Introduction

Second generation approaches to bioethanol production from lignocellulosic materials such as agricultural and forestry biomass residues have the potential to substitute substantially for petroleum-based transportation fuels (Naik et al. 2010). The annual global production of lignocellulose amounts to approximately 50 billion tons (Claassen et al. 1999), and that volume is quite enough to develop environmentally friendly and sustainable bioethanol production to help meet current transportation fuel demands (Naik et al. 2010; Sims et al. 2010). The key technology to production of bioethanol is the development of an efficient system of degradation of lignocellulosic materials and its fermentation by the industrial workhorse Saccharomyces cerevisiae.

Numerous pretreatment technologies have been developed to degrade lignocellulosic materials. These include sulfuric acid treatment, alkaline treatment, steam explosion, and enzymatic degradation (Hu and Ragauskas 2012; Yang and Wyman 2008). However, these methods have several drawbacks. For example, sulfuric acid treatment generates enormous hazardous wastes (Yang and Wyman 2008; Yu et al. 2007). In contrast, hot-compressed water treatment is recognized as the one of the novel and advantageous pretreatment technologies (Yu et al. 2007). Water in either a subcritical or a supercritical stage or at temperatures above 150 °C with various pressures (5–22 MPa) is defined as hot-compressed water. It breaks down lignocelluloses into various compounds through pyrolytic cleavage, swelling, and dissolution of lignocellulose (Lu et al. 2009). A key advantage of using hot-compressed water is that its use during pretreatment directly breaks lignocelluloses down to fermentable sugars without generating hazardous wastes (Yu et al. 2007).

However, inhibitory compounds are produced during pretreatment of lignocellulosic materials (Lu et al. 2009), which is one of the greatest bottlenecks of progress in ensuring the efficiency of bioethanol production. So far, substances such as furan derivatives (furfural and 5-hydroxymethylfurfural (5-HMF)), acetic acid, methylglyoxal, and phenolic compounds have been extensively studied as the inhibitory compounds (Almeida et al. 2007; Olsson and Hahn-Hägerdal 1993; Palmqvist and Hahn-Hägerdal. 2000; Van Maris et al. 2006), and glycolaldehyde was not paid attention as an inhibitory compound until our former studies (Jayakody et al. 2011, 2012). Glycolaldehyde is formed through retro-aldol condensation of simple sugars such as glucose, fructose, mannose, and xylose when treated with hot-compressed water (Lu et al. 2009; Yu et al. 2007). The concentration of glycolaldehyde depends largely on the hydrolysis conditions as well as the type of biomass, with concentrations of glycolaldehyde generally ranging between 2 and 28 mM (0.12–1.68 g/l) after treatment of lignocelluloses with hot-compressed water (Ehara and Saka 2002; Lu et al. 2009). We first reported that the inhibitory activity of glycolaldehyde on the growth of S. cerevisiae is stronger than those of furfural and 5-HMF (Jayakody et al. 2011, 2012). Furthermore, we showed that the ADH1-overexpressing strain shows improved tolerance against glycolaldehyde and enhanced ethanol productivity in the presence of glycolaldehyde (Jayakody et al. 2012). However, growth of the developed ADH1-overexpressing strain was not fully recovered in the presence of glycolaldehyde. On the contrary, redox balance is now emerging as an intensive target of metabolic engineering (Celton et al. 2012; Heer et al. 2009; Kitagaki et al. 2009).

Here we report engineering redox cofactor utilization for further improvement of tolerance of engineered yeast to glycolaldehyde. Analyses of changes in the redox cofactors, certain metabolites, and gene expression enabled us to show that increased NADPH-dependent glycolaldehyde reduction activity is responsible for the improved conversion of glycolaldehyde and tolerance to glycolaldehyde of the strain overexpressing both GRE2 and ADH1, especially at 5 mM glycolaldehyde. To our best knowledge, this is the first successful attempt for engineering a yeast strain towards glycolaldehyde by introducing a NADPH-dependent reduction system. These novel findings and insights will likely suggest novel targets to alleviate the toxicities of fermentation inhibitors that hinder the efficient production of economically viable levels of bioethanol.

