Redirection of pyruvate flux toward desired metabolic pathways through substrate channeling between pyruvate kinase and pyruvate-converting enzymes in Saccharomyces cerevisiae

Spatial organization of metabolic enzymes allows substrate channeling, which accelerates processing of intermediates. Here, we investigated the effect of substrate channeling on the flux partitioning at a metabolic branch point, focusing on pyruvate metabolism in Saccharomyces cerevisiae. As a platform strain for the channeling of pyruvate flux, PYK1-Coh-Myc strain was constructed in which PYK1 gene encoding pyruvate kinase is tagged with cohesin domain. By using high-affinity cohesin-dockerin interaction, the pyruvate-forming enzyme Pyk1 was tethered to heterologous pyruvate-converting enzymes, lactate dehydrogenase and α-acetolactate synthase, to produce lactic acid and 2,3-butanediol, respectively. Pyruvate flux was successfully redirected toward desired pathways, with a concomitant decrease in ethanol production even without genetic attenuation of the ethanol-producing pathway. This pyruvate channeling strategy led to an improvement of 2,3-butanediol production by 38%, while showing a limitation in improving lactic acid production due to a reduced activity of lactate dehydrogenase by dockerin tagging.

In recent decades, metabolic engineering of microbial cells has received a great attention for the production of chemicals, fuels, and pharmaceuticals [1][2][3] . Amplification of the metabolic pathway toward desired compound and inhibition of the competing pathways are generally adopted engineering strategies to achieve high titer and yield of production. To this end, various methods of regulating gene expression levels and gene knockout/down systems have been applied to metabolic engineering 2,4-6 .
On the other hand, several efforts have been focused on the spatial organization of metabolic enzymes, leading to a substrate channeling effect, which has several potential advantages such as preventing the loss of intermediates to competing pathways, reducing the accumulation of toxic intermediates, and protecting unstable intermediates, as well as improving conversion rates [7][8][9] . Spatial proximity of the pathway enzymes can be achieved by several strategies such as making a fusion protein or a scaffold-based enzyme complex, and localization of the enzymes to subcellular compartments or organelles 4,10 . For example, expression of a fusion protein of 4-coumaroyl-CoA ligase and stilbene synthase in Saccharomyces cerevisiae resulted in up to 15-fold increase in resveratrol production compared to co-expression of these enzymes 11 . Synthetic protein scaffold was applied to the heterologous mevalonate pathway in Escherichia coli, resulting in a 77-fold increase in mevalonate titer with the optimized scaffold architecture 9 . In addition to protein scaffold, RNA-and DNA-based scaffold systems have been constructed and used to produce various target products [12][13][14] . The strategy of compartmentalization of metabolic pathways to specific organelles, especially mitochondria, using N-terminal targeting sequences, has also been successfully applied to the production of isobutanol, itaconic acid, and acetoin in S. cerevisiae, Aspergillus niger, and Candida glabrata, respectively [15][16][17] .
The feasibility of the high-affinity cohesin-dockerin interaction, found in cellulosome complex, has been demonstrated in various applications such as protein purification and enzyme immobilization [18][19][20] . Previously, three enzymes converting pyruvate to 2,3-butanediol, including α -acetolactate synthase (AlsS) and α -acetolactate decarboxylase (AlsD) from Bacillus subtilis and endogenous 2,3-butanediol dehydrogenase (Bdh1), were expressed as dockerin-fused proteins, and then assembled onto a scaffold containing multiple cohesin domains 21 . Although such a substrate channeling strategy was effective in channeling α -acetolactate produced by AlsS into 2,3-butanediol production pathway, it was not efficient to compete with the native ethanol production pathway for pyruvate availability. In this study, we aimed to test the effect of substrate channeling on the flux partitioning of pyruvate, an important metabolite at a metabolic branch point. In S. cerevisiae, pyruvate is mainly converted to ethanol in the presence of fermentable carbon sources such as glucose. Pyruvate kinase, which catalyzes the conversion of phosphoenolpyruvate (PEP) to pyruvate, is the enzyme for the last step of glycolysis. By tethering heterologous pyruvate-converting enzymes to pyruvate kinase (Pyk1) using cohesin-dockerin interaction, pyruvate flux was successfully redirected to desired pathways producing lactic acid or 2,3-butanediol.

