Central carbon metabolism of Saccharomyces cerevisiae explored by biosynthetic fractional 13C labeling of common amino acids


  • Hannu Maaheimo,

    1. Department of Chemistry, University at Buffalo, The State University of New York, NY, USA;
    2. VTT Biotechnology, Espoo, Finland;
    Search for more papers by this author
    • *Note: these authors contributed equally to this paper

  • Jocelyne Fiaux,

    1. Institut für Molekularbiologie und Biophysik and
    Search for more papers by this author
    • *Note: these authors contributed equally to this paper

  • Z. Petek Çakar,

    1. Institut für Biotechnologie, Eidgenössische Technische Hochschule Hönggerberg, Zürich, Switzerland
    Search for more papers by this author
    • Present address: Department of Biochemistry, Biozentrum, University of Basel, CH-4056 Basel, Switzerland

  • James E. Bailey,

    1. Institut für Biotechnologie, Eidgenössische Technische Hochschule Hönggerberg, Zürich, Switzerland
    Search for more papers by this author
  • Uwe Sauer,

    1. Institut für Biotechnologie, Eidgenössische Technische Hochschule Hönggerberg, Zürich, Switzerland
    Search for more papers by this author
  • Thomas Szyperski

    1. Department of Chemistry, University at Buffalo, The State University of New York, NY, USA;
    Search for more papers by this author

T. Szyperski, Department of Chemistry, University at Buffalo, The State University of New York, Buffalo, USA. Fax:. + 1716 6457338, Tel.: + 1716 6456800 ext. 2245, E-mail: szypersk@chem.buffalo.edu


Aerobic and anaerobic central metabolism of Saccharomyces cerevisiae cells was explored in batch cultures on a minimal medium containing glucose as the sole carbon source, using biosynthetic fractional 13C labeling of proteinogenic amino acids. This allowed, firstly, unravelling of the network of active central pathways in cytosol and mitochondria, secondly, determination of flux ratios characterizing glycolysis, pentose phosphate cycle, tricarboxylic acid cycle and C1-metabolism, and thirdly, assessment of intercompartmental transport fluxes of pyruvate, acetyl-CoA, oxaloacetate and glycine. The data also revealed that alanine aminotransferase is located in the mitochondria, and that amino acids are synthesized according to documented pathways. In both the aerobic and the anaerobic regime: (a) the mitochondrial glycine cleavage pathway is active, and efflux of glycine into the cytosol is observed; (b) the pentose phosphate pathways serve for biosynthesis only, i.e. phosphoenolpyruvate is entirely generated via glycolysis; (c) the majority of the cytosolic oxaloacetate is synthesized via anaplerotic carboxylation of pyruvate; (d) the malic enzyme plays a key role for mitochondrial pyruvate metabolism; (e) the transfer of oxaloacetate from the cytosol to the mitochondria is largely unidirectional, and the activity of the malate–aspartate shuttle and the succinate-fumarate carrier is low; (e) a large fraction of the mitochondrial pyruvate is imported from the cytosol; and (f) the glyoxylate cycle is inactive. In the aerobic regime, 75% of mitochondrial oxaloacetate arises from anaplerotic carboxylation of pyruvate, while in the anaerobic regime, the tricarboxylic acid cycle is operating in a branched fashion to fulfill biosynthetic demands only. The present study shows that fractional 13C labeling of amino acids represents a powerful approach to study compartmented eukaryotic systems.






biosynthetically directed fractional


metabolic flux ratio


tricarboxylic acid








acetyl-coenzyme A






ribose 5-phosphate


erythrose 4-phosphate


pentose phosphate pathway


serine hydroxymethyltransferase


glycine cleavage pathway


malate dehydrogenase.

Biosynthetically directed fractional (BDF) 13C labeling of proteinogenic amino acids combined with 2D 13C-1H correlation NMR spectroscopy (COSY) for analysis of 13C–13C scalar coupling fine structures and balancing of contiguous carbon fragments within metabolic networks, has been used as a powerful tool to investigate intermediary metabolism of prokaryotic cells [1]. The desired fractional 13C labeling is achieved by feeding a mixture of uniformly 13C labeled and unlabeled compounds as the sole carbon source for cellular growth [2–4]. An outstanding strength of this approach is its ability to enable both a comprehensive characterization of the network of active biosynthetic pathways [1,5–7] and the accurate determination of metabolic flux ratios [1,6,8–11] in a single experiment (reviewed in [12]). This led us to name the approach metabolic flux ratio (METAFoR) analysis by NMR [9,14]. Additional attractive features with regard to routine applications are an inherently small demand for 13C labeled source compounds, the high sensitivity of 2D 13C-1H COSY, and the availability of a user-friendly program, fcal, for rapid data analysis [14].

Previous applications of the BDF 13C labeling protocol included its combination with metabolic flux balancing in order to derive net fluxes in Bacillus subtilis[6], the exploration of amino-acid biosynthesis in the halophilic archaeon Haloarcula hispanica[7], the investigation of Escherichia coli central carbon metabolism in microaerobic bioprocesses [8], and the larger scale evaluation of the impact of genetic (or environmental) modulations [9] and the introduction of D2O into the growth medium [11] on E. coli central carbon metabolism. These studies neatly exemplified the value of BDF 13C labeling experiments for studying prokaryotic central metabolism in both fundamental and applied biotechnological research.

In the present publication, we describe the extension of the BDF 13C labeling protocol for investigating eukaryotic systems. The key challenge is quite evidently the compartmentation of the eukaryotic cell, which leads to a dissection of central carbon metabolism in subnetworks localized in either the cytosol or in organelles, e.g. mitochondria or peroxisomes [15]. Yeasts, such as Saccharomyces cerevisiae, are eukaryotic model organisms that are pivotal for many areas of biology, biomedicine and biotechnology. Moreover, an outstanding body of experimental data with respect to central and amino-acid metabolism is available [16–18], and the genome of S. cerevisiae has been completely sequenced [19–21]. Hence, we decided to explore central carbon metabolism of the yeast S. cerevisiae when grown in batch cultures under glucose repression, in both aerobic or anaerobic regimes.

Materials and methods

Biosynthetically directed fractional 13C labeling

S. cerevisiae strain CEN.PK 113–7D (MATa, URA3, HIS3, LEU2, TRP1, MAL2–8c, SUC2) was obtained from P. Kötter (Institute of Microbiology, Johann Wolfgang Goethe-University, Frankfurt, Germany). The cells were grown on a yeast minimal medium, containing 6.7 g·L−1 of yeast nitrogen base lacking amino acids (Difco), and 0.5% (w/v) glucose consisting of a mixture of 10% (w/w) U-13C labeled (Martek Biosciences Corporation, Columbia, USA) and 90% (w/w) unlabeled glucose. Aerobic cultivations were performed using a rotary shaker at 30 °C and 300 r.p.m., in 1-L baffled shake flasks containing 100 mL medium. Anaerobic cultivations were carried out under moderate shaking (140 r.p.m.) at 30 °C in 125 mL flasks with tight rubber caps and 125 mL medium. The flasks were flushed with nitrogen prior to cultivation.

