Effects of Furfural on the Respiratory Metabolism of Saccharomyces cerevisiae in Glucose-Limited Chemostats

Effects of furfural on the aerobic metabolism of the yeast Saccharomycescerevisiae were studied by performing chemostat experiments,and the kinetics of furfural conversion was analyzed by performingdynamic experiments. Furfural, an important inhibitor presentin lignocellulosic hydrolysates, was shown to have an inhibitoryeffect on yeast cells growing respiratively which was much greaterthan the inhibitory effect previously observed for anaerobicallygrowing yeast cells. The residual furfural concentration inthe bioreactor was close to zero at all steady states obtained,and it was found that furfural was exclusively converted tofuroic acid during respiratory growth. A metabolic flux analysisshowed that furfural affected fluxes involved in energy metabolism.There was a 50% increase in the specific respiratory activityat the highest steady-state furfural conversion rate. Higherfurfural conversion rates, obtained during pulse additions offurfural, resulted in respirofermentative metabolism, a decreasein the biomass yield, and formation of furfuryl alcohol in additionto furoic acid. Under anaerobic conditions, reduction of furfuralpartially replaced glycerol formation as a way to regenerateNAD⁺. At concentrations above the inlet concentration of furfural,which resulted in complete replacement of glycerol formationby furfuryl alcohol production, washout occurred. Similarly,when the maximum rate of oxidative conversion of furfural tofuroic acid was exceeded aerobically, washout occurred. Thus,during both aerobic growth and anaerobic growth, the abilityto tolerate furfural appears to be directly coupled to the abilityto convert furfural to less inhibitory compounds.

INTRODUCTION

Top

Introduction

An important problem in fermentative conversion of lignocelluloseto ethanol is the severe inhibitory effects often exerted bylignocellulosic hydrolysates (17, 21). Furfural is a characteristiccompound present in dilute acid hydrolysates, particularly inhydrolysates from deciduous woods, in which the furfural concentrationcan be about 2 to 3 g/liter (37). Furfural has been found toseverely inhibit metabolism in the yeast Saccharomyces cerevisiaeunder anaerobic conditions both during batch cultivation (5,7, 27) and in glucose-limited chemostats (9, 12, 35). Furfuralhas been reported to have inhibitory effects on the specificgrowth rate, as well as on the fermentation rate of yeasts (4,28). However, only few studies have described the effects offurfural under aerobic conditions. These effects are important,since inhibition caused by furfural during respiratory growthhas a great impact on yeast propagation in an ethanol productionplant based on a lignocellulosic feedstock.

Taherzadeh et al. (39) compared the levels of conversion offurfural in anaerobic and aerobic batch cultures of S. cerevisiaegrowing on glucose. It was found that furfural was mainly convertedto furfuryl alcohol by exponentially growing cells under bothconditions and that the specific conversion rate was 0.6 g/g· h. Almost the same value for the maximum specific conversionrate of furfural was found in a previous study, in which pulseadditions of furfural were made to anaerobic glucose-limitedcontinuous cultures of S. cerevisiae (35). To investigate inhibitioneffects outside glycolysis, S. cerevisiae was also grown onthe nonfermentable carbon sources ethanol and acetic acid (38).Pulse addition of furfural under these conditions was foundto result in strong inhibition of cell growth until completeconversion of furfural occurred. Furthermore, more than 85%of the furfural was converted to furfuryl alcohol, and furoicacid was a minor conversion product when ethanol was used asthe carbon and energy source.

The objective of the present work was to study the physiologicaleffects of furfural on the yeast S. cerevisiae during respiratorygrowth on glucose. Complete respiratory sugar metabolism canbe achieved only at specific growth rates below a critical value.Therefore, chemostat experiments were performed with a dilutionrate of 0.10 h^-1 and different concentrations of furfural inthe inlet medium. The global effects of furfural on metabolismwere analyzed by estimating the steady-state metabolic fluxdistributions by metabolite balancing. Metabolic flux analysisshows the degrees of participation of various pathways in theoverall metabolism. Comparisons of metabolic flux maps obtainedunder different conditions can provide important informationregarding substrate utilization, product formation, and energymetabolism (25, 41). Results obtained under steady-state conditionsin the present work and steady-state results obtained previouslyunder anaerobic conditions (35) provided the basis for a metabolicflux analysis in which the effects of furfural on aerobic metabolismand anaerobic metabolism of the yeast S.cerevisiae were compared.Transient experiments were also performed, in which pulse additionsof furfural were made to continuous cultures. In this way, thedynamics of furfural conversion could be studied at high furfuralconcentrations.

