Applied Microbiology and Biotechnology

, Volume 71, Issue 3, pp 339–349

Tolerance to furfural-induced stress is associated with pentose phosphate pathway genes ZWF1, GND1, RPE1, and TKL1 in Saccharomyces cerevisiae


    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
  • B. S. Dien
    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
  • N. N. Nichols
    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
  • P. J. Slininger
    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
  • Z. L. Liu
    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
  • C. D. Skory
    • National Center for Agriculture Utilization Research, Agriculture Research ServiceUSDA
Applied Microbial and Cell Physiology

DOI: 10.1007/s00253-005-0142-3

Cite this article as:
Gorsich, S.W., Dien, B.S., Nichols, N.N. et al. Appl Microbiol Biotechnol (2006) 71: 339. doi:10.1007/s00253-005-0142-3


Engineering yeast to be more tolerant to fermentation inhibitors, furfural and 5-hydroxymethylfurfural (HMF), will lead to more efficient lignocellulose to ethanol bioconversion. To identify target genes involved in furfural tolerance, a Saccharomyces cerevisiae gene disruption library was screened for mutants with growth deficiencies in the presence of furfural. It was hypothesized that overexpression of these genes would provide a growth benefit in the presence of furfural. Sixty two mutants were identified whose corresponding genes function in a wide spectrum of physiological pathways, suggesting that furfural tolerance is a complex process. We focused on four mutants, zwf1, gnd1, rpe1, and tkl1, which represent genes encoding pentose phosphate pathway (PPP) enzymes. At various concentrations of furfural and HMF, a clear association with higher sensitivity to these inhibitors was demonstrated in these mutants. PPP mutants were inefficient at reducing furfural to the less toxic furfuryl alcohol, which we propose is a result of an overall decreased abundance of reducing equivalents or to NADPH's role in stress tolerance. Overexpression of ZWF1 in S. cerevisiae allowed growth at furfural concentrations that are normally toxic. These results demonstrate a strong relationship between PPP genes and furfural tolerance and provide additional putative target genes involved in furfural tolerance.


The role of agriculture as a supplier of renewable energy has increased as the need for an alternative and environmentally friendly energy source becomes more apparent. In the 1970s, large-scale industrial production of fuel ethanol by fermentation of agricultural products began in Brazil and North America (Wheals et al. 1999). However, the net cost of feedstock for producing a gallon of ethanol can exceed the selling price of ethanol. Alternatively, various sources of lignocellulosic biomass such as agriculture wastes, wood, municipal solid wastes, and wastes from pulp and paper industries can serve as low-cost and abundant feedstocks for fuel ethanol production (Bothast and Saha 1997; Zaldivar et al. 2001). Nonetheless, many technological obstacles exist that need to be addressed before an economical biomass-to-ethanol process can exist.

One major constraint is the presence of multiple inhibitors, including furan derivatives, weak acids, and phenolic compounds, produced during biomass-to-ethanol processing. These inhibitors usually form during acid pretreatment of lignocellulose, which is necessary for efficient hemicellulose hydrolysis and enzymatic saccharification (Olsson and Hahn-Hagerbal 1996; Palmqvist and Hahn-Hagerdal 2000; Saha 2003). Two major inhibitors produced are furfural and 5-hydroxymethylfurfural (HMF), which are degradation products of xylose and glucose, respectively (Larsson et al. 1999; Palmqvist and Hahn-Hagerdal 2000). The concentration of these inhibitors can vary depending on the type of biomass and hydrolysis condition. In a typical corn stover-acid-treated hydrolysate, a concentration of around 2.2 g/l (∼23 mM) of furfural is observed (Maciel de Mancilha and Karim 2003). High concentrations of furfural and HMF affect yeast survival, growth rates, cell budding, ethanol yield, biomass yield, and biochemical enzyme activity, with furfural causing a more severe phenotype (Banerjee et al. 1981; Horvath et al. 2001; Modig et al. 2002; Palmqvist and Hahn-Hagerdal 2000; Sanchez 1988; Taherzadeh et al. 1999, 2000b). Saccharomyces cerevisiae is capable of detoxifying low concentrations of furfural and HMF during anaerobic growth by reduction to furfuryl alcohol or HMF alcohol (Diaz de Villegas et al. 1992; Horvath et al. 2001; Liu et al. 2004; Taherzadeh et al. 2000a; Villa et al. 1992). Furfural is reduced using a nicotinamide adenine dinucleotide (NADH)-dependent alcohol dehydrogenase, whereas HMF requires nicotinamide adenine dinucleotide phosphate (NADPH) (Palmqvist et al. 1999; Wahlbom and Hahn-Hagerdal 2002). As a consequence, free NADH and NADPH must be available to detoxify these inhibitors, and thus, pathways regulating reducing equivalents are likely to be important in furfural tolerance.