Cell growth was monitored spectrophotometrically by measuring the optical density of the cultures at 600 nm (D600). The cells were harvested during the mid-exponential phase at D600 = 0.8–1.0 by centrifugation at 3000 g for 10 min using a benchtop centrifuge (Beckman, GS-6R). The resulting pellets were resuspended in Tris/HCl (20 mm, pH 7.6) and then centrifuged again at the previous settings, after which the supernatant was discarded. The pellets were vacuum-dried (Hetovac, VR-1) overnight to constant mass and the amount of dried biomass was determined. The biomass was then resuspended in 3 mL Tris/HCl (20 mm, pH 7.6). Upon addition of 6 mL 6 m HCl, hydrolysis was performed in sealed glass tubes at 110 °C for 24 h, after which the solutions were filtered through 0.2-µm filters (Millipore, Millex-GP). Finally, the hydrolysed biomass was lyophilized. Approximately 60–70 mg of dried hydrolysate were obtained from each cultivation.

NMR spectroscopy

The dried hydrolysates were dissolved in 0.1 m DCl in D2O, and 2D 13C-1H COSY spectra [22,23] were acquired for both aliphatic and aromatic resonances as described [1,5,6,8] at 40 °C on a Bruker DRX 500 spectrometer. The measurement time was ≈ 8 h for each 2D spectrum, and the spectra were processed using the program prosa[24]. The 13C abundance was determined from the resolved 13C satellites in 1D 1H NMR spectra [1], and an overall degree of 13C labeling of 10% was obtained for both the aerobic and the anaerobic preparation. The degree of 13C labeling was in close agreement with the composition of the minimal medium, and was confirmed by analysis of the 13C scalar coupling fine structure of Leu-β[1].

Data analysis

The program fcal[14] (R. W. Glaser; fcal v2.3.0) was used for data analysis including both the integration of 13C scalar fine structures and the calculation of relative abundances,  f, of intact carbon fragments arising from a single source molecule of glucose [1]. Notably, the use of fcal allowed integration of the fine structure of Lys-α, which largely overlaps with the signal of Arg-α[1]. In contrast to prokaryotic cells [1,12], the fine structure of Lys-α provides unique information: only this carbon encodes the 13C labeling pattern of cytosolic AcCoA (see below). The equations relating the 13C fine multiplet intensities to the f values allow derivation of ‘fragment balancing’ equations, which can provide accurate flux information largely independent of the overall degree of 13C incorporation [1,12].

The nomenclature used here to identify such fragments has been described previously [1,5]. Briefly,  f (1) represents the fraction of molecules in which the observed carbon atom and its neighbouring carbons originate from different source molecules of glucose, and f (2) the fraction of molecules in which the observed carbon atom and at least one neighbouring carbon originate from the same source molecule. For a central carbon in a C3 fragment that exhibits different 13C–13C scalar coupling constants with the two attached carbons,  f (2) represents the fraction of molecules for which the central carbon and the carbon with the smaller coupling come from the same source molecule, while f (2*) is used if the carbon with the larger coupling comes from the same source molecule as the observed carbon. f (3) denotes the fraction of molecules in which the observed carbon atom and both neighbours in the C3 fragment originate from the same glucose molecule.

Previous experiments with prokaryotic cells have provided ample support for the suggestion that central carbon metabolism operates in a quasi steady-state manner during the exponential growth phase [7,9,11,14]. Consequently, the quasi steady-state assumption was adopted for the present study.

Biochemical reaction network for S. cerevisiae

At the outset of this study, we compiled the currently available knowledge about central carbon metabolism of S. cerevisiae from literature [15–18] and databases (e.g. http://www.expasy.ch/cgi-bin/search-biochem-index; http://wit.mcs.anl.gov/WIT2/CGI/index.cgi; http://www.genome.ad.jp/kegg/metabolism.html; http://www.rz.uni-frankfurt.de/FB/fb16/mikro/euroscarf/index.html) in order to construct a biochemical reaction network suited for the presently chosen growth conditions where glucose served as the sole carbon source (Fig. 1). The salient feature of eukaryotic central carbon metabolism is apparently its dissection into cytosolic and mitochondrial subnetworks, connected by intercompartmental transport of metabolites [15–18]. Glycolysis and the pentose phosphate pathway (PenPp) are located in the cytosol, and the tricarboxylic acid (TCA) cycle operates in the mitochondria where respiration takes place [25–27]. Oxaloacetate (OxAc), pyruvate (Prv) and acetyl-CoA (AcCoA) are present in both compartments, and systems for their transport across the mitochondrial membrane have been identified [27–30]. Hence, cytosolic (cyt) and mitochondrial (mt) pools are distinguished throughout for these three key intermediates.

Figure 1.

Network of active biochemical pathways constructed for S. cerevisiae cells grown with glucose as the sole carbon source. The network was inferred from current literature and data deposited in databases (see text). The dissection of central carbon metabolism into two subnetworks is apparent: glycolysis and the pentose phosphate pathway take place in the cytosol, while the TCA cycle takes place in the mitochondrial matrix. Irreversible reactions, represented by unidirectional arrows, have been inferred from (a) estimations of in vivo metabolite concentrations [86, 87] and in vitro free enthalpies (http://wwwbmcd.nist.gov:8080/enzyme), and (b) the experimental NMR data presented in this publication. 13C labeling patterns of metabolites shown in ellipses can be assessed through analysis of proteinogenic amino acids (Fig. 2). Carrier systems for transport of metabolites between cytosol and mitochondrial matrix are represented by grey circles. For the Prv and OxAc carrier systems, active transport by the mitochondrial proton motive force is indicated, and the corresponding back-fluxes are represented by dashed arrows. The AcCoA carrier system (AT) operates by facilitated diffusion. Abbreviations: AcCoA, Acetyl-Coenzyme A; AcH, acetaldehyde; AcOH, acetic acid; EtOH, ethanol; E4P, erythrose 4-phosphate; F6P, fructose 6-phosphate; G6P, glucose 6-phosphate; Fum, fumarate; Gly, glycine; G3P, glyceraldehyde 3-phosphate; GriP, 3-phosphoglycerate; Mal, malate; OxGlt, 2-oxoglutarate; Prv, pyruvate; PPrv, phosphoenolpyruvate; Ri5P, ribose-5-phosphate; Ru5P, ribulose 5-phosphate; S7P, seduheptulose 7-phosphate; Ser, serine; Suc, succinate; Xu5P, xylulose 5-phosphate. For Prv, AcCoA and OxAc, cytosolic (cyt) and mitochondrial (mt) pools are indicated separately.