MATERIALS AND METHODS

Top

Materials and Methods

Yeast strain and media.
S.cerevisiae CBS 8066 obtained from the Centraalbureau voorSchimmelcultures (Delft, The Netherlands) was used in this study.Agar plates containing YEPD medium (2% yeast extract, 1% peptone,2% D-glucose, 2% agar) were used to maintain the strain. Inoculumcultures were grown in 300-ml cotton-plugged conical flaskswith a liquid volume of 100 ml on a rotary shaker at 170 rpmand 30°C for 24 h. The growth medium was a defined syntheticmedium described by Taherzadeh et al. (40) containing glucose(10 g/liter) as the carbon and energy source.

Cultivation conditions.
Aerobic continuous cultivation was performed at 30°C ina Belach BR 0.4 bioreactor (Belach Bioteknik AB, Solna, Sweden)with a stirrer speed of 800 rpm. The working volume of the culturewas kept at 0.4 liter, and the pH of the medium was controlledat 5.00 ± 0.05 by addition of 2 M NaOH. The fermentorwas flushed with air at a flow rate of 400 ml/min, which wascontrolled by a mass flow controller (Hi-Tech, Ruurlo, The Netherlands).At all times the dissolved oxygen tension under steady-stateconditions was more than 60% of the air saturation value. Aspreviously reported for other strains of S. cerevisiae, theCBS 8066 population exhibited spontaneous oscillations duringaerobic cultivation. This phenomenon was presumably due to spontaneouscell synchronization (11, 29). The oscillations eventually dampedout, and no measurements were obtained until a completely nonoscillatorysteady state had been reached.

In the steady-state experiments, no furfural was present inthe inlet medium at the first steady state. Subsequently, theconcentration of furfural in the feed solution was graduallyincreased until washout occurred. Between different inlet concentrationsthe system was allowed to reach a steady state. We assumed thatsteady-state conditions had been reached when the biomass concentrationand the specific rates of carbon dioxide production and oxygenconsumption remained constant for at least five residence times.All experiments were performed at a dilution rate of 0.10 h^-1.

In the pulse addition experiments, different amounts of furfural(resulting in concentrations of 4, 8, and 12 g/liter in thefermentor) were injected directly into steady-state continuouscultures growing on furfural-free medium. These experimentswere also performed at a dilution rate of 0.10 h^-1.

Analytical methods. (i) Gas analysis.
The carbon dioxide and oxygen contents in the outlet gas weremeasured continuously with an acoustic gas monitor (model 1308;Bruel & Kjaer, Naerum, Denmark) (8). The gas measurementsignals were averaged for 30 s. The instrument was calibratedwith gas containing 5% CO₂, 20% O₂, and nitrogen as an inertgas.

(ii) Metabolites and biomass.
Samples taken from the broth were centrifuged, and the supernatantswere analyzed by high-performance liquid chromatography. Theamounts of glucose, ethanol, glycerol, pyruvic acid, succinicacid, furfural, furfuryl alcohol, and furoic acid were determinedby using an Aminex ion-exchange column (HPX-87H; Bio-Rad, Richmond,Calif.) (37).

Cell concentrations were determined from measurements of absorbanceat 610 nm after samples were diluted to obtain an optical densityof less than 0.5. In this range, the absorbance values werefound to be linearly related to cell dry weight. The opticaldensity was calibrated with dry weight measurements by usingduplicate 3-ml samples, which were centrifuged, washed oncewith distilled water, and dried at 103°C for 24 h. Duplicatesamples differed by no more than 2%.

The total cellular protein content was determined by a modifiedbiuret method by using bovine serum albumin as the standard(44). The total cellular RNA content was determined as describedby Benthin et al. (6), except that the concentrations of coldperchloric acid were changed. Thus, samples were washed anddeproteinized three times with cold 0.2 M perchloric acid, solubilizedin 0.3 M KOH at 37°C, and acidified by addition of cold1.7 M perchloric acid (13). The precipitate was washed twicewith 0.2 M perchloric acid, and after centrifugation the absorbanceat 260 nm of the supernatant was determined with 0.2 M perchloricacid as the reference (18).