Results
Construction of PYK1-Coh-Myc strain as a platform strain for the channeling of pyruvate flux. Pyruvate is a key metabolite of the central metabolism and serves as an important precursor for biosynthesis of several products such as carboxylic acid, alcohols, and amino acids. In S. cerevisiae, pyruvate produced by glycolysis is mainly converted to ethanol via acetaldehyde during fermentative growth (Fig. 1). Therefore, if pyruvate serves as a precursor for the production of target products, as in the case of producing lactic acid or 2,3-butanediol, it is very important to redirect the carbon flux from ethanol toward desired target products.
In this study, high-affinity cohesin-dockerin interaction was applied to recruit pyruvate-converting enzymes such as lactate dehydrogenase (LdhA) or α -acetolactate synthase (AlsS) to a pyruvate-forming enzyme, Pyk1 (Fig. 1). Although S. cerevisiae has another gene encoding pyruvate kinase, PYK2, Pyk1 is known as a main pyruvate kinase 22 . As a platform strain, we generated PYK1-Coh-Myc strain in which PYK1 gene in the chromosome is C-terminally tagged with cohesin domain and Myc epitope. As a control strain, PYK1-Myc strain where PYK1 is tagged only with Myc was also constructed.
Identification of the interaction between cohesin-fused Pyk1 and dockerin-fused enzyme by co-immunoprecipitation. To ensure the interaction between cohesin and dockerin domains, fused to Pyk1 and a pyruvate-converting enzyme, respectively, we performed co-immunoprecipitation (co-IP) assays Pyruvate kinase (Pyk1) catalyzes the conversion of phosphoenolpyruvate (PEP) to pyruvate, the last step of glycolysis. Pyruvate is mainly converted to ethanol by the sequential actions of pyruvate decarboxylase (Pdc) and alcohol dehydrogenase (Adh) in S. cerevisiae. The PYK1-Coh-Myc strain expressing cohesin-tagged Pyk1 was constructed as a platform strain for the pyruvate channeling. (a) Overexpression of LdhA for lactate production or AlsS, AlsD, and Bdh1 for 2,3-butanediol production. Because cohesin-tagged Pyk1 cannot interact with native LdhA or AlsS, LdhA or AlsS competes with Pdc for the utilization of pyruvate, the common substrate. (b) Overexpression of LdhA-Doc for lactate production or AlsS-Doc, AlsD, and Bdh1 for 2,3-butanediol production. In PYK1-Myc-Coh strain, cohesin-dockerin interaction provides spatial proximity between cohesin-tagged Pyk1 and the dockerin-tagged LdhA or AlsS, resulting in pyruvate channeling to the interacting enzyme.
Lactate production using the pyruvate channeling system. In order to test whether co-localization of Pyk1 and pyruvate-converting enzymes through cohesin-dockerin interaction could bring to metabolic channeling of pyruvate to desired metabolic pathways, we first adopted this system to lactate production in S. cerevisiae. Lactic acid is a valuable chemical that can be used as a starting material for biodegradable polymers and as an acidulant and preservative in food industry. Although S. cerevisiae cannot produce lactate naturally, due to their robust characteristics such as high acid tolerance, there have been many attempts to engineer S. cerevisiae for lactate production by introducing heterologous lactate dehydrogenase [23][24][25] .
In this study, to produce lactate from pyruvate, lactate dehydrogenase LdhA from Leuconostoc mesenteroides subsp. mesenteroides ATCC 8293 was introduced into S. cerevisiae. This LdhA is specific for the production of D-lactate isomer 26 . First, we verified the effect of introducing native LdhA into PYK1-Myc or PYK1-Coh-Myc strain. Although the rates of glucose consumption and lactate production in PYK1-Coh-Myc strain were slightly lower than those of PYK1-Myc strain, there were no significant differences in the final production titers of lactate and ethanol between these two strains (Fig. 3a). Next, we investigated the substrate channeling effect on pyruvate distribution between lactate and ethanol by expressing dockerin-tagged LdhA (LdhA-Doc) in PYK1-Myc or PYK1-Coh-Myc strain (Fig. 3b). The PYK1-Myc control strain harboring p415GPD-ldhA Doc produced 4.2 g/L lactate and 18.2 g/L ethanol after 36 h cultivation in SC-Leu medium containing 50 g/L glucose. On the other hand, PYK1-Coh-Myc strain harboring p415GPD-ldhA Doc exhibited a 91% increase in lactate production (8.0 g/L) and a 9% decrease in ethanol production (16.5 g/L) compared with the control strain, demonstrating a successful substrate channeling effect between Pyk1 and LdhA (Fig. 3b). However, unfortunately, the final titer and yield of lactate decreased in both strains expressing LdhA-Doc compared with those in the strains expressing native LdhA (Fig. 3c). These results suggest that dockerin tagging might reduce the catalytic activity of LdhA, which limits the utilization of our substrate channeling strategy to improve lactate production. To confirm this, we determined the effect of dockerin tagging on the lactate dehydrogenase activity with purified enzymes, His 6 -LdhA and His 6 -LdhA-Doc. As expected, the specific activity of His 6 -LdhA-Doc (11.6 U/nmol) was lower than that of His 6 -LdhA (27.9 U/nmol) by more than two-fold.