Oxaloacetate, pyruvate and acetyl-CoA

In S. cerevisiae, OxAc is produced both as an intermediate of the TCA cycle in the mitochondrial matrix from malate, thus yielding mt-OxAc [25,26], and from cyt-Prv by the action of pyruvate carboxylase in the cytosol [31,32] yielding cyt-OxAc (Fig. 1). To enable the anaplerotic supply of the TCA cycle, cyt-OxAc is transferred across the mitochondrial membranes by the oxaloacetate carrier protein [30] in order to refill the mt-OxAc pool. The transport is driven by the proton motive force at the inner mitochondrial membrane, i.e. oxaloacetate is actively transferred across the membrane in symport with protons (Fig. 1). This mode of operation indicates directional transfer from the cytosol to the mitochondria.

cyt-Prv is produced in the cytosol by glycolysis [26] and a proportion of it is transported into the mitochondria [27] yielding mt-Prv, which may also be synthesized from malate by the malic enzyme. In S. cerevisiae, this enzyme can involve either NADH or NADPH as a cofactor [33]. The gene encoding the malic enzyme has recently been identified, and the enzyme was shown to be located exclusively in mitochondria [34]. Pyruvate is transported from the cytosol into the mitochondria by the pyruvate (monocarboxylate) carrier protein [35]. The transport is actively driven by the mitochondrial proton motive force (Fig. 1), suggesting a largely unidirectional transport from the cytosol into the mitochondria.

mt-AcCoA and cyt-AcCoA can be derived from mt-Prv and cyt-Prv, respectively, either by the pyruvate dehydrogenase complex in the mitochondria [36], or via a cytosolic ‘by-pass pathway’, the first reaction of which is catalyzed by pyruvate decarboxylase [37]. When S. cerevisiae cells have access to glucose, AcCoA is expected to originate exclusively from Prv [25]. AcCoA can cross the inner mitochondrial membrane via the ‘carnitine shuttle’[28,29]. This shuttle consists of the carnitine O-acetyltransferase [38,39] catalyzing the exchange of acetyl groups between carnitine and AcCoA (which itself cannot pass the mitochondrial membrane), and the acetyl-translocase. The acetyl-translocase represents a carnitine/acetylcarnitine antiport system which serves to balance the cytosolic and mitochondrial AcCoA pools by facilitated diffusion.

Glyoxylate pathway

When S. cerevisiae cells are grown on a medium containing glucose as the sole carbon source, the cytosolic enzymes constituting the glyoxylate pathway [40,41] are not expressed [40,42]. Notably, both cytosolic and peroxisomal malate dehydrogenase isozymes (MDH2 and MDH3, respectively), which solely serve to establish the glyoxylate cycle, are glucose-repressed along with the other glyoxylate cycle enzymes [43].

C1 metabolism

The major one-carbon donor in S. cerevisiae is serine, which is reversibly cleaved into glycine and a C1 unit by serine hydroxymethyl transferase (SHMT). Mitochondrial and cytoplasmic SHMT isozymes exist [44,45], and their action is complemented by the mitochondrial glycine cleavage pathway (GCV) which interconverts glycine into a C1 unit and CO2[46].

Biosynthesis of proteinogenic amino acids in S. cerevisiae

In order to recruit proteinogenic amino acids as probes to study central carbon metabolism [1,4], their biosynthetic pathways must be available. When applying this approach to eukaryotic cells, it is necessary to identify which of the cytosolic or mitochondrial pools of Prv, AcCoA or OxAc are used for the biosynthesis of which amino acids. Moreover, the amount of protein synthesized within the mitochondria is negligible when compared with the amount generated in the cytosol. In yeast, the mitochondrially synthesized proteins have been estimated to represent only 5–10% of the total mitochondrial protein [47], which itself represents only a small fraction of the total cellular protein. Hence, potentially different 13C labeling patterns for mitochondrial and cytosolic amino-acid pools cannot be detected, and f values refer, to a very good approximation, to cytosolic amino-acid pools only.

Except for Lys and Gly, the common amino acids are synthesized in S. cerevisiae from the same precursors (Fig. 2A) as in E. coli[15]. Specifically, Lys is synthesized via the α-aminoadipate pathway from OxGlt and cyt-AcCoA, while Gly is generated through both Ser cleavage via SHMT [15] and Thr cleavage via threonine aldolase [48], as well as the mitochondrial GCV (see above). During routine acidic hydrolysis of biomass, tryptophan and cysteine are lost due to oxidation. Hence, these amino acids are not considered in the framework of the present study.

Figure 2.

Carbon fragments originating from a single intermediate molecule present in proteinogenic amino acids, provided that only anabolic pathways are active in S. cerevisiae cells [ 15–18 ]. (A) Representation of the carbon skeleton of the intermediates of glycolysis, TCA and the pentose phosphate pathway with circles, squares and triangles, respectively. Acetyl-Coenzyme A (AcCoA) is also represented with circles, and phosphoryl groups are denoted by ‘P’. Intermediates in the solid box are generated in the cytosol, and intermediates in the dashed box are synthesized in the mitochondria. Hence, pyruvate, acetyl-CoA and oxaloacetate are generated in both compartments, and the exchange flux between the mitochondrial (mt) and cytosolic (cyt) pools need to be considered for proper data interpretation. (B) Representation of the carbon skeletons of amino acids, as well as the origin of their carbon atoms with respect to the metabolic intermediates displayed in (A) and the compartment in which the intermediate is recruited for amino-acid biosynthesis (dashed and solid boxes indicate mitochondrial and cytosolic origin, respectively). Ala is marked with an asterisk (*) because the mitochondrial origin has been inferred from the present study (see text). Thin lines denote carbon bonds that are formed between fragments arising from different intermediate molecules, while thick lines indicate carbon–carbon connectivities in intact fragments arising from a single intermediate molecule. Dashed lines connect fragments arising from the same intermediate molecule that are not directly attached in the amino-acid carbon skeleton. Unlabeled carbon atoms in the amino acids are C′ carbons. It is indicated that in S. cerevisiae Gly can be synthesized either via serine or threonine cleavage. In (B), those amino acids that are lost due to oxidation (Cys, Trp) or that are deamidated (Asn, Gln) during routine hydrolysis of cellular protein are not considered. Note that Ser and Gly are affected by C1 metabolism so that their 13C fine structures usually do not reflect the isotopomeric composition of GriP. The superscript ‘x’ indicates that the two carbons δ1 and δ2, and ε1 and ε2, respectively, of Tyr and Phe give rise to only one 13C fine structure each [1]. The nomenclature of the carbon atoms follows IUPAC-IUB recommendations [88].

Identification of the subcellular location of intermediates involved in amino-acid biosynthesis (Fig. 2B) requires knowledge about the location of the enzymes catalysing their incorporation into the amino-acid carbon skeleton [15]. Ser, Gly, Asp, Met, Thr, Tyr, Phe and His are exclusively derived from metabolite pools located in the cytosol. Ala, Val, Leu and Ile synthesis requires Prv. For Ala, the first step is the amination of pyruvate catalysed by alanine aminotransferase. For Val and Leu, the condensation of two molecules of Prv to 2-acetolactate catalysed by acetolactate synthase, and for Ile, the condensation of Prv and 2-oxobutyrate (arising from Thr) to 2-aceto-2-hydroxy-butyrate (Ile) catalysed by acetolactate synthase, serve as the first steps. While the subcellular localization of the alanine aminotransferase has not yet been identified, acetolactate synthase is located in the mitochondria [49,50]. Moreover, the incorporation of AcCoA into Leu is catalysed by mitochondrial α-isopropylmalate synthase [49]. Glu, Pro and Arg are synthesized from OxGlt, which is exclusively generated in the mitochondria. Lys is synthesized from OxGlt and mt-AcCoA via the cytosolic homocitrate synthase [51]. (It should be added here that Ala can also be synthesized from glyoxylate via glycine aminotransferase [52]. However, this pathway is not active when S. cerevisiae is grown on glucose [52], and was thus not further considered for the present study.)

