Synthetic production of prenylated naringenins in yeast using promiscuous microbial prenyltransferases

Reconstitution of prenylflavonoids using the flavonoid biosynthetic pathway and prenyltransferases (PTs) in microbes can be a promising attractive alternative to plant-based production or chemical synthesis. Here, we demonstrate that promiscuous microbial PTs can be a substitute for regiospecific but mostly unidentified botanical PTs. To test the prenylations of naringenin, we constructed a yeast strain capable of producing naringenin from l-phenylalanine by genomic integration of six exogenous genes encoding components of the naringenin biosynthetic pathway. Using this platform strain, various microbial PTs were tested for prenylnaringenin production. In vitro screening demonstrated that the fungal AnaPT (a member of the tryptophan dimethylallyltransferase family) specifically catalyzed C-3′ prenylation of naringenin, whereas SfN8DT-1, a botanical PT, specifically catalyzed C-8 prenylation. In vivo, the naringenin-producing strain expressing the microbial AnaPT exhibited heterologous microbial production of 3′-prenylnaringenin (3′-PN), in contrast to the previously reported in vivo production of 8-prenylnaringenin (8-PN) using the botanical SfN8DT-1. These findings provide strategies towards expanding the production of a variety of prenylated compounds, including well-known prenylnaringenins and novel prenylflavonoids. These results also suggest the opportunity for substituting botanical PTs, both known and unidentified, that display relatively strict regiospecificity of the prenyl group transfer.


Introduction
The polyphenolic compounds known as flavonoids constitute a large family of secondary metabolite that are recognized as for their healthpromoting properties. To date, over 7,000 flavonoids have been identified from plants; most of these molecules are promising compounds that show a variety of biological activities (Grotewold, 2006). Within this family, prenylflavonoids form a unique subclass of phytochemicals that have one or more prenyl groups, such as dimethylallyl (C 5 ) and geranyl (C 10 ) moieties, attached to their flavonoid backbone. The structural diversity of prenylflavonoids arises from the addition of varied position, length, and number of prenyl moieties and associated modifications such as cyclization, oxidation, and side chain elongation (Zhao, 2003). Indeed, to date, over 1,000 prenylflavonoids have been isolated from members of the plant kingdom. A number of prenylflavonoids have been shown to possess valuable biological and pharmacological activities, including (for example) anti-oxidant, anti-bacterial, anti-tumor, anti-inflammatory, enzyme inhibitor, and estrogenic activities (Botta et al., 2005;Chen et al., 2014;Terao and Mukai, 2014;Yazaki et al., 2009). Notably, prenylation endows flavonoid backbones with a range of enhanced and distinct biological activities. For instance, prenylation increases the estrogenic activity of naringenin or genistein approximately 100-to 1000-fold (Kretzschmar et al., 2010), while 3-prenylation of luteolin renders the resulting molecule an inhibitor of tyrosinase (Arung et al., 2010).
As noted above, prenylflavonoids have attracted significant attention for use in clinical, research, and industrial applications. However, the limited supply of these compounds prevents their wider use for these purposes; strategies for increasing the availability of prenylflavonoids are strongly desired. Although production of prenylflavonoids relies primarily on isolation from plants or organic synthesis (Yang et al., 2015), plants contain small quantities of these compounds, making extraction and purification arduous and expensive (Grienke et al., 2016;Kuete et al., 2014). Organic synthesis also is a challenge in most cases, given the structural complexity of the compounds of interest (Kawamura et al., 2014;Neves et al., 2011).
Reconstitution of prenylflavonoid production in microorganisms using the combination of a flavonoid biosynthetic pathway and a prenyltransferase (PT) can be a promising attractive alternative to plantbased production or chemical synthesis. Recombinant microbes have great potential for cost-effective and selective mass production of prenylflavonoids. In vitro biosynthesis of flavonoid backbones and prenyl side chains using appropriate purified enzymes also is expected to reproduce the structural diversity of prenylflavonoids. Naringenin, one of the most basic flavonoid backbones, is derived from L-phenylalanine by a fermentation process; the corresponding biosynthetic pathway is composed of five metabolic steps requiring six enzymes, including a cytochrome P450-mediated reaction (Fig. 1). The naringenin biosynthetic pathway also can contribute to the production of other flavonoids such as isoflavonoids, flavonols, and anthocyanins (Grotewold, 2006). Thus, once a naringenin-producing microbe has been constructed, other related flavonoids can be produced by the introduction of additional modifying enzymes.
In the biosynthesis of prenylflavonoids, the PT is a key enzyme, given that the position of the prenyl group determines the biological activity of the prenylflavonoid . Three prenylated naringenin derivatives , 3 0 -prenylnaringenin (3 0 -PN), and 6-prenylnaringnin (6-PN)) have been isolated from plants; these compounds represent modifications with dimethylallyl groups at the C-8, C-3 0 , and C-6 position (respectively) of the backbone (Fig. 1). Despite sharing the same parent backbone, these prenylnaringenins show different biological activities: 8-PN exhibits estrogenic activity (Kretzschmar et al., 2010) and prevents atrophy of disused muscle when provided as a dietary supplement (Mukai et al, 2012(Mukai et al, , 2016; 3 0 -PN inhibits the activity of protein tyrosine phosphatase-1B (PTP1B), a target for the treatment of Type 2 diabetes (Cui et al., 2007); and 6-PN is anti-oxidant and preferentially enhances the estrogen 2-hydroxylation pathway (leading to decreased risk for breast cancer) by up-regulating expression of cytochrome P450 1A1 (Wang et al., 2016).
