Saccharomyces cerevisiae CEN.PK113-1A was grown in glucose-limited chemostat culture with 0%, 0.5%, 1.0%, 2.8% or 20.9% O2 in the inlet gas (D=0.10 h−1, pH 5, 30°C) to determine the effects of oxygen on 17 metabolites and 69 genes related to central carbon metabolism. The concentrations of tricarboxylic acid cycle (TCA) metabolites and all glycolytic metabolites except 2-phosphoglycerate+3-phosphoglycerate and phosphoenolpyruvate were higher in anaerobic than in fully aerobic conditions. Provision of only 0.5–1% O2 reduced the concentrations of most metabolites, as compared with anaerobic conditions. Transcription of most genes analyzed was reduced in 0%, 0.5% or 1.0% O2 relative to cells grown in 2.8% or 20.9% O2. Ethanol production was observed with 2.8% or less O2. After steady-state analysis in defined oxygen concentrations, the conditions were switched from aerobic to anaerobic. Metabolite and transcript levels were monitored for up to 96 h after the transition, and this showed that more than 30 h was required for the cells to fully adapt to anaerobiosis. Levels of metabolites of upper glycolysis and the TCA cycle increased following the transition to anaerobic conditions, whereas those of metabolites of lower glycolysis generally decreased. Gene regulation was more complex, with some genes showing transient upregulation or downregulation during the adaptation to anaerobic conditions.
The physiology of Saccharomyces cerevisiae under fermentative, respiratory and respirofermentative conditions has always attracted considerable attention, both because it is one of the few eukaryotic organisms that can grow in truly anaerobic conditions, and because of the industrial importance of growth in anaerobic or aerobic conditions for the production of ethanol, proteins, cell biomass, and other products. The ability of S. cerevisiae to ferment glucose to ethanol even in aerobic conditions (the Crabtree effect) has made the understanding of respirofermentative growth important for optimizing aerobic processes in which ethanol is an undesirable byproduct. Furthermore, the development of S. cerevisiae as a host organism for the production of novel products such as lactic acid or 3-hydroxypropanoic acid may lead to an increase in the application of well-controlled, oxygen-restricted industrial processes, because the energy balance is probably negative under completely anaerobic conditions, whereas high oxygen concentrations lead to a loss in product yield (van Maris et al., 2004). Inadequate mixing in high-cell-density and very high-cell-density cultures may result in regions of very low oxygen provision, to which processes operating under oxygen-restricted conditions will be particularly vulnerable, and it is important to understand how cells respond to small changes in oxygen concentration and variable oxygen provision.
Although respirofermentative physiology and the transition between respirofermentative and purely fermentative growth can be studied in batch cultures (e.g. Burke et al., 1997; Lai et al., 2006), it is difficult to separate the effects of changing nutrient concentrations and specific growth rate from the effects of oxygen provision. Furthermore, purely respiratory growth in batch cultures is obtained by growing cells on nonfermentable or poorly fermentable carbon sources, adding carbon source to the variables. Respirofermentative physiology has also been obtained by growing cells in glucose-limited chemostat cultures, in which respirofermentative growth occurs at high specific growth rates, and respiratory growth at low specific growth rates (e.g. Frick & Wittmann, 2005). However, in this case, respirofermentative physiology is not distinguished from growth rate-related physiology. Franzén (2003) determined metabolic fluxes in S. cerevisiae in anaerobic and very low oxygen [0.1–2 mmol O2 (g biomass)−1 h−1] continuous flow cultures at various dilution rates, but there is a general lack of information on the effects of low oxygen concentration on yeast cell physiology in constant conditions.
In order to understand the response of S. cerevisiae to oxygen, we assessed the physiology of CEN.PK113-1A cells in terms of 17 metabolites and 69 genes related to central carbon metabolism during glucose-limited steady states with various concentrations of provided oxygen. After assessment of the steady states, anaerobic conditions were imposed in the same cultures to determine how the cells achieved their new, anaerobic steady state. By maintaining the cells under otherwise constant conditions, the effects of glucose repression and changes in specific growth rate that occur in batch cultures were minimized.
This work also highlights some of the problems and possibilities of identifying biomarkers for aerobic/anaerobic physiology. The increasing demand for information on microbial physiology that can be obtained directly from a production system, and then used to optimize the actual production process, requires the identification of one or a few marker genes or metabolites that would be indicative of the critical physiology for the specific process. Data obtained from both steady-state and transient conditions will be useful for their identification.
Materials and methods
Strain and medium
Saccharomyces cerevisiae CEN.PK113-1A (MATα, URA3, HIS3, LEU2, TRP1, MAL2-8c, SUC2) was kindly provided by Dr P. Kötter (Institut für Mikrobiologie, J.W. Goethe Universität Frankfurt, Germany; de Jong-Gubbels et al., 1998) and stored in glycerol (30% v/v) at −80°C.
Yeasts were grown in the defined minimal medium described by Verduyn (1992), with 10 g glucose L−1 as carbon source, and supplemented with 10 mg ergosterol L−1 and 420 mg Tween-80 L−1. BDH silicone antifoam (0.5 mL L−1) was used to prevent foam production in the cultures.
