Applied Microbiology and Biotechnology

, Volume 93, Issue 1, pp 131–141

Effect of alternative NAD+-regenerating pathways on the formation of primary and secondary aroma compounds in a Saccharomyces cerevisiae glycerol-defective mutant


  • Vishist K. Jain
    • Institute for Wine BiotechnologyStellenbosch University
  • Benoit Divol
    • Institute for Wine BiotechnologyStellenbosch University
    • Institute for Wine BiotechnologyStellenbosch University
  • Florian F. Bauer
    • Institute for Wine BiotechnologyStellenbosch University
Biotechnological Products and Process Engineering

DOI: 10.1007/s00253-011-3431-z

Cite this article as:
Jain, V.K., Divol, B., Prior, B.A. et al. Appl Microbiol Biotechnol (2012) 93: 131. doi:10.1007/s00253-011-3431-z


Saccharomyces cerevisiae maintains a redox balance under fermentative growth conditions by re-oxidizing NADH formed during glycolysis through ethanol formation. Excess NADH stimulates the synthesis of mainly glycerol, but also of other compounds. Here, we investigated the production of primary and secondary metabolites in S. cerevisiae strains where the glycerol production pathway was inactivated through deletion of the two glycerol-3-phosphate dehydrogenases genes (GPD1/GPD2) and replaced with alternative NAD+-generating pathways. While these modifications decreased fermentative ability compared to the wild-type strain, all improved growth and/or fermentative ability of the gpd1Δgpd2Δ strain in self-generated anaerobic high sugar medium. The partial NAD+ regeneration ability of the mutants resulted in significant amounts of alternative products, but at lower yields than glycerol. Compared to the wild-type strain, pyruvate production increased in most genetically manipulated strains, whereas acetate and succinate production decreased in all strains. Malate production was similar in all strains. Isobutanol production increased substantially in all genetically manipulated strains compared to the wild-type strain, whereas only mutant strains expressing the sorbitol producing SOR1 and srlD genes showed increases in isoamyl alcohol and 2-phenyl alcohol. A marked reduction in ethyl acetate concentration was observed in the genetically manipulated strains, while isobutyric acid increased. The synthesis of some primary and secondary metabolites appears more readily influenced by the NAD+/NADH availability. The data provide an initial assessment of the impact of redox balance on the production of primary and secondary metabolites which play an essential role in the flavour and aroma character of beverages.


RedoxSaccharomyces cerevisiaeHigher alcoholOrganic acidsFermentation


Yeast is known to contribute significantly to the final flavour and aroma profile or bouquet of wine and other fermentation-derived beverages (Lambrechts and Pretorius 2000; Vanderhaegen et al. 2003; Xu et al. 2006). The most relevant yeast-derived flavour and aroma compounds in these products include organic acids, esters, volatile fatty acids and higher alcohols. Production of many of these metabolites is thought to impact on or be impacted by the ratio of NADH/NAD+ and therefore may contribute to the maintenance of the redox balance of the cell (Fig. 1) (van Dijken and Scheffers 1986; Schoondermark-Stolk et al. 2005). However, another hypothesis regarding the production of some of these compounds includes the removal of toxic compounds such as medium chain fatty acids (Boulton et al. 1996).
Fig. 1

Diagrammatic representation of pathways associated with primary and secondary metabolite production in S. cerevisiae (Rossouw et al. 2008). Dashed arrows are used when one or more intermediates or reactions are omitted. Red font is used to identify relevant aroma compounds. The genes encoding enzymes in the biosynthesis are ATF alcohol acetyltransferase, ADH alcohol dehydrogenase, ALD aldehyde dehydrogenase, PDC pyruvate decarboxylase, SFA alcohol dehydrogenase, BAT branched chain amino acid transaminase, THI α-ketoisocaproate decarboxylase, ARO aromatic amino acid decarboxylase, CHA threonine dehydratase

Formation of ethanol results in the oxidation of NADH formed during glycolysis. However, excess NADH is generated when the intermediates of glycolysis are withdrawn for synthesis of cellular material and also when oxidized by-products such as acetate and succinate are formed (van Dijken and Scheffers 1986). This accumulation of NADH which cannot be re-oxidized through ethanol formation will lead to metabolism cessation unless the surplus NADH can be converted back to NAD+. Aerobically, this conversion occurs via the electron transport chain where oxygen is the final electron acceptor, but under anaerobic conditions, most re-oxidation occurs by NADH-coupled reduction of dihydroxyacetone phosphate (DHAP) to glycerol-3-phosphate which is ultimately converted to glycerol (Albertyn et al. 1994).

