intracellular carbon fluxes in riboffavin-producing bacillus subtilis

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Apr. 2002, p. 1760–1771 Vol. 68, No. 40099-2240/02/$04.00�0 DOI: 10.1128/AEM.68.4.1760–1771.2002Copyright © 2002, American Society for Microbiology. All Rights Reserved.

Intracellular Carbon Fluxes in Riboflavin-ProducingBacillus subtilis during Growth on Two-Carbon

Substrate MixturesMichael Dauner,1 Marco Sonderegger,1 Michel Hochuli,2 Thomas Szyperski,2†

Kurt Wüthrich,2 Hans-Peter Hohmann,3 Uwe Sauer,1*and James E. Bailey1

Institute of Biotechnology1 and Institute of Molecular Biology and Biophysics,2 ETH Zürich,CH-8093 Zürich, Switzerland, and Roche Vitamins Inc., Nutley,

New Jersey 07110-11993

Received 24 July 2001/Accepted 8 January 2002

Metabolic responses to cofeeding of different carbon substrates in carbon-limited chemostat cultures wereinvestigated with riboflavin-producing Bacillus subtilis. Relative to the carbon content (or energy content) of thesubstrates, the biomass yield was lower in all cofeeding experiments than with glucose alone. The riboflavinyield, in contrast, was significantly increased in the acetoin- and gluconate-cofed cultures. In these twoscenarios, unusually high intracellular ATP-to-ADP ratios correlated with improved riboflavin yields. Nuclearmagnetic resonance spectra recorded with amino acids obtained from biosynthetically directed fractional 13Clabeling experiments were used in an isotope isomer balancing framework to estimate intracellular carbonfluxes. The glycolysis-to-pentose phosphate (PP) pathway split ratio was almost invariant at about 80% in allexperiments, a result that was particularly surprising for the cosubstrate gluconate, which feeds directly intothe PP pathway. The in vivo activities of the tricarboxylic acid cycle, in contrast, varied more than twofold. Themalic enzyme was active with acetate, gluconate, or acetoin cofeeding but not with citrate cofeeding or withglucose alone. The in vivo activity of the gluconeogenic phosphoenolpyruvate carboxykinase was found to berelatively high in all experiments, with the sole exception of the gluconate-cofed culture.

During batch growth on mixtures of carbon substrates, bac-teria frequently first consume almost exclusively their pre-ferred substrate. Consumption of any other substrate(s) occursonly after depletion of the preferred one, leading to a diauxicpattern of growth. The primary molecular mechanisms thatinhibit the simultaneous utilization of substrates are cataboliterepression (9) and inducer exclusion (47). Under carbon-lim-ited conditions in chemostat cultures, in contrast, mixtures ofcarbon sources are often utilized simultaneously at low andintermediate dilution rates (D) (18). Thus, for the most fre-quently used paradigm catabolite repression sugar, glucose,such coutilization occurs at concentrations below a criticalrepression level (29). Consequently, many catabolic enzymesare relieved from repression so that alternative substrates canbe catabolized (32, 34).

While many biotechnological processes operate on a singlecarbon source, substrate mixtures may help to diagnose poten-tial bottlenecks in the biosynthetic pathways to a desired prod-uct (10). For the production of riboflavin (vitamin B2), which isa commercially important additive in the feed and food indus-tries, such cofeeding experiments were used to identify poten-tial limitations in building block supply (55) (Fig. 1). Generally,cofeeding experiments are evaluated on the basis of physiolog-ical analysis, with a primary focus on production. Much infor-

mation on the underlying metabolic network response, how-ever, is not accessible from extracellular physiological data butrequires knowledge of intracellular carbon fluxes.

The classical approach to analyzing intracellular carbonfluxes is based on so-called metabolite balancing, which usuallyrequires assumptions about redox or energy balances thatstrongly affect the flux estimates (8, 62). With the developmentof appropriate labeling experiments as well as extensive iso-tope isomer (isotopomer) models (11, 53, 63), a powerful toolfor flux analysis was introduced that allows researchers toavoid such assumptions about redox and energy balances andthus increases the resolution and reliability of the estimates.The term “isotopomer” describes the different positional com-binations of 12C and 13C atoms within a single molecule. Themetabolic flux state determines the distribution of isotopomersin the intracellular metabolites which, in turn, determines thedistribution of isotopomers in the amino acids (Fig. 2). Usingnuclear magnetic resonance (NMR) or mass spectroscopy,subsets of these isotopomer pools may then be analyzed, forexample, in the amino acids of hydrolyzed proteins (12, 58).Balancing the isotopomer pools of all intermediates in a bio-chemical reaction network enables an accurate mathematicaldescription of the label distribution within the network. Com-bined with data on extracellular rates of substrate consumptionand product formation, such rigorous accounting of 13C labeldistribution in the entire metabolic system provides novel in-formation on intracellular reaction rates (fluxes) (Fig. 2).

The aim of this study was to investigate metabolic flux re-sponses to cofeeding of different carbon substrates in chemo-stat cultures of a recombinant, riboflavin-producing Bacillus

* Corresponding author. Mailing address: Institute of Biotechnol-ogy, ETH Zürich, CH-8093 Zürich, Switzerland. Phone: 41-1-633 3672. Fax: 41-1-633 10 51. E-mail: [email protected].

† Present address: Department of Chemistry, State University ofNew York at Buffalo, Buffalo, NY 14260.

1760

on April 5, 2019 by guest

http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

subtilis strain (13, 41). For this purpose, four cosubstrates werechosen that enter central carbon metabolism at different posi-tions. Gluconate and citrate provide direct intermediates ofthe pentose phosphate (PP) pathway and the tricarboxylic acid(TCA) cycle, respectively. Hence, they may serve as a carbonand/or energy source for riboflavin biosynthesis. Acetoin andacetate, in contrast, are catabolized via acetyl coenzyme A(acetyl-CoA) and, due to a lack of the glyoxylate shunt in B.subtilis (30), can serve only as an energy source by catabolismin the TCA cycle. To elucidate the intracellular flux responsesto two-substrate catabolism, we performed biosynthetically di-rected 13C labeling experiments by growing chemostat cultureson mixtures of 90% unlabeled glucose, 10% [U-13C6]glucose,and one of the above unlabeled cosubstrates. The labelingpatterns were assessed by two-dimensional (2D) NMR analysisof amino acids obtained by hydrolysis of cellular proteins (58).In contrast to previous analyses based on 2D NMR data (50,51), we use here a recently developed comprehensive iso-topomer model for data interpretation that enables more rig-orous accounting of all labeling data (11, 14, 19).

MATERIALS AND METHODS

Bacterial strain. The recombinant, riboflavin-producing B. subtilis strainRB50::pRF69 was used throughout this study. The host strain, RB50 (purA60Azr-11 Dcr-15 MSr-46 RoFr-50 spo0A), is characterized by several randomlyintroduced purine and riboflavin analog resistance mutations (41). To constructRB50::pRF69, one copy of the constitutively expressed, recombinant B. subtilisrib operon, pRF69, was integrated in the native chromosomal rib operon (41). Toincrease gene dosage, the pRF69 operon was amplified by selection in serialbatch cultures up to a concentration of 80 mg of chloramphenicol liter�1. Theresulting strain was denoted RB50::[pRF69]n, where n is the number of amplifiedpRF69 rib operons. While the exact number of amplifications is not known, allresults presented are directly comparable because we inoculated all culturesfrom the same frozen stock of RB50::[pRF69]n, which was also used in previouslyreported experiments (11, 13, 14).

Growth conditions and media. All media were supplemented with 80 mg ofchloramphenicol liter�1. Frozen stocks were used to inoculate a complex me-dium, which contained 5 g of glucose liter�1, 25 g of veal infusion broth liter�1,and 5 g of yeast extract liter�1. After growth for 12 h in the complex medium, 2.5ml was used to inoculate 50 ml of minimal medium in a 500-ml baffled shakeflask; this mixture was cultivated for another 12 h and used entirely for reactorinoculation. The minimal medium used for the batch culture was similar to thatused for the chemostat culture but did not contain H2SO4 and was supplementedwith 0.1 M sodium phosphate buffer (pH 6.8).

Continuous cultivation in aerobic, glucose-limited chemostat cultures was con-ducted at 38°C with a working volume of 1 liter in a 2-liter LH discovery 210series reactor (Adaptive Biosystems) equipped with pH, dissolved oxygen, tem-perature, optical density, and foam probes. The medium of the glucose-limitedchemostat cultures contained (per liter of bidistilled water) the following: glu-cose, 5 g; NH4Cl, 4.75 g; (NH4)2SO4, 6 g; Na2HPO4 · 12H2O, 0.99 g; KH2PO4,4 g; MgSO4 · 7H2O, 0.42 g; CaCl2 · H2O, 46 mg; FeSO4 · 7H2O, 9 mg; and 40 mlof a trace element solution. The trace element solution had the following com-position (per liter of bidistilled water): MnCl2 · 4H2O, 2.25 g; ZnCl2, 1.32 g;CuCl2 · 2H2O, 0.34 g; CoCl2 · 6H2O, 0.5 g; Na2MoO4 · 2H2O, 0.5 g; andAlCl3 · 6H2O, 1.25 g. The medium was acidified to a pH of between 2 and 3 bythe addition of H2SO4 (95 to 97%) and was sterilized by passage through a0.2-�m-pore-size filter. For cofeeding experiments, 2 g of acetoin, gluconate,acetate, or citrate liter�1 was added. During cultivation, the pH was controlledat 6.8 � 0.1, and the volume was kept constant by a weight-controlled pump. Aconstant airflow of 1 liter min�1 was achieved by a mass flow meter, and theagitation speed was set to between 600 and 1,000 rpm, ensuring dissolved oxygenlevels above 40% under all conditions. All reported chemostat experiments wereperformed at the physiological steady state, which was defined as at least fivevolume changes under the same conditions and stable optical density and exhaustgas readings for at least three volume changes.