Cells were grown in 0.8–1 L of medium in B. Braun Biotech International (Sartorius AG, Germany) Biostat CT (2.5 L working volume) bioreactors. Cultures were inoculated to an initial OD600 nm of c. 0.5, and maintained as batch cultures for 6–9 h, after which continuous medium feed was started while the cells were still growing exponentially. Chemostat cultures were maintained at D=0.10±0.02 h−1, pH 5.0, and 30°C, with 1.5 volume gas [volume culture]−1 min−1 (vvm). For cultures that received <20.9% O2 in the gas stream, O2 was replaced with the equivalent volume of N2, so that total gas flow was kept constant for all experiments. The kLa (overall oxygen transfer coefficient) for the bioreactor under these conditions was 0.035–0.039 s−1 (in pure water), and the solubility of oxygen under these conditions is given in Table 1, to facilitate comparison with other published data. The dissolved oxygen tension in cultures receiving 20.9% O2 was 83%, but it was 0% in cultures receiving 2.8% O2 or less. Steady-state samples were taken after the cultures had been in constant conditions for a minimum of four residence times (six generations). Steady states were assessed over four to nine residence times (six to 13 generations) for constant biomass production, carbon dioxide evolution and oxygen uptake rates (CER and OUR), alkali utilization, and extracellular metabolites, as well as constant intracellular metabolites and gene transcription.
Biomass concentration, yield on glucose (moles of carbon, Cmol, in biomass per mole of carbon in substrate) and specific rates for oxygen uptake (OUR), glucose consumption, carbon dioxide evolution (CER), and ethanol and glycerol production in steady-state glucose-limited chemostat cultures (D=0.10 h−1, pH 5.0, 30°C, 1.5 vvm gas flow) of Saccharomyces cerevisiae CEN.PK113-1A
O2 provided (%)
O2 solubility (μM)
Biomass (g L−1)
Yield (x/C) (Cmol Cmol−1)
Specific OUR [mmol (g DW)−1 h−1]
Specific CER [mmol (g DW)−1 h−1]
Specific glucose consumption rate [Cmmol (g DW)−1 h−1]
Specific ethanol production rate [Cmmol (g DW)−1 h−1]
Specific glycerol production rate [Cmmol (g DW)−1 h−1]
5.0 ± 0.05
0.60 ± 0.01
2.7 ± 0.04
2.6 ± 0.03
6.6 ± 0.5
4.8 ± 0.05
0.56 ± 0.01
2.5 ± 0.04
3.0 ± 0.03
8.0 ± 0.3
0.2 ± 0.01
3.0 ± 0.03
0.36 ± 0.01
1.7 ± 0.02
3.7 ± 0.04
11.4 ± 0.5
3.2 ± 0.2
2.1 ± 0.02
0.27 ± 0.01
1.2 ± 0.02
4.6 ± 0.06
14.3 ± 1.1
5.5 ± 0.5
1.0 ± 0.02
0.12 ± 0.03
11.3 ± 0.30
37.1 ± 3.0
16.7 ± 1.6
3.0 ± 0.3
Values are mean ± SEM (n=33–88 for biomass and yield, 658–4199 for OUR and CER, 11–23 for ethanol and glycerol, and 13–24 for glucose, from steady states during two or four cultivations, using compound SEM for specific rates). Cmmol, mmole carbon.
* Solubility of O2 in pure water, provided for comparison with, for example, Lai (2005).
† ND indicates not detectable, i.e. <0.04 g glycerol L−1 and <0.03–0.06 Cmmol (g DW)−1 h−1, depending on biomass concentration.
Cultures that were fed 2.8% or 20.9% O2 were subject to oscillations. To prevent these, c. 5% of the total cell concentration in the bioreactor was added to the culture as cells in mid-exponential to late exponential phase at the time when continuous medium feed was started (Zamamiri et al., 2001).
The gas concentration (CO2, 13CO2, O2, N2, and Ar) was analyzed continuously in an Omnistar quadrupole mass spectrometer (Balzers AG, Liechtenstein), calibrated with 3% CO2 in Ar. Washout kinetics were determined as described by Esener (1981).
Biomass was measured as OD600 nm and as cell dry weight (DW). For DW determination, cells were collected by centrifugation and washed with one to two sample volumes of distilled water. Duplicate (5 mL) or triplicate (2 mL) samples were taken for all DW measurements. Cells were dried to a constant weight at 100°C.
Metabolite and chemical analyses
Extracellular metabolites (ethanol, glycerol, pyruvate, and acetate) and glucose were analyzed by HPLC on an Aminex HPX-87 H column (BioRad Laboratories, Hercules, CA) with 2.5 mM H2SO4 as eluant and a flow rate of 0.5 mL min−1. The column was maintained at 55°C. Peaks were detected using a Waters 410 differential refractometer and a Waters 2487 dual-wavelength UV (210 nm) detector. Glucose concentrations were also sometimes determined enzymatically using the Roche (Germany) Glucose GOD-PAP measurement kit (Cat. no. 1448676 216).
Intracellular metabolites were extracted from cells that had been transferred to 60% v/v methanol at −40°C immediately after their removal from the bioreactor, collected by centrifugation at 2000 g at −19°C for 5 min, washed once with 60% v/v cold (−40°C) methanol at −19°C (de Koning & van Dam, 1992), frozen in liquid N2, and stored at −80°C. Metabolites were subsequently extracted in boiling ethanol (Gonzalez et al., 1997), and analyzed by liquid chromatography (LC)-MS/MS (van Dam et al., 2002) using a Waters HT-Alliance HPLC coupled with a Micromass Quattro Micro triple quadrupole mass spectrometer. Internal standards were derived from a [13C]glucose fed-batch culture, as described by Wu (2005).
