Membrane Fluidity of Saccharomyces cerevisiae from Huangjiu (Chinese Rice Wine) Is Variably Regulated by OLE1 To Offset the Disruptive Effect of Ethanol

Yeasts with superior ethanol tolerance are desirable for winemakers and wine industries. In our previous work, strain F23 was evolved with superior ethanol tolerance and fermentation activity to improve the flavor profiles of Chinese rice wine. Therefore, exploring the genomic variations and ethanol tolerance mechanism of strain F23 could contribute to an understanding of its effect on the flavor characteristics in the resulting Chinese rice wine. The cellular membrane plays a vital role in the ethanol tolerance of yeasts; however, how the membrane is regulated to fight the toxic effect of ethanol remains to be elucidated. This study suggests that the membrane fluidity is variably regulated by OLE1 to offset the disruptive effect of ethanol. Current work will help develop more ethanol-tolerant yeast strains for wine industries and contribute to a deep understanding of its high flavor-producing ability.

ABSTRACT

An evolution and resequencing strategy was used to research the genetic basis of Saccharomyces cerevisiae BR20 (with 18 vol% ethanol tolerance) and the evolved strain F23 (with 25 vol% ethanol tolerance). Whole-genome sequencing and RNA sequencing (RNA-seq) indicated that the enhanced ethanol tolerance under 10 vol% ethanol could be attributed to amino acid metabolism, whereas 18 vol% ethanol tolerance was due to fatty acid metabolism. Ultrastructural analysis indicated that F23 exhibited better membrane integrity than did BR20 under ethanol stress. At low concentrations (<5 vol%), the partition of ethanol into the membrane increased the membrane fluidity, which had little effect on cell growth. However, the toxic effects of medium and high ethanol concentrations (5 to 20 vol%) tended to decrease the membrane fluidity. Under high ethanol stress (>10 vol%), the highly tolerant strain was able to maintain a relatively constant fluidity by increasing the content of unsaturated fatty acid (UFA), whereas less-tolerant strains show a continuous decrease in fluidity and UFA content. OLE1, which was identified as the only gene with a differential single-nucleotide polymorphism (SNP) mutation site related to fatty acid metabolism, was significantly changed in response to ethanol. The role of OLE1 in membrane fluidity was positively validated in its overexpressed transformants. Therefore, OLE1 lowered the rate of decline in membrane fluidity and thus enabled the yeast to better fight the deleterious effects of ethanol.

IMPORTANCE Yeasts with superior ethanol tolerance are desirable for winemakers and wine industries. In our previous work, strain F23 was evolved with superior ethanol tolerance and fermentation activity to improve the flavor profiles of Chinese rice wine. Therefore, exploring the genomic variations and ethanol tolerance mechanism of strain F23 could contribute to an understanding of its effect on the flavor characteristics in the resulting Chinese rice wine. The cellular membrane plays a vital role in the ethanol tolerance of yeasts; however, how the membrane is regulated to fight the toxic effect of ethanol remains to be elucidated. This study suggests that the membrane fluidity is variably regulated by OLE1 to offset the disruptive effect of ethanol. Current work will help develop more ethanol-tolerant yeast strains for wine industries and contribute to a deep understanding of its high flavor-producing ability.

INTRODUCTION

Huangjiu, as a Chinese national unique and traditional alcoholic beverage with more than 4,000 years of history, is produced from rice by simultaneous saccharification with qū (a saccharification starter, which contains abundant fungal, bacteria, and enzymes, like glucoamylase) and fermentation with yeast (1). Huangjiu is generally fermented for 3 to 5 days at 28°C (primary fermentation), followed by 10 to 20 days at 10 to 20°C (secondary fermentation). The production and quality of huangjiu are closely related to the yeast used for fermentation. Simultaneous saccharification and fermentation during the brewing process avoid exposing the yeast cells to a high concentration of sugar and may contribute to ethanol yields as high as 14 to 20% (vol/vol) in the final fermentation mash (2). During secondary fermentation, however, the yeast cells are exposed to ethanol, which is toxic at high concentrations. Under these conditions, it is difficult to support yeast survival and further fermentation (3). Thus, a further increase in ethanol tolerance of the yeast would theoretically result in a more complete fermentation and higher quality of huangjiu (4). In our previous work, Saccharomyces cerevisiae F23 with 25 vol% ethanol tolerance was obtained, and it yielded a higher content of volatile flavor compounds than did yeast with poor ethanol tolerance during huangjiu making. Strain F23 was bred by using strategies of ethanol domestication, UV mutagenesis, and protoplast fusion; therefore, exploring the genetic basis of strain F23 would contribute to an understanding of the molecular mechanism of its high ethanol tolerance and its role in the flavor characteristics in the resulting huangjiu.