During the labeling experiments, the feed medium was replaced by an identicalmedium but with 10% (wt/wt) of the total glucose present as [U-13C6]glucose(13C, �98%; Isotech, Miamisburg, Ohio). Biomass aliquots for NMR analysiswere withdrawn after 0.8 volume change, so that according to first-order washoutkinetics, 45% of the biomass was fractionally labeled.

FIG. 1. Purine and riboflavin biosynthesis pathway. Produced or consumed cofactors or building blocks are shown below the pathway.Abbreviations: P5P, pentose-5-phosphate; PRPP, phosphoribosylpyrophosphate; �P, number of equivalents.

FIG. 2. The metabolic flux state is one determinant of the iso-topomer pools in metabolic intermediates and amino acids. Addition-ally, the isotopomer pools are influenced by the factors shown in gray.By use of isotopomer balancing, the metabolic flux state may be esti-mated (broken arrows) from NMR- or mass spectroscopy-based iso-topomer data and knowledge of the additional influential factors (shownin gray).

VOL. 68, 2002 CARBON FLUXES IN RIBOFLAVIN-PRODUCING B. SUBTILIS 1761


http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

Analytical techniques. Cell dry weight was determined from at least eightparallel 10-ml cell suspensions that were harvested by centrifugation, washedwith distilled water, and dried at 80°C for 24 h to a constant weight. Protein,RNA, and glycogen contents of the cells were determined by colorimetric andenzymatic assays as described elsewhere (13). Concentrations of carbon dioxideand oxygen in the bioreactor feed medium and effluent gas were determined witha mass spectrometer (Prima 600; Fisons Instruments).

Extracellular glucose, succinate, pyruvate (PYR), and phosphoenolpyruvate(PEP) concentrations were determined enzymatically with commercial kits(Beckman) or as described by Bergmeyer (5). Acetate, acetoin, and 2,3-butane-diol concentrations were determined by gas chromatography (5890E; Hewlett-Packard) with a Carbowax MD-10 column (Macherey-Nagel) and with butyrateas the internal standard. Organic acid, acetoin, and diacyl concentrations inculture supernatants were determined by high-pressure liquid chromatographywith a Supelcogel C8 column (4.6 by 250 mm) (Sigma) and a diode array detector(Perkin-Elmer). Phosphoric acid (0.2 N) was used as the mobile phase at a flowrate of 0.3 ml min�1 and 30°C. Hexosamine concentrations in supernatantsoriginating from both peptidoglycan and teichoic acid were determined by acolorimetric assay (24) with glucosamine for calibration. Riboflavin concentra-tions were determined by measuring the absorption at 444 nm (A444) in cell-freeculture broth.

Intracellular concentrations of ATP and ADP were determined as describedpreviously (60) with an ATP bioluminescence kit (HS II; Boehringer Mann-heim). Briefly, a 10-ml cell suspension was rapidly (within 0.7 s) withdrawn witha syringe containing precooled glass beads (�20°C). Aliquots (40 �l) weretransferred to an Eppendorf cup and immediately quenched by mixing with 160�l of dimethyl sulfoxide. After the addition of 800 �l of ice-cold 25 mM HEPESbuffer (pH 7.75), the samples were stored at �80°C and analyzed within 1 week.ATP concentrations were measured directly with a spectrofluorophotometer(RF-5001PC; Shimadzu) by detecting the light emission at 562 nm. ADP wasconverted to ATP by adding PYR kinase and PEP to 1,250 U/ml and 1 mM,respectively. After incubation for 30 min at 37°C, the resulting total ATP levelwas determined, yielding the ADP level as the difference between the two values.Standard solutions of ATP-ADP mixtures at a molar ratio of 8:1 in the range of2 to 800 nM were used for calibration.

Description of the biochemical reaction network. The previously describedcomprehensive isotopomer model of B. subtilis central carbon metabolism (11),with an H�-to-ATP ratio of 4 (20, 48), was extended to accommodate thesituations arising with cofeeding of multiple alternative carbon sources. Specif-ically, gluconate is assumed to enter the central carbon metabolism via theconsecutive reactions of gluconate permease (GntP), gluconate kinase (GntK),and 6-phosphogluconate dehydrogenase (GntZ) (21). Since B. subtilis lacks theEntner-Doudouroff pathway, it is assumed that gluconate is further metabolizedthrough the PP pathway (40). The additional catabolic genes are organized in thegntRKPZ operon, which is transcribed from a gluconate-inducible promoterupstream of gntR. Expression is controlled by the regulator GntR, which isinactivated by gluconate (22). Like that in Escherichia coli, gluconate uptake inB. subtilis occurs via a proton symport mechanism (57).

Under glucose-limited conditions, acetyl-CoA synthetase is responsible foracetate utilization (27). In a single-step reaction, acetate is converted to acetyl-CoA at the expense of one equivalent of ATP, which is cleaved into AMP andPPi. Due to the large negative reaction enthalpy under physiological conditions(�G°�, �88.2 kJ/mol), this reaction is assumed to be irreversible. Exchangefluxes, however, are considered to occur via phosphotransacetylase and acetatekinase, which usually catalyze acetate formation (54). The uptake of acetate mostlikely occurs via passive diffusion and/or a monocarboxylate transporter (28).Because acetate either diffuses across the cytoplasmic membrane in its proton-ated form or is taken up by an H� symporter (28), we assume that one equivalentof H� is transported per transported molecule of acetate.

Generally, acetoin degradation is assumed to be mediated by the 2,3-butane-diol cycle (26). However, there is also experimental evidence that acetoin canundergo direct oxidative cleavage (33); hence, the existence of two alternativeenzymatic routes, encoded by the acu and the aco gene clusters, was proposed(27). Later evidence indicated that the aco gene cluster, which encodes theacetoin dehydrogenase enzyme system, is the primary pathway for acetoin utili-zation (31). This pathway involves the cleavage of acetoin into acetate andacetaldehyde, with the simultaneous reduction of an acceptor molecule. In thenext step, which is catalyzed by aldehyde dehydrogenase, the acetaldehyde gen-erated is converted to acetate, with the reduction of one NADH. The tworesulting acetate molecules are then introduced into the central carbon metab-olism by acetyl-CoA synthetase. In our model, NAD� is considered the electronacceptor for both reduction steps. Acetoin transport is assumed to occur by

either passive or facilitated diffusion; thus, no metabolic energy expenditures areconsidered.

Citrate enters directly into the TCA cycle. In an aconitase mutant of B. subtilis,the intracellular accumulation of citrate was reported to lead to the induction ofa divalent cation-dependent citrate transport system (64). This transport systemwas described later as a citrate-Mg2� symporter (6). Because transport dependson the delta pH component of the proton motive force, the number of symportedprotons was assumed to be 2 (6).

NMR spectroscopy data analysis. To prepare samples for NMR spectroscopy,a culture aliquot was harvested and centrifuged at 1,200 g for 20 min at 4°C.The cell pellet was washed once with 20 mM Tris-HCl (pH 7.6), centrifugedagain, resuspended in 6 ml of 6 M HCl, and hydrolyzed by incubation in sealedPyrex glass tubes for 24 h at 110°C. The hydrolysate was filtered through a0.2-�m-pore-size filter and lyophilized. The dried material was dissolved in 600�l of 20 mM deuterochloric acid in 2H2O, incubated for 2 h at room tempera-ture, centrifuged, and used to record the NMR spectra.

2D heteronuclear 13C-1H correlation NMR spectra ([13C,1H]COSY spectra)were recorded at 40°C and a 13C resonance frequency of 150.8 MHz by using aBruker DRX500 spectrometer as described previously (58). For each experi-ment, [13C,1H]COSY spectra were recorded for the aliphatic 13C-1H moieties(data size, 4,096 by 512 complex points; t1max 373 ms; t2max 72 ms) and thearomatic 13C-1H moieties (data size, 1,536 by 512 complex points; t1max 393ms; t2max 72 ms) (58), with measurement times of 4.5 to 6 h and 2.5 h perspectrum, respectively. As described previously, the relative intensities of thedifferent 13C-13C scalar coupling fine structures in the [13C,1H]COSY spectrawere evaluated by integration with the program FCAL, version 2.3.0 (59).

Estimation of intracellular carbon fluxes. For the quantification of carbonfluxes in central metabolism, the relative abundances of the 13C-13C scalarcoupling fine structures corresponding to the individual amino acid carbon po-sitions, the experimentally determined macromolecular biomass composition,and the physiological data were combined within an isotopomer balancingmodel. Precursor demands for biomass formation and the experimentally deter-mined macromolecular composition were deduced from a previously publishedgrowth model (13). Briefly, the isotopomer balances of all metabolites that arerepresented in the model are calculated starting from a randomly chosen fluxdistribution. Superposition of 13C-13C scalar coupling fine structures correspond-ing to this isotopomer distribution is then used for simulation and comparison tothe experimentally determined NMR spectra. The quality of the fit is judged bythe �2 (error) criterion. Through an iterative process of flux estimation and signalfitting, a flux solution is sought that corresponds to a minimal �2 value. Thisoptimal solution represents the maximum-likelihood flux distribution in theinvestigated metabolic system that reflects both the physiological data and the2D NMR data.

