Improving l -phenylacetylcarbinol production in Saccharomyces cerevisiae by in silico aided metabolic engineering

Journal Pre-proof Improving L-phenylacetylcarbinol production in Saccharomyces cerevisiae by in silico aided metabolic engineering Elham Iranmanesh, Mohammad Ali Asadollahi, Davoud Biria

PII:

S0168-1656(19)30918-6

DOI:

https://doi.org/10.1016/j.jbiotec.2019.11.008

Reference:

BIOTEC 8547

To appear in:

Journal of Biotechnology

Received Date:

26 June 2019

Revised Date:

13 October 2019

Accepted Date:

11 November 2019

Please cite this article as: Iranmanesh E, Asadollahi MA, Biria D, Improving L-phenylacetylcarbinol production in Saccharomyces cerevisiae by in silico aided metabolic engineering, Journal of Biotechnology (2019), doi: https://doi.org/10.1016/j.jbiotec.2019.11.008

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.

Improving L-phenylacetylcarbinol production in Saccharomyces cerevisiae by in silico aided metabolic engineering

Elham Iranmanesh1, Mohammad Ali Asadollahi1,*, Davoud Biria1

1

Department of Biotechnology, Faculty of Advanced Sciences and Technologies, University of Isfahan, Isfahan

ro of

8174673441, Iran

*Corresponding author: e-mail: [email protected]

-p

Highlights:

Effect of gene deletions on L-PAC formation in S. cerevisiae was assessed



Six mutants, namely ∆rpe1, ∆pda1, ∆adh3, ∆adh1, ∆zwf1 and ∆pdc1, were predicted

re



in-silico

∆zwf1 exhibited the highest L-PAC formation (2.48 g/L) by using 2 g/L of benzaldehyde

Abstract:

na

The yield of L-PAC on benzaldehyde was equivalent to 88% of the theoretical yield

ur



lP



Jo

L-Phenylacetylcarbinol (L-PAC) which is used as a precursor for the production of ephedrine and pseudoephedrine is the first reported biologically produced α-hydroxy ketone compound. L-PAC is commercially produced by the yeast Saccharomyces cerevisiae. Yeast cells transform exogenously added benzaldehyde into L-PAC by using the action of pyruvate decarboxylase (PDC) enzyme. In this work, genome-scale model and flux balance analysis were used to identify novel target genes for the enhancement of L-PAC production in yeast.

1

The effect of gene deletions on the flux distributions in the metabolic model of S. cerevisiae was assessed using OptGene and minimization of metabolic adjustments. Six single gene deletion strains, namely ∆rpe1, ∆pda1, ∆adh3, ∆adh1, ∆zwf1 and ∆pdc1, were predicted in silico and further tested in vivo by using knock-out strains cultivated semi-anaerobically on glucose and benzaldehyde as substrates. ∆zwf1 mutant exhibited the highest L-PAC formation (2.48 g/L) by using 2 g/L of benzaldehyde which is equivalent to 88 % of the

ro of

theoretical yield.

Keywords: Metabolic engineering, L-phenylacetylcarbinol, Saccharomyces cerevisiae, Flux

-p

balance analysis, OptGene

1. Introduction

re

Enantiopure α-hydroxy ketones comprise important building blocks for the synthesis of

lP

valuable fine chemicals and pharmaceuticals (Hoyos et al., 2010). The first report on biologically production of a famous α-hydroxy ketone, L-phenylacetylcarbinol (L-PAC), was published around a century ago by Neuberg & Hirsch (1921). It described the transformation

na

of exogenously added benzaldehyde into L-PAC (also commonly referred to in the literature as (R)-PAC) by using free yeast cells in a microbiological process. This compound is a

ur

precursor of several decongestant, antiasthmatic and bronchodilator medicines such as

Jo

ephedrine, pseudoephedrine and norephedrine (Abourashed et al., 2003). Pyruvate decarboxylase (PDC) which is found in most yeast species was reported as the main enzyme for L-PAC production (Rosche et al., 2002), while alcohol dehydrogenases (ADHs) (or other oxidoreductases) produce benzyl alcohol and 1-phenyl-1,2-propanediol (PAC-diol) as unwanted by-products (Long & Ward, 1989). In practice, quantitative conversion of benzaldehyde into L-PAC has never been achieved (Shukla & Kulkarni, 2000). This can be

2

mainly because of the competition of benzyl alcohol and benzoic acid with L-PAC in using benzaldehyde as the substrate, and also reduction of L-PAC to PAC diol. Although the metabolic pathway is intrinsic to the microorganism, the system suffers from inhibition of substrate, product and by-products. Previous works aimed to improve biotechnological production of L-PAC are evidence of how long this obstacle has been recognized (Agarwal et al., 2015; Andreu & ·lí del Olmo, 2018; Bruder & Boles, 2017; Gupta et al., 1979; Li et al., 2017; Rosche et al., 2002). In order to overcome the inhibitory effects and enhance the yield,

ro of

various strategies such as screening of robust microorganisms (Rosche et al., 2003), employing immobilized cells (Doostmohammadi et al., 2015), utilization of purified enzymes (Gunawan et al., 2008; Sehl et al., 2017) and application of two-phase systems (Khan &

-p

Daugulis, 2010) have been examined. Moreover, attempts to decrease the by-product

formation or increase the carboligation activity have had limited success (Agarwal et al.,

re

2015; Bruder & Boles, 2017; Nikolova & Ward, 1991).

lP

The use of microbes in industrial processes often requires deregulation of natural metabolism in order to increase their efficiency and economic yield (Li et al., 2019; Stephanopoulos & Stafford, 2002). As the cellular metabolism is a complex system, finding the best target genes

na

for engineering the strains is not a straightforward task (Alper & Stephanopoulos, 2004). There are numerous successful examples where genome-scale constraint-based modeling has

ur

been applied for strain improvements which are reviewed elsewhere (Kerkhoven et al., 2015).

