Elimination of aromatic fusel alcohols as by-products of Saccharomyces cerevisiae strains engineered for phenylpropanoid production by 2-oxo-acid decarboxylase replacement

Engineered strains of the yeast Saccharomyces cerevisiae are intensively studied as production platforms for aromatic compounds such as hydroxycinnamic acids, stilbenoids and flavonoids. Heterologous pathways for production of these compounds use l-phenylalanine and/or l-tyrosine, generated by the yeast shikimate pathway, as aromatic precursors. The Ehrlich pathway converts these precursors to aromatic fusel alcohols and acids, which are undesirable by-products of yeast strains engineered for production of high-value aromatic compounds. Activity of the Ehrlich pathway requires any of four S. cerevisiae 2-oxo-acid decarboxylases (2-OADCs): Aro10 or the pyruvate-decarboxylase isoenzymes Pdc1, Pdc5, and Pdc6. Elimination of pyruvate-decarboxylase activity from S. cerevisiae is not straightforward as it plays a key role in cytosolic acetyl-CoA biosynthesis during growth on glucose. In a search for pyruvate decarboxylases that do not decarboxylate aromatic 2-oxo acids, eleven yeast and bacterial 2-OADC-encoding genes were investigated. Homologs from Kluyveromyces lactis (KlPDC1), Kluyveromyces marxianus (KmPDC1), Yarrowia lipolytica (YlPDC1), Zymomonas mobilis (Zmpdc1) and Gluconacetobacter diazotrophicus (Gdpdc1.2 and Gdpdc1.3) complemented a Pdc− strain of S. cerevisiae for growth on glucose. Enzyme-activity assays in cell extracts showed that these genes encoded active pyruvate decarboxylases with different substrate specificities. In these in vitro assays, ZmPdc1, GdPdc1.2 or GdPdc1.3 had no substrate specificity towards phenylpyruvate. Replacing Aro10 and Pdc1,5,6 by these bacterial decarboxylases completely eliminated aromatic fusel-alcohol production in glucose-grown batch cultures of an engineered coumaric acid-producing S. cerevisiae strain. These results outline a strategy to prevent formation of an important class of by-products in ‘chassis’ yeast strains for production of non-native aromatic compounds.


Introduction
The aromatic amino acids L-phenylalanine and L-tyrosine are precursors of many industrially relevant compounds belonging to the phenylpropanoid family of aromatic compounds (Neelam et al., 2020), including hydroxycinnamic acids (Vannelli et al., 2007), stilbenoids (Trantas et al., 2009) and flavonoids (Falcone Ferreyra et al., 2012). These compounds have diverse applications in the food, chemical, pharmaceutical and cosmetic industries (Neelam et al., 2020). Current production processes mostly depend on petroleum-based chemical processes (Das et al., 2007) or direct extraction from plants (Trantas et al., 2015). However, the chemical processes involved are often inefficient and unsustainable (Chemler and Koffas, 2008;Bhan et al., 2013;Zha et al., 2019) while plant extraction processes are limited by biomass availability, low extraction yields and low purity of the final products (Zhang, 2007;Rodriguez et al., 2017). To overcome these pitfalls and meet the increasing demand for biologically and renewably sourced aroma and flavour compounds, microbial production from renewable feedstocks offers a promising alternative (Trantas et al., 2015).
Deletion of the three pyruvate-decarboxylase genes PDC1, PDC5, PDC6 renders S. cerevisiae unable to grow on glucose in batch cultures. During aerobic glucose-limited growth, Pdc − strains are auxotrophic for C₂-compounds (ethanol or acetate) due to involvement of pyruvate decarboxylases in cytosolic acetyl-CoA synthesis (Flikweert et al., 1996(Flikweert et al., , 1999. Although metabolic engineering strategies have been published to bypass this acetyl-CoA requirement (reviewed by Van Rossum et al. (van Rossum et al., 2016)) and to mitigate the glucose sensitivity of Pdc − strains (Oud et al., 2012), deletion of all four 2-OADC genes in S. cerevisiae is not a straightforward strategy to eliminate formation of aromatic fusel alcohols and acids. Still, complete elimination of these by-products, without negative impacts on growth on glucose or a need to rewire central carbon metabolism, would be an attractive attribute of S. cerevisiae 'chassis' strains for production of phenylpropanoid.