To date, only thirteen flavonoid PTs have been identified in plants; these botanical PTs are membrane bound enzymes that display relatively strict regiospecificity for prenyl group transfer (Akashi et al., 2009;Chen et al., 2013;Li et al., 2014;Sasaki et al, 2008Sasaki et al, , 2011Shen et al., 2012;Wang et al., 2014;Yoneyama et al., 2016). The small number of botanical PTs and their characteristic features (membrane-bound nature and strict regiospecificity) limits the microbial production of various prenylflavonoids. One such PT, SfN8DT-1 from Sophora flavescens, was characterized as a naringenin PT that catalyzes C-8 prenylation (Sasaki et al., 2008). SfN8DT-1 also catalyzes the prenylation of liquiritigenin and hesperetin; however, this enzyme regiospecifically transfers a prenyl group only onto the C-8 of the A-ring in these flavonoid backbones. PTs that catalyze C-3' or C-6 prenylation of naringenin have not been identified yet.
In contrast, actinomycetes or fungi employ promiscuous PTs in the biosynthesis of secondary metabolites. Such promiscuous PTs might be alternatives to botanical PTs for the production of prenylflavonoids. There are (to our knowledge) no reports of the endogenous biosynthesis or activity of prenylflavonoids in microorganisms; indeed, under natural conditions, these promiscuous microbial PTs would not encounter flavonoids. The native role of these microbial PTs is to transfer the prenyl group onto the aromatic ring of substrate such as naphthalene-, phenylalanine-, and tryptophan-derived compounds. Nonetheless, these PTs have been also shown to catalyze the in vitro prenylation of flavonoids, consistent with the relaxed substrate specificity of microbial PTs (in contrast to botanical PTs) (Kumano et al., 2008;Ozaki et al., 2009;Zhou et al., 2015). Microbial PTs are soluble enzymes and belong to the ABBA prenyltransferase superfamily, which shares a unique protein fold (termed the PT-barrel) comprising five repetitive αββα-motifs. In a barrel fold protein family, TIM-barrel is well known to have the (β/α) 8 -barrel structure widely conserved in many proteins (Gerlt and Raushel, 2003). In contrast to the TIM-barrel that forms an open barrel by the parallel β-strands, the PT barrel forms a central β-barrel fold consisting of ten antiparallel β-strands and contains the active central core with a spacious cavity (Bonitz et al., 2011;Tello et al., 2008). SCO7190 (Kumano et al., 2008), NovQ (Ozaki et al., 2009), and NphB (Kumano et al., 2008), which are actinomycete members of this superfamily, have been shown to catalyze the prenylation of flavonoids. Specifically, transgenic plants Fig. 1. Schematic illustration of de novo reconstitution of prenylnaringenin biosynthetic pathway in the yeast Saccharomyces cerevisiae. This approach includes two important strategic aspects: a naringenin biosynthesis platform (surrounded by the dashed line) and a prenyltransferase (PT) module for catalyzing the prenylation of the naringenin product. In the naringenin biosynthesis platform, six enzymes serve to produce naringenin from ʟ-phenylalanine through five metabolic reaction steps. Synthesized naringenin then is prenylated by PTs of varying specificities, which catalyze this key step of prenylnaringenin biosynthesis. The botanical naringenin 8prenyltransferase (SfN8DT-1) and promiscuous microbial (actinomycete and fungal) PTs are used for prenylation of naringenin. Structures of known prenylnaringenins isolated from plants are presented, and include 8-prenylnaringenin (8-PN), 3 0 -prenylnaringenin (3 0 -PN), and 6-prenylnaringenin (6-PN). PAL, phenylalanine ammonia lyase; C4H, trans-cinnamate 4-monooxygenase (a cytochrome P450 enzyme); CPR, cytochrome P450 reductase; 4CL, 4-coumaric acid-CoA ligase; CHS, chalcone synthase; CHI, chalcone isomerase. expressing SCO7190 and NovQ produced prenylflavonoids (6-PN and 6-prenylgenistein) (Sugiyama et al., 2011); NphB purified from Escherichia coli provided in vitro transfer of a geranyl group onto C5' of the B-ring of the flavonoid backbone (dihydroxychrysin) (Shindo et al., 2011). Thus, promiscuous PTs from microbes such as actinomycetes and fungi are expected to be active in other microbes, permitting the in vivo biosynthesis of prenylflavonoids.
The budding yeast Saccharomyces cerevisiae is a leading microorganism for heterologous production of valuable metabolites. The eukaryotic nature of budding yeast is expected to facilitate functional expression of plant-derived genes such as those encoding cytochrome P450 genes (Brown et al., 2015;Koopman et al., 2012;Shin et al., 2012;Vos et al., 2015) and membrane-bound PTs (Sasaki et al., 2008). If microbes like yeast enable the synthetic in vivo production of various prenylflavonoids from comparatively inexpensive amino acids as substrate feedstocks, the limited supply of these secondary metabolites could be resolved.
In the present study, we demonstrated the synthetic microbial production of prenylated flavonoids from L-phenylalanine in recombinant yeast strains. Naringenin was targeted as the major flavonoid backbone, and we reconstituted in yeast the naringenin biosynthetic pathway (consisting of five metabolic steps and six enzymes, including a cytochrome P450 reaction). To test the artificial biosynthesis of prenylated naringenin in microbes, a regiospecific botanical PT (SfN8DT-1) and a variety of promiscuous microbial PTs were introduced into the naringenin-producing yeast strain. We successfully obtained two yeast strains, one each producing 8-PN or 3 0 -PN from L-phenylalanine. The latter strain, which employs a promiscuous microbial PT from Neosartorya fischeri, is the first (to our knowledge) demonstration of heterologous microbial production of 3 0 -PN. This result suggests that microbial PTs with relaxed substrate specificity have great potential to replace regiospecific botanical PTs, most which remain unidentified, and to expand the available position of prenyl moieties in the flavonoid modification step.

