Eukaryotic Adaptation to Years-Long Starvation Resembles that of Bacteria.

The Growth Advantage in Stationary Phase (GASP) phenomenon, described in bacteria, reflects the genetic adaptation of bacteria to stress, including starvation, for a long time. Unlike in stationary phase where no cell division occurs, GASP harbors active cell division, concurrent with genetic adaptation. Here we show that GASP occurs also in eukaryotes. Two strains of Saccharomyces cerevisiae (Sc404 and Sc424) have been isolated from 2-year-old sealed bottles of beer. These strains presented advantage in survival and growth over the parent during stress. The differences between the strains are irreversible and therefore genetic in origin rather than epigenetic. Direct competition assays show that Sc404 and Sc424 outcompete the parent in direct competition. DNA sequencing shows changes of the genome: the TOR complexes are mutated, and DNA repair gene mutations confer a mutator phenotype. The differences between the strains are reflected in a difference in taste between beers brewed from them.

Introduction

Cells respond to their environment with surprising efficiency by using various mechanisms. The richness of cellular response to environmental cues was first described in bacteria and studied in detail with regard to how bacteria develop resistance and tolerance to antibiotics (Brauner et al., 2016) and other stress conditions (Orruno et al., 2017) (Harms et al., 2016). Among the strategies that exist to create resistance or tolerance, some particularly efficient ones involve growth arrest of bacterial cells. Clearly, starvation, stress, and growth arrest are the conditions that the vast majority of living cells experience in any environment most of the time as reflected from the fact that addition of nutrients to sea water or soil particles would increase the concentration of cells by several logs. Therefore, the understanding of how cells use long-term stationary phase (LTSP) strategies to overcome challenges in their environment for long times is important and has many practical uses. LTSP cultures of bacteria with the added challenge of antibiotics or abiotic stresses have highlighted phenomena exhibited by bacteria facing these conditions. GASP (growth advantage in stationary phase) cells arise in LTSP cultures as resistant mutants that overtake the entire culture in waves (Farrell and Finkel, 2003) (Finkel and Kolter, 1999). It is vital to clarify at this point that there is a difference between regulated stationary phase and GASP. During stationary phase, cell division is inhibited by tight regulation (Gray et al., 2004). In GASP, however, the situation is quite different. During continuous growth of a culture, after nutrients have been depleted from the culture, and the culture has been maintained in stationary phase for several days, without further addition of medium, death phase occurs. At this stage, more than 90% of the cell population dies. In the few percent of living cells, cycles of cell death and cell division occur. Net cell number does not change dramatically for a long time, but cell division does occur (see later discussion). It is at this stage that cells start to genetically adapt (through mutation-selection cycles) to the harsh culture conditions. Researchers have found (Zambrano et al., 1993) (Zambrano and Kolter, 1996), (Brauner et al., 2016) that, when GASP cells are faced with a chemical or physical challenge, resistant mutants start to emerge. Since GASP division rate is quite low (see later discussion), much time is needed for the resistant mutants to take over the entire culture. During that time, secondary mutants arise, with better resistance against the challenge than the first resistant mutants. Those also start the slow process of taking over the culture. Cleverly designed experiments have been able to demonstrate the existence of these mutations—selection waves in planktonic cultures (Zambrano et al., 1993) (Zambrano and Kolter, 1996) (Farrell and Finkel, 2003) (Finkel and Kolter, 1999) and inside agar (Baym et al., 2016). The hallmark of these mutants is the ability of the end point isolated strain in this experiment to successfully compete with and outgrow the original parental strain in a direct competition experiment. This hallmark gave the phenomenon its name: GASP (growth advantage in stationary phase) (Finkel, 2006).

Bacterial strains exhibiting GASP have been sequenced, and mutations repeatedly found in GASP have been isolated (Zambrano et al., 1993) (Zinser and Kolter, 2004) (Zinser and Kolter, 1999) (Zinser and Kolter, 2000) (Zinser et al., 2003). Surprisingly, many of these genes were found to be related to generalized stress responses such as the RNA polymerase subunit Sigma S (RpoS, stationary phase Sigma factor). Evidently, strains harboring these mutations alone are able to successfully compete with their naive parental strains (Zambrano et al., 1993). It is important to note that, classically, GASP mutants are also resistant to different stresses in addition to being more fit for competition.