The genes involved in the conversion of DHAP to glycerol are GPD1 and GPD2 (Ansell et al. 1997). The two isoenzymes for yeast NAD+-dependent glycerol-3-phosphate dehydrogenase encoded by GPD1 and GPD2 have distinct roles in osmoadaptation and redox regulation. GPD1 is normally induced by high osmolarity (Albertyn et al. 1994), and GPD2 induction is governed by the NADH/NAD+ ratio (Eriksson et al. 1995). A gpd1Δgpd2Δ double mutant is unable to grow under strict anaerobic conditions because NAD+ regeneration through glycerol production is no longer possible and NADH is accumulated intracellularly (Ansell et al. 1997). In principle, this excess NADH can be used to drive other NADH-dependent reduction reactions. For this purpose, alternative NADH-coupled pathways were introduced into the gpd1Δgpd2Δ mutant strain by overexpressing alternative oxidoreductase genes as described in Jain et al. (2011). Here, we investigate the impact of these manipulations on the formation of other primary and secondary metabolites by comparing production between the wild-type and gpd1Δgpd2Δ strains when cultivated in a high sugar medium under self-generated anaerobic conditions similar to those found in the production processes of alcoholic beverages (Fig. 2; Jain et al. 2011). Indeed, while the biochemical pathways leading to the production of metabolites in yeast have been reasonably well studied and documented (Lambrechts and Pretorius 2000), many questions regarding the physiological role and the genetic and metabolic regulation of this network remain to be addressed. These pathways (Fig. 1) form a complex network of shared intermediates and reversible reactions (Rossouw et al. 2008), and little information exists regarding the impact of redox balancing on in particular flavour and aroma active compounds. The aim of this study therefore is to specifically investigate the link between the requirement for NAD+ regeneration and the formation of biotechnologically relevant metabolites.
Fig. 2

Pathways of glucose and fructose catabolism by S. cerevisiae strains: 1 wild-type (WT), 2 gpd1Δgpd2Δ (DM) with glycerol-3-phosphate dehydrogenase encoding genes deleted in WT, 3 gpd1Δgpd2Δ expressing a sorbitol dehydrogenase [DM(SOR1)], 4 gpd1Δgpd2Δ expressing a sorbitol-6-phosphate dehydrogenase [DM(srlD)], 5 gpd1Δgpd2Δ expressing glycerol dehydrogenase, aldose reductase and methylglyoxal synthase [DM(gldA, GRE3, mgsA)], 6 gpd1Δgpd2Δ expressing pyruvate decarboxylase and the ALD6 gene encoding acetaldehyde dehydrogenase deleted [DM(ald6Δ, PDC1)]. Mean molar concentrations (±standard deviations of quadruplicate determinations) of glycerol, sorbitol and 1,2 propanediol determined after 20 days of fermentation are shown

Materials and methods

Yeast strains and strain construction

Saccharomyces cerevisiae BY4742 (wild type) and BY4742 gpd1Δ strains were obtained from the Euroscarf deletion library ( The gpd1Δgpd2Δ strain (DM) was created by deleting GPD2 gene from the BY4742 gpd1Δ strain as described previously (Jain et al. 2011). DM was the recipient strain for various gene constructs (Fig. 2). The DM(srlD), DM(SOR1) and DM(gldA, GRE3, mgsA) strains were constructed by inserting genes encoding respectively a sorbitol-6-phosphate dehydrogenase, a sorbitol dehydrogenase and a glycerol dehydrogenase, aldose reductase and methylglyoxal synthase (Jain et al. 2011). The DM(ald6Δ) strain was constructed by deleting the ALD6 gene in DM by using LYS2 as auxotrophic marker. The LYS2 gene with a 50-bp region homologous to the 5′ and 3′ parts of ALD6 gene in Ydpk vector (Berben et al. 1991) was amplified and integrated into the DM to obtain DM(ald6Δ). The primers used to amplify the genes are as described previously (Jain et al. 2011).

