Intracellular characterization of aerobic glucose metabolism in seven yeast species by 13C flux analysis and metabolomics


  • Editor: Jens Nielsen

Correspondence: Uwe Sauer, Institute of Molecular Systems Biology, ETH Zurich, Wolfgang-Pauli-Strasse 16, 8093 Zurich, Switzerland. Tel.: +414 4633 3672; fax: +414 4633 1051; e-mail:


Key distinguishing characteristics of yeast glucose metabolism are the relative proportions of fermentation and respiration. Crabtree-positive yeast species exhibit a respirofermentative metabolism, whereas aerobic species respire fully without secretion of fermentation byproducts. Physiological data suggest a gradual transition in different species between these two states. Here, we investigate whether this gradual transition also occurs at the intracellular level by quantifying the intracellular metabolism of Saccharomyces cerevisiae, Saccharomyces bayanus, Saccharomyces exiguus, Kluyveromyces thermotolerans, Yarrowia lipolytica, Pichia angusta and Candida rugosa by 13C-flux analysis and metabolomics. Different from the extracellular physiology, the intracellular fluxes through the tricarboxylic acid cycle fall into two classes where the aerobic species exhibit much higher respiratory fluxes at otherwise similar glycolytic fluxes. More generally, we found the intracellular metabolite concentrations to be primarily species-specific. The sole exception of a metabolite-flux correlation in a species-overarching manner was found for fructose-1,6-bisphosphate and dihydroxyacetone-phosphate, indicating a conservation of the functional properties around these two metabolites.


Although the reaction network topology of the central carbon metabolism is conserved among different yeasts, magnitude and distribution of flux through these pathways varies among them (Flores et al., 2000; Fiaux et al., 2003; Cannizzaro et al., 2004; Blank et al., 2005). This is particularly true for aerobic catabolism of glucose, the preferred carbon source for yeast and many other organisms. Aerobic glucose catabolism is classified into the groups of fermentative, respirofermentative, respiratory and obligate aerobic metabolism (van Dijken et al., 1993; Pronk et al., 1996). Respirofermentative yeasts, such as Saccharomyces cerevisiae, are characterized by the occurrence of the long-term Crabtree effect (Zimmermann & Entian, 1997); i.e. they exhibit high glucose uptake rates that are routed to ethanol under aerobic, glucose excess conditions. In contrast, respiratory and obligate respiratory yeasts, summarized as aerobic yeasts, exhibit low glucose uptake rates that are fully channelled into respiration without secretion of fermentation byproducts.

The metabolic trait of Crabtree-positive metabolism evolved to a large extent by horizontal gene transfer and whole genome duplication, through which metabolism acquired the ability for anaerobic growth, including high glycolytic and fermentation rates (Seoighe & Wolfe, 1999; Piskur & Langkjaer, 2004; Conant & Wolfe, 2007; Merico et al., 2007; van Hoek & Hogeweg, 2009). In present-day yeasts, metabolism is subject to additional levels of regulation. These include glucose repression of the tricarboxylic acid (TCA) cycle and respiration (De Deken, 1966; Gancedo, 2008) and overflow metabolism at the pyruvate decarboxylase branch point due to its high capacity and limited flux through pyruvate dehydrogenase (van Urk et al., 1989; Pronk et al., 1996). Additionally different metabolites with regulatory roles affect respiration and other cellular functions. This regulation might occur either directly, via the redox or energy state of the cell, or via glucose signalling (Muller et al., 1995; Tisi et al., 2002; Vemuri et al., 2007; Diaz-Ruiz et al., 2008; Gancedo, 2008).

Here, we ask whether the evolution of metabolism and its regulation led to a gradual transition between Crabtree-positive and aerobic steady-state metabolism in modern yeasts. Physiological studies of S. cerevisiae and other yeasts revealed a gradual transition from respiratory to fermentative metabolism with increasing glucose uptake rates (van Urk et al., 1989; Blank et al., 2005), which would suggest the existence of intermediate metabolic states. At present, it remains unclear whether the intracellular distribution of fluxes reflects such a gradual transition between respirofermentative and fully respiratory, aerobic states. Additionally, we ask whether absolute metabolite concentrations are species-specific or rather correlate with fluxes in a species-overarching manner, as one might conclude from the good flux correlations of glycolytic intermediate concentrations in S. cerevisiae under different conditions (Elbing et al., 2004; Tai et al., 2007; Bosch et al., 2008; Kresnowati et al., 2008a, b). Furthermore, it was shown qualitatively that the metabolite levels in different organisms respond similarly to environmental perturbations (Brauer et al., 2006). Whether or not such circumstantial observations are more generally true is investigated here.

