. venezuelae. These enzymes are involved in the hydroxylation of naringenin to obtain dihydrokaempferol (N3DOX), and in the further formation of a double bond to form kaempferol (FLS1). So, it was required to supplement the cultures with naringenin precursor. The production levels reached 0.2 mg/L [42]. In this work we were able to produce kaempferol by cloning the complete biosynthetic pathway into S. albus without feeding the cultures with any precursor. However, the levels achieved were 0.212 μM. Also, smaller levels of its precursor dihydrokaempferol were observed, indicating that the enzyme FLS1 is not completely efficient. In S. coelicolor, no kaempferol was detected but small amounts of its precursor, dihydrokaempferol, and its hydroxylated derivative, quercetin, were detected. This may be due to the presence of an extra P450 hydroxylase naturally found in this strain that is able to use kaempferol as a substrate, thus, converting it into quercetin. This affirmation is supported by the fact that P450 systems are really well developed in Streptomyces genus, and P450 from this actinomycete has been described for regioselective hydroxylation of diverse flavonoids [52][53][54].
Regarding quercetin biosynthesis, it was achieved by the same groups in both E. coli and S. cerevisisae following the strategy employed to produce kaempferol [39,41]. In E. coli, the complete biosynthetic pathway, except for the first gene (encoding for TAL), was cloned, and cultures supplemented with either p-coumaric acid or naringenin. The production levels were 0.05 mg/L and 0.18 mg/L respectively [39].
Quercetin production in S. cerevisiae was higher. In this species, all the genes involved in quercetin biosynthesis were cloned. Like in the case of kaempferol production, the cultures were feed with L-Phe, p-coumaric acid and naringenin. Nonetheless, production levels were smaller. In the case of L-Phe feeding there were only traces of quercetin. When p-coumaric acid and naringenin were added, the production was up to 0.26 mg/L and 0.38 mg/L respectively [41].
In this paper, we describe the biosynthesis of quercetin in Streptomyces for the very first time. Moreover, it was de novo biosynthesis as no precursors were added to these cultures. However, only traces of quercetin were detected in the recombinant strain of S. coelicolor. In the case of S. albus, 0.1 mg/L of quercetin were observed, as well as relatively high amounts of kaempferol and dihydrokaempferol, indicating the incomplete efficiency of the P450 hydroxylases employed in this study. Although the amount of quercetin produced in this work is rather small, it is comparable to the one achieved in E. coli after being feed with naringenin [39], evidencing that our biosynthetic system is more effective.
As far as the biosynthesis of myricetin in microorganisms is concerned, this flavonol was only produced by the group of Leonard et al., in E. coli, and all the genes required to produce myricetin were introduced in E. coli and then, the cultures were supplemented with naringenin and eriodictyol. Despite this feeding, production levels only reached 0.01 mg/L [39].
In our case, we demonstrate the feasibility of de novo myricetin biosynthesis in both S. albus and S. coelicolor. Moreover, production levels (0.146 μM) in S. albus were better than those achieved in E. coli even after feeding experiments. In S. albus, not only myricetin was detected by HPLC-MS, but also high levels of its precursor, quercetin (1.984 μM). This elevated amount of precursor in comparison to the final compound may be due to a 3'-hydroxylation activity of the enzyme F3'5'H in the C3'. In fact, it is known that P. hybrida F3'5'H performs both 3'-and 3',5'-hydroxylation reactions and can use flavonols as well as dihydroflavonols and flavanones as a substrate [39,55]. Taking into account these two considerations, the high yield of quercetin could be due to a 3'-hydroxylation of the intermediate dihydrokaempferol generating dihydroquercetin as a product which could be easily converted in quercetin by the FLS1 enzyme expressed in pMYR plasmid. Moreover, more quercetin can be produced from kaempferol due to, once again, a 3'-hydroxylation of the F3'5'H. For all these reasons, only a little quantity of kaempferol is available to be converted in myricetin by F3'5'H. Finally, it should be pointed out that P. hybrida F3'5'H has a broad substrate specificity towards dihydrokaempferol, kaempferol and quercetin but competition as well as inhibition may occur when more than one substrate is present, leading to a lower myricetin yield and quercetin accumulation [55].
In our experiments with S. coelicolor there are only traces of these compounds. These results, together with the obtained ones for the biosynthesis of kaempferol and quercetin, reveal that S. albus is a better host for flavonols biosynthesis than S. coelicolor. Also, low production levels in S. albus could be improved by metabolic engineering of the strain, facilitating the incorporation of malonyl-CoA to the flavonols biosynthetic pathway, as it is a limiting factor in flavonoid production [56]. Other authors confirmed that this strategy is useful to increase flavonoids yields [37,40,57,58]. However, the experiments with S. albus-pMYR-pREC4, where plasmid pREC4 was used to try to further increase intracellular malonyl-CoA precursor levels, and therefore to generate higher myricetin titers, were unsuccessful. In these experiments, instead of higher myricetin levels, due to the expected higher malonyl-CoA intracellular levels, a deviation from naringenin precursor towards apigenin (a shunt product in this study) was observed. A possible explanation for this result is that the enzyme N3DOX, in charge of converting naringenin intermediate towards hihydrokaempferol, shows a 79.18% identity with the enzyme FNS, which usually converts naringenin towards apigenin [59,60]. This means that genes present in pMYR can explain the generation of apigenin in S. albus-pMYR-pREC4 as a shunt product from naringenin. Once apigenin is present at those levels in S. albus-pMYR-pREC4, further feeding here with exogenous naringenin facilitates apigenin production and the low levels of the apigenin 3'-hydroxy derivative luteolin observed. This 3'hydroxylation can be easily explained by the presence of F3'5'H hydroxylase in pMYR.
Alternatively, production levels could be also improved by facilitating hydroxylation steps carried out by F3'H and F3'5'H enzymes. These two enzymes need reducing power provided by a single redox partner (cytochrome P 450 enzyme) whose presence is critical to show their optimal or maximal activities [39,54], and this reductase is not included in our plasmid constructions. In fact, there is a big hurdle to combine specific P450s with the right redox partners. Nevertheless, the partial activity of the P450 hydroxylases used in this study could be explained by a recognition of soluble endogenous redox partners since P450 systems are specially well developed in Streptomyces genus [52,54]. Further modifications of these two enzymes involving their transformation in a soluble chimera protein that fuses the P450s with suitable P450 reductases are being carried out.

Conclusions
Using a combinatorial biosynthesis approach and reconstituted plant flavonol pathways in the bacteria S. albus, this work describes the heterologous biosynthesis of the important nutraceuticals myricetin and quercetin by the first time in actinomycetes, according to public literature. Also, kaempferol biosynthesis has been achieved in this bacterium for the first time without feeding with precursors. These experiments open the way to heterologous production in actinomycetes of other flavonols and flavonoids.