Indeed, several mechanisms of drug and stress resistance and some tolerance mechanisms have been demonstrated and documented in eukaryotes (Casalinuovo et al., 2004) (Bojsen et al., 2017) (Huang et al., 2001) (Huang and Houghton, 2001). However, to the best of our knowledge, the GASP phenotype was not described in eukaryotes.

Here we studied the dynamics of Saccharomyces cerevisiae yeast cells in long-term stress conditions, as a model for how eukaryotic cells develop stress resistance and tolerance in LTSP cultures. To this end, we looked for isolated environments where yeast cells remained in starvation conditions and where natural GASP survivors might have evolved. One such environment was found in “old” unfiltered beer bottles left for 2 years in a Jerusalem beer brewery, where yeast cells endured harsh conditions, which include low nutrients, high ethanol concentration, and mild temperature (>10°C, <25°C for two years). By comparison of the yeast cells isolated from these bottles with their parental strains, we were able to demonstrate that GASP does indeed appear in eukaryotic LTSP cultures. We have recently shown (Aouizerat et al., 2019) that yeast can survive for thousands of years as micro-colonies in the nanopores of ancient pottery. The GASP mechanism shown here could significantly contribute to such survival.

Results

Isolation of GASP Yeast Strains from Aged Beer Bottles

To determine whether the GASP phenomenon occurs in eukaryotes, we have isolated two strains of Saccharomyces cerevisiae that sustained a long period of stressful environment, from two different 2-year-old sealed bottles of beer. The isolated strains were termed Sc404 and Sc424, and their phenotypes in various conditions were characterized and compared with the parental commercial beer strain, Saccharomyces cerevisiae Safale S-04 (Fermentis Division of S.I.Lesaffre, France). Sc404 and Sc424 showed significantly smaller colonies when growing on rich YPD plates than the parental strain, and smaller cells both in micrographs and by fluorescence-activated cell sorting (FACS) analysis (Figures 1A, 1C, and 1D). In addition, we have found that, although Sc424 and Sc404 presented similar growth rates in optimal conditions of rich media and optimal temperature, they showed different growth parameters in comparison with the commercial strain in a nutrient-depleted filtrate of the beer from which it was isolated (Figures 1B and S1). These differences have prompted us to further examine the growth of both starvation-adapted stains in various conditions, to understand better the changes that the strains have gone through during adaptation, and check whether this adaptation is similar to GASP.

Figure 1

Sc404 and Sc424 Are Yeast Strains Isolated from Long-Term Stationary Phase in Stressful Conditions, Presenting Different Characteristics Than the Parental Strain Safale S-04

(A) Sc404 and Sc424, isolated from 2-year-old sealed bottles of beer, present significantly smaller colonies than Safale S-04, as shown in cell cultures grown for 72 h on solid YPD plates in 30°C.

(B) Growth kinetics of all three strains were tested for 72 h by measuring the turbidity of the cultures in a 96-well plate reader every 20 min. Note that the three strains show similar growth rates in rich YPD medium at 30°C.

Data are represented as mean ± SEM. (C) Flow cytometry for cell size: cell size quantification was preformed using flow cytometry, based on forward scatter and side scatter measurements. (D) microscope images of cells from the different strains. Scale bar is 10 μM.

Resistance and Growth in Stressful Environments

To investigate stress adaptation, we compared the growth of Sc404 and Sc424 and the parental strain in various stress conditions as one of the hallmarks of GASP in bacteria is their ability to tolerate and survive better such conditions. In all prolonged stress conditions tested: 5% or 10% of ethanol, high osmolality levels (1 M NaCl), and alkaline conditions (pH 10), Sc404 and Sc424 presented significantly better growth over Safale S-04 (Figures 2A–2C). Moreover, Sc404 and Sc424 were also able to tolerate better an acute stress for a short time. This included high ethanol concentration (27.5% EtOH) and high temperature (55°C) (Figure 2D). Importantly, the tested stress conditions extend beyond the environment of the beer bottle from which Sc404 and Sc424 were isolated.