To address these questions, we chose seven yeast species, including S. cerevisiae, the biotechnologically relevant Yarrowia lipolytica (Gellissen et al., 2005; Vakhlu & Kour, 2006; Abbott et al., 2009; Beopoulos et al., 2009), and the less-characterized biotechnologically emerging Candida rugosa (Dominguez de Maria et al., 2006; Lee & Park, 2009). The other four species were chosen for a gradual coverage of extracellular rates between the Crabtree-positive S. cerevisiae and the obligate aerobic Y. lipolytica (Blank et al., 2005; Merico et al., 2007). Specifically, we quantified their glucose excess metabolism in batch culture by 13C flux analysis (Sauer, 2006) and metabolomics (Roessner & Bowne, 2009).

Materials and methods

Strain medium and cultivation conditions

The investigated yeasts are listed in Table 1. All liquid cultivations were carried out in minimal medium with 10 g L−1 glucose (Blank & Sauer, 2004). After precultivation overnight in glucose minimal medium, 25–50 mL cultures were inoculated to a starting OD600 nm of about 0.05 and grown in 500-mL shake flasks at 30 °C and 250 r.p.m. Aliquots were withdrawn during the exponential growth phase on glucose. For flux analysis experiments, natural abundance glucose was replaced by either 100% of the 1-13C glucose or a mixture of 20% of the U-13C isotopologue and 80% natural abundance glucose (13C-enrichment ≥99%, Cambridge Usotope Laboratories, Andover).

Table 1.   Yeast species used in this study
S. cerevisiaeFY4 Mata – wild typeWinston et al. (1995)
S. bayanus var. uvarumWild typeCLIB 251*
S. exiguousWild typeCLIB 179*
K. thermotoleransWild typeCLIB 292*
Y. lipolyticaWild typeH222
P. angustaWild typeCLIB 421*
C. rugosaWild typeIFO 0750

Biomass and extracellular metabolite concentrations

Biomass concentrations were determined by recording OD600 nm with a spectrophotometer (Novaspec II, Pharmacia Biotech, Uppsala, Sweden). For each species, we determined mass to OD600 nm conversion factors by determining cellular dry weight from 5–10 mL filtrate with predried and preweighed membranes (0.45 μM, Sartorius, Goettingen, DE), followed by three wash steps with 4 °C ddH2O. These membranes were dried overnight at 85 °C and the weight difference was measured.

Extracellular metabolite concentrations were determined with an HPX-87C Aminex, ion-exclusion column (Biorad, Munich, Germany) as described in Heer & Sauer (2008) on an HPLC HP1100 system (Agilent Technologies, Santa Clara). The column temperature was 60 °C and a flow rate of 0.6 mL min−1 of 5 mM H2SO4 as the eluant was used.

Biomass yields were obtained from a linear fit of substrate or byproduct concentrations during exponential growth as a function of corresponding biomass concentrations. Multiplication with the growth rate then yielded specific glucose uptake and byproduct secretion rates. The physiological parameters were determined from at least two independent biological replicates.

13 C-flux analysis

For labelling experiments, at least two replicate cultures were inoculated to an OD600 nm of 0.05 or less. During sampling, 1 mL of culture was harvested during exponential growth followed by two wash steps with ddH2O and stored at −20 °C for further analysis. The processing for GC-MS analysis was performed as described previously (Zamboni et al., 2009). The pellets were hydrolyzed with 6 M HCl overnight at 105 °C and then dried at 95 °C under a constant air stream. We dissolved the hydrolysates in 30 μL of the solvent DMF (Sigma-Aldrich, Buchs, Switzerland) and added 30 μL of the derivatization agent N-(tert-butyldimethylsilyl)-N-methyl-trifluoroacetamide with 1% tert-butyldimethyl-chlorosilane (Sigma-Aldrich). Upon incubation at 85 °C for 1 h, mass isotopomer distributions of the protein-bound amino acids were determined on a 6890N GC system (Agilent Technologies) combined with a 5875 Inert XL MS system (Agilent Technologies).