Jo

However, most of the efforts in this area have revolved around the production of amino acids, organic acids, isoprenoids or petroleum-alternative biochemicals (Alper et al., 2005; Bro et al., 2006; Brochado et al., 2010; Hendry et al., 2016; Kaushal et al., 2018; Lee & Wendisch, 2017; Ng et al., 2012; Otero et al., 2013). Extensive knowledge about Saccharomyces cerevisiae and its complete genome sequence in combination with mathematical tools have enabled using genome scale models (GEMs) for

3

the metabolic engineering applications. There are ten published versions of GEMs for S. cerevisiae with different scopes and applications which have been reviewed by Österlund et al. (2012). Pereira et al. (2016) observed that the oldest version of GEM i.e. iFF708 (Förster et al., 2003) provides the best predictions of central carbon metabolism. This model is still preferred by many researchers in metabolic engineering studies. For instance, it was employed to identify novel target genes in S. cerevisiae for the enhanced production of sesquiterpenes (Asadollahi et al., 2009), vanillin (Brochado et al., 2010) and succinic acid

ro of

(Agren et al., 2013) or improved utilization of glycerol (Strucko et al., 2018). In this research, new target genes for enhancing biotransformation of benzaldehyde to L-PAC were identified by the aid of in silico metabolic engineering strategy based on single gene

-p

deletions. OptGene (Patil et al., 2005) as a modeling framework and minimization of

metabolic adjustment (MOMA) as an objective function were used to identify the gene

re

knockouts leading to improved L-PAC production. These mutations along with the parental

2. Materials and methods

lP

strain were examined experimentally under semi-anaerobic conditions for L-PAC production.

2.1. Model and computational tools

na

The iFF708 GEM (Förster et al., 2003) including 708 genes, 1175 reactions, 733 metabolites with two cellular compartments was used in this work. The reactions for conversion of

ur

benzaldehyde to L-PAC and corresponding by-products which are catalyzed by isoenzymes

Jo

of PDC or ADH were added to the original GEM (Supplementary 1). The modified model was initially validated by simulating the wild type yeast growth in the aerobic and anaerobic glucose fermentation conditions and comparing the simulation results with the experimental data of the S. cerevisiae CEN.PK113-7D. In order to improve the accuracy of the in silico investigation, the limits of the uptake and secretion rates for some metabolites including glucose, benzaldehyde, ethanol and some organic acids were constrained by experimentally

4

measured values. Nevertheless, ammonia, phosphate, and sulfate which are necessary for the growth were assumed to be non-limiting and available as much as needed. ATP for maintenance requirement was adjusted to 1 mmol gDCW-1 h-1, as predicted by Förester et al. (2003). For simulations referred to as anaerobic or semi-anaerobic, the oxygen uptake rate was constrained to zero or 2 mmol gDCW-1 h-1, respectively. The glucose uptake rate was fixed to 10 mmol gDCW-1 h-1, while the benzaldehyde uptake rate was adjusted to 1 mmol gDCW-1 h-1. These values were experimentally obtained during the biotransformation of

ro of

benzaldehyde to L-PAC by S. cerevisiae CEN.PK113-7D. The calculation method has been reported previously (van Hoek et al., 2000).

After adding the reactions and mass balances corresponding to L-PAC production pathway,

-p

flux balance analysis (FBA) with a maximum growth objective function was performed to

predict the growth rate, intracellular flux distribution and byproduct secretion amounts. All

re

FBA calculations were carried out with the COBRA Toolbox 2.0 (Schellenberger et al.,

lP

2011) in MATLAB 2015 with implementing TOMLAB CPLEX (Tomopt, Inc.) as the optimization solver. The gene knockouts leading to enhanced L-PAC production were found using the OptGene algorithm with MOMA as the objective function. We excluded all the

na

essential, exchange and transport reactions from the target knockouts. The viability of these knockouts was verified by Saccharomyces genome database (http://www.yeastgenome.org).

ur

Hence, the knockouts leading to higher L-PAC yields were selected and further tested

Jo

experimentally.