Construction of plasmids
KOD -Plus-Neo (TOYOBO, Osaka, Japan) and PrimeSTAR HS DNA polymerase (TaKaRa Bio, Shiga, Japan) were used for PCR amplification. The sets of forward and reverse primers are listed in Supplementary  Table S2. The scheme of plasmid construction is illustrated in Supplementary Figs. S1 and S2. The pPF42x series are multi-copy plasmids harboring naringenin biosynthetic genes. The AtPAL1, AtC4H, and AtCHS3 genes were amplified from the Arabidopsis full-length cDNAs (pda01171, pda03082, and pda05357, respectively) and introduced between the AvrII and FseI sites of the pATP424, pATP425, and pATP426 vectors (Ishii et al., 2014) (respectively) using the In-Fusion® HD Cloning Kit (Takara Bio). PCR amplification was performed using the Arabidopsis full-length cDNA pda02355 as the template to generate a DNA fragment containing AtCPR1; this fragment was employed for In-Fusion cloning between the SalI and NotI sites of pATP425, yielding plasmid pPF005. The sequence-validated coAt4CL3 gene was digested with SalI and NotI; the resulting coAt4CL3 DNA fragment was ligated into the same restriction sites of plasmid pPF4240, yielding pPF4241. An equivalent set of reactions was used to insert the coAtCHI1 gene into pATP426 harboring the AtCHS3 gene, yielding plasmid pPF4262. The coAtCHS3 gene was amplified by overlap PCR using Fragment-1 (1-700 bp) and À2 (601-1188 bp) and exchanged for the AtCHS3 gene of pPF4262, yielding plasmid pPF4263.
The pPF00x series are genome integration plasmids and the strategy of construction is shown in Supplementary Fig. S2. A fragment including T ADH1 -AtPAL1-P ADH1 -T TDH3 -coAt4CL3-P TDH3 was prepared by the XhoI-SpeI digestion of pPF4242 and ligated into the similarly digested pRS404red vector (a pRS404-based vector in which the marker region was modified to enable integration into the trp1-Δ63 locus; under preparation for submission elsewhere), yielding plasmid pPF0041. pPF425 was digested with FseI and NotI and the AtC4H-P ADH1 -T TDH3 -At4CPR1 DNA fragment was obtained. This fragment was ligated into the similarly digested pATP405, yielding plasmid pPF005. Construction of pPF006 was accomplished by the removal of the 2μ origin from pPF4263 by AatII digestion following by self-ligation.

Yeast transformation
Transformation of S. cerevisiae YPH499 was performed using the lithium-acetate method (Gietz et al., 1992). The genomic integration plasmids were linearized before transformation, using HpaI for pPF0042 and pPF005, EcoRV for pPF0063, and AflII for pPT002x (where x ¼ 1-9, A, or B). The maintenance of genome-integrated plasmids in transformants was checked by colony-direct PCR using KOD FXneo (TOYOBO) and four independent transformants were utilized for the subsequent experiments. Supplementary Fig. S3 shows the strains constructed in this study and descriptions of these strains.

Cultivation of transformants and extraction of metabolites
The transformants were cultured overnight at 30 C in SD selection medium (for strains harboring multi-copy plasmids) or in YPDA medium (for strains harboring genome-integration plasmids). The pre-culture was inoculated into 40 mL of the production medium (SD selection medium or YPDA medium containing 10 mM ʟ-phenylalanine) in a 100-mL flask at an initial optical density at 660 nm (OD 660 ) of 0.03. After cultivation at 30 C under rotary shaking of 150 revolutions per minute (rpm) for four days, the yeast cells in 30 mL of culture broth were harvested by centrifuging at 4,800Âg for 5 min at room temperature. The resulting yeast cell pellet and the supernatant were separated and stored at À30 C and À80 C, respectively. At the same time, the yeast cells in 1 mL of the culture broth were collected by centrifuging at 20,630Âg for 5 min. The resulting cell pellet was lyophilized, and the dry cell weight (DCW) was measured.
Next, the collected yeast cells obtained from the 30-mL aliquot of culture broth were suspended in 3 mL of Milli-Q and combined with an equal volume (approximately total 6 mL) of glass beads. The yeast cells were disrupted by shaking at 1,500 rpm for 10 min using a ShakeMaster Neo (Biomedical Science, Tokyo, Japan). The disrupted yeast cells were extracted with 6 mL of ethyl acetate. After centrifugation at 9,400Âg for 1 min, 4 mL of the organic fraction were collected. Extraction was carried out twice and the resulting collected ethyl acetate solution (approximately 8 mL) was evaporated using a centrifugal evaporator at 30 C for 60 min. The residue was resuspended in 0.5 mL of methanol, and the resulting methanol suspension then was filtered using a Millex-LH 0.45mm (PTFE) filter (Millipore, Billerica, MA, USA). The resulting filtrates were stored at À30 C until analysis.
For prenylnaringenin production experiments, 300 μL of the supernatant were extracted with 600 μL of ethyl acetate. An aliquot (400 μL) of the resulting ethyl acetate layer was concentrated in vacuo and then resuspended in 200 μL of methanol. The resulting methanol solution was centrifuged at 20,630Âg for 5 min before analysis to remove any insoluble components.

HPLC and LC-MS analysis of metabolites produced by yeast transformants
The amounts of L-phenylalanine and 2-phenylethanol in the supernatant and 2-phenylethanol in the cell extract were measured using a HPLC system. Prior to HPLC analysis, the supernatant was diluted fivefold with 25% methanol and then filtered using a Mini-UniPrep™ Syringeless Filter 0.45-μm PTFE membrane (Whatman/GE Healthcare, Marlborough, MA, USA). The resulting 20% methanol solution and the cell extract were analyzed (separately) on a Prominence HPLC system (Shimadzu) equipped with a COSMOSIL 5 C 18 -MSII column (4.6 Â 150 mm, 5 μm; column temperature, 30 C; Nacalai Tesque) under the following conditions: mobile phase A, water þ 0.1% TFA; mobile phase B, acetonitrile þ 0.1% TFA; 10-50% B over 24 min, 50-98% B over 1 min, 98% B for 5 min, and then 10% B for 15 min; at a flow rate of 1.0 mL min À1 . The metabolites were monitored at 215 nm using an SPD-20A photodiode array (Shimadzu). For quantitative analysis, authentic reference samples of L-phenylalanine and 2-phenylethanol were used as external standards.
