Establishment of a yeast platform strain for production of p-coumaric acid through metabolic engineering of aromatic amino acid biosynthesis

Aromatic amino acids are precursors of numerous plant secondary metabolites with diverse biological functions. Many of these secondary metabolites are already being used as active pharmaceutical or nutraceutical ingredients, and there are numerous exploratory studies of other compounds with promising applications. p-Coumaric acid is derived from aromatic amino acids and, besides being a valuable chemical building block, it serves as precursor for biosynthesis of many secondary metabolites, such as polyphenols, flavonoids, and some polyketides. Here we developed a p-coumaric acid-overproducing Saccharomyces cerevisiae platform strain. First, we reduced by-product formation by knocking out phenylpyruvate decarboxylase ARO10 and pyruvate decarboxylase PDC5. Second, different versions of feedback-resistant DAHP synthase and chorismate mutase were overexpressed. Finally, we identified shikimate kinase as another important fluxcontrolling step in the aromatic amino acid pathway by overexpressing enzymes from Escherichia coli, homologous to the pentafunctional enzyme Aro1p and to the bifunctional chorismate synthase-flavin reductase Aro2p. The highest titer of p-coumaric acid of 1.9370.26 g L 1 was obtained, when overexpressing tyrosine ammonia-lyase TAL from Flavobacterium johnsoniaeu, DAHP synthase ARO4, chorismate mutase ARO7 and E. coli shikimate kinase II (aroL) in Δpdc5Δaro10 strain background. To our knowledge this is the highest reported titer of an aromatic compound produced by yeast. The developed S. cerevisiae strain represents an attractive platform host for production of pcoumaric-acid derived secondary metabolites, such as flavonoids, polyphenols, and polyketides. & 2015 International Metabolic Engineering Society Published by Elsevier Inc. On behalf of International Metabolic Engineering Society. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Aromatic amino acids are precursors of many secondary metabolites produced in plants, where they have a key role in the plant development, adaptation and defense mechanisms (Maeda and Dudareva, 2012;Perez-Gregorio et al., 2014). Among the plant secondary metabolites many alkaloids, flavonoids, tannins and lignins find applications as nutraceutical and pharmaceutical ingredients (Scotti, 2012). Exploratory research of this type of compounds shows promising results (Winkel-Shirley, 2001;Hawkins and Smolke, 2008;Bhan et al., 2013;Leonard et al., 2009), but the limiting factor for a wider use of these secondary metabolites is the lack of efficient extraction systems from plants, or a competent microbial biosynthetic alternative that can produce these compounds in high yields (Santos et al., 2011). There is therefore much interest in developing a microbial cell factory platform that can be used for production of secondary metabolites derived from aromatic amino acids. As the biosynthesis of many plant secondary metabolites involves P450 enzymes, which are often difficult to express in bacteria, the yeast Saccharomyces cerevisiae is well suited as a cell factory platform for production of these products. Since many of the flavonoids are intended for nutraceutical applications, using S. cerevisiae as the host may be a further advantage due to the long history of its application in food and beverage production (Krivoruchko et al., Siddiqui et al., 2012) Furthermore, S. cerevisiae has proven well amenable for genetic engineering and industrial-scale fermentation and is currently used for production of active pharmaceutical ingredients, dietary supplements, chemicals and fuels (Hong and Nielsen, 2012;Nielsen et al., 2013;Borodina and Nielsen, 2014;Li and Borodina, 2015;Borodina et al., 2015).
In order to enable efficient production of secondary metabolites derived from aromatic amino acids it is necessary to optimize aromatic amino acid biosynthesis. S. cerevisiae has been metabolically engineered for improved production of aromatic amino acids through the elimination of feedback inhibition of key enzymes and elimination of by-product formation. Luttik et al. (2008) explored the use of mutated DAHP synthase ARO4 and chorismate mutase ARO7 to avoid feedback inhibition and got an increment of 200-fold of aromatic compounds in comparison to the reference strain. Furthermore, through expression of heterologous pathways, it has been possible to produce secondary metabolites derived from aromatic amino acids. Thus, Koopman et al. (2012) produced the flavonoid naringenin from glucose using a background strain with a triple knockout of the most active phenylpyruvate decarboxylases, which prevented formation of the by-product phenylethanol, and overexpressing feedback-resistant DAHP synthase and chorismate mutase.