2,3-Butanediol production using the pyruvate channeling system. The effect of dockerin tagging on enzymatic activity might be different depending on enzymes. Therefore, to further confirm the effect of metabolite channeling for pyruvate redistribution, we attempted to use PYK1-Coh-Myc strain to produce 2,3-butanediol (Fig. 1). 2,3-Butanediol is a valuable platform chemical used in extensive industrial applications such as a liquid fuel, solvent, and monomer of synthetic rubber. 2,3-Butanediol production from pyruvate using heterologous acetoin production pathway in S. cerevisiae has been described previously 27 . In our previous study to verify substrate channeling using a protein scaffold, three enzymes converting pyruvate to 2,3-butanediol including α -acetolactate synthase (AlsS) and α -acetolactate decarboxylase (AlsD) from B. subtilis and endogenous 2,3-butanediol dehydrogenase (Bdh1), were expressed as dockerin-fused proteins, and then assembled onto a scaffold containing multiple cohesin domains 21 . Because present study aimed to demonstrate the effect of metabolite channeling on the flux partitioning at a metabolic branch point, we focused on AlsS, a pyruvate converting enzyme. Therefore, AlsS was expressed as dockerin-fusion enzyme (AlsS-Doc) that enables its recruitment to Pyk1 via cohesin-dockerin interaction (Fig. 1).
We tested the effect of introducing native AlsS together with AlsD and Bdh1 into PYK1-Myc or PYK1-Coh-Myc strain. As shown in Fig. 4a, there were no significant differences between these two strains,  ( Fig. 4a). These metabolite profiles during fermentation were also maintained in PYK1-Myc strain overexpressing AlsS-Doc, AlsD, and Bdh1 from p413-S Doc DB, which produced 9.8 g/L 2,3-butanediol and 7.9 g/L ethanol after 30 h, suggesting that dockerin tagging does not affect AlsS activity (Fig. 4b). In contrast, PYK1-Coh-Myc strain overexpressing AlsS-Doc, AlsD, and Bdh1 produced 13.6 g/L 2,3-butanediol, which indicates about 38% increase compared with that produced in PYK1-Myc strain. Consistent with this increase in 2,3-butanediol titer, ethanol production decreased by 46% (4.3 g/L after 30 h fermentation) (Fig. 4b,c). These results demonstrate that the pyruvate flux was successfully redistributed from ethanol toward 2,3-butanediol through the spatial organization of Pyk1 and AlsS.

Discussion
As a proof of concept study, we investigated the effect of metabolite channeling on the flux partitioning at a metabolic branch point, focusing on pyruvate, a key intermediate of the central metabolism in S. cerevisiae. Especially, S. cerevisiae favors ethanol production from pyruvate in the presence of fermentable carbon sources even under aerobic conditions. Therefore, redirecting pyruvate flux to desired pathways in competition with ethanol production pathway is essential to produce pyruvate-derived chemicals. We constructed PYK1-Coh-Myc strain as a platform strain for pyruvate channeling and overexpressed pyruvate-converting enzymes with dockerin fusion. In the case of both lactate production and 2,3-butanediol production, pyruvate flux toward the target products was significantly enhanced, successfully competing with the native central metabolic pathway toward ethanol production even without genetic attenuation of the ethanol pathway enzymes such as pyruvate decarboxylase (Pdc) or alcohol dehydrogenase (Adh). As a result, it has been demonstrated that spatial organization of enzymes located at metabolic branch point could provide competitive advantages over the other pathways, within the frame of metabolite channeling effect.