For LC-MS analysis, the supernatant and the cell extract (separately) were diluted with methanol to appropriate concentrations. After the insoluble components were removed by centrifugation, each sample was analyzed using an LCMS-2020 system (Shimadzu) on a COSMOSIL 5 C 18 -MSII column (2.0 Â 150 mm, 5 μm; column temperature, 30 C; Nacalai Tesque), at a flow rate of 0.2 mL min À1 . Gradient elution was performed under the following conditions: 20%-60% B over 19 min, 60%-98% B over 1 min, 98% B for 5 min, and then 20% B for 15 min (mobile phase A: water þ 0.1% acetic acid; mobile phase B: acetonitrile þ 0.1% acetic acid). MS analysis was performed using electro-spray ionization in the negative ion mode. The amount of product was calculated from a standard curve, which was obtained from the peak area of authentic reference samples using the selected ion monitoring (SIM) mode: m/z 147. The LC-MS analysis of the metabolites produced by the PT-expressing strains (YPNG008-010 and YPNG015-022) was performed as follows. The supernatant extract and the cell extract were prepared as described above ("Cultivation of transformants and extraction of metabolites"). If needed, the extracts were diluted with methanol to appropriate concentrations. For quantification of trans-cinnamic acid, p-coumaric acid, and naringenin, the extract samples of the supernatant and cell were analyzed using the LCMS-2020 under the conditions described above. Prenylated compounds in the cell extracts were analyzed using the LCMS-2020 system (Shimadzu) equipped with a COSMOSIL 5 C 18 -MSII column (2.0 Â 150 mm, 5 μm; column temperature, 30 C; Nacalai Tesque) at a flow rate of 0.2 mL min À1 . Gradient elution was carried out under the following conditions: 20%-80% B over 34 min, 80%-98% B over 1 min, 98% B for 5 min, and then 20% B for 15 min (mobile phase A: water þ 0.1% acetic acid; mobile phase B: acetonitrile þ 0.1% acetic acid). Detection was carried out in SIM mode, with m/z 339.2 [M-H]for prenylnaringenin. The identity of prenylated compounds were confirmed by comparing the mass spectrum and the retention time of each sample with authentic 8-PN, 3 0 -PN, and 6-PN. For quantitative analysis, a standard curve was generated using the peak area of MS chromatograms corresponding to 8-PN, 3 0 -PN, and 6-PN; the amount of each product then was calculated from the respective standard curve.

Prenylation assay of PTs with naringenin and LC-MS analysis of reaction products
YPNG007-009 and YPNG015-018 were grown overnight in YPDA medium at 30 C and then harvested. The collected yeast cells were suspended in a buffer containing 50 mM Tris-HCl (pH 7.5) and 100 mM NaCl, and the suspension was adjusted to an OD 660 ¼ 50 with the same buffer. An aliquot (250 μL) of the cell suspension was disrupted by vortexing with glass beads. After removal of disrupted cell debris by centrifugation at 20,630Âg for 5 min, 50 μL of the cell lysate were incubated with 0.2 mM naringenin and 0.2 mM DMAPP. The YPNG008-010 cell lysate was adjusted to 5 mM MgCl 2 -6H 2 O because SfN8DT-1 requires Mg 2þ ion for enzyme activity. Similarly, the YPNG015-022 cell lysate was adjusted to 10 mM CaCl 2 -2H 2 O because Ca 2þ ions enhance the enzyme activity of DMATS-family PTs. The enzyme reactions were carried out at 30 C for 6 h. The reaction products were extracted with 200 μL of ethyl acetate. An aliquot (150 μL) of the resulting organic phase was concentrated in vacuo, and the residue then was dissolved in 50 μL of methanol. The resulting methanol solution was analyzed directly using an LCMS-2020 system on a COSMOSIL 5 C 18 -MSII column (2.0 Â 150 mm, 5 μm; column temperature, 30 C; Nacalai Tesque) at a flow rate of 0.2 mL min À1 . Gradient elution was performed under the following conditions: 20%-98% B over 34 min, 98% B for 5 min, and then 20% B for 15 min (mobile phase A: water þ 0.1% acetic acid; mobile phase B: acetonitrile þ 0.1% acetic acid). Detection was carried out in SIM mode, using m/z 271.1 [M-H]for naringenin and m/z 339.2 [M-H]for prenylnaringenin.
Yeast strains harboring these multi-copy plasmids were designated YPN003, YPN004, YPN010, and YPN015, corresponding to YPH499 with pPF4240, pPF4240/pPF425, pPF4240/pPF425/pPF4262, and pPF4240/ pPF425/pPF4263, respectively (Table 2); control strains consisting of YPH499 harboring empty vector(s) (designated YPN001, YPN002, and YPN006) were constructed in parallel ( Table 2). All of these yeast strains were cultured in synthetic minimal medium containing 10 mM ʟphenylalanine, and the produced metabolites then were extracted and analyzed. Fig. 2A shows the amounts of the target metabolite (naringenin) and intermediates (trans-cinnamic acid and p-coumaric acid) detected in the various strains. The negative control strains (YPN001, YPN002, and YPN006) produced no naringenin or intermediates (data not shown). YPN003 (AtPAL1 only) reproducibly generated trans-cinnamic acid ( Fig. 2A). YPN004 (AtPAL1, AtC4H, and AtCPR1) synthesized both trans-cinnamic acid and p-coumaric acid, consistent with the corresponding genotypes of the inserted gene(s), although the amounts of trans-cinnamic acid and p-coumaric acid varied in different colonies (transformants). This observation indicated that the cytochrome P450 reaction proceeded successfully in S. cerevisiae. Both YPN010 (all six genes; including non-codon-optimized AtCHS3) and YPN015 (all six genes; including codon-optimized coAtCHS3) produced naringenin as expected, although the amounts of the measured compounds (trans-cinnamic acid, p-coumaric acid, and naringenin) further varied among the different colonies. Interestingly, YPN015, which harbors the codonoptimized coAtCHS3 gene, accumulated naringenin to nominally higher levels than did YPN010. These results indicated that de novo reconstitution of the naringenin biosynthetic pathway in yeast was possible using A. thaliana-derived genes, as previously demonstrated (Koopman et al., 2012). Notably, we found that codon optimization of the AtCHS3 gene was effective in increasing naringenin production.

