Short-Term Metabolome Dynamics and Carbon, Electron, and ATP Balances in Chemostat-Grown Saccharomyces cerevisiae CEN.PK 113-7D following a Glucose Pulse

The in vivo kinetics in Saccharomyces cerevisiae CEN.PK 113-7Dwas evaluated during a 300-second transient period after applyinga glucose pulse to an aerobic, carbon-limited chemostat culture.We quantified the responses of extracellular metabolites, intracellularintermediates in primary metabolism, intracellular free aminoacids, and in vivo rates of O₂ uptake and CO₂ evolution. Withthese measurements, dynamic carbon, electron, and ATP balanceswere set up to identify major carbon, electron, and energy sinksduring the postpulse period. There were three distinct metabolicphases during this time. In phase I (0 to 50 seconds after thepulse), the carbon/electron balances closed up to 85%. The accumulationof glycolytic and storage compounds accounted for 60% of theconsumed glucose, caused an energy depletion, and may have ledto a temporary decrease in the anabolic flux. In phase II (50to 150 seconds), the fermentative metabolism gradually becamethe most important carbon/electron sink. In phase III (150 to300 seconds), 29% of the carbon uptake was not identified inthe measurements, and the ATP balance had a large surplus. Theseresults indicate an increase in the anabolic flux, which isconsistent with macroscopic balances of extracellular fluxesand the observed increase in CO₂ evolution associated with nonfermentativemetabolism. The identified metabolic processes involving majorcarbon, electron, and energy sinks must be taken into accountin in vivo kinetic models based on short-term dynamic metabolomeresponses.

INTRODUCTION

Top

Introduction

Mathematical models of in vivo enzyme kinetics in microorganismsare important for understanding metabolic control mechanismsoperating on the level of the metabolome and can be used toassist the rational redesign of metabolic pathways to enhancedesired functionalities of microbes (54). Kinetic parametersin this kind of model can be obtained by stimulus-response experiments,in which cells grown in a (quasi-)steady state are perturbedby an external stimulus and the dynamic responses of intra-and extracellular metabolites are monitored. The time windowof observation is usually within tens to a few hundred secondsafter the application of the stimulus, and the responses areusually attributed to rapid (allosteric) enzyme-metabolite interactions(38). Kinetic parameters can be estimated from the measuredresponses, based on a set of (dynamic) material balances (7,30, 49), as follows: dx/dt = Sv (equation 1), where x is a vectorof the metabolite concentrations, S is the stoichiometry matrix,and v is a vector of the reaction rates as a function of (yet)unknown kinetic parameters. The stimulus-response methodologyis an ideal tool for obtaining kinetic information and has beenapplied to various microorganisms under different growth conditions(7, 26, 32, 38, 52).

For Saccharomyces cerevisiae, a frequently applied perturbationis the addition of a concentrated glucose solution, i.e., aglucose pulse, to a glucose-limited chemostat culture, therebyinducing a short-term Crabtree effect. Theobald et al. (38)first quantified for a 180-second postpulse period the responsesof extracellular metabolites (glucose, ethanol, acetate, andglycerol) and intracellular intermediates of and cofactors involvedin glycolysis. These measurements were used to establish a kineticmodel focused on glycolysis (30). In addition, Vaseghi et al.(49, 50) quantified responses of the pentose phosphate pathwayand the protein kinase A/phosphofructokinase 2 (PKA/PFK2) signaltransduction cascade to such a glucose pulse to extend thiskinetic model. Models developed so far incorporate detaileddescriptions of metabolic subnetworks, such as glycolysis, forwhich metabolome responses have been quantified. Other metabolicprocesses, such as the tricarboxylic acid (TCA) cycle, storagecarbon metabolism, and anabolism, are represented by lumpedreactions with highly simplified kinetics due to a lack of experimentalinformation.

The importance of various metabolic processes in a kinetic modelcan be evaluated by the relative contributions of metabolitesinvolved in these processes to the overall material balance,which forms the basis for kinetic parameter estimation (equation1). Under quasi-steady-state conditions, for example, materialbalances of extracellular metabolites are required for metabolicflux analysis (33). Previous work has evaluated dynamic carbonbalances in a time frame of tens of minutes to hours (14, 26,45), based on measured extracellular fluxes and an assumed intracellularquasi-steady state. Under dynamic conditions and within thetime frame of interest for in vivo kinetic analysis, settingup material balances requires comprehensive quantification ofextra- and intracellular metabolite concentrations as well asthe in vivo O₂ uptake rate and CO₂ evolution rate.

The objectives of this study were (i) to quantify an extendedrange of metabolome responses following a glucose pulse to anaerobic, carbon-limited chemostat culture of S. cerevisiae,(ii) to construct complete carbon, electron, and energy balancesduring a 300-second postpulse period from these measurements,(iii) to identify major carbon/electron sinks and energy production/consumptionprocesses in this period, and (iv) to determine if major unquantifiedmetabolic processes remain. This information will increase ourunderstanding of short-term physiological responses of S. cerevisiaeto glucose excess and will assist in building more realisticand comprehensive kinetic models.

MATERIALS AND METHODS

Top

Materials and Methods

Yeast strain and chemostat cultivation.
The haploid S. cerevisiae strain CEN.PK 113-7D was cultivatedin an aerobic, carbon-limited chemostat with a working volumeof 4 liters at a dilution rate of 0.051 h^–1, as previouslydescribed (18). The medium used for chemostat cultivation wasbased on a previously described medium (18), with 0.15 M (27.1g liter^–1) glucose and 0.03 M (1.42 g liter^–1) ethanolto obtain a biomass of $~$ 15 g dry weight liter^–1 duringsteady state. The dissolved oxygen tension (DOT) in the fermentorand the O₂ and CO₂ contents in the fermentor off-gas were analyzedas previously described (55). The DOT was maintained above 70%in the steady state.