Plasmids and cell transformation

The SOR1, srlD, gldA and PDC1 PCR products were cloned into pDMPM multicopy shuttle vector (Malherbe 2010), whereas the GRE3 gene was cloned into pSTAH (Gururajan et al. 2007) integrating vector and the mgsA gene was cloned into pSTAL vector (Jain et al. 2011). The complete description of the vectors including the insertion of 2-μ origin of replication gene in pSTAH was described previously (Jain et al. 2011). The pDMPM(SOR1) and pDMPM(srlD) plasmids were transformed into DM resulting in the DM(SOR1) and DM(srlD) strains, respectively. The pDMPM(gldA), pSTAH(GRE3) and pSTAL(mgsA) plasmids were transformed into DM to obtain DM(gldA, GRE3, mgsA) strain. Lastly, pDMPM(PDC1) plasmid was transformed into DM(ald6Δ) to obtain DM(ald6Δ, PDC1) strain. Prior to transformation, the genes were sequenced at the Central Sequencing Facility of Stellenbosch University. The sequences for the SOR1, srlD, gldA, GRE3, mgsA, PDC1 genes were 100% homologous to the respective sequence accession numbers P35497 and P05707, P0A9S5, P38715, P0A731, P06169 lodged in GenBank ( Subcloning in Escherichia coli DH5α, yeast and bacterial transformations and isolation of genomic DNA from E. coli and S. cerevisiae were done using standard protocols (Gietz and Schiestl 2007; Harju et al. 2004).

Medium and fermentation conditions

The fermentations were conducted in quadruplicate as described previously (Jain et al. 2011). Briefly, the fermentations were carried out in 250-ml Erlenmeyer flasks with a 100-ml working volume without agitation and fitted with a fermentation cap filled with sterile water to restrict air permeation. A self-generated anaerobic system was created during fermentation in the flasks by O2 consumption and CO2 release and confirmed by a methylene blue (2 mg/l) colour change from blue to colourless (Gordon and Dubos 1970). The liquid medium for the growth of strains contained 10% total sugar (5% glucose and 5% fructose as is found under winemaking conditions) and 6.7 g/l yeast nitrogen base (YNB; Difco, Franklin Lakes, NJ, USA) without amino acids. The medium was supplemented with amino acids or nucleotides required for the growth of these auxotrophic strains. Supplements added were uracil, leucine, lysine and histidine to a final concentration of 24, 72, 36 and 24 mg/l, respectively. The pH and temperature were maintained at 3.5 and 30°C. The media were inoculated at an initial absorbance (600 nm) of 0.5, and fermentation was monitored by weight loss. The cultures were sampled each alternate day over 20 days, and metabolite analysis was carried out. The samples were centrifuged at 5,000 rpm for 10 min, the supernatants were filtered (0.2-μm membrane filter) and stored at 4°C until further analysis. Biomass was determined by resuspending cells in 1 ml water and centrifuging in 1.5-ml microcentrifuge tubes. The supernatant was removed, and the preweighed microcentrifuge tubes containing cells were dried at 60°C overnight until constant weight.

Plate assays of strain growth under aerobic and anaerobic conditions

Two sets of plates containing 6.7 g/l YNB without amino acids supplemented with amino acids as required and 2% agar were prepared. One set contained 2% glucose and the other 5% glucose and 5% fructose. For anaerobic plate assays, the medium was supplemented with 10 mg/l ergosterol and 420 mg/l Tween 80 (Bro et al. 2006). A culture grown aerobically at 30°C for 48 h in 2% glucose, 6.7 g/l YNB without amino acids supplemented with amino acids as required was diluted to an A600nm of 1, and 5-μl aliquots of tenfold dilution series were spotted on the plates. Plates were incubated for respectively 8 and 20 days under aerobic and anaerobic conditions at 30°C. Anaerobic conditions were maintained in 3.5-l anaerobic jars using the AnaeroGenTM system (Oxoid, Basingstoke, UK).

Chemical analyses

Metabolites formed were analysed using a Waters HPLC system equipped with an Aminex HPX-87H column (BioRad, Hercules, CA, USA) connected to a refractive index–ultraviolet detector (RID-6A, Shimadzu, Kyoto, Japan). A mobile phase of 5 mM H2SO4 at a flow rate of 0.6 ml/min and a column temperature of 45°C were used. The components measured using a refractive index detector were ethanol, sorbitol and glycerol, and the components measured using the ultraviolet detector were pyruvate, acetate, succinate and malate. Propane-1,2-diol was extracted (Louw et al. 2009) and the concentration determined by gas chromatography (Agilent 6890; Agilent Technologies, Santa Clara, CA, USA) with a Nukol free fatty acid phase fused-silica capillary column coupled to a mass spectrometer (Agilent 5975C) (Mulligan 1996). Higher alcohols, esters and minor acids formed by yeasts were extracted (Louw et al. 2009), and concentrations were determined with a gas chromatograph equipped with a flame ionization detector and Nukol free fatty acid phase fused-silica capillary column (Agilent 6890; Ng 2002).