Figure 2 summarizes the biochemical data outlined in this paragraph and shows a mapping of the carbon skeletons of the proteinogenic amino acids to the intermediates of central carbon metabolism involved in their biosynthesis. This mapping thus assigns the 13C fine structure observed for a given carbon in the amino acids to carbon atoms in the intermediates. For the sake of simplicity, and when appropriate, we may refer f values experimentally determined for the amino acids directly to the corresponding carbon of the metabolic intermediates. Figure 2 provides a guide to make this reasoning readily transparent to the reader.

Analysis of 13C scalar coupling fine structures in S. cerevisiae

In order to cope with cellular compartmentation as encountered in eukaryotic organism S. cerevisiae (Fig. 1), the formalism derived to analyse prokaryotic metabolism [1,5–9] needs to be extended accordingly. An established but key assumption is that the metabolic network is operating in a quasi steady-state (see above). Moreover, fluxes of glycolysis and PenPp, both of which are entirely located in the cytosol, can be treated as described for prokaryotic cells [1,6–9,12] (see above).

Anaplerosis of TCA cycle

The contribution of the anaplerotic interconversion of cytosolic PPrv into mt-OxAc, which is participating in the TCA cycle, shall be calculated. The synthesis of OxAc from PPrv via the joint action of pyruvate kinase and pyruvate carboxylase is restricted to the cytosol in S. cerevisiae[31,32], and cyt-OxAc thus needs to be imported into the mitochondria to warrant anaplerosis. However, only cyt-OxAc, but not mt-OxAc, can be assessed via Asp (Fig. 2), and the required f values of mt-OxAc must be inferred from OxGlt. For growth of S. cerevisiae on glucose, it can be safely assumed that the condensation of mt-AcCoA, assessed via Leu-α (Fig. 2), and mt-OxAc is irreversible [15]. Hence, the fraction of intact C2-C3 connectivities in mt-OxAc, which is equal to the fraction of mt-OxAc stemming from anaplerotic carboxylation [1], denoted Xana=X(mt-OxAc ← PPrv), can be derived from OxGlt and PPrv and is given by Eqn (1):


The f values of Glu-α and Phe/Tyr-α represent C2 of OxGlt and PPrv, respectively (Fig. 2).

Exchange flux between OxAc and fumarate

Intact C1-C2-C3 fragments originating from a single source molecule of glucose in OxGlt must arise from mt-OxAc that was converted to mt-fumarate or mt-succinate and subsequently reacted back to mt-OxAc [1]. Hence, the fraction of mt-OxAc molecules that were at least once reversibly interconverted to mt-fumarate, denoted mt-Xexch, is given by Eqn (2):


The same fraction can be calculated for the cytosolic pool [1] using Eqn (3):


In the case that (a) cyt-Xexch is nonzero and (b) the transfer from OxAc into the mitochondria is unidirectional, Eqn (2) can readily be extended to:


Calculation of Xexch according to Eqns (2) and (3) assumes complete symmetrization of 13C labeling patterns about the C2-C3 bond of succinate or fumarate [1] arising from unrestricted rotation in solution. In cases where metabolite channeling [53–57], which may restrict rotational reorientation, had to be invoked for the exchange reaction, the value for Xexch would represent a lower bound (see below).

Flux via the malic enzyme

The interconversion of mt-malate to mt-Prv via the malic enzyme is manifested by the flux of excess intact C1-C2 fragments (represented by f (2*)) and C3-C2 fragments (represented by f (2)) into the mt-Prv pool when compared with the PPrv pool [6]. Upper and lower bounds for the fraction of mt-Prv from mt-malate can be derived under the assumptions that malate is entirely derived from either mt-OxAc or OxGlt, respectively [6]. The calculation of the upper bound requires the f values for mt-OxAc (see below).

If Xana and Xexch are known from Eqns (1) and (2), the f values can be calculated from Asp and Glu, representing cyt-OxAc and OxGlt, respectively:


where i = 1, 2, 2*, or 3 in these and the following equations.

Eqns (4) were derived in the limit of complete metabolite channeling [53–57], i.e. symmetrization of 13C labeling patterns about the C2-C3 bond of succinate or fumarate that arise from free rotation in solution, is not considered. Assuming full symmetrization, i.e. neglecting any metabolite channeling, the contributions multiplied by the (1 − Xana) term in Eqns (4) are averaged between C2 and C3 of OxAc, yielding Eqns (5):


Finally, the symmetrization of 13C labeling patterns arising from reversible interconversion from OxAc to fumarate, characterized by mt-Xexch, is considered, yielding Eqns (6):


Provided that Ala is synthesized from mt-Prv, Eqns (7) and (8) yield the aforementioned upper, Xub, and lower bounds, Xlb, assuming that malate is entirely derived from either mt-OxAc or OxGlt, respectively. With f (2*){mt-OxAc-C2} from Eqn (6a) this yields:




Considering that f (2){mt-malate-C2} ≤ f (2){mt-OxAc-C2}, a second lower bound can be derived with f (2){mt-OxAc-C2} from Eqn (6a) when introduced into balancing Eqn (9):


Intercompartmental exchange fluxes

Balancing equations for relative abundances of intact carbon fragments (i.e. f values), which characterize the cytosolic and mitochondrial pools of Prv, AcCoA and OxAc, can provide estimations of metabolic flux ratios involving intercompartmental fluxes. Figure 3 displays the selected subnetwork as well as metabolic flux definitions chosen to derive the desired equations. Reactions that were considered to be irreversible are indicated by unidirectional arrows (see legend of Fig. 1).

Figure 3.

Sub-network for analysis of fluxes at the interface between glycolysis and TCA cycle considering also the compartmentation in S. cerevisiae cells ( Fig. 1 ). Metabolic fluxes, ν, and flux ratios, X, are labeled with ‘mt’ for mitochondrial and with ‘cyt’ for cytosolic depending on the compartment in which the biochemical reactions takes place. The flux ratios which involve the exchange fluxes (see text) are depicted in grey. Abbreviations for metabolites are as given in the legend of Fig. 1. Single-headed arrows represent reactions that are assumed to be unidirectional under the present growth conditions (see legend of Fig. 1). Metabolites which are assessible by BDF 13C labeling of proteinogenic amino acids are encircled.

1. Prv. Assuming that the mt-Prv pool is experimentally assessible via Ala (Fig. 2), i.e.

inline image

one obtains with inline image = inline image/[inline image + inline image]Eqn (10):


and estimations of f (i) {mt-malate-C2} and f (i) {cyt-Prv-C2} yield the flux ratio inline image. For the bioreaction network of Fig. 3, inline image corresponds to X(mt-Prv ← mt-malate) introduced in Eqns (79) to estimate the involvement of the malic enzyme to synthesize mt-Prv.

