Copyright � 2010 by the Genetics Society of AmericaDOI: 10.1534/genetics.110.121871

Exploiting Natural Variation in Saccharomyces cerevisiae to Identify Genesfor Increased Ethanol Resistance

Jeffrey A. Lewis,*,† Isaac M. Elkon,*,† Mick A. McGee,†,‡ Alan J. Higbee†,‡

and Audrey P. Gasch*,†,§,1

*Laboratory of Genetics, †Great Lakes Bioenergy Research Center, ‡Biotechnology Center, and §Genome Center of Wisconsin,University of Wisconsin, Madison, Wisconsin 53706

Manuscript received August 5, 2010Accepted for publication September 18, 2010

ABSTRACT

Ethanol production from lignocellulosic biomass holds promise as an alternative fuel. However,industrial stresses, including ethanol stress, limit microbial fermentation and thus prevent costcompetitiveness with fossil fuels. To identify novel engineering targets for increased ethanol tolerance,we took advantage of natural diversity in wild Saccharomyces cerevisiae strains. We previously showed that anS288c-derived lab strain cannot acquire higher ethanol tolerance after a mild ethanol pretreatment, whichis distinct from other stresses. Here, we measured acquired ethanol tolerance in a large panel of wildstrains and show that most strains can acquire higher tolerance after pretreatment. We exploited thismajor phenotypic difference to address the mechanism of acquired ethanol tolerance, by comparing theglobal gene expression response to 5% ethanol in S288c and two wild strains. Hundreds of genes showedvariation in ethanol-dependent gene expression across strains. Computational analysis identified severaltranscription factor modules and known coregulated genes as differentially expressed, implicating geneticvariation in the ethanol signaling pathway. We used this information to identify genes required foracquisition of ethanol tolerance in wild strains, including new genes and processes not previously linkedto ethanol tolerance, and four genes that increase ethanol tolerance when overexpressed. Our approachshows that comparative genomics across natural isolates can quickly identify genes for industrialengineering while expanding our understanding of natural diversity.

CELLULOSIC materials are an attractive source forbiofuel production, given the availability of

agricultural residues that do not directly compete withfood sources (Solomon 2010). However, fermentationof cellulosic biomass is problematic. Stressful by-products generated during preprocessing, coupledwith the unique composition of pentose and hexosesugars, limit microbial ethanol production. Significantattention is therefore being dedicated toward engi-neering stress-tolerance microbes for cellulosic fer-mentation.

Saccharomyces cerevisiae has been the organism ofchoice for ethanol production, because of its inherentethanol tolerance. However, high ethanol levels can stillinhibit viability and fermentation, and engineeringgreater ethanol resistance has led to improved bioetha-nol production (Alper et al. 2006). Ethanol affects manycellular processes, including membrane fluidity, pro-tein stability, and energy status (reviewed recently inStanley et al. 2010). Recent genetic screens have

implicated additional genes important for ethanoltolerance, including those involved in vacuolar, perox-isomal, and vesicular transport, mitochondrial function,protein sorting, and aromatic amino acid metabolism(Kubota et al. 2004; Fujita et al. 2006; Van Voorst et al.2006; Teixeira et al. 2009; Yoshikawa et al. 2009). Yetdespite the attention to the mechanism of ethanoltolerance, significant gaps in our knowledge remain.

Several studies have also investigated the global geneexpression response to ethanol (Alexandre et al. 2001;Chandler et al. 2004; Fujita et al. 2004; Hirasawa et al.2007). However, mutational analysis shows that mostgenes upregulated by ethanol are not required forethanol tolerance (Yoshikawa et al. 2009). Thus, geneexpression responses in a single strain are poor pre-dictors of genes important for tolerance of the initialstressor. Instead, we have argued that the role of stress-dependent gene expression changes is not to survive theinitial stress, but rather to protect cells against impend-ing stress in a phenomenon known as acquired stressresistance (Berry and Gasch 2008). When cells arepretreated with a mild stress, they often acquire toler-ance to what would otherwise be a lethal dose of thesame or other stresses. Consistently, the gene expressionresponse triggered by a single stress treatment has no

Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.110.121871/DC1.

1Corresponding author: Department of Genetics, 425 Henry Mall,University of Wisconsin, Madison, WI 53706.E-mail: agasch@wisc.edu

Genetics 186: 1197–1205 (December 2010)

impact on surviving the initial stress, but instead iscritical for the increased resistance to subsequent stress(Berry and Gasch 2008). However, it remains true thatrelatively few of the expression changes are importantfor subsequent tolerance of a particular stress. Thus,identifying the important genes remains a challenge.

Our understanding of the physiological and tran-scriptional response to ethanol has been further nar-rowed since most studies focus on laboratory-derivedstrains. While ethanol tolerance and adaptation havebeen explored in sake, wine, and industrial yeast strains(Rossignol et al. 2003; Wu et al. 2006), we have onlyrecently begun to appreciate the physiological diversityof natural yeast isolates. Wild yeast isolates from diverseenvironments have widely varying phenotypes undervarious conditions, and many of these phenotypes maybe related to variation in gene expression (Cavalieri

et al. 2000; Fay et al. 2004; Kvitek et al. 2008). Here weexploited strain-specific differences in the physiologicaland transcriptional response to ethanol. We comparedstrains with and without the ability to acquire increasedethanol tolerance after ethanol pretreatment, thenidentified corresponding gene expression differencesacross strains. This rapidly revealed genes that wereinvolved in acquired ethanol tolerance and identifiedseveral new genes that increase ethanol tolerance whenoverexpressed. By applying systems biology approachesto the analysis of phenotypic diversity, we have gener-ated a new understanding of the transcriptional re-sponse to ethanol and have identified novel genesinvolved in its tolerance.

MATERIALS AND METHODS

Strains, culture media, and growth conditions: Strains usedare listed in supporting information, File S1. All chemicalswere purchased from Sigma (St. Louis, MO). Gene deletionswere created by homologous recombination that replaced thegene-coding sequence with KanMX3 drug resistance cassettes.The HO gene was replaced with the HygMX3 cassette togenerate a haploid YPS163 upon dissection, and this was usedas the background in all YPS163 strain knockouts. The haploidstrain behaved similarly to the diploid strain in all ethanolresistance assays (compare Figure 1D [diploid] and Figure S3[haploid]). All mutations were confirmed by diagnostic PCR.