The mass isotopomer distribution of the amino acid fragments was corrected for the amount of naturally occurring stable isotopes and unlabeled biomass. From the corrected mass isotopomer distribution of the amino acids, ratios of converging fluxes were calculated with the analytical equations described in Blank & Sauer (2004) and Zamboni et al. (2005). These ratios together with the extracellular fluxes and a general stoichiometric network were then used as constraints for the netto subprogram of the fiatflux software to iteratively identify an absolute flux solution that best described the data (Zamboni et al., 2005, 2009).

Intracellular metabolite concentrations

For rapid quenching of metabolism (Buscher et al., 2009; Ewald et al., 2009), a 1-mL culture aliquot was transferred to 4 mL of −40 °C 60% methanol and 10 mM ammonium acetate (pH 7.5) within 10 s. The quenching was followed by 3 min of centrifugation with a swing-out rotor at 4500 g and −9 °C (Centrifuge 5804R, Eppendorf, Germany). Pellets were stored at −80 °C until extraction. The extraction was performed at 80 °C in 75% boiling ethanol and 10 mM ammonium acetate (pH 7.5). At this step, 100 μL of fully labelled 13C-biomass was added as an internal standard (Wu et al., 2005). The extracts were dried using a vacuum centrifuge (Christ-RVC 2–33 CD plus, Kuehner AG, Birsfeld, Switzerland). The dried extracts were dissolved in 50–100 μL ddH2O before being separated by ion pairing-reverse phase liquid chromatography coupled to a ultrahigh-performance system.

We used a Waters Acquity UPLC (Waters Corporation, Milford, MA) with a Waters Acquity T3 end-capped reverse-phase column with dimensions 150 mm × 2.1 mm × 1.8 μm (Waters Corporation) for metabolite separation as described in detail in Buscher et al. (2009). The chromatography was coupled to a Thermo TSQ Quantum Ultra triple quadrupole mass spectrometer (Thermo Fisher Scientific, Waltham, MA) with a heated electrospray ionization source (Thermo Fisher Scientific) in negative mode with multiple reaction monitoring. Acquisition and peak integration was performed with an in-house software (B. Begemann & N. Zamboni, unpublished data) and the peak areas were further normalized to fully 13C-labeled internal standards and the amount of biomass.

The metabolome of all species was measured with at least four separately quenched replicates. To exclude artifacts, we performed an outlier detection on the raw data. With four data-points, the outlier detection discards the data point furthest away from the mean. With more data points, we calculated the 0.2 and 0.8 quantiles. From these data points, we determined the mean and the SD. The points being higher/lower than the average ±2 times the SD were discarded.


Physiology and growth

To verify the presence and to quantify the degree of the respirofermentative metabolism in the seven selected yeast species, we characterized their physiology by determining specific growth, uptake and secretion rates in aerobic batch cultures with 10 g L−1 glucose (Table 2). The seven species covered a wide range of glucose uptake and ethanol secretion rates. As expected, the Crabtree-positive species S. cerevisiae, Saccharomyces bayanus, Saccharomyces exiguus and Kluyveromyces thermotolerans exhibit high glucose uptake rates between 6.3 and 15.4 mmol g−1 h−1, coupled with ethanol secretion and to a minor extent with acetate and glycerol secretion. The aerobic species Pichia angusta and Y. lipolytica exhibited much lower glucose uptake rates, ranging from 2.8 to 4.4 mmol g−1 h−1, without detectable secretion of any metabolic byproducts. Consequently, the aerobic species grew more efficiently with a yield of at least 0.5 gbiomass gglucose−1, whereas the Crabtree-positive species achieved yields between 0.13 and 0.27 gbiomass gglucose−1 during exponential growth on glucose. Despite their slower overall metabolism, the aerobic species P. angusta, Y. lipolytica and C. rugosa grew faster than the Crabtree-positive species.

