PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of aemPermissionsJournals.ASM.orgJournalAEM ArticleJournal InfoAuthorsReviewers
 
Appl Environ Microbiol. 2012 December; 78(23): 8340–8352.
PMCID: PMC3497381

Oxygen Response of the Wine Yeast Saccharomyces cerevisiae EC1118 Grown under Carbon-Sufficient, Nitrogen-Limited Enological Conditions

Abstract

Discrete additions of oxygen play a critical role in alcoholic fermentation. However, few studies have quantitated the fate of dissolved oxygen and its impact on wine yeast cell physiology under enological conditions. We simulated the range of dissolved oxygen concentrations that occur after a pump-over during the winemaking process by sparging nitrogen-limited continuous cultures with oxygen-nitrogen gaseous mixtures. When the dissolved oxygen concentration increased from 1.2 to 2.7 μM, yeast cells changed from a fully fermentative to a mixed respirofermentative metabolism. This transition is characterized by a switch in the operation of the tricarboxylic acid cycle (TCA) and an activation of NADH shuttling from the cytosol to mitochondria. Nevertheless, fermentative ethanol production remained the major cytosolic NADH sink under all oxygen conditions, suggesting that the limitation of mitochondrial NADH reoxidation is the major cause of the Crabtree effect. This is reinforced by the induction of several key respiratory genes by oxygen, despite the high sugar concentration, indicating that oxygen overrides glucose repression. Genes associated with other processes, such as proline uptake, cell wall remodeling, and oxidative stress, were also significantly affected by oxygen. The results of this study indicate that respiration is responsible for a substantial part of the oxygen response in yeast cells during alcoholic fermentation. This information will facilitate the development of temporal oxygen addition strategies to optimize yeast performance in industrial fermentations.

INTRODUCTION

Oxygen is discretely added during winemaking to avoid sluggish and stuck fermentations (55). This is generally achieved through pump-overs, where dissolved oxygen concentrations could transiently reach up to 100 μM afterwards. The oxygen is consumed completely in approximately 1 h but depends on the wine variety and the oxygen addition method employed (M. I. Moenne, P. Saa, J. R. Perez-Correa, F. Laurie, and E. Agosin, unpublished data). The wide range of resulting dissolved oxygen concentrations has a deep impact on the physiology of wine yeast cells, improving the yeast fermentative rate as well as yeast viability (55). These effects are partially explained because oxygen is required for sterol biosynthesis, proline uptake, and unsaturated lipid biosynthesis (29, 52, 62). Previous studies attempted to assess the fate of oxygen under winemaking culture conditions, focusing on the behavior of lipid and sterol biosynthesis (52, 53). However, a global approach is required for a quantitative understanding of the impact of dissolved oxygen on the physiology of yeast cells grown under enological conditions, i.e., sugar-sufficient, nitrogen-limited, acidic cultures.

The presence of oxygen in the culture medium can shift the central yeast metabolism from fermentative to respiratory (30, 32, 50, 60). However, Saccharomyces cerevisiae shows a fully respiratory metabolism only at low growth rates with low sugar concentrations (20). Above a critical specific growth rate or sugar concentration, yeast cells synthesize ethanol, regardless of the oxygen level (60). The latter is known as the Crabtree effect (66). Several hypotheses have been suggested to explain this phenomenon, such as the catabolic repression of respiratory enzymes (21), a bottleneck of the carbon flux toward the tricarboxylic acid (TCA) cycle (47), and a limitation of yeast respiratory enzymes to reoxidize the NADH produced by glycolysis (65). In all cases, carbon is channeled toward ethanol biosynthesis in the cytosol, instead of the TCA cycle, causing “overflow metabolism,” the foundation of alcoholic fermentation (39). This overflow normally occurs under nitrogen-limited and carbon-sufficient culture conditions, e.g., the operation conditions of winemaking (60). However, to the best of our knowledge, it is not clear how the specific oxygen uptake rate (OUR) influences overflow metabolism under nitrogen-limited conditions and whether it is compatible or not with respiratory activity.

An understanding of overflow metabolism requires an understanding of the redox biochemistry of alcoholic fermentation in yeast. Under anaerobic conditions, the NADH spent in the glycolytic pathway must be reoxidized. Ethanol production is a way to reoxidize this NADH, although there is always an electron surplus, caused by the release of fully oxidized carbon as CO2. Redox balance is restored by diverting carbon toward glycerol production, which also reoxidizes NADH (64). This is the only way to achieve redox balance under anaerobic conditions. On the contrary, under aerobic conditions, oxygen can act as an electron sink and help to establish this balance. For this, two requisites are needed under enological conditions: respiration has to be active under high-sugar, nitrogen-limited conditions, and there must be functional shuttles of NADH from the cytosol to the mitochondria.

Three redox shuttles have been proposed to be at work in yeast mitochondria: the glycerol-3-phosphate (Gut2p) shuttle, the external NADH dehydrogenase (Nde1p) shuttle, and the mitochondrial alcohol dehydrogenase (Adh3p) shuttle (5, 49). The Adh3p reaction consumes the ethanol that diffuses from the cytoplasm and produces acetaldehyde and NADH in the mitochondrial matrix. Acetaldehyde can diffuse back into the cytoplasm, while NADH is reoxidized by oxygen in the electron transport chain. The ethanol-acetaldehyde conversion is potentially reversible, and its ΔG value is low (33.5 kJ/mol at pH 5 at 25°C) (3). The Nde1p reaction, on the contrary, directly couples cytosolic NADH oxidization to the respiratory chain. The latter is similar to Gut2p, which shuttles electrons from glycerol-3-phosphate oxidization to the mitochondrial quinone pool. Adh3p has been regarded as the main shuttle (12, 38). However, there has been only one study on the function of these shuttles under nitrogen-limited conditions (41), emphasizing that this issue deserves further investigation.

To tackle such issues, enological conditions need to be simulated on a laboratory scale. Continuous cultures, unlike batch cultures, allow a tightly controlled environment while keeping yeast cells at a defined specific growth rate. Thus, we can simulate culture conditions prevailing at the end of the exponential phase of wine fermentation, when assimilable nitrogen is running out (i.e., nitrogen limited) and the effect of the addition of oxygen is maximal (55). Chemostats, as they operate in the steady state, are highly reproducible, making them ideal for “-omics” experiments (28, 60).

The simulation of different oxygen conditions in a winemaking setting requires the characterization of the relationship between OURs and the dissolved oxygen concentrations under these culture conditions. The specific oxygen consumption rate by yeast cells increases linearly with oxygen availability up to a critical dissolved oxygen concentration, at which the OUR is maximal, i.e., the “critical” OUR (16). When an oxygen impulse is carried out during fermentation, yeast cells are subjected to oxygen contents within the oxygen-limiting regime; the whole oxygen concentration range achieved during an oxygen impulse under enological conditions could be simulated by increasing the dissolved oxygen concentrations in nitrogen-limited continuous cultures.

Once the conditions are set, the OUR and other rates measured at the steady state are valuable inputs for metabolic flux analysis (MFA). MFA allows the determination of intracellular metabolic fluxes, difficult to determine in vivo, from a set of extracellular uptake and production rates, which are easily measurable. MFA has been successfully used to understand yeast metabolism under anaerobic conditions in both carbon- and nitrogen-limited settings (40, 63). The incorporation of oxygen consumption pathways into these networks is a task that could help to explain the metabolic effects and the fate of oxygen in yeast industrial fermentations.

In this research, we combined the use of carbon-sufficient, nitrogen-limited continuous cultures; MFA models; and DNA microarrays to simulate and examine the response of S. cerevisiae to increasing dissolved oxygen concentrations under enological conditions. MFA analysis indicated that yeast cells underwent respirofermentative metabolism under wine fermentation conditions above a certain dissolved oxygen threshold. Transcriptome analyses supported these findings, as they showed that respiratory genes were induced by oxygen. Other enologically relevant processes, such as proline uptake and cell wall remodeling, were also affected by oxygen. Altogether, this research provides an understanding of the mechanisms through which oxygen can influence yeast performance in industrial fermentations.

MATERIALS AND METHODS

Yeast strain and culture conditions.

Saccharomyces cerevisiae EC1118 (Lalvin, Switzerland), an industrial strain used worldwide by the wine industry, was employed throughout this study. Initial seed cultures were grown in 50 ml yeast extract-peptone-dextrose (YPD) broth at 28°C under aerobic conditions to a mid-logarithmic growth phase. For continuous cultivation, a 2.0-liter Biostat B bioreactor (Sartorius Biotech, Germany), with a working volume of 1.5 liters, was inoculated with enough of the microbial broth to obtain an initial cell density of 106 cells ml−1. The culture was allowed to grow in the batch mode until it reached the early to mid-exponential growth phase. Constant feeding was then initiated with a defined artificial must that limited growth by nitrogen, with a constant residual carbon concentration (see below). The dilution rate (D) was set at 0.1 h−1. Agitation, temperature, and pH were maintained at 200 rpm, 20°C, and 3.5, respectively, to simulate white wine fermentation. All experiments were performed in triplicate.