The goal of the present study was to identify heterologous pyruvate decarboxylases that show activity with pyruvate but not with aromatic 2-oxo acids and to investigate whether they can functionally replace the native yeast enzymes and thereby prevent formation of aromatic Ehrlich-pathway products. To this end, a set of 11 2-oxo acid decarboxylases from yeasts (K. lactis, K. marxianus and Y. lipolytica) and bacteria (Z. mobilis and G. diazotrophicus) were expressed in a 2-OADCdeficient S. cerevisiae strain. Enzyme assays with cell extracts of the resulting strains were used to assess substrate specificities and affinities of each of the decarboxylases for pyruvate and phenylpyruvate. Subsequently, they were used to replace the native 2-OADC in a S. cerevisiae strain engineered for the production of the phenylpropanoid compound, coumaric acid. The impact of replacing the native yeast 2-OADC with the two best-performing heterologous pyruvate-decarboxylases was evaluated in aerobic, pH-controlled bioreactor cultures.

Molecular biology techniques
DNA templates for cloning were amplified with Phusion high-fidelity polymerase (Thermo Fisher Scientific, Landsmeer, Netherlands) according to manufacturer's protocol, with the exception that a primer concentration of 200 nM and 0.04 U μL − 1 of polymerase were used. The YeaStar genomic DNA kit (Zymo Research, Irvine, CA) was used to isolate genomic DNA as template for PCR amplification. The Zymoclean kit (Zymo Research) was used to purify PCR products by gel purification according to manufacturer's recommendations using milliQ water as eluent. Alternatively, PCR products were first incubated for 1 h with DpnI FastDigest enzyme (Thermo Fisher Scientific) to digest template DNA and subsequently purified using the GenElute™ PCR clean-Up Kit (Sigma-Aldrich). Diagnostic PCR was performed with DreamTaq PCR mastermix (Thermo Fisher Scientific) and with oligonucleotide primers shown in Table S1 The GenElute plasmid miniprep kit (Sigma-Aldrich) was used to isolate plasmids from E. coli.

Construction of plasmids and expression cassettes
Plasmids used and constructed in this study are shown in Table 2. Constructed plasmids were transformed to E. coli (XL1-Blue) cells according to the supplier's recommendations and grown under selective conditions.
The expression cassettes used in this study were constructed using the Yeast Toolkit (Lee et al., 2015). In brief, promoter, gene and terminator fragments (parts) are amplified with part type specific overhangs containing restriction sites (BsmBI and BsaI). Using Golden Gate assembly with the corresponding restriction enzyme, BsmBI, the individual parts are initially assembled in an universal entry vector, resulting in a part plasmid. Next, a promoter, gene and terminator part plasmid are assembled into an expression cassette using BsaI-mediated golden gate assembly, resulting in an expression cassette containing a transcriptional unit.
The GFP dropout plasmid pGGKd017 (URA3) (Wronska et al., 2020) was used as backbone to construct expression cassettes expressing a single 2-oxo acid decarboxylase. As example, the Golden Gate assembly of pGGkp035 (TDH3pr), pGGkp182 (CYC1t) and pGGKp211 (Zmpdc1) using pGGKd017 as a backbone resulted in the construction of pUDE827 (URA3, TDH3pr-Zmpdc1-CYC1t). A full overview of all part plasmids that were used to construct the expression cassettes is presented in Table 2. Correct construction was verified by diagnostic PCR and restriction analysis.

Growth studies
Shake-flask cultures were grown in 500 mL shake flasks containing 100 mL medium and incubated at 30 • C in an Innova incubator shaker (New Brunswick Scientific, Edison, NJ). Precultures on SMEG were inoculated from frozen stock cultures. These precultures were used to inoculate shake flasks containing SMEG and SMD, at an initial OD 660 of 0.2. Independent duplicate cultures were grown for each combination of yeast strain and medium composition. Specific growth rates were calculated from a minimum number of six data points collected during exponential growth and covering 3-4 doublings of OD 660 . Ehrlich pathway products were quantified in supernatant samples of triplicate stationary phase (72 h) shake-flask cultures.