2.2. Strains and chemicals All single gene deletion strains used in this study are derived from S. cerevisiae BY4741 strain family (MAT a, his3Δ1, leu2Δ0, met15Δ0, ura3Δ0). This reference strain and its single gene deletion derivatives were purchased from the European S. cerevisiae Archive for Functional Analysis (Euroscarf, Frankfurt, Germany). Table 1 presents the genotypes and

5

descriptions of the six single gene deletion mutants. The wild type strain S. cerevisiae CEN.PK113-7D (MATa, URA3, HIS3, LEU2, TRP1, MAL2-8C, SUC2) was used to obtain the uptake and secretion rates of some metabolites and make an accurate simulation. Among the vast variety of S. cerevisiae strains, the BY family is particularly important because the widely used deletion collections are based on this background, while CEN.PK113-7D is a popular strain for physiological studies and functional genomics. Standard L-PAC was a gift from Embio Limited (Mumbai, India). All other reagents were of

ro of

analytical grade. 2.3. Media composition

To compare in silico and in vivo L-PAC production, the biotransformation of benzaldehyde to

-p

L-PAC was conducted in 100 mL serum bottles for the selected single gene deletion strains. YPD medium (10 g/L yeast extract, 20 g/L peptone and 20 g/L glucose) was used for the

re

growth of strains, then the biomass was harvested for L-PAC production in a

lP

biotransformation medium containing 30 g/L glucose, 6 g/L peptone and 2 g/L benzaldehyde. At first, pre-culture was prepared by growing the strains on sterile YPD agar plates. After incubating at 30 °C for 48 h, the culture on agar plates was kept at 4 °C in order to subculture

na

every month. A parallel set of cultures was inoculated in the centrifuge tubes containing 5 mL of the sterile YPD medium with a single colony of each yeast strain. The inoculated tubes

ur

were incubated overnight on a rotary shaker at 30 °C and 180 rpm. This culture was

Jo

inoculated further in 50 mL YPD medium and incubated overnight at the same conditions. After reaching the appropriate cell density, cells were centrifuged and introduced to the biotransformation medium in 100 mL serum bottles with initial OD600 of 9-10 for all strains (except ∆adh1) and final working volume of 50 mL. The serum bottles were then sealed with rubber stoppers and aluminum crimp caps to support an oxygen-limited environment and fermentative metabolism. The culturing media were incubated in a shaker incubator at 200

6

rpm and 30 °C for 1 h adaptation of cells. Then, the reactions were started by addition of 2 g/L benzaldehyde. The bottles were incubated at 30 °C and 200 rpm for 24 h. Samples were taken aseptically with a syringe pierced through the rubber stopper for analysis of cell growth and metabolites. 2.4. Analytical methods Cell growth was monitored by the UV/Vis spectrophotometer (Eppendorf) at 600 nm. Dry weights were determined from 1 mL sample of each culture strain transferred to pre-weighted

ro of

micro centrifuge tubes. After centrifugation at 6000 rpm for 5 min at 4 °C, the supernatant was transferred to new tube, used for the metabolite assay or frozen at -20 °C if not measured directly. The pellets were washed twice with deionized water and dried at 85 °C until a

-p

constant weight was obtained. The correlation between dry weight and OD600 was measured. Mean values and standard deviations were calculated from three different samples.

re

Concentrations of benzaldehyde, benzyl alcohol and L-PAC were determined by gas

lP

chromatography (GC). Samples were extracted with 1:1 ratio of dichloromethane and vortexed for 2 min. One μL of the bottom organic layer was injected into GC (Agilent Technologies 6890N, USA) equipped with flame ionization detector (FID) and HP-5

na

capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness). Helium was used as the carrier gas at a flow rate of 1.5 mL/min and a split ratio of 20:1. Injector and detector

ur

temperatures were set at 200 and 250 °C, respectively. Initial oven temperature was 50 °C

Jo

and was ramped to 150 °C by 10 °C/min slope, then to 210 °C with 20 °C/min and kept constant for 5 min. Quantification of compounds was carried out using standard curves of a reference compound. The concentrations of ethanol and acetate were also measured by the same GC system but with HP-Innowax capillary column (60 m × 0.32 mm i.d., 0.5 μm film thickness). The injector and detector temperatures were set at 220 and 260 °C, respectively. Initial oven

7

temperature was 45 °C and was increased to 125 °C at a rate of 10 °C/min followed by a ramp of 5 °C/min to 220 °C, and held at this temperature for 1 min. Glucose content was analyzed enzymatically by a glucose kit (GOD-PAP, Man Lab., Iran). 3. Results and Discussion 3.1. Characterization of the wild type yeast strain One of the crucial points in yeast metabolism is the balance of NAD(P)H; because the condition for consumption of all produced NAD(P)H is not always provided. In glycolysis

ro of

pathway, two NADH molecules are produced which are re-oxidized to NAD+ via reduction of acetaldehyde to ethanol by ADH. Especially, at low rates of oxygen consumption, ethanol production pathway takes place in order to balance NADH. In L-PAC biotransformation

-p

pathway, the activated-acetaldehyde derived from glycolysis, is condensed with exogenous benzaldehyde by PDC to produce L-PAC, hence it is not available further for reduction by

re

ADH. Instead, reduction of benzaldehyde to benzyl alcohol takes place to balance the

lP

NAD+/NADH ratio and some of the added benzaldehyde is also oxidized to benzoic acid. A simplified schematic pathway of S. cerevisiae has been provided in Fig. 1 with details of biotransformation of benzaldehyde to L-PAC.

na

Prior to in silico design, substrate consumption and product formation by S. cerevisiae CEN.PK113-7D were investigated in the biotransformation medium with initial cell density

ur

(OD600) of 10. Concentrations of substrates and byproducts, as well as extracellular

Jo

metabolites are shown in Fig. 2. The biotransformation was started by addition of benzaldehyde at the optimum concentration of 2 g/L, according to our previous work (Doostmohammadi et al., 2016). These experimentally measured values were used to limit the uptake and secretion rates of glucose, benzaldehyde, ethanol and acetate to improve the accuracy of the in silico investigation. After approximately 10 h, the highest L-PAC formation was detected around 1.4 g/L which is 50% of the theoretical yield.