Figure 2

Sc404 and Sc424 Show Increased Ability to Tolerate Various Stressful Environments Compared with Safale S-04

The growth kinetics of Sc404, Sc424, and Safale S-04, represented by measurements of the cultures turbidity in a 96-well plate reader every 20 min for 48 h, were monitored in different stressful conditions.

(A–D) These conditions include (A) ethanol in 5%(v/v) and 10%(v/v), (B) high alkaline environment (pH = 10), and (C) medium with high osmotic pressure (YPD containing 1 or 2 M NaCl). In all tested conditions, the growth of Sc404 and Sc424 was affected significantly less than Safale S-04, which was inhibited almost completely. Data are represented as mean ± SEM. (D) Sc404, Sc424, and Safale S-04 were exposed to identical sublethal conditions, where Sc404 and Sc424 presented noticeably better tolerance than Safale S-04. The conditions include high temperature (55°C for 10 min) and ethanol (27.5% v/v for 1 h). The cultures had been grown in the same conditions to equal optical density and, after exposure to the stress conditions, spotted on YPD plates.

GASP Predominantly Consists of Genetic Changes Rather Than Epigenetic Changes

The increased ability of LTSP isolates to withstand stress could originate from permanent genetic changes such as mutations, deletions, and loss or gain of heterozygosity, as previously described in bacteria, or from changes that are epigenetic and reversible in nature such as the changes leading to persisters formations (Fisher et al., 2017).

Researchers have demonstrated the predominant genetic changes in bacterial GASP strains, by showing that the modified strains have maintained their phenotype even after repeated rounds of serial passage through log-phase growth. The Sc404 and Sc424 strains have been passed through a similar process of serial passages for six consecutive days on rich medium (YPD), to select against epigenetic modifications. The yeast strains created in this process were termed 404St1, 404St2, and 404St3; and 424St1, 424St2, 424St3; and 424St4, and their phenotypes were tested and compared with Sc404. Even after serial passages on rich medium, the 404St and 424St strains still showed a significant advantage over the original Safale S-04 strain both in growth kinetics and survival in stressful environments (Figures 3 and S2). It is interesting to note that despite all 404St and 424St showing a significant advantage over the original Safale S-04 strain, there is some phenotypic variability among the tested strains. It is not clear if this variability originates from experimental conditions or if it expresses epigenetic variation among the progeny of Sc404 and Sc424. We conclude that genetic changes, which accumulated during a long starvation period, are the predominant factor in the ability of Sc404 and Sc424 to tolerate stress, rather than epigenetic changes.

Figure 3

Sc404 Cells Maintain Their Stress-Driven Characters after Serial Dilutions

(A–C) 404St strains were evolved by diluting overnight cultures of Sc404 through a period of 6 days (see Methods). The 404St strains present similar qualities to those of Sc404 such as in (A) growth in YPD and (B) in the presence of ethanol (10% v/v), represented by measurements of the culture turbidity in a 96-well plate reader every 20 min for 48 h. Data are represented as mean ± SEM. (C) 404St strains together with Sc404 and Safale S-04 were grown to the same optical density and exposed to high temperature (55°C for 10 min) and then spotted on YPD plates. In these conditions, 404St strains maintained their advantage in stress conditions over the parental strain Safale S-04.