Adenosine nucleotides (ATP, ADP, and AMP) were separated and quantified using ion-pairing LC–electrospray ionization (ESI)–MS with diisopropyl amine (DIPA) as the ion-pairing reagent. HPLC was carried out in an Agilent 1100 (Santa Clara, CA) with an Xterra MS C18 (1 × 150 mm) column (Waters, Milford MA). Mobile phases were: (1) 60 mM DIPA (pH 7) (solution A), and (2) methanol/DIPA (6 mM) 80 : 20 (pH 7) (solution B), the pH being adjusted with formic acid. Elution was carried out with 5% (v/v) solution B at 80 μL min−1, for a 5-μL injection volume. Isopropanol (40 μL min−1) was then added to enhance the ionization efficiency in the mass spectrometer. The nucleotides were detected by ESI-MS (positive ionization mode) with a Micromass Quattro II triple quadrupole mass spectrometer (UK). The adduct ions formed by DIPA with nucleotides in ESI-MS were used for quantification: single ion monitoring quantification ions were 551 m/z for AMP (AMP+2 DIPA+H+), 732 m/z for ADP (ADP+2 DIPA+H+), and 812 m/z for ATP (ATP+2 DIPA+H+).
Transcriptional analysis was performed with the TRanscript analysis with aid of Affinity Capture (TRAC) assay described by Rautio (2006a) for 69 genes involved in central carbon and related metabolism. mRNA was extracted from 10-mL samples (10–50 mg DW) that had been rapidly frozen in liquid N2 and stored at −80°C. GeneScan-120LIZ size standard (Applied Biosystems, Foster City, CA) was added to each sample to calibrate the separation of the detection probes by size. In addition, in vitro synthesized mRNA (MEGAscript transcription kit; Ambion, Austin, TX) of the Escherichia coli traT gene was also added to each sample [1.5 fmol (100 μL)−1] so that the results for each probe in any analysis could be correlated with this internal standard, eliminating experimental variation in different hybridizations and samples. Probes were divided into seven probe pools with eight to 11 probes per pool. The identity of the probes was determined by the migration behavior, and the quantity by the peak area.
Total polyA RNA was quantified from the cell lysate after elution of polyA RNA in dimethyl pyrocarbonate-treated H2O, using the RiboGreen RNA quantification kit (Molecular Probes, Leiden, the Netherlands). mRNA expression levels are given as the standardized (using traT internal standard) amount per total polyA RNA.
Data are given as mean±SEM. Where appropriate, values were compared by anova, and significant differences were determined using Fisher's multiple range test.
Results and discussion
Steady-state metabolite and transcript levels in aerobic (20.9% O2) and anaerobic cultures
The physiology of S. cerevisiae CEN.PK113-1A was strongly affected by changes in oxygen concentration, as expected. The metabolite concentrations observed during aerobic steady states (Fig. 1) were similar to those reported by Mashego (2005), Wu (2005) and Visser (2004) for S. cerevisiae CEN.PK113-7D (the MATa strain equivalent to CEN.PK113-1A) in glucose-limited chemostats at D=0.05 h−1 in the presence of 31 mM ethanol. Metabolites of upper glycolysis [glucose 6-phosphate (G6P), fructose 6-phosphate (F6P), and fructose 1,6-bisphosphate (F1,6BP)] were maintained at significantly higher concentrations in anaerobic than in aerobic conditions, as was pyruvate (Fig. 1). 2-Phosphoglycerate (2PG)+3-phosphoglycerate (3PG) and phosphoenolpyruvate (PEP) concentrations were significantly lower in anaerobic than in aerobic conditions (Fig. 1). A similar increase in the concentrations of hexose phosphates and reduction in the concentrations of 2PG+3PG and PEP is observed during respirofermentative growth following addition of a pulse of glucose (Visser et al., 2004; Wu et al., 2005), when the low concentrations of 2PG+3PG and PEP are thought to reflect a dynamic response to F1,6BP regulation of pyruvate kinase.
Metabolite concentrations in steady-state glucose-limited chemostat cultures (D=0.10 h−1, pH 5.0, 30°C, 1.5 vvm gas flow) of Saccharomyces cerevisiae CEN.PK113-1A. Error bars indicate ±SEM for seven to 24 samples taken during steady states in two (0.5% and 2.8% O2) or four (0%, 1.0% and 20.9% O2) cultivations. Data points with the same letter (a–e) did not differ significantly (P>0.05, Fisher's multiple range test) from data points for the same metabolite showing the same letter. M6P, maltose 6-phosphate; T6P, trehalose 6-phosphate.
Tricaboxylic acid cycle (TCA) metabolites (citrate, succinate, fumarate, and malate) concentrations were also significantly higher in anaerobic than in aerobic conditions (Fig. 1), as observed by Villas-Bôas (2005a) in anaerobic, as compared to aerobic, batch cultures of CEN.PK113-7D. The high concentrations of TCA cycle metabolites in anaerobic conditions are maintained while there is low flux through the pathway (Nissen et al., 1997; Franzén, 2003) and low or zero oxoglutarate dehydrogenase, isocitrate dehydrogenase (Machado et al., 1975) and succinate dehydrogenase (Camarasa et al., 2003) activities. TCA cycle metabolites provide precursors for amino acid biosynthesis during anaerobic growth, but it is not clear why relatively high concentrations of these metabolites are maintained during anaerobic growth.
Individual pentose phosphates were not separated by the LC-MS/MS method used here, but together showed higher concentrations under anaerobic than in any aerobic condition (Fig. 1).
The adenylate energy charge [AEC, calculated as ([ATP]+1/2[ADP])/([ATP]+[ADP]+[AMP])] was the same (0.83±0.01; P>0.05) under anaerobic and aerobic conditions, and comparable to values observed in both anaerobic and aerobic batch (0.8–0.9; Ball & Atkinson, 1975) and aerobic chemostat (0.8–0.9; D=0.05 h−1; Mashego et al., 2005) cultures of S. cerevisiae.