The cytosolic PPrv pool is experimentally assessible via Phe or Tyr (Fig. 2), i.e. f (i) {PPrv-C2} = f (i) {Phe-α, Tyr-α}. Assuming that the mt-Prv pool is observed at Ala, i.e. that f (i) {mt-Prv-C2} = f (i) {Ala-α}, one obtains with inline image = inline image/[inline image+ inline image]Eqn (11):


from which the flux ratio inline image can be calculated.

2. AcCoA. The mt-AcCoA pool is experimentally assessible via Leu-α (Fig. 2), i.e. f (2) {mt-AcCoA-C2} =f (2) {mt-AcCoA-C1} = f (2*) {Leu-α} (Fig. 2), and the cytosolic values f (i) {cyt-AcCoA-C2} are obtained correspondingly from analysis of Lys-α. One then obtains with inline image = inline image/[inline image+ inline image]Eqns (12):


yielding the flux ratio inline image.

Analogous equations are valid for the cyt-AcCoA pool, and with inline image = inline image/[inline image + inline image]Eqn (13) is obtained:


The f values for cyt-Prv-C2 need to be derived from PPrv (i.e. Phe-α, Tyr-α) or cyt-OxAc (i.e. Asp-α).

3. OxAc. The mt-OxAc pool has been considered for assessing the anaplerosis of the TCA cycle: a flux ratio invoking inline image can be derived in a straightforward fashion when considering that anaplerosis is solely due to the transport of cyt-OxAc into the mitochondria (see above).


where inline image represents the net flux providing mt-OxAc from OxGlt through operation of the TCA cycle. For unidirectional interconversion of PPrv into cyt-OxAc (Fig. 1), one obtains with Xana from Eqn (1) and the bioreaction network of Fig. 3 that


The cyt-OxAc pool is experimentally assessible via Asp (Fig. 2), i.e. f (i){cyt-OxAc-C2} = f (i) {Asp-α}. With inline image = inline image/[inline image + inline image] one arrives at the balancing Eqns (15):


Hence, estimations of f (i) {cyt-Prv-C2} (see above) and the f (i) {mt-OxAc-C2} values yield the flux ratio inline image.

Results and discussion

Biosynthesis of proteinogenic amino acids in S. cerevisiae

The 2D 13C-1H COSY spectra were analysed as described previously [1,14] yielding the desired f values (Table 1). During hydrolysis of the biomass, Cys and Trp were oxidized and could thus not be evaluated, Asn and Gln were deamidated to Asp and Glu, and the ring carbons of phenylalanine were not considered because of strong 13C–13C scalar coupling effects [1]. A survey of cross sections taken along ω1(13C) from the 2D spectra is afforded by Fig. 4.

Table 1.  Relative abundances of intact C2 and C3 fragments in proteinogenic amino acids. The first column indicates the carbon position for which the 13C fine structure was observed. The f values were calculated as described [1], and are given for the aerobic and the anaerobic preparation in columns 2–5 and 6–9, respectively. Note, firstly, that for terminal carbons f (2*) and f (3) are not defined, and, secondly, that for Tyr the two carbons δ1 and δ2, and ε1 and ε2, respectively, give rise to only one 13C fine structure each [1].
 Relative abundance of intact carbon fragments
atom f (1) f (2) f (2*) f (3) f (1) f (2) f (2*) f (3)
Figure 4.

13C scalar coupling fine structures observed for BDF 13C labeled amino acids obtained during growth of the S. cerevisiae cultures on a minimal medium containing glucose as the sole carbon source. In the left-most column the carbon position is listed for which the 13C fine structure was observed. The right-most column indicates the carbon positions for which closely similar 13C fine structures were observed (for a quantitative comparison see Table 1). The columns two and three display the ω1(13C) cross sections as observed in the 2D 13C-1H COSY spectrum recorded for the aerobic and anaerobic preparation, respectively. The carbon atoms are grouped according to the metabolic intermediates from which they are derived (Fig. 2). The section on the left comprises intermediates generated in the cytosol, while the section on the right contains those synthesized in the mitochondria. The doublet of doublets arising from histidine 13Cγ-13Cβ-13Cα is further split by the two-bond coupling constant 2JCβCδ, indicating the presence of intact C5 fragments in histidine. Note that due to the influence of 13C isotope effects on the 13C chemical shifts the doublet, as well as the doublet of doublets is shifted by ≈ 1–2 Hz to higher field relative to the singlet line [89]. Hence, the superposition of the multiplets arising from different isotopomers is not symmetrical with respect to the singlet line.

Analysis of the f values [1,7] obtained in both the aerobic and anaerobic regimes (Table 1) provided strong evidence for the hypothesis that the proteinogenic amino acids in S. cerevisiae are synthesized according to known pathways [15–18] (Fig. 2). Except for a yet unexplained slight deviation of the f values of Leu-δ1, which was previously also observed for prokaryotes [1], inconsistencies could not be detected: f values predicted [1,7] from (a) the bioreaction network (Fig. 1), (b) the location of intermediates serving for amino acid biosynthesis (Fig. 2), and (c) the known biosynthetic routes [15–18] (Fig. 2) were in excellent agreement with those determined experimentally. As (a) f (i){Lys-δ} = f (i){Glu-γ} = f (i){mt-OxAc-C4}, (b) f (i){Lys-γ} =f (i){Glu-β} = f (i){mt-OxAc-C3} and (c) [  f (2) + f (3)]{Lys-α} = 0 (Fig. 2; Table 1), Lys synthesis must occur from homocitrate and AcCoA, i.e. via the α-aminoadipate pathway.

When grown on glucose, S. cerevisiae cells are using both Thr and Ser cleavage to generate Gly [44,48]. However, as Ser is derived from GriP, which can be assumed to exhibit virtually the same f values as cyt-Prv [1] (from which Thr is derived via carboxylation to cyt-OxAc), we have that [  f (2*) + f (3)]{Ser-α} = [  f (2*) + f (3)]{Thr-α} (Table 1). Cleavage of the Cα-Cβ bond of either Ser or Thr would thus lead to the same f (2) {Gly-α} values, i.e. the relative contribution of the two cleavage pathways cannot be determined using the presently employed 13C labeling protocol. In contrast, we find that f (1){Ser-β} − f (1){Phe-β} = 0.19 in both the anaerobic and the aerobic regime. This provides evidence for reversible cleavage of 19% of the Ser molecules into Gly and a C1 unit [1], while f (1){Thr-β}≈ 0 shows that Thr cleavage is, if present, not reversible. Hence, we are left with evidence for the operation of SHMT, but we cannot judge on the relative importance of Thr cleavage for Gly biosynthesis. Considering that Thr cleavage has, in principle, been shown to be of importance when S. cerevisiae grows on glucose [44,48], we tentatively conclude that this reaction is irreversible under the currently growth conditions. Consistently, ΔG°′ ≈ 10 kJ·mol−1 (http://wwwbmcd.nist.gov:8080/enzyme/ename.html) has been measured for the Thr cleavage, while ΔG°′ ≈ 20 kJ·mol−1 was obtained for the Ser cleavage reaction. This indicates that the Ser cleavage reaction tends to be more on the side of the intact amino acid than the Thr cleavage.