Gases were provided by means of a GFC mass flow controller (Aalborg), at a rate of 0.25 liter min−1 in all experiments, except under conditions with the highest dissolved oxygen concentration (21 μM) (detailed below), where the gas flow was provided at a rate of 1 liter min−1. Polyurethane tubing and butyl rubber septa were used to minimize oxygen diffusion into the anaerobic cultures. The inlet gas entered through an in-line 0.2-μm-pore size filter to maintain sterility. Gaseous carbon dioxide and oxygen concentrations were measured online with Gascard NG (Edinburgh Instruments, United Kingdom) and Parox 1000 (Messtechnick Engineering, Switzerland) gas analyzers, respectively. Afterwards, a condenser connected to a cryostat set at 2°C cooled the off-gas. All the data were acquired on a Simatic PCS7 distributed control system with a monitoring station (Siemens, Germany).

Dissolved oxygen levels.

To achieve different levels of dissolved oxygen, we sparged the cultures with different oxygen-nitrogen mixtures. For anaerobic experiments, ultrapure nitrogen (certified 99.999% N2; Indura, Chile) was passed through an HPIOT3-2 oxygen trap (Agilent) to reduce residual oxygen levels to below 15 ppb. Other oxygen levels were achieved by the sparging of 1%, 5%, and 21% oxygen-nitrogen mixtures (Indura, Chile) directly into the culture, reaching concentrations of 1.2 μM, 2.7 μM, and 5.0 μM dissolved oxygen at the steady state, respectively (Table 1). For the last oxygen concentration, the gaseous flow with the 21% oxygen mixture was increased from 0.25 to 1 liter min−1. At this point, stirring was increased from 200 to 400 rpm, reaching a concentration of 21 μM dissolved oxygen at the steady state. Experiments at all dissolved oxygen concentrations were performed in triplicate. In all cases, the steady state was reached within 5 residence times (60 h) and kept for at least 1.5 more residence times. The dissolved oxygen concentration was measured online with an InPro model 6950 probe (Mettler, Toledo, OH). This probe has a detection limit of 0.03 μM.

Table 1
Steady-state metabolite concentrations and metabolite consumption and production rates in S. cerevisiae EC1118 grown under different dissolved oxygen culture conditions

Culture medium.

A defined medium simulating standard white grape juice was used in the bioreactor fermentations. The residual glucose concentration in the chemostat outlet was targeted to 40 g liter−1 (Table 1) to sustain glucose repression at the same level in all experiments (45, 60). Briefly, the feed glucose concentrations were 80 g liter−1 for the cells grown anaerobically and with 1.2 μM dissolved oxygen and 85 g liter−1 and 95 g liter−1 for cells grown at 2.7 μM and with 5 μM dissolved oxygen, respectively. Fructose was not included in the formulation. The compositions of the other nutrients were based on MS300 medium, as described previously by Salmon and Barre (56), but was modified to run continuous cultures limited in nitrogen, since MS300 medium was designed for batch fermentations, and its use in chemostats resulted in 100 mg liter−1 of residual nitrogen (data not shown). The modified medium, called cMS300 medium, was different from MS300 medium in its phosphate, sulfate, and biotin contents. cMS300 medium contained 3 g of KH2PO4, 5.75 g of K2SO4, 500 mg of MgSO4 · 7H2O, and 0.05 mg of biotin per liter, according to the mineral medium described previously by Pizarro et al. (45). Other vitamins and micronutrients were the same as those in MS300 medium. The yeast assimilable nitrogen (YAN) concentration corresponded to 380 mg N liter−1 under all culture conditions. However, under anaerobic conditions, the YAN content corresponded to 300 mg N liter−1, since proline is not assimilable by yeast under anaerobic growth conditions (29). The nitrogen sources employed were 20.5% (wt/wt) ammoniacal nitrogen (NH4Cl), 16.9% (wt/wt) l-proline, 1.25% (wt/wt) l-glutamine, 6% (wt/wt) l-arginine, 4.9% (wt/wt) l-tryptophan, 4% (wt/wt) l-alanine, 2.6% (wt/wt) l-glutamic acid, 2.6% (wt/wt) l-serine, 1.6% (wt/wt) l-threonine, 1.5% (wt/wt) l-leucine, 1.5% (wt/wt) l-aspartic acid, 1.3% (wt/wt) l-valine,1.1% (wt/wt) l-phenylalanine, 1.1% (wt/wt) l-isoleucine, 1.1% (wt/wt) l-histidine, 0.6% (wt/wt) l-methionine, 0.6% (wt/wt) l-tyrosine, 0.6% (wt/wt) l-glycine, and 0.4% (wt/wt) l-lysine and l-cysteine (63).

All culture media were supplemented with 10 mg of ergosterol and 30 mg of oleic acid, in the form of Tween 80. The final concentrations of these anaerobic factors were lower than those in the original MS300 formulation (15 mg liter−1 ergosterol and 90 mg liter−1 oleic acid) for all experiments performed, to avoid the masking effect of oxygen on the synthesis of these compounds. We determined the biomass yields of both ergosterol and oleic acid in S. cerevisiae strain EC1118 in chemostats, as described previously (37). The ergosterol and oleic acid concentrations in the final culture medium were set 3-fold above their biomass yields to ensure an excess but were still lower than the concentrations in the original anaerobic factor formulation. We confirmed ergosterol and oleic acid excesses by detecting these anaerobic factors in the supernatants of all cultures at the steady state; i.e., continuous cultures were limited only in nitrogen.

Metabolite sampling and analysis.

For each of the three biological replicates corresponding to the dissolved oxygen levels evaluated, steady-state culture samples were taken after at least six residence times by using an ice-cold sterile 50-ml plastic syringe plugged into the sampling device of the bioreactor. The samples were rapidly transferred into an ice-cold 50-ml sterile plastic tube, where either the cell dry weight (DW) of the culture was determined (see below) or the cultures were transferred into Eppendorf tubes. The latter cultures were centrifuged at 10,000 × g for 3 min, and 1-ml supernatant aliquots were stored at −80°C until further analysis. Biomass was determined on a dry weight basis by filtering 10-ml culture samples through preweighed 0.2-μm-pore-size acetate ester filters (Whatman). The filter was washed twice with 10 ml milliQ water and dried in an infrared drier-equipped balance (Precisa, Switzerland) to a constant weight at 65°C.

Extracellular metabolites were determined by high-performance liquid chromatography (HPLC). Twenty microliters of the supernatant samples was injected into a LaChrom L-7000 HPLC system (Hitachi, Japan). Organic acids, alcohols, and sugars were separated by using an Aminex HPX-87H anion-exchange column (Bio-Rad), with 5 mM H2SO4 as the mobile phase. Organic acids were detected by using a LaChrom L-7450A diode array detector (Hitachi, Japan) at 210 nm. Sugars and alcohols were detected by a LaChrom L-7490 refraction index detector (Hitachi, Japan). Each compound was quantified by using a calibration curve made with known concentration standards.

The residual assimilable nitrogen in the cultures was monitored by the σ-phthaldehyde–N-acetyl-l-cysteine spectrophotometric assay (NOPA) procedure, as described previously (63).

Proline was detected by HPLC, as described previously (36). Briefly, amino acids in the supernatant samples were derivatized with 6-aminoquinoleyl-N-hydroxysuccinimidyl carbamate. The amino acids were then separated in a AccQ Tag column (Waters). Once separated, the derivatized amino acids were detected by fluorescence and quantified by using known standards.

Sampling and RNA isolation.

For RNA isolation, samples from each one of the three biological replicates under the five oxygen conditions were extracted by rapidly placing 20 ml of culture into a Falcon tube filled with 35 to 40 ml of crushed ice (45). From this tube, 1 ml of sample was transferred into a 1.5-ml Eppendorf tube; after centrifugation at 10,000 × g for 3 min, the supernatant was removed, and the pellet was directly frozen in liquid nitrogen and stored at −80°C until RNA isolation.

Before RNA isolation, all water-based reagents were treated with diethyl pyrocarbonate (DEPC). RNA was extracted from frozen cell pellets by using the AxyPrep Multisource RNA kit (Axygen Biosciences), modified for use with glass beads for cell lysis. Briefly, we resuspended the cell pellet in the R-I kit buffer and added it to a screw-cap tube containing 250 μl of acid-washed glass beads (500-μm diameter) (Sigma). We stirred the tubes for three cycles of 45 s in a Mini Bead Beater (Biospec), with the tubes standing for an equivalent time in ice between cycles. Later, the lysate was centrifuged at 10,000 × g for 1 min, and the supernatant was transferred into a tube with 250 μl of isopropanol. After this point, we followed the instructions provided by the kit manufacturer. RNA was checked for integrity in a 1.3% agarose gel, prepared in a buffer with 20 mM morpholinepropanesulfonic acid (MOPS), 5 mM sodium acetate, and 1 mM EDTA. After heating, we added 2% (vol/vol) formaldehyde. This was also used as the electrophoresis running buffer. RNA was quantified with a NanoDrop instrument (Thermo Scientific), and samples with absorbance ratios at 260 nm/280 nm and 260 nm/230 nm of >1.8 were used for further processing.