Genomic Sequence Analysis and Validations

To examine the changes the survivor strains underwent during stress adaptation, we first investigated the genome content of the parental and survivor strains by FACS (Figure 4A). According to our analysis, the parental strain is tetraploid and the survivor is diploid (Sc404), or close to diploid (Sc424, a little higher than diploid). To test whether chromosomes are present in homolog pairs in the survivors, we performed a sporulation test. Sc404 and Sc424 sporulated efficiently, as did Safale S-04 (Figure S3). It is possible that at least one round of meiosis occurred during adaptation, although stress-related non-meiotic ploidy reduction has also been observed in many fungi (Gerstein et al., 2017) (Hickman et al., 2015). It is important to note that this reduction in ploidy does not necessarily lead to homozigotization of all mutations. Rather, we find many heterozygous mutations in the genome of Sc404 and Sc424 (see later discussion). Furthermore, we sequenced the genomes of Safale S-04 Sc404 and Sc424 (Accessions GeneBank:SRR6706477, GeneBank:SRR6706475, and GeneBank:SRR6706476 correspondingly, and Tables S1 and S2 showing genomic changes and Table S3 showing heterozygosity) and determined the differences between these strains. An analysis of chromosome copy numbers showed that in Safale S-04 and Sc404 there was no over-representation of any chromosome in the genome (Figure 4B and Table S4). In contrast, the genome of Sc424 shows an over-representation of chromosomes I, III, and VI (Figure 4B and Table S4), a result that is consistent with previous reports of a possible adaptive duplication of chromosomes (Gerstein et al., 2017) (Hickman et al., 2015). The summary of genomic changes is presented in Figure 4C. The main mechanism (more than 65% of genomic changes in both strains) of change in the genomes of both Sc404 and Sc424 are changes in heterozygosity (loss of heterozygosity and gain of heterozygosity, LOH + GOH). As shown in Figure 4C the genomes of Sc404 and Sc424 also contain deletions and insertions compared with Safale S-04. The sequencing results were also validated by Sanger sequencing in several representative loci, with complete identity to the predicted sequence by next generation sequencing (data not shown).

Figure 4

Genomic Analysis of Safale S-04, Sc404, and Sc424

(A) Flow cytometry for DNA content: DNA content was evaluated by flow cytometry following permeabilization and fixation using methanol and treatment with propidium iodide before measurements.

(B) Average sequencing coverage per chromosome in the three sequenced strains shows uniform coverage in all chromosomes in strains Safale S-04 and Sc404. Sc424 shows an over-representation of chromosomes I, III, and VI.

(C) All genomic differences between Safale S-04, Sc404, and Sc424 were divided to appropriate categories according to the change caused by the mutation. The relative abundance of these categories is presented in the pie chart. The list of genomic changes can be found on Tables S1, S2, and S3.

A functional annotation (https://david.ncifcrf.gov/) performed on all changed genes in Sc404 and Sc424 shows an enrichment in DNA repair, cell cycle, metabolism, autophagy, aging, and meiosis genes (Tables 1, 2, and S5). Many genes with mutations are mutated in both Sc404 and Sc424 (see Figure 5A). (Genes included are ones with a SNP mutation in the “type” column and “false” in the “heterozygote” column in Tables S1 and S2. For a detailed list of genes included in this analysis, see tab “genes included in 5A” in Tables S1 and S2, and the functional enrichment of these common genes) Tables 3 and S5 show enrichment for the same functional categories as earlier. These results suggest that convergent evolution is at play in this genetic adaptation process.

Table 1

Selected Functional Enrichment Categories of Genomic Changes in Sc404 vs. Safeale S-04

Functional Category	# Genes in List	# Genes in Category	p Value	Fold Enrichment
DNA repair	38	191	6.4 × 10−5	2
Protein localization to pre-autophagosomal structure	7	9	6.7 × 10−5	7.7
RNA splicing	22	111	3.3 × 10−3	2
Phosphorelay signal transduction system	4	5	8.8 × 10−3	7.9
Budding cell apical bud growth	7	20	1.2 × 10−2	3.5
Chromatin remodeling	13	59	1.3 × 10−2	2.2
Replicative cell aging	10	39	1.4 × 10−2	2.5
Late nucleophagy	6	17	2.3 × 10−2	3.5
Positive regulation of actin cytoskeleton reorganization	4	7	2.6 × 10−2	5.6
Steroid biosynthetic process	7	25	3.5 × 10−2	2.8

Genes where genomic changes occur (see Tables S1 and S2) were subjected to functional enrichment analysis using the DAVID tool (https://david.ncifcrf.gov/summary.jsp). Names of selected categories with their enrichment statistics are listed.