Escherichia coli and S. cerevisiae have been extensively engineered in order to obtain flavonoid-producing cell factories. Among the flavonoids produced in E. coli from glucose are naringenin, pinocembrin, and kaempferol 3-O-rhamnoside, with titers of 84 mg L À 1 , 40 mg L À 1 , and 57 mg L À 1 , respectively (Santos et al., 2011;Wu et al., 2014;Yang et al., 2014). Several other flavonoids were produced by supplementing the broth with intermediate compounds, e.g., pinocembrin (429 mg L À 1 ), naringenin (119 mg L À 1 ), eriodictyol (52 mg L À 1 ), quercetin (23.78 mg L À 1 ), and resveratrol (2.3 g L À 1 ) (Leonard et al., 2007;Pandey and Sohng, 2013;Lim et al., 2011). Flavonoids, such as resveratrol (0.31 mg L À 1 ), genistein (7.7 mg L À 1 ), kaempferol (4.6 mg L À 1 ), and quercetin (0.38 mg L À 1 ), have been produced by engineered S. cerevisiae, when supplemented with naringenin (Trantas et al., 2009). Using glucose as carbon source, Koopman et al. (2012) produced 102 mg L À 1 of naringenin. Jendresen et al. (2015) has reported several novel highly active tyrosine ammonia-lyases and shown their activity in E. coli, Lactococcus lactis and S. cerevisiae. These examples prove the potential of E. coli and S. cerevisiae for flavonoids production, but also evidence the need for development of a platform strain capable of high-level production of aromatic metabolites, e.g. p-Coumaric acid from which many secondary metabolites are derived (Santos et al., 2011). We therefore here developed a S. cerevisiae strain that overproduces p-coumaric acid, and besides representing a starting point for further development of a process for commercial p-coumaric acid, we believe this strain can be used as a platform strain for production of flavonoids and other coumaric-acid derived secondary metabolites.

Plasmids and strains construction
The background strain for this research was S. cerevisiae CEN. PK102-5B. Cloning was carried out using E. coli strain DH5α. All the fragments used for overexpression of genes were amplified by PCR using primers and templates as described in Supplementary Tables 1 and 2. The fragment encoding chorismate mutase from C. guilliermondii was identified through BLAST search, by comparing the full amino acids sequence of Aro7p from S. cerevisiae against C. guilliermondii protein sequences in GenBank. Homology of 60% was found between Aro7p from S. cerevisiae and the hypothetical protein PGUG_00476 from C. guilliermondii (Genbank accession number: XM_001487049.1).
Tyrosine ammonia-lyase TAL from F. johnsoniaeu was as described before (Jendresen et al., 2015). The amplified products were cloned along with strong constitutive promoters into Easy-Clone integrative plasmids by USER cloning (Jensen et al., 2014). The clones with correct inserts were confirmed by sequencing. The list of the constructed vectors can be found in Table 1 and the details on the cloning are given in Supplementary Table 3.
Transformation of yeast cells was carried out by the lithium acetate method (Gietz et al., 2002). The strains were selected on synthetic drop-out medium (Sigma-Aldrich), selecting for URA, HIS and LEU markers. The yeast strains constructed in this study are listed in Table 1.