Despite many efforts having been made to explain the mechanism of ethanol tolerance, it remains poorly understood, and little overlap in results has been found. Some studies have suggested that the backgrounds of strains considerably affect ethanol tolerance (3, 5, 6). Generally, researchers have tended to use the laboratory type strain S. cerevisiae S288c in studies of ethanol tolerance, even though this strain has a relatively lower ethanol tolerance than do natural and industrial strains (7–9). Accordingly, little is known about the genomic backgrounds of industrial yeast strains with regard to ethanol tolerance. Therefore, the responses of natural or industrial yeast strains isolated from various fermented environments to ethanol stress should be investigated at the whole-genome level. The tolerance mechanisms of sake (which is similar to huangjiu) yeast have been studied for years (10–13); however, far fewer studies have investigated the yeast used in huangjiu production, let alone studies of the tolerance mechanisms of yeast generated from huangjiu. Knowledge about the ethanol tolerance of yeast obtained from huangjiu will be useful for a better understanding of the mechanisms of yeast tolerance.

Ethanol tolerance refers to the ability of yeasts to grow and survive in the presence of ethanol. It is a widely accepted that ethanol tolerance is a complex phenotype involving multiple layers of interaction and function (3, 14). Whole-genome analysis of yeast strains exposed to ethanol can facilitate a comprehensive identification of the key genes and cellular processes that contribute to ethanol tolerance. The development of multiple omics tools has greatly facilitated our understanding of genetic variations that arise in response to ethanol stress. Taking advantage of high-throughput sequencing technology, whole-genome sequencing and transcriptome analysis were performed with two yeast strains obtained from huangjiu (BR20 and F23), and the laboratory type strain S288c was used as the reference. In our previous study, strain BR20, which has an ethanol tolerance of 18 vol%, was naturally isolated from huangjiu. Strain F23, which has an ethanol tolerance of 25 vol%, was obtained by using strategies of adaptive evolution, UV mutagenesis, and protoplast fusion from the progenitor strains BR20 and BR30 (with high fermentation performance) (15). In the current study, the genomic variations between the two strains were investigated to identify the key genes and metabolic pathways responsible for the genomic evolution and enhanced ethanol tolerance of strain F23. As some studies suggested that the mechanism underlying ethanol tolerance was dependent on the ethanol concentration (16–18), transcriptome responses against various ethanol concentrations were investigated in the two strains. The phenotypic characteristics of these strains, including growth ability, ultrastructure, and cell membrane fluidity and integrity, were also evaluated. Differential variations related to cell membrane functions were determined from the whole-genome sequencing results. The overexpression of differential genes related to cell membrane function was then analyzed to validate the effects on ethanol tolerance.

RESULTS

Genome-wide identification of differential variations.

The whole genomes of strain F23 and its progenitor strain BR20 were sequenced on an Illumina HiSeq platform. The total raw sequencing read lengths of strains F23 and BR20 were 1,008.19 and 936.43 Mb, respectively. After quality control processing, 78.12% and 77.37% of the reads in strains F23 and BR20, respectively, had minimal average Phred scores of 30. The sequenced reads were mapped to the reference genome S288c, and the mappable rates of both strains exceeded 90% (Table 1).

TABLE 1.

Summary of sequencing reads in strains F23 and BR20

Sample	Raw data (Mb)	% reads with ≥Q30	GC content (%)	% mapped reads	No. of SNPs
F23	1,008.19	78.12	37.62	95.03	88,007
BR20	936.43	77.37	38.22	93.81	87,580

After mapping to the reference genome S288c, variant calling was then used to identify variations in single-nucleotide polymorphisms (SNPs) and indels between BR20 and F23. A slightly higher number of SNPs was discovered in F23 (88,007) than in BR20 (87,580). The differential SNPs between F23 and BR20 were identified using the ANNOVAR tool and classified into 1,769 exonic mutations and 2,245 intronic mutations. Notably, 39.45% of the exonic mutations caused amino acid changes. The differential indels between the two strains included 10 frameshift mutations, which may have significant impacts on the phenotypic differences between the two strains (see Table S2 in the supplemental material). The differential SNPs and indels between the two strains were widely distributed across the 16 yeast chromosomes and the mitochondrion genome and were particularly enriched in chromosomes II, IV, VII, XIV, and XVI (Fig. 1a). Additionally, a total of 2,540 genes were shared genes among strains F23, BR20, and S288c, while the numbers of species-specific genes in F23, BR20, and S288c were 2,634, 322, and 500, respectively. The species-specific genes contained genes missed completely in other species and gene sequences that were altered during evolution. Thus, the large number of species-specific genes in strain F23 might be responsible for its improved ethanol tolerance.

FIG 1.

Genomic variations between S. cerevisiae BR20 and F23 compared to reference strain S288c. (a) Distribution of differential SNPs and indels between the two strains across the genome. Genes related to differential variations in the chromosome were viewed in the outlet cycle. The second and third layers are the distribution of differential SNPs and indels, respectively. Genetic homology genes among BR20, F23, and S288c is illustrated as a Venn diagram in the center. (b) Metabolic pathways associated with the differential variations between S. cerevisiae BR20 and F23 in the KEGG map.