Ethanol resistance assays: Acquired ethanol resistance wasassayed as in Berry and Gasch (2008). Briefly, cultures weregrown in YPD (1% yeast extract, 2% peptone, 2% glucose) forat least eight generations to an optical density (OD600) of 0.3.Each culture was split into two cultures and received either asingle dose of 5% (v/v) ethanol or 5% water as a mock control.Mock-treated cells were thereafter handled identically. Cellswere exposed to a panel of severe ethanol doses (ranging from5 to 25% v/v depending on the experiment) in YPD for 2 hr in96-well plates. A 50-fold dilution of each culture was spottedonto YPD agar plates and grown for 48 hr, after which viabilityat each dose was scored by visual inspection using a four-pointscale to score 100%, 50–100%, 10–50%, or 0% survivalcompared with the no stress (YPD) control. An overall ethanoltolerance score was calculated as the sum of scores over 11doses of stress.

Cycloheximide experiments were performed as above,except that 10 mg/ml cycloheximide was added to the culture20 min before and throughout the ethanol pretreatment. Amock-treated culture received inhibitor treatment but noprimary stress. Long-term ethanol tolerance was scored byplating cells on YPD 1 8% (v/v) ethanol. Growth was scoredafter 3 days (or 2 days in the case of controls).

To measure the effects of gene overexpression, BY4741 cellsharboring galactose-inducible, GST-tagged constructs (OpenBiosystems, Huntsville, AL; Sopko et al. 2006) were grownovernight on SC -Ura containing 2% dextrose, and thensubcultured for at least 8 generations in SC -Ura containing2% galactose to induce overexpression before exposure toethanol as described above. All overexpression strains werecompared to the isogenic BY4741 containing the vector onlycontrol (pEGH). Ethanol tolerance was scored using both thespot assay described above and flow cytometry. For flowcytometry, viability was assayed using the LIVE/DEAD Funga-Light yeast viability kit (Invitrogen, Carlsbad, CA) on a GuavaEasyCyte flow cytometer (Millipore, Billerica, MA) accordingto both manufacturers’ instructions. Briefly, mock and etha-nol-treated cells were diluted 10-fold into 10 mm HEPES–NaOH (pH 7.2 at 25�) supplemented with 2% dextrose andthe viability dye reagents (SYTO 9 and propidium iodide). Theproportion of prodium iodide negative cells was reported aspercentage of viable cells.

Array hybridization and analysis: Cells were grown over-night for at least eight generations to an OD600 of 0.3–0.6. Asample of cells was collected (time 0), and ethanol was addedto a final concentration of 5% (v/v). Cells were collected at 15,30, 45, and 60 min post-ethanol addition. A single biologicalreplicate was collected for each strain during the time course.For detailed analysis of the 30-min time point, biologicaltriplicates were collected using a paired experimental design.Cell collection, RNA isolation, and microarray labeling wereperformed as described (Gasch 2002), using cyanine dyes(Flownamics, Madison, WI), Superscript III (Invitrogen,Carlsbad, CA), and amino-allyl-dUTP (Ambion, Austin, TX).Microarrays were spotted in house using 70mer oligonucleo-tides representing each of the yeast ORFs (Qiagen, Chats-worth, CA). We previously showed that ,5% of measuredexpression differences could be affected by hybridizationdefects due to polymorphism (Kvitek et al. 2008). Arrays werescanned using a scanning laser (GenePix 4000B) fromMolecular Devices (Sunnyvale, CA). Inverse dye labeling wasused in replicates to control for dye-specific effects. Data werefiltered (retaining unflagged spots with R2 . 0.1) andnormalized by regional mean centering (Lyne et al. 2003).Genes with significant expression differences in response toethanol were identified separately for each strain by perform-ing a t-test using the BioConductor package Limma v. 2.9.8(Smyth 2004) and FDR correction (Storey and Tibshirani

2003) (see File S3 for the Limma output). Expression differ-ences in YPS163 or M22 relative to S288c, both with andwithout ethanol treatment, were identified in a similarmanner. Gene clustering was done in Cluster 3.0 (http://bonsai.ims/u-tokyo.ac.jp/�mdehoon/software) using hierar-chical clustering and uncentered Pearson correlation as themetric (Eisen et al. 1998). Arrays were weighted using a cutoffvalue of 0.4 and an exponent value of 1. Enrichment of geneontology (GO) functional categories was performed usingGO-TermFinder (http://go.princeton.edu/cgi-bin/GOTerm-Finder) hosted by the Lewis–Sigler Institute for IntegrativeGenomics (Boyle et al. 2004), with Bonferroni-correctedP-values , 0.01 taken as significant. Clusters were analyzedfor enrichment of known transcription factor ChIP-Chiptargets (Harbison et al. 2004) using Fisher’s exact test andby upstream motif identification using multiple em for motif

1198 J. A. Lewis et al.

elicitation (MEME; Bailey and Elkan 1994). MEME param-eters were model, tcm; minimum width, 6; maximum width,12; maximum number of motifs, 3. All microarray data areavailable through the NIH Gene Expression Omnibus (GEO)database under accession number GSE22904.

Lipidomic gas chromatography mass spectrometry: Cellswere grown in synthetic complete (SC; Sherman 2002)medium for at least 8 generations to an OD600 of 0.3–0.6.Acquired ethanol tolerance was similar in SC vs. YPD (datanot shown). Two technical replicate samples were collectedfor each biological sample; biological triplicates were col-lected. Cells were collected immediately prior to the additionof 5% (v/v) ethanol (time 0) and at 60 min after the ethanoladdition. For the collections, 2 ml of cells were added directlyto 200 ml concentrated HCl (final concentration of 1.1 m)and heated to 95 � for 1 hr. Total lipids were then extracted bythe method of Bligh and Dyer (1959). Preparation of fattyacid methyl esters (FAMEs) was performed using the methodof Christie(1989). FAMEs were analyzed by gas chromatog-raphy mass spectrometry (GC–MS) using a Pegasus 4DGCxGC-TOF gas chromatograph–mass spectrometer (LecoCorp. St. Joseph, MI) fitted with a Rx1-5MS column (30-m,0.25-mm ID, 0.25u df; Restek, Inc., Bellefonte, PA). In-strument parameters were: He carrier gas flow rate: 1 ml/min; split ratio: 5:1; injector temperature: 250�, GC oven: 50�for 1 min initially, increased at 20�/min to 330�, and held at330� for 5 min.