Standardized biplot

A standardized biplot based on principal component analysis was used to graphically display relationships between variables (metabolite concentrations) as well as clustering of samples from different strains. A significant correlation between two variables is likely if p < 0.05. This implies that no matter how close or far the variables are on the biplot, the significance of the correlation between them is determined by the p value.


NAD+-regenerating pathways partially restore the growth of DM

All strains grew well under aerobic conditions on 2% glucose agar medium with the gpd1Δgpd2Δ strain however showing slightly reduced growth as compared to all other strains (Fig. 3). On 10% sugar agar medium under aerobic conditions, all the manipulated strains showed reduced growth when compared to the wild type (WT; Fig. 3), indicating a more significant requirement for NAD regeneration under these conditions. However, growth of the sorbitol producing DM(srlD) and DM(SOR1) and 1,2 propanediol producing DM(gldA, GRE3, mgsA) strains was significantly improved over the gpd1Δgpd2Δ strain unable to produce glycerol, indicating that expression of oxidoreductases and other genes had some impact on the ability of the strain to compensate for the absence of the glycerol NAD+ regeneration pathway in such conditions. No suppression of the growth defect was observed in the DM(ald6Δ, PDC1) strain whose growth was similar to gpd1Δgpd2Δ strain under these conditions (Fig. 3). Under anaerobic conditions, deletion of GPD1 and GPD2 genes resulted in little visible growth on the 2% glucose agar plate and no visible growth on the 10% sugar agar plate (Fig. 3). The slight growth of gpd1Δgpd2Δ strain on 2% glucose agar plate under anaerobic conditions may be due to the residual oxygen that was present in the aerobic inoculation culture. When complemented with SOR1 and srlD, the gpd1Δgpd2Δ strain showed some growth on both high and low sugar agar plates. On the other hand, no growth could be observed for the DM(gldA, GRE3, mgsA) and DM(ald6Δ, PDC1) strains under these anaerobic conditions. When the growth of the cultures was evaluated in liquid media containing 10% sugar in a self-generated anaerobic system as previously described (Jain et al. 2011), the biomass produced by gpd1Δgpd2Δ strain over 20 days was approximately half of that of the wild-type strain. As observed on agar plates, growth of the gpd1Δgpd2Δ strain complemented with SOR1, and srlD was substantially improved after 20 days, whereas growth by DM(gldA, GRE3, mgsA) was only slightly improved and DM(ald6Δ, PDC1) not at all when compared to the gpd1Δgpd2Δ strain (Table 1). Measurement of CO2 during fermentation confirmed the results of biomass measurement (Fig. 4). As previously described (Jain et al. 2011), the growth by the strains in liquid media was deemed to be adequate to evaluate the effect of the NAD+-generating pathways on the production of metabolites.
Fig. 3

Plate assay of S. cerevisiae strains under aerobic and anaerobic conditions in 2% glucose and 5% glucose–5% fructose and 6.7 g/l yeast nitrogen base

Table 1

Concentrations (mean ± standard deviation of four independent cultivations) of primary (grams per liter) and secondary (milligrams per liter) metabolites and dry biomass (grams per liter) formed by WT and genetically manipulated strains after cultivation for 20 days in 5% glucose–5% fructose and 6.7 g/l YNB under oxygen-limited conditions







DM(gldA, mgsA, GRE3)

DM(ald6Δ, PDC1)