The mutant genes, scARO4 fbr (K229L), scARO7 fbr (G141S), ecar-oG fbr1 (L175D) and ecaroG fbr2 (S180F), were constructed by sitedirected mutagenesis method (Zheng et al., 2004), using primers and templates described in Supplementary Tables 1 and 3. The ARO4, ARO7 and aroG wild-type genes were amplified from the genomic DNA of S. cerevisiae and E. coli NST 74. The DNA fragments were gel-purified and cloned into vector pESC-URA-ccdB-USER or pESC-HIS-ccdB-USER, the derived plasmids pCfB761, pCfB775, pCfB1075 and pCfB1076 were confirmed by DNA sequencing. These plasmids were used as the templates for site-directed mutagenesis reactions. The complementary primers with nucleotide substitutions for mutagenesis were designed for each mutation according to the guidelines stated by Zheng et al., 2004. The reactions were incubated with DpnI for 1 h before transformation into competent DH5α E. coli cells, the strains were grown overnight on LB agar (Amp) plates at 37 1C. Colonies were then selected, and plasmid DNA was extracted and sequenced over the region of the mutation. Successful mutants were verified by DNA sequencing and then re-cloned into the integrative expression vectors (Supplementary Tables 2 and 3).

Deletions of ARO10 and PDC5
The double knockout strain Δaro10Δpdc5 was constructed by an iterative replacement of the targeted genes with the URA3 cassette in the strain CEN.PK102-5B by bi-partite method (Erdeniz et al., 1997). The knockout fragments were transformed into S. cerevisiae and transformants were selected on SC-Ura yeast synthetic drop-out media. The knockouts were confirmed by PCR on genomic DNA preparations. The URA3 marker was looped-out via direct repeats by growing the yeast on 5-fluoroorotic acid (5-FOA) plates, and the second gene was knocked-out in the same way and the URA3 marker was removed again.
For the single knockout strains (Δaro10 and Δpdc5), the target genes were replaced by a LEU2 cassette in the strain CEN.PK102-5B. The knockout fragments were transformed into S. cerevisiae and transformants were selected on SC-Leu yeast synthetic dropout media. The knockouts were confirmed by PCR on genomic DNA preparations.
The gene fragments, carrying the upstream and downstream fragments of the marker cassettes URA3 and LEU2 and of the targeted genes ARO10 and PDC5, were generated by PCR amplification using the method developed by Reid et al. (2002). Primers and templates used for targeting the genes are indicated in Supplementary Tables 1 and 2. The upstream fragments of the targeted genes (PDC5_UP or ARO10_UP), the downstream fragments of the targeted genes (PDC5_DOWN or ARO10_DOWN), the upstream and downstream fragment of the markers (2/3_URA3_UP, 2/3_URA3_DOWN, for the double knockout strain and 2/3_LEU2_UP and 2/3_LEU2_DOWN for the single knockout strain construction) from Kluyveromyces lactis were amplified using primers described in Supplementary Table 1. To generate a Table 1 Plasmids and strains used in this study.

Media and cultivations
Synthetic complete (SC) medium as well as drop-out media (SC-Ura, SC-Leu, SC-His) and agar plates were prepared using premixed drop-out powders from Sigma-Aldrich. Synthetic fed-batch medium for S. cerevisiae M-Sc.syn-1000 (FIT) was purchased from M2P labs GmbH (Germany). The medium was supplemented with the supplied vitamins solution (final 1% v/v) and the enzyme mix (final concentration 0.5% v/v) immediately prior to use.
At least six single colonies originating from independent transformants were inoculated in 0.5 ml drop-out SC liquid medium without uracil, histidine, and/or leucine in 96-deep well At the end of the cultivation OD 600 was measured as following: 10 ml of the sample was mixed with 190 ml water and absorbance was measured at 600 nm wavelength in microplate reader BioTek Synergy MX (BioTek). The culture broth was spun down and the supernatant was analyzed for p-coumaric acid concentration using HPLC.

Quantification of p-coumaric acid
Quantification of p-coumaric acid was performed on HPLC (Thermo), equipped with a Discovery HS F5 150 mm Â 2.1 mm column (particle size 3 mm). Samples were analyzed using a gradient method with two solvents: 10 mM ammonium formate pH 3.0 (A) and acetonitrile (B) at 1.5 ml min À 1 . The program started with 5% of solvent B (0-0.5 min), after which its fraction was increased linearly from 5% to 60% (0.5-7.0 min) and maintained at 60% for 2.5 min (7.0-9.5 min). Then the fraction of solvent B was decreased back to 5% (9.5-9.6 min) and remained at 5% until the end (9.6-12 min). p-Coumaric acid was detected by absorbance at 277 nm and the peak (retention time 4.7 min) area was integrated with Chromeleon 7 and used for quantification by fitting with a standard curve. For all the strains at least three biological replicates were analyzed.