RESULTS

Natural variation in acquired ethanol tolerance indiverse yeast strains: We previously showed that anS288c-derived lab strain, pretreated with individual mildstressors, can acquire increased tolerance to either thesame or different stresses (Berry and Gasch 2008).However, ethanol was the only pretreatment that did notincrease resistance to subsequent stresses, includingethanol itself (Figure 1A and Berry and Gasch 2008).This raised the question of whether ethanol was uniqueas a stressor, or whether the S288c laboratory strain wasanomalous. To test this, we performed acquired ethanoltolerance assays on 47 diverse strains from vineyards,oak exudate, sake and wine fermentations, clinicalsettings, and other natural environments. Cells wereexposed to 5% ethanol for 60 min, then exposed to apanel of 11 high doses of ethanol (Figure 1 andmaterials and methods). Intriguingly, most (but notall) strains tested could acquire further ethanol toler-ance after mild pretreatment (Figure 1B). The majorprogenitor strain of S288c, EM93 (Mortimer andJohnston 1986), showed some acquisition of ethanoltolerance after a pretreatment (Figure 1B and File S1),suggesting that S288c lost this ability relatively recently.We subsequently focused on two wild strains—oak-soilstrain YPS163 and the vineyard strain M22—to probethe physiology of acquired ethanol resistance.

Variation in the genomic expression response toethanol: Acquired resistance to several stresses requiresnascent protein synthesis during the mild-stress pre-treatment (Berry and Gasch 2008). Consistently, wefound that acquired ethanol resistance in wild strains

also requires protein synthesis during pretreatment(Figure 1, C and D). We therefore suspected thatS288c may have an altered genomic expression responseto ethanol. We used whole-genome DNA microarrays tomeasure the gene-expression response of S288c, YPS163,and M22 responding to 5% ethanol over a 60-min timecourse (Figure 2A). To identify statistically significantdifferences between strains, we performed biologicaltriplicates before and at 30 min after ethanol treatment,which encompassed the peak response.

As expected, ethanol induced a dramatic remodelingof the yeast transcriptome. Overhalf of the genome (3941genes, false discovery rate, FDR, of 0.01) was significantlyaffected by ethanol in any of the three strains, with similarkinetics (Figure S1). Genes induced more than threefoldwere enriched for certain functional categories, includingvacuolar catabolic processes, response to temperaturestimulus, glucose metabolism, alcohol catabolism, andmetabolism of energy reserves including glycogen andtrehalose (Bonferroni-corrected P , 0.01 in all cases).The genes significantly repressed more than threefoldby ethanol were strongly enriched for ribosome bio-genesis and protein synthesis. Together, these resultsare consistent with activation of the yeast environmentalstress response (Gasch et al. 2000) and largely agreewith the previous literature (Alexandre et al. 2001;Chandler et al. 2004; Fujita et al. 2004; Hirasawa et al.2007).

We next identified genes with larger ethanol-responsiveinduction in wild strains compared to S288c, reasoningthat they may account for the phenotypic difference inacquired ethanol tolerance. We therefore identified ex-pression differences between each wild strain compareddirectly to S288c (FDR , 0.05). There were 1555 genes(25%) and 1662 genes (27%) differentially expressed inresponse to ethanol in M22 and YPS163, respectively,compared to S288c—875 of these genes were commonto both comparisons (Figure 2B). In contrast, the twowild strains compared to each other showed differentialethanol response at only 735 genes, revealing a large frac-tion of S288c-specific differences. A fraction (38–45%)of the 875 ethanol-responsive genes that distinguishS288c from the wild strains also showed underlying dif-ferences in basal gene expression (393/875 in M22 com-pared toS288cand329/875 in YPS163vs.S288c).However,there was little overlap in functional groups enriched ingenes with basal expression differences compared to geneswith variation in ethanol response (File S2). Together, thisindicates significant variation in the gene expressionresponse to ethanol.

Network analysis implicates transcription factorsunderlying expression differences: To identify patternsin the data set, we hierarchically clustered �2300 geneswith ethanol-dependent expression differences in ei-ther strain compared to S288c (FDR , 0.05, Figure 2A).We systematically scored enrichment of GO functionalcategories for each cluster (File S2). Several gene

Natural Variation in Yeast Ethanol Response 1199

clusters with higher induction in both wild strains wereenriched for functional categories, including vacuolarprotein catabolism, trehalose biosynthesis, response tooxidative stress, alcohol metabolism, and proteolysis(cluster J), and transposition (clusters L and M). Severalgene clusters actually showed higher induction inS288c, such as oxidative phosphorylation and cellularrespiration (cluster H) and protein folding (cluster I).These may represent processes that are more stronglyaffected by ethanol in the S288c background.

The results of the clustering analysis suggestedupstream differences in physiology and/or ethanolsignaling that affected many genes in trans. We soughtto implicate transcription factors required for a robustethanol response and to examine whether variability intranscription factor function was responsible for S288c’sinability to mount a proper response to ethanol.

We first ruled out a known polymorphism in the S288cHAP1 gene (Gaisne et al. 1999), which encodes atranscription factor involved in heme and oxygensensing (Figure S2 and Figure S3). Clustering analysisand transcription factor-target and motif enrichment(see materials and methods) implicated three addi-tional transcriptional regulators: Rpn4, which regulates

proteasome genes, the oxidative-stress transcriptionfactor Yap1, and the stress-activated factor Msn2. Thetargets of Rpn4 and Yap1 showed weaker induction inS288c compared to both wild strains, indicating varia-tion in their responsiveness to ethanol. However, nei-ther Rpn4 nor Yap1 had an effect on acquired ethanoltolerance, as mutants lacking either gene acquiredethanol resistance at wild-type levels (Figure S4).