Microarray analysis.

We used the Yeast Genome 2.0 chip from Affymetrix. Triplicate arrays for each one of the five oxygen conditions were performed. RNA for hybridization was prepared according to the manufacturer's instructions (2). Hybridization, staining, and scanning onto the microarrays were performed according to the manufacturer's instructions (1). Gene expression data were imported from the CEL files into R, and quality was assessed by using the tools in the simpleaffy package from Bioconductor (22). After the quality check step, data were normalized in R by using robust multiarray analysis (RMA) from the affy package from Bioconductor (22). Differential expression analyses between data under the different conditions were carried out by using the rank products algorithm (8), with a cutoff false discovery rate of 0.1. We also performed hierarchical clustering on the data, using group average and Euclidian distance as a distance measure. All calculations were performed by using R statistical software (48a). The resulting lists of genes from the clusters or differential expression analyses were then submitted to AmiGO term enrichment analysis (11), to find out which Gene Ontology (GO) annotations and functions were overrepresented among the genes regulated by oxygen levels. The full gene annotation was then extracted from the Saccharomyces Genome Database (SGD) (13). Alternatively, the GO enrichment analysis was performed by using Blast2GO software (14). This software has the Gossip package integrated for statistical assessment, which employs Fisher's exact test and corrects for multiple testing to find enriched GO terms between two sets of sequences (6). The complete set of annotated genes for Saccharomyces cerevisiae from the SGD was used as the reference control group, with a false-positive discovery rate value of 0.05.

Clustering analysis.

A hierarchical clustering of gene expression levels was performed by using Euclidean distance to measure differences between expression vectors and the group average clustering algorithm. This clustering method was selected because it exhibited the highest cophenetic correlation coefficient (CCC) (CCC = 0.61) compared to single linkage (CCC = 0.41), complete linkage (CCC = 0.58), and Ward's method (CCC = 0.55). The CCC values were calculated as previously described (58). A distance cutoff value of 2.26 was selected to make the partition of the clustered data into 56 groups with different gene expression patterns under the tested conditions. The specific clustering algorithm and distance cutoff value were defined after a partition-based cluster structure consistency analysis using the Silhouette measure (43). Briefly, all possible distance cutoff values to split the clustered data into k different groups were tested, and Silhouette values for each partition were calculated. According to this analysis, a maximum average value of 0.304 for the Silhouette measure was obtained when using the group average clustering algorithm and a distance cutoff value of 2.26, thus producing 56 independent gene groups. The resulting partition was then visually inspected for consistency in the dendrogram with the associated heat maps of the clustered genes in each group. The dendrogram with the heat maps for each gene was built by using the iTOL Web server (http://itol.embl.de/) (33). A color-coded strip illustrating the independent gene clusters was added to the dendrogram to facilitate the mapping and analysis of individual genes.

Metabolic flux analysis.

We built a stoichiometric matrix based on a model developed previously by Varela et al. (63), which includes glycolysis, TCA cycle, fermentative, and anaplerotic reactions as well as essential anabolic pathways. We added the following oxygen-related reactions to this matrix: proline consumption, NADH- and reduced flavin adenine dinucleotide (FADH2)-dependent respiration pathways (each one lumped into one reaction), and ergosterol and unsaturated fatty acid synthesis. These reactions were extracted from the YeastCyc database (http://pathway.yeastgenome.org/). We used an approach described previously by Nissen et al. (40) for the biomass equation, leaving every biomass component (protein, DNA, RNA, carbohydrate, and lipids) as individual products. Moreover, we separated the lipids into sterols and unsaturated and saturated fatty acids, in order to have a better description of the oxygen-dependent pathways. The resulting matrix had 47 reactions and 44 metabolites and is detailed in the supplemental material. The condition number of the matrix was 44, indicating that the model is numerically robust (39).

Flux estimation.

Since the model has three degrees of freedom, the system becomes overdetermined if more than three extracellular rates are measured. We evaluated inputs, including all of the possible combinations of rates (from 3 to 15), selected from the following set of experimentally measured extracellular rates of substrate uptake or metabolic product production: glucose, ethanol, glycerol, succinate, acetate, CO2, oxygen, proline, biomass components (carbohydrates, DNA, RNA, and proteins), and the three classes of lipids mentioned above. The procedures for the measurement of biomass components are stated in the supplemental material. Since the resulting matrix is not square, we carried out a “pseudoinverse” operation (59) to solve the mass balance equation and estimate the flux vector, including the unselected rates. An evaluation of the rate combinations indicated that the more rates selected for input, the more accurate the estimations of the model. This prompted us to use 13 extracellular rates as an input, leaving ethanol and CO2 as calculated rates. This was validated by sensitivity analyses (see below). The redistribution of intracellular fluxes under different oxygen conditions was assessed by determining the flux vector for the three individual replicates under each condition and comparing the resulting fluxes to each other by using a t test.

Consistency and sensitivity analysis.

Data consistency was checked by using a method described previously by Wang and Stephanopoulos (67). After confirming the absence of gross measurement errors, we performed a custom sensitivity analysis, based on the normalized error distribution, which was calculated as the differences between the estimated fluxes and the fluxes calculated by using modified rates (specific rates ± experimental errors). We then calculated the square sum of these differences and took their square root to obtain the error value. Finally, we normalized the errors by the value of the corresponding flux calculated with the unmodified rates. The specific rates were modified by the experimental error one at a time, and the sum of the errors on all the fluxes of the stoichiometric matrix was calculated. We found that ethanol and CO2 fluxes were the rates whose errors contributed the most to the overall flux error of the model. Therefore, we used 13 rates as the input and estimated the specific rates of ethanol and CO2 fluxes to validate the model. These estimations had less than an 11% error (see Table S2 in the supplemental material). All the preceding calculations were carried out by using custom scripts in the R computer language.

Microarray data accession number.

The raw microarray data were submitted to the Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/), where they are available under accession number GSE34964.

RESULTS

Simulation of dissolved oxygen concentrations and oxygen uptake rates under enological conditions.

To characterize wine yeast physiology at the different dissolved oxygen concentrations found after oxygen impulses in winemaking, we built a specific OUR-versus-dissolved-oxygen curve (Fig. 1). The former showed a constant increase with concentrations of dissolved oxygen up to 21 μM. From this critical value, the OUR was constant at 3.9 mmol g DW−1 h−1. This empirical critical value was close to the value reported previously for the OUR in fully aerobic, nitrogen-limited, continuous cultures (3.5 mmol g DW−1 h−1) (32). Thus, five different levels of dissolved oxygen spanning were selected to cover the complete range from anaerobic conditions to the maximum OUR (Table 1). These concentrations covered the entire oxygen-limited range for the yeast cells under enological conditions (54).

Fig 1
Relationship between the specific oxygen uptake rate (OUR) by cells of Saccharomyces cerevisiae strain EC1118 and the dissolved oxygen concentration. Gray symbols correspond to steady-state, nitrogen-limited continuous cultures with increasing dissolved ...

Carbon balances and specific rates of substrate uptake and product formation.

Accounting for all the carbon inputs and outputs to and from the culture (i.e., the carbon balance) is a requirement for quantitative assessments of metabolic physiology. We measured sugar and oxygen uptake rates as well as the accumulation of several organic compounds. Carbon balances, calculated through the yields in glucose (in C mol per C mol of glucose consumed), had less than an 8% error (see Table S1 in the supplemental material), indicating that no major product or substrate had been left out.

The biomass significantly increased with the availability of dissolved oxygen, almost doubling its concentration at 5 μM oxygen, compared to that under anaerobic conditions (Table 1). This also correlated with an increase in proline consumption, in line with the limitation of all other nitrogen sources (29). The proline consumption rate and biomass increase matched each other at dissolved oxygen concentrations of up to 2.7 μM, according to the biomass yield in nitrogen for this strain under anaerobic conditions (45; this work) (see Table S4 in the supplemental material). This correspondence did not occur with higher dissolved oxygen concentrations, likely because the biomass yield in nitrogen under aerobic conditions is different from that calculated under anaerobic conditions. Consistent with the biomass increase, the glucose consumption volumetric rate increased with increasing dissolved oxygen levels (from 141 to 197 C-mmol glucose liter−1 h−1 for 0 to 5 μM dissolved oxygen, respectively). Nevertheless, a further increase in the dissolved oxygen level resulted in only a modest increase in biomass and a decrease of the glucose consumption volumetric rate (to 125 C-mmol glucose liter−1 h−1). Surprisingly, the biomass yield in glucose (Yx/Glc) increased, while ethanol and CO2 yields in glucose decreased (see Table S1 in the supplemental material).