Materials and methods

Strains

Yeast strains used in the study are listed in Table 1. NEB 5-alpha competent Escherichia coli cells, obtained from New England Biolabs (Beverly, MA, USA), were used as the hosts for recombinant DNA manipulation.

Table 1 Yeast strains used in this study
Full size table

Construction of overexpression strains

The GRE2-overexpressing plasmid was constructed by following standard methods as previously described (Jayakody et al. 2012). Briefly, the GRE2 gene fragment including its own promoters was amplified by PCR from the genome of S. cerevisiae BY4743 by using forward and reverse primers (Table 2). The purified gene fragment of GRE2 was cleaved with appropriate restriction enzymes and ligated into pAUR123 plasmid. Success of transformation was confirmed by DNA sequencing. The pAUR123-GRE2 plasmid was transformed into BY4743 using a high efficiency yeast transformation technique (Gietz and Schiestl 2007). The transformants of pAUR123-GRE2 was selected using 0.5 μg/ml of aureobasidin. The strain has been deposited to the National BioResource Project of Japan database (http://www.nbrp.jp/report) as BY25955. The strain harboring both pRS426-ADH1 and pAUR123-GRE2 plasmids was constructed by transforming the pAUR123-GRE2 plasmid into BYpADH1 strain. The transformed strains were selected in synthetic minimal medium without uracil in the presence of 0.5 μg/ml of aureobasidin. The strain has been deposited to the National BioResource Project of Japan database as BY25956.

Table 2 Primers used for genes amplification and real-time quantitative RT-PCR analysis
Full size table

Yeast growth assay with glycolaldehyde

Three independent experiments were performed using the respective independent starter cultures to examine the inhibitory effects of glycolaldehyde on the growth of different yeast strains. Yeast cells were grown in 10 ml of synthetic minimal medium containing 2 % (w/v) glucose in the presence or absence of glycolaldehyde in 15 ml glass test tubes. The initial cell concentration was set at 0.1 OD600 (corresponding to 1 × 106 cells/ml), and the growth was monitored by measuring OD600 at different time points.

Fermentation of hot-compressed water-treated cellulose hydrolysate

The hot-compressed water-treated cellulose hydrolysate was prepared by the method described previously using cellulose microcrystalline (Avicel PH-101) (Jayakody et al. 2012). The generated hydrolysate was supplemented with 790 mg/l of complete supplementary media and 0.67 % (w/v) of yeast nitrogen base without amino acids and ammonium sulfate prior to cell inoculation. The pH of the media was adjusted at 5.5 with NaOH solution. Two milliliters of the medium in 3 ml glass test tubes was inoculated with the strains and analyzed in terms of growth and fermentation products.

Analysis of fermentation products

To analyze metabolites produced by fermentation, yeast strains were grown in synthetic minimal medium containing 2 % (w/v) glucose with or without glycolaldehyde. Samples were collected at different time intervals and analyzed in terms of the cell dry weight, as described previously (Jayakody et al. 2011). Concentrations of glucose, ethanol, glycolaldehyde, ethylene glycol, glycerol, and acetic acid were quantitated according to the methods as described previously (Jayakody et al 2012).

Analysis of the glycolaldehyde-dependent NAD(P)H oxidation

Cell lysate was extracted with Y-PER (Thermo Scientific, Rockford, USA) solution by following the manufacturer’s protocol. To extract the cell lysate, cells were grown overnight by shaking at 30 °C in 10 ml of synthetic minimal medium containing 2 % (w/v) glucose. The protein concentration of the extracted lysate was assessed using Bradford reagent (Sigma Aldrich, St. Louis, USA). The glycolaldehyde-dependent NAD(P)H oxidation was measured by adding 2 μg of cell lysate to 1 ml of reaction mixture containing 100 mM potassium phosphate buffer (pH 7.2), 10 mM glycolaldehyde, and 100 μM of either NADH or NADPH. Reaction mixtures were shaken at 25 °C for 3 min, and the OD at 340 nm was assessed to measure the glycolaldehyde-dependent NAD(P)H oxidation (Moon and Liu 2012).