Resistance to furfural and HMF can be increased in S. cerevisiae by a process of mutation and selection (Liu et al. 2005). Therefore, we felt confident that altered regulation of specific native genes could provide improved resistance to furfural. In the present study, a S. cerevisiae disruption library was screened to identify potential target genes that alter furfural tolerance compared to the parent strain. It was hypothesized that efforts to increase furfural tolerance would be more effective if we focused on these identified genes. Surprisingly, more than 62 genes were found to be associated with sensitivity to furfural. For several reasons that are addressed in this paper, we focused on several mutants deficient in the pentose phosphate pathway (PPP). These mutants were evaluated at various concentrations of furfural, and a clear association with higher sensitivity to this inhibitor was demonstrated. We further showed that overexpression of one of the PPP genes, ZWF1, provided improved growth in the presence of toxic levels of furfural, thereby demonstrating the effectiveness of using this approach to identify genes involved in furfural tolerance.

Materials and methods

S. cerevisiae strains and growth conditions

Standard yeast culturing, transformation, and genetic analysis techniques were used (Guthrie and Fink 2002a,b; Sambrook and Russell 2001). The S. cerevisiae haploid Mata disruption library and its parent, BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0), were purchased from Open Biosystems (Huntsville, AL). Haploid strains were used for ease of genetic analysis purposes. Yeast were grown in either yeast extract peptone dextrose (YPD) medium (2% glucose, 2% Bacto peptone, and 1% yeast extract; 2% agar was added for solid plates) or in synthetic dextrose (SD) medium [2% glucose and 0.67% yeast nitrogen base without amino acids and with ammonium sulfate (US Biological L4092418); 2% agar was added for solid plates]. Synthetic medium was supplemented with either all essential amino acids and nucleic acids (SD-complete), without uracil (SD-URA), without leucine (SD-LEU), or without leucine and methionine (SD-LEU-MET) for selection of appropriate plasmids or with 200 μg/ml of G418 sulfate antibiotic (Geneticin, catalog number G1001, US Biological, Swampscott, MA) for selection of strains with a kanr disruption cassette.

Disruption library screening

The disruption library contained 4,848 haploid mutants, each with a different gene completely replaced by a kanr gene cassette and each placed in a separate well of 96-well plates (Winzeler et al. 1999). In addition, each plate contained parent BY4741 cells in one well. Using a 96-pin replicating tool, inocula from original 96-well plates were placed into 100 μl YPD. Plates were incubated at 30°C for approximately 24 h before transferring inocula to solid YPD plates with either 0 or 50 mM furfural and incubated at 30°C. Mutants and parent BY4741 cells were assayed daily for growth as determined by the presence or absence of colonies. Compared to parent BY4741 colonies, mutants that failed to form a colony were selected for further analysis.

Gene identification and verification

Disrupted genes in mutants of interest were identified from an annotated list (Open Biosystems) and verified by specifically amplifying disrupted loci using primer pairs homologous to the kanr disruption cassette and genomic regions outside the identified gene's open reading frame (yeast DNA sequences obtained from Saccharomyces genome database Select mutants were crossed back to the parent BY4741, sporulated, and its tetrads dissected using a Zeiss Axioskop 40 fitted with a micromanipulator (Carl Zeiss Microimaging Inc., Thornwood, NY). Sister spores were analyzed to verify that the kanr disruption marker cosegregated with furfural growth inhibition phenotypes.

Subcloning to test complementation and overexpression

ZWF1, GND1, TKL1, and RPE1 open reading frames plus promoter and terminator sequences were subcloned into pRS416 (CEN, URA3, and amp) (ATCC 87521) for complementation experiments. Primer pairs are as follows, with the first of each pair representing the forward primer, the restriction enzyme used in parentheses, and the engineered restriction site in upper case: ZWF1 (BamHI) ggggGGATCCgcgctactggaagcaccacgt and (SalI) ggggGTCGACgggccgagcgccgcagcagtt, GND1 (BamHI) ggggGGATCCggccatttcatacctcggcca and (SalI) ggggGTCGACcatcctcctctccttggtggt, RPE1 (HindIII) ggggAAGCTTcgcataacaaacgcgcttatc and (BamHI) ggggGGATCCagcgttggacctatacttcca, and TKL1 (BamHI) ggggGGATCCggttagcaagtttaagtgcta and (SalI) ggggGTCGACgccttttcgacttcatgtttg. ZWF1, GND1, TKL1, and RPE1 open reading frames were only cloned into pRS425-MET25 (multicopy, LEU2, and amp) (ATCC 87323) for overexpression experiments. Primer pairs used are as follows: ZWF1 (BamHI) ggggGGATCCatgagtgaaggccccgtcaaa and (XhoI) ggggCTCGAGctaattatccttcgtatcttc, GND1 (BamHI) ggggGGATCCatgtctgctgatttcggtttg and (XhoI) ggggCTCGAGttaagcttggtatgtagagga, RPE1 (BamHI) ggggGGATCCatggtcaaaccaattatagct and (XhoI) ggggCTCGAGctaatctagcaaatctctaga, and TKL1 (BamHI) ggggGGATCCatgactcaattcactgacatt and (XhoI) ggggCTCGAGttagaaagcttttttcaaagg. Amplified products and vectors were digested with appropriate restriction enzymes (New England BioLabs Inc., Beverly, MA), ligated together (Quick T4 DNA Ligase, New England BioLabs Inc.), and transformed into DH10B ElectroMax competent Escherichia. coli cells (Invitrogen Corp., Carlsbad, CA). Putative subclones were verified by sequencing and transformed into mutant S. cerevisiae strains (Guthrie and Fink 2002a).