8

3.2. Model validation The modified GEM containing the reactions for conversion of benzaldehyde to L-PAC and relative by-products was checked for its accuracy. The flux distribution of the wild type yeast under anaerobic and semi-anaerobic fermentation on glucose and benzaldehyde was simulated with the objective function of maximized biomass. The rates of glucose and benzaldehyde consumption during 2–8 h were calculated and inserted into the model. When oxygen uptake rate was unconstrained (aerobic fermentation), no production of L-PAC was

ro of

predicted by the model. However, high CO2 and low ethanol formation rates were observed and all of exogenously added benzaldehyde was oxidized to benzoic acid. The simulation

results showed a large increase in ethanol and benzyl alcohol production when oxygen uptake

-p

rate was set to zero in the model, but again there was no L-PAC production. By constraining ethanol and acetate secretion rates to 10 and 1 mmol gDCW-1 h-1, respectively, L-PAC

re

production was predicted by the model. At semi-anaerobic conditions (oxygen uptake rate of

lP

2 mmol gDCW-1 h-1), the ratio of L-PAC to benzyl alcohol secretion was increased. More addition of oxygen uptake rate did not further improve L-PAC production because of benzaldehyde depletion. Complete in silico design of the modified model with the flux

na

distributions are shown in the supplementary file (Supplementary 1). These simulation results are in agreement with the experimental observations and physiological characteristics of S.

ur

cerevisiae CEN.PK113-7D (Table 2).

Jo

3.3. Identifying gene deletion targets The results of flux distributions by FBA were used as wild type flux for MOMA calculations. With application of OptGene algorithm, six target genes, namely RPE1, PDA1, ADH3, ADH1, ZWF1, and PDC1 were suggested as the genes whose deletions could improve L-PAC production. These genes with their basic reactions are shown in Fig. 1 by red color and the consequence of deletion of each of these reactions can be seen. For instance, PDA1 and

9

PDC1 are genes encoding enzymes associated with pyruvic acid consumption and deletion of these genes is expected to enhance pyruvic acid availability as a substrate needed in benzaldehyde biotransformation to L-PAC. Also, deletion of ZWF1 which encodes the synthesis of first committed enzyme in the oxidative pentose phosphate pathway, leads to reduced NADPH regeneration. Hence, less NADPH will be available for the reactions leading to by-product accumulation. Deletion of ADH isozymes which use either NADH or NADPH as cofactor is expected to reduce the formation of benzyl alcohol or other reduced

ro of

forms of benzaldehyde and L-PAC. RPE1 is in the non-oxidative part of the pentose phosphate pathway and of great importance for the generation of NADPH. Insufficient NADPH reserves for biosynthesis in the null mutants of RPE1 or ZWF1, will increase

-p

NADP+ dependent mitochondrial malic enzyme flux to produce pyruvate (Blank et al., 2005) and subsequently more pyruvate will be available for L-PAC production.

re

For in silico design, the flux distributions in these mutants were obtained by setting the

lP

boundaries of related fluxes to zero (Supplementary 1). In compare with the reference strain, the flux towards L-PAC production increased around 14% (70% of the maximum theoretical value) in RPE1, PDA1, and ZWF1 null mutants. None of the other strains significantly

na

differed from the reference strain. Moreover, in silico simulation indicated no obvious influence on S. cerevisiae growth (Table 2).

ur

3.4. Experimental validation

Jo

In order to validate if the in silico identified single gene deletions could augment L-PAC production in vivo, the corresponding strains of the BY4741 background were cultivated semi-anaerobically in batch fermentation condition and the results were compared with those obtained from simulation (Table 2). For some cases, there is a significant difference between the simulated and experimental values. This could be explained by the fact that the GEM we used, is solely a stoichiometric model and lacks regulatory and kinetic information.

10

The growth curves of these mutants in YPD and biotransformation media have been shown in Fig. 3. Except for ∆adh1 mutant, all other mutants grew well and no marked changes in the growth rates were observed as compared with the parental strain. Similar results were obtained by Shirai et al. (2013) for single gene deletion of PDC1, PDA1 and ZWF1. Although it is expected that some mutations in central metabolism hinder growth on glucose, a majority of them use alternative enzymes or pathways, which obtain a robust phenotype for cell growth (Blank et al., 2005).

ro of

In line with our results, Reiner et al. (2006) reported reduced growth rate for the ∆adh1 yeast strain under anaerobic conditions probably due to low ergosterol content for this mutant

(Marisco et al. 2011). It should be noted that the initial OD600 of all deletion strains in the