Primary metabolites


45.17 ± 2.56

22.88 ± 1.68

36.31 ± 2.47

30.89 ± 2.34

31.78 ± 1.23

32.01 ± 1.68


0.16 ± 0.01

0.18 ± 0.004

0.85 ± 0.01

0.57 ± 0.01

0.72 ± 0.02

0.19 ± 0.004


0.38 ± 0.01

0.50 ± 0.06

0.43 ± 0.03

0.31 ± 0.04

0.33 ± 0.04

0.33 ± 0.01


0.67 ± 0.03

0.01 ± 0.004

0.01 ± 0.001

0.21 ± 0.03

0.01 ± 0.002

0.004 ± 0.0002


0.73 ± 0.04

0.27 ± 0.01

0.16 ± 0.02

0.28 ± 0.01

0.71 ± 0.01

0.26 ± 0.02

Secondary metabolites


18.53 ± 0.55

118.75 ± 3.02

85.52 ± 3.36

28.63 ± 1.46

282.53 ± 6.03

97.07 ± 1.72

  Isoamyl alcohol

25.09 ± 2.49

24.89 ± 2.07

71.23 ± 4.33

59.95 ± 1.06

29.38 ± 0.75

19.09 ± 2.92

  2-Phenyl ethanol

3.53 ± 0.44

3.09 ± 0.10

4.49 ± 0.14

4.83 ± 0.14

3.69 ± 0.35

2.91 ± 0.09

  Isobutyric acid

0.91 ± 0.12

2.14 ± 0.49

2.52 ± 0.37

1.03 ± 0.28

6.11 ± 0.15 1

68 ± 0.12

  Ethyl acetate

26.51 ± 1.72

4.63 ± 0.33

4.87 ± 0.09

11.22 ± 1.58

5.72 ± 0.30

1.58 ± 0.12


0.97 ± 0.008

0.41 ± 0.03

0.94 ± 0.02

0.88 ± 0.03

0.58 ± 0.02

0.38 ± 0.006
Fig. 4

Release of CO2 (mean ± standard deviation of four replicates) by the WT (open square), gpd1Δgpd2Δ (DM) (open triangle up), DM(SOR1) (open triangle down), DM(srlD) (open diamond) and DM(gldA, GRE3, mgsA) (filled hexagon) and DM(ald6Δ, PDC1) (open circle) when cultivated in 5% glucose–5% fructose and 6.7 g/l YNB

Metabolite and organic acid concentration varied among the strains

During alcoholic fermentation, ethanol is the most important primary metabolite produced in terms of re-oxidation of excess NADH and redox balancing, followed by the production of glycerol. Since the glycerol-producing pathway was deleted and replaced with other NAD+-generating pathways, we could test whether these changes might affect metabolite production and the redox balance. As expected, the WT produced the highest total amount of ethanol and the gpd1Δgpd2Δ strain the least (Table 1). Among the strains created to improve NAD+ regeneration, DM(SOR1) showed the highest final ethanol production capacity (Table 1). The insertion of the alternative NAD+-generating pathways into the gpd1Δgpd2Δ strain resulted in the production of sorbitol or propane-1,2-diol instead of glycerol by the wild-type strain. This allowed us to estimate the efficiency of NAD+ regeneration compared to the glycerol-producing pathway. The molar concentration of the DM(srlD) and DM(SOR1) strains produced respectively 27 ± 9 and 6 ± 1 mmol/g sorbitol, and DM(gldA, GRE3, mgsA) strain produced 38 ± 3 mmol/g propane-1,2-diol compared to 46 ± 1 mmol/g glycerol produced by the wild-type strain. This suggests that the inserted pathways were not as efficient in NAD+ regeneration as the glycerol-producing pathway.

The production of many organic acids during fermentation is also closely associated with redox balancing, and pyruvate, malate, acetate and succinate were measured in our conditions. The data show that pyruvate production by WT, DM and DM(ald6Δ, PDC1) strains was similar up to the second day of fermentation, and the concentration remained constant thereafter (Fig. 5a). However, the other strains such as DM(srlD), DM(SOR1) and DM(gldA, GRE3, mgsA) continued to produce pyruvate beyond the second day, and the final values were considerably higher than for the other strains (Fig. 5a, Table 1). Malate production was the highest in the gpd1Δgpd2Δ strain and fairly similar in all the other strains (Fig. 5b, Table 1). Acetate production was significantly reduced in the genetically manipulated strains compared to the WT strain, and only the DM(srlD) strain showed acetate production above 0.2 g/l (Fig. 5c, Table 1). This suggests a lack of NAD+ to oxidize acetaldehyde to acetate in the genetically manipulated strains (Fig. 1). Compared to the WT, succinate production was reduced in the genetically manipulated strains except DM(gldA, GRE3, mgsA) where the production pattern was similar to that of the WT strain (Fig. 5d, Table 1). This is rather surprising since genetic manipulation had little impact on malate production, and succinate would be expected to originate from malate (Fig. 1). The possibility exists of an alternative route to produce succinate from fumarate in S. cerevisiae which might be affected by the redox balance (Larsson et al. 1998).
Fig. 5

Pyruvate (a), malate (b), acetate (c), succinate (d), isobutanol (e), isoamyl alcohol (f), 2-phenyl ethanol (g), ethyl acetate (h) and isobutyric acid (i) production by the WT (open square), gpd1Δgpd2Δ (DM) (open triangle up), DM(SOR1) (open triangle down), DM(srlD) (open diamond), DM(gldA, GRE3, mgsA) (filled hexagon) and DM(ald6Δ, PDC1) (open circle) when cultivated in 5% glucose–5% fructose and 6.7 g/l YNB in a representative experiment