Table 2

Selected Functional Enrichment Categories of Genomic Changes in Sc424 vs. Safeale S-04

Functional Category	# Genes in List	# Genes in Category	p Value	Fold Enrichment
Positive regulation of transcription, DNA-templated	8	17	1.2 × 10−3	4.4
Double-strand break repair	12	37	1.2 × 10−3	3
DNA repair	35	191	1.8 × 10−3	1.7
Budding cell apical bud growth	8	20	3.5 × 10−3	3.7
Peptidyl-serine phosphorylation	10	32	5.1 × 10−3	2.9
Replicative cell aging	11	39	6.6 × 10−3	2.6
Chromatin remodeling	14	59	8.3 × 10−3	2.2
Methylation	18	87	9.7 × 10−3	1.9
Autophagy	15	67	1.0 × 10−2	2.1
Mannose metabolic process	4	5	1.0 × 10−2	7.5

Genes where genomic changes occur (see Tables S1 and S2) were subjected to functional enrichment analysis using the DAVID tool (https://david.ncifcrf.gov/summary.jsp). Names of selected categories with their enrichment statistics are listed.

Figure 5

Genomic Analysis of Safale S-04, Sc404, and Sc424

(A–C) (A) The number of genes with genomic changes in Sc404 and Sc424 compared with Safale S-04—note that 751 genes are changed in both strains. (B) The growth kinetics of Sc404, Sc424, and Safale S-04, represented by measurements of the cultures turbidity in a 96-well plate reader every 20 min for 48 h, were monitored in YPD with 2.2 μM Rapamycin. Data are represented as mean ± SEM. (C) EMS resistance: Overnight yeast cultures were diluted to various concentrations and then spotted on agar plates containing 6 μM of EMS in drops containing 3 mL each. Note that Sc404 is more sensitive to the drug than Safale S-04.

(D) Fluctuation test: a fluctuation test for the ability of the three strains to grow on 5-FOA-containing medium (see Methods). The calculated level of mutations per strain is shown. Note that the level of mutations in Sc404 and Sc424 is more than in Safale S-04.

Table 3

Selected Functional Enrichment Categories of Shared Genomic Changes in Both Sc404 and Sc424 vs. Safeale S-04

Functional Category	# Genes in List	# Genes in Category	p Value	Fold Enrichment
Double-strand break repair	10	37	5.2 × 10−4	4.1
DNA repair	26	191	7.2 × 10−4	2.1
Phosphorylation	26	212	3.2 × 10−3	1.8
Replicative cell aging	9	39	3.4 × 10−3	3.5
Intracellular signal transduction	10	48	3.7 × 10−3	3.1
Chromatin remodeling	11	59	4.8 × 10−3	2.8
Budding cell apical bud growth	6	20	8.4 × 10−3	4.5
Protein localization to pre-autophagosomal structure	4	9	1.8 × 10−2	6.7
Regulation of transcription from RNA polymerase II promoter	15	115	1.9 × 10−2	2
Late nucleophagy	5	17	2.3 × 10−2	4.4

Genes where genomic changes occur (see Tables S1 and S2) were subjected to functional enrichment analysis using the DAVID tool (https://david.ncifcrf.gov/summary.jsp). Names of selected categories with their enrichment statistics are listed.

Stress response can be heavily influenced by the ability of cells to activate cellular signaling. One of the major complexes contributing to this signaling is the TOR complex (Lushchak et al., 2017) (Gonzalez and Hall, 2017). In our genomic analysis we noticed that mutations occurred in the TORC1 and TORC2 complexes (see Tables S1 and S2). Specifically, the TOR1 protein has 20 SNPs between Sc404 and SafAle S-04, the LST8 gene has 5 SNPs, and the SLM1 has 1 SNP. TOR1 also has 20 SNPs between Sc424 and SafAle S-04 (the second tab in Tables S1 and S2 now enables the search of more genes with SNPs in both strains). In accordance with the mutations in TORC1 and TORC2, Sc404 was more sensitive to Rapamycin than Safale S-04. Incubation of the Sc404 and Sc424 yeast strains in growth medium containing 2.2 μM Rapamycin resulted in marked growth inhibition of Sc404 (note that Sc404 plateaus at a lower OD than Safale S-04). Interestingly, Sc424 shows no such sensitivity to Rapamycin (Figure 5B). To investigate whether these and additional mutations affected transcription of TOR subunits, we performed qRT-PCR analysis on all TOR components mRNA in rich medium and medium containing ethanol. The results (Figure S4) show no significant differences between the expression levels of TOR components in Sc404 and Safale S-04 except for the gene SLM1. It is unlikely that a modest elevation in SLM1 induction alone will lead to a global change in transcriptional control of TOR complexes.