-p

biotransformation media were in the range of 9-10, except for ∆adh1 where it was equal to 2. Benzaldehyde, L-PAC and benzyl alcohol titers were followed in batch fermentation for the

re

six single gene deletion and parental strains (Fig. 4). Compared to the parental strain, the

lP

∆rpe1, ∆adh3, ∆adh1 strains showed a lower benzaldehyde consumption rate during the first hours. With the exception of ∆pda1 and ∆pdc1, lower benzyl alcohol accumulation was observed for the mutants.

na

The highest L-PAC concentration was obtained for the ∆zwf1 which was 2.48 g/L from 2 g/L of benzaldehyde (88 % of the theoretical yield). In fact, the major role of oxidative part of the

ur

pentose phosphate pathway is the generation of NADPH. Therefore, deletion of ZWF1 which

Jo

encodes glucose dehydrogenase as the active enzyme for the first reaction of the pathway, reduces NADPH levels. It has recently been reported that benzyl alcohol formation is dependent on NADPH rather than NADH (Bruder & Boles, 2017). As the reduction of benzaldehyde to benzyl alcohol is carried out by a variety of oxidoreductases, aside to NADH regeneration, there could be several NADPH-dependent oxidoreductase activities for benzyl alcohol formation. For instance, aldo-keto reductases (AKRs) which are a superfamily of

11

enzymes that have a broad substrate specificity, convert aldehydes and ketones to alcohols in the presence of NADPH. Unlike AKRs, ADHs use either NADH or NADPH as their cofactors. Our results also confirmed lower benzyl alcohol formation in ∆zwf1 mutants as a result of less NADPH generation (Fig. 4.c). As such, it is the availability of NADPH and not just NADH necessary for benzaldehyde reduction to benzyl alcohol. Based on the fermentation results, deletion of PDA1 improves L-PAC production to 1.8 g/L, although it increased benzyl alcohol formation as well. Inactivation of the mitochondrial

ro of

pyruvate dehydrogenase complex in the PDA1 mutant is concomitant with the formation of petite mutants which lack mitochondrial DNA during growth on glucose (Wenzel et al.,

1992). We have previously reported higher L-PAC titers in yeast petite mutants because of

-p

accumulation of acetohydroxyacid synthase (AHAS) in the cytosol (Doostmohammadi et al., 2016). The ability of AHAS to catalyze condensation of pyruvate with benzaldehyde and

re

formation of L-PAC has been reported (Engel et al., 2003).

lP

Glucose was completely consumed by all strains except the ∆adh1 strain. Ethanol concentration reached 10-15 g/L and minimal quantities of acetate were formed (data not shown). Benzaldehyde was depleted after approximately 10 h and maximum L-PAC titers

na

reached after approximately 8-10 h. Fig. 5 compares the production level of L-PAC and benzyl alcohol for all strains used in this study after 10 h.

ur

Decrease in benzyl alcohol formation in ∆adh1 strain is not just because of ADH deletion, as

Jo

cell growth was severely retarded which was related to the reduced glucose uptake rate. However, deletion mutant of mitochondrial ADH3 which involves in the shuttling of mitochondrial NADH to the cytosol under anaerobic condition, did not affect growth rate significantly and this mutant produced benzyl alcohol to some extent. Previous studies showed that even in ADH1, ADH2 and ADH3 deletion mutants, benzyl alcohol was still produced (Nikolova & Ward, 1991) which indicated the presence of more than one type of

12

oxidoreductase in the yeast. Furthermore, the reduction in ADH activity increases NADH availability which may subsequently inhibit the pyruvate dehydrogenase complex and also affect the glycolytic flux (Oliver et al., 1997). Thus, the reduction in pyruvate accumulation results in lower L-PAC production. 4. Conclusions Novel target genes were identified by GEM and FBA with OptGene algorithm to improve LPAC production in the yeast S. cerevisiae. The results of this study demonstrated that L-PAC

ro of

production in S. cerevisiae can be efficiently improved with the aid of in silico metabolic engineering. L-PAC titers were obtained up to 2.48 g/L, equivalent to 88% of the theoretical yield by using the ∆zwf1 strain. Findings of the present study can provide an insight to the

-p

future works on the strain improvement to increase L-PAC production and demonstrate the

successful utilization of computationally guided genetic manipulation to enhance metabolic

lP

re

capacity.

Declaration of interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

na

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

ur

Appendix A. Supplementary data

Jo

Supplementary information related to this article can be found in online version of the paper.

13

References Abourashed, E.A., El-Alfy, A.T., Khan, I.A., Walker, L. 2003. Ephedra in perspective–a current review. Phytotherapy Research, 17, 703-712. Agarwal, P.K., Uppada, V., Swaminathan, A., Noronha, S.B. 2015. Engineering of yeast pyruvate decarboxylase for enhanced selectivity towards carboligation. Bioresource Technology, 192, 90-96. Agren, R., Otero, J.M., Nielsen, J. 2013. Genome-scale modeling enables metabolic

ro of

engineering of Saccharomyces cerevisiae for succinic acid production. Journal of Industrial Microbiology & Biotechnology, 40, 735-747.

Alper, H., Jin, Y.-S., Moxley, J., Stephanopoulos, G. 2005. Identifying gene targets for the

-p

metabolic engineering of lycopene biosynthesis in Escherichia coli. Metabolic Engineering, 7, 155-164.

re

Alper, H., Stephanopoulos, G. 2004. Metabolic engineering challenges in the post-genomic

lP

era. Chemical Engineering Science, 59, 5009-5017.