The differential variants between the two strains were subjected to iPath analysis to identify the associated metabolic pathways. At the genomic level, all differential SNPs and indels between strains F23 and BR20 were associated with to 210 proteins. These proteins were mainly involved in the amino acid metabolism, lipid metabolism, nucleotide metabolism, and carbohydrate metabolism pathways (Fig. 1b and Table S3).

Transcriptome responses to various levels of ethanol.

To further understand the effects of ethanol on intracellular processes, the transcriptome responses of strains F23 and BR20 were evaluated in the presence of 0, 10, and 18 vol% ethanol. Compared to BR20, 77.4% and 56.1% of the ethanol response gene transcripts were expressed at higher levels in F23 in response to 10 vol% and 18 vol% ethanol, respectively. Both strains exhibited higher gene expression in the presence of 18 vol% ethanol than in the presence of lower ethanol levels (Fig. 2a). Genes exhibiting significant differential expression between the two strains under various grades of ethanol stress were subjected to a KEGG enrichment analysis. The top 20 pathways are listed in Fig. 2. Under 10 vol% ethanol stress, differential gene transcriptions were mainly associated with amino acid metabolism, followed by carbohydrate metabolism and then fatty acid metabolism pathways (Fig. 2b). In the presence of 18 vol% ethanol, the differential gene transcriptions were mainly related to fatty acid metabolism and minorly to carbohydrate metabolism and amino acid metabolism pathways (Fig. 2c).

FIG 2.

Transcriptome analysis of S. cerevisiae BR20 and F23 exposed to 0, 10, and 18 vol% ethanol stress, respectively. (a) Violin plot represents the transcript levels in the two strains under 0, 10, and 18 vol% ethanol stress. (b) Top 20 metabolic pathways related to the genes with significantly (P < 0.05) different expression levels between the two strains in the presence of 10 vol% ethanol. (c) Top 20 metabolic pathways related to the genes with significantly (P < 0.05) different expression levels between the two strains in the presence of 18 vol% ethanol. TCA, tricarboxylic acid.

Membrane integrity and viability of strains exposed to various ethanol concentrations.

Figure 3 presents the membrane integrity and viability of strains F23 and BR20 under ethanol stress. The ultrastructures of yeast cells under various levels of ethanol stress were examined using transmission electron microscopy. The images showed that in a random field of view, the ratio of cells that lost membrane integrity increased with increasing concentrations of ethanol (not all data shown). Compared to strain BR20, a lower ratio of strain F23 cells lost membrane integrity in the presence of 18 vol% ethanol (Fig. 3a and A). In the fluorescein diacetate/propidium iodide (FDA/PI) staining experiment, cells with an intact membrane and those that lost membrane integrity were stained green and red, respectively. The relative membrane integrities of both strains exhibited significant declines after incubation with 10 or 18 vol% ethanol, although strain F23 exhibited better membrane integrity than did strain BR20, especially in the presence of 18 vol% ethanol (Fig. 3B to D and b to d). The two strains presented similar lag and log phases of growth in the absence of ethanol stress (Fig. 3e). Under incubation with 10 vol% ethanol, the lag phases of both strains were extended, although F23 experienced a shorter lag phase than did BR20 (Fig. 3f). Under 18 vol% ethanol stress, the growth of BR20 nearly stopped, whereas F23 continued to grow (Fig. 3g). In addition, F23 always exhibited a higher consumption rate of medium glucose than did BR20 under ethanol stress, especially under exposure to 18 vol% ethanol (Fig. S1). The growth curves and residual medium glucose were consistent with the FDA/PI staining results, which further confirmed that the highly ethanol-tolerant strain F23 exhibited greater membrane integrity than did strain BR20.

FIG 3.

Phenotypic results of S. cerevisiae BR20 and F23 subjected to 0, 10, and 18 vol% ethanol stress, respectively. (a and A) Ultrastructure of S. cerevisiae BR20 (a) and F23 (A) in the presence of 18 vol% ethanol is presented. (b to d) FDA/PI staining results of S. cerevisiae BR20 subjected to 0, 10, and 18 vol% ethanol. The pictures were viewed under a 20× lens objective. (B to D) FDA/PI staining results of S. cerevisiae F23 subjected to 0, 10, and 18 vol% ethanol. The red arrows show the impaired positions of the cell membrane. (e to g) Growth curves of S. cerevisiae BR20 and F23 grown in the presence of 0 vol% (e), 10 vol% (f), and 18 vol% (g) ethanol. (h) When the growth of a yeast cell was badly impaired by ethanol, membrane leakage would occur.

Slow alteration of membrane fluidity was induced in response to high ethanol stress.

The above-mentioned results suggest that fatty acid metabolism and membrane integrity were major contributors to high ethanol stress tolerance in yeast cells. Membrane fluidity, as a major factor in the maintenance of membrane integrity, was studied under ethanol stress conditions. After a further analysis of the variant calling data, OLE1 was found to be the only differential variant related to fatty acid metabolism between BR20 and F23 (Table S4). The same indel mutation of OLE1 was found in BR20 and F23, while F23 had one more SNP mutation site than did BR20 (Table S5), which resulted in the difference. OLE1, which encodes delta-9 desaturase in S. cerevisiae, converts saturated fatty acids to unsaturated fatty acids (UFA) (19), which are key components of the cellular membrane. Therefore, the total UFA contents under ethanol stress were analyzed. The ACC1 and HFA1 genes, which respectively encode the yeast cytoplasmic and mitochondrial acetyl-coenzyme A (acetyl-CoA) carboxylase (rate-limiting enzyme) in the biosynthesis of saturated fatty acids (20), were also investigated.