(iii) Calculation of specific uptake and production rates.
Carbon recovery values were calculated from the production ofbiomass and metabolites and the carbon dioxide evolution rate.The composition of the ash-free biomass was assumed to be CH_1.76O_0.56N_0.17(33).

The respiratory quotient (RQ) was calculated from the measuredexhaust gas mole fractions by using the following equation.

where y_j,in and y_j,out are the mole fractions ofj in the inlet and outlet gas, respectively.

Specific uptake and production rates were determined based onthe biomass concentrations. Volumetric conversion rates werecalculated from mass balances in the bioreactor at steady state.Degrees of reduction balance were calculated by the method ofRoels (33).

Stoichiometric network models.
The stoichiometric models used to calculate the metabolic fluxdistributions under aerobic and anaerobic conditions are givenin the Appendix.

The metabolic network used for anaerobic conditions comprised50 compounds in 44 intracellular reactions. The condition numberof the stoichiometric matrix was 89, indicating a numericallywell-conditioned system. The anaerobic model is presented anddiscussed in detail elsewhere (14). However, some importantpoints are also mentioned here. In order to obtain as good fitas possible with experimental data under anaerobic conditions,it was assumed that in dehydrogenase reactions NADP⁺ and NADHwere used as cofactors in cases in which the cofactor specificitywas uncertain (1, 2). Thus, in the synthesis of amino acids,the homoserine dehydrogenase (Hom6p), D1-pyrroline-5-carboxylatereductase (Pro3p), and saccharopine reductase (Lys9) were assumedto proceed with cytosolic NADH in order to minimize the amountof NADH formed in anabolism (2). For the same reason, 5-methyltetrahydrofolate was assumed to be formed via the NADPH-specificmethylene tetrahydrofolate reductase (Met13p) (32, 34). 5-Formyltetrahydrofolate formation was assumed to proceed via the NADP-specificAde3p (46). Glycine formation was assumed to proceed both viathe cytosolic serine hydroxymethyltransferase (Shm2p) and viathreonine aldolase (Gly1p) (23), in order to balance the formationof 5,10-methylene tetrahydrofolate. Acetyl-coenzyme A (acetyl-CoA)used in the tricarboxylic acid (TCA) cycle was assumed to beformed via pyruvate decarboxylase and a cytosolic NADP⁺-coupledacetaldehyde dehydrogenase, which was followed by passive transportinto the mitochondria (25). Furthermore, it was assumed thatfurfuryl alcohol was produced via an NADH-coupled alcohol dehydrogenase.

The metabolic network used for aerobic conditions comprised51 compounds in 45 intracellular reactions. The condition numberof the stoichiometric matrix was 88. Under aerobic conditions,acetyl-CoA used in the TCA cycle was assumed to be formed viapyruvate dehydrogenase, whereas cytosolic acetyl-CoA used inbiosynthesis was assumed to be formed via pyruvate decarboxylaseand NAD⁺-coupled acetaldehyde dehydrogenase (25). It was notpossible to include both the pyruvate dehydrogenase (PDH) andthe PDH bypass for formation of mitochondrial acetyl-CoA (15)since this led to a dramatic increase in the condition numberof the stoichiometric matrix. Furoic acid was assumed to beformed via an NAD⁺-coupled aldehyde dehydrogenase under aerobicconditions. The P/O ratio was assumed to be 1.1 for both theexternal and internal NADH dehydrogenases (42).

The amount of ATP used for intracellular pH regulation was estimatedon the basis of proton balance by using the plasma membraneH⁺-translocating ATPase to account for export of excess protons.An ATP balance was used to estimate the amount of ATP consumedin all unknown processes; hence, this was referred to as nonspecificATP hydrolysis.

The amino acid composition of cellular protein was obtainedfrom the study of Albers et al. (1), and the nucleotide compositionof RNA was obtained from the study of Oura (26). The cellularprotein and RNA contents were measured as described above. Thebiomass was assumed to contain 5% lipids under aerobic conditionsand 2% lipids under anaerobic conditions (average lipid compositions)(14, 31). Under anaerobic conditions, unsaturated fatty acids,ergosterol, and inositol were assumed to be taken up from themedium, while under aerobic conditions only inositol was providedin the medium. This was taken into account by modifying thereaction stoichiometry in these cases (see Appendix). The monomercompositions of proteins, RNA, and lipids were assumed to beconstant under all the conditions investigated. The remainderof the biomass was assumed to be ash (5%) and carbohydrates(balance up to 100%). The unknown intracellular reaction rateswere estimated by weighted, constrained optimization as previouslydescribed (14).