Andreu, C., ·lí del Olmo, M.. 2018. Biotransformation using halotolerant yeast in seawater: a sustainable strategy to produce R-(−)-phenylacetylcarbinol. Applied Microbiology and

na

Biotechnology, 102, 4717-4727.

Asadollahi, M.A., Maury, J., Patil, K.R., Schalk, M., Clark, A., Nielsen, J. 2009. Enhancing

ur

sesquiterpene production in Saccharomyces cerevisiae through in silico driven

Jo

metabolic engineering. Metabolic Engineering, 11, 328-334. Blank, L.M., Kuepfer, L., Sauer, U. 2005. Large-scale 13C-flux analysis reveals mechanistic principles of metabolic network robustness to null mutations in yeast. Genome Biology, 6, R49.

14

Bro, C., Regenberg, B., Förster, J., Nielsen, J. 2006. In silico aided metabolic engineering of Saccharomyces cerevisiae for improved bioethanol production. Metabolic Engineering, 8, 102-111. Brochado, A.R., Matos, C., Møller, B.L., Hansen, J., Mortensen, U.H., Patil, K.R. 2010. Improved vanillin production in baker's yeast through in silico design. Microbial Cell Factories, 9, 84. Bruder, S., Boles, E. 2017. Improvement of the yeast based (R)-phenylacetylcarbinol

ro of

production process via reduction of by-product formation. Biochemical Engineering Journal, 120, 103-112.

Doostmohammadi, M., Asadollahi, M.A., Nahvi, I., Biria, D., Ghezelbash, G.R.,

Annals of Microbiology, 66, 1049-1055.

-p

Kheyrandish, M. 2016. L-phenylacetylcarbinol production by yeast petite mutants.

re

Doostmohammadi, M., Esfandiari, M., Asadollahi, M.A., Kamali, M., Nahvi, I. 2015.

lP

Biotransformation of benzaldehyde to L-phenylacetyl carbinol using immobilized cells of Saccharomyces cerevisiae. Minerva Biotecnologica, 27, 43-49. Engel, S., Vyazmensky, M., Geresh, S., Barak, Z.e., Chipman, D.M. 2003. Acetohydroxyacid

na

synthase: A new enzyme for chiral synthesis of R-phenylacetylcarbinol. Biotechnology and Bioengineering, 83, 833-840.

ur

Förster, J., Famili, I., Fu, P., Palsson, B.Ø., Nielsen, J. 2003. Genome-scale reconstruction of

Jo

the Saccharomyces cerevisiae metabolic network. Genome Research, 13, 244-253. Gunawan, C., Breuer, M., Hauer, B., Rogers, P.L., Rosche, B. 2008. Improved (R)phenylacetylcarbinol production with Candida utilis pyruvate decarboxylase at decreased organic to aqueous phase volume ratios. Biotechnology Letters, 30, 281286.

15

Gupta, K., Singh, J., Sahni, G., Dhawan, S. 1979. Production of phenyl acetyl carbinol by yeasts. Biotechnology and Bioengineering, 21, 1085-1089. Hendry, J.I., Prasannan, C.B., Joshi, A., Dasgupta, S., Wangikar, P.P. 2016. Metabolic model of Synechococcus sp. PCC 7002: prediction of flux distribution and network modification for enhanced biofuel production. Bioresource Technology, 213, 190-197. Hoyos, P., Sinisterra, J.-V., Molinari, F., Alcántara, A.R., Domínguez de María, P. 2010. Biocatalytic strategies for the asymmetric synthesis of α-hydroxy ketones. Accounts of

ro of

Chemical Research, 43, 288-299. Kaushal, M., Chary, K.V.N., Ahlawat, S., Palabhanvi, B., Goswami, G., Das, D. 2018. Understanding regulation in substrate dependent modulation of growth and

-p

production of alcohols in Clostridium sporogenes NCIM 2918 through metabolic

network reconstruction and flux balance analysis. Bioresource Technology, 249, 767-

re

776.

lP

Kerkhoven, E.J., Lahtvee, P.J., Nielsen, J. 2015. Applications of computational modeling in metabolic engineering of yeast. FEMS Yeast Research, 15, 1-13. Khan, T.R., Daugulis, A.J. 2010. Application of solid–liquid TPPBs to the production of L-

na

phenylacetylcarbinol from benzaldehyde using Candida utilis. Biotechnology and Bioengineering, 107(4), 633-641.

ur

Lee, J.-H., Wendisch, V.F. 2017. Production of amino acids–genetic and metabolic

Jo

engineering approaches. Bioresource Technology, 245, 1575-1587. Li, H., Liu, N., Hui, X., Gao, W.-Y. 2017. An improved enzymatic method for the preparation of (R)-phenylacetyl carbinol. RSC Advances, 7, 32664-32668.

Li, X., Wang, Z., Zhang, G., Yi, L. 2019. Improving lycopene production in Saccharomyces cerevisiae through optimizing pathway and chassis metabolism. Chemical Engineering Science, 193, 364-369.