The occurrence of many mutations in the yeast genome may be explained by a phenotype with a higher mutation rate, which was also previously described as occurring during the domestication process of beer yeast (Gallone et al., 2016) and GASP in bacteria (Avrani et al., 2017). We have thus looked at the sequence divergence of our survivor in DNA repair genes. We found (Table 4) that many DNA repair genes, in most DNA repair pathways, are changed. Notable are mutations in genes that were shown to contribute to phenotypes with a higher mutation rate before such as the gene EXO1 (Tran et al., 1999). Many genes involved in homologous recombination (HR) were also mutated such as RAD50, MRE11, RAD51, and RAD54 (Lisby and Rothstein, 2015). Other genes that are components of DNA damage checkpoints such as CHK1 and MEC1 were changed during starvation (Sanchez et al., 1999) (Sanchez et al., 1996). In support of these results, we found that the survivor strains are slightly more sensitive when exposed to ethyl methanesulfonate (EMS), relative to the parental strain (Figure 5C) as expected from a phenotype with a higher mutation rate.

Table 4

Changed DNA Repair Genes

TFB4	MCM3	CHK1	SIN3
LIF1	MSH6	DPB3	TFB5
REV1	MCM2	MLH1	RAD2
ACT1	TUP1	RPT6	SMC5
DNA2	MCM7	POL2	RFC4
MSH3	TEL1	LIG4	NUP84
NDK1	RPB1	OGG1	MCM6
MLH3	UNG1	DPB2	SMC6
PAN2	NUP133	GRR1	KU70
HTA1	MCM5	TOR1	PHO85
RAD59	RFA1	SMC3	REV3
SCC2	RAD51	POL1	MSH2
TFB1	SMC1	DOT1	HTB1
RFA2	APN2	EXO1	CKA1
CUL3	SIR2	THO2	RAD17
MEC1	POL3	HAT1	DPB4
SLX4	RVB1	DMC1	SGS1
NTG2	CKB2	RPT4	SSN6
CDC28	MRE11	SUB2	PAN3
RAD52	RDH54	ESA1	RAD54
RAD50

A list of genes related to DNA repair that are changed in Sc404 and Sc424 vs. Safale S-04.

To measure more accurately the mutation rate in the adapted strains, we performed a fluctuation test for the ability of the yeast to grow in a medium containing 5-Fluoroorotic acid (5-FOA) (Green et al., 1976) (Lang, 2018), (Lang and Murray, 2008) (see Methods). With this assay we can estimate the relative abundance of mutations in the strain of origin and its two adapted decadents. Assuming that the rate of mutations in wild-type yeast strains for growth in 5-FOA is on average 5.5 × 10−8 mutations per genome per generation (Lang and Murray, 2008), the results (Figure 5D) point to an elevation of about 3-fold in Sc424 and about 6-fold in Sc404. These results confirm that the basic mutation level in the adapted strains is higher than in the original strain.

Direct Competition Assays

To determine if Sc404 and Sc424 are able to outcompete Safale S-04, which is a major hallmark of GASP in bacteria, specific primers were designed to allow the identification and quantification of Safale S-04 differentially from the survivor strains in each competition experiment. Equal numbers of cells had been mixed together in both rich medium and stressful conditions (5% EtOH), and the subpopulations were then monitored for several passages. In each passage the population was diluted 400-fold and re-grown. In each passage samples were taken for DNA extraction and qPCR. Results are presented in Figure 6. These results show that, in both rich medium and in 5% EtOH, Sc404 and Sc424 took over the culture after five to six passages. The enhanced capacity of the survivor strains in direct competition in YPD and high ethanol conditions shows that they manifest real GASP capabilities.