RESULTS

Top

Results

Aerobic continuous cultivation.
Aerobic continuous cultivation of S.cerevisiae was carried outwith glucose as the carbon and energy source at a dilution rateof 0.10 h^-1. Under these conditions the glucose metabolism wasfully respiratory; i.e., no ethanol or other fermentation productscould be detected in the reactor. The measured biomass yieldon glucose was 0.52 g/g, and the only other products obtainedin furfural-free medium were carbon dioxide and biomass. Thespecific rates of CO₂ production and O₂ consumption were approximatelyequal, resulting in an RQ of around 1 (Table 1).

View this table:
[in this window]
[in a new window]
TABLE 1. Experimentally determined and estimated specific formation rates obtained from aerobic chemostat cultures of S. cerevisiae growing on glucose (10 g/liter) as a carbon and energy source with different levels of furfural present in the inlet medium^a

The biomass of and products formed by S. cerevisiae were investigatedby using different inlet concentrations of furfural (1.14, 2.25,and 3.00 g/liter) at a dilution rate of 0.10 h^-1. The resultsare summarized in Table 1. The glucose metabolism was fullyrespiratory, and no ethanol or glycerol was formed under anyof the steady-state conditions.

Furoic acid was the only furfural-derived product found, andno residual furfural was detected in the bioreactor during thesteady states obtained in the presence of 1.14 and 2.25 g offurfural per liter in the feed solution. This is in sharp contrastto fermentative or respirofermentative growth, in which furfurylalcohol and a recently identified probable acyloin condensationproduct of furfural and pyruvate, 3-(2-furfuryl)-2-hydroxy-2-methyl-3-oxo-propanoicacid (FHMOPA), were found to be the dominant products (35, 39).Washout occurred when the furfural concentration in the feedsolution was increased to 3.00 g/liter, suggesting that themaximum specific conversion rate of furfural is between 0.054and 0.072 g/g · h under steady-state conditions at adilution rate of 0.10 h^-1.

Flux distributions in aerobic and anaerobic continuous cultures.
The systemic effects of furfural were investigated by estimatingthe metabolic flux distributions for the aerobic steady-statecultures with and without furfural in the medium by using acompartmentalized, stoichiometric network model (Fig. 1). Toenable a comparison between the furfural effects on aerobicmetabolism and the furfural effects on anaerobic metabolism,anaerobic steady-state results for a dilution rate 0.10 h^-1published previously (35) were also analyzed by using the stoichiometricmodel (Fig. 2). The net production and consumption rates predictedby the model are shown in Tables 1 and 2 for aerobic and anaerobicconditions, respectively. The predicted rates fulfill mass andredox balances and are weighted according to the variances inthe measurements (14) (Tables 1 and 2). They are therefore betterestimates of the true rates, provided that a correct model hasbeen postulated.

View larger version (38K):
[in this window]
[in a new window]
FIG. 1. Flux distribution during respiratory growth of S. cerevisiae at a dilution rate of 0.10 h^-1 calculated on the basis of steady-state data obtained in furfural-free medium (top values) and in the presence of 2.25 g of furfural per liter in the feed solution (bottom values in italics). The data are expressed in C millimoles per C mole of glucose.

View larger version (35K):
[in this window]
[in a new window]
FIG. 2. Flux distribution during anaerobic growth of S. cerevisiae at a dilution rate of 0.10 h^-1 calculated on the basis of steady-state data obtained in furfural-free medium (top values) and in the presence of 5.8 g of furfural per liter in the feed solution (bottom values in italics). The data are expressed in C millimoles per C mole of glucose.

View this table:
[in this window]
[in a new window]
TABLE 2. Experimentally determined and estimated specific formation rates obtained from anaerobic chemostat cultures of S. cerevisiae growing on glucose (50 g/liter) as a carbon and energy source with different levels of furfural present in the inlet medium^a

For furfural-free medium, the model describes known metabolicbehavior (Fig. 1). This means that during respiratory growth,biomass was the major carbon-containing product. The pentosephosphate pathway was very active, providing NADPH and precursorsfor biosynthesis; glycolysis provided biosynthetic precursors,as well as pyruvate, and the TCA cycle was highly active.