The fluorescence anisotropy (1/membrane fluidity) of 1,6-diphenyl-1,3,5-hexatriene (DPH)-labeled cells was measured to evaluate membrane fluidity. In all three strains, the anisotropy decreased under low ethanol (<4 vol%) stress and increased under medium ethanol (4 to 10 vol%) and high ethanol (>10 vol%) stress (Fig. 4a). Under low ethanol stress, strain S288c exhibited a sharp fluctuation in membrane fluidity, whereas only slight increases were observed in strains BR20 and F23. Under medium ethanol stress, the three strains exhibited similar declines in membrane fluidity. Under high ethanol stress, strain F23 exhibited minor fluctuations in membrane fluidity, whereas strains BR20 and S288c exhibited greater rates of decline. Because fatty acids, especially UFA, strongly affect membrane fluidity, the relative UFA contents were analyzed. In all three strains, the relative UFA content decreased as the ethanol stress increased. Although the three strains had similar UFA contents in the absence of ethanol, S288c and F23 had the lowest and highest UFA contents, respectively, in the presence of ethanol. In addition, the greatest rate of decline in the UFA content was observed in S288c, followed by in BR20 and F23 (Fig. 4b).

FIG 4.

(a to c) Anisotropy (a), UFA contents (b), and transcript levels of OLE1, ACC1, and HFA1 (c) in S. cerevisiae BR20 and F23 subjected to 0, 4, 8, 10, 14, 18, and 20 vol% ethanol. *, P < 0.05; **, P < 0.01.

In the presence of ethanol, strain S288c expressed much lower levels of OLE1 mRNA than did strains BR20 and F23 (Fig. 4c). This may be attributed to the haploid status of S288c, as the other two strains are diploid. The expression of OLE1 tended to decrease in S288c in response to increased ethanol concentrations. However, the expression of OLE1 was significantly upregulated in BR20 and F23 in response to low and medium ethanol stress and downregulated in response to high ethanol stress. F23 and BR20 expressed similar levels of OLE1 mRNA in response to 4 and 8 vol% ethanol stress, which had little effect on the growth of either strain. However, at ethanol stress levels exceeding 10 vol%, the OLE1 mRNA expression level was significantly higher in F23 than in BR20. Moreover, F23 exhibited superior growth compared to BR20 (Fig. 3f and g). Although both ACC1 and HFA1 encode acetyl-CoA carboxylase, the expression levels of these genes differed under the same level of ethanol stress, indicating that the two genes might compete for the same substrate. Under low ethanol stress, HFA1 was expressed more strongly than ACC1 in S288c and BR20, whereas ACC1 was more strongly expressed in F23. In response to medium and high ethanol stress, ACC1 and HFA1 were expressed at similar levels in S288c and BR20. These differences in the expression patterns of the two genes among the strains suggest that acetyl-CoA carboxylase formation depends on strain specificity.

Effect of OLE1 overexpression on membrane fluidity and ethanol tolerance.

To validate the effects of OLE1 on ethanol tolerance, a plasmid carrying OLE1 was used to transform strains S288c (poor ethanol tolerance) and F23 (high ethanol tolerance) to generate the overexpression transformants F23-pOLE1 and S288c-pOLE1, respectively. A colony PCR assay revealed the successful generation of the transformants (Fig. S2).

Under ethanol stress levels of <10 vol%, S288c-pOLE1 and F23 had similar relative UFA contents. However, the membrane fluidity decreased to a lesser extent in S288c-pOLE1 than in S288c, F23, and F23-pOLE1 (Fig. 5a and b). At higher levels of ethanol stress, both S288c and S288c-pOLE1 exhibited sharp decreases in the UFA content and membrane fluidity. However, OLE1 overexpression improved the response of S288c to ethanol stress, indicating the importance of this gene in the variability of membrane fluidity in response to ethanol stress. As shown in Fig. 5c, improved ethanol tolerance was observed in F23-pOLE1 and particularly in S288c-pOLE1. Overexpression of OLE1 significantly abrogated the rate of decline in membrane fluidity in S288c-pOLE1 than in S288c under medium ethanol stress. Moreover, F23-pOLE1 exhibited a significant difference in membrane fluidity under high ethanol stress (Fig. 5a). The relative UFA contents in the overexpressed transformants also tended to decrease as the level of ethanol stress increased. However, the relative UFA contents in both transformants were higher than those in the corresponding wild-type strains (Fig. 5b). Therefore, OLE1 overexpression might help slow the changes in membrane fluidity by increasing the content of UFA under high ethanol stress.

FIG 5.