In contrast, deletion of msn2 in YPS163 impaired bothacquired ethanol resistance and gene expression. TheYPS163 msn2D mutant showed reduced acquisition ofethanol tolerance after pretreatment but no differencein basal ethanol tolerance (Figures 3A and 4A). Tran-scriptional profiling of the YPS163 msn2D mutantresponding to 5% ethanol identified 244 genes withattenuated gene induction (FDR , 0.01; File S3),confirming involvement of Msn2 in the ethanol re-sponse. Of the 239 Msn2-regulated genes, 106 (44%, p¼4x10�9, Fisher’s exact test) also had significantly lowerinduction in S288c responding to ethanol compared toYPS163 (Figure 3B). This suggests that Msn2 activation byethanol is partially defective in S288c (see discussion),and implicates one or more Msn2 targets as likely directeffectors of acquired ethanol tolerance.

Figure 1.—Acquired eth-anol tolerance in diverseyeast strains. (A) A repre-sentative acquired ethanoltolerance assay is shown.S288c (left) or YPS163(right) was exposed to 5%ethanol or mock pretreat-ment for 60 min. Cells wereexposed to one of seven in-dicated severe doses of eth-anol for 2 hr and thenplated onto a YPD plate toscore viability. (B) Basal(red) and acquired (blue)percentage ethanol toler-ated is shown for strains col-lected from diverse niches(clin, clinical; ferm, fermen-tations). The maximal dosesurvived was based on.50% spot density com-pared to the no-ethanolcontrol. Data represent theaverage of biological dupli-cates. Strains and scoresare found in File S1. (C)Cycloheximide blocks ac-quired ethanol tolerancein both YPS163 and M22.Error bars represent stan-dard deviation of biologicaltriplicates. Cycloheximide(CHX), if present, wasadded 20 min prior to ei-ther the mock or ethanol

(5%) primary stress. (D) Results of a representative spot assay from the experiment shown in C. The ethanol doses were 12.5–17.5% (v/v) in 0.5% increments.

1200 J. A. Lewis et al.

Identifying mutants with defects in acquired ethanolresistance: To identify additional effectors of acquiredethanol resistance, we generated deletion mutants of 20manually chosen genes, implicated by their reducedinduction in S288c (see materials and methods).Strikingly, over 50% of the genes interrogated (Table 1)were required for normal acquisition of ethanol toler-ance, indicating that our method produced a ‘‘hit rate’’significantly higher than that of other studies (3–6%)(Kubota et al. 2004; Fujita et al. 2006; Van Voorst et al.2006; Teixeira et al. 2009; Yoshikawa et al. 2009). Weidentified eight genes (in addition to MSN2) that werenecessary for acquired ethanol resistance (ELO1, SLA1,AIP1, TPS1, EDE1, GPB2, PEP4, and OAC1; Figure 4A).These genes participate in a variety of cellular functionsincluding RAS signaling (GPB2), cytoskeleton/endocy-tosis (SLA1, AIP1, EDE1), vacuolar protein degradation(PEP4), mitochondrial transport (OAC1), fatty acidlipid elongation (ELO1), and trehalose biosynthesis(TPS1). Several of these genes (ELO1, EDE1, PEP4,and OAC1) had never been linked to ethanol resistancebefore and showed no discernable ethanol sensitivity inprior ethanol screens (see discussion).

To further characterize the behavior of the mutantstrains, we analyzed long-term growth on agar platescontaining ethanol (Figure 4B). Three of the strains(YPS163 gpb2D, pep4D, and tps1D mutants) were unableto grow in the presence of ethanol, and two additional

strains had weak defects (sla1D, sod2D). However, theremaining mutants (amounting to 50%) had no dis-cernable defect in surviving a single dose of ethanol(Figure 4B). Thus, these genes play a specific role inacquired ethanol tolerance, highlighting that the mech-anism is overlapping with but distinct from the mech-anism of basal ethanol tolerance.

Figure 2.—Variation in gene expres-sion between S288c, M22, and YPS163.(A) Log2 ethanol-responsive expres-sion changes of 2203 genes differen-tially expressed in either wild-strain vs.S288c (FDR is 0.05, paired t-test). Basalexpression differences in M22 (M) orYPS163 (Y) vs. S288c (S) are shownon the left; time courses of the expres-sion changes in response to ethanol areshown in the middle; and difference be-tween ethanol response in each wildstrain vs. S288c is shown on the right.Each row represents a gene and eachcolumn represents a strain or condi-tion, with time-course samples indi-cated by triangles. Genes wereorganized by hierarchical clustering ofthe combined basal expression andtime-course data. Differences in etha-nol response across strains were subse-quently added to the figure. Redindicates induced and green indicatesrepressed expression in response toethanol. Blue indicates higher and yel-

low indicates lower expression in S288c relative to the wild strains. Complete GO categories enriched in each cluster are foundin File S2. The top GO categories for each cluster listed below were chosen on the basis of lowest P-value for clearly nonoverlappingfunctional groups: (A) RNA localization, (B) translation, (C) regulation of translation, (D) cellular amine metabolism, (E) ox-idative phosphorylation, (F) flocculation, (G) methionine biosynthesis, (H) oxidative phosphorylation, cell death, (I) proteinfolding, ( J) vacuolar protein catabolism, trehalose metabolism, proteolysis, (K) catabolic process, (L) transposition, (M) trans-position, (N) polyphosphate metabolism. (B) Venn diagram of differentially expressed genes across all possible pairwise straincomparisons. This comparison includes genes with significant differential expression in M22 vs. YPS163, which were omitted fromthe clustering analysis.