Quantification of intracellular NADH and NADP+ contents

Yeast cells were grown for 24 h at 30 °C in 10 ml of synthetic minimal medium with or without different concentrations of glycolaldehyde. The NADH and NADP+ contents of the extracted cell lysate were analyzed using the AmpliteTM Colorimetric NAD/NADH and NADP/NADPH assay kits “Blue color” (AAT Bioquest Inc, Sunnyvalve, CA, USA), respectively, according to the manufacturer’s protocols. After 1 h incubation of the assay mixture in the dark, the absorbance was monitored at 545 nm, using the ChroMate plate reader (Awareness Technology, Inc.). To detect the NADH content, samples were heated for 30 min at 60 °C to deactivate the NAD+, prior to the assay.

Total RNA extraction and real-time quantitative reverse transcriptase (RT)-PCR

Total RNA was extracted using the hot phenol method (Kohrer and Domdey 1991) after 6 h of incubation of yeast strains with or without different concentrations of glycolaldehyde in 10 ml of synthetic minimal medium containing 2 % (w/v) glucose. The extracted RNA was further purified by using an RNeasy Mini Kit (Qiagen, Valencia, CA, USA) and was frozen at −80 °C until further analysis. Expression of the GRE2, ZNF1, and GND1 genes were analyzed by real-time quantitative RT-PCR using the Roche LightCyclerTM (Roche Applied Biosystem) with the one-step SYBR® Prime scriptTM PLUS RT-PCR KIT (Perfect Real Time) (TaKaRa Biotechnology [Dalian] Co., Ltd), according to the manufacturer’s protocol. The purified 50 ng of template total RNA and the respective primers were used for real-time quantitative RT-PCR analysis (Table 2). The RDN1 gene was used as an internal control for normalization. The RT reaction was carried out at 42 °C for 5 s, followed by heating at 95 °C for 10 s, and the genes were amplified in 40 cycles at 95 °C for 5 s followed by 60 °C for 20 s. The relative gene expressions of ZWF1, GND1, and GRE2 genes were calculated by following the method described by Livak and Schmittgen (2001).

Statistical analysis

All experiments were performed independently in triplicate from the respective independent starter cultures. The results were expressed as mean values and SEM. A one-tailed Student’s t test was adapted for pairwise comparison of the differences between the sample averages of two groups. A one-way analysis of variance (ANOVA) and Tukey’s post hoc honest significance difference test were implemented for multiple comparisons.

Results

ADH1-overexpressing strain has an increased NADH-dependent, but not NADPH-dependent, glycolaldehyde reduction activity

First, in order to investigate the cause of the incomplete recovery of the ADH1-overexpressing strain in the presence of glycolaldehyde, the capacities of cell lysates of strains for the glycolaldehyde-dependent in vitro oxidation of NADH and NADPH were assessed. Results in Fig. 1a show that cell lysates of ADH1-overexpressing strains (BYpADH1) had significantly higher glycolaldehyde-reducing activity using NADH as a cofactor than the control strain (BYp426) (p < 0.01). On the other hand, the glycolaldehyde-reducing activity using NADPH was not significantly different between the two strains (p > 0.05, Fig. 1b). These results indicated that glycolaldehyde reduction in the ADH1-overexpressing strain mainly occurs with NADH rather than NADPH as a cofactor. This result prompted us to hypothesize that the reduced fermentative capacity of the ADH1-overexpressing strain in the presence of a high concentration of glycolaldehyde (≥5 mM) could be restored through activation of a NADPH-dependent reduction system.

Fig. 1
figure 1

Glycolaldehyde-dependent NAD(P)H oxidation activities of cell lysate of engineered strains. Ten millimolar solution of glycolaldehyde was added to the reaction mixture containing 100 mM potassium phosphate buffer (pH 7.2) and 100 μM NADH or NADPH. The reaction was started by adding 2 μg of cell lysate and shaken at 25 °C for 3 min. The optical density at 340 nm was assessed to measure NADH- and NADPH-dependent reduction activities. a NADH oxidation activity. b NADPH oxidation activity. The results are expressed as the mean ± SEM of the independent triplicate experiments from the respective independent starter cultures. Bars labeled with different letters indicate statistically significant values (p < 0.05; one-way ANOVA followed by Tukey’s post hoc honest significance difference test)