To identify the global error minimum in the solution space by this iterativeprocedure, a range-restricted evolutionary algorithm was used (3). The finalsolution was obtained by restarting a modified direction set search algorithmaccording to Powell’s quadratic convergent method at the optimal flux solutionidentified by the global search evolutionary algorithm (44). The parametersearch was initiated from at least five different starting points, which alwaysyielded similar results, with fluxes varying maximally �10%. The solution withthe lowest �2 value was then subjected to detailed statistical analysis. For thispurpose, the Jacobian matrix of the overall output function, i.e., the linearapproximation of the nonlinear isotopomer balancing model, was calculatednumerically at the flux solution obtained. Conclusions could then be drawn abouthow the measured state variables would be influenced by differential changes inthe flux estimates. After the model was linearized, linear statistical theory couldbe applied (39) to calculate confidence intervals for single parameters (11).

Determination of the biochemical energy content. To determine the biochem-ical energy content of carbon substrates, the steady-state flux balance was for-mulated as an optimization problem in which the production of a particularproduct or quotients of the production of pairs of products were maximized,subject to

S � � � b (1)

where S is the stoichiometric matrix, � is the vector of fluxes, and b is the ratevector of metabolite production. For this purpose, either a variant of Mehrotra’spredictor-corrector algorithm (35), a primal-dual interior point method, or asequential quadratic programming method was used, as implemented in thelinprog and fmincon functions of the MATLAB Optimization Toolbox (TheMathWorks, Inc.).

1762 DAUNER ET AL. APPL. ENVIRON. MICROBIOL.


http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

RESULTS

Physiological analysis. Continuous cultures of B. subtilisRB50::[pRF69]n were grown in minimal medium at D of 0.10to 0.12 h�1 with either glucose or mixtures of glucose andgluconate, acetoin, acetate, or citrate as the growth-limitingsubstrates. In all instances, cometabolism of the additionalsubstrates with glucose was observed (Fig. 3). While gluconateand acetate were consumed completely, only a fraction of thesupplemental acetoin and citrate was used, with residual con-centrations of 16 mM (71% of the supplied acetoin) and 0.9mM (8% of the supplied citrate) in the effluent medium, re-spectively. The specific substrate uptake rate in terms of car-bon moles (C-mol) was enhanced in all cofeeding experimentscompared to the single-substrate experiments with glucose atthe same D (Fig. 3). This result was also reflected in thespecific oxygen consumption and carbon dioxide evolutionrates (data not shown). Generally, the formation of by-prod-ucts accounted for 4 to 8% of metabolized total molar carbon,and no significant differences were observed between single-substrate and mixed-substrate cultures (data not shown).

As may be expected, the biomass yield on glucose was sig-nificantly higher in the cofeeding experiments because addi-tional carbon sources were metabolized. The sole exceptionwas a slightly reduced yield with acetoin as the cofed substrate.

Specifically, the following yields (in grams mole�1) were ob-tained: 71 (acetate; D, 0.1 h�1), 64 (citrate; D, 0.1 h�1), 73(gluconate; D, 0.12 h�1), and 55 (acetoin; D, 0.12 h�1); incomparison, yields were 56 and 57 with glucose only at D of 0.1and 0.12 h�1, respectively. When normalized to the total car-bon content in the substrates, however, the biomass yield wasconsistently lower in all instances (Fig. 4). To account for thedifferent degrees of reduction (or energy) in the substrates aswell, two different yield coefficients were calculated to attemptto normalize the biomass yield to the energy content of thesubstrate. The first yield coefficient, YX/e, is defined as theweight of dry cell mass produced per electron equivalents on nsubstrates i(Si) according to

YX/e � �i1

n YX/Si

Ye/Si(2)

where YX/Si is the molar yield coefficient on Si during thecofeeding experiment and Ye/Si is the number of electronsavailable from complete combustion of the substrate to CO2

(7). Ye/Si can be deduced from the degree of reduction ( S)(45) by multiplication with the number of carbon atoms con-tained in Si; the S values are 4 for glucose, 3.67 for gluconate,4 for acetate, 5 for acetoin, and 3 for citrate. The second yieldcoefficient, YX/BEC, is based on the same principle but consid-ers the specific cellular biochemistry and is a more accuratemeans to differentiate energy-deficient and energy-excess sub-strates (1, 2), according to

YX/BEC � �i1

n YX/Si

YBEC/Si(3)FIG. 3. Specific uptake rates for glucose, cofed substrates, and total

carbon in carbon-limited chemostat cultures. Cultures were growneither on glucose only or on mixtures of glucose and the indicatedcosubstrates.

FIG. 4. Biomass yields and riboflavin yields in chemostat cultures.The biomass yields are calculated per mole of carbon (left, dark graybars), mole of electrons (middle, light gray bars), or mole of ATP(right, black bars) available in the substrates.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

where YBEC/Si is the biochemical energy content of Si.YBEC/Si

was calculated as the maximum amount of ATP that could begenerated from a substrate molecule within the biochemicalreaction network of B. subtilis. On average, per carbon atom, amaximum of 2.50 molecules of ATP were found for glucose,2.21 were found for gluconate, 1.13 were found for acetate,1.75 were found for acetoin, and 1.58 were found for citrate.Consistent with the carbon-based yield, in almost all instanceswhen the energy content was considered (YX/e and YX/BEC), thebiomass yields were lower for the mixed-substrate culturesthan for the cultures with glucose alone. Thus, growth on themixtures was obviously less efficient than that on glucose alone.

While the biomass yield decreased in all cofeed scenarios,the yield of riboflavin showed no such consistent trend (Fig. 4).Cofeeding of acetoin or gluconate, respectively, resulted in 64or 29% higher yields of riboflavin per mol of carbon in theconsumed substrates. While citrate had no significant effect,the acetate-glucose mixture reduced the riboflavin yield byabout 50%.

Metabolic flux analysis. To gain deeper insight into theintracellular carbon flux distribution during dual-substrate lim-ited growth, metabolic flux analysis was performed for each ofthe aforementioned experiments. First, to assess the buildingblock requirements for biomass formation, we determined thecontents of cellular protein, RNA, and glycogen. Glycogen wasnot detected, while the protein and RNA contents were virtu-ally identical in all cultures, accounting for 50 to 55% and 6 to8% of cellular dry weights, respectively (data not shown). Onthe basis of these data and a previously developed structuredbiomass model for chemostat growth of B. subtilisRB50::[pRF69]n (13), we calculated the biomass compositionand the corresponding detailed building block requirements.Using the linear function of biomass composition with D fromthis structured model, which is based on 17 chemostat runs, wereduce potential inaccuracies from individual experimentalanalyses. In the single-substrate experiment, the carbon bal-ance was 106.1% � 4.3%, and in the cofeeding experimentswith gluconate, acetoin, acetate, and citrate, the carbon bal-ances were 98.2% � 3.8%, 98.4% � 3.4%, 103.6% � 4.0%,and 107.9% � 4.4%, respectively.

In the physiological steady state, all cultures were subjectedto a labeling experiment during which the normal medium wasreplaced by a similar medium containing 10% (wt/wt) uni-formly 13C-labeled glucose. After about 0.8 volume change,aliquots of cells were harvested and [13C,1H]COSY spectrawere recorded for the total hydrolyzed biomass (58). The in-tracellular flux distribution was calculated as the best fit to therelative intensities of the 13C-13C scalar coupling fine structuremultiplets for 44 carbon atoms in the amino acids (Table 1 andAppendix), detailed building block requirements, and extracel-lular fluxes. This procedure was carried out with a previouslydescribed, comprehensive isotopomer model of B. subtilis me-tabolism (11). The flux distribution in the two glucose-limitedexperiments at D of 0.10 and 0.12 h�1 were virtually identical,and in the following experiments, the latter was used as thereference experiment (Fig. 5A).

As a measure of in vivo enzyme activity, the absolute fluxes(millimoles gram�1 hour�1) were found to vary significantlybetween the different experiments, most markedly for the PPpathway (glucose-6-phosphate to pentose-5-phosphate), PEP

carboxykinase (oxaloacetate [OAA] to PEP), malic enzyme(malate to PYR), and the anaplerotic PYR carboxylase (PYRto OAA) (Fig. 5). The latter activity was particularly low withcitrate cofeeding (Fig. 5E), presumably because citrate is theonly substrate that can directly replenish TCA cycle interme-diates; thus, the anaplerotic reaction is not required. Althoughcitrate supply was more than sufficient to replenish the with-drawal of TCA cycle intermediates for building block biosyn-thesis, we found low but significant anaplerotic PYR carboxy-lase activity. Significant malic enzyme activity was seen onlyduring cofeeding with gluconate, acetate, and acetoin. Thegluconeogenic PEP carboxykinase was active in all experi-ments, except during gluconate cofeeding. Generally, the invivo activity of the TCA cycle was higher in all cofeedingexperiments than in the reference experiment, with the highestactivity occurring during cofeeding with acetate or acetoin(Fig. 5C and D).

In some instances, it is informative to consider fluxes relativeto the glucose uptake rate to allow a more direct comparisonbetween experiments with different uptake rates. Although theabsolute fluxes through the oxidative PP pathway were shownto vary considerably, from 0.10 to 0.42 mmol g�1 h�1 (Fig. 5Aand C), the normalized fluxes were rather similar, ranging from7% (acetate cofed) to 20% (glucose limited). Because glu-conate is catabolized entirely via the oxidative PP pathway, itgreatly increases the fluxes downstream of glucose-6-phos-phate. The split ratios between the fluxes through the anaple-rotic PYR carboxylase and the catabolic PYR dehydrogenase(PYR to acetyl-CoA) were found to span 1 order of magnitudeover the five environmental conditions considered, from 1:1.7to 1:17.5, illustrating the remarkable flexibility of the PYRbranch point.