The impact of oxygen on cell physiology was more evident when specific rates were analyzed (Table 1). A negative correlation with dissolved oxygen was observed for glucose consumption and ethanol- and glycerol-specific production rates. However, in all cases, significant ethanol production was found, consistent with the presence of the Crabtree effect. Accordingly, the respiratory quotients (RQ) were all higher than 1 under all oxygenated conditions (Table 1). Organic acid production was also affected by oxygen availability. For instance, acetic acid was produced only under conditions of strict anaerobiosis (0.3 C-mmol g DW−1 h−1). On the other hand, a striking and significant (P < 0.01) increase in the level of succinic acid production occurred between the 1.2 and 2.7 μM dissolved oxygen conditions (from 0.02 to 0.27 C-mmol g DW−1 h−1). This increase was unexpected, as there is no known mechanism of succinic acid export in yeast, although succinic acid production has been consistently reported in previous studies (44, 45).

Metabolic flux analysis.

The redistribution of intracellular carbon fluxes in the central metabolic pathways of Saccharomyces cerevisiae occurring in response to the increasing availability of dissolved oxygen under enological conditions was determined by using the stoichiometric model. The carbon flux toward fermentation progressively decreased, with a concomitant increase of the flux through the TCA cycle, as the dissolved oxygen content increased (Fig. 2). We confirmed that the TCA cycle was not functional and operated in two branches under conditions of anaerobiosis, as suggested previously for anaerobic, carbon-limited conditions (10, 40); a similar situation was observed for the 1.2 μM dissolved oxygen level. However, at dissolved oxygen concentrations of 2.7 μM and higher, the TCA cycle followed its canonical direction, with a large increase in carbon flux circulation through it (P < 0.05) (Fig. 2). Another remarkable feature was the increase of the flux toward carbohydrate synthesis under conditions with 21 μM dissolved oxygen, increasing more than 60% compared with the flux under the other conditions (Fig. 2). This finding is supported by experimental data (Table 2).

Fig 2
Flux distributions in S. cerevisiae strain EC1118 grown in nitrogen-limited chemostats at a D of 0.1 h−1 from anaerobic (top) to 21 μM (bottom) dissolved oxygen conditions. Numbers indicate specific reactions related with oxygen consumption: ...
Table 2
Specific production rates of major biomass components at different dissolved oxygen concentrations

Sources and sinks of nucleotide cofactors.

MFA showed that mitochondria played an increasingly important role in NADH and ATP turnover rates at concentrations above 2.7 μM dissolved oxygen. Fluxes through respiration and cytosol to mitochondrial NADH exchange rose significantly as dissolved oxygen concentrations increased (Fig. 2). Mitochondrial NADH production and utilization significantly increased with increasing oxygen concentrations (P < 0.05), rising from 25% to 60% of the total NADH production rate in the cell when the dissolved oxygen level was augmented from 2.7 to 21 μM, respectively. Analyses of NADH reoxidization pathways showed that ethanol production was still the dominating form for the reoxidization of cytosolic NADH produced by glycolysis under all the conditions tested (Table 3). Below a concentration of 2.7 μM dissolved oxygen, glycerol is the only other electron sink, albeit a minor one (~4% of NADH reoxidization). Above concentrations of 2.7 μM dissolved oxygen, the reoxidization of cytosolic NADH through the mitochondrial redox shuttle significantly exceeded its reoxidization by glycerol synthesis (P < 0.05 by t test); moreover, glycerol synthesis showed almost no contribution under 21 μM dissolved oxygen conditions.

Table 3
Contribution to NADH and ATP turnover of different metabolic pathways in S. cerevisiae EC1118 grown under different dissolved oxygen culture conditions

Oxygen also extended the mitochondrial contribution to ATP production. Under anaerobic conditions, ATP was produced in glycolysis by substrate-level phosphorylation. On the other hand, the ATP produced by oxidative phosphorylation had a noticeable contribution, increasing from 5.7% to 21% of the total ATP when the dissolved oxygen content increased from 2.7 μM to 21 μM, respectively. The glycolytic pathway was still the main contributor, producing more than 70% of the total ATP (Table 3). Nevertheless, the amount of ATP produced by any means was significantly reduced, decreasing by approximately 30%, at the highest level of dissolved oxygen. This was also associated with an increase in the fraction of ATP used for maintenance, which increased from 43% to 54% under the 2.7 μM and 21 μM dissolved oxygen conditions, respectively.

NADPH metabolism was slightly affected by oxygen levels. The most remarkable change occurred in the aldehyde dehydrogenase reaction. The level of NADPH production by this reaction decreased from 4% to 1% when the available oxygen concentration increased from 0 μM to 1.2 μM dissolved oxygen. The rest of the NADPH was produced through the pentose phosphate pathway, a flux that slightly increased when comparing the anaerobic conditions to the rest of the conditions. However, the flux through this pathway represents only 2% of the carbon under all conditions (see Table S5 in the supplemental material), confirming results reported previously by Varela et al. (63).

Gene expression analysis.

To complement the metabolic flux data, global gene expression under the five dissolved oxygen conditions was determined. As a whole, the effect of oxygen on the transcriptome was limited, since only 8.8% of the transcripts (507 genes) were affected by oxygen availability. The latter was calculated by taking into account the differentially expressed genes in all possible comparisons of the different oxygen conditions. The comparison between the anaerobic conditions and the conditions with the highest concentration of dissolved oxygen (21 μM) yielded 313 differentially expressed genes. These data were consistent with data from previous reports (60), where 371 genes were differentially expressed between anaerobic and aerobic conditions in nitrogen-limited, continuous cultures.

Analysis of oxygen level transitions.

To simulate the oxygen consumption kinetics of yeast cells after an oxygen impulse during winemaking aeration operations (M. I. Moenne and E. Agosin, unpublished data), we simulated a pseudodynamic setting by comparing the data for one condition with those for conditions with the next highest concentration of dissolved oxygen (Table 4). The largest effect (200 differentially expressed genes) occurred upon the onset of the addition of oxygen, i.e., between the 0 and 1.2 μM dissolved oxygen concentrations. The transition between the 5 and 21 μM concentrations had an equivalent impact, causing differential changes in 195 genes. Remarkably, this was the transition between oxygen-limited and oxygen-saturated conditions (Fig. 1). The effect was much smaller when the intermediate transitions were compared (1.2 with 2.7 μM and 2.7 with 5 μM dissolved oxygen), affecting the expression of 25 and 19 genes, respectively.

Table 4
Characterization of differentially expressed genes across different oxygen levels

Genes differentially expressed between the 0 and 1.2 μM dissolved oxygen conditions.

The genes differentially expressed between 0 and 1.2 μM dissolved oxygen conditions can be classified into two groups. The first group consisted of six genes, annotated as “siderophore transport” genes, that participate in iron uptake. The putative increase in iron uptake could be related to the overexpression of hemoprotein-related genes, such as cytochromes. The second group comprised nine genes annotated as “amino acid transport” genes including the proline transporter PUT4. This highlights both the induction of respiratory genes and the impact of oxygen on the regulation of the nutrient uptake of yeast.

Genes differentially expressed between the 1.2 and 2.7 μM dissolved oxygen conditions.

Between the 1.2 and 2.7 μM dissolved oxygen conditions, key mitochondrial genes, such as CYC1 (cytochrome c) and NDE1 (external NADH dehydrogenase), were induced. NDE1p is responsible for the shuttling of cytosolic redox equivalents directly into the respiratory chain. GUT2, another gene responsible for a shuttle mechanism (49), was also induced between the 0 and 2.7 μM dissolved oxygen additions. Among the repressed genes, two members of the TIR genes and the HPF1 gene, all encoding cell wall mannoproteins, were found. Note that the repression of mannoproteins could have several consequences on wine quality, modulating astringency and other organoleptic properties (24). The AUS1 gene, involved in fatty acid and sterol uptake, was also repressed.

Genes differentially expressed between the 5 and 21 μM dissolved oxygen conditions.

The transition between dissolved oxygen concentrations of 5 and 21 μM consolidated the trend in the induction of respiratory genes. Induced genes included more respiratory genes, such as CYC1 and COX7 (subunit of complex IV), supporting the idea of active respiration, even under these high-glucose conditions. Other induced genes were stress response genes, such as CTT1 (catalase), HSP12 (membrane heat shock protein), and GRX4 (glutaredoxin), hinting at a potential oxidative stress occurring under this condition. The repressed genes at this transition included genes involved in iron metabolism as well as those involved in sulfate and phosphate ion transport (Table 4).