Construction of naringenin-producing yeast platform by genome integration
Although de novo reconstitution of naringenin biosynthesis in yeast succeeded, the amount of produced naringenin varied substantially and showed low reproducibility among different colonies of YPN015 transformants (n ¼ 4) ( Fig. 2A). This variability probably was due to the unstable maintenance of multiple plasmids caused by the use of the same plasmid origin (2μ) (Futcher and Cox, 1984). To overcome this issue, genome-integration plasmids, designated pPF0041 (AtPAL1 and coAt4CL3), pPF005 (AtC4H and AtCPR1), and pPF0063 (coAtCHS3 and coAtCHI1), were constructed (Supplementary Table S2 and Supplementary Fig. S2). Yeast strains harboring genomically integrated plasmids were designated as YPNG002, YPNG004, and YPNG006 (YPH499 with integrated pPF0041, pPF0041/pPF005, and pPF0041/pPF005/pPF0063, respectively) ( Table 2). Control strains harboring genomically integrated empty vector(s), designated YPNG001, YPNG003, and YPNG005 (Table 2), were constructed in parallel. All of these strains were grown in YPDA rich medium containing 10 mM ʟ-phenylalanine, and the produced metabolites then were extracted and analyzed.
The negative control strains (YPNG001, YPNG003, and YPNG005) produced no naringenin or intermediates (data not shown). As shown in Fig. 2B, YPNG002 synthesized comparatively less trans-cinnamic acid (2.30 AE 0.113 mg L À1 ) (n ¼ 4) than YPN003 (6.52 AE 0.248 mg L À1 ) (n ¼ 4), presumably due to the lower copy number in the strain carrying the single-copy genomic integration. Similarly, while YPNG004 produced both intermediates (trans-cinnamic acid and p-coumaric acid) at similar amounts (2.08 AE 0.283 mg L À1 and 2.70 AE 0.419 mg L À1 , respectively) (n ¼ 3; one colony was omitted due to clearly impaired performance), the levels were nominally lower than those obtained with YPN004 (4.58 AE 2.05 mg L À1 and 7.76 AE 5.19 mg L À1 , respectively) (n ¼ 4). However, the variations in the amounts of these intermediates among different colonies were considerably decreased in YPNG004, presumably because the genomically integrated gene expression cassettes stabilized the output phenotypes. Lastly, although YPNG006 (expressing all six genes, including codon-optimized coAtCHS3, from genomically integrated plasmids) also showed nominally lower naringenin production (0.210 AE 0.0824 mg L À1 ) (n ¼ 4) than did YPN015 (harboring multi-copy plasmids) (0.468 AE 0.269 mg L À1 ) (n ¼ 4) (Fig. 2C), the clonal variation among different colonies was much improved (Fig. 2B and C). Therefore, we decided to use YPNG006 as the naringenin biosynthesis platform strain because of its superior reproducibility.
3.3. Heterologous production of 8-prenylnaringenin (8-PN) in yeast using naringenin biosynthesis platform and regiospecific botanical prenyltransferase (SfN8DT-1) Heterologous production of prenylnaringenins in S. cerevisiae requires identifying a microbial PT that can catalyze the prenylation of naringenin, given that the only known naringenin PTs are the botanical SfN8DT-1 (Sasaki et al., 2008) and SfFPT (Chen et al., 2013) proteins. To demonstrate the utility of our naringenin biosynthesis platform yeast for prenylnaringenin production, a regiospecific botanical PTs was expressed in the YPNG006 yeast strain. Specifically, SfN8DT-1 (Sasaki et al., 2008), a protein derived from the plant S. flavescens, was used as the botanical PT; SfN8DT-1 has been shown to function as a naringenin PT by catalyzing the C-8 prenylation of naringenin in the source plant Sasaki et al, 2008Sasaki et al, , 2009. A codon-optimized SfN8DT-1 gene (coSfN8DT-1) was synthesized and subcloned into a genome-integration vector. To increase the malonyl-CoA supply, which is necessary for the CHS (chalcone synthase) reaction (Fig. 1), the yeast ACC1 gene, which encodes the endogenous acetyl-CoA carboxylase (ACC) that catalyzes acetyl-CoA carboxylation to generate malonyl-CoA (Roggenkamp et al., 1980), was concurrently cloned and inserted into the same plasmid (Table 1 and Supplementary  Fig. S2). The resulting plasmid (pPT0022) then was introduced into YPNG006 (resulting in strain YPNG009) ( Fig. 3A and Table 2).
Before testing this strain in an in vivo production experiment, the cells were used for an in vitro enzyme reaction. Harvested YPNG007 (negative control) ( Table 2) and YPNG009 (coSfN8DT-1) cells were disrupted, and the resulting cell lysates were incubated with naringenin, DMAPP, and Mg 2þ , all of which are factors necessary for the prenylation reaction. The reaction products were extracted and analyzed by LC-MS. Incubation of these factors with the YPNG009 cell extract resulted in the synthesis of 8-PN, along with much smaller peak that matched the molecular mass and the retention time with 6-PN (Fig. 3B).
Subsequently, an in vivo prenylnaringenin production experiment was performed using the same YPNG007 and YPNG009 yeast strains. The cells were cultured in YPDA rich medium containing 10 mM ʟ-phenylalanine, and the produced metabolites were analyzed by LC-MS. Consistent with our expectation, YPNG009 demonstrated the reasonably selective heterologous production of 8-PN from phenylalanine; the strain possibly might have also produced another putative prenylated compounds ( Fig. 4A and B).