Using TRAC analysis, we found that all except six (HAP3, HAP4, HXK1, MDH2, PYK2, and URA1) of the 69 genes related to central carbon metabolism that were considered here showed significant differences (P<0.05) in expression in aerobic and anaerobic cultures (Fig. 2). Only ADH1, COX5b, ACS1 and PYC1 were more highly expressed in anaerobic than in aerobic (20.9% O2) steady states (Fig. 2), whereas the other 59 genes showed lower expression in the anaerobic conditions. ter Linde (1999) identified 219 genes that showed higher transcription in aerobic glucose-limited chemostat cultures of S. cerevisiae CEN.PK113-7D than in anaerobic cultures, and 140 genes that had lower expression in the aerobic glucose-limited chemostat cultures. However, of the 69 genes considered here, ter Linde (1999) observed significant differences in expression between aerobic and anaerobic glucose-limited chemostat cultures for only 26, of which five (ADH1, CIT2, COX5b, MAE1, and GPP1) showed higher expression under anaerobic than under aerobic conditions. Thus, both studies identified ADH1 and COX5b as genes more strongly expressed in anaerobic than aerobic conditions. We did find that both GPP1 and MAE1 were more highly expressed in anaerobic conditions than with 0.5% or 1.0% O2, although they had higher expression in fully aerobic conditions. The relatively high expression of PYC1, which was not identified by ter Linde (1999) as an anaerobically upregulated gene (less than twofold induction), that we observed in anaerobic conditions presumably reflects its role in providing C4 intermediates for growth (Brewster et al., 1994). None of these anaerobically upregulated genes was mentioned in the study published by Piper (2002), which did not list anaerobic genes with less than an eightfold increase in expression, and would thus not include this subset. Low expression of ADH2, CIT3 and LSC1 in anaerobic, as compared to aerobic, conditions was observed by both Piper (2002) and ter Linde (1999), as well as in this study (Fig. 2), and the low expression of most genes involved in the TCA cycle in anaerobic conditions was also observed by Piper (2002).
Relative mRNA levels for genes involved in central carbon metabolism, and some genes involved in transcription, redox regulation and stress responses in Saccharomyces cerevisiae CEN.PK113-1A in steady-state, glucose-limited chemostat cultures (D=0.10 h−1, pH 5.0, 30°C, 1.5 vvm gas flow) with various concentrations of oxygen provision. Error bars indicate ± SEM for four to eight samples taken during steady states in duplicate cultivations. Lines indicate (adjacent) data points that did (lines with positive or negative slope; P<0.05, Fisher's multiple range test) or did not (lines with slope=0; P>0.05) differ significantly from each other.
Biomass concentration and specific consumption or production of glucose, oxygen, ethanol and glycerol during steady-state growth of S. cerevisiae CEN.PK113-1A in glucose-limited culture are given in Table 1.
Respirofermentative growth in low oxygen
Provision of only 0.5% O2 resulted in significantly lower concentrations of the metabolites of upper glycolysis, such as G6P, F6P, and F1,6BP, in comparison with anaerobic cultures, whereas the concentrations of the lower glycolysis metabolites 2PG+3PG and PEP were increased (Fig. 1). The concentrations of TCA cycle metabolites such as citrate, succinate, malate and fumarate decreased as compared to anaerobic cultures (Fig. 1). Glycolytic gene transcription, in contrast to metabolite concentration, was largely unaffected (relative to anaerobic cultures) by provision of up to 1% O2, whereas most of the TCA cycle genes were upregulated with provision of only 0.5% O2 (Fig. 2).
Reduction of the oxygen input from 20.9% to 2.8% was sufficient to allow net ethanol production, even though there was sufficient oxygen to maintain a high yield of biomass on glucose and a high OUR (Table 1). The high level of expression of ADH1 did not result in substantial ethanol production, but ADH2 was strongly repressed, so it seemed unlikely that substantial simultaneous consumption was occurring (Fig. 2). Approximately 50% of the 69 genes considered here showed either significantly higher (23%) or lower (31%) expression with 2.8% O2, as compared to 20.9% O2 (Fig. 2). All glycolytic and TCA cycle genes analyzed and most genes of the pentose phosphate pathway (PPP) showed higher expression levels with 2.8% O2, as compared to more anaerobic conditions. However, the metabolite pools of the cells grown in 2.8% O2 were more similar to those of cells grown in 0.5% or 1.0% O2 than that of cells grown in 20.9% O2 (Fig. 1). Cells growing in sufficient but low oxygen concentration were clearly very different from cells growing in abundant oxygen.
The difference in the metabolic condition of cells growing with 2.8% O2 as compared to those grown with 20.9% O2 was also apparent in the low AEC, which was only 0.54±0.03 when 2.8% O2 was supplied (D=0.10 h−1, pH 5.0). Unlike E. coli, which requires an AEC of at least 0.8 for growth (Chapman et al., 1971), S. cerevisiae remains able to grow in conditions supporting an AEC as low as 0.4 (Polakis & Bartley, 1966).
Shifting from aerobic conditions to anaerobic conditions
When conditions were switched from aerobic (0.5–20.8% O2) to anaerobic (0% O2), ethanol and glycerol concentrations in the culture started to increase within an hour of the shift, requiring c. 36 h to reach the steady-state values observed in anaerobic cultures (Fig. 3). Biomass concentrations also began to decrease almost immediately (Fig. 3). Although biomass concentration decreased after the switch to anaerobic conditions, cells continued to grow at c. 0.06 h−1 during the washout.