The obtained �2 values, indicating the quality of the fit of theabove flux estimates to the experimental data (Fig. 3 and 4 andAppendix), were 197 (acetoin), 624 (citrate), 646 (gluconate),and 481 (acetate); the value obtained with glucose as the solecarbon source was 119. The �2 values in the citrate, gluconate,and acetate cofeeding experiments were relatively high, sincetypical values for glucose-grown B. subtilis or E. coli are about100 when considered in the same type of labeling experiment

TABLE 1. Carbon atoms subjected to 2D COSY NMR analysis infractionally 13C-labeled amino acids and their corresponding

precursors in central metabolism

Carbon atom(s) analyzed Precursor(s)a

His-�, His-�, His-�2 .........................................................P5PTyr-ε ...................................................................................E4PTyr-�...................................................................................E4P and PEPSer-�, Ser-�.......................................................................SERGly-� ..................................................................................GLYPhe-�, Phe-�, Tyr-�, Tyr-� .............................................PEPAla-�, Ala-�, Leu-�, Leu-�1, Leu-�2, Ile- 2, Val-�,

Val- 1, Val- 2 ...............................................................PYRLeu-�..................................................................................ACAAsp-�, Asp-�, Thr-�, Thr-�, Ile-�, Ile- 1, Ile-�,

Met-�, Thr- .................................................................OAALys-�, Lys- , Lys-�, Lys-ε ................................................OAA and PYRGlu-�, Glu-�, Glu- , Pro-�, Pro-�, Pro- , Pro-�,

Arg-�, Arg-�..................................................................OGA

a P5P, pentose-5-phosphate; E4P, erythrose-4-phosphate; SER, serine; GLY,glycine; ACA, acetyl-CoA; OGA, oxoglutarate.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

and 2D COSY analysis (11, 19). Considering the degree offreedom within the single-substrate experiments, values ofabout 100 and 120 at confidence levels of 68 and 95%, respec-tively, typically would be expected (11). Closer inspection re-veals that more than 50% of the total error criterion originatesfrom deviation between simulated and measured 2D COSY

FIG. 5. Metabolic flux distribution of B. subtilis RB50::[pRF69]n incarbon-limited chemostat cultures with glucose (A), glucose and glu-conate (B), glucose and acetate (C), glucose and acetoin (D), andglucose and citrate (E). D was 0.1 h�1 (C and E) or 0.12 h�1 (A, B, andD). Gray arrows indicate precursor withdrawal for biomass biosynthe-sis. Net and exchange fluxes (millimoles gram�1 hour�1) are given insquare and oval boxes, respectively. Due to the fitting procedure usedfor intracellular flux estimation, extracellular fluxes can show smalldeviations from the experimentally determined data (Fig. 3). Abbre-viations: G6P, glucose-6-phosphate; F6P, fructose-6-phosphate; 6PG,6-phophogluconate; E4P, erythrose-4-phosphate; T3P, triose-3-phos-phate; PGA, phosphoglycerate; ACA, acetyl-CoA; OGA, oxogluta-rate; MAL, malate; FUM, fumarate; CIT, citrate; n.d., not determined.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

data at very few carbon positions (Appendix). Specifically,these carbon positions are Tyr-�x and εx for the citrate cofeed;His-�, Lys- , Ile- 1, Pro- , Tyr-εx, and Glu- for the gluconatecofeed; and six-His-�, Arg-�, Leu-�, Lys-ε, Tyr-εx, and Ile- 2

for the acetate cofeed. In almost all instances, poorly fittedsignals originate from carbon positions that are located at thejunction of different metabolic precursors for a particularamino acid. Obviously, NMR multiplet signals are very sensi-tive to slight variations in the degree of labeling in metabolicprecursors; this situation occurs only in the cofeeding experi-ments. This problem is less apparent in the acetoin cofeedingexperiment because the additional substrate is not used exten-sively for anabolism. This phenomenon was not considered inthe present error model (11), so that the reported �2 values arelikely overestimated in the cofeeding experiments. Since themisfits vary between individual experiments, structural errorsin the model are less likely than difficulties in automatic de-termination of multiplet NMR signals (59, 61). This view isalso consistent with experimental signal differences betweenPro- and Glu- and among Ile- 2, Val- 1, and Leu-�1 in thegluconate and acetate cofeeding experiments, respectively(Appendix). These carbon positions originate from the sameprecursor metabolite; thus, one would expect similar signal ratios.

Bidirectional reaction steps and their impact on the labelingstate of the system are included in the flux model, and theobtained estimates for the exchange fluxes are given in Fig. 5.For the experiments shown here, these estimates provide atbest qualitative information, simply because variations in theexchange cannot be discerned from the available data. This facthas, however, only a marginal influence on the estimated netfluxes.

The reliability of the flux estimates was verified by two dif-ferent procedures. First, the iterative fitting procedure wasrepeated at least five times with each data set. Although theparameter search was initiated from randomly chosen startingpoints, the resulting flux distributions were very similar in allinstances (data not shown). Thus, there is a high probabilitythat a global minimum was identified. Second, the linearizedmodel was used to compute the confidence regions around theflux estimates of the reported solutions with the lowest �2

values (11). The 68% confidence intervals of the oxidative PPpathway flux were �20% the specific glucose uptake rate. Netfluxes in glycolysis (PGA to PEP) and the TCA cycle weremore precisely determined, with 68% confidence intervals of�10% and �6%, respectively.

It is commonly observed that fluxes in the PEP-PYR-OAAtriangle are difficult to resolve, and estimates typically showhigh confidence intervals (11, 43). As shown before, a linear-ized statistical model is not well suited to assessing the statis-tical relevance of these particular fluxes, since the confidenceregions thus calculated are much larger than those obtained byother approaches (11). It appears, therefore, that the confi-dence intervals for the estimated fluxes in this triangle are onthe order of magnitude of those for the other fluxes in thenetwork. For more detailed analysis of these particular fluxes,it would be necessary to use labeled cosubstrates as well (43).

Redox and energy metabolism. The estimated flux distribu-tion also allows conclusions about intracellular redox and en-ergy metabolism to be drawn. The majority of biosynthetic

NADPH (26) was produced by the isocitrate dehydrogenasereaction in the TCA cycle. In the reference experiment and thegluconate cofeeding experiment, however, significant propor-tions of total NADPH were produced in the oxidative PPpathway (Fig. 6). On the demand side, the relative NADPHrequirements for biomass formation and riboflavin biosynthe-sis are reduced in the cofeeding experiments (Fig. 6). Conse-quently, higher proportions of NADPH are converted toNADH by the transhydrogenase reaction, which was previ-ously shown to be active in B. subtilis (13). This flux of reducingequivalents is a direct function of the estimated carbon fluxes;hence, it is determined within the confidence region of theestimated fluxes, in particular, the oxidative PP pathway flux.Although the 68% confidence interval of this particular flux is�20% the specific glucose uptake rate, the data presenteddemonstrate clearly that a transhydrogenase or a transhydro-genase-like reaction must operate in B. subtilis metabolism.

Unlike NADPH production, metabolic production of ATP isnot directly accessible from the flux distribution because thestoichiometry of ATP generation via respiration and ATPase isnot known exactly and may vary with environmental conditions(49). It is not unreasonable, however, to assume identical ef-ficiencies of ATP generation in all the carbon-limited samplesinvestigated here, thus allowing for a direct comparison ofATP production in the five experiments. Specifically, we useda P-to-O ratio of about 1, which corresponds to the generationof 1 ATP molecule per NADH (13, 49). Depending on theestimated flux through the succinate dehydrogenase that pro-duces the energetically less valuable FADH, however, the P-to-O ratio may be lower than 1. From the fluxes depicted inFig. 5, one thus obtains the ATP production rate (Fig. 7). Analmost invariant fraction of the produced ATP is required forthe biosynthesis of biomass and riboflavin, and a comparativelyminor fraction is required for futile cycle reactions, such asthose of the PEP carboxykinase and the malic enzyme (Fig. 7).By far the largest fraction of ATP produced, however, is notrequired for any specific purpose and is thus referred to asexcess ATP. The absolute excess production of ATP wasslightly but significantly increased in all but one of the cofeed-ing experiments, with the acetate-cofed culture being the onlyexception. This result explains the higher YX/BEC with acetateand the lower YX/BEC in the other cofed cultures compared tothe data in the reference experiment. The different behavior ofthe acetate-cofed culture is not related to a potentially lowerP-to-O ratio, for example, through a decoupling of the protongradient, because the residual acetate concentration was belowthe detection level and the proton-coupled uptake of acetatewas already considered in the calculation.

To investigate whether reduced growth efficiency and in-creased excess ATP production were correlated with the cel-lular energy state, we determined the intracellular ATP andADP concentrations in the physiological steady state prior tothe labeling experiment. All investigated cofeeding experi-ments were characterized by high ATP levels, which were morethan doubled in the gluconate and acetoin cofeeding experi-ments compared to the reference experiment (Fig. 8). Thesetwo cultures exhibited unusually high ATP-to-ADP ratios aswell.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

DISCUSSION

In this study, we used an isotopomer model to obtain abest-fit flux solution to all physiological data and the relativeabundances of the 13C-13C scalar coupling fine structure mul-tiplets for 44 amino acid carbon positions obtained from bio-synthetically directed fractional 13C labeling experiments. Thesplit ratio of glycolysis to PP pathway was relatively invariant atabout 80% in all experiments for growth on one- and two-carbon substrate mixtures. Surprisingly, coutilization of glu-

conate did not reduce this split ratio significantly, althoughgluconate supply exceeded by far the cellular demands ofNADPH and pentoses. Hence, it appears that flux into thePP pathway of B. subtilis is not controlled by the cellular de-mand for NADPH and/or pentoses but rather is determined bythe kinetic properties of the enzymes at this branch point, aswas previously shown for Corynebacterium glutamicum as well(36). The PYR node, in contrast, was rather flexible (56),because cofeeding with citrate reduced the anaplerotic fluxvia PYR carboxylase almost sixfold. The in vivo activity ofthe TCA cycle can respond in a flexible manner to cellularrequirements. This fact is illustrated by more than twofoldvariations in the estimated TCA cycle flux on different sub-strate mixtures.