Full size image

Pentose phosphate pathway plays an important role in glycolaldehyde tolerance

To further confirm the NADPH-dependent recovery hypothesis, expressions of genes involved in generating intracellular NADPH challenged with different concentrations of glycolaldehyde were investigated. The pentose phosphate pathway is the major intracellular source of NADPH and pentoses, which includes components such as Zwf1, Gnd1, Gnd1, Sol3, and Tkl1. Hence, we analyzed the relative gene expressions of ZWF1 and GND1 by real-time quantitative RT-PCR. Zwf1 encodes glucose-6-phosphate dehydrogenase and Gnd1 encodes 6-phosphogluconate dehydrogenase, both of which consume NADP+ and generate NADPH. As a result, the expressions of these genes were significantly increased in the presence of 5 mM glycolaldehyde than when grown in the absence or in the presence of 2 mM glycolaldehyde (p < 0.05, Fig. 2a, b). This result suggests that NADPH is the key to tolerance to high concentrations of glycolaldehyde (5 mM) and cells are upregulating components involved in NADPH production. To further confirm the NADPH-dependent recovery hypothesis, we analyzed the glycolaldehyde sensitivities of mutants deficient in the pentose phosphate pathway enzymes encoded by the GND1, SOL3, and TKL1 genes. As a result, all of these mutants were significantly more sensitive to glycolaldehyde than their parent strain at 5 mM glycolaldehyde (p < 0.05, Fig. 2c). Altogether, these results are consistent with our hypothesis that activation of the pentose phosphate pathway, which causes NADPH or pentose production, is the key phenomena for tolerance to a high concentration of glycolaldehyde.

Fig. 2
figure 2

Involvement of the pentose phosphate pathway in response to glycolaldehyde. Relative expression levels of a ZWF1 and b GND1 of cells incubated with 2 and 5 mM glycolaldehyde (GA) were analyzed by real-time quantitative RT-PCR. c Glycolaldehyde sensitivity of mutants defective in pentose phosphate pathway. Mutant strains and their parental strain were grown for 24 h at 30 °C in 2 ml of synthetic minimal medium with different concentrations of glycolaldehyde. The results are expressed as mean values ± SEM of independent triplicate experiments from the respective independent starter cultures. Bars labeled with different letters indicate statistically significant values (p < 0.05; one-way ANOVA followed by Tukey’s post hoc honest significance difference test)