Biomass, ethanol and glycerol concentrations during glucose-limited chemostat cultures of Saccharomyces cerevisiae CEN.PK113-1A following a shift from aerobic (0.5%, 1.0%, 2.8% or 20.9% O2) to anaerobic conditions. Cultures were maintained at D=0.10 h−1, pH 5.0, 30°C, and 1.5 vvm gas flow throughout the culture, with N2 replacing air in the gas to maintain a constant gas flow. Error bars for biomass indicate ±SEM (n=3–5).
On the basis of the steady-state concentrations, most glycolytic and TCA cycle metabolites were expected to increase to reach steady-state anaerobic concentrations, following a switch to anaerobic conditions, as was observed (Figs 4 and 5). An initial increase in metabolite concentration occurred within <10 min, except for citrate, but new steady-state concentrations generally required at least 30 h (i.e. three retention times or four to five generations) to become established (Fig. 5). 2PG+3PG and PEP of lower glycolysis both decreased in concentration within 10 min of the switch from aerobic to anaerobic conditions (Fig. 5). The gradual adjustment of most metabolites to the anaerobic steady-state concentration was generally independent of the initial oxygen concentration provided, reflecting the fact that metabolite concentrations generally differed significantly between anaerobic conditions and any level of oxygen provision (Fig. 1).
Relative changes in the concentration of intracellular metabolites during glucose-limited chemostat cultures of Saccharomyces cerevisiae CEN.PK113-1A, following a shift from aerobic (0.5%, 1.0%, 2.8% or 20.9% O2) to anaerobic conditions, with time increasing from left to right. For each metabolite, the bars indicate relative concentrations in cultures that initially received oxygen as follows: top two bars 0.5%, the next two bars 1.0%, the next bar 2.8%, and the bottom two bars 20.9%. Yellow indicates no change in concentration. Red indicates increasing concentrations of the metabolite (from light red indicating a 1.5–2-fold increase, to bright red indicating a >10-fold increase), and green indicates reduced concentrations (from light green indicating 1.4–2-fold lower concentration, to bright green indicating >sevenfold lower). Cultures were maintained at D=0.10 h−1, pH 5.0, 30°C, and 1.5 vvm gas flow throughout the culture, with N2 replacing air in the gas to maintain a constant gas flow.
Metabolite concentrations in Saccharomyces cerevisiae CEN.PK113-1A after a shift from growth in 20.9% O2 to 0% O2 in glucose-limited chemostat cultures (D=0.10 h−1, pH 5.0, 30°C, 1.5 vvm gas flow). (a) Glycolytic metabolites that increase. (b) TCA cycle metabolites that increase. (c) Metabolites that decrease. (d) Glycolytic metabolites, G6P green, F1,6BP blue, F6P black, pyruvate red. Data from two cultures are shown (connecting lines indicate data from one culture, and separate data points data from the other culture).
For several of the genes considered here, the shift to anaerobic conditions had no effect on transcription during the first hour (e.g. SDH1, ADR1, PDA1, and PDB1), and in some cases even for 8 h (e.g. HAP1 and KGD1), after the change (Fig. 6), being comparable to the ‘chronic’ response to anaerobiosis described by Lai (2005) for batch cultures of S. cerevisiae JM43. We also observed that several genes involved in carbohydrate utilization and reserve energy metabolism (e.g. GLK1, HXK1, PGI1, and TPS1) showed weak transient responses to the anaerobic shift in the glucose-limited chemostats. These genes showed transient upregulation in response to anaerobiosis in batch cultures with galactose (but not glucose) as the carbon source (Lai et al., 2005), leading to the suggestion that these genes could be regulated by oxygen provision when glucose was not repressing this. Our data support this suggestion while confirming the importance of balancing the energy supply during the transition to fermentative growth.
Relative changes in the level of mRNA expression for genes involved in central carbon metabolism, transcription or respiration in glucose-limited chemostat cultures of Saccharomyces cerevisiae CEN.PK113-1A following a shift from aerobic (0.5%, 1.0%, 2.8% or 20.9% O2) to anaerobic conditions, with time increasing from left to right. For each gene, the top two bars indicate relative transcription in cultures that initially received 0.5% O2, the next two bars cultures that initially received 1.0% O2, the next bar a culture that initially received 2.8% O2, and the bottom two bars cultures that initially received 20.9% O2. Yellow indicates no change in expression. Red indicates increasing levels of expression (from light red indicating a 1.5–2-fold increase, to bright red indicating a >10-fold increase), and green indicates decreased expression (from light green indicating 1.4–2-fold lower expression, to bright green indicating >sevenfold lower expression), as illustrated in Fig. 4. Cultures were maintained at D=0.10 h−1, pH 5.0, 30°C, and 1.5 vvm gas flow throughout the culture, with N2 replacing air in the gas to maintain a constant gas flow.
The three genes that showed higher expression in anaerobic than aerobic chemostat cultures (COX5b, ADH1, and PYC1; Fig. 2) also showed stable upregulation following a transition from 20.9% O2 to anaerobic conditions (Figs 6 and 7). However, many other genes showed transient upregulation (Figs 6 and 7), typically during the first 8 h (approximately one generation) after the change. Genes with higher expression in 2.8% than in 20.9% O2 (e.g. GPD2, SDH2, CYC1, PDA1, and HAP1) did not exhibit transient upregulation following the shift to anaerobic conditions, indicating that cells did not pass through a physiological condition corresponding to that observed in the steady state with 2.8% O2 when adjusting to anaerobiosis. Transient downregulation was also observed (Figs 6 and 7), with rapid downregulation, but more gradual subsequent recovery. The duration of transient responses was affected by the initial oxygen concentration before the change (Fig. 6).