The gluconeogenic PEP carboxykinase was found to be ac-tive under all conditions investigated. Although the expressionof this enzyme had been shown to be glucose repressed in B.subtilis batch cultures (16), very low residual glucose expres-sion in our chemostat cultures was obviously not sufficient toexert complete repression. Consistent with this view, it wasreported previously that in vivo PEP carboxykinase activityincreased with decreasing glucose concentrations at lowergrowth rates (11, 50). Regulation of this in vivo activity, how-ever, is more complex because ammonia- or phosphate-limitedchemostat cultures exhibited intermediate or absent PEP car-boxykinase fluxes, respectively, although high extracellular glu-cose concentrations were present in both cultures (14). Similarto in vitro results obtained with E. coli (37), we found signifi-cant in vivo malic enzyme activity on a glucose-acetate mixture.Likewise, high in vivo activities occurred with gluconate andacetoin as cosubstrates but not with citrate-glucose cofeedingor on glucose alone. From these results, we conclude thatneither increased futile cycle activity via PEP carboxykinase ormalic enzyme nor substantially altered pathway fluxes, for ex-ample, through altered splitting between glycolysis and theoxidative PP pathway, are primarily responsible for the re-duced growth efficiency in the mixed-substrate cultivations.

In addition to carbon fluxes, intracellular concentrations ofintermediates are an important component of comprehensivemetabolic analyses. Of particular importance are the concen-trations of ATP and ADP, which reflect the cellular energy

FIG. 6. Specific rates of NADPH production and consumption incarbon-limited chemostat cultures. NADPH production is accom-plished via the oxidative PP pathway (lower, dark gray bars) andisocitrate dehydrogenase (upper, light gray bars). NADPH consump-tion occurs via biomass formation, transhydrogenase, and riboflavinbiosynthesis (dark gray, light gray, and black bars, respectively).

FIG. 7. Specific ATP production rates in carbon-limited chemostatcultures. The ATP fraction consumed for biomass (including ribofla-vin) formation and futile cycling via malic enzyme and PEP carboxyki-nase and the excess ATP fraction are indicated by light gray, black, anddark gray bars (top to bottom), respectively.

FIG. 8. Cellular contents of ATP (left, light gray bars) and ADP(right, dark gray bars) in carbon-limited chemostat cultures. CDW, celldry weight.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

APPENDIX. Measured and simulated values of the 2D COSY spectra of amino acidsa

Aminoacid

Type ofvalue

Value for the following cultureb:

Glucose � Gluconate � Acetate � Acetoin � Citrate

s d1 d2 dd s d1 d2 dd s d1 d2 dd s d1 d2 dd s d1 d2 dd

Ala� Meas 0.18 0.09 0.06 0.66 0.19 0.08 0.05 0.68 0.19 0.08 0.07 0.66 0.19 0.09 0.07 0.64 NC NC NC NC

Sim 0.19 0.10 0.07 0.66 0.23 0.05 0.05 0.67 0.26 0.06 0.09 0.63 0.22 0.04 0.07 0.64 NC NC NC NC� Meas 0.26 0.74 — — 0.26 0.74 — — 0.28 0.72 — — 0.27 0.73 — — 0.25 0.75 — —

Sim 0.25 0.74 — — 0.30 0.73 — — 0.35 0.68 — — 0.30 0.71 — — 0.33 0.71 — —

Arg� Meas 0.56 0.37 — 0.07 0.54 0.40 — 0.06 0.64 0.32 — 0.04 0.55 0.40 — 0.05 0.74 0.21 — 0.05

Sim 0.58 0.34 — 0.03 0.59 0.32 — 0.03 0.65 0.29 — 0.03 0.59 0.32 — 0.02 0.74 0.21 — 0.02� Meas 0.29 0.71 — — 0.35 0.65 — — 0.48 0.52 — — 0.36 0.64 — — 0.37 0.63 — —

Sim 0.25 0.73 — — 0.32 0.66 — — 0.31 0.60 — — 0.35 0.64 — — 0.31 0.66 — —

Pro� Meas 0.36 0.24 0.32 0.08 0.43 0.24 0.30 0.03 0.47 0.27 0.26 0.00 0.42 0.24 0.27 0.07 NC NC NC NC

Sim 0.36 0.23 0.32 0.10 0.43 0.24 0.28 0.10 0.46 0.21 0.31 0.06 0.41 0.22 0.28 0.11 NC NC NC NC� Meas 0.57 0.42 — 0.01 0.50 0.38 — 0.11 0.62 0.37 — 0.01 0.53 0.40 — 0.07 0.79 0.21 — 0.00

Sim 0.61 0.36 — 0.04 0.55 0.30 — 0.02 0.66 0.30 — 0.03 0.58 0.31 — 0.02 0.79 0.22 — 0.02 Meas 0.22 0.67 — 0.11 0.40 0.49 — 0.10 0.37 0.54 — 0.09 0.41 0.49 — 0.10 0.29 0.62 — 0.09

Sim 0.23 0.67 — 0.08 0.27 0.56 — 0.04 0.29 0.58 — 0.07 0.30 0.55 — 0.05 0.29 0.62 — 0.07� Meas 0.26 0.74 — — 0.28 0.72 — — 0.45 0.55 — — 0.34 0.66 — — 0.32 0.68 — —

Sim 0.25 0.74 — — 0.33 0.69 — — 0.32 0.62 — — 0.36 0.65 — — 0.32 0.68 — —

Leu� Meas 0.25 0.00 0.7 0.05 0.24 0.00 0.67 0.08 0.38 0.00 0.53 0.08 0.30 0.00 0.60 0.09 0.25 0.00 0.67 0.08

Sim 0.24 0.02 0.7 0.09 0.28 0.02 0.66 0.06 0.30 0.03 0.57 0.07 0.33 0.02 0.59 0.07 0.31 0.02 0.64 0.08� Meas 0.83 0.16 — 0.01 0.69 0.20 — 0.11 0.95 0.04 — 0.01 0.74 0.19 — 0.07 0.77 0.18 — 0.05

Sim 0.82 0.19 — 0.01 0.71 0.12 — 0.01 0.92 0.12 — 0.01 0.75 0.13 — 0.01 0.77 0.16 — 0.01�1 Meas NC NC — — 0.38 0.62 — — 0.37 0.63 — — 0.37 0.63 — — 0.38 0.62 — —

Sim NC NC — — 0.27 0.66 — — 0.33 0.65 — — 0.28 0.67 — — 0.30 0.66 — —�2 Meas 0.89 0.11 — — 0.87 0.13 — — 0.89 0.11 — — 0.87 0.13 — — 0.87 0.13 — —

Sim 0.89 0.10 — — 0.87 0.07 — — 0.89 0.10 — — 0.87 0.09 — — 0.87 0.09 — —

His� Meas 0.18 0.00 0.05 0.77 0.20 0.00 0.05 0.76 0.17 0.00 0.07 0.76 0.16 0.00 0.05 0.79 0.15 0.00 0.05 0.80

Sim 0.17 0.02 0.06 0.77 0.27 0.01 0.03 0.73 0.19 0.02 0.05 0.75 0.19 0.01 0.02 0.78 0.24 0.02 0.02 0.77� Meas 0.23 0.35 0.02 0.4 LR LR LR LR 0.33 0.31 0.10 0.26 0.33 0.50 0.02 0.15 0.19 0.35 0.03 0.43

Sim 0.20 0.37 0.01 0.4 LR LR LR LR 0.17 0.41 0.01 0.16 0.18 0.54 0.01 0.18 0.22 0.44 0.01 0.26�2 Meas SNR SNR — — 0.44 0.56 — — LR LR — — 0.39 0.61 — — 0.45 0.55 — —

Sim SNR SNR — — 0.46 0.55 — — LR LR — — 0.37 0.62 — — 0.44 0.56 — —

Ser� Meas 0.21 0.05 0.34 0.39 0.26 0.04 0.34 0.36 0.19 0.06 0.24 0.51 0.19 0.07 0.24 0.51 0.19 0.06 0.32 0.44

Sim 0.24 0.06 0.33 0.38 0.27 0.03 0.32 0.36 0.23 0.04 0.24 0.49 0.23 0.02 0.23 0.50 0.27 0.03 0.27 0.42� Meas 0.56 0.44 — — 0.55 0.45 — — 0.44 0.56 — — 0.44 0.56 — — 0.52 0.48 — —

Sim 0.55 0.45 — — 0.56 0.44 — — 0.43 0.56 — — 0.43 0.57 — — 0.53 0.46 — —

Gly� Meas 0.29 0.71 — — 0.30 0.70 — — 0.30 0.70 — — 0.29 0.71 — — 0.28 0.72 — —

Sim 0.33 0.69 — — 0.35 0.67 — — 0.32 0.69 — — 0.33 0.69 — — 0.36 0.68 — —

Asp� Meas 0.35 0.09 0.34 0.22 0.33 0.10 0.31 0.26 0.41 0.08 0.27 0.23 0.32 0.12 0.27 0.28 0.38 0.09 0.34 0.19

Sim 0.34 0.09 0.33 0.24 0.36 0.09 0.28 0.24 0.41 0.06 0.31 0.19 0.37 0.09 0.27 0.22 0.43 0.04 0.29 0.08� Meas 0.36 0.22 0.33 0.09 0.33 0.23 0.31 0.12 0.43 0.21 0.26 0.09 0.36 0.25 0.27 0.12 0.39 0.17 0.33 0.10