In vitro screening of microbial prenyltransferases for prenylnaringenin production
To screen promiscuous microbial PTs for their function in yeast, an in vitro naringenin prenylation assay was conducted. Eleven promiscuous PTs were selected from among actinomycete or fungal secondary metabolite biosynthetic enzymes (Table 3). These microbial PTs belong to the ABBA prenyltransferase superfamily, which can be classified into the two families, the CloQ/NphB family and the DMATS/CymD family. These families are known to correspond to phenol/phenazine PTs and indole PTs, respectively (Bonitz et al., 2011). Within the CloQ/NphB family, the bacterial (actinomycete) PTs (dimethylallyltransferases) SCO7190 (Kumano et al., 2008) and NovQ (Ozaki et al., 2009) are known Fig. 2. Production of naringenin and intermediates using a naringenin-producing yeast platform. Engineered yeasts (characterized as four separate transformant colonies each) were cultured in synthetic minimal SD selection medium (for strains harboring multi-copy plasmids) or rich YPDA medium (for strains harboring genomically integrated plasmids) containing 10 mM ʟ-phenylalanine for 96 h; internal metabolites then were extracted. The amounts of naringenin and its intermediates (trans-cinnamic acid and p-coumaric acid) were analyzed by LC-MS. Colored arrows shown under each bar graph indicate gene(s) harbored by each strain. (A) Engineered yeast strains harboring naringenin biosynthetic genes on multi-copy plasmid(s). (B) Engineered yeast strains harboring naringenin biosynthetic genes on plasmids integrated into the genome. Left axis indicates the amounts of trans-cinnamic acid and naringenin; right axis indicates the amount of p-coumaric acid. (C) Comparison of naringenin production among YPN010 (multi-copy plasmids, using non-codon-optimized AtCHS3 gene), YPN015 (multi-copy plasmids, using codonoptimized coAtCHS3 gene), and YPNG006 (genomically integrated plasmid, using codon-optimized coAtCHS3 gene). Dashed bars indicate the mean naringenin amounts.
Among these microbial PTs, SCO7190, NovQ, 7-DMATS, and AnaPT have been shown to be able to catalyze the prenylation of naringenin in vitro, either at the C-6 of the A-ring (SCO7190 (Kumano et al., 2008) and 7-DMATS (Zhou et al., 2015)), the C-3 0 and O-4 0 of B-ring (NovQ (Ozaki et al., 2009)), or the C-3' of the B-ring (AnaPT (Zhou et al., 2015)) ( Table 3). Although other microbial PTs of the DMATS group have not been reported to catalyze the prenylation of naringenin, some of these enzymes have been shown to accept the naphthalene backbone as a substrate, consistent with their relaxed substrate specificity (Yu et al., 2011). Therefore, these actinomycete and fungal PTs were selected as candidate enzymes to test the potential for transferring a prenyl group to naringenin when expressed in yeast.
These eleven microbial PT genes were codon optimized and subcloned into a genome integration vector along with the ACC1 gene (Table 1 and Supplementary Fig. S2); the resulting plasmids were designated pPT0021, pPT0023-pPT0029, pPT002A, and pPT002B. Each of these genome integration plasmids then was introduced into YPNG006 (the naringenin-producing yeast platform strain) (Fig. 3A), yielding strains YPNG008, YPNG010, and YPNG015-022). First, in vitro enzyme reactions were conducted (Fig. 3B). Cultured YPNG007 (negative control lacking PT), À008, À010, and À015 to À022 cells were harvested and disrupted, and cell lysates were incubated with naringenin, DMAPP, and Mg 2þ or Ca 2þ (cofactors that are necessary for PT activity). The reaction products were extracted and analyzed by LC-MS.
The YPNG018 (harboring coAnaPT) cell extract catalyzed C-3 0 Fig. 3. In vitro prenylation assay for naringenin using various prenyltransferases (PTs) expressed in yeast. Engineered yeasts were cultured overnight in YPDA rich medium, and cells then were disrupted for in vitro assays using expressed enzymes. Crude cell lysates were mixed with 0.2 mM naringenin, 0.2 mM DMAPP, and divalent ions (5 mM MgCl 2 for YPNG008-010 or 10 mM CaCl 2 for YPNG015-022); the reaction mixtures then were incubated at 30 C for 6 h. The reaction products were extracted and analyzed by LC-MS. (A) Schematic illustration of genes (colored arrows) and yeast strains used for prenylation assays. These strains were based on YPNG006 (naringenin biosynthesis platform strain with genomically integrated plasmids), but additionally include the yeast ACC1 gene and various PT genes. (B) LC-MS analyses of reaction products. MS chromatograms of m/z 339.2 are illustrated. Botanical and actinomycete PTs were incubated with MgCl 2 ; fungal PTs were incubated with CaCl 2 . A peak in the MS chromatogram of the YPNG010 reaction products (indicated by a white triangle) exhibited a retention time similar to that of 3 0 -PN, but the MS spectrum of this peak was distinct from those of known prenylnaringenins. 8-PN and 6-PN were detected in the YPNG009 reaction product. Right axis indicates the intensity for reaction products of the YPNG018 cell extract; left axis indicates the intensity for reaction products of the other strains. (C) LC-MS analyses of the reaction products of YPNG018 (expressing coAnaPT) and YPNG007 (negative control) cell extracts incubated with naringenin and Ca 2þ . MS chromatograms of m/z 339.2 are illustrated. The YPNG018 cell extract catalyzed the C-3 0 prenylation of naringenin, producing 3 0 -PN. (D) LC-MS analysis of reaction products of YPNG019 (expressing coCdpC3PT), YPNG020 (coCdpNPT), and YPNG007 (negative control) cell extracts incubated with naringenin and Ca 2þ . MS chromatograms of m/z 339.2 are illustrated. The YPNG019 cell extract catalyzed the production of 8-PN, 3 0 -PN, and 6-PN; the YPNG020 cell extract catalyzed the production of 3 0 -PN and 6-PN. Three other putative (but unidentified) prenylated naringenins also were detected in the YPNG019 reaction products, while one such putative (but unidentified) prenylated naringenin was detected in the YPNG020 reaction products (indicated by blue triangles).