Figure 6

Direct Competition Assay

(A and B) The relative fractions of Sc404 (green) (A), Sc424 (blue) (B), and Safale S-04 (red) in the culture (after starting from 50-50) was measured at different conditions by qPCR after each dilution (see Methods). Note that in both YPD and Ethanol conditions, Sc404 and Sc424 take over the culture. Data presented in each experiment are the average of three biological replicates.

Yeast GASP and Cell Divisions

In the previously described bacterial GASP cultures, the growth of new mutants in the culture occurred in waves, each mutant presumably arising from the previous culture and almost entirely taking over the culture owing to growth advantage. This mode of adaptation thus requires active cell division. To test whether active cell division indeed occurs during prolonged starvation in yeast, we subjected the starved cultures (after 2 years of starvation in the same medium as in the starved culture) to live cell imaging. In all fields filmed (see Videos S1 and S2), active cell division was observed (see Figure 7). We also quantified the number of dividing cells in the population, and the average length of the cell cycle during starvation conditions (Figure 7). These measures were used in our estimation of the number of mutations per dividing cell, the number of available mutations in the culture per generation, and the number of total mutations that accumulated in the culture during these 2 years of starvation (see later discussion and Figure 7). To test whether mutations that arise during the course of adaptation indeed increase their representation in the culture (eventually representing the vast majority of the culture) owing to selective power, we sequenced by Sanger sequencing two representative loci (in the coding sequence of ALK1 and in the coding sequence of ERV46). These loci were selected on the basis of having conserved primers that can be used for amplifying all strains: Sc404 Sc424 and SafAle S-04, while having multiple SNPs between the strains in the amplified regions themselves (For the primer sequences, see Methods section, primer design paragraph). 30 independent clones from the original starved cultures were sequenced. The results show no SNPs in both regions in all 30 clones of Sc404 (700- and 1,000-bp-length regions). In the colonies of Sc424 we found 10 SNPs in 30 colonies in each of the 700- and 1,000-bp regions we sequenced. The SNPs were found each in 1 of the 30 colonies (unrelated SNPs). This number of SNPs amounts to a ratio of ∼4 × 10−4 of SNPs diverting from the dominant sequence in the culture. These results show that the mutations that are specific to each strain and arose in one clone during the course of the genetic adaptation are now present in almost all the independent clones that we checked. This means these mutations were selected for in the culture, and it supports the appearance and selection of secondary mutations in waves in the culture. It thus seems that GASP in yeast follows the same basic principles and dynamics as the parallel process in bacteria.

Figure 7

Cell Divisions

(A and B) Examples of still images from live cell imaging experiments of starving cells. Note that cells are dividing (arrows point to the mother cells). See Videos S1 and S2. Scale bar is 10 μM. (C) Percentage of cells that divide in each of the fields imaged by live cell imaging. The red line marks the median level.

(D) Time of division of each dividing cell from bud emergence to the daughter bud emergence. The red line marks the median level.

Video S1. Division in Starvation, Related to Figure 7

Cells starved for 2 years were filmed in starvation conditions. Note that cell divisions occur in starvation and that many dead cells appear around the dividing cells. Also note that cell division is clonal.

Video S2. Division in Starvation, Related to Figure 7

Cells starved for 2 years were filmed in starvation conditions. Note that cell divisions occur in starvation and that many dead cells appear around the dividing cells. Also note that cell division is clonal.

Beer Production from GASP Yeast Strains

Understanding the mechanism of yeast stress tolerance may contribute to more efficient ethanol fermentation and beer production. Sc404 and Sc424 have acquired modifications that improve their ability to tolerate these exact stresses, compared with the well-known commercial Safale S-04. This fact has made Sc404 and Sc424 perfect candidates for the production of beer. Sc404, Sc424, and Safale S-04 have been used simultaneously for producing beers in the same conditions. The beers that were brewed together under the same conditions were tasted by professional beer tasters (of the Israeli beer association). The three beers differ considerably in their taste characteristics, as shown in the scoring results (Figure S5). Of note is that fact that the tastes and aromas of Sc404 and Sc404 are closer to each other in comparison with those of beer brewed from Safale S-04.