Secondary metabolite production by strains

The secondary metabolites isobutanol, isoamyl alcohol, 2-phenyl ethanol, isobutyric acid and ethyl acetate were produced in detectable concentrations (Table 1). The production of other metabolites such as propanol, butanol, ethyl butyrate, ethyl hexanoate, hexyl acetate, ethyl lactate, ethyl caprylate, propionic acid, ethyl caprate, isovaleric acid, valeric acid, 2-phenylethyl acetate, isoamyl acetate, hexanoic acid, octanoic acid, decanoic acid, butyric acid and amyl alcohol has been reported in other alcoholic fermentations by S. cerevisiae (Fig. 1, Lambrechts and Pretorius 2000; Rossouw et al. 2008). However, in this study, the concentrations of these compounds were below detection sensitivity of our instrumentation which might be due to the relatively low amino acid concentrations in the medium. Production of higher alcohols may also be driven by the necessity to regenerate NAD+ in the genetically manipulated strains. Indeed, the data showed that isobutanol production by all the genetically manipulated strains, especially DM(gldA, GRE3, mgsA), was higher than that of the WT strain, suggesting that the production of this compound could be an important route to regenerate NAD+ (Fig. 1). The isobutanol concentration produced by DM(srlD) was only slightly higher than the WT strain (Fig. 5e, Table 1) pointing to the relative efficiency of this oxidoreductase in maintaining the redox balance.

However, isoamyl alcohol production differed in the genetically manipulated strains compared to the WT strain with DM(ald6Δ, PDC1) showing lower production and gpd1Δgpd2Δ and DM(gldA, GRE3, mgsA) strains showing similar production compared to WT (Fig. 5f, Table 1). Only DM(SOR1) and DM(srlD) strains showed considerably greater production of this higher alcohol compared to WT strain. Interestingly, the conversion of isobutanal to isobutanol and 3-methyl butanal to isoamyl alcohol is catalyzed by the enzymes encoded by ADH1/5 and SFA1 yet the strains behaved substantially differently in their production (Hazelwood et al. 2008) (Fig. 1). Production of 2-phenyl ethanol was higher in DM(srlD) and DM(SOR1) strains, whereas gpd1Δgpd2Δ and DM(ald6Δ, PDC1) strains showed lower production compared to WT. DM(gldA, GRE3, mgsA) and WT strains produced similar concentrations of this compound (Fig. 5g, Table 1).

The ethyl acetate concentration produced by the WT strain was higher than that of all the other strains. This indicates that these genetic manipulations negatively impacted the ability of the strains to produce this metabolite (Fig. 5h, Table 1) similar to the reduced acetate levels (Fig. 5c) as there was apparently a lack of NAD+ to drive the catalysis (Fig. 1). Among genetically manipulated strains, DM(srlD) strain showed the highest capability and DM(ald6Δ, PDC1) strain the lowest capability to produce ethyl acetate. On the other hand, an opposite behaviour was observed for the production of isobutyric acid as the concentrations increased significantly in cultures inoculated with genetically manipulated strains as compared to the WT strain albeit to a different extent (Fig. 5i, Table 1). DM(gldA, GRE3, mgsA) showed a high production of isobutyric acid within the first 10 days after which the production ceased. In the other mutant strains, the production was complete within the first 4 days. This pattern of isobutyric acid production mirrored in part the pattern of isobutanol production (Fig. 5e) and may reflect the availability of excess isobutanal as a precursor for the synthesis of both compounds.

Comparative study of primary and secondary metabolite production by strains using standardized biplot analysis