Changes in relative gene expression in Saccharomyces cerevisiae CEN.PK113-1A after a shift from growth in 20.9% O2 to growth in 0% O2 during glucose-limited chemostat cultures (D=0.10 h−1, pH 5.0, 30°C, 1.5 vvm gas flow), showing: (a) upregulation, (b) downregulation, (c) transient upregulation, and (d) transient downregulation. Insets for (c) and (d) show extended (120 h) time frames. Data from two cultures are shown (connecting lines indicate data from one culture, and separate data points data from the other culture).
The effect of oxygen provision on glycolysis and the PPP
Although glycolytic genes were generally more highly expressed with 2.8% or 20.9% O2 than with less or no O2 (Fig. 2), these genes generally showed no downregulation following a shift from aerobic to anaerobic conditions until at least 24 h after the shift. GPM1 and PYK1, in fact, were initially upregulated following the shift (Fig. 6). In contrast, most genes involved in the TCA cycle were downregulated, as expected, but generally only after 2 or 3 h in anaerobic conditions. Downregulation of PPP genes also occurred (Fig. 6).
As the amount of oxygen available for cellular metabolism is reduced, the relative flux through the PPP diminishes (cf. Gombert et al., 2001; Fiaux et al., 2003; Franzén, 2003; van Winden et al., 2005), and the flux through glycolysis increases. However, mRNA transcripts of the genes involved in glycolysis and those involved in the PPP are maintained at higher levels in conditions in which oxygen is readily available than in conditions with little or no oxygen (Fig. 2). Thus, although the slight increase in transcription of glycolytic genes following a shift from aerobic to anaerobic conditions may reflect the cells' initial means of increasing the flux through the pathway and increasing energy provision through ethanol production, it can only be transitory, as transcription levels need to be reduced to achieve the anaerobic steady state. The downregulation of PPP genes presumably contributes to the shift in flux towards glycolysis, but much of the control for the glycolytic flux must occur posttranscriptionally, as is also implied by the sharp changes in glycolytic metabolite concentrations, particularly those of lower glycolysis. Daran-Lapujade (2004) also concluded that regulation of glycolytic genes for CEN.PK113-7D grown aerobically on different carbon sources was primarily posttranscriptional, whereas there was evidence that both transcriptional and posttranscriptional regulation occurred for genes of the PPP. This conclusion is further supported by the observation that glycolytic enzyme activity is higher under anaerobic conditions than under aerobic conditions in glucose-limited chemostat cultures of S. cerevisiae CEN.PK113-7D (van Hoek et al., 2000).
It is also interesting to note that the expression of GND1, TKL1, and ALD6, which are all subject to Stb5p induction (Larochelle et al., 2006), showed similar responses (transient downregulation) to a change from aerobic (20.9%, 1.0% or 0.5% O2) to anaerobic conditions, whereas other genes under Stb5p regulation (GND2, TAL1, ALD4, GOR1, and PGI1) were either unaffected by the change (PGI1, ZWF1 and TAL1 in 20.9% O2) or downregulated (ALD4, GOR1 and TAL1 in 0.5% and 1.0% O2).
Induction and repression of oxygen-dependent genes
Several genes (e.g. COX5a, COX5b, CYC1, CYC7, and ROX1) were included in this study because their expression was known or expected to be strongly influenced by extracellular oxygen. However, expression of these genes did not necessarily respond to oxygen concentration as predicted on the basis of previous reports.
COX5b and CYC7 have been highlighted as anaerobic genes that are only expressed when oxygen solubility is below 0.5 μM (Burke et al., 1997). COX5b was indeed only detected in anaerobic cultures (0.5% O2, resulting in <6 μM solubility, should be sufficient to repress expression); however, expression levels were very low, and reduced expression would not have been detectable even if some expression still occurred. Others in our laboratory have observed higher expression levels for COX5b in an industrial brewer's yeast, even when oxygen was available (J.J. Rautio, pers. commun.), which suggests that expression of this gene is subject to strain variation, with CEN.PK113-1A not being the optimal strain in which to study its response to oxygen, because of its low anaerobic expression. Furthermore, CYC7 was more highly expressed in aerobic than in anaerobic conditions (Fig. 2), rather than being repressed by the presence of oxygen, as in batch cultures of S. cerevisiae JM43 (Burke et al., 1997). Similarly, ter Linde (1999) did not observe induction of CYC7 expression in CEN.PK113-7D in anaerobic conditions, and nor was CYC7 subject to transient induction following the shift from aerobic to anaerobic conditions (Fig. 6). This difference in CYC7 expression may reflect differences in gene regulation between batch (Burke et al., 1997) and chemostat cultures, but may also again indicate strain differences. Such variation, however, has implications for the selection of biomarkers for assessment of industrial processes.
CYC1 and COX5a were more highly expressed in aerobic (2.8% and 20.9% O2) than in low (0.5% or 1.0%) or no O2, as expected (Burke et al., 1997; ter Linde et al., 1999), and downregulation was observed following a shift from aerobic to anaerobic conditions. Expression was reduced in 1.0% (COX5a) or 0.5% (CYC1) O2, providing soluble oxygen at a concentration (6–12 μM) comparable to that in which reduced expression of these genes was observed in batch cultures (Burke et al., 1997). The regulatory gene ROX1 only had high expression in 20.9% O2. However, although genes such as CYC7 and COX5b are under ROX1 regulation, a high level of expression of CYC7 was observed in 20.9% O2 when ROX1 expression was also high, and COX5b remained undetectable in 0.5–2.8% O2, even though ROX1 expression was low in these conditions.