Rapid sampling, quenching, and metabolite extraction.
Culture supernatant for the analysis of extracellular metaboliteswas obtained by rapidly quenching the broth with stainless steelbeads (4-mm diameter) at –20°C, followed by filtration,as described previously (21), for a sample volume of 2 ml. Foranalysis of intracellular metabolites, sampling and sample preparationwere carried out as previously described (22). Briefly, $~$ 1 mlof broth was rapidly quenched in 5 ml of –40°C 60:40(vol/vol) methanol-H₂O. After centrifugation (2,000 x g, –20°C,5 min), the supernatant was decanted, and the pellet was resuspendedin 5 ml of –40°C 60:40 (vol/vol) methanol-H₂O andcentrifuged again. Intracellular metabolites were extractedfrom the pellet with boiling 75:25 (vol/vol) ethanol-H₂O. Theethanol extracts were evaporated to dryness. The dried sampleswere resuspended in 0.5 ml deionized water and centrifuged (3,000x g, 4°C, 5 min). The supernatant and pellet were separatedand stored at –80°C until they were analyzed.

Determination of culture dry weight.
Culture samples (5 ml, in triplicate) were filtered throughpredried, tared nitrocellulose filters (pore size, 0.45 µm;Gelman Science, Ann Arbor, MI). Filters were washed twice with5 ml demineralized water, dried at 70°C for 48 h, and weighed.

Extracellular metabolite analysis.
The amounts of glucose, acetate, and glycerol in the supernatantwere measured with Enzytec kits (Scil Diagnostics GmbH, Martinsried,Germany). The concentration of ethanol in the supernatant wasanalyzed with an Enzymatic Bioanalysis kit (R-Biopharm AG, Darmstadt,Germany). Absorbance was read on an Agilent 8453 UV-visiblespectroscopy system (Agilent Technologies Deutschland GmbH,Waldbronn, Germany).

MS analysis of intracellular metabolites.
The concentrations of glucose 6-phosphate (G6P), fructose 6-phosphate(F6P), fructose 1,6-bisphosphate (F1,6bP), 2-phosphoglycerate(2PG), 3-phosphoglycerate (3PG), and phosphoenol pyruvate (PEP)in the cell extract were analyzed by liquid chromatography-electrosprayionization-tandem mass spectrometry (LC-ESI-MS/MS) accordingto the method of van Dam et al. (40). The same analysis methodand conditions were applied for the analysis of pyruvate (Pyr),citrate (Cit), isocitrate (iCit), ${alpha}$ -ketoglutarate ( ${alpha}$ KG), succinate(Suc), fumarate (Fum), malate (Mal), glucose 1-phosphate (G1P),mannose 6-phosphate (M6P), trehalose 6-phosphate (T6P), fructose2,6-bisphosphate (F2,6bP), and orthophosphate (P_i) in the cellextract. Neither 2PG and 3PG nor Cit and iCit could be resolvedwith this procedure, so only the sums of these compounds weredetermined. For F2,6bP, only the peak area was quantified insteadof the absolute amount in the sample.

The concentrations of the adenine nucleotides (AMP, ADP, andATP) in the cell extract were quantified by LC-ESI-MS/MS. Thenucleotides were separated by an ion-pairing reversed-phasehigh-performance liquid chromatography (HPLC) method adaptedfrom a previously described method (8), using an XTerra MS C₁₈column (100 mm x 1 mm) equipped with a guard column (10 mm x2.1 mm) (both from Waters, Milford, MA). The standard solutionand sample injection volume was 10 µl; all standard solutionsand samples were mixed 24:1 with a 2 M solution of the ion-pairingreagent tetrabutylammoniumacetate (Sigma-Aldrich, Steinheim,Germany) to obtain a final tetrabutylammoniumacetate concentrationof 80 mM. Chromatography was performed at 20°C with an AllianceHT 2795 pump system (Waters), giving an isocratic elution flowof 0.1 ml min^–1. The mobile phase was a 10 mM NH₄H₂PO₄solution that was adjusted to pH 6.4 with NH₄OH, after which2 mM tetrabutylammoniumhydroxide (Sigma-Aldrich) was added,giving a pH of 6.8, followed by the addition of 15% (vol/vol)acetonitrile solution. The postcolumn eluent was mixed with0.1 ml min^–1 80% (vol/vol) acetonitrile solution. Thesolutions were filtered (0.45 µm; Gelman Sciences) priorto use to remove particles and bubbles. The HPLC instrumentwas coupled to an MS/MS instrument (Quattro Ultima Pt; MicromassLtd., Manchester, United Kingdom) via a flow splitter whichsent a flow of 0.06 ml min^–1 to the ESI instrument, whichwas operated in positive mode with a nebulizer gas (nitrogen)flow of 50 liter h^–1 and a desolvation gas (nitrogen)flow of 500 liter h^–1 at 300°C. The source block temperaturewas 120°C. The capillary voltage was set at 3.5 kV, andthe cone voltage was set at 35 V. The [M + H]⁺ ions (m/z 348,428, and 508 for AMP, ADP, and ATP, respectively) were fragmentedby collision-induced dissociation with argon (collision energy,24 V). The strongest daughter fragment of m/z 136 was monitoredfor each of them.