With furfural present in the medium under aerobic conditions,the specific rates of glycolysis, TCA cycle reactions, and respirationincreased by 30, 50, and 50%, respectively, for an inlet furfuralconcentration of 2.25 g/liter (Fig. 1). The specific rates ofthe biosynthetic fluxes changed only slightly due to changesin the cellular macromolecular composition. This resulted ina more-than-threefold increase in the nonspecific ATP utilization.The biomass yield on ATP (Y_ATP) decreased from 16.0 g of biomass/molof ATP to 10.6 g/mol, and the nonspecific hydrolysis of ATPincreased from 0.4 to 1.2 mol of ATP/C-mol of biomass (i.e.,the amount of biomass containing 12 g of carbon) when the mediumwas changed from furfural-free medium to a medium containing2.25 g of furfural per liter (Table 1). During respiratory growththe only product of the furfural conversion reaction was furoicacid. In the model this reaction was assumed to proceed withNAD⁺ as a cofactor. However, the additional NADH produced inthis reaction accounted for only 27% of the additional NADHrespired, and the rest originated from the increasing ratesof the glyceraldehyde dehydrogenase and TCA cycle reactions.

During anaerobic growth, glycolysis was highly active as anATP-producing pathway, and ethanol was the main product; onlya small fraction of the glucose was converted in the pentosephosphate pathway (Fig. 2). Glycerol formation provided additionalreoxidation of cytosolic NADH. The Adh3p provided a redox shuttlefor mitochondrial NADH (3), and the TCA cycle operated in aforked mode, providing biosynthetic precursors with minimumnet NADH formation (25).

Under anaerobic conditions the main product of the furfuralconversion reaction was furfuryl alcohol. This compound wasassumed to be produced with NADH as a cofactor, possibly byalcohol dehydrogenase I (10, 16, 24). As a consequence, furfurylalcohol production led to an almost stoichiometric decreasein the flux to glycerol (Fig. 2). In the furfural-free medium,24 mmol of NADH/C-mol of glucose was converted in glycerol synthesis,whereas 26 mmol of NADH/C-mol of glucose was converted via furfuralreduction at the highest furfural concentration. In the furfural-freemedium there was a split in the TCA cycle into an oxidativebranch and a reductive branch (Fig. 2). When furfural was presentin the medium, all the TCA cycle reactions instead appearedto be slightly active in the oxidative direction. This alsoled to an 85% decrease in the succinate yield. Furthermore,there was an increase in the movement of reducing equivalentsout of the mitochondria via the Adh3p shuttle (25).

At intermediate furfural concentrations in the medium, the biomassyield actually increased by 13%, while the specific glycolyticrate decreased by 11%. This resulted in an increase in the Y_ATPfrom 12.6 to 14.2 g of biomass/mol of ATP and a decrease inthe nonspecific ATP consumption from 0.9 to 0.6 mol of ATP/C-molof biomass (Table 2). However, at the highest furfural concentration,the Y_ATP dropped to 10.1 g/mol, and the nonspecific ATP consumptionincreased by 50% (from 0.9 to 1.3 mol of ATP/C-mol of biomass)compared to the nonspecific ATP consumption in the furfural-freemedium (Table 2). This was accompanied by a 12% increase inthe specific ethanol production rate (Table 2) and a 9% increasein the ethanol yield (Fig. 2). The calculated specific rateof ethanol production was about 8% higher than the measuredrate, which can be explained by evaporation of ethanol (36).

Pulse addition of furfural to respiratory continuous cultures.
The specific rate of conversion of furfural and the inhibitoryeffects on cell growth were studied by making pulse additionsof furfural (4, 8, and 12 g/liter) to continuous cultures growingat a respiratory steady state. By making direct pulse additionsof furfural, it was possible to obtain a high concentrationof furfural in the bioreactor that could not be obtained incontinuous cultures at a steady state. Furfural significantlyaffected the specific growth rate. Furthermore, the metabolismimmediately switched to mixed respiration and fermentation,as shown by both the amount of ethanol formed and increasingRQ values. Formation of ethanol was measured within 15 min afterthe pulse additions of furfural (Fig. 3). The RQ immediatelyincreased as a result of furfural addition. The maximum value,4.5, was reached within 0.5 h, and the RQ subsequently decreasedslowly to the initial steady-state value after complete conversionof furfural (data not shown). No transient formation of glycerolwas detected.