Sim 0.36 0.23 0.32 0.10 0.37 0.23 0.27 0.10 0.42 0.19 0.29 0.06 0.39 0.21 0.26 0.10 0.44 0.09 0.29 0.04

Thr� Meas 0.35 0.09 0.34 0.22 0.33 0.10 0.31 0.25 0.46 0.08 0.27 0.19 0.33 0.12 0.28 0.27 0.38 0.09 0.33 0.20

Sim 0.34 0.09 0.33 0.24 0.36 0.09 0.28 0.24 0.42 0.06 0.32 0.20 0.37 0.09 0.27 0.22 0.43 0.04 0.29 0.08� Meas 0.32 0.62 — 0.06 0.34 0.55 0.11 — NC NC NC — 0.37 0.53 0.10 — 0.65 0.27 0.09 —

Sim 0.37 0.58 — 0.10 0.38 0.52 0.10 — NC NC NC — 0.41 0.50 0.10 — 0.51 0.44 0.05 — Meas 0.58 0.42 — — 0.57 0.43 — — 0.64 0.36 — — 0.60 0.40 — — 0.56 0.44 — —

Sim 0.58 0.42 — — 0.59 0.41 — — 0.62 0.38 — — 0.60 0.40 — — 0.60 0.38 — —

Met� Meas 0.33 0.10 0.31 0.26 SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR

Sim 0.34 0.09 0.33 0.23 SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR SNR

Ile� Meas 0.39 0.02 0.55 0.04 0.34 0.07 0.47 0.12 0.49 0.04 0.44 0.03 0.38 0.07 0.45 0.10 0.38 0.06 0.46 0.10

Sim 0.41 0.04 0.53 0.06 0.38 0.03 0.44 0.04 0.45 0.05 0.47 0.06 0.41 0.04 0.43 0.05 0.46 0.04 0.36 0.04

Continued on following page



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

APPENDIX—Continued

Aminoacid

Type ofvalue

Value for the following cultureb:

Glucose � Gluconate � Acetate � Acetoin � Citrate

s d1 d2 dd s d1 d2 dd s d1 d2 dd s d1 d2 dd s d1 d2 dd

1 Meas 0.47 0.48 — 0.05 0.45 0.48 0.07 — 0.50 0.41 0.08 — 0.52 0.44 0.04 — 0.47 0.47 0.06 —Sim 0.52 0.42 — 0.04 0.53 0.37 0.03 — 0.54 0.36 0.04 — 0.56 0.38 0.04 — 0.53 0.38 0.04 —

2 Meas 0.26 0.74 — — 0.28 0.72 — — 0.26 0.74 — — 0.29 0.71 — — 0.28 0.72 — —Sim 0.25 0.74 — — 0.29 0.72 — — 0.35 0.69 — — 0.30 0.71 — — 0.32 0.70 — —

� Meas 0.54 0.46 — — 0.53 0.47 — — 0.60 0.40 — — 0.57 0.43 — — 0.52 0.48 — —Sim 0.57 0.42 — — 0.58 0.40 — — 0.61 0.38 — — 0.59 0.40 — — 0.59 0.38 — —

Val� Meas 0.26 0.01 0.65 0.09 0.28 0.01 0.65 0.06 0.25 0.03 0.63 0.09 0.28 0.02 0.63 0.07 0.27 0.02 0.65 0.07

Sim 0.26 0.02 0.65 0.08 0.26 0.02 0.65 0.06 0.28 0.03 0.62 0.08 0.24 0.02 0.64 0.08 0.27 0.02 0.64 0.08 1 Meas 0.25 0.75 — — 0.26 0.74 — — 0.28 0.72 — — 0.28 0.72 — — 0.28 0.72 — —

Sim 0.25 0.75 — — 0.29 0.72 — — 0.35 0.68 — — 0.30 0.71 — — 0.33 0.70 — — 2 Meas 0.88 0.12 — — 0.86 0.14 — — 0.89 0.11 — — 0.86 0.14 — — 0.86 0.14 — —

Sim 0.89 0.10 — — 0.87 0.07 — — 0.89 0.10 — — 0.87 0.09 — — 0.87 0.09 — —

Phe� Meas 0.22 0.09 0.04 0.64 0.23 0.09 0.04 0.64 0.22 0.08 0.04 0.66 0.23 0.09 0.05 0.62 0.28 0.08 0.06 0.57

Sim 0.19 0.09 0.07 0.65 0.21 0.04 0.04 0.65 0.23 0.06 0.06 0.65 0.20 0.03 0.05 0.63 0.24 0.03 0.05 0.60� Meas 0.26 0.69 0.00 0.06 0.24 0.59 0.05 0.12 0.25 0.62 0.06 0.07 0.26 0.60 0.04 0.10 0.24 0.59 0.05 0.12

Sim 0.24 0.69 0.02 0.08 0.23 0.60 0.01 0.06 0.26 0.61 0.02 0.08 0.24 0.61 0.02 0.07 0.28 0.58 0.02 0.07

Tyr� Meas 0.21 0.10 0.04 0.65 0.21 0.11 0.04 0.65 0.20 0.08 0.05 0.67 0.21 0.10 0.05 0.64 0.25 0.09 0.06 0.60

Sim 0.19 0.09 0.07 0.65 0.21 0.04 0.04 0.65 0.23 0.06 0.06 0.66 0.21 0.03 0.05 0.64 0.24 0.03 0.05 0.61� Meas 0.24 0.69 0.00 0.07 0.19 0.56 0.07 0.17 0.25 0.63 0.04 0.09 0.23 0.61 0.05 0.11 0.28 0.59 0.05 0.09

Sim 0.24 0.69 0.02 0.08 0.21 0.57 0.01 0.05 0.27 0.62 0.02 0.08 0.24 0.61 0.02 0.07 0.29 0.59 0.02 0.07�X Meas 0.22 0.72 — 0.06 0.22 0.69 — 0.08 0.30 0.65 — 0.04 0.22 0.71 — 0.07 0.47 0.45 — 0.08

Sim 0.20 0.73 — 0.04 0.25 0.69 — 0.02 0.22 0.68 — 0.04 0.22 0.71 — 0.04 0.22 0.55 — 0.03εX Meas 0.37 0.27 — 0.36 0.53 0.11 — 0.36 LR LR — LR 0.45 0.16 — 0.38 0.69 0.10 — 0.21

Sim 0.36 0.23 — 0.39 0.44 0.14 — 0.44 LR LR — LR 0.44 0.16 — 0.40 0.47 0.20 — 0.41

Glu� Meas 0.34 0.23 0.33 0.10 0.34 0.24 0.33 0.09 0.46 0.22 0.26 0.06 0.37 0.25 0.28 0.10 0.43 0.17 0.32 0.08

Sim 0.35 0.23 0.32 0.10 0.40 0.22 0.26 0.09 0.44 0.20 0.30 0.06 0.40 0.21 0.27 0.10 0.46 0.09 0.30 0.05� Meas 0.63 0.37 — 0.00 0.54 0.38 — 0.07 0.74 0.25 — 0.01 0.68 0.26 — 0.06 0.84 0.08 — 0.08

Sim 0.63 0.37 — 0.04 0.58 0.31 — 0.03 0.71 0.32 — 0.03 0.64 0.34 — 0.02 0.80 0.22 — 0.02 Meas 0.22 0.01 0.69 0.07 0.21 0.03 0.66 0.10 0.40 0.03 0.53 0.05 0.30 0.00 0.61 0.08 0.29 0.03 0.62 0.07

Sim 0.24 0.02 0.69 0.08 0.31 0.01 0.61 0.05 0.30 0.03 0.57 0.07 0.34 0.02 0.60 0.06 0.30 0.02 0.61 0.08

Lys� Meas 0.26 0.64 — 0.10 0.28 0.64 — 0.09 NC NC — NC 0.29 0.62 — 0.09 0.43 0.52 — 0.06

Sim 0.30 0.63 — 0.09 0.33 0.61 — 0.07 NC NC — NC 0.34 0.59 — 0.08 0.40 0.53 — 0.06 Meas 0.54 0.43 — 0.03 0.68 0.27 — 0.05 0.62 0.38 — 0.00 0.54 0.42 — 0.04 0.65 0.34 — 0.01

Sim 0.53 0.44 — 0.05 0.56 0.43 — 0.04 0.59 0.42 — 0.04 0.54 0.41 — 0.04 0.58 0.42 — 0.05� Meas 0.30 0.64 — 0.07 NC NC — NC 0.33 0.62 — 0.05 NC NC — NC 0.32 0.61 — 0.07

Sim 0.30 0.63 — 0.09 NC NC — NC 0.39 0.58 — 0.06 NC NC — NC 0.41 0.54 — 0.06ε Meas 0.46 0.54 — — 0.45 0.55 — — 0.46 0.54 — — 0.46 0.54 — — 0.48 0.52 — —

Sim 0.46 0.54 — — 0.46 0.54 — — 0.53 0.45 — — 0.48 0.52 — — 0.56 0.40 — —Considered

long-rangecouplings

His-�Glucose �

gluconateMeas 0.19 0.29 0.03 0.35 0.02 0.05 0.01 0.06

Glucose �gluconate

Sim 0.16 0.31 0.00 0.01 0.00 0.01 0.00 0.10

His-�2Glucose �

acetateMeas 0.43 0.40 0.14 0.03

Glucose �acetate

Sim. 0.44 0.36 0.04 0.13

Tyr-εXGlucose �

acetateMeas 0.41 0.06 0.05 0.27 0.00 0.02 0.10 0.09

Glucose �acetate

Sim 0.41 0.05 0.01 0.25 0.01 0.00 0.11 0.14

a Simulated values are from flux estimates given in Fig. 5.b Plus symbols indicate that the substrate shown was added to glucose. NC, not considered in parameter estimation; LR, long-range coupling was used; SNR,

signal-to-noise ratio was below detection levels.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

state and are also common regulators in many cellular activi-ties (26, 52). All cofed cultures exhibited significantly higherconcentrations of ATP and higher ratios of ATP to ADP thanthe glucose-limited culture (Fig. 8). Thus, cofeeding of alter-native substrates improves the energy charge of B. subtilis. Theflux results indicate, however, that this apparent energeticprosperity is mostly used for excess ATP formation which, inturn, is used for maintenance metabolism and energy-spillingreactions (46) that cannot be assessed by the present method-ology (Fig. 7). This view would be consistent with the reducedbiomass yield in all cofeeding experiments. Yet another expla-nation could be that the actual P-to-O ratio in B. subtilis islower than the value of about 1 used in our model, a situationwhich would lead to reduced excess ATP production.