Table 2.   Growth parameters of the different yeast species in glucose minimal medium batch
rate (h−1)
Glucose uptake
(mmol g−1 h−1)
Glycerol secretion
(mmol g−1 h−1)
Acetate secretion
(mmol g−1 h−1)
Ethanol secretion
(mmol g−1 h−1)
Yield (g g−1)
S. cerevisiae0.35 ± 0.0115.4 ± 1.11.26 ± 0.150.69 ± 0.1622.2 ± 6.80.13 ± 0.01
S. bayanus var uvarum0.22 ± 0.048.8 ± 2.10.5 ± 0.10.04 ± 0.0212 ± 2.40.13 ± 0.03
S. exiguus0.28 ± 0.047.2 ± 0.10.3 ± 0.020.37 ± 0.0510.2 ± 10.18 ± 0.02
K. thermotolerans0.30 ± 0.056.3 ± 1.10.1 ± 0.020.41 ± 0.186.2 ± 0.20.27 ± 0.04
Y. lipolytica0.46 ± 0.034.2 ± 1.20000.51 ± 0.02
P. angusta0.42 ± 0.054.4 ± 0.80000.52 ± 0.08
C. rugosa0.45 ± 0.092.8 ± 0.40000.79 ± 0.02

Intracellular flux distribution

Given the differences in macroscopic physiological rates, we ask whether the intracellular distribution of fluxes relied on the same pathways within each group of aerobic glucose metabolism. For this purpose, separate isotopic tracer experiments with [U-13C]- and [1-13C]-labelled glucose were performed in at least duplicate batch cultures for each case. Specifically, we determined the mass isotopomer distributions of proteinogenic amino acids to calculate ratios of converging fluxes at key branch points in central metabolism (Blank et al., 2005; Zamboni et al., 2009). The largest variability between species was in the pentose phosphate pathway and the TCA cycle, indicated by large differences in the split ratios serine derived through glycolysis and mitochondrial oxaloacetate originating from anaplerosis, respectively (Fig. 1a).

Figure 1.

 Flux distribution in the seven yeast species on glucose minimal medium. (a) Ratios of converging fluxes at branch points of the central carbon metabolism. The ratios are given by the division of the flux indicated by the solid line over the flux indicated by the dotted line. (b) The quantitative flux distribution of Crabtree-positive (left) and aerobic (right) yeasts in mmol g−1 h−1. The thickness of the arrows indicates the glucose uptake normalized flux distribution with a variability that is depicted in grey.

For a more detailed insight, we estimated network-wide absolute fluxes using these intracellular flux ratios (Fig. 1a) and secretion rates (Table 2) as input values for net flux analysis using the fiatflux software (Zamboni et al., 2005, 2009). Specifically, this method estimates the best flux distribution by iteratively fitting intracellular fluxes to the experimental data within a model of yeast metabolism (Supporting Information, Appendix S1). These absolute flux values clearly differed between Crabtree-positive and -negative metabolism (Fig. 1b). In the Crabtree-positive species, glucose is almost exclusively catabolized through glycolysis and finally to the fermentative product ethanol, while the respiratory TCA cycle flux was negligible. This single-pathway metabolism was essentially independent of the overall flux rate that increased about threefold from K. thermotolerans to S. cerevisiae (Fig. 1b). In aerobic species, in contrast, the pentose phosphate pathway and TCA cycle became major pathways that equal the glycolytic flux. However, because the overall flux was much lower in the aerobic species, the pentose and NADPH-producing pentose phosphate pathway remained largely constant in all species at 1–1.7 mmol g−1 h−1, and contributed between 50% and 80% to the overall NADPH required for biomass production. Because we considered only the NAD-dependent isoform of the isocitrate dehydrogenase to be active, we could not further distinguish the isocitrate dehydrogenase or acetaldehyde dehydrogenase contribution to NADPH regeneration.