Aerobic bioreactor batch cultures on SMD supplemented with 0.2 g L − 1 antifoam C (Sigma-Aldrich) were grown in 2L bioreactors (Applikon, Delft, Netherlands) with a working volume of 1.0 L. Oxygen was supplied by continuously sparging the culture with pressurized air at 0.5 L min − 1 . Exponentially growing shake-flask cultures on SMD were used to inoculate the bioreactors at an initial biomass concentration of around 0.1 g L − 1 . Cultures were grown at 30 • C and stirred at 800 rpm with a Rushton impeller. The culture pH was maintained at 5.0 by automated addition of 2 M KOH or 2 M H 2 SO 4 . Optical density at 660 nm was measured with a Jenway 7200 spectrophotometer (Jenway, Staffordshire, United Kingdom). Biomass dry weight was measured as described previously (Postma et al., 1989a). Off-gas from the bioreactors was cooled using a condenser and dried using a Permapure MD-110-48P-4 dryer (Permapure, Lakewood, NJ). CO 2 and O 2  (Gottardi et al., 2017b) E.-J. Hassing et al. concentrations in the off-gas were measured with a NGA 2000 Rosemount gas analyser (Rosemount, Analytical, Irvine, CA).
Concentrations of glucose, ethanol and extracellular organic acids in culture supernatants were measured by high performance liquid chromatography (HPLC) as described before (Hassing et al., 2019). The Ehrlich pathway metabolites 2-phenylethanol, p-hydroxyphenylethanol, phenylacetate, phenylpyruvate, coumaric acid and cinnamic acid were also measured by HPLC as described before (Hassing et al., 2019). Aromatic compounds were detected by a diode-array multiple-wavelength detector (Agilent G1315C), at wavelengths of 200 nm for phenylacetate, 210 nm for phenylpyruvate, 214 nm for 2-phenylethanol and p-hydroxyphenylethanol, 270 nm for cinnamic acid and 280 nm for coumaric acid.

Enzyme-activity assays in cell extracts
Cell extracts of S. cerevisiae strains were prepared from late exponential phase (OD 660 of approximately 8) shake-flask cultures grown on SMEG or SME medium. After 10 min centrifugation at 4696×g, cell pellets were washed twice with 20 mL 10 mM potassium phosphate buffer (pH 7.5) containing 2 mM EDTA, resuspended in 4 mL buffer and stored at − 20 • C. Prior to the enzyme assays, biomass samples were thawed, resuspended and washed with 100 mM potassium phosphate buffer (pH 7.5) containing 2 mM MgCl 2 and 2 mM dithiothreitol. When cell extracts were prepared for experiments to estimate kinetic parameters, which took several hours, complete(TM), Mini Protease Inhibitor Co. (Sigma-Aldrich) was added as protease inhibitor according to manufacturer's recommendations. Cell extracts were prepared by sonication with 0.7 mm diameter glass beads using a MSE sonicator (150-W output, 7-nm peak-to-peak amplitude) at 0 • C. After four bursts of 30 s with 30 s cooling intervals, debris was removed by centrifugation using a Sorvall SS34-rotor (Thermo Fisher Scientific) for 20 min at 47.000×g operated at 4 • C. The clear supernatants were used as cell extracts and kept on ice during experiments.
Pyruvate-decarboxylase activity in cell extracts was measured as described previously (Postma et al., 1989b). Phenylpyruvatedecarboxylase activity was assayed essentially as described before (Vuralhan et al., 2003) but with 5 mM instead of 2 mM phenylpyruvate. K m values for pyruvate were obtained by measuring pyruvatedecarboxylase activities at concentrations ranging from 0.1 mM to 50 mM, followed by nonlinear regression of the obtained results with GraphPad Prism (version 9.02, GraphPad Software, San Diego, CA). Datasets were fitted with Michaelis-Menten as well as allosteric sigmoidal kinetics.

Protein homology and phylogenetic tree
The amino acid sequences (Supplemental Dataset S.1) of the 2-oxo acid decarboxylases used in this study were aligned using Clustal Ώ (Sievers et al., 2011). A heat map displaying sequence similarity was generated using GraphPad Prism. A phylogenetic tree of the aligned protein sequences was constructed with SeaView5 (Gouy et al., 2010) applying the LG model (Le and Gascuel, 2008) with default parameter settings using 100 Bootstrap replicates as support level for internal branches.
