Polyphenols are one of the largest and widely natural products distributed in plant cells, since more than 10,000 phenolic compounds have been described so far in higher plants, with several 100 found in edible plants (Manach et al., 2004; Tsao, 2010; Li et al., 2014). Flavonoids (Latin flavus, “yellow”) represent 60% of these polyphenols and cover more than 6000 compounds ubiquitously distributed in plants (Chaudhuri et al., 2013; González-Vallinas et al., 2013).

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, Figure 1) (Kumar and Pandey, 2007; Verma and Pratap, 2010; Liu, 2013; Ravishankar et al., 2013; Fantini et al., 2015). This basic skeleton can have multiple substituents, as hydroxyl groups as well as sugars (Crozier et al., 2009). 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 focus on two groups: flavones as luteolin (present in Thai chili, broccoli and leaves of onion and celery) and apigenin (present in parsley, celery, onion, garlic, red pepper, and chamomile tea); and flavanones (as eriodictyol found in citrus fruits such as orange, lemon, or grapefruit) (Manach et al., 2004; Tsao, 2010; Liu, 2013).

The biological activities of these compounds depend both on their structural differences and glycosylation patterns (Kumar and Pandey, 2007; Tsao, 2010). In fact, diet flavonoids can promote health and prevent certain diseases such as cancer by acting against cell oxidation processes and by stopping cell degradation and aging (Luo et al., 2015).

The effects of flavonoids in humans can be attributed to antioxidant, antitumor, anti-inflammatory, antimicrobial, anticancer, and cardiovascular actions. For example, eriodictyol performs a key role against the pathogenesis of diabetes mellitus, inhibiting immunoglobulin E/antigen-produced type I hypersensitivity. This flavanone also shows antinociceptive, hyperthermic, and anticancer effects (Zhu L. et al., 2014). Apigenin has a growing importance due to its inhibitory effect on the PI3K/Akt/mTOR signaling pathways (Tong and Pelling, 2013). Luteolin is able to inhibit angiogenesis or to reduce tumor size in vivo since it is able to modulate protein kinases and growth factor receptors as EGFRs or VEGFRs (López-Lázaro, 2006). Also, the antioxidant capacity of these and other polyphenols acts as master key in preventing numerous diseases associated with oxidative damage such as cardiovascular, neurodegenerative and cancer (Ravishankar et al., 2013; Hollman, 2014).

Flavonoids are synthesized in planta by the phenylpropanoid pathway, which converts L-Phe in 4-coumaroyl-CoA in three steps, the common intermediate for all flavonoids (Figure 1) (Falcone Ferreyra et al., 2012). These first three common steps are catalyzed by phenylalanine ammonia lyase (PAL), cinnamate 4-hydroxylase (4CH) and 4-coumaroyl CoA ligase (4CL) (Figure 1). Then the chalcone synthase (CHS) condenses a molecule of 4-coumaroyl-CoA with three molecules of malonyl-CoA, generating naringenin chalcone, the basic skeleton for all flavonoids (Tsao, 2010; Wang et al., 2011; Falcone Ferreyra et al., 2012). The heterocycle C closure is catalyzed by chalcone isomerase (CHI), which generates naringenin, the precursor for apigenin and luteolin (flavones) and for eriodictyol (flavanone). In order to generate apigenin, the flavone synthase (FNS) oxidoreductase is required. Apigenin is then the substrate for the flavonoid 3′-hydroxylase (F3′H), giving rise to luteolin (Figure 1). On the other hand, naringenin is also the substrate for F3′H, generating eriodictyol without FNS activity (Figure 1) (Crozier et al., 2009).

Flavonoids are valuable compounds for use in the pharmaceutical, chemical and nutraceutical industry, but the amounts that can be obtained directly from plants are very limited. One solution is the chemical synthesis, but when dealing with metabolites with complex chemical structures, their production is economically not feasible in most cases. Therefore, an attractive alternative is to express these plant biosynthetic gene pathways in microbial factories, using combinatorial biosynthesis, where genes from different organisms are grouped in an artificial gene cluster directed to the production of the natural bioactive compound (Falcone Ferreyra et al., 2012; Trantas et al., 2015). Microorganisms that are most commonly used as cell factories for flavonoids are Saccharomyces cerevisiae, Escherichia coli and in some cases Streptomyces venezuelae, where different techniques are applied in order to increase production levels.