Other gene expression changes.

The genes annotated as belonging to the “TCA cycle” GO term showed no differential expression in the transitions analyzed. However, when anaerobic conditions and 21 μM dissolved oxygen conditions were compared, the TCA cycle term was statistically enriched among induced genes (P < 0.01), including the ACO1 (aconitase) and CIT1 (citrate synthase) genes. On the other hand, the FRD1 gene was significantly repressed when anaerobic conditions were compared with 5 μM dissolved oxygen conditions. FRD1 encodes a soluble fumarate reductase that is responsible for the operation of the reductive branch of the TCA cycle. This observation was in line with the redistribution of metabolic carbon fluxes observed by MFA with increasing dissolved oxygen concentrations: an increase of TCA fluxes and a disappearance of the anaerobic, two-branch operation of the TCA cycle.

Clustering analysis.

Genes that responded to oxygen under at least one condition were classified into 56 clusters (Fig. 3), by means of hierarchical clustering. We classified the clusters according to their major tendencies with regard to the dissolved oxygen concentration. We analyzed the 12 clusters that showed at least one enriched GO term according to the GO term enrichment analysis (see Table S3 in the supplemental material). We classified the clusters according to their general trends in relation to the dissolved oxygen concentration. We found four broad categories: clusters with genes downregulated with 21 μM dissolved oxygen, clusters with genes downregulated with 1.2 μM dissolved oxygen, and genes negatively and positively correlated with dissolved oxygen. We describe the genes in these clusters below.

Fig 3
Hierarchical clustering of the S. cerevisiae transcriptome data obtained under the five dissolved oxygen conditions. From the center to the outside, the dendrogram depicts (i) a hierarchical clustering tree, (ii) color ribbons indicating each of the 56 ...

Genes downregulated with 21 μM dissolved oxygen.

Several gene clusters were downregulated with 21 μM dissolved oxygen. Four clusters (clusters 2, 8, 26, and 52) related to iron metabolism were repressed with 21 μM dissolved oxygen. From these, clusters 2, 26, and 52 were induced with 1.2 μM dissolved oxygen, confirming the induction of iron uptake at low oxygen levels, a trend that is reversed at high oxygen concentrations. Besides the iron metabolism genes, several genes involved in ergosterol metabolism were downregulated at high oxygen levels. This suggests possible product repression when enough oxygen is present to synthesize these compounds (15).

Genes downregulated with 1.2 μM dissolved oxygen.

Several clusters showed a common negative response to low levels of oxygen. Gene expression levels in clusters 25 and 29 dropped significantly with 1.2 μM dissolved oxygen, although they increased concomitantly with increasing oxygen concentrations for the higher oxygen levels (Fig. 3). Several transcription factors belong to these clusters, most notably two positive regulators of nitrogen catabolism, DAL81 and GZF3 (23). Other genes, such as those from cluster 37, showed very low expression levels. This is characteristic of silenced transposable elements, the expression levels of which were further decreased with increasing oxygen concentrations. Cluster 43 also showed this pattern. The latter cluster includes FRD1, the fumarate reductase gene, confirming data from the two-branch TCA transition analysis, as well as SLC1, a key enzyme in phospholipid metabolism (4).

Genes negatively correlated with dissolved oxygen.

The family of TIR genes appears both in cluster 53 and with the 1.2 to 2.7 μM oxygen transition, showing a dose-dependent response to oxygen (Fig. 3). These results suggest that cell wall remodeling is taking place as the oxygen level increases, which could also be linked to increasing levels of ergosterol production (Table 2). Consistently, the UPC2 transcription factor, which regulates ergosterol biosynthesis and TIR genes (15), was also repressed as the oxygen level increased (cluster 13).

Genes positively correlated with dissolved oxygen.

Supporting the idea of a respiratory metabolism under these conditions, several genes related to complex IV of the respiratory chain were induced together with dissolved oxygen, and they are grouped into cluster 32 (Fig. 3).

Despite these findings, some metabolic pathways showed phenotypic changes that were not reflected in gene expression levels, such as acetate production. For instance, we observed that ALD genes, coding for aldehyde dehydrogenases responsible for acetate production (7), were not affected by oxygen levels despite the fact that acetate is detected only in anaerobic cultures.

DISCUSSION

This research addresses the impact of different levels of dissolved oxygen on the physiology of an industrial strain of S. cerevisiae under enological (i.e., carbon-sufficient, nitrogen-limited) conditions. We experimentally captured a subset of dissolved oxygen concentrations, aiming to represent the oxygen-limiting range of dissolved oxygen concentrations found in discrete enological aeration operations (54). We found that oxygen had major metabolic effects, reflected by changes in the levels of production of several extracellular compounds and the metabolic flux redistribution within the cell. For instance, ethanol- and glycerol-specific production rates decreased when the oxygen level was increased, along with the respiratory quotient. This is an indication of a transition from fermentative to mixed respirofermentative metabolism. Nevertheless, the respiratory quotient values indicate that fully respiratory metabolism was never achieved. Although this finding was reported previously for laboratory strains under nitrogen-limited conditions (60), the evidence of active respiratory metabolism under enological conditions is striking, since there is a general belief that respiration is under catabolic repression under these conditions (21, 51, 66).

Respiratory quotients showed a large decrease between concentrations of 1.2 and 2.7 μM dissolved oxygen. Consistently, metabolic flux analysis predicts two very different metabolic configurations depending on the concentration of dissolved oxygen: fully fermentative metabolism (including 0 and 1.2 μM dissolved oxygen) and mixed respirofermentative metabolism (dissolved oxygen concentration of 2.7 μM and higher). Fully fermentative conditions featured low carbon fluxes and a two-branch operation of the TCA cycle, ethanol and glycerol as the only redox sinks, and glycolysis as the major source for ATP and NADH. The operation of the two-branch TCA cycle under anaerobic conditions was initially suggested by MFA (40) and later proven by a 13C-based metabolomic analysis (10) under carbon-limited conditions. To the best of our knowledge, this is the first report of the operation of a two-branch TCA cycle under carbon-sufficient, nitrogen-limited conditions.

Cultures with more than 2.7 μM dissolved oxygen showed a mixed respirofermentative metabolism despite the high external sugar concentration (40 g/liter). It is worth noting that the model assumes that respiration is working under these conditions. Under this assumption, the estimations of ethanol and CO2 production were reliable (see Table S2 in the supplemental material), providing further support for a functional and active respiratory pathway under these culture conditions. The main features of the putative mixed respirofermentative metabolic configuration were significant respiratory activity, TCA cycle operation in its canonical direction, increased levels of succinic acid production, and a mitochondrial redox shuttle working as a significant cytosolic NADH sink. This effect of oxygen on yeast metabolism is mirrored by gene expression, e.g., the induction of the COX and CYC1 genes (respiration), the repression of the fumarate reductase gene (reductive branch of the TCA cycle), and the induction of the NDE1 and GUT2 genes (mitochondrial electron shuttles). Moreover, TCA genes such as the ACO1 (aconitase) and CIT1 (citrate synthase) genes are induced at the highest level of oxygen tested. Altogether, the data confirm the occurrence of a respiratory metabolism. Nevertheless, ATP and NADH were still produced mainly by glycolysis, confirming that this is the major pathway regarding carbon flow operating under these conditions.

One of the hallmarks of mixed respirofermentative metabolism was the large increase (approximately 10-fold) in the level of succinic acid production between 1.2 and 2.7 μM dissolved oxygen. Most likely, this resulted from the much larger flux toward the TCA cycle than with the two-branch TCA cycle, since this was not observed at concentrations of dissolved oxygen above 1.2 μM. Nevertheless, how succinic acid is exported remains unclear. Several succinic acid transporters of the mitochondrial membrane that could account for export out of this organelle are known (42). However, no transporter for succinic acid at the plasma membrane has yet been identified. Therefore, succinic acid could be exported by diffusion and/or active transport, although the diffusion mechanism implicates an actual production rate 6-fold higher than the one observed (see the supplemental material). Preliminary experiments argued in favor of an active transport mechanism, since batch cultures of S. cerevisiae EC1118 with approximately 2.7 μM dissolved oxygen were able to export succinic acid despite being supplemented with exogenous succinic acid.

The export of acetic acid also showed an interesting trend, being present only under strict anaerobic conditions. With a low dissolved oxygen level (1.2 μM), acetic acid production disappeared. This correlated with a decrease of the flux through the aldehyde dehydrogenase reaction and a slight increase of the flux through the pentose phosphate pathway. Therefore, it is likely that acetate production in anaerobiosis is necessary to provide NADPH to the cell, a function that is taken over by the pentose phosphate pathway when oxygen is available. In fact, both pathways are complementary, and together, they are the only source of NADPH in glucose-containing media (26). While the mechanism for the coordination of these two pathways is unclear, the lack of a change in the expression level of the ALD6 gene (encoding aldehyde dehydrogenase) in response to oxygen suggests a nontranscriptional mechanism. The involvement of Ald6p in a calcium/calmodulin-dependent signaling pathway supports this hypothesis (9).