Intracellular trehalose analysis.
Intracellular trehalose was analyzed by quantitative ¹H nuclearmagnetic resonance. Analysis was performed at 360 MHz on a BrukerAMX 360 spectrometer (Bruker Analytik, Karlsruhe, Germany).To 0.5 ml of cell extract, an equal amount of a standard solutioncontaining 5.6 g liter^–1 maleic acid and 10 g liter^–1EDTA (natrium salt) was added. After lyophilization, the residuewas dissolved in 100% D₂O, and the ¹H nuclear magnetic resonancespectrum was measured, with a relaxation delay of 30 secondsto ensure full relaxation of all hydrogen atoms between pulses.The integrals of the ¹H protons of trehalose (doublet at 5.24ppm) and the internal standard (singlet at 6.1 ppm) were measured,and the content of trehalose was calculated.

Intracellular glycogen analysis.
Intracellular glycogen was measured in the cell extracts andthe insoluble residues as previously described (27). Glycogenin the cell extracts was directly hydrolyzed with amyloglucosidase(from Aspergillus niger; Sigma-Aldrich) without alkaline extraction.

Intracellular free amino acid analysis.
Intracellular concentrations of free amino acids were measuredby HPLC using the AccQ-tag system, as previously described (53).

Estimation of extracellular P_i carryover to cell extracts.
About 0.25 ml of the 5 ml of quenching or washing solution remainedin the test tube after decanting, which corresponds to a washingefficiency of 1 – (0.25/5)², or 99.75%, and a carryoverof 0.25% after the quenching and washing steps. A steady-stateextracellular P_i concentration of 39.3 mM was calculated basedon the biomass phosphate content (19) and the medium composition.Assuming a constant extracellular P_i concentration during thepostpulse period, the carryover is no more than 8% of the measuredP_i concentration in the cell extract.

Glucose pulse experiment.
S. cerevisiae was grown for 10 generations in an aerobic, carbon-limitedchemostat, and 15 ml of a glucose solution (1.48 M, thus containing22.2 mmol of glucose) was injected into the fermentor. Samplesfor extra- and intracellular metabolite analyses were takenduring the steady state shortly before the glucose pulse andduring the 300 seconds following the glucose pulse. During thepostpulse period, the steady-state feed to the fermentor wascontinued, but no effluent was removed via the effluent pumpsince the chemostat was weight controlled and the fermentorweight decreased slightly due to sampling.

Steady-state metabolic flux analysis.
The measured extracellular fluxes in the steady state were usedafter reconciliation (41) to estimate intracellular fluxes byconventional metabolic flux analysis (33). The biomass compositiondetermined for a growth rate of 0.05 h^–1 was used (19),and 1 Cmol biomass (equivalent to the amount of biomass thatcontains 1 mol of carbon) corresponds to 26.15 g cell dry weight.The noncompartmented network of S. cerevisiae was based on previouswork (34). We assumed that acetaldehyde dehydrogenase (AADH)is NADP⁺ specific and that isocitrate dehydrogenase is NAD⁺specific. The inclusion of acetyl coenzyme A synthetase createsa parallel route for the synthesis of acetyl coenzyme A frompyruvate, which can proceed via either pyruvate dehydrogenaseor the pyruvate bypass (Fig. 1). In the analyzed network, NADPHneeded for growth was presumed to be regenerated by AADH andthe pentose phosphate (PP) pathway; hence, intracellular fluxescan be calculated by specifying the PP pathway flux (or, equivalently,the PP split ratio). To estimate the PP split ratio, specificNADPH regeneration (mmol of NADPH g dry weight^–1) viathe PP pathway was assumed to decrease linearly with increasingNADPH regeneration by metabolizing ethanol via the NADP⁺-dependentAADH in a glucose-ethanol mixed-substrate feed when the ethanolfraction was below 0.4 Cmol of ethanol Cmol^–1. Above thislimit, the metabolic network would change due to the inductionof gluconeogenic enzymes (9, 44). For the Saccharomyces strainused in this study, the PP split ratio was estimated to be 0.44and 0.24 in two ¹³C-labeling studies (3, 48), for ethanol fractionsof 0 and 0.28 Cmol ethanol Cmol^–1, respectively, at agrowth rate of 0.1 h^–1. For the feed used in this study(0.06 Cmol ethanol Cmol^–1), a PP split ratio of 0.31 wasinterpolated and used to estimate the intracellular fluxes.

View larger version (13K):
[in this window]
[in a new window]
FIG. 1. Estimated intracellular fluxes in the reference steady state (normalized against a glucose influx of 0.013 mol Cmol_x^–¹ h^–1).

Estimation of q_O₂ and q_CO₂.
The in vivo O₂ uptake rate (q_O₂) and the CO₂ evolution rate(q_CO₂) during the postpulse period were calculated with a masstransport model as described previously (5). The model describesliquid/gas mass transfer in the fermentor, mixing/delay effectsin the headspace and tubing system, and sensor dynamics. Theparameters of this model were either derived from physical knowledgeor computed by fitting the model to an identification data set.The q_O₂ and q_CO₂ values during the postpulse period were calculatedfrom the off-gas and DOT measurements and the model by extendedKalman filtering. The cumulative O₂ uptake and CO₂ evolutionrates were obtained by numerically integrating the obtainedq_O₂ and q_CO₂ values.