View larger version (36K):
[in this window]
[in a new window]
FIG. 3. Metabolite concentrations following pulse additions of furfural (4, 8, and 12 g/liter) to respiratory chemostat cultures of S. cerevisiae growing at a dilution rate of 0.10 h^-1.

Pulse addition of furfural resulted in a dramatic decrease inthe biomass concentration (Fig. 3). The reason for this couldhave been the much lower biomass yield as a result of respirofermentativegrowth or an almost complete cessation of growth in the presenceof furfural. Due to the short transition time, it was not possibleto distinguish between these two explanations. After all ofthe furfural was consumed, cell growth resumed, and the samerespiratory steady-state conditions that were present beforethe pulse addition of furfural were obtained.

The residual glucose concentration was not affected by pulseaddition of the lowest concentration of furfural (4 g/liter).However, at higher concentrations of furfural (8 and 12 g/liter),the residual glucose concentration transiently increased duringfurfural conversion (Fig. 3). Also, metabolites, such as succinicacid, acetic acid, and pyruvic acid, were affected. Acetateand pyruvate were transiently excreted into the medium, butthey were consumed again when growth was restored. Succinateexcretion also increased, but succinate was not consumed again(Fig. 3).

The decrease in the concentration of furfural following a pulseaddition was due to both dilution and conversion by the cells.In contrast to the conversion of furfural at a steady state,the main conversion products of furfural in this case were furfurylalcohol, furoic acid, and FHMOPA. The maximum specific rateof conversion of furfural was estimated to be 0.26 ±0.01 g/g · h by using a Michaelis-Menten-type kineticmodel (35). Most of the furfural was converted to furfuryl alcohol,for which the maximum specific rate of production was 0.19 ±0.01 g/g · h. The maximum specific rate of productionof furoic acid was estimated to 0.040 ± 0.005 g/g ·h. The Michaelis-Menten constants for furfural were found tobe about 0.1 g/liter for conversion to furfuryl alcohol andalmost zero for conversion to furoic acid. This suggests thatthe affinity for furfural is high, and this suggestion is supportedby the fact that the furfural concentrations were below thedetection limit under the steady-state conditions.

DISCUSSION

Top

Discussion

Furfural has previously been found to have profound effectson the growth of and metabolite formation by S. cerevisiae duringanaerobic growth on glucose (27, 35), respirofermentative growthon glucose (39), and respiratory growth on ethanol (38). Therefore,the strong effect of furfural on respiratory growth on glucosefound in the present study was expected. More surprising, however,was the finding that the only conversion product obtained fromfurfural in chemostat cultures was furoic acid. The main conversionproduct of furfural in all previously reported studies was furfurylalcohol. Furfuryl alcohol was the main product of furfural conversionduring anaerobic continuous cultivation, and no furoic acidwas detected (35). Under transient conditions, after pulse additionof furfural, FHMOPA was found in addition to furfuryl alcohol,but no furoic acid was formed during anaerobic batch cultivation(39) or continuous cultivation on glucose (35). Biotransformationof furfural to furoic acid has been reported only for yeastcells growing on a nonfermentable carbon source, such as ethanolor acetic acid (38). However, furfuryl alcohol was also thedominant conversion product in those cases.

The maximum rate of conversion of furfural to furoic acid understeady-state conditions was between 0.054 and 0.072 g/g ·h (at the latter conversion rate washout occurred). On the otherhand, at high concentrations of furfural (such as the concentrationafter a pulse addition), a maximum specific rate of conversionof about 0.26 g/g · h was found. In this case, however,predominantly furfuryl alcohol was obtained (Fig. 3). Consequently,the maximum specific rate of conversion of furfural could bedivided into a maximum specific rate of reduction to furfurylalcohol of 0.19 g/g · h and a maximum specific rate ofoxidation to furoic acid of 0.04 g/g · h. The maximumspecific rate of furoic acid production was similar to the maximumfurfural conversion rate obtained under steady-state conditionsduring respiratory growth, in which case the only product ofthe furfural conversion was furoic acid. This may indicate thatthe oxidation of furfural to furoic acid proceeded at the maximumrate and that, in addition, overflow conversion of furfuralto furfuryl alcohol occurred during the pulse addition experiments.Although the maximum value for the aerobic specific rate ofconversion of furfural (0.26 g/g · h) is higher thanthe aerobic chemostat conversion rate, this value is clearlylower than the corresponding maximum conversion rate obtainedpreviously during fermentative metabolism (0.62 g/g ·h) both in continuous cultures (35) and in batch cultures (39)of S. cerevisiae.