Another feature of the mixed respirofermentative metabolism under the culture conditions of this study was the increase in the shuttling of redox equivalents from the cytoplasm to the mitochondria. MFA suggested that the functioning of the Adh3p shuttle is necessary to explain the ethanol reduction under 21 μM dissolved oxygen conditions. However, the function of neither Nde1p nor Gut2p in respirofermentative metabolism can be ruled out, since the replacement of Adh3p with Nde1p does not change model estimations. Moreover, the inclusion of the Gut2p shuttle in the MFA model yielded good estimations under all aerobic conditions (data not shown). Experimental evidence for Nde1p and Gut2p suggested that both mechanisms are active, as the enzymatic activities of both mechanisms were detected under aerobic, nitrogen-limited conditions (41). Moreover, at the oxygen metabolic threshold (between 1.2 and 2.7 μM dissolved oxygen), we found an induction of the Nde1p gene at the transcriptional level. The expression of the Gut2p gene was also gradually induced by oxygen, contradicting previous studies where this gene was reported to be catabolically repressed (49). Therefore, the increased mitochondrial capacity for cytosolic NADH reoxidization could be attributed to the presence of these shuttles, as well as of the Adh3p shuttle, which showed constant, significant gene expression regardless of the oxygen level.

Despite the increase in the mitochondrial shuttle activity, mitochondria showed a limited reoxidization capacity, as reflected by the large contribution of the ethanol production pathway to NADH reoxidization, even when the yeast was at its fastest OUR (Table 3). This limited capacity, and not catabolic repression, was the most likely cause of the Crabtree effect (65) and, overall, of the inability of Saccharomyces cerevisiae to develop a fully respiratory metabolism under conditions of nitrogen limitation. Whether this limitation occurs at the shuttle level or at the level of the activity of the respiratory enzymes is unclear from our experiments. Nevertheless, the shuttle hypothesis is in line with recent reports that proposed that mitochondrial membrane surface availability is crucial in regulating the respirofermentative transition (69). Therefore, focusing on the shuttles can be a valuable and novel approach to shift yeast metabolism to a more oxidative state. This could be useful, for example, to lower the level of ethanol production in enological fermentations, an active area of research in metabolic engineering (for example, see reference 18).

The hypothesis of limited mitochondrial reoxidization as the major cause of the Crabtree effect was reinforced by data from gene expression analyses. We found little evidence of glucose catabolic repression of respiratory enzymes, as the COX and cytochrome b and c (CYB2 and CYC1) genes are responsive to oxygen despite high external sugar concentrations. This was also observed previously by Tai et al. (60). Even under anaerobic conditions, the expression levels of these genes were higher than the average (Fig. 3), discarding any repressive effect from the high external glucose concentration. The data also suggest that oxygen was able to override glucose catabolic repression, as the expressions of many responsive genes (such as COX) were positively correlated with dissolved oxygen concentrations (Fig. 3). Therefore, heme-dependent oxygen induction, through the HAP transcription factors (46), could be able to bypass catabolic repression. However, it was not possible to detect significant changes in transcription factor activities by using network component analysis (35), suggesting that the mechanisms involved in this phenomenon are too elaborate to be inferred only from gene expression profiles. This regulatory landscape appears to encompass all conditions in the presence of oxygen. However, at the highest dissolved oxygen level (21 μM), a puzzling gene expression scheme occurred. Respiratory genes were induced, but those encoding ion transporters, such as iron and copper, were repressed. This is the opposite of what occurs upon oxygen exposure, where both iron transport and respiratory genes are induced coordinately, since iron is a requirement for the building of the essential hemoproteins of the respiratory chain. The situation with 21 μM dissolved oxygen could be a major physiological reconfiguration when the maximal OUR capacity of yeast cells is reached (Fig. 1). One component of this reconfiguration could be oxidative stress, which would also explain the iron uptake restriction, as an excess of iron generates more free radicals in the cell (19). Furthermore, some oxidative stress marker genes are induced, such as GRX4. Grx4p negatively regulates the activity of Aft1p (48), the master transcriptional factor regulating iron metabolism. The latter provides a mechanism to explain iron transporter repression. Aft1p could also influence nitrogen metabolism (57), which is indeed the case under high-oxygen conditions. For instance, DAL81 (regulator of allantoin utilization), a transcription factor that positively regulates the utilization of alternative nitrogen sources, shows the highest expression levels under anaerobic conditions and with 21 μM dissolved oxygen as well, both conditions under which nitrogen is effectively unavailable. Aft1p could influence the metabolisms of other nutrients (57), for which we also found transporter repression (Table 4).

The repression of the nutrient transporters could explain the modest biomass increase when 21 μM dissolved oxygen conditions were compared with 5 μM dissolved oxygen conditions. Also, limited nutrient availability could play a role in establishing OUR saturation at 21 μM dissolved oxygen (Fig. 1), since nutrient limitation can impair the cell's capacity to build more respiratory complexes and/or mitochondria. This is supported by the fact that the critical OUR is much higher in carbon-limited cultures (32), indicating that the catalytic capacity of the respiratory chain can sustain a higher OUR. Therefore, the low critical OUR under nitrogen-limited culture conditions is more likely related to the total respiratory complexes available.

The evidence presented suggests some degree of both oxidative and nutritional stress under oxygen-saturated conditions (21 μM). Consistently, we observed that with dissolved oxygen concentrations higher than 21 μM (Fig. 1), the biomass can drop as low as 4 g liter−1 (data not shown), suggesting that at 21 μM dissolved oxygen, yeast cells are at the edge of their biomass-producing capacities. Additionally, under this condition, there is a strong increase in carbohydrate synthesis (Table 2), which is a landmark of physiological stress responses in microorganisms (34). Altogether, these data suggest that yeast cells are possibly under multifactorial stress under nitrogen-limited conditions with the OUR saturation regime. Increased carbohydrate synthesis may also explain the simultaneous increase of the biomass/glucose yield and decrease of ethanol and CO2/glucose yields, another puzzling feature of the 21 μM dissolved oxygen conditions (see Table S1 in the supplemental material).

The changes in dissolved oxygen concentrations also impact the cell wall. For instance, SLC1, encoding a key enzyme in phospholipid metabolism, is repressed with 1.2 μM dissolved oxygen. At the highest oxygen concentrations, these changes can be another source of stress. For example, ergosterol and unsaturated lipid biosynthetic gene expression levels were significantly decreased with 21 μM dissolved oxygen. In fact, the ergosterol content and synthesis rate decreased by 77% when the 5 and 21 μM dissolved oxygen conditions were compared (Table 2). This reduction in the level of ergosterol could also contribute to the establishment of a stress phenotype, as ergosterol has been regarded as a protective compound against oxidative stress (31). On the other hand, oxygen represses several mannoprotein-encoding genes (TIR) in a dose-dependent fashion This could be caused by Upc2p, an inducer of ergosterol biosynthesis and of TIR genes (15). The Upc2p level decreases at high dissolved oxygen levels (see Table S3 in the supplemental material) by an as-yet-unknown mechanism. Moreover, oxygen represses another cell wall-related gene, MUC1 (also called FLO11), which is critical for yeast flocculation (25).

From a winemaking perspective, metabolic flux analysis and gene expression data suggest that elevated dissolved oxygen concentrations could affect yeast performance during and after fermentation. During fermentation, reaching oxygen levels above 2.7 μM would reduce the ethanol yield (see Table S1 in the supplemental material), and reaching levels of 21 μM or higher would induce a severe stress on the yeast cell, further decreasing its fermentative capacity. These oxygen levels are easily achieved in industrial winemaking practice (Moenne et al., unpublished). Therefore, these results indicate that it is advisable not to keep wine yeast cells at these oxygen levels for an extended period of time. Furthermore, the repression of mannoprotein genes by oxygen could also be damaging, since their presence in wine has been linked to many beneficial effects, such as increased mouthfeel, aroma retention, and astringency reduction (24). Furthermore, one of these mannoproteins (Hpf1p) is a key contributor to white wine clarification by protein haze removal (17). Another repressed gene product, Flo1p, is crucial for flocculation, a process needed for the removal of yeast from wine. The repression of these proteins could be a novel mechanism of how oxygen can affect wine quality, besides its known oxidative effect on phenolic compounds (68).