Another pair of genes that have been described as aerobic or nonaerobic are the acetyl coenzyme A synthetase genes ACS1 and ACS2 (van den Berg et al., 1996). In this case, the function of the enzymes encoded by the genes is linked to anaerobic conditions, rather than the expression, with ACS2, the ‘nonaerobic’ form, being expressed both aerobically and anaerobically, with higher anaerobic expression than ACS1 (van den Berg et al., 1996). We found that ACS1 was not highly expressed in CEN.PK113-1A, and ACS2 expression was always higher than that of ACS1 (Fig. 2). ACS2 expression was lower in cultures receiving 0%, 0.5% or 1.0% O2 than in cultures receiving 2.8% or 20.9% O2 (Fig. 2), and was downregulated when conditions were changed from aerobic to anaerobic (Fig. 6). Thus, if ACS activity was reduced in anaerobic as compared to aerobic glucose-limited chemostats in CDN.PK113-1A, as has been observed in strain T2-3D (van den Berg et al., 1996), this would be explained by the reduction in ACS2 expression, rather than a reduction in ACS1 expression.
All four of the dehydrogenases located in the mitochondrial intermembrane space (NDE1, NDE2, GUT2, and CYB2) that are considered here had low expression under anaerobic conditions and were downregulated following a shift from aerobic to anaerobic conditions.
Transcriptional regulation of ethanol and byproduct formation
Ethanol production occurred in conditions of low (0.5–2.8%) and no O2, as expected (Fig. 3;Table 1). However, ADH1 expression did not simply increase with decreasing oxygen provision, as noted above, and upregulation of ADH1 following a switch to anaerobic conditions was only seen when the initial oxygen concentration was high (20.9%) or very low (0.5%).
Genes involved in ethanol consumption were consistently downregulated, as were genes involved in respiration (ROX1, COX5a, CYB2, and CYC1) and some genes involved in acetate metabolism (ALD4, ALD6, ACS2, and ACH1). Low expression of genes involved in ethanol consumption (ADH2 and ALD4) has also been observed during respirofermentative growth at high specific growth rates in comparison with respiratory growth at lower specific growth rates (Regenberg et al., 2006). Regulation of genes involved in ethanol consumption (and respiration) following a shift to anaerobic conditions followed the pattern suggested for HAP4-regulated genes (Raghevendran et al., 2006), even though HAP4 itself was not transcriptionally regulated with change in oxygen provision (Figs 2 and 6).
No extracellular acetate was observed under any condition. Glycerol was produced only in completely anaerobic conditions (Table 1;Fig. 3), and was not produced in proportion with ethanol production, indicating that 0.5% O2 still provided sufficient oxygen to avoid cytoplasmic NADH accumulation. GPD1 and GPP1 showed transient upregulation following the shift to anaerobic conditions, but expression of GPD2 was reduced (Fig. 6), even though its expression increases following a shift to anaerobic conditions in batch cultures and it is the primary glycerol-3-phosphate dehydrogenase involved in redox balancing in respiratory-deficient mutants (Ansell et al., 1997; Valadi et al., 2004). Both GPD1 and GPD2 had lower expression in anaerobic than in aerobic conditions (Figs 2 and 6).
With the development of cost-efficient, sensitive methods for analysis of transcripts and metabolites, bioprocess monitoring can be extended to include monitoring of process-specific biomarkers from these groups of compounds. Optimally, such biomarkers would be used to identify when a culture was shifting towards an unproductive physiology, in order to facilitate intervention to correct the problem. Identification of suitable biomarkers is thus important for the full exploitation of physiologic data.
From the steady-state data, it is easy to identify gene transcripts whose expression level is unique to specific environmental conditions, with the clearest marker genes indicating only the presence of too much (for anaerobic or low-air cultures) or too little oxygen (for high-air or low-air cultures). In particular, genes such as COX5b, TAL1, YGR043C, ACH1, CIT3 and ADH2 would appear to be suitable for distinguishing cultures experiencing higher or lower oxygen supply than expected. This may be most relevant in cultures receiving limiting concentrations of oxygen, such that fluctuations would not be seen in the dissolved oxygen tension, which is always zero, or when assessing the extent to which oxygen-poor regions may exist in a large bioreactor. However, in either situation, the conditions of low oxygen are likely to be experienced only periodically (over intervals of a few minutes in poorly mixed bioreactors, to several hours when oxygen provision is variable), and it is necessary to consider how these genes respond to changing conditions. Thus, TAL1 and YGR043C, whose expression showed transient upregulation and downregulation following a shift to anaerobic conditions, would not be suitable, as a high expression level could be indicative either of a high oxygen supply, or of a low oxygen supply or poor oxygen mixing. ADH2 and COX5b, on the other hand, showed consistent downregulation and upregulation, respectively, in response to the switch to anaerobiosis, and would give more reliable information. In addition, the fact that many genes show transient responses to change means that it would be difficult to identify genes that would serve as markers for a range of conditions (the actual concentration of available oxygen) rather than just the contrast between two (too much or too little). GUT2 and GOR1, with decreasing expression with decreasing oxygen concentration below 2.8% O2 and consistent downregulation following a shift to anaerobic conditions, might be suitable biomarkers. Genes that did show transient upregulation or downregulation (Fig. 7) in response to the shift to anaerobic conditions, but did not show large expression differences in steady states, may be suitable candidates as markers for poor mixing and localized variation in conditions within the bioreactor (e.g. GPM1 and NDE2). High (or low) expression levels would indicate that cells were experiencing periodic transient conditions. As the recovery is somewhat quicker (Fig. 7), upregulated genes (e.g. PYK1) would appear to be better biomarkers for heterogeneous conditions than downregulated genes, but it should be noted that none of the genes considered here would be expected to serve as an effective biomarker for transients of duration <10 min.