Full size image

A GRE2-overexpressing strain shows partially improved glycolaldehyde tolerance

In our former study, acetic acid production was increased by the addition of glycolaldehyde (Jayakody et al. 2012). Together with the above results and the fact that production of acetic acid is catalyzed by Ald6 by using NADP+ in the cytosol as a cofactor (Saint-Prix et al. 2004), it was considered that intracellular NADPH is consumed to attenuate the toxicity of glycolaldehyde. In order to verify the hypothesis to relieve glycolaldehyde toxicity by augmenting NADPH-dependent reduction pathway, we searched for a NADPH-dependent oxidoreductase that could reduce glycolaldehyde. Overexpression of ADH6, which encodes a NADPH-dependent oxidoreductase, has been reported to improve tolerance against furfural and 5-HMF (Almeida et al. 2008; Liu and Slininger 2006; Liu 2011). Therefore, the ADH6-overexpressing strain was constructed and its tolerance to glycolaldehyde was investigated. Although the strain showed tolerance to furfural and 5-HMF as reported earlier (Almeida et al. 2007, 2008; Petersson et al. 2006), it did not show significantly different tolerance to glycolaldehyde relative to the control strain (p > 0.05, Supplementary Fig. S1). Therefore, we shifted to a genome-wide survey of genes that confer tolerance to glycolaldehyde. Our former study of a genome-wide survey of gene deletion mutants sensitive to glycolaldehyde showed that the Δgre2 mutant is one of the most sensitive mutants to glycolaldehyde (Jayakody et al. 2011). This result suggested that Gre2, a NADPH-dependent reductase (Chen et al. 2003), could confer tolerance to glycolaldehyde. Moreover, Gre2 has been reported to reduce aldehyde inhibitors such as fufural, 5-HMF (Liu et al. 2005; Liu and Slininger 2006), and carbonyl compounds (Katz et al. 2003; Rodrigues-Pousada et al. 2010). Therefore, the GRE2-overexpressing strain was constructed and its fermentation profile in the presence of glycolaldehyde was analyzed. The GRE2-overexpressing strain (BYpGRE2) grew significantly faster than the control strain (BYp123) in the presence of glycolaldehyde (p < 0.05, Fig. 3a, Table 3). These results indicate that the GRE2-overexpressing strain has improved growth ability in the presence of glycolaldehyde. In order to understand the physiology of the GRE2-overexpressing strain in detail, we next analyzed the fermentation products produced by the strains in the presence of glycolaldehyde. The results shown in Table 4 revealed that the GRE2-overexpressing strain had slightly higher rates of ethanol production and glucose consumption rate than the control strain, although the differences were not statistically significant. The GRE2-overexpressing strain had a significantly higher rate of acetic acid production (p < 0.01) and a lower rate of glycerol production (p < 0.01) than the control strain (Table 4). Production of ethylene glycol of the GRE2-overexpressing strain was significantly higher than that of the control strain at 48 h (p < 0.05, Fig. 3c). The rates of ethylene glycol production by the GRE2-overexpressing strain and the control strain were 11.5 ± 3.2 and 5.9 ± 1.0 μM/g/h, respectively, which were significantly different (p < 0.05). These results are consistent with the improved growth profiles of the GRE2-overexpressing strain in the presence of glycolaldehyde (Fig. 3a and Table 3). Moreover, the GRE2-overexpressing strain showed higher growth in the actual hot-compressed water-treated cellulose than the control strain (p < 0.05, Fig. 3d). However, the growth parameters of specific cell growth rate and maximum cell growth of the GRE2-overexpressing strain were lower than those of the ADH1-overexpressing strain (Table 3) (Jayakody et al. 2012). These results indicate that the GRE2-overexpressing strain shows an improved growth and tolerance in the presence of glycolaldehyde and hot-compressed water-treated cellulose, although not as effective as the ADH1-overexpressing strain.

Fig. 3
figure 3

GRE2-overexpressing strain shows improved fermentation profiles in glycolaldehyde- and hot-compressed water-treated cellulose-containing media. The growth and glycolaldehyde reduction profiles of the GRE2-overexpressing strain and the control strain were assessed in a medium containing 5 mM glycolaldehyde (GA) (a–c). a Growth profile. b Glycolaldehyde concentration. c Ethylene glycol concentration. d Growth of yeast cells in cellulose hydrolysate treated with hot-compressed water. Empty triangles, BYp123 (nontreated); empty squares, BYpGRE2 (nontreated); filled triangles, BYp123 (glycolaldehyde-treated); and filled squares, BYpGRE2 (glycolaldehyde-treated). The results are expressed as mean values ± SEM of independent triplicate experiments from the respective independent starter cultures. Statistically significant levels of the difference of two strains are indicated (*p < 0.05, **p < 0.01)

Full size image
Table 3 Specific growth rate of strains in the presence of 5 mM glycolaldehyde
Full size table
Table 4 Production rates of metabolites per biomass of strains in the presence of 5 mM glycolaldehyde
Full size table

Simultaneous overexpression of ADH1 and GRE2 confers hyper-resistance to glycolaldehyde