On the contrary, low oxygen levels could be beneficial for winemaking. For instance, with 1.2 μM dissolved oxygen, ethanol production is similar to anaerobiosis, with no acetic acid production, which is beneficial since acetic acid is a common “off-flavor” in wine. The viability and stress resistance of the wine yeast might also increase, as the specific ergosterol content, a protective compound against stresses in wine fermentation (31), increases to its maximum (M. Orellana, F. F. Aceituno, and E. Agosin, unpublished data). Moreover, nitrogen-deficient musts can be more efficiently utilized in winemaking, since the proline carrier (PUT4) is induced and proline assimilation is effectively increased with 1.2 μM oxygen (Table 2), in turn increasing biomass synthesis. Further research will be aimed at finding a tradeoff between oxygen addition and limitation under winemaking conditions.

In conclusion, we found that in a nitrogen-limited setting, oxygen exerted a large metabolic effect on yeast mitochondria, and there is a threshold that separates fermentative and respirofermentative metabolisms. This is related to the expressions of some key genes, such as COX, NDE1, GUT2, and FRD1. Changes in the expression levels of these genes could explain most of the flux changes estimated in relation to respiration, cytosolic NADH shuttling to the mitochondria, and two-branch cycle operation. Furthermore, gene induction casts doubts on the operation of glucose catabolic repression under nitrogen-limited conditions, since it can be overridden by oxygen. Other genes affected by oxygen were mannoprotein-encoding genes, which were repressed as part of the global remodeling of the cell wall. This repression could have negative consequences in winemaking, highlighting the dual role of oxygen in “making or breaking wines.”

Supplementary Material

Supplemental material:

ACKNOWLEDGMENTS

This research was supported by FONDECYT grant number 1090520 from CONICYT, Chile, to E.A.; doctoral thesis support grant AT-24100170 from CONICYT, Chile, to F.F.A.; and an ICM (Iniciativa Científica Milenio, Chile) grant (no. P09-016-F) to F.M. We are grateful to Lallemand, Inc. (Canada), for financial support and Indura S.A. (Chile) for providing gas mixtures. F.F.A., M.O., and A.W.S. were supported by CONICYT and VRI-UC doctoral fellowships.

We thank Paulina Torres, Pablo Cañón, and Camila Orellana for technical support; Leonardo I. Almonacid (Molecular Bioinformatics Laboratory, Millennium Institute on Immunology and Immunotherapy) for bioinformatics support and GO enrichment analysis; and Elena Vidal and Rodrigo A. Gutierrez (Department of Molecular Genetics and Microbiology, Faculty of Biological Sciences, Pontificia Universidad Católica de Chile) for the use of microarray facilities and technical support.

Footnotes

Published ahead of print 21 September 2012

Supplemental material for this article may be found at http://aem.asm.org/.