It was found that low concentrations of furfural did not haveas strong an effect on cellular metabolism during fermentativegrowth (Fig. 2) as during respiratory growth (Fig. 1). The mainproduct of the conversion of furfural is furfuryl alcohol underanaerobic conditions. The reduction of furfural can thereforeact as an alternative redox sink, redirecting the NADH poolto furfuryl alcohol formation, which results in a substantialdecrease in the glycerol yield and a slight increase in biomassformation (27, 35). This was also reflected in an increase inthe Y_ATP at the intermediate levels of furfural addition (Table2). Since glycerol formation is accompanied by net ATP consumption,an externally supplied redox sink (e.g., in the form of furfural)could result in more ATP being available for biomass synthesis.Furfural concentrations that are too high, however, result ininhibition of biosynthesis, and growth stops. A limiting factorfor furfuryl alcohol production could be that the high demandfor NADH for the reduction of furfural cannot be met duringanaerobic growth (35).

The analysis also showed that furfural affects fluxes involvedin energy metabolism (Fig. 1). The flux model showed that therewas 50% higher specific respiratory activity, including allthe reaction steps involved, together with an almost threefoldincrease in nonspecific ATP consumption when 2.25 g of furfuralper liter was present in the inlet medium. This was also reflectedin the lower Y_ATP (Table 1). The increase in the energy requirementcould be a direct effect of the furoic acid formed. In the modelused, furoic acid was assumed to be exported via diffusion inits undissociated form. On the other hand, if the dissociatedacid is exported via an active anion transporter, additionalprotons would be liberated both due to the dissociation of theacid and due to the ATP hydrolysis linked to the transport.The export of the anions could tentatively be achieved via theATP binding cassette transporter Pdr12p, which has been shownto confer resistance to water-soluble, monocarboxylic acidswith chains that are one to seven C atoms long, including sorbicacid and benzoic acid (19, 30). In total, 0.8 mol of ATP/C-molof furoic acid would be used in conjunction with the formationand expulsion of the furoic acid and liberated protons, assuminga stoichiometry of 1 ATP per furoate exported via the Pdr12ptransporter. When this information was included in the model,the level of nonspecific ATP hydrolysis was reduced to 1.0 molof ATP/C-mol of glucose at the highest inlet furfural concentration,a value which is still 2.5-fold higher than the value obtainedin the absence of furfural. Furthermore, the low pK_a of furoicacid (pK_a 3.17) makes a futile cycle due to rediffusion of theprotonated acid across the plasma membrane unlikely (20, 43).Thus, the increase in the nonspecific energy requirement cannotbe explained solely by transport energy requirements.

The variation in the preferred conversion products obtainedfrom furfural may be due to differences in enzyme activitiesand/or intracellular metabolite levels. Reduction of furfuralto furfuryl alcohol is most likely catalyzed by an NADH-coupledalcohol dehydrogenase (5, 28, 38, 45). Assuming that NADPH isa cofactor for furfuryl alcohol production failed to explainthe observed metabolite yields and resulted in completely unreasonablemetabolic flux distributions. Since the oxidation of furfuralto furoic acid is an oxidation of an aldehyde to a carboxylicacid, the reaction is likely to proceed via an aldehyde dehydrogenase.Lower aerobic alcohol dehydrogenase activity, in combinationwith a lower intracellular NADH level, could well explain thepreference for furoic acid formation over furfuryl alcohol formationunder aerobic conditions. The cofactor requirement of this reactionhas not been determined. However, several facts indicate thatNAD⁺ is the preferred cofactor in the oxidation of furfuralto furoic acid. Modig et al. showed that commercially availableNAD⁺-dependent aldehyde dehydrogenase was active in convertingfurfural to furoic acid (22). Furthermore, using NAD⁺ as a cofactorfor the aldehyde dehydrogenase reaction in our model resultedin a metabolic flux distribution which agrees better with thedistribution described by Gombert et al. (15). By using ¹³Clabeling methodology, these authors estimated that the fluxthrough the oxidative part of the pentose phosphate pathwayis 44.2% on a molar basis. Importantly, this was done withoutany use of cofactor balances in the study. When we used NAD⁺in our stoichiometric model, we obtained a value of 35.4% (Fig.1), while the use of NADP⁺ resulted in a much lower value, 28.6%(data not shown). Furthermore, if NADP⁺ was used, the pentosephosphate pathway flux decreased dramatically when furfuralwas present, whereas with NAD⁺ it decreased only in proportionto the decrease in biomass formation (Fig. 1). Moreover, theincreased respiratory activity obtained when furfural was presentindicates that there is an increased rate of NADH turnover inthe respiratory chain. As shown in the metabolic flux analysis,the additional NADH accompanying furoic acid production actuallyexplained only 27% of the increase in the respiratory activity.The rest of the NADH was formed due to an increase in TCA cycleactivity. Taken together, these observations support the hypothesisthat there is an NAD⁺-coupled reaction mechanism.