Several studies on rRNA synthesis control have indicatedthat the levels of nucleoside triphosphate pools increase lin-early with the growth rate and, in particular, that the ATP poolmay be responsible for growth rate-dependent regulation ofrRNA synthesis (4, 23, 38). Although this view has recentlybeen questioned (42), all these studies relied on alternativecarbon sources to achieve different growth rates in batch cul-tures. Our findings indicate that differences in cellular ATPand ADP contents may depend not only on the specific growthrate but also on the substrates and the associated flux distri-bution. However, we cannot exclude the possibility that theapparent divergence with our results is partly related to our useof an industrial B. subtilis strain with several deregulatory mu-tations in the purine biosynthesis pathway (41).

B. subtilis does not contain the enzymes of the glyoxylateshunt and therefore cannot grow on acetate or acetoin as asingle carbon source (30). Hence, these cosubstrates are ex-pected to influence primarily energy metabolism, while glu-conate can serve as a precursor for riboflavin biosynthesis.Acetoin and gluconate cofeeding increased the riboflavin yieldby 64 and 29%, respectively, but citrate had no significanteffect, and acetate addition even reduced the riboflavin yield byabout 50% (Fig. 4). In particular, the very different metabolicresponses in the acetate and acetoin cofeeding experiments aresurprising because both substrates enter metabolism at thesame intermediate, acetyl-CoA. The low yield with the acetatecofeed cannot be due to toxic effects of acetate because theeffective acetate concentration in the chemostat culture wasclose to zero. It is important to note in this context that the twobeneficial cosubstrates increased the intracellular ATP-to-ADP ratio most strongly. Due to the phosphorylation stepsinvolved in the conversion of GMP to GTP, a strong interde-pendence on the ATP-to-ADP ratio can be suspected. Thus,the high ATP-to-ADP ratio in the gluconate and acetoincofeeding experiments may be the primary reason for the im-proved riboflavin yield.

Cofeeding experiments are very useful to redirect carbonflow, such that by-product formation can be reduced and im-portant overflow reactions may be identified (10). For exam-ple, cometabolism of glucose and citrate in batch cultures andcontinuous B. subtilis cultures was shown to prevent the for-mation of overflow products and to increase the carbon yieldmore than twofold (25). Moreover, the simultaneous use ofglucose and gluconate in batch cultures of a recombinant D-ribose-producing B. subtilis strain led to higher yields of D-ribose and by-products (15). Our interest, however, was to

evaluate the influence of cofeeding on yields under process-relevant conditions. Hence, we conducted the experimentssuch that little or no by-product formation occurred. It isdifficult to compare the efficiencies of biomass and productformation on mixtures of substrates because carbon and energycontents are usually different. For industrial purposes, themost appropriate method is the calculation of product andbiomass yields on a cost basis. For the discovery of mismatchesin the regulation of carbon flow, however, physiologicallymeaningful concepts are required. The optimal approachwould be the determination of biomass and product yield co-efficients for individual substrates and the evaluation ofwhether the mixtures have purely additive effects (17). How-ever, many substrates, such as acetate and acetoin in thepresent investigation, do not support growth when used as theonly carbon and energy source (30). Thus, in addition to cal-culating biomass yields on a per-mole-of-carbon basis, we usedtwo alternative methods that consider the energy content aswell, YX/e and YX/BEC. Generally, all calculated biomass yieldsfor B. subtilis RB50::[pRF69]n were reduced in the cofeedingexperiments, indicating less efficient utilization of the mixturesthan of glucose alone (Fig. 4). From a bioenergetic perspective,glucose alone is clearly the most efficient substrate for biomassformation by B. subtilis. For producing riboflavin, however,cofeeding acetoin is a pertinent strategy to improve energeticprosperity and thus the production of the desired compound.

ACKNOWLEDGMENTS

This work was supported by a scholarship from the BoehringerIngelheim Fonds to M.D., by Roche Vitamins Inc., and by the SwissPriority Program in Biotechnology (SPP BioTech).

REFERENCES

1. Arbidge, M. V., B. A. Bulthuis, J. Schultz, and D. Crabb. 1993. Fermentationof Bacillus, p. 871–895. In A. L. Sonenshein, J. A. Hoch, and R. Losick (ed.),Bacillus subtilis and other gram-positive bacteria: biochemistry, physiology,and molecular genetics. American Society for Microbiology, Washington,D.C.

2. Babel, W., and R. H. Müller. 1985. Mixed substrate utilization in micro-organisms: biochemical aspects and energetics. J. Gen. Microbiol. 131:39–45.

3. Bäck, T., and H.-P. Schwefel. 1993. An overview of evolutionary algorithmsfor parameter optimization. Evol. Comput. 1:1–23.

4. Bagnara, A. S., and L. R. Finch. 1973. Relationships between intracellularcontents of nucleotides and 5-phosphoribosyl 1-pyrophosphate in Esche-richia coli. Eur. J. Biochem. 36:422–427.

5. Bergmeyer, H. U. 1985. Methods of enzymatic analysis, vol. IV. VCH Pub-lishers, Deerfield Beach, Fla.

6. Bergsma, J., and W. N. Konings. 1983. The properties of citrate transport inmembrane vesicles from Bacillus subtilis. Eur. J. Biochem. 134:151–156.

7. Blanch, H. W., and D. S. Clark. 1996. Biochemical engineering. MarcelDekker, Inc., New York, N.Y.

8. Bonarius, H. P. J., G. Schmid, and J. Tramper. 1997. Flux analysis ofunderdetermined metabolic networks: the quest for the missing constraints.Trends Biotechnol. 15:308–314.

9. Chambliss, G. H. 1993. Carbon source-mediated catabolite repression, p.213–219. In A. L. Sonenshein, J. A. Hoch, and R. Losick (ed.), Bacillussubtilis and other gram-positive bacteria: biochemistry, physiology, and mo-lecular genetics. American Society for Microbiology, Washington, D.C.

10. Cocaign, M., C. Monnet, and N. D. Lindley. 1993. Batch kinetics of Coryne-bacterium glutamicum during growth on various carbon substrates: use ofsubstrate mixtures to localise metabolic bottlenecks. Appl. Microbiol. Bio-technol. 40:526–530.

11. Dauner, M., J. E. Bailey, and U. Sauer. 2001. Metabolic flux analysis with acomprehensive isotopomer model in Bacillus subtilis. Biotechnol. Bioeng.76:144–156.

12. Dauner, M., and U. Sauer. 2000. GC-MS analysis of amino acids rapidlyprovides rich information for isotopomer balancing. Biotechnol. Prog. 16:642–649.

13. Dauner, M., and U. Sauer. 2001. Stoichiometric growth model for riboflavin-producing Bacillus subtilis. Biotechnol. Bioeng. 76:132–143.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

14. Dauner, M., T. Storni, and U. Sauer. 2001. Bacillus subtilis metabolism andenergetics in carbon-limited and excess-carbon chemostat culture. J. Bacte-riol. 183:7308–7317.

15. De Wulf, P., W. Soetaert, D. Schwengers, and E. J. Vandamme. 1996. D-Glucose does not catabolite repress a transketolase-deficient D-ribose-pro-ducing Bacillus subtilis mutant strain. J. Ind. Microbiol. 17:104–109.

16. Diesterhaft, M. D., and E. Freese. 1973. Role of pyruvate carboxylase, phos-phoenolpyruvate carboxykinase, and malic enzyme during growth and sporu-lation of Bacillus subtilis. J. Biol. Chem. 248:6062–6070.

17. Dijkhuizen, L., and W. Harder. 1979. Regulation of autotrophic and hetero-trophic metabolism in Pseudomonas oxalaticus OX1: growth on mixtures ofacetate and formate in continuous culture. Arch. Microbiol. 123:47–53.

18. Egli, T., U. Lendenmann, and M. Snozzi. 1993. Kinetics of microbial growthwith mixtures of carbon sources. Antonie Leeuwenhoek 63:289–298.

19. Emmerling, M., M. Dauner, A. Ponti, J. Fiaux, M. Hochuli, T. Szyperski, K.Wüthrich, J. E. Bailey, and U. Sauer. 2002. Metabolic flux responses topyruvate kinase knockout in Escherichia coli. J. Bacteriol. 184:152–164.

20. Fillingame, R. H. 1997. Coupling H� transport and ATP synthesis in F1F0-ATP synthases: glimpses of interacting parts in a dynamic molecular ma-chine. J. Exp. Biol. 200:217–224.