Figure 3.—Acquired ethanol tolerance depends on Msn2 .(A) Acquired tolerance defect of an msn2D strain. Cells werepretreated with 5% (v/v) ethanol for 1 hr and then subjectedto severe ethanol doses (x-axis) for 2 hr. Colony-forming unitsindicated percentage viability. Error bars represent standarddeviation of biological triplicates. (B) Average log2 expressionchange of 106 Msn2-dependent genes with significantly lowerexpression in S288c vs. YPS163, in S288c (S), M22 (M), YPS163(Y), and YPS163 (Y) msn2D strains responding to ethanol.

Natural Variation in Yeast Ethanol Response 1201

We hypothesized that genes involved in acquiredethanol resistance are potential engineering targetsfor improving ethanol resistance. We used galactose-inducible overexpression constructs in S288c to test forincreased basal ethanol tolerance. Growth in galactoseincreased basal ethanol tolerance in even the controlstrain. However, overexpression of MSN2 in S288cfurther increased ethanol resistance beyond that ofthe control strain, as was expected (Watanabe et al.2009). In fact, half of the other overexpressing con-structs tested (TPS1, EDE1, and ELO1) significantlyincreased basal ethanol resistance compared to thecontrol (Table 1 and Figure 4, C and D). This indicatesthat effectors of acquired ethanol resistance are pro-ductive targets for directed engineering.

Strain-specific differences in lipid composition: Therequirement for fatty acid elongase I (Elo1) raised thepossibility that S288c may not properly remodel itsplasma membrane in response to the fluidizing effectsof ethanol. We therefore performed GC–MS analysis of

the total membrane fatty acids from S288c, YPS163, andthe YPS163 elo1D strain, in either the presence or theabsence of 5% ethanol.

In response to ethanol, YPS163 increased the pro-portion of oleic acid (18:1) in the membrane, with acommensurate decrease in palmitic acid (16:0) (Figure5). Indeed, higher levels of oleic acid are known tocorrelate with higher ethanol tolerance (You et al.2003). The membrane lipid profile of S288c contrastedwith YPS163, since basal levels of palmitic acid werehigher while oleic acid was lower in S288c. Uponethanol treatment, S288c was able to increase its oleicacid content but not to levels seen in YPS163 (Figure 5).Thus, the difference in lipid content in S288c correlateswith its inability to acquire ethanol resistance after amild pretreatment.

Given that the YPS163 strain lacking ELO1 had adefect in acquired ethanol tolerance, we expected itwould have lower levels of long-chain fatty acids, andspecifically oleic acid (C18:1). Starting levels of 14:0

Figure 4.—Genes necessary for acquired ethanol resistance. (A) Basal and acquired ethanol tolerance is shown for variousmutant strains. The average and standard deviation of ethanol tolerance scores (see materials and methods) is shown for strainspretreated with 5% ethanol (blue) versus the mock-treated control (orange). Error bars represent standard deviation of biologicaltriplicates. Asterisks denote significant differences in acquired ethanol resistance relative to the YPS163 wild-type strain (*, P ,0.05; ** ¼ P , 0.01, t-test). (B) Basal ethanol sensitivity in mutants with defects in acquired ethanol resistance. Cells were grown atleast eight generations to an OD600 of 0.3–0.6, after which cells were normalized to an OD600 of 0.15 and 10-fold serial dilutionswere plated onto either a YPD (control) plate or YPD 1 8% (v/v) ethanol (YPDE). Growth on the YPD plate was scored at 2 days,while growth on the YPDE plate was scored at 3 days. (C) A representative experiment showing strain basal tolerances to 2 hrexposure of indicated ethanol doses in S288c containing the indicated galactose-inducible plasmid constructs (see materials

and methods). Data are as shown in Figure 1A. (D) As in C except viability was scored after a 2-hr exposure to 19% ethanolusing flow cytometry to determine the proportion of propidium iodide negative (i.e., live) cells. Error bars represent standarddeviation of biological triplicates. Asterisks denote significant differences in percentage viability relative to S288c carrying thevector only control (**, P , 0.01, t-test).

1202 J. A. Lewis et al.

were slightly higher than the YPS163 parent, similar tothe S288c strain (Figure 5). However, following ethanoltreatment the membrane lipid profile of the YPS163elo1D strain did not differ measurably from wild-typeYPS163. The effect of Elo1 on lipid profiles may beobscured by technical limitations of our study, since wewere unable to observe subclasses of these lipids.Nonetheless, this result shows that the effect of Elo1 ismore complicated than simply elongating fatty acidchains (see discussion).

DISCUSSION

By taking advantage of the phenotypic diversity inyeast ethanol responses, we have identified new genesand processes required for ethanol tolerance whileproviding a glimpse into natural variation in S. cerevisiae.A key feature of our approach is to compare andcontrast strains with unique phenotypes, to quicklyimplicate the genetic basis of the phenotypic difference.We identified hundreds of gene expression differencesin S288c that correlate with its inability to acquireethanol tolerance after a mild pretreatment. Strikingly,over 50% of the genes interrogated in our pilot knock-out study were important for the phenomenon ofacquired ethanol resistance, a far higher proportionthan that of library screens for basal ethanol resistancemutants (3–6%) (Kubota et al. 2004; Fujita et al. 2006;Van Voorst et al. 2006; Teixeira et al. 2009; Yoshikawa

et al. 2009). This dramatic improvement in identifyingrelevant genes resulted not only from the cross-straincomparison, but also from assaying the phenotype mostdependent on gene expression changes—that of ac-quired, rather than basal, stress resistance. Our resultshighlight the promise of this approach in rapidly

identifying new targets for biofuels engineering, partic-ularly for engineering increased microbial tolerance torelevant stresses.