Although cohesin tagging to Pyk1 and dockerin tagging to AlsS did not exert noticeable effects on enzyme activities, LdhA activity was severely reduced by dockerin tagging. Such a tagging-dependent inactivation of the protein function is one of the limitations of using substrate channeling strategies based on domain tagging. This problem might be solved by trying different protein-protein interaction domains or different tagging sites and linkers, which minimize the perturbation of the enzyme activity. The PYK1-Coh-Myc platform strain could also be applied to metabolic engineering strategies involving other pyruvate-converting enzymes whose activities are not largely diminished by dockerin tagging.

Strains and media.
Yeast strains used in this study are listed in Table 1. S. cerevisiae CEN.PK2-1C was used as a parental strain. PYK1-Myc and PYK1-Coh-Myc strains were constructed by PCR-mediated homologous recombination based on Cre/loxP recombination system 28 . The integration cassettes were obtained by PCR amplification from pUG27-Myc and pUG27-Coh-Myc as templates, using a primer pair of 5′-CGGTGCTGGTCACTCCAACACTTTGCAAGTCTCTACCGTT GGTGGTTCTGGTGCTAGT-3′ and 5′-TTCAAAAAAATAATATCTTCATTCAATCATGATTCTTTTT GCATAGGCCACTAGTGGATC-3′ (sequences for homologous recombination were shown in italics). After confirmation of the correct integration of the cassette at the 3′ end of PYK1 target gene locus through PCR analysis using the confirmation primers (5′-TGTCCACTTCCGGTACCACC-3′ and 5′-ATGCAACACCTCATCGTT-3′), the marker gene was removed by transformation of Cre recombinase-expression vector, p414GPD-Cre.
Yeast cells were cultured in YPD medium (10 g/L yeast extract, 20 g/L bacto-peptone, and 20 g/L glucose) or in synthetic complete (SC) medium (6.7 g/L yeast nitrogen base without amino acids and 1.4 g/L amino acids dropout mixture lacking His, Trp, Leu, and Ura) supplemented with auxotrophic amino acids as required and 20 or 50 g/L glucose.
Plasmid construction. Plasmids used in this study are listed in Table 1. To construct C-terminal tagging vector, pUG27 plasmid containing loxP-his5 + -loxP cassette was used as a starting plasmid. The DNA fragment sequentially consisting of linker sequence (GGSG), Myc epitope tag, and CYC1 terminator (T CYC1 ) was cloned into HindIII and SalI sites of pUG27, resulting in pUG27-Myc plasmid. The cohesin domain from Clostridium thermocellum CipA was obtained by PCR amplification using p416GPD-GST-[Coh] 2 as template 21 . The PCR product was cloned into the BamHI site located between linker sequence and Myc epitope tag in pUG27-Myc plasmid, resulting in pUG27-Coh-Myc. In order to omit the procedure of galactose induction for the expression of Cre recombinase, the DNA fragment encoding cre recombinase was obtained by cutting the pSH63 plasmid with SpeI and XhoI and cloned into p414GPD plasmid, resulting in p414GPD-Cre. HA-tagged BDH1 was amplified by using PCR primer containing HA tag sequence and cloned into p415GPD, resulting in p415GPD-HA-BDH1. A plasmid p415GPD-HA-BDH1-Doc, expressing N-terminal HA-tagged Bdh1 with C-terminal dockerin fusion, was previously reported 21 . p415GPD-ldhA 29 , containing ldhA (LEUM_1756, D-lactate dehydrogenase) gene from L. mesenteroides ATCC 8293, was used for lactate production. The PCR product encoding ldhA without stop codon was prepared by PCR amplification and cloned into p415GPD-Doc 21 , resulting in p415GPD-ldhA Doc . For the expression of N-terminal His-tagged proteins, E. coli expression vector pET28b was used. The DNA fragment encoding ldhA or dockerin-tagged ldhA (ldhA-Doc) was obtained by cutting the p415GPD-ldhA or p415GPD-ldhA Doc with SpeI and XhoI, and cloned into NheI and XhoI sites of pET28b plasmid, resulting in pET28b-ldhA or pET28b-His-ldhA-Doc, respectively. The construction of p413-SDB, a multigene-expression vector for 2,3-butanediol pathway, was described previously 30 . The DNA fragment encoding