21. Fujita, Y., T. Fuhita, Y. Miwa, J. Nihashi, and Y. Aratani. 1986. Organizationand transcription of the gluconate operon, gnt, of Bacillus subtilis. J. Biol.Chem. 261:13744–13753.

22. Fujita, Y., and T. Fujita. 1987. The gluconate operon gnt of Bacillus subtilisencodes its own transcriptional negative regulator. Proc. Natl. Acad. Sci.USA 84:4524–4528.

23. Gaal, T., M. S. Bartlett, W. Ross, C. L. Turnbough, Jr., and R. L. Gourse.1997. Transcription regulation by initiating NTP concentration: rRNA syn-thesis in bacteria. Science 278:2092–2097.

24. Gerhardt, P., R. G. E. Murray, W. A. Wood, and N. R. Krieg (ed.). 1994.Methods for general and molecular bacteriology. ASM Press, Washington,D.C.

25. Goel, A., M. M. Domach, W. Hanley, J. W. Lee, and M. M. Ataai. 1996.Coordination of glycolysis and TCA cycle reaction networks. Citrate-glucosecometabolism eliminates acids and reveals potential metabolic engineeringstrategies. Ann. N. Y. Acad. Sci. 782:1–16.

26. Gottschalk, G. 1986. Bacterial metabolism, 2nd ed. Springer-Verlag, NewYork, N.Y.

27. Grundy, F. J., D. A. Waters, T. Y. Takova, and T. M. Henkin. 1993. Identi-fication of genes involved in utilization of acetate and acetoin in Bacillussubtilis. Mol. Microbiol. 10:259–271.

28. Halestrap, A. P., and N. T. Price. 1999. The proton-linked monocarboxylatetransporter (MCT) family: structure, function and regulation. Biochem. J.343:281–299.

29. Harder, W., and L. Dijkhuizen. 1976. Mixed substrate utilization, p. 297–314.In A. C. R. Dean, D. C. Ellwood, C. G. T. Evans, and J. Melling (ed.),Continuous culture 6: applications and new fields. Ellis Horwood Ltd.,Chichester, England.

30. Hederstedt, L. 1993. The Krebs citric acid cycle, p. 181–197. In A. L. Sonen-shein, J. A. Hoch, and R. Losick (ed.), Bacillus subtilis and other gram-positive bacteria: biochemistry, physiology, and molecular genetics. Ameri-can Society for Microbiology, Washington, D.C.

31. Huang, M., F. B. Oppermann-Sanio, and A. Steinbüchel. 1999. Biochemicaland molecular characterization of the Bacillus subtilis acetoin catabolic path-way. J. Bacteriol. 181:3837–3841.

32. Lendenmann, U., and T. Egli. 1995. Is Escherichia coli growing in glucose-limited chemostat culture able to utilize other sugars without lag? Microbi-ology 141:71–78.

33. Lopez, J. M., B. Thoms, and H. Rehbein. 1975. Acetoin degradation inBacillus subtilis by direct oxidative cleavage. Eur. J. Biochem. 57:425–430.

34. Matin, A. 1979. Microbial regulatory mechanisms at low nutrient concentra-tions as studied in chemostat, p. 323–339. In M. Shilo (ed.), Strategies ofmicrobial life in extreme. Verlag Chemie, Weinheim, Germany.

35. Mehrotra, S. 1992. On the implementation of a primal-dual interior pointmethod. SIAM J. Optimiz. 2:575–601.

36. Moritz, B., K. Striegel, A. A. de Graaf, and H. Sahm. 2000. Kinetic propertiesof the glucose-6-phosphate and 6-phosphogluconate dehydrogenases fromCorynebacterium glutamicum and their application for predicting pentosephosphate pathway flux in vivo. Eur. J. Biochem. 267:3442–3452.

37. Murai, T., M. Tokushige, J. Nagai, and H. Katsuki. 1971. Physiologicalfunctions of NAD- and NADP-linked malic enzymes in Escherichia coli.Biochem. Biophys. Res. Commun. 43:875–881.

38. Neidhardt, F. C., and D. G. Fraenkel. 1961. Metabolic regulation of RNAsynthesis in bacteria. Cold Spring Harbor Symp. Quant. Biol. 16:63–74.

39. Pazman, A. 1993. Nonlinear statistical models. Kluwer Academic Publishing,New York, N.Y.

40. Peekhaus, N., and T. Conway. 1998. What’s for dinner? Entner-Doudoroffmetabolism in Escherichia coli. J. Bacteriol. 180:3495–3502.

41. Perkins, J. B., A. Sloma, T. Hermann, K. Theriault, E. Zachgo, T. Erden-berger, N. Hannett, N. P. Chatterjee, I. V. Williams, G. A. Rufo, Jr., R.Hatch, and J. Pero. 1999. Genetic engineering of Bacillus subtilis for thecommercial production of riboflavin. J. Ind. Microbiol. Biotechnol. 22:8–18.

42. Petersen, C., and L. B. Moller. 2000. Invariance of the nucleoside triphos-phate pools of Escherichia coli with growth rate. J. Biol. Chem. 275:3931–3935.

43. Petersen, S., A. A. de Graaf, L. Eggeling, M. Möllney, W. Wiechert, and H.Sahm. 2000. In vivo quantification of parallel and bidirectional fluxes in theanaplerosis of Corynebacterium glutamicum. J. Biol. Chem. 275:35932–35941.

44. Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery. 1995.Numerical recipes in C: the art of scientific computing, 2nd ed. CambridgeUniversity Press, New York, N.Y.

45. Roels, J. A. 1980. Application of macroscopic principles to microbial metab-olism. Biotechnol. Bioeng. 22:2457–2514.

46. Russell, J. B., and G. M. Cook. 1995. Energetics of bacterial growth: balanceof anabolic and catabolic reactions. Microbiol. Rev. 59:48–62.

47. Saier, M. H., Jr., and M. Crasnier. 1996. Inducer exclusion and the regula-tion of sugar transport. Res. Microbiol. 147:482–489.

48. Saraste, M. 1999. Oxidative phosphorylation at the fin de siecle. Science283:1488–1493.

49. Sauer, U., and J. E. Bailey. 1999. Estimation of P-to-O ratio in Bacillussubtilis and its influence on maximum riboflavin yield. Biotechnol. Bioeng.64:750–754.

50. Sauer, U., V. Hatzimanikatis, J. E. Bailey, M. Hochuli, T. Szyperski, and K.Wüthrich. 1997. Metabolic fluxes in riboflavin-producing Bacillus subtilis.Nat. Biotechnol. 15:448–452.

51. Sauer, U., D. R. Lasko, J. Fiaux, M. Hochuli, R. Glaser, T. Szyperski, K.Wüthrich, and J. E. Bailey. 1999. Metabolic flux ratio analysis of genetic andenvironmental modulations of Escherichia coli central carbon metabolism. J.Bacteriol. 181:6679–6688.

52. Schlegel, H. G. 1985. Allgemeine Mikrobiologie, 6th ed. Thieme, Stuttgart,Germany.

53. Schmidt, K., M. Carlsen, J. Nielsen, and J. Villadsen. 1997. Modeling iso-topomer distributions in biochemical networks using isotopomer mappingmatrices. Biotechnol. Bioeng. 55:831–840.

54. Speck, E. L., and E. Freese. 1973. Control of metabolite secretion in Bacillussubtilis. J. Gen. Microbiol. 78:261–265.

55. Stahmann, K. P., J. L. Revuelta, and H. Seulberger. 2000. Three biotechnicalprocesses using Ashbya gossypii, Candida famata, or Bacillus subtilis competewith chemical riboflavin production. Appl. Microbiol. Biotechnol. 53:509–516.

56. Stephanopoulos, G., and J. J. Vallino. 1991. Network rigidity and metabolicengineering in metabolite overproduction. Science 252:1675–1681.

57. Stülke, J., and W. Hillen. 2000. Regulation of carbon catabolism in Bacillusspecies. Annu. Rev. Microbiol. 54:849–880.

58. Szyperski, T. 1995. Biosynthetically directed fractional 13C-labeling of pro-teinogenic amino acids. An efficient analytical tool to investigate intermedi-ary metabolism. Eur. J. Biochem. 232:433–448.

59. Szyperski, T., R. W. Glaser, M. Hochuli, J. Fiaux, U. Sauer, J. E. Bailey, andK. Wüthrich. 1999. Bioreaction network topology and metabolic flux ratioanalysis by biosynthetic fractional 13C labeling and two-dimensional NMRspectroscopy. Metab. Eng. 1:187–197.

60. Tran, Q. H., and G. Unden. 1998. Changes in the proton potential and thecellular energetics of Escherichia coli during growth by aerobic and anaerobicrespiration or by fermentation. Eur. J. Biochem. 251:538–543.

61. van Winden, W., D. Schipper, P. Verheijen, and J. Heijnen. 2001. Innova-tions in generation and analysis of 2D [13C,1H] COSY NMR spectra formetabolic flux analysis purposes. Metab. Eng. 3:322–343.

62. Varma, A., and B. O. Palsson. 1994. Metabolic flux balancing: basic concepts,scientific and practical use. Bio/Technology 12:994–998.

63. Wiechert, W., M. Möllney, N. Isermann, M. Wurzel, and A. A. de Graaf.1999. Bidirectional reaction steps in metabolic networks. III. Explicit solu-tion and analysis of isotopomer labeling systems. Biotechnol. Bioeng. 66:69–85.

64. Willecke, K., and A. B. Pardee. 1971. Inducible transport of citrate in agram-positive bacterium, Bacillus subtilis. J. Biol. Chem. 248:807–814.



http://aem.asm

.org/D

ownloaded from

http://aem.asm.org/

intracellular carbon fluxes in riboffavin-producing bacillus subtilis

Documents