However, in the last years, focus has been made in de novo processes for flavonoids microbial production, without adding expensive precursors to the culture medium, therefore allowing cheap production platforms necessary for broad commercial use of flavonoids as nutraceuticals (Fowler and Koffas, 2009; Wang et al., 2011; Wu et al., 2014). In this work, we have achieved de novo production of apigenin, luteolin, and eriodictyol by means of synthetic biosynthetic pathways heterologous expressed in Streptomyces albus. The use of Gram-positive bacterial factories, as this actinomycete, facilitates the further industrialization of bioactive compounds for the pharmaceutical industry, as on the one hand, actinomycetes are the most diverse and rich producers of bioactive compounds (antibacterials, antifungals, antitumors, herbicides, immunosuppressors, etc.), belonging to many different biochemical families (lactams, polyketides, non-ribosomal peptides, terpenoids, flavonoids, etc.), and providing all the necessary metabolic precursors and biosynthetic machinery for these purposes, including also all post-translationally modification enzymes (Newman and Cragg, 2007). Actually, about two thirds of all known bioactive polyketides are produced by actinomycetes. This huge biosynthetic capability is accompanied also by a huge diversity in resistance mechanisms to allow production of such a wide range of bioactive compounds (Zhang et al., 2008; Hwang et al., 2014). This role as useful bacterial factories for bioactives is reinforced in the case of some actinomycetes as Saccharopolyspora erythraea, S. coelicolor, S. lividans, S. avermitilis, S. venezuelae and S. albus, as their genomes have been sequenced and multiple genetic manipulation tools are available for them. Specifically, in the case of S. albus, this host is more convenient as it shows a fast and nice disperse mycelial growth, instead of the most common mycelial dense aggregates shown in the case of other actinomycetes. This disperse growth facilitates scaling-up at industrial bioreactors, as less complications arise from local cellular density and cell lysis. Also, S. albus can be transformed with methylated plasmid DNA, in contrast to other actinomycetes, which also facilitates genetic engineering in this species (Kieser et al., 2000).

In this work, S. albus has been able to de novo biosynthetize four flavonoids: naringenin (a flavanone precursor in this family of nutraceuticals), apigenin (its oxidized flavone derivative), luteolin (a 3′-hydroxylated apigenin) and eriodictyol (a 3′-hydroxylated version of the flavanone naringenin).

De novo naringenin production has been previously described in recombinant strains of E. coli (Hwang et al., 2003; Miyahisa et al., 2006; Santos et al., 2011), S. cerevisiae (Yan et al., 2005; Trantas et al., 2009), and S. clavuligerus (Álvarez-Álvarez et al., 2015). Other authors also achieved its production in other microorganisms, but after feeding with precursors (Park et al., 2009).

Apigenin was previously produced de novo only in E. coli (Miyahisa et al., 2006). Here, the complete biosynthetic pathway was cloned in E. coli and production levels reached up to 13 mg/L. However, the addition to culture medium of the main precursor L-tyrosine at 543 mg/L (3 mM) was required (Miyahisa et al., 2006) (Table 3). In a more recent work, apigenin yields were increased to 23 mg/L in E. coli. But, as in the work developed by Miyahisa et al. (2006), its synthesis needed the supplementation with p-coumaric acid (Lee et al., 2015). Higher production levels in E. coli, 110 mg/L, have been described after feeding with p-coumaric acid and malonate (Leonard et al., 2008). This flavone was also produced in S. venezuelae at 1.5 mg/L, but only after feeding with 0.5 mM (136 mg/L) naringenin a S. venezuelae strain which contained only the gene coding for FNS oxidoreductase (Park et al., 2010) (Table 3).

In this work, however, the complete biosynthetic pathway for apigenin was heterologous expressed in S. albus. Besides, it was not supplemented with any precursor, although the levels of apigenin obtained (0.089 mg/L) were smaller than in other previously described works. Only small amounts (0.001 mg/L) of its immediate precursor naringenin were detected, indicating that the activity of the plant oxidoreductase enzyme FNS is good in this microorganism.