Gly can also be synthesized from a C1 unit and CO2 via mitochondrial GCV. This reaction generates increased f (1) {Gly-α} values [1], provided that mt-Gly is transported into the cytosol where the bulk part of the cellular protein is generated. As f (1){Gly-α} − f (1){Phe-α} = 0.05 and 0.19 for the aerobic and anaerobic preparation, respectively (Table 2; Fig. 2), we have evidence for (a) the reversible action of mitochondrial GCV in both the aerobic and the anaerobic regime, and for (b) a significant flux of mt-Gly into the cytosolic pool (we also find that f (1) {Ser-α} =f (1) {Gly-α}[  f (1){Ser-β} − f (1){Phe-β}], which is consistent with the view that cytosolic Ser and Gly are used for the generation of the bulk part of the cellular protein). We cannot determine the fraction of cyt-Gly arising from mt-Gly because the 13C labeling patterns of the latter are not available. However, the value of f (1){Gly-α} − f (1){Phe-α} constitutes a lower bound for cyt-Gly arising from mt-Gly, which would be reached only if all mt-Gly molecules were reversibly cleaved by the GCV. Considering that GCV probably cleaves only a minor fraction of the mitochondrial Gly pool, lower bounds of 5 and 19% in the aerobic and anaerobic regime, respectively, indicate a significant flux of Gly from the mitochondria into the cytosol (Table 2). This finding is consistent with the earlier observation that mammalian cells with active cytosolic SHMT but inactive mitochondrial SHMT are glycine auxotrophs [58], i.e. the cytosolic Gly production is not sufficient to meet cellular demands.

Table 2.  Origin of intermediates in S. cerevisiae central carbon metabolism inferred from analysis of BDF 13C labeled amino acids. The fraction of the total pool of the metabolite indicated on the left is given. Values obtained for the cytosol, the mitochondria and C1 metabolism are listed separately. The abbreviations of the metabolites are given in the legend of Fig. 1, and the Eqns employed to calculate the fractions are given in the corresponding footnotes. Entries containing NA could not be obtained for the present system due to degeneracy of f values (see text; Table 1). Experimental errors have been estimated from analysis of redundant 13C scalar coupling fine structures [1] and the signal-to-noise ratio of the cross peaks in the 2D 13C-1H COSY spectra (Fig. 4). Ranges of fractions were derived either from bounds, or from multiple determinations employing the balancing Eqns (11), (14) and (15) for different f (i) values (or linear combinations thereof). For the definition of the flux ratios, X, see Fig. 3. According to these definitions, the fractions inline imageinline image and inline image, which can readily be derived from the corresponding X values given in this table, represent ‘cyt-Prv from mt-Prv’, ‘cyt-AcCoA from mt-AcCoA’ and ‘cyt-OxAc from mt-OxAc’, respectively. (1 − inline image) (1 − inline image) and (1 − inline image) represent ‘mt-Prv from cyt-Prv’, ‘mt-AcCoA from cyt-AcCoA’ and ‘mt-OxAc from mt-OxGlt’, respectively.
 Fraction of total pool (%)
  1. a   Eqn (11); b Eqn (13); c Eqn (15); d Eqn (3); e Eqn (7); f Eqn (8); g Eqn (9); h Eqn (12); k Eqns (1) and (14); and l Eqn (2b)

 P-Prv from pentoses (upper bound) 0–4 0–4
 Ri5P from glucose (lower bound)32 ± 225 ± 2
 Ri5P from G3P and S7P (transketolase reaction)68 ± 275 ± 2
 Ri5P from E4P (transketolase and transaldolase reactions)10 ± 211 ± 2
 E4P from fructose (lower bound)15 ± 417 ± 4
 cyt-Prv from aP-Prv ( inline image)70–10050–100
 cyt-AcCoA from b cyt-Prv ( inline image)NANA
 cyt-OxAc from c cyt-Prv ( inline image)88–10071–100
 cyt-OxAc reversibly converted d to fumarate (cyt-Xexch) 0–4 0–4
 mt-Prv from malate (X1mt)
 (upper e bound I, Xub-I)
 (lower f bound I, Xlb-I)
 (lower g bound II, Xlb-II)
≈ 25
 ≈ 5
≈ 30
 mt-AcCoA from h mt-Prv (X2mt)NANA
 mt-OxAc from kPPrv (Xana/X3mt)76 ± 492–100
 mt-OxAc reversibly l converted to fumarate (mt-Xexch)35 ± 417 ± 4
 C1 metabolism
 Ser from Gly and C1 unit19 ± 219 ± 2
 Gly from CO2 and C1 unit 4 ± 319 ± 3
 Cyt-Gly from mt-Gly > 4> 19

Novel insight from the 13C labeling experiment was also obtained with respect to the location of the pyruvate pool serving for Ala biosynthesis: the f values of Ala-α (Fig. 2) reflect the action of the malic enzyme (see below; Table 1) which itself has been shown to be exclusively localized in the mitochondria [34]. Our data thus show that alanine aminotransferase is located in the mitochondria.

Overall, the biosynthetically directed fractional 13C labeling experiment yielded the assignment of amino-acid 13C-fine structures to carbon atoms in intermediates as well as their subcellular locations as shown in Fig. 2. This allows an investigation of the central carbon metabolism of S. cerevisiae, and in particular the determination of metabolic flux ratios.

Cytosol: pentose phosphate pathway and glycolysis

Intermediates of the pentose phosphate pathway (PenPp) and glycolysis are entirely located in the cytosol, and the biosynthetic origin of these intermediates (Table 2) was determined as previously described for prokaryotic cells [1,6–9,11]. Firstly, the data show that the flux ratios of the PenPp are astonishingly similar when comparing the aerobic (glucose repressed), and the anaerobic growth regime. Secondly, the employment of PenPp for generation of PPrv is beyond detection (< 4%), i.e. PPrv stems from glycolysis only, suggesting an exclusively anabolic role of the PenPp when S. cerevisiae cells are grown under the current conditions. In fact, this is consistent with a previous study [59] estimating that S. cerevisiae catabolizes less than 1% of the glucose via PenPp to PPrv. Notably, the inability of glycolytic enzyme knock-out mutants to employ the PenPp for glucose degradation is apparently not due to an insufficient capacity of the pathway ensuring appropriate metabolic fluxes for glucose catabolism [60], but rather is a result of the lack of a transhydrogenase in S. cerevisiae[61] for equilibration of the NADH and NADPH pools. Thirdly, PenPp is fed via both the oxidative and nonoxidative branches. The activity of the oxidative branch [1,6,8] could be directly assessed from registering intact C5 fragments in Ri5P (Fig. 2,4) which yields lower bounds around 30% for pentose synthesis by glucose oxidation (Table 2). The operation of the nonoxidative branch is implied by a significant flux from fructose 6-phosphate to erythrose 4-phosphate (≈ 15%) catalysed by transketolase and exchange fluxes catalysed by transaldolase and transketolase combined with the notion that PPrv generation through PenPp is not detected.