We also generated new insights into the mechanism ofethanol defense by identifying both downstream andupstream effectors. We identified several downstreamprocesses as important for acquired ethanol tolerance,including trehalose metabolism, vacuolar function,actin cytoskeleton, endocytosis, transport, and fatty acidelongation/metabolism. While some of these were pre-viously implicated in ethanol tolerance, their preciseroles in acquired ethanol tolerance are not entirely clear.However, many of these processes are likely to affect thecellular membrane. We identified several genes thataffect endocytosis, which may be important in remodel-ing the membrane following ethanol exposure, and isknown to be inhibited by ethanol (Meaden et al. 1999).Vacuolar processes, including protein degradation byPep4, may help to turn over internalized membrane andother proteins. We confirmed trehalose as an importantplayer, since the YPS163 tps1D strain had the strongestdefect of all the mutants tested, and overexpression ofTps1 conferred higher ethanol resistance. Intriguingly,trehalose can also protect endocytosis from the inhibi-tory effects of ethanol (Lucero et al. 2000), in additionto its traditional role in protein and membrane stabili-zation (Singer and Lindquist 1998).

Our lipidomic analysis suggests a complicated role formembrane fatty acid metabolism in ethanol tolerance.While fatty acid desaturation has been previouslyimplicated in ethanol resistance (You et al. 2003), weidentified a role for fatty acid elongase in the acclima-tion to ethanol. ELO1 was induced to higher levels inwild strains responding to ethanol and was required for

TABLE 1

Genes screened for ethanol resistance phenotypesin a YPS163 background

Acquisitiondefect

Nodefect

Increased tolerancein S288c

No increasedtolerance in S288c

aip1D aim2D EDE1 AIP1ede1D ctt1D ELO1 GPB2elo1D gas2D MSN2 OAC1gpb2D hap1D TPS1 PEP4msn2D hsp12D SFA1oac1D icy1D SOD2pep4D ino1D

sla1D rpn4D

tps1D sfa1D

sod2D

yap1D

ycf1D

Underlined genes denote strains with a basal growth defecton YPD 1 8% (v/v) ethanol plates. With the exception of thetps1D strain, none of the other underlined strains had a short-term basal defect.

Figure 5.—Differences in membrane lipid composition inresponse to ethanol in S288c and YPS163. GC–MS analysis ofthe total membrane lipids in response to ethanol in S288c,YPS163, or the YPS163 (Y) elo1D mutant. The x-axis representslipid chain length and level of saturation. Error bars representstandard error of biological triplicates. Asterisks denote signif-icantly different comparisons between YPS163 and S288c(*, P , 0.05; **, P , 0.01; ***, P ,0.001, paired t-test).

Natural Variation in Yeast Ethanol Response 1203

normal acquisition of ethanol resistance. That ELO1 wasnot required for basal ethanol tolerance is consistentwith a specific role in acquired ethanol tolerance.Despite the implication of Elo1 in this response, theYPS163 elo1D strain did not show a discernable defect inoleic acid accumulation (C18:1) compared to its wild-type parent. This may be due to Elo1’s effect on lipidsubclasses, which we were unable to measure here.Indeed, variations in lipid polar head groups or mem-brane sterol composition has been shown to affectethanol tolerance (Walker-Caprioglio et al. 1990;Sajbidor et al. 1995). Future detailed studies of Elo1will increase our understanding of the role of fatty acidelongation in the acclimation to ethanol stress.

We have also identified upstream regulators of theadaptive response to ethanol, including Msn2. Theincreased ethanol tolerance afforded by Msn2 over-expression confirms its importance in ethanol resis-tance. However, the msn2D strain had a relatively milddefect in acquired ethanol resistance, suggesting thatother regulators play a role. Interestingly, another genewe identified—GPB2—is a negative regulator of the RASpathway, which suppresses Msn2 activity (Lu andHirsch 2005). Notably, the gpb2D mutation produceda stronger defect in both basal and acquired ethanolresistance compared to the msn2D mutation, suggestingthat Gpb2 and/or PKA signaling play an additional,Msn2-independent role in ethanol resistance.

Our results have also reflected upon the diversity in S.cerevisiae strains, in terms of stress tolerance, geneexpression, and membrane lipid content. Several mod-ules of transcription-factor targets, including genesregulated by Rpn4, Yap1, and Msn2, show variableresponses across strains. Most of these genes show nodifference in expression in the absence of stress and arerevealed only as variably affected upon environmentalshift. Although future studies will be required to dissectthe precise genetic basis for these expression differ-ences, this work demonstrates the substantial variationin or above regulatory networks that coordinate envi-ronmental responses. This, in turn, further underscoresthe importance of considering multiple strain back-grounds to identify the mechanisms of stress resistance.

Author contributions: J.A.L. and A.P.G designed research; J.A.L.and I.M.E. performed experiments; M.A.M. and A.J.H. performedlipidomic analysis; J.A.L. and A.P.G analyzed data; and J.A.L. andA.P.G wrote the article. This work was funded by the Department ofEnergy (DOE) Great Lakes Bioenergy Research Center (DOE Officeof Biological and Environmental Research Office of ScienceDE-FC02-07ER64494).

LITERATURE CITED

Alexandre, H., V. Ansanay-Galeote, S. Dequin and B. Blondin,2001 Global gene expression during short-term ethanol stressin Saccharomyces cerevisiae. FEBS Lett. 498: 98–103.

Alper, H., J. Moxley, E. Nevoigt, G. R. Fink and G. Stephanopoulos,2006 Engineering yeast transcription machinery for improvedethanol tolerance and production. Science 314: 1565–1568.

Bailey, T. L., and C. Elkan, 1994 Fitting a mixture model by expec-tation maximization to discover motifs in biopolymers. Proc. Int.Conf. Intell. Syst. Mol. Biol. 2: 28–36.

Berry, D. B., and A. P. Gasch, 2008 Stress-activated genomic ex-pression changes serve a preparative role for impending stressin yeast. Mol. Biol. Cell 19: 4580–4587.

Bligh, E. G., and W. J. Dyer, 1959 A rapid method of total lipid ex-traction and purification. Can. J. Biochem. Physiol. 37: 911–917.

Boyle, E. I., S. Weng, J. Gollub, H. Jin, D. Botstein et al.,2004 GO:TermFinder: open source software for accessing GeneOntology information and finding significantly enriched GeneOntology terms associated with a list of genes. Bioinformatics20: 3710–3715.

Cavalieri, D., J. P. Townsend and D. L. Hartl, 2000 Manifoldanomalies in gene expression in a vineyard isolate of Saccharomy-ces cerevisiae revealed by DNA microarray analysis. Proc. Natl.Acad. Sci. USA 97: 12369–12374.