In the case of eriodictyol, according to the literature, its biosynthesis in actinomycetes had not been described before. However, two research groups have reported production of this flavanone in E. coli from glucose or after feeding with p-coumaric acid plus malonate, caffeic acid or L-tyrosine, reaching production levels of 50, 5.70, 0.02, and 42.6 mg/L respectively (Leonard et al., 2007, 2008; Zhu S. et al., 2014) (Table 3). In both cases, metabolic engineering in E. coli was carried out afterward to enhance the availability of the flavonoids precursor malonyl-CoA, increasing the eriodictyol production levels to 52 and 107 mg/L respectively (Leonard et al., 2007; Zhu S. et al., 2014) (Table 3). In our case, despite the low eriodictyol production levels in S. albus, this biosynthesis is de novo and has been achieved in an actinomycete for the first time. This low production levels, as well as the accumulation of its precursors naringenin and p-coumaric acid, show a low activity for the hydroxylase F3′H (see below).

In the case of the flavone luteolin, its heterologous biosynthesis has been described in E. coli (4 mg/L) and S. cerevisiae (2 mg/L) after feeding with p-coumaric acid plus malonate, or caffeic acid respectively (Leonard et al., 2005, 2008) (Table 3). In our case, we demonstrate its de novo biosynthesis in Streptomyces. Nonetheless, as in the case of eriodictyol, luteolin yields were low. This was expected after the observed eriodictyol results, as luteolin biosynthesis requires also the activity of F3′H hydroxylase (Figure 1).

These experiments showed that F3′H enzyme in S. albus is not able to completely transform naringenin into eriodictyol nor apigenin into luteolin. In S. albus-pLUT, no eriodictyol was observed, indicating that the FNS enzyme is quite more efficient than F3′H, as apigenin is detected in higher amounts here than eriodictyol. This idea is supported also by the fact that eriodictyol levels in S. albus-pERI were almost 300 times lower than apigenin levels in S. albus-pAPI. Other authors have described also that P450 hydroxylases like F3′H from eukaryotic origin may present low solubility in prokaryotic hosts, as in this case S. albus, because a membrane anchorage domain may not be suitable in prokaryote systems (Park et al., 2011). The hydroxylase F3′H is a cytochrome P450-dependent monooxygenase which requires its integration into the membrane of the endoplasmic reticulum of plant cells as well as the presence of a P450-reductase carrying electrons from a donor to a NADPH P450 heme-core complex. Functional expression of these plant membrane-bound enzymes in microorganisms is a challenge due to several factors, such as protein insolubility (prokaryotes do not have endoplasmic reticulum membranes to bound) and cofactor incorporation (Burbulis and Winkel-Shirley, 1999; Chang et al., 2007; Fowler and Koffas, 2009). It has been shown that microbial expression of this family of enzymes is enhanced by engineering P450 reductases, in a way that these reductases are fused to the corresponding P450-dependent monooxygenase, and the membrane binding domain of these ones is deleted to increase solubility in prokaryotic cytoplasm. Following this strategy, soluble and functional chimeras have been obtained (Leonard et al., 2006; Leonard and Koffas, 2007; Zhu S. et al., 2014). These reasons for the monooxygenases counterparts could explain the very low F3′H activity shown in this study as shown with the low lutein and eriodictyol levels in S. albus. Therefore, these studies in S. albus will pave the way in order to conduct further enzyme engineering in this type of enzymes, to achieve higher production titers (Zhu S. et al., 2014) (Table 3).