Cytosol: glyoxylate cycle and pyruvate carboxylation/decarboxylation

The exchange fluxes between cyt-OxAc and cyt-fumarate, cyt-Xexch, calculated with Eqn (3), are zero in both the aerobic and anaerobic regime (Table 2). This exchange reaction is catalyzed by enzymes constituting the glyoxylate cycle. Hence, our data strongly support the previously published finding that the glyoxylate pathway enzymes [40–42], including the cytosolic malate dehydrogenase [43], are not expressed when S. cerevisiae cells grow under glucose repression.

f (3) {Asp-α} > 0.9 and f (2*) {Lys-α} > 0.9 show that the majority of the cyt-OxAc and cyt-AcCoA pools arise, respectively, from carboxylation and decarboxylation of cyt-Prv in both the aerobic and the anaerobic regime.

Mitochondria: TCA cycle and malic enzyme activity

The relative supply of the TCA cycle with either AcCoA for generation of reduction equivalents or OxAc to fulfill biosynthetic demands is a pivotal parameter characterizing the regulatory state of central metabolism. Employing Eqn (1), we obtained Xana= 0.76 and 1.00 for the aerobic and anaerobic regime, respectively (Table 2). In the anaerobic regime, the reactions constituting the TCA cycle are evidently operating in a branched fashion as it is typically observed for prokaryotic cells [1]. Hence, the flux from OxGlt to mt-OxAc is zero. Consistently, OxGlt dehydrogenase activity cannot be detected under anaerobic conditions [62,63]. The aerobic value of Xana= 0.76 is quite high when compared with that of prokaryotic cells [1,6,8,9,11,12,14], and reflects that S. cerevisiae cells grown under glucose repression generate a large fraction of the cellular ATP through fermentative metabolism. This implies a diminished role of ATP synthesis by respiration, so that the TCA cycle functions primarily for biosynthesis. Consistently, Polakis & Bratley [64] found reduced TCA enzyme activities for S. cerevisiae cells grown in the presence of 0.9% glucose, and Wales et al. [62] obtained similar results for Saccharomyces carlsbergensis.

A second parameter that can be readily obtained from BDF 13C labeling of amino acids is the exchange flux between mt-OxAc and mt-fumarate. As cyt-Xexch= 0 (see above), we obtained mt-Xexch= 0.35 and 0.17 for the aerobic and anaerobic regime, respectively, using Eqn (2).

In order to assess the activity of the malic enzyme, the f values of mt-OxAc were calculated with Eqns (46) (Table 3). Due to the prevalence of the anaplerotic generation of mt-OxAc, potential channeling of metabolites during interconversion of OxGlt to mt-OxAc [53–57] would have little impact (compare f values in Eqns 4–6 in Table 3), i.e. the following results are largely independent of the actual degree of metabolite channeling in the mitochondria. Employing Eqns (7), (8) and (9), we obtain Xub-I(mt-Prv ← mt-malate) = 0.25, Xlb-I(mt-Prv ← mt-malate) = 0.05 and Xlb-II(mt-Prv ← mt-malate) = 0.30 for the aerobic regime. As the first upper and the second lower bound are nearly equally large, i.e. Xub-I(mt-Prv ← mt-malate) ≈Xlb-II(mt-Prv ← mt-malate), we conclude that 25–30% of the mt-Prv pool is synthesized via the malic enzyme, i.e. X(mt-Prv ← mt-malate) ≈ 0.25–0.30. In turn, this indicates that the f values of mt-malate and mt-OxAc are very similar, if not identical. Consistently, ΔG°′≈−30 kJ·mol−1 (http://wwwbmcd.nist.gov:8080/enzyme/ename.html) has been measured for the interconversion of mt-OxAc into mt-malate, i.e. the production of malate is strongly favoured in this equilibrium. Under the anaerobic regime, Eqns (7) and (8) do not apply as f (2*) {mt-OxAc-C2} = 0. However, Eqn (9) yields Xlb-II(mt-Prv ← mt-malate) = 0.70. This rather high value is consistent with the observation that malic enzyme gene expression under anaerobic growth is about threefold increased at the transcriptional and about fourfold increased translational level when compared with the aerobic regime [34]. Obviously, the malic enzyme plays a pivotal role for mt-Prv metabolism in both the aerobic and anaerobic regime.

Table 3.  f values calculated for mitochondrial oxaloacetate (mt-OxAc). The values are calculated from f values observed for Glu (Table 1). Xana and mt-Xexch were calculated using Eqns (1) and (2b), respectively. Calculations for Eqns (4), (5) and (6) were based on conditions (a), (b) and (c), respectively. (a) Assuming complete metabolite channeling during synthesis of mt-OxAc from OxGlt and neglecting reversible exchange from mt-OxAc to mt-fumarate, (b) neglecting both metabolite channeling and the reversible exchange, and (c) neglecting metabolite channeling and considering the reversible exchange.
 Aerobic Xana= 0.76a mt-Xexch= 0.35Anaerobic Xana= 1.00 mt-Xexch= 0.17
Carbon f (1) f (2) f (2*) f (3) f (1) f (2) f (2*) f (3)
Eqns (4)        
Eqns (5)
Eqns (6)

Intercompartmental fluxes: pyruvate, acetyl-CoA and oxaloacetate

Fig. 3 shows the metabolic subnetwork that was selected to analyse the exchange fluxes of Prv, AcCoA and OxAc connecting cytosolic and mitochondrial pools (Table 2). For this network we have that inline image of Eqn (10) is equal to X(mt-Prv ← mt-malate) ≈ 25–30% in the aerobic and X(mt-Prv ← mt-malate) > 70% in the anaerobic regime (see above).

inline image, i.e. the fraction of cyt-Prv arising from PPrv, can be estimated using Eqn (11). Assuming that f (3) {cyt-Prv-C2} = f (3){cyt-OxAc-C2} one obtains inline image > ≈ 70% for the aerobic and inline image > ≈ 50% for the anaerobic regime. As f (3){cyt-Prv-C2} might be somewhat reduced by the combination of the flux from mt-OxAc to cyt-OxAc followed by the decarboxylation of cyt-OxAc (see below), these values constitute, in principle, lower bounds (Table 2). Although a more accurate determination of inline image is prevented for the current system by rather similar f (3) values, the data provide evidence that a large fraction of cyt-Prv is derived from PPrv in both regimes.

Recruiting Leu-α and Glu-γ to obtain f (2){mt-AcCoA-C1}, Lys-α to derive f (2){cyt-AcCoA-C1}, and Ala-α to get [  f (2) + f (2*)]{mt-Prv-C2}, employment of Eqn (12) shows that, for the present growth conditions, inline image cannot be obtained because of degeneracy of f values. Similarly, with f (2) {Lys-α}=f (2) {cyt-AcCoA-C1}=f (2) {Leu-α}=f (2) {mt-AcCoA-C1}, degeneracy of f values prevents from calculating inline image using Eqn (13). The identity of the 13C labeling labeling patterns indicates that the two AcCoA pools are either fully equilibrated, or that one is derived from the other. Future studies need to identify whether the cytosolic and mitochondrial AcCoA pools are possibly quite generally in rapid exchange due to the facilitated diffusion catalysed by the carnitine shuttle.