Chandler, M., G. A. Stanley, P. Rodgers and P. Chambers, 2004 Agenomic approach to defining the ethanol stress response in theyeast Saccharomyces cerevisiae. Ann. Microbiol. 54: 427–454.

Christie, W. W., 1989 Gas Chromatography and Lipids: A PracticalGuide, pp. 36–38. Oily Press, Bridgwater, England.

Eisen, M. B., P. T. Spellman, P. O. Brown and D. Botstein,1998 Cluster analysis and display of genome-wide expressionpatterns. Proc. Natl. Acad. Sci. USA 95: 14863–14868.

Fay, J. C., H. L. McCullough, P. D. Sniegowski and M. B. Eisen,2004 Population genetic variation in gene expression is associ-ated with phenotypic variation in Saccharomyces cerevisiae. GenomeBiol. 5: R26.

Fujita, K., A. Matsuyama, Y. Kobayashi and H. Iwahashi,2004 Comprehensive gene expression analysis of the responseto straight-chain alcohols in Saccharomyces cerevisiae using cDNAmicroarray. J. Appl. Microbiol. 97: 57–67.

Fujita, K., A. Matsuyama, Y. Kobayashi and H. Iwahashi,2006 The genome-wide screening of yeast deletion mutantsto identify the genes required for tolerance to ethanol and otheralcohols. FEMS Yeast Res. 6: 744–750.

Gaisne, M., A. M. Becam, J. Verdiere and C. J. Herbert, 1999 A’natural’ mutation in Saccharomyces cerevisiae strains derived fromS288c affects the complex regulatory gene HAP1 (CYP1). Curr.Genet. 36: 195–200.

Gasch, A. P., 2002 Yeast genomic expression studies using DNAmicroarrays. Methods Enzymol. 350: 393–414.

Gasch, A. P., P. T. Spellman, C. M. Kao, O. Carmel-Harel, M. B. Eisen

et al., 2000 Genomic expression programs in the response of yeastcells to environmental changes. Mol. Biol. Cell 11: 4241–4257.

Harbison, C. T., D. B. Gordon, T. I. Lee, N. J. Rinaldi, K. D.Macisaac et al., 2004 Transcriptional regulatory code of a eu-karyotic genome. Nature 431: 99–104.

Hirasawa, T., K. Yoshikawa, Y. Nakakura, K. Nagahisa, C.Furusawa et al., 2007 Identification of target genes conferringethanol stress tolerance to Saccharomyces cerevisiae based on DNAmicroarray data analysis. J. Biotechnol. 131: 34–44.

Kubota, S., I. Takeo, K. Kume, M. Kanai, A. Shitamukai et al.,2004 Effect of ethanol on cell growth of budding yeast: genesthat are important for cell growth in the presence of ethanol.Biosci. Biotechnol. Biochem. 68: 968–972.

Kvitek, D. J., J. L. Will and A. P. Gasch, 2008 Variations in stresssensitivity and genomic expression in diverse S. cerevisiae isolates.PLoS Genet. 4: e1000223.

Lu, A., and J. P. Hirsch, 2005 Cyclic AMP-independent regulationof protein kinase A substrate phosphorylation by Kelch repeatproteins. Eukaryot. Cell 4: 1794–1800.

Lucero, P., E. Penalver, E. Moreno and R. Lagunas,2000 Internal trehalose protects endocytosis from inhibitionby ethanol in Saccharomyces cerevisiae. Appl. Environ. Microbiol.66: 4456–4461.

Lyne, R., G. Burns, J. Mata, C. J. Penkett, G. Rustici et al.,2003 Whole-genome microarrays of fission yeast: characteris-tics, accuracy, reproducibility, and processing of array data.BMC Genomics 4: 27.

Meaden, P. G., N. Arneborg, L. U. Guldfeldt, H. Siegumfeldt andM. Jakobsen, 1999 Endocytosis and vacuolar morphology inSaccharomyces cerevisiae are altered in response to ethanol stressor heat shock. Yeast 15: 1211–1222.

1204 J. A. Lewis et al.

Mortimer, R. K., and J. R. Johnston, 1986 Genealogy of principalstrains of the yeast genetic stock center. Genetics 113: 35–43.

Rossignol, T., L. Dulau, A. Julien and B. Blondin, 2003 Genome-wide monitoring of wine yeast gene expression during alcoholicfermentation. Yeast 20: 1369–1385.

Sajbidor, J., Z. Ciesarova and D. Smogrovicova, 1995 Influenceof ethanol on the lipid content and fatty acid composition of Sac-charomyces cerevisiae. Folia Microbiol. 40: 508–510.

Sherman, F., 2002 Getting started with yeast. Methods Enzymol.350: 3–41.

Singer, M. A., and S. Lindquist, 1998 Multiple effects of trehaloseon protein folding in vitro and in vivo. Mol. Cell 1: 639–648.

Smyth, G. K., 2004 Linear models and empirical bayes methods forassessing differential expression in microarray experiments. Stat.Appl. Genet. Mol. Biol. 3: 1–26.

Solomon, B. D., 2010 Biofuels and sustainability. Ann. N. Y. Acad.Sci. 1185: 119–134.

Sopko, R., D. Huang, N. Preston, G. Chua, B. Papp et al.,2006 Mapping pathways and phenotypes by systematic geneoverexpression. Mol. Cell 21: 319–330.

Stanley, D., A. Bandara, S. Fraser, P. J. Chambers and G. A. Stanley,2010 The ethanol stress response and ethanol tolerance of Sac-charomyces cerevisiae. J. Appl. Microbiol. 109: 13–24.

Storey, J. D., and R. Tibshirani, 2003 Statistical significancefor genomewide studies. Proc. Natl. Acad. Sci. USA 100:9440–9445.