Based on the above results, we speculated that overexpression of both ADH1 and GRE2 might substantially enhance the ability to eliminate glycolaldehyde toxicity of yeast cells. In order to verify this hypothesis, we developed a strain that overexpresses ADH1 and GRE2 under their own promoters (BYpADH1pGRE2). Analysis of growth in the presence of glycolaldehyde revealed that the developed strain exhibits improved resistance to glycolaldehyde relative to the ADH1-overexpressing strain (BYpADH1p123), especially after 36 h (p < 0.05, Fig. 4a). The strain overexpressing both ADH1 and GRE2 showed significantly higher rates of ethanol production (p < 0.05, Fig. 4b), glucose consumption (p < 0.05, Fig. 4c), acetic acid production (Fig. 4d), and ethylene glycol production (p < 0.05, Fig. 4g) than the ADH1-overexpressing strain. It should be noted that especially at the period of vigorous growth (12 to 24 h), the concentration of glycolaldehyde of the strain overexpressing both ADH1 and GRE2 was significantly lower than that of the ADH1-overexpressing strain (p < 0.05, Fig. 4f). Glycerol production was not significantly different between the two strains (p > 0.05, Fig. 4e). The rates of glycolaldehyde conversion of the strain overexpressing both ADH1 and GRE2 and the control strain were 101.0 ± 11.2 and 90.3 ± 15.0 μM/g/h, respectively, which were significantly different (p < 0.05). The conversion rates of glycolaldehyde into ethylene glycol were 75.5 ± 2.5 and 69.4 ± 5.7 % for the strain that overexpressed both ADH1 and GRE2 and the control strain, respectively.

Fig. 4
figure 4

The strain that overexpresses both ADH1 and GRE2 shows hyper-resistance and enhanced fermentation profile in the presence of glycolaldehyde. The growth and fermentation profiles of the strains BYpADH1p123 and BYpADH1pGRE2 with or without glycolaldehyde (GA). a Growth profile of the strains at 10 mM glycolaldehyde. b–g Fermentation profiles of the strains at 5 mM glycolaldehyde. b Ethanol concentration. c Glucose concentration. d Acetic acid concentration. e Glycerol concentration. f Glycolaldehyde concentration. g Ethylene glycol concentration. Empty triangles, BYpADH1p123 (nontreated); empty squares, BYpADH1pGRE2 (nontreated); filled triangles, BYpADH1p123 (glycolaldehyde-treated); and filled squares, BYpADH1pGRE2 (glycolaldehyde-treated). The results are expressed as mean values ± SEM of independent triplicate experiments from the respective independent starter cultures. Statistically significant levels of the difference of two strains are indicated (*p < 0.05, **p < 0.01)

Full size image

To better understand the mechanism that enables glycolaldehyde conversion in these strains, we analyzed the glycolaldehyde-dependent NADH and NADPH oxidation activities of the cell lysates extracted from the strain overexpressing both ADH1 and GRE2 and the control strain. The results indicated that the strain overexpressing both ADH1 and GRE2 has significantly higher conversion activities of glycolaldehyde using NADH and NADPH as cofactors than those of the control strain (p < 0.05, Fig. 1a, b). These results indicate that simultaneous overexpression of ADH1 and GRE2 confers an increased NADH- and NADPH-dependent reduction activity of glycolaldehyde.

Decrease of NADH and increase of NADP+ in the strain overexpressing both ADH1 and GRE2

In order to obtain insight into the redox cofactor preference of the glycolaldehyde-reducing activity of these strains, we analyzed the intracellular redox profiles of the cells by monitoring the intracellular NADH and NADP+ levels. In all of the strains studied, NADH decreased in proportion to the concentration of glycolaldehyde (Fig. 5a). The intracellular level of NADH decreased in the ADH1-overexpressing strain treated with 5 mM glycolaldehyde relative to the control strain (p < 0.05, Fig. 5a), which is consistent with the Adh1-catalyzed oxidation of glycolaldehyde using NADH as a cofactor (Fig. 1a). The intracellular level of NADH further decreased in the strain overexpressing both ADH1 and GRE2 treated with 5 mM glycolaldehyde (p < 0.05, Fig. 5a), suggesting a role of Gre2 in NADH-mediated glycolaldehyde reduction. NADP+ was increased in the GRE2-overexpressing strain and the strain overexpressing both ADH1 and GRE2 treated with 5 mM glycolaldehyde relative to the control strain (p < 0.05, Fig. 5b), suggesting Gre2-catalyzed glycolaldehyde reduction using NADPH as a cofactor at 5 mM glycolaldehyde. In contrast, at 2 mM glycolaldehyde, NADP+ was not significantly increased in the GRE2-overexpressing strain and the strain overexpressing both ADH1 and GRE2 relative to the control strains (p > 0.05, Fig. 5b), suggesting that glycolaldehyde reduction at 2 mM glycolaldehyde does not predominantly occur by using NADPH as a cofactor.