REFERENCES

1. Affymetrix 2009. Gene Chip expression analysis technical manual. Affymetrix, Santa Clara, CA
2. Affymetrix 2010. GeneChip 3′ IVT Express kit manual. Affymetrix, Santa Clara, CA
3. Alberty RA. 2006. Biochemical thermodynamics: applications of mathematica. Methods Biochem. Anal. 48:1–458 [PubMed]
4. Athenstaedt K, Daum G. 1997. Biosynthesis of phosphatidic acid in lipid particles and endoplasmic reticulum of Saccharomyces cerevisiae. J. Bacteriol. 179:7611–7616 [PMC free article] [PubMed]
5. Bakker BM, et al. 2001. Stoichiometry and compartmentation of NADH metabolism in Saccharomyces cerevisiae. FEMS Microbiol. Rev. 25:15–37 [PubMed]
6. Bluthgen N, et al. 2005. Biological profiling of gene groups utilizing Gene Ontology. Genome Inform. 16:106–115 [PubMed]
7. Boubekeur S, Camougrand N, Bunoust O, Rigoulet M, Guerin B. 2001. Participation of acetaldehyde dehydrogenases in ethanol and pyruvate metabolism of the yeast Saccharomyces cerevisiae. Eur. J. Biochem. 268:5057–5065 [PubMed]
8. Breitling R, Armengaud P, Amtmann A, Herzyk P. 2004. Rank products: a simple, yet powerful, new method to detect differentially regulated genes in replicated microarray experiments. FEBS Lett. 573:83–92 [PubMed]
9. Butcher RA, Schreiber SL. 2004. Identification of Ald6p as the target of a class of small-molecule suppressors of FK506 and their use in network dissection. Proc. Natl. Acad. Sci. U. S. A. 101:7868–7873 [PubMed]
10. Camarasa C, Grivet JP, Dequin S. 2003. Investigation by 13C-NMR and tricarboxylic acid (TCA) deletion mutant analysis of pathways for succinate formation in Saccharomyces cerevisiae during anaerobic fermentation. Microbiology 149:2669–2678 [PubMed]
11. Carbon S, et al. 2009. AmiGO: online access to ontology and annotation data. Bioinformatics 25:288–289 [PMC free article] [PubMed]
12. Celton M, Goelzer A, Camarasa C, Fromion V, Dequin S. 2012. A constraint-based model analysis of the metabolic consequences of increased NADPH oxidation in Saccharomyces cerevisiae. Metab. Eng. 14:366–379 [PubMed]
13. Christie KR, et al. 2004. Saccharomyces Genome Database (SGD) provides tools to identify and analyze sequences from Saccharomyces cerevisiae and related sequences from other organisms. Nucleic Acids Res. 32:D311–D314 doi:10.1093/nar/gkh033 [PMC free article] [PubMed]
14. Conesa A, et al. 2005. Blast2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 21:3674–3676 [PubMed]
15. Davies BSJ, Rine J. 2006. A role for sterol levels in oxygen sensing in Saccharomyces cerevisiae. Genetics 174:191–201 [PubMed]
16. Doran PM. 1995. Bioprocess engineering principles. Academic Press, London, United Kingdom
17. Dupin IVS, et al. 2000. Saccharomyces cerevisiae mannoproteins that protect wine from protein haze: evaluation of extraction methods and immunolocalization. J. Agric. Food Chem. 48:1086–1095 [PubMed]
18. Ehsani M, Fernandez MR, Biosca JA, Julien A, Dequin S. 2009. Engineering of 2,3-butanediol dehydrogenase to reduce acetoin formation by glycerol-overproducing, low-alcohol Saccharomyces cerevisiae. Appl. Environ. Microbiol. 75:3196–3205 [PMC free article] [PubMed]
19. Eide DJ. 1998. The molecular biology of metal ion transport in Saccharomyces cerevisiae. Annu. Rev. Nutr. 18:441–469 [PubMed]
20. Frick O, Wittmann C. 2005. Characterization of the metabolic shift between oxidative and fermentative growth in Saccharomyces cerevisiae by comparative C-13 flux analysis. Microb. Cell Fact. 4:30 doi:10.1186/1475-2859-4-30 [PMC free article] [PubMed]
21. Gancedo JM. 1998. Yeast carbon catabolite repression. Microbiol. Mol. Biol. Rev. 62:334–361 [PMC free article] [PubMed]
22. Gentleman RC, et al. 2004. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 5:R80 doi:10.1186/gb-2004-5-10-r80 [PMC free article] [PubMed]
23. Georis I, Feller A, Vierendeels F, Dubois E. 2009. The yeast GATA factor Gat1 occupies a central position in nitrogen catabolite repression-sensitive gene activation. Mol. Cell. Biol. 29:3803–3815 [PMC free article] [PubMed]
24. Gonzalez-Ramos D, Cebollero E, Gonzalez R. 2008. A recombinant Saccharomyces cerevisiae strain overproducing mannoproteins stabilizes wine against protein haze. Appl. Environ. Microbiol. 74:5533–5540 [PMC free article] [PubMed]
25. Govender P, Kroppenstedt S, Bauer FF. 2011. Novel wine-mediated FLO11 flocculation phenotype of commercial Saccharomyces cerevisiae wine yeast strains with modified FLO gene expression. FEMS Microbiol. Lett. 317:117–126 [PubMed]
26. Grabowska D, Chelstowska A. 2003. The ALD6 gene product is indispensable for providing NADPH in yeast cells lacking glucose-6-phosphate dehydrogenase activity. J. Biol. Chem. 278:13984–13988 [PubMed]
27. Hazelwood LA, Daran JM, van Maris AJ, Pronk JT, Dickinson JR. 2008. The Ehrlich pathway for fusel alcohol production: a century of research on Saccharomyces cerevisiae metabolism. Appl. Environ. Microbiol. 74:2259–2266 [PMC free article] [PubMed]
28. Hoskisson PA, Hobbs G. 2005. Continuous culture—making a comeback? Microbiology 151:3153–3159 [PubMed]
29. Ingledew WM, Magnus CA, Sosulski FW. 1987. Influence of oxygen on proline utilization during the wine fermentation. Am. J. Enol. Vitic. 38:246–248
30. Jouhten P, et al. 2008. Oxygen dependence of metabolic fluxes and energy generation of Saccharomyces cerevisiae CEN.PK113-1A. BMC Syst. Biol. 2:60 doi:10.1186/1752-0509-2-60 [PMC free article] [PubMed]
31. Landolfo S, et al. 2010. Oleic acid and ergosterol supplementation mitigates oxidative stress in wine strains of Saccharomyces cerevisiae. Int. J. Food Microbiol. 141:229–235 [PubMed]
32. Larsson C, Vonstockar U, Marison I, Gustafsson L. 1993. Growth and metabolism of Saccharomyces cerevisiae in chemostat cultures under carbon-limiting, nitrogen-limiting, or carbon-limiting and nitrogen-limiting conditions. J. Bacteriol. 175:4809–4816 [PMC free article] [PubMed]
33. Letunic I, Bork P. 2007. Interactive Tree of Life (iTOL): an online tool for phylogenetic tree display and annotation. Bioinformatics 23:127–128 [PubMed]
34. Li L, et al. 2009. The induction of trehalose and glycerol in Saccharomyces cerevisiae in response to various stresses. Biochem. Biophys. Res. Commun. 387:778–783 [PubMed]
35. Liao JC, et al. 2003. Network component analysis: reconstruction of regulatory signals in biological systems. Proc. Natl. Acad. Sci. U. S. A. 100:15522–15527 [PubMed]
36. Liu H, Sanuda-Pena MC, Harvey-White JD, Kalra S, Cohen SA. 1998. Determination of submicromolar concentrations of neurotransmitter amino acids by fluorescence detection using a modification of the 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate method for amino acid analysis. J. Chromatogr. A 828:383–395 [PubMed]
37. Mateles RI, Battat E. 1974. Continuous culture used for media optimization. Appl. Microbiol. 28:901–905 [PMC free article] [PubMed]
38. Murray DB, Haynes K, Tomita M. 2011. Redox regulation in respiring Saccharomyces cerevisiae. Biochim. Biophys. Acta 1810:945–958 [PubMed]
39. Nielsen J, Villadsen J, Liden G. 2003. Bioreaction engineering principles, 2nd ed Kluwer Academic/Plenum Publisher, New York, NY
40. Nissen TL, Schulze U, Nielsen J, Villadsen J. 1997. Flux distributions in anaerobic, glucose-limited continuous cultures of Saccharomyces cerevisiae. Microbiology 143(Pt 1):203–218 [PubMed]
41. Pahlman IL, Gustafsson L, Rigoulet M, Larsson C. 2001. Cytosolic redox metabolism in aerobic chemostat cultures of Saccharomyces cerevisiae. Yeast 18:611–620 [PubMed]
42. Palmieri L, Runswick MJ, Fiermonte G, Walker JE, Palmieri F. 2000. Yeast mitochondrial carriers: bacterial expression, biochemical identification and metabolic significance. J. Bioenerg. Biomembr. 32:67–77 [PubMed]
43. Pearson RK, Zylkin T, Schwaber JS, Gonye GE. 2004. Quantitative evaluation of clustering results using computational negative controls, p 188–199 In Berry MW, Umeshwar D, Kamath C, Skilliconr D, editors. (ed), Proceedings of the Fourth SIAM International Conference on Data Mining Society for Industrial and Applied Mathematics, Lake Buena Vista, FL
44. Pizarro F, et al. 2007. Coupling kinetic expressions and metabolic networks for predicting wine fermentations. Biotechnol. Bioeng. 98:986–998 [PubMed]
45. Pizarro FJ, Jewett MC, Nielsen J, Agosin E. 2008. Growth temperature exerts differential physiological and transcriptional responses in laboratory and wine strains of Saccharomyces cerevisiae. Appl. Environ. Microbiol. 74:6358–6368 [PMC free article] [PubMed]
46. Plakunov VK, Shelemekh OV. 2009. Mechanisms of oxygen regulation in microorganisms. Microbiology (Russia) 78:535–546
47. Pronk JT, Steensma HY, vanDijken JP. 1996. Pyruvate metabolism in Saccharomyces cerevisiae. Yeast 12:1607–1633 [PubMed]
48. Pujol-Carrion N, Belli G, Herrero E, Nogues A, de la Torre-Ruiz MA. 2006. Glutaredoxins Grx3 and Grx4 regulate nuclear localisation of Aft1 and the oxidative stress response in Saccharomyces cerevisiae. J. Cell Sci. 119:4554–4564 [PubMed]
48a. R Development Core Team 2012. R: a language and environment for statistical computing. The R Foundation for Statistical Computing, Vienna, Austria
49. Rigoulet M, et al. 2004. Organization and regulation of the cytosolic NADH metabolism in the yeast Saccharomyces cerevisiae. Mol. Cell. Biochem. 256-257:73–81 [PubMed]
50. Rintala E, et al. 2011. Transcriptional responses of Saccharomyces cerevisiae to shift from respiratory and respirofermentative to fully fermentative metabolism. OMICS 15:461–476 [PMC free article] [PubMed]
51. Rosenfeld E, Beauvoit B. 2003. Role of the non-respiratory pathways in the utilization of molecular oxygen by Saccharomyces cerevisiae. Yeast 20:1115–1144 [PubMed]
52. Rosenfeld E, Beauvoit B, Blondin B, Salmon JM. 2003. Oxygen consumption by anaerobic Saccharomyces cerevisiae under enological conditions: effect on fermentation kinetics. Appl. Environ. Microbiol. 69:113–121 [PMC free article] [PubMed]
53. Rosenfeld E, Beauvoit B, Rigoulet M, Salmon JM. 2002. Non-respiratory oxygen consumption pathways in anaerobically-grown Saccharomyces cerevisiae: evidence and partial characterization. Yeast 19:1299–1321 [PubMed]
54. Saa PA, Moenne MI, Perez-Correa JR, Agosin E. 2012. Modeling oxygen dissolution and biological uptake during pulse oxygen additions in oenological fermentations. Bioprocess Biosyst. Eng. 35:1167–1178 [PubMed]
55. Sablayrolles JM, Julien A, Roustan JL, Dulau L. 2000. Comparison of nitrogen and oxygen demands of enological yeasts: technological consequences. Am. J. Enol. Vitic. 51:215–222
56. Salmon JM, Barre P. 1998. Improvement of nitrogen assimilation and fermentation kinetics under enological conditions by derepression of alternative nitrogen-assimilatory pathways in an industrial Saccharomyces cerevisiae strain. Appl. Environ. Microbiol. 64:3831–3837 [PMC free article] [PubMed]
57. Shakoury-Elizeh M, et al. 2004. Transcriptional remodeling in response to iron deprivation in Saccharomyces cerevisiae. Mol. Biol. Cell 15:1233–1243 [PMC free article] [PubMed]
58. Sokal RR, Rohlf FJ. 1962. The comparison of dendrograms by objective methods. Taxon 11:33–40
59. Stephanopoulos G, Aristidou A, Nielsen J. 1998. Metabolic engineering: principles and methodologies. Academic Press, San Diego, CA
60. Tai SL, et al. 2005. Two-dimensional transcriptome analysis in chemostat cultures. Combinatorial effects of oxygen availability and macronutrient limitation in Saccharomyces cerevisiae. J. Biol. Chem. 280:437–447 [PubMed]
61. Reference deleted.
62. Valero E, Millan C, Ortega JM. 2001. Influence of oxygen addition during growth phase on the biosynthesis of lipids in Saccharomyces cerevisiae (M330-9) in enological fermentations. J. Biosci. Bioeng. 92:33–38 [PubMed]
63. Varela C, Pizarro F, Agosin E. 2004. Biomass content governs fermentation rate in nitrogen-deficient wine musts. Appl. Environ. Microbiol. 70:3392–3400 [PMC free article] [PubMed]
64. Vargas FA, Aceituno FF, Agosin E. 2010. Biochemistry and molecular biology of yeast alcoholic fermentation, p 201–217 In Pandey A, et al., editors. (ed), Comprehensive food fermentation and biotechnology, vol 2 Asiatech Publishers, New Delhi, India
65. Vemuri GN, Eiteman MA, McEwen JE, Olsson L, Nielsen J. 2007. Increasing NADH oxidation reduces overflow metabolism in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. U. S. A. 104:2402–2407 [PubMed]
66. Walker GM. 1998. Yeast physiology and biotechnology. John Wiley & Sons, West Sussex, United Kingdom
67. Wang NS, Stephanopoulos G. 1983. Application of macroscopic balances to the identification of gross measurement errors. Biotechnol. Bioeng. 25:2177–2208 [PubMed]
68. Waterhouse AL, Laurie VF. 2006. Oxidation of wine phenolics: a critical evaluation and hypotheses. Am. J. Enol. Vitic. 57:306–313
69. Zhuang K, Vemuri GN, Mahadevan R. 2011. Economics of membrane occupancy and respiro-fermentation. Mol. Syst. Biol. 7:500 doi:10.1038/msb.2011.34 [PMC free article] [PubMed]

Articles from Applied and Environmental Microbiology are provided here courtesy of American Society for Microbiology (ASM)