Characterization of mutants

Optical density (OD) measurements (absorbance at 600 nm; 1/3 cm) were taken using a Power Wave X 340 96-well plate spectrophotometer (BIO-TEK Instruments Inc., Winooski, VT). Selected gene disruption mutants were grown overnight (approximately 18 h) in 3 ml of SD-complete medium, at 30°C, in airtight culture tubes. Overnight cultures were diluted to 0.4 OD units (A600nm) in SD-complete medium and grown under the same conditions until cultures reached an exponential phase, approximately 3 h. Exponential phase cultures were diluted to 0.2 OD units (A600nm) in fresh SD-complete medium under the same growth conditions with either 5–50 mM furfural (Sigma, St. Louis, MO, catalog 185914, stored under nitrogen gas), 20–40 mM HMF (Sigma, catalog H40807), 8–12% ethanol (AAPER Alcohol and Chemical Co., Shelbyville, KY), or no inhibitor. Optical densities were measured over 192 h, and growth curves were constructed. Cultures failing to double twice were characterized as unable to exit lag and enter the exponential phase. All growth experiments were repeated 2–8 times.

High-performance liquid chromatography (HPLC) analysis was conducted on a Spectra System HPLC system (Thermo Electron Corp., Waltham, MA) composed of an autosampler, quaternary pump, dual-wavelength UV detector, and data processor. Separation was performed on an Econosphere C18 5 m column (250×4.6 mm, Alltech Assoc. Inc., Deerfield, IL) at ambient temperature and eluted with a 43% (v/v) methanol 0.25% (v/v) acetic acid solution at a flow rate of 0.8 ml/min. Furfural was detected at 277 nm, and furfuryl alcohol was detected at 215 nm.


Screening for mutants with increased furfural sensitivity

To identify S. cerevisiae genes involved in furfural tolerance, a yeast library containing 4,848 complete gene disruptions was screened on YPD solid medium supplemented with 50 mM furfural. This concentration in YPD was determined to be low enough to have minimal effects on growth of the parent BY4741. Compared to BY4741, mutants that grew poorly or not at all on YPD + furfural but grew on YPD alone were selected. No mutants were selected that grew better in the presence of furfural compared to BY4741.

A total of 229 mutants identified as having increased sensitivity to furfural were further analyzed for their ability to grow (i.e., exit the lag phase and enter the exponential phase) in liquid synthetic complete medium containing dextrose supplemented with either 0, 25, or 50 mM furfural at 30°C. In synthetic medium, none of the strains were able to grow in 50 mM furfural (data not shown), and in 25 mM furfural, the parent BY4741 grew but only after a lag of 2 h. We have previously observed that S. cerevisiae is less sensitive to furfural in richer medium (i.e., YPD), but YPD is highly variable and not optimal for controlled experiments (data not shown). When mutants were grown in the presence of 25 mM furfural, four categories of growth behavior were observed: (1) failure to grow, (2) longer lag phases of 12–72 h and often lower final cell densities than wild type, (3) poor growth even in the absence of furfural, and (4) growth in a similar manner as the parent BY4741 (Table 1, categories I and II mutants only). It should be noted that in the absence of furfural, categories I and II mutants had similar growth patterns as BY4741 (examples in Fig. 1a and Table 2). Categories III and IV mutants were not characterized further.
Table 1

Single gene disruption mutants selected based on a failure or delay to exit lag phase when grown in synthetic medium supplemented with 25 mM furfural and the corresponding cellular function of the gene when present


Cellular function

Carbohydrate and metabolite metabolism



6-phosphogluconate dehydrogenase, generates NADPH (pentose phosphate pathway)


Ribulose-phosphate 3-epimerase (pentose phosphate pathway)


Glucose 6-phosphate 1-dehydrogenase, generates NADPH (pentose phosphate pathway)


Transketolase (pentose phosphate pathway)


Ribose phosphate pyrophosphokinase, pyrimidine and histidine biosynthesis, stress cell cycle regulator


C-5 sterol desaturase activity and ergosterol biosynthesis


Methionine biosynthesis


Purine biosynthesis, IMP cyclohydrolase, phosphoribosylaminoimidazole-carboxamide


Pyridoxine metabolism, cellular response to nutrient limitation and growth arrest


Trehalose-phosphate synthase activity, carbohydrate metabolism, response to stress