Construction of cumulative carbon, electron, and ATP balances.
The cumulative carbon balance, in Cmol per Cmol of biomass (CmolCmol_x^–¹), compares the cumulative uptake of carbon sourceswith the accumulation of known carbon sinks, calculated at eachsampling time point relative to time zero, when the glucosepulse was given (14). During the postpulse period, glucose (addedvia both the pulse and the feed) was the sole carbon source,assuming that the small amount of ethanol in the feed was notmetabolized. The known carbon sinks were secreted extracellularmetabolites, intracellular metabolites, CO₂, and flux towardsanabolism. The anabolic flux was assumed to be equal to thebiomass growth rate. Since possible changes in the growth rateduring the postpulse period are difficult to quantify accurately,we assumed that it was the same as in steady state, i.e., 0.05h^–1. The adenine nucleotides were excluded from the carbonbalances, assuming that the adenine moiety was conserved duringthe short time window. A constant total broth volume was alsoassumed, since within 300 seconds, effluent via sampling ( $~$ 0.1liter) and influent via the feed pump (0.02 liter) were negligiblecompared to the total broth volume (4 liters). These assumptionswere also made for the cumulative electron and ATP balances.

The cumulative electron balance (in mol Cmol_x^–¹) was setup in a manner similar to that of the cumulative carbon balance,but with the degree of reduction of the metabolites (24), andit compares the cumulative uptake of electron sources with theaccumulation of known electron sinks. During the postpulse period,glucose was the sole electron source. The known electron sinksinclude the carbon sinks and O₂.

The cumulative ATP balance (in mol ATP equivalents Cmol_x^–¹)(Table 1) was also set up in a manner similar to that of thecumulative carbon balance. The P/O ratio (the ratio of ATPsproduced to the number of atmospheric oxygen atoms convertedto water) was assumed to be constant during the postpulse period.Intermediates in the PP pathway and the TCA cycle, as well asfree amino acids, were excluded from the ATP balance becausethey account for only a very small proportion (<3%) of thetotal carbon uptake.

View this table:
[in this window]
[in a new window]
TABLE 1. Energy production or consumption associated with accumulation of metabolites

RESULTS

Top

Results

Changes in metabolic fluxes and intracellular metabolites following a glucose pulse.
Three independent glucose pulse experiments yielded highly reproducibletrends in metabolic responses. The average variations in theaccumulation (in mCmol Cmol_x^–¹) of extracellular carbonsinks (e.g., ethanol) and groups of intracellular carbon sinks(e.g., intracellular metabolites in glycolysis) (Table 2) relativeto the steady-state levels were within 10% and 15%, respectively,in these experiments. Results from one representative experimentare presented here and were used for setting up material balances.Prior to the glucose pulse, the culture of S. cerevisiae wasmaintained in a reference steady state at a growth rate of 0.05h^–1 (Fig. 1; Table 2). Perturbation of this referencestate by a glucose pulse at time zero, which increased the residualglucose concentration from 0.15 mM to 5.56 mM, led to dynamicsin the biomass-specific glucose uptake rate (q_glucose) (Fig.2A), the secretion of ethanol, acetate (Fig. 2B), and a smallamount of glycerol, and dynamic changes in the biomass-specificrates of O₂ uptake and CO₂ evolution (q_O₂ and q_CO₂, respectively)(Fig. 2C). The respiratory quotient (RQ) was >1 during thepostpulse period, consistent with respiratory fermentative metabolism(Fig. 2D).

View this table:
[in this window]
[in a new window]
TABLE 2. Steady-state concentrations of different classes of intracellular metabolites

View larger version (14K):
[in this window]
[in a new window]
FIG. 2. Estimated biomass-specific fluxes of extracellular metabolites. (A) q_glucose; (B) q_ethanol ( ${square}$ ) and q_acetate ( ${blacklozenge}$ ); (C) q_O₂ (solid line) and q_CO₂ (dotted line); (D) RQ. The vertical dotted lines indicate the three metabolic phases.

Nonfermentative CO₂ evolution and O₂ uptake provide a measureof changes in the flux towards respiratory, nonfermentativemetabolism, i.e., the conversion of glucose to CO₂ and biomasscomponents coupled to NADH generation in lower glycolysis andthe TCA cycle. The q_O₂ and q_CO₂ values associated with nonfermentativemetabolism can be calculated, along with the apparent RQ, fromestimated extracellular fluxes, as follows:

Formula

Formula
assuming an NADP⁺-dependent acetaldehyde dehydrogenase(Fig. 3). Notably, both q_O₂ and q_CO₂ are closely related toflux through the TCA cycle, which is a major source of NADHand nonfermentative CO₂ and accounts for 69% and 82% of thesteady-state q_O₂ and q_CO₂, respectively.

View larger version (13K):
[in this window]
[in a new window]
FIG. 3. Estimated nonfermentative q_O₂ (+) and q_CO₂ ( ${square}$ ) (upper panel) and apparent RQ (lower panel). The vertical dotted lines indicate the three metabolic phases.

We divided the 300-second postpulse period into three sequentialmetabolic phases that are characterized by distinct metabolicresponses for a number of metabolites (see Fig. 4 to 7).

View larger version (18K):
[in this window]
[in a new window]
FIG. 4. Dynamic responses of intermediates in glycolysis, the PP pathway, and storage carbon metabolism and thence-derived biosynthetic precursors, in x-fold changes relative to the reference steady state. (A) G6P ( ${square}$ , left axis), F6P (+, left axis), and F1,6bP ( ${blacklozenge}$ , right axis); (B) 2PG + 3PG (+) and PEP ( ${square}$ ); (C) G1P (+, left axis) and T6P ( ${square}$ , right axis); (D) His (+), Phe ( ${square}$ ), and Tyr ( ${blacklozenge}$ ); (E) 6PG; (F) trehalose. The relative standard deviation of metabolite concentrations at each time point is within 10%, based on duplicate determinations. The vertical dotted lines indicate the three metabolic phases.