The reference strain without deletions was able to produce 0.24 70.03 g L À 1 of p-coumaric acid (Fig. 2). The strain with the single deletion of PDC5 produced 0.30 7 0.09 g L À 1 of p-coumaric acid, while the strain carrying the knockout of ARO10 produced 0.29 70.04 g L À 1 . The highest production (0.55 7 0.13 g L À 1 ) was obtained in the strain with the double knockout of ARO10 and PDC5.
3.2. Effect of the elimination of the feedback inhibition of DAHP synthase and chorismate mutase on p-coumaric acid production The enzymes DAHP synthase and chorismate mutase from the aromatic amino acids pathway are feedback-inhibited by L-tyrosine and L-phenylalanine (Hartmann et al., 2003;Luttik et al., 2008). In order to enhance the activity of these enzymes, we overexpressed feedback-resistant variants of DAHP synthase and chorismate mutase in the Δaro10Δpdc5 strain. For this we selected 4 variants of DAHP synthase, a mutated feedbackresistant ARO4 K229L from S. cerevisiae, an aroF from E. coli NST 74 (ATCC 31884) and two mutant variants of aroG, also from E. coli, which were constructed through replacement of the residues L175D and S180F of the hydrophobic domain. For chorismate mutase, there were selected a mutated feedback-resistant ARO7 G141S from S. cerevisiae, a naturally feedback-resistant chorismate mutase from C. guilliermonii and tyrA from the E. coli strain NST 74 (ATCC 31884) (Tribe, 1987). The chorismate mutase from C. guilliernondii had been reported as non-feedback inhibited (Bode and Birnbaum, 1991). The variants of chorismate mutase from S. cerevisiae and C. guilliermondii were complemented with prephenate dehydrogenase TYR1 from S. cerevisiae in order to get equivalent overexpressions to the bifunctional chorismate mutase-prephenate dehydrogenase TyrA from E. coli NST 74 (ATCC 31884). The mutations were selected from previous studies, where they had been reported as feedback-insensitive mutations: the mutations in ARO4 and ARO7 were reported by Luttik et al. (2008), the mutation of aroG (L175D) was reported by Hu et al. (2003) and the mutation of aroG (S180F) was reported by Ger et al. (1994).
All the strains overexpressing DAHP synthase and chorismate mutase were evaluated for their ability to produce p-coumaric acid and a two-way ANOVA was conducted to analyze the effect of DAHP synthases and chorismate mutases. The p-coumaric acid production was normally distributed for all the combinations of DAHP synthases and chorismate mutases as assessed by a Shapiro-Wilk's test (p-Value40.05). The results showed that the overexpression of chorismate mutase alone did not have a significant effect on the production of p-coumaric acid (p-Value 0.399), while the overexpression of DAHP synthase and the combined overexpression of DAHP synthase and chorismate mutase had a significant effect on the production of the compound (p-Values 0.010 and 0.0005 correspondingly) (Supplementary Tables 4 and 5).
From the strains overexpressing scARO4 fbr , the best combination was obtained, when overexpressing at the same time scAR-O7 fbr , for the strains overexpressing ecaroF the best producer was obtained in combination with cgARO7 fbr . Although the two residues replaced in the aroG strains are located in the same region, the p-coumaric acid production after the overexpression of the mutated AroG enzymes was different. The replacement L175D seems to generate a more active AroG, since two of the strains carrying this mutation (ecaroG fbr1 -ectyrA and ecaroG fbr1 -cgARO7) had production of over 0.9 g L À 1 in contrast to the strains with the replacement S180F, where the titer did not exceed 0.8 g L À 1 . It was not possible to see a general trend of the effect of Tyr1p in the production of p-coumaric acid, but in 4 of the 8 strains overexpressing this enzyme a negative effect was observed (Fig. 3).