Uracil permease, uracil transport

Transcription, chromatin modification, and mRNA export



Transcription and mRNA export under stress conditions


Transcription factor, response to xenobiotic stimulus


RNA polymerase II transcription factor that regulates zinc tolerance genes during zinc stress


Chromatin modification, histone acetylation, transcription from pol II promoter


Adaptor protein in SAGA complex, chromatin modification, histone acetylation, transcription factor


Acetyltransferase activity, chromatin modification


Nicotinate phosphoribosyltransferase, NAD + regulation, chromatin silencing at rDNA and telomere


Subunit of NuA4 histone acetyltransferase complex with unknown role

Protein synthesis, modification, and degradation



Structural constituent of ribosome


Structural constituent of ribosome


Ubiquitin–protein ligase activity


Ubiquitin-dependent protein catabolism


Ubiquitin-specific protease


Peptide alpha-N-acetyltransferase activity, catalytic subunit of NatB N-terminal acetyltransferase


Protein phosphorylation, proteolysis, sporulation, response to stress


mRNA processing


Protein tyrosine phosphatase


Heat shock protein (yeast dnaJ homolog)


RNA methyltransferase

Replication and DNA damage repair



Required for survival upon exposure to K1 killer toxin; proposed to regulate double-strand break repair


Sister chromatid cohesion and replication


Ring finger protein involved in DNA damage response with possible role in recombination


RNA polymerase II degradation in response to DNA damage


Negative regulator of RNA pol III in response to changes to cellular environment

Vacuole, mitochondrion, and cytoskeleton function



Hydrogen-transporting ATPase activity, vacuolar acidification


Vacuolar ATPase


Chaperone activity and vacuolar acidification


Required for a functional vacuolar ATPase


Protein vacuolar targeting, Golgi-to-vacuole trafficking


Vacuolar ATPase


Vacuole inheritance and vacuolar protein sorting


Mitochondrial 3-oxoacyl reductase, fatty acid metabolism, aerobic respiration


Mitochondrial organization and biogenesis


Epsilon subunit of ATP synthase, hydrogen-transporting ATP synthase activity


Mitochondrial ATPase


Enoplasmic reticulum morphology and inheritance


Endocytosis, vacuoles are fragmented, molecular function unknown


Cytoskeletal regulator activity, mitochondrial inheritance, subunit of NatB N-terminal acetyltransferase


Putative role in microtubule function and mitochondrial distribution


Class I myosin, cell wall biogenesis, endocytosis, exocytosis, polar budding, response to osmotic stress

Bud site selection and cell division



Bud site selection


Bud site selection


Cytokinesis, protein binding, contractile ring

Uncharacterized genes


YDR049Wa, YLR218C, YKL056C, YFL043C, YKR070W, YLL007C, YDR333C, YJL055W

Identified genes are placed into seven functional groups

aMutants that failed to exit lag phase

bGene disruption verified by PCR amplification, kanR disruption cassette cosegregated with growth phenotype, and wild-type allele complements disruption strain
Fig. 1

ac Growth and percent growth inhibition of pentose phosphate pathway mutants in the presence of 0 or 25 mM furfural. Cell densities (absorbance at 600 nm) for parent BY4741, gnd1, zwf1, rpe1, tkl1, gnd2, tkl2, tal1, and ygr043c are shown for time points 0, 24, and 96 h in SD-complete with either 0 mM furfural (a) or 25 mM furfural (b). Percent growth inhibitions at 24 h (underlined) and 96 h (bold) are included above each respective strain. In (c), cell densities and percent growth inhibition for gnd1, zwf1, rpe1, and tkl1 cells containing low copy plasmids (pRS416) either without (light gray) or with (dark gray) corresponding wild-type genes (GND1, ZWF1, RPE1, or TKL1) are shown for the 24-h time point grown in SD-URA + 25 mM furfural. Optical density of BY4741 + pRS416 alone was 0.81 at 24 h (dashed line). Percent growth inhibitions are shown below each bar, with percentages representing mutants containing vector only underlined. Percent inhibition = 100−(A600 25 mM furfural cultures/A600 0 mM furfural cultures×100). Data points represent averages from two to four experiments

Table 2

Average lag timesa in hours (standard deviation) of parent BY4741 cells and pentose phosphate pathway mutants in the presence of no inhibitor, furfural, hydroxymethylfurfural (HMF), or ethanol (n=minimum of two replicates)


No inhibitor




5 mM

10 mM

20 mM

25 mM

35 mM

45 mM

40 mM


Wild type


0 (0)

0 (0)

0 (0)

2 (2)

0 (0)


0 (0)



1 (2)

4 (7)

12 (0)

6 (8)

0 (0)



6 (5)

20 (7)

2 (2.8)



4 (7)

16 (28)

24b (n/a)

0 (0)

0 (0)



1 (2)

5.3 (6)

30 (25)

144b (n/a)

0 (0)