Finally, we have seen accumulation of p-coumaric acid in our three producing strains, the second metabolite in the pathway (after TAL activity on L-Tyr), which is common to all flavonoid biosynthetic pathways. p-Coumaric acid is converted to coumaroyl-CoA by the action of 4CL enzyme, which requires malonyl CoA to synthetize the corresponding flavonoid (naringenin chalcone in our case, Figure 1). Malonyl-CoA is therefore a limiting factor in flavonoid production, as it is also precursor for fatty acids biosynthesis (Winkel-Shirley, 2001). In fact, different authors consider that this influx of malonyl-CoA is the biggest bottleneck in the flavonoids biosynthesis, since three molecules of malonyl-CoA are needed to generate a molecule of eriodityol, apigenin, or luteolin in our case. Basal levels of this metabolite inside cells are relatively low, as it is regularly consumed during fatty acids biosynthesis, so low intracellular malonyl-CoA availability may restrict the efficient production of flavonoids in both E. coli and Streptomyces strains (Santos et al., 2011; Rathnasingh et al., 2012). Accordingly, diverse described strategies to increase flavonoids yields focused on increasing the intracellular pool of malonyl-CoA by overexpressing the acetyl-CoA carboxylase enzyme complex (ACC) in the host strain (Hwang et al., 2003; Miyahisa et al., 2006; Park et al., 2011; Wang et al., 2011). ACC overexpression increased cell concentration malonyl-CoA in a 278% in relation to the E. coli wild strain (Zhu S. et al., 2014). In our case, with so high intracellular p-coumaric acid accumulation in comparison to final flavonoids, we can hypothesize that the reason may be also a low supply of malonyl-CoA, due to its consumption by fatty acids biosynthesis in S. albus. However, feeding with pure 13.5 mM sodium malonate final concentration in flasks did not generated a significant increase in production levels. The cause of this may be that initially identified similar genes in S. albus chromosome to matB (membrane malonate carrier protein) and matC (malonyl-CoA synthetase) from S. coelicolor counterparts (sco2443, sco2445) (Park et al., 2011), actually do carry out other enzymatic functions in S. albus, and therefore this bacterium cannot incorporate extracellular sodium malonate toward flavonoids biosynthesis. As this step is the bottleneck in flavonoids biosynthesis, addition of p-coumaric acid nor both precursors (p-coumaric acid plus sodium malonate) does not increase apigenin final titers (Figure 5). In order to check if feeding with the precursor naringenin was an alternative for increasing final apigenin levels, 0.1 mM of this flavanone were added to S. albus-pAPI cultures, and in this case, apigenin production levels were increased 468% (Figure 5). Once we have achieved biosynthesis of these four flavonoids in S. albus, future metabolic engineering in this actinomycete, where the genome information is available, will be able to increase carbon flux toward malonyl-CoA, most probably leading to higher flavonoids production levels. Also, it will be possible to enhance intracellular production levels for p-coumaric acid, or limit at some degree possible catabolism pathways or enzymes affecting these precursors or flavonoid intermediates.

With respect to other heterologous hosts for production of bioactives as flavonoids, actinomycetes as S. albus, do possess the same degree of available genetic tools for genome and metabolic engineering, in comparison to E. coli, but more suitable and easier genetic tools than in the case of the eukaryote S. cerevisiae, which has the disadvantage that two copies for each gene function must be modify. As an extra advantage, the huge genetic and metabolic richness derived from large actinomycetes chromosomes allows these bacteria to possess a high diversity of anabolic and catabolic pathways in charge of providing different precursors needed for heterologous biosynthesis of a wide fan of bioactives, as it has been described in “Introduction” (Newman and Cragg, 2007).

With respect to the Gram-negative E. coli host, production of nutraceuticals like flavonoids for the pharmaceutical industries is favored in Gram-positives as S. albus. The reason for this is that S. albus, as a canonical Gram-positive bacterium, lacks the presence of important endotoxins for human cells, as those derived from E. coli outer membrane. Industrial removal of these lipopolysaccharide toxins involves costly procedures (Lopes et al., 2010).

As an actinomycete, S. albus shows has a complex cell cycle that can be modified in order to further increase production yields of secondary metabolism heterologous bioactives. Secondary metabolism is activated during the MII stage (Manteca et al., 2008). In this work, we have carried out de novo production of apigenin in S. albus-pAPI strain by using three alternative cultivation approaches, with and without spore conditioning (Manteca et al., 2008). It is noticeable that similar production levels were obtained in conditioned and not conditioned cultures, but production was anticipated up to 2 days in the conditioned cultures, even with a delayed growth during the first 24-h. Consequently, conditioned cultures can be useful to manipulate production rates in eventual S. albus industrial fermentations, as it was reported in other streptomycetes (Sánchez and Braña, 1996; Rioseras et al., 2014).

LM and IG-d-R were involved in creation of the different recombinant plasmids and heterologous production strains. CV and FL were supervisors of these experiments. PY and AM contributed to the perform conditioning experiments. FL is principal investigator of the research project were these experiments were planned.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.