A phylogenetic tree of the amino-acid sequences of the selected 2-OADCs generated by multiple-sequence alignment using Clustal Ώ (Sievers et al., 2011) showed a clear segregation of the eukaryotic (yeast) and bacterial sequences (Fig. 2). As anticipated, sequences of the S. cerevisiae pyruvate decarboxylases Pdc1 and Pdc5 clustered with those of the K. marxianus and K. lactis Pdc1 orthologs. Interestingly, KlPdc5 and KmPdc5 that had 76% similarity to one another showed only 34% similarity to ScPdc5. Despite the phylogenetic distance of the yeasts S. cerevisiae and Y. lipolytica, YlPdc1 was more similar to the Pdc1 cluster comprising the Pdc1 orthologs from S. cerevisiae and Kluyveromyces species as well as the S. cerevisiae Pdc1 paralogs Pdc5 and Pdc6 than to the other selected proteins. Table 3 Heterologous 2-oxo-acid decarboxylase (2-OADC) genes investigated in this study. Published information on activity with pyruvate and with the aromatic 2-oxo acids phenylpyruvate (PPY) and p-hydroxyphenylpyruvate (pOHPPY) is presented. The right-hand column indicates for which of these genes the coding sequences were codon-optimized (Co) for expression in S. cerevisiae in the present study.
The platform strain CENPK711-7C used to individually express the 2oxo acid decarboxylases carried a deletion of THI3, a gene that was originally assumed to encode a fifth S. cerevisiae 2-oxo acid decarboxylase (Dickinson et al., 1998(Dickinson et al., , 2000Vuralhan et al., 2003) but was later shown to instead encode a protein involved in thiamine homeostasis (Mojzita and Hohmann, 2006;Nosaka et al., 2008). To investigate if inactivation of THI3 was responsible for the unexpectedly low specific growth rate of the tested strains, strain IME668 (CEN.PK711-7C ScPDC1↑ ScTHI3↑) was constructed. Its specific growth rate on SMD was only 10% higher than that of strain IME667 (CENPK711-7C ScPDC1↑) and therefore still much lower than that of strain CEN.PK113-7D (Fig. 3).
As anticipated, cell extracts of strains expressing yeast Aro10 orthologs showed phenylpyruvate-decarboxylase activity, although activities were two orders of magnitude lower than pyruvatedecarboxylase activities observed in cell extracts of strains expressing yeast or Z. mobilis Pdc1 homologs ( Fig. 4 and Table S2). Three of the heterologous 2-oxo acid decarboxylases with demonstrated in vivo and in vitro pyruvate-decarboxylase activity upon expression in S. cerevisiae (GdPdc1.2, GdPdc1.3 and ZmPdc1) showed no activity with 5 mM phenylpyruvate as substrate (Table 4). These enzymes were therefore (Gdpdc1.3↑), IME615 (KlPDC1↑) and IME667 (ScPDC1↑). The dotted line indicates the detection limit for decarboxylase activity, which was <0.04 μmol mg protein − 1 ⋅min − 1 for pyruvate as substrate and <20 nmol mg of protein − 1 min − 1 for phenylpyruvate as substrate. This results in the visualization of 4 classes: enzymes with decarboxylase activity for A) PPY but not pyruvate, B) both PPY and pyruvate C) no activity for either substrates and D) activity for pyruvate, but not PPY. Table 4 Specific pyruvate decarboxylase activity, K m and the Hill coefficient for cell free extracts of S. cerevisiae strain CEN.PK711-7C (pdc1Δ pdc5Δ pdc6Δ aro10Δ thi3Δ) expressing individual 2-OADC genes. All strains were grown aerobically at 30 • C, 200 RPM in shake flasks containing 100 mL synthetic medium with 2% w/v ethanol as carbon source (SME). The cell extracts were prepared from late-exponential-phase shake-flask cultures. Different pyruvate concentrations were used as substrate for measuring pyruvate decarboxylase activity ranging from 0.1 mM to 50 mM. Enzyme activities were assayed from duplicate cultures.
To estimate the Michaelis constant (K m ) of the heterologous pyruvate decarboxylases for pyruvate, enzyme activity assays with cell extracts of strains expressing the prokaryotic enzymes and ScPdc1 yeast orthologs (KlPdc1, KmPdc1 and YlPdc1) were performed at pyruvate concentrations ranging from 0.1 to 50 mM (Fig. S1). To investigate whether, similar to S. cerevisiae pyruvate decarboxylase (Hübner et al., 1978), the heterologous pyruvate decarboxylases exhibit cooperativity, the data was fitted by non-linear regression to substrate-saturation Michaelis-Menten kinetics as well as to sigmoidal allosteric Hill kinetics (Table 4). Consistent with literature (Romagnoli et al., 2012), cell extracts containing ScPdc1 showed a Hill coefficient of 2.4, while a similar cooperativity was observed for cell extracts containing the Kluyveromyces enzymes KlPdc1 and KmPdc1. In contrast, assays with cell extracts containing either Y. lipolytica Pdc1 or one of the three bacterial enzymes (Zmpdc1, Gdpdc1.2 or Gdpdc1.3), yielded a Hill coefficient close to one and absence of a sigmoidal relation between substrate concentration and enzyme activity (Table 4, Fig. S1), thus indicating absence of cooperativity (Table 4). In these assays, the Z. mobilis pyruvate decarboxylase ZmPdc1 showed a 4-fold lower K m than ScPdc1 and a higher V max /K m ratio than the G. diazotrophicus pyruvate decarboxylases.