A standardized biplot was drawn with primary and secondary metabolites formed by different strains as variables and projected on two dimensional axes. The first component on the x-axis of the biplot accounts for 42% of the variance, while the second component on the y-axis accounts for 34% of the variance, and together, they account for 76% of the variance of the data (Fig. 6). WT strain is placed separately from other strains and is characterized by higher production of acetate and ethyl acetate. DM(SOR1) and DM(srlD) strains are characterized by higher production of 2-phenyl ethanol and isoamyl alcohol and low production of succinate. The DM(gldA, GRE3, mgsA) strain is characterized by the higher production of isobutanol and isobutyric acid and lower production of ethyl acetate, whereas gpd1Δgpd2Δ and DM(ald6Δ, PDC1) strains cluster closely together and are characterized by the lower production of isoamyl alcohol and 2-phenyl ethanol as compared to other strains. Moreover, similar clustering of the strains was obtained in another biplot based on concentrations of primary and secondary metabolites and concentrations (millimoles per gram) of glycerol, sorbitol or propane-1,2-diol represented as NAD+-generating efficiency (data not shown). This suggests that clustering of strains was impacted not only by the production of NAD+ equivalents in the form of glycerol, sorbitol or propane-1,2-diol but could also be controlled by production of other metabolites which are not included as variables in biplot analysis shown in Fig. 6. One such example is methylglyoxal which is formed as an intermediate in the production of propane-1,2-diol by DM(gldA, GRE3, mgsA) strain.
Fig. 6

Standardized biplot of quadruplicate data reveals graphically the relationships between concentrations of metabolites formed by WT (diamond), gpd1Δgpd2Δ (DM) (triangle), DM(SOR1) (circle), DM(srlD) (filled square), DM(gldA, GRE3, mgsA) (cross) and DM(ald6Δ, PDC1) (open square)


This study has shown that the ability of a pathway to regenerate NAD+ has a major impact on the metabolite profile and biomass produced by S. cerevisiae when grown under fermentative conditions. The significantly lower biomass concentration produced by the DM(gldA, GRE3, mgsA) strain compared to WT and DM(SOR1) and DM(srlD) strains might be related to the toxicity of methylglyoxal (an intermediate of the inserted pathway) (Kalapos 1999) or due the NADPH specificity of the aldose reductase (Kuhn et al. 1995) which would direct this cofactor away from biomass synthesis. The much lower biomass of the gpd1Δgpd2Δ strain and DM(ald6Δ, PDC1) suggests that both of these strains lack NAD+ regeneration ability to enable growth to occur.

Influence of NAD+ regeneration capability on organic acid production

Under fermentative conditions, several factors such as NAD+ or NADH accessibility or availability of precursors might determine the production of organic acids (Fig. 1). For example, pyruvate production depends on the availability of glycolytic precursors. Once formed, most pyruvate is converted to ethanol to maintain the redox balance. However, some pyruvates might be exported from the cell due to low capacity of alcohol dehydrogenase at the start of fermentation. This hypothesis might be valid for the gpd1Δgpd2Δ, WT and DM(ald6Δ, PDC1) strains where pyruvate production increased until approximately the second day, and the concentration remained constant thereafter. However, in other strains, pyruvate export was much greater which could be due to inhibition of the ADH1 activity by sorbitol or propane-1,2-diol (Pronk et al. 1996). Moreover, oxidation of a non-glycolytic precursor such as methylglyoxal using 2-oxoaldehyde dehydrogenase (Murata et al. 1986) in DM(gldA, GRE3, mgsA) might also contribute to higher pyruvate production and export. Although 2-oxoaldehyde dehydrogenase activity has not yet been detected in S. cerevisiae (Murata et al. 1986), the catalysis of 2-keto aldehydes (such as methylglyoxal) by aldehyde dehydrogenases should be investigated.

The much lower concentrations of acetate in genetically manipulated strains compared to the WT strain point to lower availability of NAD+ (due to deletion of GPD1 and GPD2 genes). Among genetically manipulated strains, the DM(srlD) strain formed the highest concentration of acetate, suggesting that the sorbitol-6-phosphate dehydrogenase was most efficient in regenerating NAD+ compared to the other NAD+-generating pathways inserted into the other strains. The possibility of NADP+ production in the sorbitol-6-phosphate dehydrogenase catalyzed reaction together with acetate production through a NADP+-dependent aldehyde dehydrogenase (Saint-Prix et al. 2004) is unlikely as the E. coli sorbitol-6-phosphate dehydrogenase is strictly NAD+ dependent (Novotny et al. 1984).

Succinate is formed through the reductive pathway of the tricarboxylic acid cycle (Camarasa et al. 2003). The greater production of this compound by WT than the genetically manipulated strains was unexpected. Strains with the GPD1 and GPD2 genes deleted would need other routes to oxidize NADH, and the reductive pathway from pyruvate to succinate could assist (Fig.1). However, only the DM(gldA, GRE3, mgsA) strain produced similar succinate concentrations as the WT strain, suggesting that the other manipulated strains might lack other intermediates required to produce succinate. An alternative route of succinate production via the {gamma}-aminobutyric acid pathway has been described in S. cerevisiae (Bach et al. 2009), but it is unlikely to be significant under the conditions used in this study.