0 (0)



0 (0)

0 (0)

0 (0)

6 (7)

2 (3)





0 (0)

0 (0)

6 (8)

15 (6)

14 (14)





0 (0)

0 (0)

6 (8)

12 (0)

0 (0)





0 (0)

0 (0)

2 (2)

8 (5)

2 (3)



aLag times were determined as the time point before cultures entered exponential growth taken from logarithmic scaled growth curves (minimum of two replicates)

bIn all but one experiment, cells failed to exit lag phase; since these numbers represent the one experiment that exited lag phase, standard deviations were not calculated

∞Cultures failed to exit lag in all experiments (i.e., cultures failed to have two doublings)

Nd Not determined

The 62 remaining categories I (8) and II (54) mutants were organized into seven groups based on the corresponding function of the disrupted gene as defined by the Saccharomyces genome database (1) carbohydrate or metabolite metabolism (11 genes), (2) transcription, chromatin modification, or mRNA export (8 genes), (3) protein synthesis, modification, or degradation (11 genes), (4) replication and DNA repair (5 genes), (5) vacuole, mitochondrion, and cytoskeleton functions (16 genes), (6) bud site selection and cytokinesis (3 genes), and (7) previously uncharacterized (8 genes) (Table 1). Amongst the first group are four PPP gene mutants, zwf1, gnd1, rpe1, and tkl1. In the oxidative branch of the PPP, ZWF1 and GND1 encode glucose-6-phosphate dehydrogenase (Zwf1p) and 6-phosphogluconate dehydrogenase (Gnd1p), respectively. Zwf1p and Gnd1p are responsible for producing most of the cytoplasmic NADPH by catalyzing the two irreversible reactions, glucose-6-phosphate → 6-phospho-gluconate → ribulose-5-phosphate, respectively. In the nonoxidative branch, RPE1 and TKL1 encode d-ribulose 5-phosphate 3-epimerase (Rpe1p) and transketolase-1 (Tkl1p), respectively, which are important in producing precursors needed for aromatic amino acid and nucleic acid biosynthesis (Flores et al. 2000). These PPP mutants were selected for further analysis for reasons described in the “Discussion.”

Growth characteristics in furfural of mutants lacking PPP genes

In addition to the four PPP disruption mutants identified (zwf1, gnd1, rpe1, and tkl1), four other PPP mutants [gnd2, tkl2, tal1, and ygr043c (hypothetical tal2)] were reexamined to determine if they exhibited a furfural-sensitive phenotype that was not detected in the original screen. Polymerase chain reaction (PCR) amplification of the disrupted regions for these eight mutants verified that the kanR disruption cassettes were positioned at the appropriate gene loci. Cultures of parent BY4741 cells and PPP mutants in the exponential phase were diluted into fresh SD-complete medium containing either 0 or 25 mM furfural, and their ability to grow (exit the lag phase and enter the exponential phase) was determined. Each culture started with the same cell concentration. Cell densities were monitored (absorbance at 600 nm) over a period of 196 h; however, after 96 h, no significant increase in growth was observed. In the growth without furfural, the absence of lag times (Table 2) confirmed that cultures were growing exponentially at the time of inoculation. Furthermore, in the absence of furfural similar doubling times (Table 3), cell densities at 24 and 96 h (Fig. 1a) are observed between parent and PPP mutant cultures, demonstrating that these disruptions do not affect control growth in the absence of furfural. However, differences were observed when furfural was present. Parent BY4741 cultures in 25 mM furfural exited the lag phase at 2 h with a doubling time of 5 h. In contrast, only a slight growth was observed in zwf1, gnd1, and rpe1 mutants, but they never reached exponential growth even after 192 h of incubation. In tkl1 mutants, a similar growth pattern was observed, although in one experiment, they were able to exit the lag phase, but not until 144 h, and even then, growth was poor, with a 21-h doubling time (Tables 2, 3, and Fig. 1b). Moreover, with 25 mM furfural present, the percent growth inhibition in cells lacking ZWF1, GND1, RPE1, or TKL1 at 24 h was between 74 and 86%, and at 96 h, the percent growth inhibition was between 72 and 90% compared to parent BY4741, which was 41 and 12% at 24- and 96-h time points, respectively (Fig. 1b, below bars). Conversely, the other PPP mutants (gnd2, tkl2, tal1, and ygr043c) in 25 mM furfural had only slightly longer lag times (6–15 h) (Table 2) and doubling times (6–11 h) (Table 3), but they were able to reach cell density levels similar to BY4741 by 96 h (Fig. 1b).
Table 3

Average doubling times in hoursa (standard deviation) of parent BY4741 cells and pentose phosphate pathway mutants in the presence of no inhibitor, furfural, hydroxymethylfurfural (HMF), or ethanol


No inhibitor




5 mM

10 mM

15 mM

20 mM

25 mM

30 mM

35 mM

45 mM

40 mM


Wild Type

1.9 (0.4)