View larger version (9K):
[in this window]
[in a new window]
FIG. 7. Dynamic response of F2,6bP (in x-fold changes relative to the reference steady state). The vertical dotted lines indicate the three metabolic phases.

In phase I (0 to 50 seconds after the glucose pulse), highlydynamic changes occurred in metabolic fluxes and intracellularmetabolite concentrations. q_glucose increased to 17x the steady-stateflux (q_glucose⁰) at 10 seconds, followed by a sharp drop to $~$ 3x q_glucose⁰ at 50 seconds. Similar dynamic patterns were foundfor q_O₂ and q_CO₂. Both nonfermentative q_O₂ and q_CO₂ decreasedafter an initial increase, indicating a similar change in fluxthrough the TCA cycle. Ethanol production in this phase wasnot significant and was much lower than q_acetate. Intracellularmetabolite responses in phase I were characterized by accumulationsin the hexose phosphate pool, T6P, and trehalose (relative toalready high steady-state levels) (Table 2) and decreases in6PG and the phosphorylated C₃ pool, i.e., 2PG, 3PG, and PEP(Fig. 4). In addition, peaks in the concentrations of pyruvateand several TCA cycle intermediates were observed (Fig. 5).The ATP concentration and the energy charge, defined as (ATP+ 1/2 ADP)/(ATP + ADP + AMP), sharply decreased (Fig. 6). TheADP and AMP concentrations both decreased after an initial increase.

View larger version (18K):
[in this window]
[in a new window]
FIG. 5. Dynamic responses of pyruvate, TCA cycle intermediates, and thence-derived free amino acids, in x-fold changes relative to the reference steady state. (A) Pyr; (B) Cit + iCit (+) and ${alpha}$ KG ( ${square}$ ); (C) Suc (+), Fum ( ${square}$ ), and Mal ( ${blacklozenge}$ ); (D) Ala (+) and Val ( ${square}$ ); (E) Glm (+) and Gln ( ${square}$ ); (F) Asn (+) and Asp ( ${square}$ ). The relative standard deviation of metabolite concentrations at each time point is within 10%, based on duplicate determinations. The vertical dotted lines indicate the three metabolic phases.

View larger version (13K):
[in this window]
[in a new window]
FIG. 6. (Upper panel) Dynamic responses of ATP (+, left axis), ADP ( ${square}$ , left axis), and AMP ( ${blacklozenge}$ , right axis), in x-fold changes relative to the reference steady state. (Lower panel) Change in energy charge relative to the steady state. The energy charge in the steady state is 0.86, as calculated from the steady-state ATP, ADP, and AMP concentrations in Table 2. The relative standard deviation of metabolite concentrations at each time point is within 10%, based on duplicate determinations. The vertical dotted lines indicate the three metabolic phases.

In phase II (50 to 150 seconds after the pulse), gradual changesin the extracellular fluxes and intracellular metabolite concentrationswere observed, and q_glucose increased from 3x to 8x q_glucose⁰.Fermentative metabolism began, as evidenced by strong increasesin q_ethanol, q_CO₂, and the RQ. The ATP level and the energycharge recovered due to increased ATP generation in both fermentativemetabolism, i.e., increased q_ethanol, and oxidative phosphorylation,i.e., increased q_O₂. In the cell, the concentrations of thehexose phosphates decreased, whereas those of the TCA cycleintermediates, T6P, and F2,6bP (Fig. 7) increased gradually.

In phase III (150 to 300 seconds after the pulse), the extracellularfluxes and metabolite concentrations remained rather constant.The magnitude of metabolite concentration changes, i.e., |dx/dt|,was negligible compared to the magnitude of extracellular fluxes,indicating that an intracellular quasi-steady state had beenestablished. During phases II and III, nonfermentative q_O₂ andq_CO₂ continued to increase, up to 1.5x and 1.7x their steady-statevalues, respectively, indicating an increase in flux throughthe TCA cycle. The apparent RQ remained close to 1.

No significant changes were observed in intracellular free aminoacid concentrations in all three phases (Fig. 4 and 5). Theonly exception was aspartate, with a 2x decrease in phase IIIrelative to the reference steady state.

Cumulative carbon, electron, and ATP balances.
The available intra- and extracellular metabolite measurementswere combined with the in vivo q_O₂ and q_CO₂ values to constructthe cumulative carbon, electron, and ATP balances. The measuredglycogen concentrations in the insoluble residues and the cellextracts were $~$ 3.1 mCmol Cmol_x^–¹ and $~$ 0.5 mCmol Cmol_x^–¹,respectively, which is much lower than the glycogen contentof 102 mCmol Cmol_x^–¹ reported for S. cerevisiae grownat 0.05 h^–1 (13). The alkaline extraction seemed to beincomplete, presumably due to the use of insoluble residue obtainedafter boiling ethanol treatment rather than making the extractiondirectly from yeast cells, as previously described (27). Hence,we considered the glycogen measurements unreliable and excludedthem from the balance calculations. The cumulative carbon, electron,and ATP balances (Fig. 8) were evaluated for each of the threemetabolic regimens during the postpulse period.

View larger version (13K):
[in this window]
[in a new window]
FIG. 8. Cumulative carbon (A), electron (B), and ATP (C) balances. Symbols: +, carbon/electron source and ATP production; ${square}$ , carbon/electron sink and ATP consumption. The vertical dotted lines indicate the three metabolic phases.