dockerin-tagged alsS was obtained from the p413GPD-alsS-Doc plasmid 21 and cloned into p414_P TDH3 /T PYK1 , resulting in p414_ P TDH3 -alsS-Doc-T PYK1 . [alsS-Doc]-expression cassettes (P TDH3 -alsS-Doc-T PYK1 ) flanked by MluI and NotI sites was obtained using the universal primers 30  After an equal volume of acid-washed glass beads was added to each sample, cells were lysed by vortexing 10 times for 1 min at full speed with 1 min intervals on ice. Whole-yeast lysates (700 μg of total protein) were incubated with 1 μg c-Myc antibody (Santa Cruz Biotechnology) for 3 h at 4 °C followed by addition of 15 μL protein G PLUS-Agarose beads (Santa Cruz Biotechnology). After further incubation for 90 min at 4 °C, samples were washed three times with IP150 buffer and the bound proteins were eluted by boiling in sample buffer for 10 min. Samples were resolved by SDS-PAGE and proteins were detected by immunoblotting with the use of antibodies against Myc (Santa Cruz Biotechnology) and HA epitope tag (Roche Life Science).
Protein expression and purification. Escherichia coli Rogetta gami2 (DE3) pLysS was used as a host strain to express N-terminal His-tagged proteins. The bacteria harboring pET28b-His-ldhA or pET28b-His-ldhA Doc were grown to OD 600 of 0.8 in LB medium supplemented with 50 μg/mL kanamycin at 37 °C, and induced with 1 mM isopropyl 1-thio-β -D-galactopyranoside (IPTG) at 30 °C for 4 h with shaking at 170 rpm. Cells were harvested, resuspended in binding buffer [20 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 30 mM imidazole] supplemented with 0.1% protease inhibitor cocktail (Calbiochem) and 1 mM phenylmethylsulfonyl fluoride (PMSF), and lysed by sonication on ice for a total pulsing time of 10 min (pulse on for 15 s and off for 15 s). The cell lysates were centrifuged at 8,000 × g for 15 min at 4 °C and the supernatant samples were incubated with HisPur Ni-NTA resin (Thermo Scientific) for 2 h at 4 °C. After the Ni-NTA resin was washed twice with wash buffer [20 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 45 mM imidazole], enzymes were eluted with elution buffer [20 mM Tris-HCl (pH 8.0), 500 mM NaCl, and 500 mM imidazole]. The expression and purity of the enzymes were confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and the concentrations of purified enzymes were determined by the Bradford assay (Bio-Rad) with a bovine serum albumin standard.
Lactate dehydrogenase activity assay. The enzyme activity of lactate dehydrogenase was determined as previously described 31 . Briefly, reduction of pyruvate to lactate was monitored by measuring the oxidation rate of NADH with a UV spectrophotometer at 340 nm. The enzyme activity was examined using 50 mM Tris-HCl (pH 7.0) buffer solution containing 10 mM sodium pyruvate and 0.2 mM β -NADH. One unit (U) of enzyme activity was defined as the amount that catalyzed the oxidation of 1 μmol of NADH per minute at pH 7.0.   Culture conditions. For the production of lactate, yeast cells harboring p415GPD-ldhA or p415GPD-ldhA Doc were pre-cultured in SC-Leu medium containing 20 g/L glucose and diluted to OD 600 of 0.3 in 7.5 mL of SC-Leu medium containing 50 g/L glucose in a 50 mL conical tube at 30 °C with shaking at 170 rpm. For the production of 2,3-butanediol, yeast cells harboring p413-SDB or p413-S Doc DB were cultured in the same condition of lactate production in SC-His medium instead of SC-Leu medium.
Analytical methods. The concentration of metabolites in the culture medium was determined by high performance liquid chromatography (HPLC) with BioRad Aminex HPX-87H column as previously described 32 .
Briefly, 800 μL of culture supernatants were collected and filtered through a 0.22 μm syringe filter. Analysis was performed with 5 mM H 2 SO 4 as a mobile phase at a flow rate of 0.6 mL/min and refractive index (RI) detector. The column temperature and detector temperature were maintained at 60 °C and 35 °C, respectively.