2.0 (0.5)

3.6 (0.8)

4.3 (0.1)

6.3 (0.45)

5.2 (0.9)

7.1 (3.4)

9.6 (0.3)

7.5 (0.1)

10 (1.1)


2.1 (0.4)

6.8 (2.7)

7.1 (2.6)

12 (0.9)

28 (5.4)

19 (2.4)

8.4 (1.8)


2.6 (0.9)

10 (4.6)

16.4 (1.1)

14 (4.0)



2.3 (0.7)

4.4 (1.3)

8.4 (0.9)

19 (2.0)


29 (2.0)

8.6 (2.1)


2.4 (0.8)

3.7 (1.8)

8.9 (3.4)

13 (4.0)

9.3 (4.6)



21 (2.1)

7.8 (0.1)


1.9 (0.4)

1.7 (0.2)

3.4 (1.6)


8.1 (0.4)

7.1 (2.2)

22 (13)

25 (7.1)




2.2 (0.7)

2.0 (1.1)

3.3 (1.6)


9.6 (0.7)

9.45 (5.0)

26 (7.4)

28 (5.6)




1.8 (0.5)

1.8 (0.1)

3.4 (1.6)


7.0 (2.2)

11 (3.8)

15 (3.1)

24 (11)




1.7 (0.5)

2.4 (0.7)

4.2 (2.4)


9.8 (1.3)

6.3 (2.8)

12 (4.1)

15 (5.8)



aDoubling times were calculated from the exponential phase of two to eight logarithmic scaled growth curves where cultures doubled at least twice (minimum of two replicates)

bIn all but one experiment, cells failed to exit lag phase; since these number represents the one experiment that exited lag phase, standard deviations were not calculated

∞In all experiments, cells failed to exit lag phase

Nd Not determined

Interestingly, in the presence of another fermentation inhibitor furan, 5-HMF, zwf1, gnd1, rpe1, and tkl1 mutants had longer doubling times, and zwf1 and gnd1 mutants also had longer lag times compared to parent BY4741. However, in the presence of a nonfuran inhibitor, ethanol, observed in the PPP mutants was less of a growth phenotype compared to parent BY4741 (Tables 2 and 3).

To show that growth phenotypes were not due to secondary mutations, zwf1, gnd1, rpe1, and tkl1 mutants were crossed back to the parent strain BY4741, sporulated, and dissected to verify that disrupted genes (assayed by growth on G418 antibiotic) cosegregated with furfural growth inhibitory phenotypes. All complete tetrads demonstrated a 2:2 segregation of the antibiotic resistance with the furfural sensitivity phenotype, and at least 20 tetrads were dissected (data not shown). Moreover, wild-type alleles (ZWF1, GND1, RPE1, and TKL1) contained in low copy plasmids in S. cerevisiae (pRS416) were transformed into corresponding disruption mutants to show that growth in 25 mM furfural could be restored to levels similar to that observed for BY4741 (Fig. 1c).

Influence of various furfural concentrations on growth

The effects of various concentrations of furfural (0–45 mM) on lag times, doubling times, and percent growth inhibition were determined for parent BY4741 cells and zwf1, gnd1, rpe1, and tkl1 mutants. Parent BY4741 was able to grow at furfural concentrations between 0 and 35 mM, although at increasing concentrations, lag times (Table 2) and doubling times (Table 3) were longer, and percent growth inhibition (Fig. 2) all increased. Compared to BY4741, PPP mutants had considerably longer lag and doubling times and required less furfural to achieve 25, 50, and 75% growth inhibition (Fig. 2).
Fig. 2

Percent growth inhibition of parent BY4741, zwf1, gnd1, rpe1, and tkl1 in SD-complete with varying furfural concentrations (5, 10, 20, 25, 35, and 45 mM). Percent growth inhibition was calculated using cell densities at exponential phase. Percent inhibition=100−(x mM furfural A600 (postexponential growth)/0 mM furfural A600 (postexponential growth)×100). Data points represent average percent inhibition from two to seven experiments

PPP mutant's ability to metabolize furfural to furfuryl alcohol

Mutant strains zwf1, gnd1, rpe1, and tkl1 were tested for the ability to convert toxic furfural into the less toxic furfuryl alcohol. A time-course experiment was conducted using HPLC to measure the concentration of these two compounds in PPP mutants and parent BY4741 cultures containing 25 mM furfural. Wild-type cells metabolized all detectable furfural within 24 h (Fig. 2). Conversely, zwf1, gnd1, and rpe1 mutants were only able to metabolize 55, 48, and 87%, respectively, of the original furfural by 96 h (Fig. 3). Even after 192 h, furfural conversion was similar to that seen at 96 h (data not shown). In contrast, tkl1 mutants were able to metabolize all detectable furfural but not until 96 h (Fig. 3). While most of the metabolized furfural appeared to be converted to its alcohol form (Fig. 3, right bar), furfural can be transformed alternatively to furoic acid. Furoic acid was also measured, but very little was detected regardless of genotype or furfural concentration, suggesting that furoic acid had little to do with observed furfural growth inhibition phenotypes (data not shown).
Fig. 3