prenylation of naringenin, providing the production of 3 0 -PN, as predicted from a previous in vitro study (Zhou et al., 2015) (Fig. 3C). YPNG019 (harboring coCdpC3PT) cell extract showed a mixture of peaks consistent with 8-PN, 3 0 -PN, and 6-PN (Fig. 3D). In contrast, the YPNG020 (harboring coCdpNPT) cell extract showed a mixture of peaks consistent with 3 0 -PN and 6-PN (Fig. 3D). Interestingly, additional putative prenylated compounds (distinct from the known PNs, and presumed to be modifications of naringenin or (unlikely but) of other endogenous compounds with the same molecular mass) also were detected in the reaction products of the YPNG019 and YPNG020 cell extracts (Fig. 3D). Thus, these three microbial PTs were shown to be able to catalyze the prenylation of naringenin in vitro, although other PTs may also have the potential to produce the PNs if we can find the optimal reaction conditions. YPNG009 and YPNG018 showed the same peaks with 8-PN (orange triangle; peak 2) and 3 0 -PN (red triangle; peak 3) standards, respectively. Two putative prenylated compounds were also detected (indicated by blue triangles). The MS spectrum of peak 1 (white triangle) was distinct from those of known prenylnaringenins. (C) LC-QTOF/MS analysis of metabolites produced by YPNG009 and YPNG018. MS chromatograms of m/z 339.124 AE 0.002 are illustrated. YPNG009 and YPNG018 showed the same peaks with 8-PN (orange triangle; peak 2) and 3 0 -PN (red triangle; peak 3) standards, respectively. Two putative prenylated compounds (indicated by blue triangles in Fig. 4A and B) were not detected in the chromatogram with the narrow mass window (AE0.002). The MS/MS spectra from the precursor ions of peak 2 (orange triangle) and peak 3 (red triangle) were well consistent with those of 8-PN and 3 0 -PN standards, respectively (Fig. S4). On the other hand, the MS/MS spectra obtained from the precursor ions of the putative PN compounds (blue triangles) were not similar to those of any PN standards, in addition to the difference of exact m/z of the precursor ions (339.199) (Fig. S4).

Table 3
Candidate prenyltransferases for naringenin prenylation. Eleven promiscuous microbial prenyltransferases (PTs) were selected for testing for prenylation of naringenin in yeast. A botanical naringenin PT, SfN8DT-1, was used as the positive control. A. fumigatus --

Uns€ old and Li (2005) FtmPT1
A. fumigatus -- Grundmann and Li (2005) 3.5. Heterologous production of 3 0 -prenylnaringenin in yeast using promiscuous microbial prenyltransferases In order to confirm the production of prenylnaringenins by YPNG008-010 and À015 to À022 in vivo, these strains were grown in YPDA rich medium containing 10 mM ʟ-phenylalanine; the produced metabolites then were analyzed by LC-MS. Fig. 4A shows the analytical data for the prenylated products. YPNG018 (harboring coAnaPT) produced 3 0 -PN from phenylalanine in vivo, consistent with outcome of the in vitro assay (Fig. 4B). This result was the first demonstration (to our knowledge) of the in vivo heterologous production of 3 0 -PN in a microorganism. In contrast to our in vitro results, YPNG019 (harboring coCdpC3PT) and YPNG020 (harboring coCdpNPT) synthesized none of the well-known prenylnaringenins (8-PN, 3 0 -PN, and 6-PN), but almost every prenyltransferase-introducing yeast including YPNG009 and YPNG018 (8-PN and 3 0 -PN producing strains, respectively) produced one or two putative prenylated compounds of unknown identity ( Fig. 4A and B). To identify whether these compounds are prenylated naringenin or not, the metabolites produced by YPNG009 and YPNG018 were further analyzed by LC-QTOF/MS. The MS chromatograms of LC-QTOF/MS analysis at m/ z 339.124 AE 0.002 (m/z 339.2 AE 0.6 was used for LC-MS analysis) demonstrated that YPNG009 and YPNG018 produced the compounds with the same accurate m/z as 8-PN and 3 0 -PN, respectively (Fig. 4C). Additionally, the MS/MS spectra were well consistent with 8-PN and 3 0 -PN standards (Fig. S4). The putative prenylated compounds were not detected at m/z 339.124 AE 0.002 (Fig. 4C). These results indicate that YPNG009 and YPNG018 surely produced 8-PN and 3 0 -PN, respectively, and the unknown peaks were not PNs.
Supplementary Fig. S5 shows the amount of consumed substrate (Lphenylalanine), byproduct (2-phenylethanol), and intermediates (transcinnamic acid, p-coumaric acid, and naringenin) in the various strains. Naringenin productivity of YPNG009 and YPNG018 ( Supplementary  Fig. S5C) was greatly increased compared with YPNG006 (Fig. 2C). Since YPNG009 and YPNG018 additionally expressed ACC1 and PT genes in YPNG006, the increase of naringenin was probably attributed to the overexpression of ACC1, which is involved in the supply of malonyl-CoA. The initial concentration of L-phenylalanine was 2230 mg L À1 , of which 70% was consumed by YPNG007, whereas approximately one third of the L-Phe was consumed by YPNG009 and YPNG018. A portion of the Lphenylalanine was converted to 2-phenylethanol by yeast endogenous enzymes, representing the production of 110 mg L À1 of 2-phenylethanol by YPNG007. On the other hand, the amount of 2-phenylethanol was nominally lower in YPNG009 and YPNG018 than in YPNG007. This result indicated that exogenous introduction of the naringenin biosynthesis pathway shifted the carbon flux from synthesis of 2-phenylethanol toward production of prenylnaringenin (naringenin), although the amounts of (prenyl)naringenin and intermediates produced were relatively low compared with the amount of L-phenylalanine consumed. This result suggested that most of the consumed L-phenylalanine might be used for cell growth or other by-products, and further improvement of (prenyl)naringenin production would be needed for future development.