Fig. 5
figure 5

Intracellular NADH and NADP+ levels in ADH1- and GRE2-overexpressing strains and strain overexpressing both ADH1 and GRE2 treated with or without glycolaldehyde. The intracellular NADH and NADP+ concentrations of extracted cell lysate of yeast cells grown at 30 °C for 24 h in 10 ml of synthetic minimal media with or without different concentrations of glycolaldehyde were analyzed. a Intracellular NADH content. b Intracellular NADP+ content. The results are expressed as mean values ± SEM of independent triplicate experiments from the respective independent starter cultures. Bars labeled with different letters indicate statistically significant values (p < 0.05; one-way ANOVA followed by Tukey’s post hoc honest significance difference test)

Full size image

Together, these results indicated that at a lower concentration (2 mM) of glycolaldehyde, most of the conversion of glycolaldehyde was mediated by using NADH as a cofactor. In contrast, at 5 mM glycolaldehyde, the relative contribution of NADPH in the glycolaldehyde-reducing system significantly increased relative to 2 mM in all strains (p < 0.05, Fig. 5b). These results indicate that the conversion of glycolaldehyde to ethylene glycol occurs mainly using NADH at 2 mM glycolaldehyde and mainly using NADPH at 5 mM glycolaldehyde.

The levels of expressions of GRE2 of yeast cells challenged with different concentrations of glycolaldehyde were investigated by using real-time quantitative RT-PCR (Fig. 6). As a result, the expression of GRE2 significantly increased in the presence of 5 mM glycolaldehyde relative to when grown in the absence or in the presence of 2 mM glycolaldehyde (p < 0.05, Fig. 6). As we expected, the strain overexpressing GRE2 under its own promoter also increased the expression of GRE2 in response to glycolaldehyde (p < 0.05, Fig. 6). Together, these results strongly indicate that both GRE2 expression and the NADPH-dependent glycolaldehyde reduction by Gre2 are increased at 5 mM glycolaldehyde.

Fig. 6
figure 6

Expression of GRE2 with different concentrations of glycolaldehyde. Relative expression levels of GRE2 incubated with 2 and 5 mM glycolaldehyde (GA) were analyzed by real-time quantitative RT-PCR. The results are expressed as mean values ± SEM of independent triplicate experiments from the respective independent starter cultures. Bars labeled with different letters indicate statistically significant values (p < 0.05; one-way ANOVA followed by Tukey’s post hoc honest significance difference test)

Full size image

Discussion

In this study, we have improved the glycolaldehyde tolerance of the ADH1-overexpressing strain, which we developed in an earlier study (Jayakody et al. 2012), by engineering redox cofactor utilization. This is the first report to confer tolerance to glycolaldehyde by introducing a NADPH-dependent reducing mechanism. Moreover, although it has been reported that the reducing activities of furfural and 5-HMF are dependent on NADH and NADPH (Heer et al. 2009), this is the first report that artificially augmented both NADH- and NADPH-dependent reducing mechanisms to confer tolerance to glycolaldehyde.

Expression of GRE2 is reported to be upregulated under multiple stress conditions, such as osmotic stress, heat shock, oxidative stress, and heavy metal stress (Krantz et al. 2004; Liu and Slininger 2006; Liu and Moon 2009; Liu 2011; Rep et al. 2001). Furthermore, Moon and Liu (2012) have engineered Gre2 to confer tolerance to furfural and 5-HMF using NADH as a cofactor. The metabolic role of glycolaldehyde reduction by Gre2 elucidated in this study will provide insights into the knowledge of aldehyde reductase activities obtained in the former studies on Gre2.