Conversion of furfural to furfuryl alcohol in PPP mutants. Furfural and furfuryl alcohol concentrations (mM) were determined by HPLC at set time points. Percent furfural removed=100−(furfural at 24 or 96 h) mM/(furfural at 0 h) mM×100 and percent furfuryl alcohol formed=[(furfuryl alcohol) mM/(furfural at 0 h) mM×100] at 96 h are shown. Data points represent averages from four experiments

ZWF1 overexpression increases growth in the presence of furfural

Parent BY4741 cells overexpressing PPP wild-type genes individually from the MET25 promoter were tested for improved growth in the presence of furfural. ZWF1 overexpression did not provide a growth advantage when 30 mM furfural was present. However, a modest growth advantage in the presence of 40 mM furfural was observed in cells overexpressing ZWF1. Concentrations of 50 mM furfural typically prevent growth; however, ZWF1 overexpression allowed cells to grow at this lethal concentration of furfural (Table 4). GND1, RPE1, and TKL1 overexpression did not provide an equal growth advantage.
Table 4

Average lag and doubling times and percent growth inhibition (standard deviations) in BY4741 cells, with ZWF1 overexpressed (pRS425-MET25promoter-ZWF1) compared with ZWF1 not overexpressed (pRS425-MET25promoter)


0 mM furfural

30 mM furfural

40 mM furfural

50 mM furfural









Lag timea

0 (0)

0 (0)

6.5 (1.5)

2.8 (2.9)

12 (0)

7.5 (0.58)

19 (18)

Doubling timeb

2.7 (0.56)

2.7 (0.27)

8.3 (1.5)

9.1 (1.1)

13 (4.7)

6.2 (1.6)

13 (3.7)

Growth inhibitionc (%)



44 (5.6)

44 (8.7)

62 (7.5)

52 (8.9)

85 (1.5)d

64 (4)

aLag times were determined as the time point before cultures entered exponential growth taken from logarithmic scaled growth curves

bDoubling times were calculated during exponential growth from logarithmic scaled growth curves

c100−(A600×mM furfural (average of postexponential time points)/A600 0 mM furfural (average of postexponential time points)×100)

dCultures that failed to grow, percent growth inhibition calculated using late time points where cell densities had increased slightly

∞Cultures that failed to have at least two doubling were characterized as failing to exit lag into exponential growth


In this work, we have demonstrated that furfural tolerance is a complex process, as revealed by the broad physiological roles of the 62 genes identified (Table 1), which represent about 1% of the entire S. cerevisiae genome. Speculating on the functional relationship of many of these genes to furfural tolerance was challenging. None of the genes appeared to have any direct correlation to pathways involved in furfural detoxification, although many had previously been connected to other stresses. We had anticipated identifying one or more alcohol dehydrogenase genes due to their likely role in the reduction of low concentrations of furfural (Modig et al. 2002; Palmqvist et al. 1999). Surprisingly, no ADH genes were identified. However, on closer examination of screening data, adh6 and adh1 mutants did grow a little slower than BY4741, but adh6 quickly recovered, and adh1 grew poorly even on medium lacking furfural. Mutants with such weak phenotypes were not selected due to the rules of our selection criteria, [(1) mutants have a clear growth defect, (2) mutants grow similarly as wild type in medium lacking furfural, and (3) growth phenotypes are reproducible]. Nevertheless, further evaluation of adh6 and adh1 might be justified, as adh6 is a NADPH-dependent alcohol dehydrogenase with specificity for aldehyde reduction (Larroy et al. 2002), and the adh1 mutant from the disruption library is not a correct isolate (Open Biosystems).

We chose to focus our efforts on the PPP disruption mutants (zwf1, gnd1, rpe1, and tkl1) partly because almost half of the PPP's genes were identified in the present screen. In addition, previous studies by other laboratories have shown a correlation between several PPP genes and specific stresses such as oxidative (Juhnke et al. 1996), sorbic acid (Mollapour et al. 2004), and osmotic (Krems et al. 1995). Moreover, the PPP is an important pathway in carbohydrate metabolism, producing a substantial fraction of the cytoplasmic NADPH required for anabolic pathways as well as precursors for the biosynthesis of aromatic amino acids and nucleic acids (Flores et al. 2000). The PPP has also been shown to be a necessary component for engineered S. cerevisiae strains capable of metabolizing xylose (Jeppsson et al. 2003; Karhumaa et al. 2005; Kuyper et al. 2005; Wahlbom et al. 2003; Zaldivar et al. 2002), a predominant sugar found in lignocellulosic biomass (Lynd 1996).