For phase I (0 to 50 seconds), $~$ 85% of the measured carbon andelectron uptake can be accounted for by the known carbon andelectron sinks (Tables 3 and 4). More than 60% of the consumedglucose was found in (phosphorylated) intracellular metabolitesin upper glycolysis and storage carbon metabolism, while 20%was converted to secreted compounds (ethanol, acetate, and CO₂)and drained towards anabolism. The ATP balance (Table 5) showsthat the ATP consumption exceeded its production, which is consistentwith the sharp drop in the ATP level and energy charge. However,by the end of phase I, the resulting energy depletion (1.2 mmolATP equivalents Cmol_x^–¹) largely exceeded the observedtotal decrease in ATP (0.1 mmol ATP Cmol_x^–¹). Since anabolismrepresents a major fraction of ATP consumption, but only a minorfraction of the carbon balance, the calculated difference betweenATP consumption and production might indicate a temporary decreaseof the anabolic flux (q_anabolism).

View this table:
[in this window]
[in a new window]
TABLE 3. Relative contributions of different carbon sinks to cumulative glucose uptake relative to time zero

View this table:
[in this window]
[in a new window]
TABLE 4. Relative contributions of different electron sinks to cumulative electron uptake relative to time zero

View this table:
[in this window]
[in a new window]
TABLE 5. Cumulative energy generation and consumption relative to time zero

In summary, in phase I intense dynamics in glycolysis lead toelevated pools of phosphorylated compounds and storage material,leading to a large energy drain and possible decrease in q_anabolism.Only 2% of the consumed glucose was converted to ethanol. Thisresult suggests that in biomass-directed industrial applicationsof S. cerevisiae, undesired ethanol formation due to inhomogeneousdistribution and mixing in large-scale fermentors (29) couldbe reduced if the mixing time, i.e., the maximum time that yeastcells encounter a local glucose excess, is reduced to <50seconds.

In phase II (50 to 150 seconds), a gap between the carbon/electronsources and sinks emerged at about 100 seconds. At 150 seconds,26% of the consumed glucose had been channeled into unquantifiedsinks, and another 26% was involved in fermentative metabolism,which was the main carbon and electron sink at the end of phaseII. With the onset of fermentative metabolism, q_O₂ increasedand ATP generation accelerated. The ATP balance shows a slightATP excess at 150 seconds, which corresponds to the measuredincrease in the ATP concentration and the recovery of the energycharge.

In phase III (150 to 300 seconds), the difference between thecarbon/electron sources and sinks continued to increase. At300 seconds, 29% of the consumed glucose was channeled intounquantified sinks. Fermentative metabolism was the largestcarbon and electron sink, with ethanol, acetate, and the relatedCO₂ evolution accounting for 36% of the total glucose uptakeat the end of phase III. The ATP balance has a large energysurplus, which indicates that the unquantified carbon/electronsinks might be related to an unidentified energy-consuming process.

Estimation of q_anabolism in phase III.
The amount of carbon/electrons and energy withdrawn by anabolicprocesses is constant in the balance calculations due to anassumed constant growth rate of 0.05 h^–1 during the postpulseperiod. An increase in q_anabolism could potentially accountfor the missing carbon and electrons and the surplus ATP equivalents.However, such an increase would produce a practically undetectablechange in the biomass dry weight, since even assuming that allof the carbon that had not been accounted for was convertedto biomass, it would increase the dry weight only $~$ 7.6 mCmol_xliter^–1 (0.2 g dry weight liter^–1), relative toa steady-state biomass concentration of 573.6 mCmol_x liter^–1(15 g dry weight liter^–1).

To quantify possible changes in q_anabolism in phase III, weused a reaction network model, which contains six macroscopicprocesses (Table 6) and assumes a quasi-steady state in phaseIII. This type of model has been used successfully to describecatabolic and anabolic processes in postpulse periods lastingseveral hours (12). The quasi-steady-state fluxes were estimatedfrom metabolite concentration profiles at the end of phase III(Table 6). We used q_ethanol, q_acetate, q_glycerol, q_O₂, and twometabolite balances, i.e., CO₂ and NADH (assuming a constantintracellular NADH concentration), to calculate the remainingtwo unknown fluxes, q_anabolism and q_catabolism. The calculatedq_anabolism corresponds to a growth rate of 0.14 h^–1, relativeto a steady-state growth rate of 0.05 h^–1. Thus, an increasedq_anabolism is consistent with the measured extracellular fluxesand the known macroscopic stoichiometry.

View this table:
[in this window]
[in a new window]
TABLE 6. Stoichiometry of macroscopic model and estimated fluxes at the end of phase III

The measured glucose uptake rate and the ATP balance were notutilized in the flux calculation and can be used as an independentcheck of the model. Two different q_anabolism values were compared;one corresponds to an accelerated growth rate of 0.14 h^–1,and the other corresponds to a steady-state growth rate of 0.05h^–1 (Table 7). Assuming increased q_anabolism, both thecarbon and electron recovery improved by 16% to 85% at the endof phase III. In addition, the increase in q_anabolism can beenergetically sustained, since ATP production now closely matchesits consumption, while a q_anabolism value corresponding to anunchanged growth rate would result in a large ATP surplus. Thus,an increased rather than unchanged q_anabolism is consistentwith the ATP balance; moreover, this increase is consistentwith increased flux through nonfermentative respiratory metabolismin phases II and III, as indicated by increased q_O₂, nonfermand q_CO₂, nonferm.

View this table:
[in this window]
[in a new window]
TABLE 7. Comparison of carbon, electron, and ATP balances in phase III under two assumptions^a

Intracellular free and bound phosphate.
Intracellular phosphate exists in the free ion form (P_i) orcan be bound in phosphorylated pathway intermediates, nucleosidephosphates, and pyro- and polyphosphate. We constructed a balancefor the sum of free and bound intracellular phosphates withmeasured concentrations of adenine nucleotides, phosphorylatedintermediates, and intracellular P_i (Fig. 9). Increased concentrationsof phosphorylated intermediates are compensated for by decreasedconcentrations of adenine nucleotides and P_i, leading to a relativelyconstant amount of the measured free and bound phosphate inthe cell during the postpulse period. This implies that thehydrolysis of pyro- or polyphosphate is not significant duringthe postpulse period for energy generation.