To elucidate whether there was a gradual shift between respiratory and fermentative metabolism, we correlated key fluxes from all species. We also included previously published S. cerevisiae fluxes during growth on the respiratory substrates galactose and pyruvate (Fendt & Sauer, 2010). On these substrates, the glycolytic flux was low and the TCA cycle flux high, representing an in-between state of metabolism. Consistent with their role as the key pathways, substrate uptake rate and yield correlated with both glycolytic flux and ethanol secretion, independently of whether respiratory metabolism was triggered genetically or environmentally (Fig. 2a). Including the data of S. cerevisiae on galactose and pyruvate further underlines the key role of the pentose phosphate pathway in supplying the anabolic cofactor NADPH for biomass formation in yeasts (Fig. 2b), which differs from its primarily catabolic role in some bacteria (Fischer & Sauer, 2005). These monotonous changes in various fluxes across genetic and environmental changes suggest similar underlying regulatory mechanisms. A similarly good correlation was found between the substrate uptake rate or glycolytic flux with the respiratory TCA cycle flux among all Crabtree-positive yeast on glucose and S. cerevisiae on respiratory substrates. The aerobic yeasts, however, fall completely out of these correlations; i.e. they exhibited several-fold higher TCA cycle fluxes than would be expected for the same glycolytic flux in galactose-grown S. cerevisieae (Fig. 2c). This suggests a different mode of regulation in the aerobic species relative to the environmentally triggered regulation of respiratory metabolism in S. cerevisiae. Generally, the increased TCA cycle flux of the Crabtree-positive yeasts is a function of higher biomass yield as their metabolism becomes more efficient with reduced ethanol secretion. In aerobic species, in contrast, the biomass yield is not increased any further.

Figure 2.

 Inter-relationship between fluxes of the central carbon metabolism in Crabtree-positive species (•), aerobic species (○) and Saccharomyces cerevisiae under galactose and pyruvate conditions (+). The fitted values from the net flux analysis were used. (a) Correlations to glucose uptake and glycolysis. (b) Correlations to growth rate. (c) Correlations to the TCA cycle flux.

Relations between fluxes and metabolites

Finally, we asked whether intracellular metabolite concentrations are species-specific or generally related to the flux through their pathway. For this purpose, we quantified 26 intermediates of the central metabolism plus 10 amino acids by LC-MS/MS (Buscher et al., 2009) from mid-exponential growing aerobic batch cultures of the seven species (Fig. 3). Generally our metabolite concentrations were in good agreement with published data (Bolten & Wittmann, 2008; Canelas et al., 2009; Fendt et al., 2010). The sole exception was a systematically higher AMP concentration in the Crabtree-positive strains that resulted in a slightly lower energy charge of about 0.7 compared with 0.9 in the other yeasts (Wiebe & Bancroft, 1975). Our data do not differentiate whether this difference was biological or a putative artifact from imperfect quenching by cold methanol. None of the species featured systematically lower or higher concentrations than the others, indicating a comparable quality of extraction by the hot ethanol procedure used. Generally, the intracellular pool of amino acids accounted for about 90% of all determined metabolite concentrations, with glutamate having concentrations up to 150 nmol mg−1 (Fig. 4). In Y. lipolytica and C. rugosa, glutamate was the dominating metabolite, accounting for about 50% of the total determined metabolite pool. In the two Saccharomyces species S. bayanus and S. cerevisiae, three amino acids each contributed about one-quarter to the total pool; i.e. glutamate, glutamine and alanine or arginine (Fig. 4). All other concentrations varied significantly between species at much lower concentrations between 0.02 and about 10 nmol mg−1 (Fig. 4).

Figure 3.

 Intracellular metabolite concentrations of the seven yeast species. CR, Candida rugosa; PA, Pichia angusta; YL, Yarrowia lipolytica; KT, Kluyveromyces thermotolerans; SE, Saccharomyces exiguus; SB, Saccharomyces bayanus; SC, Saccharomyces cerevisiae. The species are sorted according to their glucose uptake rate, and hence the Crabtree effect. The concentrations are given in nmol mg−1 or arbitrary units (AU).

Figure 4.

 Fractionation of the measured metabolome into the different metabolites. The amino acids make up 90% of the total metabolite concentrations measured. Nevertheless, the portioning of this pool between the different amino acids is quite diverse.