The overexpression of DAHP synthase and chorismate mutase mostly had a positive effect on production of p-coumaric acid, however there were some exceptions: scARO4-cgARO7, scARO4-ectyrA, ecaroG-ARO7 and ecaroG-scARO7-scTYR1 all resulted in the same or lower titer than in the reference strain.
3.3. p-Coumaric acid production after overexpression of ARO1 and ARO2 from S. cerevisiae and their analogous from E. coli In S. cerevisiae, the five steps to synthetize the aromatic intermediate compound 5-enolpyruvylshikimate-3-phosphate (EPSP) from DAHP are catalyzed by the pentafunctional enzyme Aro1p, while in other organisms such as plants and bacteria each step is performed by monofunctional enzymes (Fig. 1). In order to find flux-controlling steps in this common branch of the aromatic amino acid pathway, analogous enzymes to Aro1p and Aro2p from E. coli were overexpressed in S. cerevisiae. Strains overexpressing ARO1 and ARO2 from S. cerevisiae were also constructed with the purpose of making a comparative analysis between the strains overexpressing enzymes from E. coli and S. cerevisiae. The background strain for this experiment was the strain ST3213 (aro10Δpdc5Δ ARO4 K229L ARO7 G141S ). The control strain for this experiment was the strain ST3213 transformed with the empty integrative plasmid pCfB257 instead of plasmids carrying overexpression cassettes for ARO1, ARO2 or their analogous.
The overexpression of the monofunctional enzymes from E. coli had a positive effect on the p-coumaric acid production; the only exception was the overexpression of shikimate kinase AroK (Fig. 4A).  The strains overexpressing dehydroquinate synthase (AroB), shikimate dehydrogenase (YdiB), EPSP synthase (AroA) or shikimate kinase (AroL) produced more than 1.6 g L À 1 of p-coumaric acid. The strain with the highest improvement was the strain overexpressing the isoenzyme of shikimate kinase AroL, producing 1.9370.26 g L À 1 of p-coumaric acid.
Overexpression of the pentafunctional enzyme Aro1p and the bifunctional chorismate synthase-flavin reductase Aro2p from S. cerevisiae had a positive effect in the p-coumaric acid production. The strain overexpressing Aro1p produced 1.6870.19 g L À 1 of p-coumaric acid, the strain overexpressing Aro2p produced 1.4070.12 g L À 1 , and the simultaneous overexpression of Aro1p and Aro2p increased the production of p-coumaric acid to 1.7170.12 g L À 1 . The production of p-coumaric acid was very similar between the strain overexpressing Aro1p and the strain overexpressing Aro1p and Aro2p simultaneously. None of the strains expressing native versions of Aro1p and Aro2p produced more p-coumaric acid than the strain overexpressing AroL from E. coli (Fig. 4B).

Discussion
This study describes engineering of S. cerevisiae for production of p-coumaric acid from glucose, leading to a final production titer of 1.93 7 0.26 g L À 1 on feed-in-time medium in deep-well plates, which represents a 7.9-fold improvement in comparison to the non-optimized strain. The optimized strain also produced 1.89 g L À 1 p-coumaric acid in controlled fed-batch fermentation on mineral medium ( Supplementary Fig. 1). To the best of our knowledge, this is the highest titer of de novo production of an aromatic compound reported for S. cerevisiae. There are studies of flavonoids production from glucose in E. coli and S. cerevisiae; however none of them has p-coumaric acid as the final product. Some studies reported the accumulation of p-coumaric acid in parallel to the production of other aromatic compounds. In E. coli, Santos et al. (2011) and Wu et al. (2014) reported accumulation of p-coumaric acid to 79 and 70 mg L À 1 , respectively, when they were producing 84 and 101 mg L À 1 of naringenin. In S. cerevisiae, Koopman et al. (2012) reported around 61 mg L À 1 of accumulated p-coumaric acid, when they were producing 65 mg L À 1 of naringenin in bioreactors.