Teixeira, M. C., L. R. Raposo, N. P. Mira, A. B. Lourenco andI. Sa-Correia, 2009 Genome-wide identification of Saccharomyces

cerevisiae genes required for maximal tolerance to ethanol. Appl.Environ. Microbiol. 75: 5761–5772.

van Voorst, F., J. Houghton-Larsen, L. Jonson, M. C. Kielland-Brandt and A. Brandt, 2006 Genome-wide identification ofgenes required for growth of Saccharomyces cerevisiae under etha-nol stress. Yeast 23: 351–359.

Walker-Caprioglio, H. M., W. M. Casey and L. W. Parks,1990 Saccharomyces cerevisiae membrane sterol modifications inresponse to growth in the presence of ethanol. Appl. Environ.Microbiol. 56: 2853–2857.

Watanabe, M., D. Watanabe, T. Akao and H. Shimoi,2009 Overexpression of MSN2 in a sake yeast strain promotesethanol tolerance and increases ethanol production in sake brew-ing. J. Biosci. Bioeng. 107: 516–518.

Wu, H., X. Zheng, Y. Araki, H. Sahara, H. Takagi et al.,2006 Global gene expression analysis of yeast cells during sakebrewing. Appl. Environ. Microbiol. 72: 7353–7358.

Yoshikawa, K., T. Tanaka, C. Furusawa, K. Nagahisa, T. Hirasawa

et al., 2009 Comprehensive phenotypic analysis for identifica-tion of genes affecting growth under ethanol stress in Saccharomy-ces cerevisiae. FEMS Yeast Res. 9: 32–44.

You, K. M., C. L. Rosenfield and D. C. Knipple, 2003 Ethanol tol-erance in the yeast Saccharomyces cerevisiae is dependent on cellu-lar oleic acid content. Appl. Environ. Microbiol. 69: 1499–1503.

Communicating editor: F. Winston

Natural Variation in Yeast Ethanol Response 1205

GENETICSSupporting Information

http://www.genetics.org/cgi/content/full/genetics.110.121871/DC1

Exploiting Natural Variation in Saccharomyces cerevisiae to Identify Genesfor Increased Ethanol Resistance

Jeffrey A. Lewis, Isaac M. Elkon, Mick A. McGee, Alan J. Higbeeand Audrey P. Gasch

Copyright � 2010 by the Genetics Society of AmericaDOI: 10.1534/genetics.110.121871

J. A. Lewis et al. 2 SI

FILES S1-S3

Files S1-S3 are available for download as Excel files at http://www.genetics.org/cgi/content/full/genetics.110.121871/DC1.

File S1: Strains Used in this Study

File S2: Functional Enrichment Analysis of Gene Clusters

File S3: Normalized Microarray Data and Statistical Analyses

J. A. Lewis et al. 3 SI

FIGURE S1.—Gene expression kinetics of the ethanol response in S288c, M22, and YPS163. The average log2 expression ratio

for each time point was calculated using only the genes that were induced or repressed by > 2-fold.

Ave

rage

log2

Exp

ress

ion

Cha

nge

S288cM22

YPS163

Time (min)15 30 45 60

0

1.0

-1.0

-2.0

2.0

Figure S1

J. A. Lewis et al. 4 SI

FIGURE S2.—Hierarchical clustering of gene expression differences in a YPS163 hap1 strain. The diagrams show the average

log2 expression differences in the denoted strains, as shown in Figure 2. Genes were organized independently for panels shown on

the left and on the right by hierarchical clustering. A red color indicates genes induced by ethanol and a green color indicates

genes repressed by ethanol. A blue color indicates genes with higher expression in S288c or the hap1 strain relative to YPS163,

and a yellow color indicates higher expression in the YPS163 relative to S288c or the hap1 strain. GO categories enriched in

each cluster are found in Dataset S2. The left panel denotes 2304 genes with either basal gene expression differences in either

wild strain compared to S288c (FDR = 0.05, t-test) or significant differences in gene expression in the YPS163 hap1 strain

compared to YPS163 (FDR = 0.05, paired t-test). The results show that deletion of hap1 only accounts for a small fraction of basal

gene expression differences between wild strains and S288c, the latter of which contains a known polymorphism that reduces

Hap1 function. The right panel denotes 2590 genes with either differences in ethanol-induced gene expression in either wild

strain versus S288c (FDR = 0.05, paired t-test), or significant differences in gene expression in the YPS163 hap1 strain compared

to YPS163 (FDR = 0.05, paired t-test). Again, Hap1 accounts for a small number of ethanol-dependent gene expression

differences between the wild strains and S288c.

S288

c

YPS1

63

Y v.

Y hap1Δ

Y hap1Δ

Y v.

S

M/S

Y/S

Y v.

Y hap1Δ

Basal (time 0)

Differences in EtOH Responses

EtOH (30 min)Basal (time 0)

M/S

Y/S

Figure S2

B.

A.

D.

H.

F.E.

G.

C.

I.

J.

K.

M22 M v

. S

B.A.

D.

H.

F.

E.

G.

C.

I.J.

K.

L.

M.

>3-foldinduced

>3-foldrepressed

>3-foldhigher in

wild strain

>3-foldlower in

wild strain

J. A. Lewis et al. 5 SI

FIGURE S3.—Hap1 is not required for acquired ethanol tolerance. The left panel depicts results of a spot assay for acquired

ethanol tolerance as described in Figure 1 and Methods. Error bars represent standard deviation of biological triplicates. The

right panel shows a representative spot assay from the experiment.

hap1Δ

Treatment

0

0

mock

Treatment

YPS163

mock

EtOH

EtOH

[EtOH] (12.5 - 17.5% v/v)

[EtOH] (12.5 - 17.5% v/v)hap1ΔYPS163

5

15

25

mock EtOH

Figure S3E

than

ol T

oler

ance

Sco

re

J. A. Lewis et al. 6 SI

FIGURE S4.—Rpn4 and Yap1 are not required for acquired ethanol tolerance. Results of a spot assay for acquired ethanol

tolerance as described in Figure 1 and Methods. Error bars represent standard deviation of biological duplicates.

rpn4ΔYPS163

10

20

0

mock pretreatment EtOH pretreatment

Figure S4

Eth

anol

Tol

eran

ce S

core

30 yap1Δ