16

Long, A., Ward, O. 1989. Biotransformation of benzaldehyde by Saccharomyces cerevisiae: characterization of the fermentation and toxicity effects of substrates and products. Biotechnology and Bioengineering, 34, 933-941. Marisco, G., Saito, S.T., Ganda, I.S., Brendel, M., Pungartnik, C. 2011. Low ergosterol content in yeast adh1 mutant enhances chitin maldistribution and sensitivity to paraquat‐ induced oxidative stress. Yeast 28, 363-373. Neuberg, C., Hirsch, J. 1921. An enzyme which brings about union into carbon chains

ro of

(Carboligase). Biochemische Zeitschrift, 115, 282-310. Ng, C.Y., Jung, M.-y., Lee, J., Oh, M.-K. 2012. Production of 2,3-butanediol in

Saccharomyces cerevisiae by in silico aided metabolic engineering. Microbial Cell

-p

Factories, 11, 68.

Nikolova, P., Ward, O.P. 1991. Production of L-phenylacetyl carbinol by biotransformation:

re

Product and by-product formation and activities of the key enzymes in wild-type and

lP

ADH isoenzyme mutants of Saccharomyces cerevisiae. Biotechnology and Bioengineering, 38, 493-498.

Oliver, A.L., Roddick, F.A., Anderson, B.N. 1997. Cleaner production of

na

phenylacetylcarbinol by yeast through productivity improvements and waste minimisation. Pure and Applied Chemistry, 69, 2371-2385.

ur

Österlund, T., Nookaew, I., Nielsen, J. 2012. Fifteen years of large scale metabolic modeling

Jo

of yeast: Developments and impacts. Biotechnology Advances, 30, 979-988. Otero, J.M., Cimini, D., Patil, K.R., Poulsen, S.G., Olsson, L., Nielsen, J. 2013. Industrial systems biology of Saccharomyces cerevisiae enables novel succinic acid cell factory. PLOS One, 8, e54144. Patil, K.R., Rocha, I., Förster, J., Nielsen, J. 2005. Evolutionary programming as a platform for in silico metabolic engineering. BMC Bioinformatics, 6, 308.

17

Pereira, R., Nielsen, J., Rocha, I. 2016. Improving the flux distributions simulated with genome-scale metabolic models of Saccharomyces cerevisiae. Metabolic Engineering Communications, 3, 153-163. Reiner, S., Micolod, D., Zellnig, G., Schneiter, R. 2006. A genomewide screen reveals a role of mitochondria in anaerobic uptake of sterols in yeast. Molecular Biology of the Cell, 17, 90-103. Rosche, B., Breuer, M., Hauer, B., Rogers, P.L. 2003. Screening of yeasts for cell-free

ro of

production of (R)-phenylacetylcarbinol. Biotechnology Letters, 25, 841-845. Rosche, B., Leksawasdi, N., Sandford, V., Breuer, M., Hauer, B., Rogers, P. 2002. Enzymatic (R)-phenylacetylcarbinol production in benzaldehyde emulsions. Applied

-p

Microbiology and Biotechnology, 60, 94-100.

Schellenberger, J., Que, R., Fleming, R.M., Thiele, I., Orth, J.D., Feist, A.M., Zielinski, D.C.,

re

Bordbar, A., Lewis, N.E., Rahmanian, S. 2011. Quantitative prediction of cellular

lP

metabolism with constraint-based models: the COBRA Toolbox v2. 0. Nature Protocols, 6, 1290-1307.

Sehl, T., Bock, S., Marx, L., Maugeri, Z., Walter, L., Westphal, R., Vogel, C., Menyes, U.,

na

Erhardt, M., Müller, M., Pohl, M., Rother, D. 2017. Asymmetric synthesis of (S)phenylacetylcarbinol–closing a gap in C–C bond formation. Green Chemistry 19,

ur

380-384.

Jo

Shirai, T., Matsuda, F., Okamoto, M., Kondo, A. 2013. Evaluation of control mechanisms for Saccharomyces cerevisiae central metabolic reactions using metabolome data of eight single-gene deletion mutants. Applied Microbiology and Biotechnology, 97, 35693577. Shukla, V., Kulkarni, P. 2000. L-Phenylacetylcarbinol (L-PAC): biosynthesis and industrial applications. World Journal of Microbiology and Biotechnology, 16, 499-506.

18

Stephanopoulos, G., Stafford, D.E. 2002. Metabolic engineering: a new frontier of chemical reaction engineering. Chemical Engineering Science, 57, 2595-2602. Strucko, T., Zirngibl, K., Pereira, F., Kafkia, E., Mohamed, E.T., Rettel, M., Stein, F., Feist, A.M., Jouhten, P., Patil, K.R. 2018. Laboratory evolution reveals regulatory and metabolic trade-offs of glycerol utilization in Saccharomyces cerevisiae. Metabolic Engineering, 47, 73-82. van Hoek, P., van Dijken, J.P., Pronk, J.T. 2000. Regulation of fermentative capacity and

Enzyme and Microbial Technology, 26, 724-736.

ro of

levels of glycolytic enzymes in chemostat cultures of Saccharomyces cerevisiae.

Wenzel, T.J., van den Berg, M.A., Visser, W., vn den Berg, J.A., Steensma, H.Y. 1992.