The mt-OxAc pool is characterized by the flux ratio Xana, which provides the fraction of mt-OxAc arising from PPrv via Prv. As the carboxylation occurs in the cytosol, Xana also establishes a ratio involving the transport flux of cyt-OxAc into the mitochondria (Eqns 14). The f values calculated for mt-OxAc (Table 3) allow determination of inline image using Eqn (15). Selecting f (3), which exhibits the largest difference when comparing Asp-α (Table 1), and mt-OxAc-C2 (Table 3) for calculation of inline image, we obtain that inline image > 0.88 for the aerobic and inline image > 0.71 in the anaerobic regime, i.e. the transfer of mt-OxAc into the cytosol is of minor importance for the generation of cyt-OxAc (Table 2). Hence, the transport of OxAc from the cytosol to the mitochondria appears to be largely unidirectional, and this finding is in agreement with the view that the OxAc carrier protein is actively transporting OxAc into the mitochondria by symport of protons. Moreover, the data provide evidence that the employment of the malate–aspartate shuttle [65–67], which serves for transfer of cytosolic NADH into the mitochondria, is small as it would be expected to export mt-OxAc, which is not detected (Table 1). The same holds for the succinate–fumarate carrier [68,69]: as the glyoxylate cycle is switched off, transport of succinate into the mitochondria (as well as concomitant export of mitochondrial fumarate into the cytosol) is apparently not required to balance cytosolic and mitochondrial demands of succinate.

Finally, we have not invoked the import of citrate (or isocitrate) from the cytosol into the mitochondria for data interpretation, although S. cerevisiae can express a mitochondrial citrate transfer protein [70]. However, as described for mammalian mitochondria [71], this protein primarily serves for catalysing citrate efflux into the cytosol in order to fulfill biosynthetic demands. As we find no indication of a substantial transfer of citrate that was generated in the cytosol into the mitochondria, our data support the above indicated functional assignment of the citrate transfer protein.

Comparison with previous analyses of E. coli cells

Previously reported METAFoR analyses of E. coli[1,9] were obtained under very similar growth conditions as for the S. cerevisiae cultivations described here, i.e. aerobic and anaerobic shake flask cultures and glucose-supplemented minimal media were used to study central carbon metabolism under exponential growth. Hence, these studies provide an attractive framework to compare the salient features of E. coli and S. cerevisiae central carbon metabolism when glucose is supplied as the sole carbon source.

Most strikingly, and in sharp contrast to E. coli, the PenPp in S. cerevisiae operates exclusively for generating NADPH and pentoses, i.e. virtually no PPrv is synthesized through PenPp. Moreover, S. cerevisiae exhibits a significantly higher fraction of mt-Prv originating from malate than E. coli, suggesting a higher in vivo activity of the malic enzyme. In both organisms, OxAc is entirely generated via anaplerosis in the anaerobic regime, illustrating a branched mode of operation for the TCA cycle. In the aerobic regime, however, the contributions of anaplerosis are rather different: about 75% and 30–50% contribution for OxAc synthesis were observed in S. cerevisiae and E. coli, respectively. This observation reflects the phenomenon of ‘glucose repression’ of the TCA cycle in S. cerevisiae, which is not observed for the E. coli. Furthermore, Prv cleavage through pyruvate–formate lyase, a characteristic feature of anaerobic E. coli metabolism, is not detected, simply because S. cerevisiae does not possess such a lyase. Finally, we found that the glyoxylate shunt is generally not active for both E. coli and S. cerevisiae when glucose is provided as the sole carbon source.


We have extended to application of BDF 13C labeling of proteinogenic amino acids, an efficient analytical tool to study intermediary metabolism [1], to the compartmented eukaryotic organism S. cerevisiae. This allowed (a) unravelling the network of biosynthetic pathways activated under glucose repression in both cytosol and mitochondria; (b) determination of a large number of metabolic flux ratios in the two compartments, and (c) characterization of intercompartmental transport fluxes. The wealth of new insights into compartmented yeast metabolism (Table 2) obtained from the present investigation complements previous 13C-NMR studies conducted for S. cerevisiae[72–76], and enhances our understanding of yeast metabolism in general [77].

In a manner previously demonstrated for prokaryotes [1,5–12,14], the current methodology paves the way for investigating the metabolic response of unicellular eukaryotes such as yeasts to genetic and environmental modulations. Future applications may include: (a) high-throughput analyses in the framework of metabolic engineering in biotechnology research [12,78,79]); (b) the joint employment of BDF 13C labeling with metabolic flux balancing (reviewed in [12]) in order to obtain reliable estimates for net fluxes [6]; and (c) the synergistic application of mass spectrometry for analysis of the 13C labeling patterns [12,80–82].

The rapid assessment of active biosynthetic pathways on the level of the metabolic intermediates afforded by BDF 13C labeling [1] is clearly of keen interest for future metabolic engineering efforts. The fact that this approach yields quantitative information about the in vivo malic enzyme activity in S. cerevisiae is a telling example. Engineered expression of malic enzyme may serve to improve food quality [83], and heterologous malic enzymes using both NADH and NADPH as cofactors may effectively act as transhydrogenases, thus expanding the enzymatic portfolio of S. cerevisiae. Evidently, such metabolic engineering projects would greatly profit from monitoring malic enzyme activity via determination of metabolic flux ratios.

Overall, ‘metabolic profiling’ using BDF 13C labeling of proteinogenic amino acids [1] may well be an outstandingly potent complement to both the ‘proteome approach’, aiming at the construction of complete protein maps [84], and transcriptome analysis for genomewide exploration of gene expression [85].


We thank Prof. Kurt Wüthrich for continuous support and again Dr M. Hochuli for helpful discussions. The work was supported by a start-up fund of the State University of New York to T. S., a postdoctoral fellowship of VTT Biotechnology and the National Technology Agency to H. M., and the Swiss Priority Program in Biotechnology.


  1. *Note: these authors contributed equally to this paper

  2. Enzymes: acetolactate synthase (EC; alanine aminotransferase (EC; carnitine O-acetyltransferase (EC; fumarase (EC; glycine aminotransferase (EC; glycine cleavage system (EC; homocitrate synthase (EC; α-isopropylmalate synthase (EC; malate dehydrogenase (EC; malic enzyme (EC; 2-oxoglutarate dehydrogenase (EC; pyruvate carboxylase (EC; pyruvate decarboxylase (EC; pyruvate dehydrogenase (EC; pyruvate formate-lyase (EC; serine hydroxymethyltransferase (EC; threonine aldolase (EC; transaldolase (EC; transketolase (EC

  3. Present address: Department of Biochemistry, Biozentrum, University of Basel, CH-4056 Basel, Switzerland