Effects of OLE1 overexpression on different yeast strains. (a) Anisotropy of OLE1-ovexpressing transformants and the host strains. (b) Relative contents of UFA in OLE1-ovexpressing transformants and the host strains exposed to different ethanol stress. (c) Spot assay of OLE1-ovexpressing transformants and the host strains under 14 vol% ethanol. (d to f and D to F) Transcript levels of OLE1 (d and D), ACC1 (e and E), and HFA1 (f and F) in OLE1-ovexpressing transformants and the host strains. *, P < 0.05; **, P < 0.01. NS, nonsignificant.

As expected, significantly greater OLE1 expression was observed in F23-pOLE1 and S288c-pOLE1 than in F23 and S288c, respectively (P < 0.05). The upregulation of OLE1 in F23-pOLE1 and S288c-pOLE1 increased by up to 27.65-fold and 14.20-fold compared with the levels in F23 and S288c, respectively. In both transformants, OLE1 expression decreased continuously as the ethanol concentration increased (Fig. 5d and D). In F23, the HFA1 and ACC1 transcript levels were similar under high ethanol stress (Fig. 5e and f). This result differed from than those presented in Fig. 4c, and the use of yeast extract-peptone–d-galactose (YPG) instead of yeast extract-peptone-glucose (YPD) medium might be responsible for this difference. Compared with the host strains, the levels of HFA1 and ACC expression were significantly higher (P < 0.05) in S288c-pOLE1 but not F23-pOLE1 under 10 vol% ethanol stress (Fig. 5e, f, E, and F). In addition, OLE1 overexpression contributed to the significant increases (P < 0.01) in ACC1 and HFA1 expression in F23 under high ethanol stress. Here, the expression advantages of ACC1 and HFA1 in F23-pOLE1 differed from those in F23 (Fig. 4c and 5e and f). Interestingly, ACC1 and HFA1, as upstream genes of UFA synthesis, were strongly expressed under high ethanol stress in F23, in contrast to the changes of OLE1. This pattern may have occurred because the conversion speed of fatty acids (FA) to UFA decreased as the level of ethanol stress increased, especially in the range of high ethanol stress (≥14 vol%).

Correlation analysis between OLE1 expression and ethanol tolerance.

Fifteen yeast strains (Table S6) isolated from various fermented samples grown in various regions of China and preserved in our laboratory were used to analyze the relationship between OLE1 expression and growth ability under medium (10 vol%) ethanol stress. As shown in Fig. 6, a strong correlation was observed between OLE1 expression and growth. The Pearson’s correlation between the two variables was 0.8054 (P < 0.05).

FIG 6.

Correlation analysis of OLE1 overexpression and growth viability of 15 yeast strains isolated from different fermented samples. Details of the 15 yeast strains are listed in Table S6.

DISCUSSION

Ethanol is an industrially important chemical, and therefore, manufacturers and researchers are greatly interested in a comprehensive understanding of the genetic mechanisms underlying yeast ethanol tolerance. The evolution and resequencing strategy has become increasingly popular in studies of the genetic basis of novel phenotypes (21, 22). In this study, we focused on a natural S. cerevisiae strain (BR20) and a strain that evolved with enhanced ethanol tolerance (F23), with the aim of uncovering the genetic mechanism underlying enhanced ethanol tolerance via whole-genome sequencing and RNA sequencing (RNA-seq).

Whole-genome sequencing suggested that the enhanced ethanol tolerance of F23 might be attributable to differences in amino acid metabolism, lipid metabolism, carbohydrate metabolism, and nucleotide metabolism. Although yeast growth was dramatically impaired by ethanol, strains with different ethanol tolerances theoretically compensated for the toxicity associated with ethanol by altering the posttranscriptional levels of gene expression (3, 23). A further transcriptome analysis of the two strains in response to various levels of ethanol stress indicated that amino acid metabolism contributed most strongly to enhanced ethanol tolerance in yeast cultured under 10 vol% ethanol stress, whereas fatty acid metabolism was a major factor in the enhanced ethanol tolerance observed in the presence of 18 vol% ethanol. These differences in cellular processes in response to medium and high ethanol stress indicate that the ethanol concentration strongly affects the mechanism of ethanol tolerance, consistent with previous reports in which different genes contributed to different levels of tolerance under low and high ethanol stress (17, 18, 24). Additionally, the transcriptome results were consistent with the whole-genome sequencing results, which highlighted the importance of fatty acid metabolism and amino acid metabolism.