A shift in the cofactor preferences of yeast cells for the reduction of aldehydes according to their concentrations has been reported in several other aldehyde-reducing reactions, including those involving the conversions of furfural and 5-HMF (Almeida et al. 2008; Heer et al. 2009; Liu 2011). The reduction of furfural into less toxic furan methanol at a concentration less than 6 mM in yeast cells involves the use of NADH as cofactor (Horvath et al. 2001). In contrast, at a higher concentration (>15 mM) of furfural, the cofactor preference of the furfural-reducing reaction shifts to NADPH, owing to an insufficient supply of NADH (Heer et al. 2009). Celton et al. (2012) have demonstrated that a similar shift in cofactor preference occurs in the acetoin-reducing reaction. However, the shift in cofactor preference occurred in our study at a lower concentration (2 to 5 mM) in the case of glycolaldehyde than the other aldehydes, possibly as a result of its strong toxicity (Jayakody et al. 2011).

Given that S. cerevisiae lacks pyridine nucleotide transhydrogenase-like activities (Bruinenberg et al. 1983), the excess NADH generated under anaerobic conditions is primarily balanced by the formation of glycerol, which is a major by-product of yeast fermentation (Zaldivar et al. 2001). Accordingly, a decrease in the glycerol concentration reflects a shortage of NADH within cells. At the same time, it has been reported that reducing the rate of glycerol production increases the ethanol yield (Nissen et al. 2000). The low glycerol production has been reported to be achieved by ablation of the GDH1 gene and overexpression of the GDH2 gene (Roca et al. 2003) and recombination of the phosphoketolase pathway (Sonderegger et al. 2004). The results reported in this study indicate that in the presence of glycolaldehyde, the strain overexpressing both ADH1 and GRE2 predominantly oxidizes surplus cytosolic NADH and generates NAD+, which substantially reduces glycerol formation and increases ethanol formation. Thus, the strain overexpressing both ADH1 and GRE2 decreased the glycerol yield by 70 % and increased the ethanol yield by around 7 % relative to the ADH1-overexpressing strain. This is also similar to the decrease of glycerol by overexpression of a water-forming NADH oxidase (Vemuri et al. 2007).

It has been reported that the increase in acetic acid is interconnected to the NADPH demand of the cells (Grabowska and Chelstowska 2003). A surplus of NADP+ increases acetic acid production (Van Dijken and Scheffers 1986). NADPH is provided mainly through the pentose phosphate pathway (Verduyn et al. 1990). In glycolaldehyde-treated cells, increases in the levels of NADPH are considered to be fueled by activation of the pentose phosphate pathway and to have been provided to the glycolaldehyde-reducing reaction using Gre2. This hypothesis is substantiated by the increased expression of ZWF1 and GND1, encoding components of the pentose phosphate pathway, at a high concentration of glycolaldehyde (≥5 mM). Moreover, the glycerol–dihydroxyacetone cycle, which has been reported to be activated by elevated levels of NADP+, converts NADH into NADPH (Celton et al. 2012; Costenoble et al. 2000). The lower glycerol content in the GRE2-overexpressing strain in glycolaldehyde-reducing condition (Table 4) suggests that the glycerol–dihydroxyacetone cycle might be activated in the GRE2-overexpressing strains. Many of the other genes have been reported to be involved in the reduction and detoxification of aldehyde inhibitors using NAD(P)H as a cofactor. Genes encoding aldehyde dehydrogenases ADH1, ADH6, ADH7, and SFA1; aldehyde reductases ALD4, ALD6, and ARI1; methylglyoxal reductases GRE2 and GRE3; and uncharacterized ORF YKL071W have been identified as genes encoding furfural and 5-HMF-reducing enzymes (Heer et al. 2009; Liu 2011). Therefore, the genes other than the ADH1 and GRE2 are candidate genes of overexpression to improve tolerance to glycolaldehyde, except ADH6, whose overexpression did not confer tolerance to glycolaldehyde in the present study (Supplementary Fig. S1).

In conclusion, for the first time, to the best of our knowledge, we have engineered a yeast strain that is hyper-resistant to glycolaldehyde by introducing a NADPH-dependent reduction pathway in addition to the NADH-dependent pathway. The strain overexpressing both GRE2 and ADH1 converted glycolaldehyde to ethylene glycol and detoxified it at 5 mM glycolaldehyde dependent on both NADH and NADPH. The redox cofactor utilization strategy for glycolaldehyde detoxification proposed in this study has the potential to influence strategies to improve tolerance of other fermentation inhibitors.