-p

Characterization of Saccharomyces cerevisiae mutants lacking the E1α subunit of the

Jo

ur

na

lP

re

pyruvate dehydrogenase complex. European Journal of Biochemistry, 209, 697-705.

19

Table 1. S. cerevisiae mutants used in this research Mutants

Genotype

Cellular

Basic Reaction Encoded Enzyme

compartment ∆rpe1

BY4741;

cytosolic

RU5P→X5P

YJL121c::kanMX4 ∆pda1

D-Ribulose 5-phosphate 3-epimerase

BY4741;

mitochondrial

Pyr→AcCoA

YER178w::kanMX4

E1 alpha subunit of the pyruvate dehydrogenase complex

BY4741;

mitochondrial

Acetald→EtOH

YMR083w::kanMX4 ∆adh1

BY4741;

isozyme III

cytosolic

Acetald→EtOH

YOL086c::kanMX4 BY4741;

cytosolic

YNL241c::kanMX4 BY4741;

cytosolic

G6P→6PG

Pyr→Acetald

re

∆pdc1

Alcohol dehydrogenase isozyme I

-p

∆zwf1

Alcohol dehydrogenase

ro of

∆adh3

Jo

ur

na

lP

YLR044c::kanMX4

20

Glucose 6-phosphate dehydrogenase

Pyruvate decarboxylase isozyme I

Table 2. Computational and experimental comparative data for CEN.PK113-7D, BY4741 and six single gene deletion strains. Specific growth rate (µmax; h-1) and specific consumption/ productivity values (q; mmol gDCW-1 h-1) are reported for major metabolites (glucose, benzaldehyde, L-PAC and benzyl alcohol). Note that in silico design for CEN.PK113-7D and BY4741 is similar; simulated qglucose and qbenzaldehyde are fixed at 10 and 1 mmol gDCW-1 h-1, respectively. µmax

qglucose

qbenzaldehyde

qL-PAC

qbenzyl alcohol

Sim.

Exp.

Sim.

Exp.

Sim.

Exp.

Sim.

Exp.

Sim.

CEN.PK113-7D

0.19

0.22

12.40

10

0.86

1

0.51

0.86

0.40

0.14

Reference

0.13

0.22

10.80

10

0.71

1

0.33

0.86

0.22

0.14

Δrpe1

0.09

0.21

7.90

10

0.69

1

Δpda1

0.08

0.21

7.90

10

0.60

1

Δadh1

0.03

0.22

2.76

10

0.34

1

Δadh3

0.15

0.22

9.82

10

0.82

1

Δzwf1

0.15

0.21

11.90

10

0.68

1

Δpdc1

0.11

0.22

9.50

10

0.50

1

ro of

Exp.

1

0.30

0

0.65

1

0.37

0

0.11

0.86

0.03

0.14

0.50

0.93

0.23

0.06

0.83

1

0.19

0

0.20

0.86

0.34

0.14

Jo

ur

na

lP

re

-p

0.42

21

Figure Captions: Fig. 1. Simplified schematic pathway of S. cerevisiae for biotransformation of benzaldehyde to L-PAC along with important side reactions. The exogenously added benzaldehyde (BZA) (blue box) reacts with activated acetaldehyde (Active-ACA) to produce L-PAC (green box). Main by-products benzoic acid (BZOA) and benzyl alcohol (BZOH) emerge respectively from oxidation and reduction of benzaldehyde, while PAC-diol as a minor by-product forms by reduction of L-PAC (orange boxes). Red descriptions are suggested by OptGene as the

ro of

genes whose deletions could improve L-PAC production. Fig. 2. Concentration profiles of substrates and products during biotransformation of

benzaldehyde to L-PAC by S. cerevisiae CEN.PK113-7D. a) Concentration profiles of

-p

benzaldehyde, benzyl alcohol and L-PAC, b) Concentration profiles of glucose, ethanol,

acetate and cell growth profile. Error bars represent the standard deviation of triple measured

re

values.

lP

Fig. 3. Growth profiles of BY4741 parental strain and six single gene deletion mutants in a) YPD media, b) Biotransformation media. Closed circles BY4741, closed squares ∆rpe1,

squares ∆pdc1.

na

closed triangles ∆pda1, closed diamonds ∆adh3, crosses ∆adh1, open circles ∆zwf1, open

Fig. 4. Time-course of a) benzaldehyde consumption; b) L-PAC production; c) benzyl

ur

alcohol production by the BY4741 parental strain and six single gene deletion mutants.

Jo

Closed circles BY4741, closed squares ∆rpe1, closed triangles ∆pda1, closed diamonds ∆adh3, crosses ∆adh1, open circles ∆zwf1, open squares ∆pdc1. The error bars indicate the standard deviations for triplicate experiments. Fig. 5. Production level of L-PAC and benzyl alcohol after 10 h by BY4741 parental strain and six single gene deletion mutants.

22

ro of

-p

re

lP

na

ur

Jo Fig. 1

23

ro of

-p

re

lP

na

ur

Jo Fig. 2

24

ro of

-p

re

lP

na

ur

Jo Fig. 3

25

ro of

-p

re

lP

na

ur

Jo Fig. 4

26

ro of

-p

re

lP

na

ur

Jo Fig. 5

27