Decarboxylase swapping in a coumaric acid-producing S. cerevisiae strain
To investigate whether replacement of the native yeast 2-OADCs (Pdc1, Pdc5, Pdc6 and Aro10) by heterologous pyruvate decarboxylases ('decarboxylase swapping') can eliminate formation of byproducts in S. cerevisiae strains engineered for phenylpropanoid production, a tester strain producing coumaric acid was constructed. To this end, ARO10, PDC5 and PDC6 were first deleted from the previously constructed strain IMX1593, which overexpresses feedback-insensitive alleles of the DAHP synthase and chorismate mutase (aro3Δ ARO4 K229L ↑ ARO7 T226I ↑) (Hassing et al., 2019). Subsequently, expression cassettes for PlstlA, Rctal1t, AtC4H and AtCPR1, which encode for respectively, a phenylalanine ammonia lyase, tyrosine ammonia lyase, cinnamic acid hydroxylase and its cytochrome p450 reductase, required for the activation of the cytochrome P450, were integrated at the X3 locus on CHRX (Mikkelsen et al., 2012). Deletion of the pyruvate decarboxylase gene PDC1 yielded the 2-OADC-negative, coumaric acid producing platform strain IMX2668. This strain was transformed with multi-copy plasmids carrying expression cassettes for the different 2-oxo acid decarboxylases with specificity for pyruvate. All these strains grew on SMD in shake-flask cultures, albeit slower than the ScPDC1-expressing reference strain (IME677) ( Table 5). Consistent with their low pyruvate-decarboxylase activities in cell extracts, the lowest specific growth rates were observed for the two strains expressing the G. diazotrophicus pyruvate decarboxylases (Table 5).
Strain IME663 (Gdpdc1.3↑) showed a 50% lower specific growth rate in the bioreactor cultures than strain IME677 (ScPDC1↑) ( Table 6). In comparison to the other two coumaric acid-producing strains, it did not produce detectable amounts of ethanol and reached 30% higher final biomass concentrations. In contrast to the shake-flask cultures of strain IME663, the bioreactor cultures consumed all glucose. Although aromatic fusel alcohols were not detected in culture supernatants, strain IME663 reached a lower coumaric acid titer than strain IME677 (1.5 mM vs 2.2 mM) and, additionally, produced nearly three-fold higher extracellular phenylpyruvate concentrations (0.3 mM).
0.29 ± 0.01 0.20 ± 0.02 0.14 ± 0.00 a Y X/S (g g − 1 ) 0.12 ± 0.00 0.11 ± 0.00 0.35 ± 0.00 a Y EtOH/S (mol mol − 1 ) 1.37 ± 0.01 1.20 ± 0.05 0.00 ± 0.00 Y X/S (g g − 1 ) 0.27 ± 0.01 0.24 ± 0.01 0.35 ± 0.00 Y 2PE/S (mmol mol − 1 ) 1.08 ± 0.16 0.00 ± 0.00 0.00 ± 0.00 Y pOH2PE/S (mmol mol − 1 ) 1.29 ± 0.04 0.00 ± 0.00 0.00 ± 0.00 Y COUM/S (mmol mol − 1 ) 18 IME658, in which the native yeast 2-OADCs were replaced by Z. mobilis pdc1, did not produce aromatic fusel alcohols and showed a higher coumaric acid yield than the congenic strain IME677 that instead expressed ScPDC1. While our study provided a clear proof of principle for the 'decarboxylase swapping' approach, the Zmpdc1-expressing strain grew 30% slower than the ScPDC1 expressing strain. These different growth rates occurred despite high and similar pyruvate-decarboxylase activities in cell extracts of SMEG-grown cultures of pdc1Δ pdc5Δ pdc6Δ aro10Δ thi3Δ strains carrying the same Zmpdc1 and ScPDC1 expression vectors (Table S2 and Fig. 4). Slower growth of the Zmpdc1-expressing strain may be related to a reported 20-fold higher sensitivity of ZmPdc1 to inhibition by its product acetaldehyde (Goetz et al., 2001). It would therefore be interesting to express the acetaldehyde-tolerant variant ZmPdc1 W392M (Bruhn et al., 1995;Yun and Kim, 2008). Alternatively, as proposed earlier for a ScPDC1-overexpressing S. cerevisiae strain (van Hoek et al., 1998), reduced growth rates on SMD of strains expressing pyruvate-decarboxylase genes from episomal-multicopy plasmids may reflect protein-burden effects. Further metabolic engineering and/or adaptive laboratory evolution  can be applied to identify optimal expression levels of these pyruvate-decarboxylases. When impacts on specific growth rate can be prevented, the 2-oxo acid decarboxylase swapping strategy outlined in this study should be applicable for reduction of by-product formation by yeast strains engineered for production of a wide range phenylpropanoids including stilbenoids, flavonoids and hydroxycinnamic acids.