View larger version (10K):
[in this window]
[in a new window]
FIG. 9. Balance of intracellular free and bound phosphate. The graph shows total free and bound phosphates present in measured intracellular metabolites ( ${circ}$ ), P_i ( ${blacksquare}$ ), phosphate bound in adenine nucleotides (+), and phosphate bound in pathway intermediates ( ${blacklozenge}$ ).

DISCUSSION

Top

Discussion

Comparison with previous glucose pulse experiments.
Responses to glucose excess have been measured for differentstrains of S. cerevisiae grown in glucose-limited chemostats(38, 45, 52). The time window of observation can be severalhundred seconds to several hours. The measured responses ofextracellular metabolites, intracellular glycolytic intermediates,and adenine nucleotides in this study are qualitatively comparablewith previous observations of short-term glucose excess (5,38, 52). The rapid increase in glucose influx in phase I, from0.013 to 0.22 mol Cmol_x^–¹ h^–1 10 seconds after thepulse, is further consistent with a calculated maximal fluxof 0.21 mol Cmol_x^–¹ h^–1 based on the Michaelis-Mentenkinetic parameters estimated from zero-trans experiments (10).

A number of metabolic responses observed in this study weredifferent from those observed in similar glucose pulse experimentswith S. cerevisiae. In phase I, an initial decrease of the 6PGconcentration was observed, although an increase has been reportedpreviously (49). The difference could be due to differencesin the S. cerevisiae strain, growth rate, or carbon source usedin the studies. Notably, glucose was the sole carbon sourceused for the previous study (49), while a small amount of ethanolwas supplied in this study, which might yield additional NADPHvia the NADP⁺-dependent acetaldehyde dehydrogenase and hencecould result in a lower steady-state PP split ratio. The dropof the 6PG concentration in our experiment might be due to aninhibition of G6P dehydrogenase by excess NADPH from NADP⁺-dependentacetaldehyde dehydrogenase that accompanied the sudden increasein q_acetate (Fig. 2).

In phase I, a considerable amount of carbon was directed towardstorage carbon metabolism. This result differs from previousreports of degradation of trehalose and glycogen by derepressedand chemostat-grown S. cerevisiae exposed to glucose excess,based on observations over a period of hours (42, 47). The differencemay be due to a mutation in adenylate cyclase in the yeast strainused in this study, which largely abolishes the glucose-inducedincrease in cyclic AMP (cAMP) concentration (43). We measureda rather constant cAMP concentration in our experiment (M. T.A. P. Kresnowati, unpublished results), in contrast with strongincreases observed for a different S. cerevisiae strain exposedto glucose excess (50). Since neutral trehalase (the major trehalose-degradingenzyme) is activated through a cAMP-dependent protein kinaseA (39), S. cerevisiae strains carrying this mutation exhibita glucose-induced accumulation of trehalose and, to a lesserextent, glycogen (43).

In phase II, F2,6bP began to accumulate, as observed by Vaseghiet al. (50) in a similar glucose pulse experiment with chemostat-grownS. cerevisiae CBS 7336. However, in their experiment, the cAMPconcentration also increased, which activated the PKA, whichin turn led to the phosphorylation of PFK2 by PKA. The absenceof cAMP accumulation by the S. cerevisiae strain used in thisstudy implies a PKA-independent regulation of phosphofructokinase2 activity by, for example, other protein kinases (11) or allostericmechanisms (15). The observed increases in F2,6bP concentrationseen previously (50) and in this study suggest that within theshort time window of observation (i.e., 300 seconds), proteinphosphorylation needs to be taken into account in kinetic models,in addition to enzyme-metabolite interactions.

Reconciliation with known kinetic mechanisms.
A number of kinetic mechanisms known from in vitro studies havebeen shown to be consistent with metabolic responses observedin glucose pulse experiments; for example, the activation ofpyruvate kinase by F1,6bP in vitro (23) serves as a possibleexplanation for the rapid decrease in PEP concentration uponan increase in F1,6bP concentration (38). In the following discussion,we evaluate the consistency of known in vitro kinetic mechanismsof primary metabolism of S. cerevisiae with the estimated metabolicfluxes and measured metabolite concentrations obtained in thisstudy.

During primary metabolism, the reversible reactions catalyzedby phosphoglucoisomerase, aldolase, enolase, and fumarase areclose to equilibrium in glucose-limited chemostat cultures ofS. cerevisiae, as confirmed by measured mass isotopomer distributionsof the reactants involved (48). With a glucose excess, G6P andF6P as well as fumarate and malate show highly correlated concentrationchanges, indicating that the reversible reactions catalyzedby phosphoglucoisomerase and fumarase are close to equilibrium.The reversible reaction catalyzed by enolase, however, is displacedfrom equilibrium, as can be seen from the more pronounced decreaseof the concentration of PEP than that of 2PG plus 3PG.

The decrease of q_glucose in phase I after the rapid initialincrease indicates that glucose transport is subject to significantinhibition. Candidate metabolites that inhibit the facilitateddiffusion of glucose are glucose (36) and G6P (2, 31), withthe latter showing a strong increase after the pulse. We cannotexclude, however, the inhibitory effect of intracellular glucose,as it was not measured.