The cellular membrane, as the most sensitive cell structure to ethanol (23, 25, 26), was found with higher membrane integrity in F23 than in BR20 in response to medium and high levels of ethanol. Thus, the enhanced ethanol tolerance was attributed to the various defense mechanisms of the cellular membrane against ethanol stress. The cell membrane, which comprises a lipid bilayer, directly senses the extracellular environment (27). The molecular packing of fatty acids is considered a direct determinant of membrane fluidity, which is a key factor in maintaining the integrity of the membrane (28). The correlation of membrane fluidity with ethanol tolerance has been extensively documented, although some controversial findings have been reported. Some researchers suggested that yeast strains with higher UFA contents and membrane fluidity levels were more ethanol tolerant. In contrast, other researchers suggested an inverse correlation between membrane fluidity and ethanol tolerance. Our results suggested that membrane fluidity increased under low ethanol stress and were consistent with previous studies (19, 29–32). However, this fluidity decreased under medium and high ethanol stress, in contrast to the results reported by Alexandre et al. (29), You et al. (19), and Dinh et al. (33) but in agreement with those described by Steels et al. (34), Swan and Watson (35), Learmonth (36), and Ishmayana et al. (37). This difference might be partly ascribed to the use of different strains. Additionally, Ishmayana et al. (37) stated that differences in the methods used to detect membrane fluidity might have led to different results. Alexandre et al. (24, 29) and Kajiwara et al. (31) used the fluorescent probe DPH, whereas Learmonth (36) and Ishmayana et al. (37) used a generalized polarization of laurdan. In this study, DPH was used to evaluate the membrane fluidity. Alexandre et al. (24, 29) suggested that membrane fluidity increased consistently as the ethanol concentration was increased from 2 to 12 vol%. However, our results showed that the membrane fluidity varied, especially under high levels of ethanol stress (Fig. 4a and 5a). Learmonth et al. (36) also used DPH anisotropy to measure the membrane fluidity of S. cerevisiae and found that this parameter increased dramatically in response to 20 vol% ethanol, followed by varied recovery. That pattern was similar to our findings. In yeast strains previously grown in YPD medium containing ethanol, Alexandre et al. (24, 29) also observed varied membrane fluidity in response to ethanol stress levels ranging from 2 to 12 vol%. As yeast produces ethanol during normal cell growth, the varied fluidity observed by Alexandre et al. is likely more approximate to actual fermentation conditions.

Ethanol, a polar solvent, is known to fluidize cell membranes. According to the hypothesis of homeoviscous adaptation, microorganisms can adapt to survive under adverse environmental conditions (38). The membrane forms the primary architecture of the cell and thus plays important roles in energy exchange, transportation, and protection from toxins (39). Our study showed that yeast cells could alter their membrane fluidity to compensate for ethanol-induced fluidization (Fig. 4 and 5). Under various levels of ethanol stress, membrane fluidity was regulated promptly to offset the disruptive effects of ethanol. At low concentrations, the partition of ethanol into the membrane increased the membrane fluidity, which had little effect on cell growth. However, the toxic effects of medium and high ethanol concentrations tended to decrease the membrane fluidity. Therefore, we hypothesized that this decrease in membrane fluidity helped minimize the fluidizing effect of high levels of ethanol. Dramatic decreases in membrane fluidity were observed in strains S288c and BR20 in response to medium and high concentrations of ethanol. In contrast, the membrane fluidity of F23 remained relatively stable. The changes in the UFA contents of the three strains support this result (Fig. 4b and 5b). Strain F23 is more ethanol tolerant than its progenitor strain BR20 or S288c because it can regulate the cell membrane more effectively and thus lower the rate of decline in membrane fluidity. Accordingly, F23 maintains a more stable membrane environment for cellular functions.

Palmitoleic acid and oleic acid were the predominant components of the UFA, and trace amounts of polyunsaturated fatty acids were detected. Similar results were reported by Kajiwara et al. (31) and Wang et al. (32). Although the tested strains exhibited a trend of decreasing UFA contents in response to increased ethanol stress, strain F23 exhibited a higher UFA content and lower rate of UFA decrease than did the other strains. To understand the underlying genetic mechanism, the genome-wide variations related to fatty acid metabolism were explored. Accordingly, we identified OLE1, which converts FA to UFA by desaturation (40). It has been reported that OLE1 could confer resistance to cadmium, activate the mitogen-activated protein kinase (MAPK) high-osmolarity glycerol (HOG) pathway to enhance stress tolerance, and greatly contribute to Atg9 delivery and isolation membrane expansion in S. cerevisiae (40–42). These studies demonstrated the key role of OLE1 in the cell functions and stress tolerance of S. cerevisiae. Higher OLE1 expression was observed in F23 than in less-ethanol-tolerant strains in the presence of ethanol, and particularly under high ethanol stress (Fig. 4c). The total ACC1 and HFA1 transcript levels were also higher in F23 than in the other two strains. Hoja et al. (20) reported that ACC1 encodes the yeast cytoplasmic acetyl-CoA carboxylase, whereas HFA1 encodes a specific mitochondrial acetyl-CoA carboxylase that provides malonyl-CoA for intraorganellar fatty acids. The differential expression patterns of HFA1 and ACC1 under various levels of ethanol stress likely represent the transportation of fatty acids between the mitochondria and cytoplasm. The S. cerevisiae strains expressed high total levels of HFA1 and ACC1 together with high OLE1 expression (Fig. 4 and 5). Accordingly, we speculated that the genes encoding acetyl-CoA carboxylase contribute to OLE1 expression and UFA synthesis and thus lower the rate of decline in membrane fluidity, thus minimizing the fluidizing impact of ethanol toxicity (Fig. 7). Upon exposure to ethanol, the membrane fluidity in yeast cells is regulated to offset the disruptive effects of ethanol. Membrane fluidity increases in response to low ethanol (<5 vol%) stress; however, this low level of stress had little effect on cell growth. In the presence of medium ethanol stress (≤10 vol%), ethanol impaired the growth of yeast, and the cells decreased their membrane fluidity to minimize the fluidizing effect of ethanol. In the presence of high ethanol (>10 vol%), the increasing toxicity of ethanol considerably perturbed the yeast cell growth, and the membrane tended to maintain a relatively constant fluidity to better counteract ethanol toxicity. Under high ethanol stress, yeast strains with better ethanol tolerance expressed higher levels of OLE1 mRNA and lower rates of decline in membrane fluidity than did strains with poorer ethanol tolerance. Therefore, OLE1 regulates the rate of decline in membrane fluidity, which allows yeast to better combat the disruptive effects of ethanol.