Expression of 2-OADCs from an episomal-multicopy plasmid in the CEN.PK711-7C background (pdc1Δ pdc5Δ pdc6Δ aro10Δ thi3Δ) resulted in specific growth rates for all strains between 0.12 and 0.15 h − 1 (Fig. 3). When CEN.PK113-7D is grown on synthetic medium with glucose as carbon source and ammonium sulphate as nitrogen source, ScPDC1 transcript levels are 10 fold higher compared to the other decarboxylases (Vuralhan et al., 2005) and serves as the main decarboxylase under these conditions. However, IME667 (ScPDC1↑) only had a specific growth rate of 0.13 h − 1 whereas a specific growth rate close to CEN.PK113-7D (0.42 h − 1 ) was expected. Simultaneous expression of ScPDC1 and ScTHI3 (IME668), did not lead to a higher growth rate. Surprisingly, expression of the same 2-OADC multicopy plasmids in a newly constructed, pdc1Δ pdc5Δ pdc6Δ aro10Δ coumaric acid producing strain (IMX2668) resulted in strains with much higher specific growth rates (0.11-0.28 h − 1 (Table 5)). Therefore a genetic defect, besides the THI3 deletion, in the CEN.PK711-7C background and its transformants is causing the low specific growth which may find its origin in the use loxP-Cre recombinase during the construction of the strain (Romagnoli et al., 2012); a method that can cause chromosomal recombination (Solis-Escalante et al., 2015). Whole genome sequencing might elucidate the exact cause.
In addition to outlining a metabolic engineering strategy for minimizing by-product formation, our results provided new insights in the diversity of microbial 2-OADCs. Except for ScAro10, KmAro10 and KlAro10, the genes evaluated in this study were annotated as structural genes encoding pyruvate decarboxylases. Based on the inability of KmPdc5 and KmPdc5 to complement the growth defect of a pyruvatedecarboxylase-negative S. cerevisiae strain and the absence of in vitro decarboxylase activity with pyruvate or phenylpyruvate, further research is required to investigate their catalytic activity. A predicted pyruvate decarboxylase (YlPdc1; YALI0D10131g) in Y. lipolytica, which showed low sequence similarity with other yeast pyruvate decarboxylases, is active and in contrast to other yeast pyruvate-decarboxylases (König et al., 2009) did not exhibit cooperativity for its substrate (Table 5). This result is intriguing in view of the inability of this yeast to produce ethanol (Gatter et al., 2016) and because it is generally assumed that, in Y. lipolytica, cytosolic acetyl-CoA, which is a key precursor for lipid synthesis by this oleaginous yeast, originates from the activity of the ATP-citrate lyase (Zhang et al., 2016). Combined with previously reported aldehyde dehydrogenases (Dujon et al., 2004) and an acetyl-CoA synthetase (Kujau et al., 1992;Gatter et al., 2016), YlPdc1 could provide an alternative, energetically less efficient (van Rossum et al., 2016), bypass. Further research should establish the physiological relevance of YlPdc1 in its native host.
The present study, which explored only a fraction of the natural diversity of 2-OADCs, illustrates for further screening, mutagenesis and targeted protein engineering to tailor catalytic and regulatory properties of these key enzymes to specific applications in biotechnology.

Financial support
This project has received funding from the European Union's Horizon 2020 Research and Innovation program under grant agreement No 720824.

Declaration of competing interest
None.