How ZWF1, GND1, RPE1, and TKL1 function in furfural and HMF tolerance is still unclear. They are not likely involved in a general stress tolerance function, as their absence did not influence growth in ethanol. A possible function can be related to PPP's role in NADPH production first catalyzed by ZWF1's gene product, glucose-6-phosphate dehydrogenase, or Zwf1p (Nogae and Johnston 1990). Supporting this is the observation that ZWF1 overexpression provided the strongest growth advantage compared to other PPP genes. This is perhaps due to Zwf1p being the first enzyme in the PPP, and it irreversibly converts glucose-6-phosphate to d-6-phospho-glucono-δ-lactone. Thus, overexpression of ZWF1 can preferentially commit glucose-6-phosphate to the PPP as opposed to other pathways (e.g., glycolysis or trehalose synthesis).

Nicotinamide adenine dinucleotide phosphate's importance in HMF detoxification is expected, as its reduction is catalyzed by an NADPH-dependent reaction (Flores et al. 2000; Wahlbom and Hahn-Hagerdal 2002). However, furfural reduction requires an NADH-dependent reaction (Palmqvist et al. 1999; Wahlbom and Hahn-Hagerdal 2002); thus, PPP genes would not initially be thought of as being directly involved in furfural tolerance. Three possible roles for the PPP during furfural stress are discussed below.

First, by decreasing cytoplasmic NADPH, there would be a decreased abundance of overall reducing equivalents. Thus, it seems likely that the available NADH would have to compensate for the lost reducing potential of the cell, making less NADH available for NADH-dependent reduction pathways, including furfural reduction and stress-induced glycerol production. In this potential scenario, the PPP would not be directly involved in detoxifying furfural but, more likely, would function indirectly by maintaining proper levels of reducing equivalents.

Second, the observed furfural-induced growth inhibition of PPP mutants could be due to a generalized growth deficiency even in the absence of furfural. This would result in a decreased flux in anabolic pathways, which would result in less NADH generation, and thus, furfural reduction would also be compromised. This possibility seems unlikely as, compared to BY4741, the PPP mutants in the absence of furfural had similar lag and doubling times and final cell densities (Tables 2, 3, and Fig. 1a). Moreover, we did not routinely isolate slow-growing mutants in our screen. Surprisingly, many slow-growing mutants grew comparatively well in the presence of furfural, suggesting that slow growth is not causing the observed furfural growth phenotype.

Finally, PPP's role in furfural tolerance might not be related to detoxifying furfural, but rather functions in protecting and repairing furfural-induced damage. Even in wild-type yeast cultures, before all the furfural is detoxified, there is a lag time (Taherzadeh et al. 2000a), and during this time, the cell must combat the possible effect of this and other furan aldehydes (Banerjee et al. 1981; Gupta et al. 1991; Janzowski et al. 2000; Modig et al. 2002; Taherzadeh et al. 1999, 2000b). Supporting this role are the findings linking PPP genes to various stresses (Jeppsson et al. 2003; Juhnke et al. 1996; Krems et al. 1995; Mollapour et al. 2004; Slekar et al. 1996), although its role, specifically in furfural and HMF stress, has previously not been reported. Moreover, there are many NADPH-dependent enzymes (e.g., glutathione reductase, thioredoxin reductase, and 3-oxoacyl reductase) that are responsible for cellular defense mechanisms against stresses (e.g., oxidative, osmotic, heat, and heavy metal stress) (Carmel-Harel and Storz 2000; Schneider et al. 1997). In this possible scenario, the PPP may function during times of stress by maintaining NADPH levels that stress enzymes require.

Interestingly, OAR1, which encodes 3-oxoacyl reductase, was amongst the genes identified in our screen. NADPH-dependent 3-oxoacyl reductase functions in fatty acid metabolism and has been linked to stress tolerance (Mollapour et al. 2004; Thorpe et al. 2004), and these stresses are known to damage cellular membranes (Swan and Watson 1999). Moreover, membranes in BY4741 cells such as mitochondrial membranes were severely compromised in the presence of furfural, which were repaired once furfural was detoxified (data not shown).

Other potential NADPH-reductase target genes could have been missed in the present screen for several reasons, including the presence of redundant genes, the stringent conditions of our screen, they may be essential genes and would not have been present in the original disruption library, or they may be among the eight uncharacterized genes (Table 1).

Understanding fermentation inhibitor tolerance in yeasts at the genetic level is of considerable importance to the fermentation industry. The screening strategy in the present study proved to be useful in identifying genes involved in furfural tolerance. Specifically, this work demonstrates a strong relationship between some PPP genes and furfural tolerance. Currently, genes identified in the present study and related genes are being analyzed to evaluate their impact on growth, inhibitor tolerance, ethanol yield, cellular physiology, and stress protection.


Special thanks to Drs. Nancy Alexander and Cletus Kurtzman for helpful discussions and suggestions and to Nicole Maroon, Stephanie Gove, Elizabeth Kilduski, Sarah Frazer, and Patricia O'Bryan for expert technical assistance.

Copyright information

© Springer-Verlag 2005