Discussion
Our prenylnaringenin-producing yeast system revealed that the promiscuous microbial PTs can replace regiospecific botanical PTs. This result is especially interesting, given that most of botanical PTs still remain unidentified. To the best of our knowledge, this work is the first demonstration of first microbial production of 3 0 -PN, and was achieved by using the promiscuous fungal PT AnaPT (Yin et al., 2009) in combination with genes encoding components of the Arabidopsis naringenin biosynthetic pathway. Cell extracts of YPNG019 (harboring coCdpC3PT) and YPNG020 (harboring coCdpNPT) also catalyzed the prenylation of naringenin in vitro. The reaction products obtained using these fungal PTs included 8-PN, 3 0 -PN, and/or 6-PN, and three putative but unidentified prenylated compounds (whether naringenins or other molecules) (Fig. 3D). These results represent the first demonstration that CdpC3PT and CdpNPT, at least, can catalyze the prenylation of naringenin in vitro. Although YPNG019 and YPNG020 did not produce well-known prenylnaringenins in vivo, an increase in the level of the naringenin precursor might permit detection of the synthesis of these prenylnaringenins. These results suggest that the promiscuous microbial PTs used in this study have great potential for the production of prenylnaringenins or other prenylated compounds, possibly expanding the available positions on the flavonoid backbone that can be modified by the prenyl moiety.
The naringenin-producing yeast system expressing botanical SfN8DT-1 also produced 8-PN from L-phenylalanine. This study used L-phenylalanine as a substrate, which is the initial compound of flavonoid biosynthesis. Notably, L-Phe is simpler, less expensive, and more water-soluble than are the substrates (naringenin ) and p-coumaric acid ) used in previous studies. Recently, the production of 8-PN from glucose was reported (Levisson et al., 2019). Compared with that report, our yeast production system showed low productivity of naringenin from L-phenylalanine and much lower bioconversion efficiency to prenylnaringenins; increasing the naringenin productivity of our system will be an important future challenge. Our metabolite analysis revealed the low productivity of the desired biological intermediates from L-phenylalanine, as indicated by comparison of L-phenylalanine consumption (~700 mg L À1 ) with the amounts of trans-cinnamic acid, p-coumaric acid, and naringenin generated (~2 mg L À1 each). These data suggest that the majority of the consumed L-phenylalanine is used for cell growth and/or other processes. Engineering of the host yeast strain is expected to improve the productivity, as has been demonstrated in previous reports (Koopman et al., 2012;Levisson et al., 2019). The low biosynthetic rate also may be caused by the insufficient expression of the biosynthetic enzymes. Codon optimization is one way of increasing the expression level of enzymes. Previous studies have demonstrated that the codon optimization of foreign gene(s) significantly increases the production of compounds (Ajikumar et al., 2010) or proteins (Norkiene and Gedvilaite, 2012), given that codon optimality contributes to translation efficiency and mRNA stability (Presnyak et al., 2015). Indeed, codon optimization of the AtCHS3 gene considerably increased naringenin production in the present study (Fig. 2). In addition, the screening of gene sources other than A. thaliana would be also effective for increasing the naringenin productivity (Mark et al., 2019).
These data indicated that the PTs used here may have low prenylation activity for naringenin (Zhou et al., 2015) and/or that the biosynthesis of prenylnaringenins may be limited by an insufficient supply of substrates.
A previous investigation of 8-PN production in S. cerevisiae demonstrated that the low availability of DMAPP impaired 8-PN production; yeast engineered to have a larger DMAPP pool exhibited increased 8-PN production . Therefore, increasing the availability of DMAPP is expected to improve the productivity of prenylnaringenins. For example, additional expression of the rate-limiting enzyme of the mevalonate (MVA) pathway, 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGR), might be an effective metabolic engineering strategy for increasing the pool of available DMAPP (Polakowski et al., 1998) in our strains. Therefore, the truncated HMG1 (tHMG1) (Polakowski et al., 1998) and the acetoacetyl-CoA synthase gene (nphT7) from Streptomyces (Okamura et al., 2010) were additionally co-expressed in YPNG009 (8-PN producer); however, we did not observe the increase of 8-PN productivity (data not shown). Thus, other strategy would be needed to improve the DMAPP supply for increasing the productivity of prenylnaringenin. This study demonstrated that promiscuous microbial PTs have great potential for use in the production of various prenylated naringenins and other molecules, including unnatural compounds, although these microbial PTs displayed low regiospecificity for prenyl group transfer (Fig. 3D). The crystal structures of several ABBA PTs have been elucidated Fan et al., 2015;Jost et al. (2010); Kuzuyama et al., 2005;Liu et al., 2013;Metzger et al., 2010;Mori et al., 2016), and mutagenesis of these enzymes has been shown to alter the prenyl donor selectivity and regiospecificity of prenylation Fan et al., 2015;Mori et al., 2016). Therefore, on-target production of various prenylnaringenins should be possible by protein engineering of promiscuous microbial PTs to enhance their prenylation selectivity.
In conclusion, we showed that promiscuous microbial PTs can be used to substitute for unidentified botanical PTs. We further demonstrated the first heterologous production of 3 0 -PN using AnaPT, a promiscuous fungal DMATS (Fig. 5). These PTs are expected to serve as useful tool for artificial biosynthesis of novel prenylflavonoids, compounds that are difficult to produce in nature. Furthermore, we also revealed that PTs with relaxed substrate specificity have great potential to provide a diversity of prenylated compounds, including both well-known and novel prenylnaringenins. These findings suggest strategies towards expanding the variety of prenylated compounds, as well as for providing substitutes for botanical PTs, most of which remain unidentified and are restricted by the relatively strict regiospecificity of their prenyl group transfer activities.

Declaration of competing interest
We confirm that there are no conflicts of interest associated with this publication. Fig. 5. Production of prenylnaringenin by expressing promiscuous microbial PTs in yeast. A schematic summary of this paper. An engineered yeast strain produced 3 0prenylnaringenin using a promiscuous microbial PT (AnaPT) in combination with the six naringenin biosynthetic enzymes.