Regulation in the production of higher alcohols

Availability of precursors such as pyruvate and amino acids determines the production of higher alcohols (Fig. 1). Moreover, production of higher alcohols under fermentative conditions has been associated with the need to regenerate NAD+ (Schoondermark-Stolk et al. 2005). While growth stopped after 4 days in the self-generated anaerobic conditions, the genetically manipulated strains continued to produce some higher alcohols, whereas production entirely ceased in the WT, indicating that higher alcohol production is indeed sensitive to the NAD+ regeneration needs of yeast cells. Increased production of isobutanol was the most significant impact observed in the genetically manipulated strains, suggesting that NAD+ regeneration was occurring through the pathway from 2-keto-isovalerate to isobutanol via isobutanal (Fig. 1). Only the DM(srlD) strain showed a similar isobutanol production to WT (Fig. 5e) pointing again to the greater efficiency of sorbitol-6-phosphate dehydrogenase in regenerating NAD+ and a reduced need to regenerate NAD+ through isobutanol production. The comparison of the isoamyl alcohol and 2-phenyl ethanol production (Fig. 5f, g) revealed a less clear pattern, although the genetically manipulated strains consistently produced greater amounts of these two alcohols than WT. Furthermore, biplot analysis revealed a close relationship between the production of isoamyl alcohol and 2-phenyl ethanol (Fig. 6). However, other factors might possibly be involved in regulation of the secondary alcohol production.

Regulation of esters and fatty acid production

Production of esters in S. cerevisiae is regulated differently as compared to production of higher alcohols in terms of cofactor requirement. Production of higher alcohols regenerates NAD+, whereas production of acetate (an intermediate in the production of esters) requires NAD+ (Fig. 1). In this regard, the higher concentration of ethyl acetate in WT compared to the other strains reflects a possibly greater availability of NAD+ and acetate. The NAD+ concentration would be expected to be lower in the mutant strains as compared to the WT strain due to the deletion of GPD1 and GPD2 genes. This may explain the lower production of ethyl acetate by the genetically manipulated strains as compared to the WT strain. Among genetically manipulated strains, DM(srlD) showed the highest production of ethyl acetate, indicating that the sorbitol-6-phosphate dehydrogenase cloned into the strain was more efficient in regenerating NAD+ than the other manipulations.

Production of isobutyric acid was boosted in all the genetically manipulated strains as compared to WT. This might be related to the greater production of isobutanol which acts as the main precursor of isobutyric acid (Fig. 1). The inability to detect butanol, propanol and amyl alcohol production by any strains could indicate that the amino acid precursors were not present in sufficient concentrations for synthesis (Fig. 1). None of the strains including the WT formed isoamyl acetate or butyric acid. This was surprising as these compounds are usually formed by S. cerevisiae under fermentative conditions in grape must or rich media (Lambrechts and Pretorius 2000).

Among the amino acids added in the growth media of the strains in this study, only leucine is used in the Ehrlich pathway and is converted to isoamyl alcohol. However, isoamyl alcohol is also synthesized from pyruvate (Fig. 1). Leucine was added only in the growth medium of WT and DM strains, whereas other strains were able to synthesize this amino acid using LEU2 gene as a selectable marker. Interestingly, both gpd1Δgpd2Δ and WT strains produced similar and lower concentrations of isoamyl alcohol compared to DM(srlD) and DM(SOR1) strains, showing that the addition of amino acids did not affect the formation of isoamyl alcohol.

This study has shown that the need to restore the redox balance in the cell plays an important but not an exclusive role in the production of metabolites. The introduction of different pathways for NAD+ regeneration also influenced the production of secondary compounds in different ways. In the absence of the ability to form glycerol and therefore to regenerate NAD+, a higher production of higher alcohols under fermentative conditions was expected in the genetically manipulated strains. However, only isobutanol production was strongly stimulated. This work was carried out on laboratory yeast strains, and whether the same effects would be observed in industrial yeast strains need further investigation. The results set a platform for future metabolic engineering or breeding of S. cerevisiae for the alteration of the secondary metabolite profile.


This work was supported by the National Research Foundation, South Africa. We would like to thank Dr. D. F. Malherbe and Dr. V. T. Gururajan for kindly providing the pDMPM and pSTAH vectors, respectively, for cloning purposes. The authors would also like to thank Dr. A. Tredoux for technical assistance and Prof. Martin Kidd for statistical analysis.

Copyright information

© Springer-Verlag 2011