To assess the relationship of intracellular metabolite concentrations with fluxes or species, we correlated the determined pathway fluxes with all metabolite concentrations (Fig. 5). There was clearly no general correlation between metabolite levels and fluxes in a species-overarching manner; hence, metabolite levels are rather species-specific. The exceptions were the good correlations of fructose-1,6-bisphosphate and dihydroxyacetone phosphate with glycolysis and ethanol secretion (R2>0.9; P<10−9). Because both metabolites span a relatively large range of concentrations of about one order of magnitude in either case, they could potentially function as general glycolytic flux indicators, a mechanism that then would be conserved across species.

Figure 5.

 Correlation between fluxes and metabolites in the seven yeast species. In the heat map, the correlation coefficients (R2) between all metabolites (left) and the representative fluxes (top) are illustrated. Fructose-1,6-bisphosphate and dihydroxyacetone-phosphate showed a high correlation with glycolysis and ethanol secretion in a species-overarching manner (right).


We quantified metabolic fluxes in seven yeast species with different strengths of the Crabtree effect by 13C flux analysis and metabolomics. For six species, the physiological metabolic states of aerobic glucose metabolism were in agreement with published data (Blank et al., 2005; Merico et al., 2007; Fendt & Sauer, 2010). As expected, decreasing glucose uptake rates concur with gradually decreasing ethanol secretion until complete absence of fermentation. Despite this gradual physiological transition, the corresponding intracellular flux distributions fall into two clearly different modes of Crabtree-positive and aerobic metabolism. Specifically, the aerobic species featured about threefold higher respiratory TCA cycle fluxes than respiring Crabtree-positive S. cerevisiae at otherwise equal rates of overall metabolism on galactose and pyruvate. This difference in respiratory rates at equal glycolytic flux are potentially explained by the stronger repression of the TCA cycle and respiration, with concomitant activation of overflow metabolism at the pyruvate decarboxylase in the Crabtree-positive species (Pronk et al., 1996; Gancedo, 2008; Zaman et al., 2008). In this case, the fermentation route is the only possibility to reoxidize NADH, hence to maintain redox homeostasis. For the aerobic species Kluyveromyces lactis and Pichia stipitis, in contrast, it was shown that glucose repression of respiratory genes is absent and overflow metabolism at the pyruvate decarboxylase branchpoint is negatively regulated in an oxygen-dependent manner (Passoth et al., 1996; Kiers et al., 1998). Thus, aerobic species channel their glycolytic flux into the TCA cycle for complete oxidation and NAD+ regeneration via respiration (Snoek & Steensma, 2007). Interestingly, aerobic species rely primarily on energy-dependent glucose transport (van Urk et al., 1989; van Dijken et al., 1993), which might further require an energy-efficient mode of metabolism.

Previous studies with S. cerevisiae revealed glycolytic intermediate concentrations to vary with differentially repressive substrates and thus with glycolytic fluxes (Elbing et al., 2004; Bosch et al., 2008). If this was indeed a general feature of glycolysis, one would expect a correlation between flux and the respective pathway intermediate concentrations across all species. Here, we show that intracellular metabolites do not generally correlate with flux and are therefore species-specific. Most strikingly, up to 90% of the total metabolite pool was made up of amino acids. Different from Escherichia coli (Bennett et al., 2009), the total amino acid pool in yeast species was not only dominated by glutamate but rather by a combination of glutamate, glutamine, arginine, alanine and aspartate. The large variability of other metabolite concentrations was not correlated with fluxes, which indicates diverse kinetic properties and expression levels of enzymes in the different yeast species. The exceptions were the good correlations between fructose-1,6-bisphosphate and dihydroxyacetone-phosphate concentrations with glycolytic flux, which indicates that the functional properties of enzymes around these two metabolites are conserved among the different species. Because fructose-1,6-bisphosphate inhibition of the respiratory complex III in Crabtree-positive species was proposed earlier (Diaz-Ruiz et al., 2008), it is tempting to speculate that this inhibition causes the negative correlation between the TCA cycle and the glycolytic flux. The almost equally good correlation of dihydroxyacetone-phosphate might then be explained by the near equilibrium with fructose-1,6-bisphosphate via aldolase.


We thank the Degussa AG and the Swiss initiative for systems biology ( project YeastX for financial support.