The same considerations apply to the choice of a metabolite as a biomarker, but the choice should be wider, as transient changes were less frequently observed (Fig. 4). F1,6BP and succinate appear to be the best biomarkers for distinguishing between cells experiencing low and no air. Pyruvate may be a marker for intermediate conditions, although concentration changes following the shift to anaerobic conditions were dependent on the initial oxygen concentration. From this work, it is not clear how metabolite concentrations are affected in cells experiencing periodic exposure to varying oxygen concentrations. It is also apparent from this work that the current methodology for transcript analysis (TRAC as shown here, but also reverse transcriptase-PCR) is more robust and more readily available than metabolite analyses, making transcripts more likely to be successful biomarkers than metabolites, despite the problems inherent in transient regulation. Metabolite analyses still suffer from questions related to quenching, leaking of cell contents during quenching, extraction method, and separation methods (Villas-Bôas et al., 2005b). Metabolic fingerprinting, either by MS or by nuclear magnetic resonance, may offer better prospects for bioprocess monitoring than quantitative metabolite measurements.
The results presented here show that, although S. cerevisiae is able to respond rapidly to changes in oxygen provision, considerable time (approximately four to five generations, i.e. 28–35 h) is required for cellular metabolism to fully adapt to anaerobic conditions, even when cells have been growing in the presence of only low amounts of air. Thus, when interpreting experiments based on short-term analysis of the transcriptome (or metabolome) following a shift in external conditions, such as a transition from aerobic to anaerobic conditions in batch culture (Lai et al., 2005), it is worth noting that physiological adaptation is likely to occur throughout the duration of the experiment. When fewer than four or five generations are monitored, the data cannot provide a direct comparison of the transcriptome in the conditions before and after the shift (e.g. in aerobic and anaerobic conditions), but rather provides an indication of the ongoing cellular response to the shift, relative to the stable condition prior to it. It is also worth noting that S. cerevisiae appears to require more time to adjust to a change in conditions than does the filamentous fungus Trichoderma reesei, which is able to attain transcriptional steady state within less than one generation (Rautio et al., 2006b). The adaptation to anaerobic conditions (four to five generations) was, however, more rapid than the time needed for S. cerevisiae to achieve transcriptional steady state following a shift from growth on galactose to growth on glucose (>10 generations; >43 h at D=0.16 h−1; Braun & Brenner, 2004). On the other hand, as the response to new conditions requires several generations, S. cerevisiae is likely to be very robust in conditions of inadequate mixing. Indeed Ronen & Botstein (2006) observed that steady-state glucose-limited cells exposed to pulses of excess glucose required only one generation (c. 3.5 h at D=0.2 h−1) or less to regain transcriptional steady state. Further work should consider the effects of periodic exposure to altered oxygen provision, to determine how quickly cells recover from short-term oxygen shortage. Whereas metabolite concentrations responded more rapidly to removal of oxygen from the culture, metabolite concentrations also continued to change for up to five generations (35 h) following the change (Fig. 5d).
This work demonstrates how provision of limiting concentrations of oxygen to S. cerevisiae CEN.PK113-1A resulted in distinctly different cellular physiology, measured in terms of metabolite pools and gene transcription. In particular, 2.8% O2 provided only limited oxygen (dissolved oxygen tension=0%), but was sufficient for high biomass production. However, the levels of gene transcription in the presence of 2.8% O2 were often higher than would have been predicted on the basis of data from aerobic and anaerobic cultures, whereas gene transcription in cultures receiving less O2 (0.5% or 1.0%) was more predictable. This is also reflected in the fact that, for many genes involved in central carbon metabolism, up to 1.0% O2 could be supplied to the culture with no effect as compared to transcription levels in anaerobic conditions. This information will be useful in the design of cultures requiring low oxygen input.
The role of oxygen in regulating cellular physiology is complex, as is also indicated by the extensive research on the topic (cf. Lascaris et al., 2004; Tai et al., 2005; Lai et al., 2006) and the diversity of approaches used. It will be important to further identify those responses that are truly oxygen dependent and those that are strain, medium or culture system specific.
We thank Outi Könönen, Pirjo Tähtinen, Eila Leino and Tarja Hakkarainen for technical assistance. We thank Tekes, the Finnish Funding Agency for Technology and Innovation (Project numbers 40135/04 and 40537/05) and the Academy of Finland (Project number 202409, and SYSBIO program, project number 207435) for financial support.
(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.
(2005) Dynamical remodeling of the transcriptome during short-term anaerobiosis in Saccharomyces cerevisiae: differential response and role of Msn2 and/or Msn4 and other factors in galactose and glucose media. Mol Cell Biol 25: 4075–4091.
(2004) Distinct intracellular localization of Gpd1p and Gpd2p, the two yeast isoforms of NAD+-dependent glycerol-3-phosphate dehydrogenase, explains their different contributions to redox-driven glycerol production. J Biol Chem 279: 39677–39685.
(2002) Analysis of glycolytic intermediates in Saccharomyces cerevisiae using anion exchange chromatography and electrospray ionization with tandem mass spectrometric detection. Anal Chim Acta 460: 209–218.