One possible reason for the lack of a respiratory steady stateat an inlet furfural concentration higher than 2.25 g/litercould be a limitation in the oxidative capacity, either in theoxidation of furfural to furoic acid or in the oxidation ofNADH to NAD⁺. A limitation in the oxidative capacity is in factsuggested by the pulse addition experiments. When the maximumoxidation capacity is reached, ethanol formation is induced,which results in a large decrease in the biomass yield. Thisin turn results in a decreased overall ability to convert theadded furfural. As a consequence, the furfural concentrationincreases, which leads to inhibition of essential enzymes, mostlikely including the pyruvate dehydrogenase and aldehyde dehydrogenaseenzymes (22). The final result is rapid washout of the culture.Whether the limitation is in the enzymatic conversion of furfuralto furoic acid or in the turnover of the NADH produced couldnot be deduced from the experiments performed in this study.

Based on these observations, we concluded that the ability tosurvive in the presence of furfural is directly related to theability to convert furfural to less inhibitory compounds underaerobic conditions, as well as under anaerobic conditions. Underaerobic conditions, a shift to respirofermentative metabolismoccurs when the capacity to oxidize furfural is exceeded, whichleads to a decrease in the biomass yield. Under anaerobic conditions,the supply of NADH appears to limit the reduction of furfuralat conversion rates above rates at which glycerol formationhas been totally replaced by furfuryl alcohol production. Inboth these cases, the result is that the concentration of furfuralincreases, probably leading to additional direct inhibitoryeffects on essential enzymatic steps in the central metabolicpathways.

APPENDIX

Top

Appendix

Biochemical reactions in the metabolic network model are expressedin C-mol stoichiometry. Water and the cofactors ADP, CoA, NAD⁺,NADP⁺, and tetrahydrofolate are not included. The main productin each reaction is indicated by boldface type; subscript mitand cyt indicate mitochondria and cytoplasm, respectively. Reactions11a, 14, 17a, 40a, and 46a occur only under anaerobic conditions,and reactions 11b, 17b, 40b, 44, 45, and 46b occur only underaerobic conditions. Glycolysis and fermentative products

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11a)

(11b)

(12)

(13)

(14)

(15)

(16)
TCA cycle

(17a)

(17b)

(18)

(19)

(20)

(21)

(22)

(23)
Pentose phosphatepathway

(24)

(25)

(26)
Proton symport of ions

(27)

(28)

(29)
Synthesis of amino acids

(30)

(31)

(32)

(33)

(34)

(35)

(36)
Biosynthesis of proteins, RNA,polysaccharides, and lipid

(37)

(38)

(39)

(40a)

(40b)
ATPhydrolysis

(41)

(42)
Respiration

(43)

(44)

(45)
Furfural conversion

(46a)

(46b)

ACKNOWLEDGMENTS

This work was financially supported by The Swedish NationalEnergy Administration.

FOOTNOTES

* Corresponding author. Mailing address: Department of Chemical Engineering, Lund Institute of Technology, Lund University, S-221 00 Lund, Sweden. Phone: 46-46-222-08-62. Fax: 46-46-14-91-56. E-mail: Gunnar.Liden{at}chemeng.lth.se.

REFERENCES

Top