The intermediates in upper glycolysis (G6P, F6P, and F1,6bP)accumulated in phase I even after the drop in q_glucose, whichimplies that the in vivo fluxes through the enzymes that metabolizethese intermediates, i.e., hexokinase (HK), PFK, and the enzymesin lower glycolysis, must have decreased. Several known kineticmechanisms can explain this seeming paradox. First, a strongdecrease in the ATP concentration and/or the energy charge canlimit the fluxes through HK and PFK (38). HK might be furtherinhibited by elevated levels of T6P (Fig. 4) (4, 16). In addition,decreased flux through PFK can be due to hyperaccumulation ofthe product F1,6bP (37) as well as to decreased levels of theallosteric activator AMP (after about 20 seconds, Fig. 6) (25).During phase I, the concentration of F2,6bP, a strong activatorof PFK (1), remained close to the steady-state level. In phaseII, the concentrations of G6P, F6P, and F1,6bP decreased, whileq_glucose increased, which implies that the in vivo fluxes throughHK, PFK, and lower glycolysis must have increased. The stimulatoryeffects leading to the increased fluxes could be a recoveredenergy charge in phase II and elevated levels of F2,6bP (Fig.7).

Pyruvate is partitioned between the TCA cycle and the fermentativepathway leading to ethanol and acetate. In phase I, q_acetateis larger than q_ethanol, which could be explained by a muchlower affinity of alcohol dehydrogenase than acetaldehyde dehydrogenasefor acetaldehyde (28). The acetaldehyde dehydrogenase couldbe saturated with acetaldehyde, leading to the rather constantq_acetate observed during the postpulse period. Furthermore,we observed that changes in pyruvate concentration paralleledthe estimated nonfermentative q_CO₂ (which closely correlateswith flux through the TCA cycle) and were similar to changesobserved in most TCA cycle intermediates (Fig. 5). The in vitroK_m estimated for pyruvate oxidation by intact mitochondria (0.3mM) (46) is much higher than the pyruvate concentrations weobserved, which suggests that the TCA cycle flux can respondquickly to changes in its precursor concentration.

The concentration increases of the more reduced fumarate andmalate metabolites were accompanied by a concentration decreaseof the more oxidized aspartate, while the sum of these C₄ metabolitesremained almost constant during the postpulse period (not shown).These results suggest that the fluxes into and out of the C₄pool are well balanced and that the redistribution of the poolcomponents towards a more reduced status might be triggeredby an increased free NADH/NAD⁺ ratio. This result is qualitativelyconsistent with an estimated increase in the cytosolic freeNADH/NAD⁺ ratio by Theobald et al. (38).

Implications from carbon, electron, and energy balances.
The carbon, electron, and energy balances suggest that q_anabolismincreased during phase III. The onset of anabolic processesdoes not seem to be related to concentration changes in freeamino acids as direct precursors, as their average levels incells were not significantly affected by the glucose pulse.In addition, during a prolonged chemostat cultivation of S.cerevisiae CEN.PK 113-7D, the growth rate remained constantfor up to 90 generations, while average concentrations of freeamino acids in the cells decreased severalfold (20). A possiblesignal for increased q_anabolism might be a net increase in cellularenergy. Although the energy charge is only slightly higher inphase III than in the steady state, the individual ratios ofATP/ADP and ATP/AMP are nearly twice their steady-state values.An increased q_anabolism also implies an excess of ribosomalproteins at the 0.05 h^–1 growth rate, as the amount ofribosomes for protein synthesis probably remains constant duringthe relatively short postpulse period. This excess capacityindicates a low ribosomal efficiency, which has been observedpreviously for low specific growth rates in both prokaryotes(17) and eukaryotes (6, 35).

The unmeasured carbon sinks account for $~$ 15% of the glucose uptakeat the end of phase III (assuming increased q_anabolism) (Table7). The carbon, electron, and ATP balances suggest that theseas yet unmeasured carbon sinks have an average degree of reductionof 4.17 per Cmol and do not involve intensively energy-consumingpathways. One such carbon sink could be glycogen (degree ofreduction per Cmol = 4.0), which has a rather low energy requirement(0.33 mol of ATP mol of glycogen^–1, compared to 1.65 molof ATP Cmol_x^–¹ for biomass). Given the high glycogen contentin yeast cells, i.e., 102 mCmol Cmol_x^–¹ (13), completeconversion of the remaining unquantified carbon sinks to glycogenwould increase the cellular glycogen content only 7%, whichwould require a highly accurate method of glycogen quantificationfor correct assessment.

In summary, this study provides a comprehensive picture of theextra- and intracellular dynamic responses of S. cerevisiaewithin 300 seconds of a glucose pulse. The constructed carbonand electron balances close >70% in the 300-second postpulseperiod and are thus consistent with these measurements representingthe major sinks. The discrepancy in the ATP balance in phasesI and III, as well as the missing carbon and electron sinksin phase III, suggests that significant changes in the anabolicflux can occur during short transient periods. Our findingsemphasize the importance of including storage and anabolic processesin kinetic models, in addition to glycolysis and fermentativemetabolism, which have already been included (30, 49). Futureresearch is needed to validate the estimated increase in q_anabolismand to quantify possible changes in glycogen content.

ACKNOWLEDGMENTS

This work was supported by The Netherlands Organization forScientific Research (NWO) and by DSM.

FOOTNOTES

* Corresponding author. Mailing address: DSM Anti-Infectives, P.O. Box 525, 2613 AX Delft, The Netherlands. Phone: 31 152792513. Fax: 31 152792779. E-mail: Liang.Wu{at}dsm.com.

REFERENCES

Top