The single knockouts of PDC5 and ARO10 had a positive effect on the production p-coumaric acid. The PDC5 strain was auxotrophic for histidine, and the effect of the PDC5 knockout in comparison to ARO10 may be different in a prototrophic strain. The simultaneous deletions lead a 2-folds improvement in comparison to the reference strain (Fig. 2).This was expected as deletion of the two genes had been previously reported to improve production of some other tyrosine-and phenylalanine-derived compounds, i.e., naringenin (Koopman et al., 2012) The results obtained from the 2-factor Anova showed that overexpression of chorismate mutase alone did not have a statistically significant effect on production of p-coumaric acid. This is consistent with results obtained by Luttik et al. (2008), where overexpression of a native or a feedback resistant Aro7 did not have significant impact on the production of the aromatic fusel alcohols, unless Aro4 was overexpressed as well.
Besides Aro4p and Aro7p that have been widely reported as feedback controlled enzymes, there must be other enzymes limiting the production of aromatic amino acids. To this end, we showed that the activity of the pentafunctional enzyme Aro1p in S. cerevisiae is also limiting for the biosynthesis of chorismate. With exception of the strain overexpressing AroK, all the strains overexpressing the Aro1p and Aro2p analogous enzymes from E. coli had a higher production of p-coumaric acid (Fig. 4A). These results show that these enzymes also exhibit flux control and further improvements can be obtained through the overexpression of heterologous enzymes with higher activity than Aro1p and Aro2p from S. cerevisiae.
The strain overexpressing AroL from E. coli had the highest production of p-coumaric acid, indicating that conversion of shikimate to shikimate-3-phosphate has the highest flux control of these five steps. These results are supported by the observation from Luttik et al. (2008), where they reported accumulation of shikimate in the culture supernatants after overexpression of scAro4 fbr in S. cerevisiae. Previous studies in E. coli also reported increased flux of intermediate compounds to L-tyrosine, when shikimate kinase II was overexpressed (Takai et al., 2005;Juminaga et al., 2012).
The overexpression of AroK did not increase the production of p-coumaric acid in the constructed strain. This may be due to the low affinity of AroK to shikimate, i.e. the K m for shikimate of AroK is more than 20 mM whereas the K m of AroL is only 0.2 mM (DeFeyter and Pittard, 1986). The contribution of AroK to the shikimate kinase activity in E. coli is therefore minimal (DeFeyter and Pittard, 1986) and its overexpression did not have a significant effect in p-coumaric acid production in E. coli. Previously there was not detected any feedback inhibition of AroL or AroK by aromatic amino acids, chorismic acid or prephenic acid (DeFeyter and Pittard, 1986). Overexpression of the pentafunctional protein (Aro1p) and the bifunctional chorismate synthase-flavin reductase (Aro2p) from S. cerevisiae had a positive effect on the production of p-coumaric acid, though the titer was still lower than of the strain overexpressing AroL alone.
The p-coumaric acid overproducing strain that we describe can be further engineered to include overproduction of malonyl-CoA, a common precursor for biosynthesis of polyphenols and flavonoids. It has been previously shown that increasing malonyl-CoA supply improves production of naringenin (Koopman et al., 2012), flavonone (Leonard et al., 2007), and 7-O-methyl aromadendrin (Malla et al., 2012). Increased flux towards malonyl-CoA can be achieved by overexpression of deregulated acetyl-CoA carboxylase (Shi et al., 2014) and by further increase of acetyl-CoA biosynthesis as described previously (Krivoruchko et al., 2015).
In conclusion, we describe metabolic engineering strategies that lead towards a platform yeast strain, producing high levels of p-coumaric acid, which besides being a product of commercial interest by itself, also serves as an intermediate compound for aromatic secondary metabolites. We also demonstrate that heterologous expression of tyrosine-ammonia lyase TAL is well suitable for evaluation of metabolic engineering targets for improving the flux through the aromatic amino acid biosynthetic pathway. Through combination of several different strategies we improved the production of p-coumaric acid 7.9-fold, and we are therefore confident that our strain represents a good platform stain for production of p-coumaric acid derived secondary metabolites by S. cerevisiae.