FIG 7.

Potential mechanism depicting the regulation of cell membrane fluidity to cope with rising ethanol toxicity. When yeast cells are exposed to ethanol, membrane fluidity is regulated to offset the disruptive effect of ethanol. Membrane fluidity increases in response to low ethanol (<5 vol%) stress; however, the stress had little effect on cell growth. In the presence of medium ethanol stress (≤10 vol%), ethanol impaired the growth of yeast, and the cells decreased their membrane fluidity to minimize the fluidizing effect of ethanol. In the presence of high ethanol (>10 vol%), the increasing toxicity of ethanol considerably perturbed the cell growth of yeast, and the membrane tended to maintain relatively stable fluidity to better counteract ethanol toxicity. Under high ethanol stress, yeast strains with better ethanol tolerance expressed higher levels of OLE1 mRNA and lower rates of decline in membrane fluidity than did strains with poorer ethanol tolerance. Therefore, OLE1 regulates the rate of decline in membrane fluidity, which allows yeast to better combat the disruptive effects of ethanol.

In conclusion, our study contributes to a deeper understanding of ethanol tolerance. Our findings will encourage the development of more-ethanol-tolerant yeast strains for wine industries. Amino acid metabolism was the second ethanol-responsive pathway, and proteins are distributed in the plasma membrane; therefore, our ongoing work will focus on the genetic differences associated with these membrane proteins and the effects of interactions between these proteins and fatty acid metabolism on ethanol tolerance.

MATERIALS AND METHODS

Strains, media, and growth conditions.

S. cerevisiae S288c was kindly donated by Siliang Zhang (East China University of Science and Technology). S. cerevisiae diploid strain BR20, which has an ethanol tolerance of 18 vol%, was isolated from huangjiu and preserved in the China General Microbiological Culture Collection (accession number CGMCC 9445). Diploid strain F23, which can survive in 25 vol% ethanol, was preserved in the same depository (accession number CGMCC 12787) (15). Plasmid pYES6/CT, which contains a blasticidin resistance gene and the GAL1 promoter, was purchased from Addgene.

YPD (2% glucose, 2% peptone, 1% yeast extract) and YPG (2% d-galactose, 2% peptone, 1% yeast extract) media were used for yeast growth and gene overexpression experiments, respectively. Blasticidin (25 μg/ml) was added to the media to screen transformants. Propidium iodide (PI) and fluorescein diacetate (FDA) were prepared as 1 mg/ml solutions. All primers used in this study are listed in Table S1.

The ethanol concentrations were classified into the following three grades: low (<4 vol%), medium (4 to 10 vol%), and high (>10 vol%). Exponential-phase yeast cells (optical density at 600 nm [OD₆₀₀], 1.0) were harvested, washed, and inoculated into YPD/YPG medium containing 0, 4, 8, 10, 14, 18, or 20 vol% ethanol at a ratio of 10%. The inocula were then incubated in a shaker at 150 rpm and 28°C for 12 h.

Plasmid construction.

The open reading frame (ORF) of OLE1 was amplified (PES-F/R primer) from strain S288c and cloned into the plasmid pYES6/CT backbone (after double digestion by BamHI and XbaI) using a ClonExpress MultiS one-step cloning kit (Vazyme Biotechnology Ltd., China). The successfully cloned plasmid containing the OLE1 gene was named pOLE1. This plasmid was then transformed into competent F23 and S288c cells using the lithium acetate (LiAc)/single-stranded (SS) carrier DNA/polyethylene glycol (PEG) method (43) and grown on YPD agar plates supplemented with 50 μg/ml blasticidin. Successful transformants were validated by PCR (GAL1-F/CYC1-R primers) amplification and sequence analysis.

Whole-genome sequencing.

Strains F23 and BR20 were subjected to whole-genome sequencing. The BWA software was used to map the sequencing reads of BR20 and F23 to the reference genome S288c (https://www.ncbi.nlm.nih.gov/assembly/GCF_000146045.2#) separately. SAMtools was used to remove the duplicate reads produced by PCR. The GATK software was used to identify single-nucleotide polymorphisms (SNPs) and indel calls, and VCFtools was used to generate SNPs by applying filters with a minimal quality of 50 and minimal sequencing depth of 10 MB. The detected SNPs and indels were classified into the coding and intergenic regions according to their positions in the reference genome. SNPs in the coding sequence were annotated using the SnpEff tool.

RNA-seq analysis.

The RNA sequencing library was sequenced using an Illumina HiSeq X Ten system (2 × 150-bp read length). The processed reads of F23 and BR20 were separately aligned to the reference genome S288c with orientation mode using the Tophat software. The expression level of each transcript was calculated according to the fragments per kilobase of exons per million mapped reads (FRKM). After aligning to reference genome S288c, BR20 and F23 were compared to each other, and then the differential expression of genes among the samples was identified. The R statistical package software program EdgeR was used for the differential expression analysis.

Cell viability, membrane integrity, and membrane fluidity analyses.

The ability of yeast to grow and survive in the presence of ethanol was used to evaluate the ethanol tolerance of yeast. Growth ability and spot assays were used to evaluate cell viability under ethanol stress. Cell growth at 28°C was monitored for 48 h using a Bioscreen C MBR apparatus (Oy Growth Curves Ab Ltd.). A spot assay was performed by spotting 5 μl of a 10-fold serially diluted suspension onto YPD agar containing ethanol.

Membrane integrity was assessed using the FDA/PI staining technique (44). Briefly, the harvested yeast cells (OD₆₀₀, 0.2) were stained with 100 μg/ml PI and 50 μg/ml FDA and incubated in a dark shaker at 150 rpm and 28°C for 30 min. FDA- and PI-labeled yeast cells were then observed under a 20× lens objective and viewed using a DM2500 fluorescence microscope (Leica Microsystems) with blue and green lasers, respectively. Three fields of view were randomly chosen.

Plasma membrane fluidity was assessed by measuring the steady-state fluorescence anisotropy of 1,6-diphenyl-1,3,5-hexatriene (DPH) (24). The yeast cells were labeled with a 10 μM DPH solution (dissolved in tetrahydrofuran) and subsequently incubated in the dark for 30 min at 28°C and 150 rpm. The fluorescence intensity was measured using a SpectraMax M5e spectrofluorometer (Molecular Devices Ltd.) equipped with polarization filters. Background fluorescence was corrected using an unlabeled cell suspension. The results were expressed as anisotropy (r) values and calculated using the SoftMax Pro 6.5.1 software.

Analysis of fatty acids.

Fatty acids were extracted using the method described by Bligh and Dyer (45), with some modifications. Briefly, the harvested yeast cells were treated with 150 U lyticase (Sigma) for 1 h at 30°C. The fatty acids were then treated with 2% (vol/vol) sulfuric acid in methanol for 2 h at 90°C to derive the corresponding fatty acid methyl esters (FAMEs). Subsequently, the FAMEs were dissolved in hexane and analyzed using a gas chromatograph-mass spectrometry system (Trace 1300-ISQ; Thermo Fisher) equipped with a DB-WAX column (30 m by 0.25 mm by 0.25 μm; Agilent Technologies). The oven was held at 40°C for 3 min and then increased at a rate of 5°C/min to 230°C for 20 min. Helium, the carrier gas, was delivered at a flow rate of 1 ml/min.

Transmission electron microscopy.

The transmission electron microscopy methods described by Wright (46) and Schiavone et al. (47) were used in this study. Briefly, yeast cells were fixed for 12 h (i.e., overnight) in 2.5% glutaraldehyde, washed, and fixed in 1% osmium tetroxide for 1 h. Subsequently, the cells were washed, dehydrated in acetone, and finally embedded in LR White resin. A Leica Ultracut UCT ultramicrotome (Leica) was used to obtain ultrathin sections (70 to 80 nm). These sections were observed using a JEM-1400 electron microscope (JEOL Ltd.) at room temperature and an acceleration voltage of 80 kV.

RT-qPCR.

Total RNA was extracted according to the manufacturer’s instructions (Qiagen). Reverse transcription was performed using the PrimeScript reverse transcriptase (RT) reagent kit and genomic DNA (gDNA) eraser (TaKaRa). Quantitative reverse transcriptase PCR (RT-qPCR) was performed using a SYBR green PCR kit (TaKaRa) and a LightCycler real-time PCR system (Roche). The mRNA levels of each target gene were normalized to the ACT1 mRNA levels and were represented according to the 2^−ΔΔCT method (48). The quantification cycle (C_q) value of the target gene in yeast strains subjected to 0 vol% ethanol was set as the untreated control.

Statistical analysis.

The results were analyzed using a one-way analysis of variance, followed by Duncan’s test to determine significant differences. Significance (P) values of <0.05 and <0.01 were considered to indicate statistical significance.

Data availability.

All genomes and transcriptomes have been deposited in GenBank (BioProject accession numbers PRJNA509008 and PRJNA511370, respectively).