Discussion

In the course of their lifespan, cells may find themselves in a variety of different environments and have therefore evolved specific mechanisms to deal with specific acute stress conditions. These mechanisms may become a burden when the cell is exposed to a chronic form of stress. Following this line of thought, an interesting adaptation mechanism to chronic starvation has been described in bacteria. This mechanism was termed GASP (Growth Advantage in Stationary Phase), as the resulting survivors outgrow the parental strain in stressful conditions. Here we describe that a similar adaptation mechanism exists in eukaryotes. As eukaryotic cells are often exposed to the same kinds of stress as bacteria, this may come as no surprise, although the mechanism itself may differ somewhat.

In the same manner as in bacteria, GASP yeast survivors outgrow their parent strain in the stressful conditions it adapted to (see Figure 6). Interestingly, GASP yeast also outgrow the original strain in YPD medium (rich medium), suggesting that YPD poses an adaptive challenge to yeast (perhaps due to fast nutrient exhaustion). The adaptation is composed of a predominantly genetic component rather than an epigenetic one (Figure 3). This point is reflected by the extent of genetic differences between the two strains (Figures 4 and 5). Since multiple ways for stress adaptation exist, including persistence, which contains a major epigenetic component (Sun et al., 2016) (Bojsen et al., 2016) (Knoechel et al., 2014) (Guler et al., 2017) (Rosenberg et al., 2018), this result constitutes a novel and interesting angle to adaptation under stress. This observation could suggest that a prolonged exposure to stress causes a shift toward a genetic adaptation and that in the first stages of adaptation epigenetic forms of adaptation would be prevalent. This hypothesis has been suggested with experimental evidence in bacteria (Levin-Reisman et al., 2017) and merits further experimental exploration in yeast and other eukaryotes in future work. Previous work has addressed the core gene mutations needed for adaptation to environmental stress, and a set of 16 regulators has been identified (Chasman et al., 2014). Interestingly, of the three main genes in this core, HOG1 and PDE2 are mutated in both Sc404 and Sc424 and MCK1 is mutated only in Sc424. Of the remaining 13 genes, 4 (RIM101, GBP2, RIM15, and SWC5) are mutated in both Sc404 and Sc424 and 1 gene (NPR2) is mutated only in Sc424. This shows a high degree of conservation with the critical core of genes needed for genetic adaptation to stress. It should be noted that the fact of using a different experimental approach than ours (mainly screening of single gene deletion libraries), could underestimate the overlap between the conditions tested by the authors and our starvation conditions.

Our genomic analysis revealed transformative changes in Sc404 and Sc424 compared with Safale S-04. These changes follow previously described adaptations in eukaryotes, with similarities to GASP in bacteria. Loss and gain of heterozygosity play important roles during eukaryote and yeast evolution (Hope et al., 2017), (Polyak et al., 2009), and the GASP adaptation process is no different. This loss and gain of heterozygosity could in principle, at least in part, originate by cycles of meiosis performed by the yeast cells, which may lead to the reduction in ploidy observed in the survivor strains (Figure 4), although mitotic mechanisms for LOH seem more likely than meiotic mechanisms (Hope et al., 2017), (Polyak et al., 2009). Finally, we also see the adaptation process operating to mutate specific processes that have important downstream influences on life in stressful conditions, such as DNA repair pathways. Mutations in DNA repair genes may manifest in a phenotype with a higher mutation rate, in a similar manner to hyper-mutability described in bacteria under starvation conditions (Avrani et al., 2017). One particularly interesting system that is under the influence of the adaptation process is the TOR complexes (Lushchak et al., 2017) (Gonzalez and Hall, 2017). It seems that TOR1 has a large number of SNPs in both adaptive strains. Additional mutations and changes in the regulation of transcription of the complexes subunits exist (LST8 and LSM1, see earlier text), which make it hard to predict the status of the entire complex and merit additional research. However, this situation bears a remarkable similarity to the mutagenesis of RpoS found in bacteria under GASP conditions (Zambrano et al., 1993) and shows that changes in stress signaling may be required for GASP adaptation.

The dynamics of the population are also similar to bacterial GASP. We show (Figure 7) that cells divide under starvation conditions, presumably feeding off the remains of their deceased brethren. We also show that these divisions together with strong selective pressure give rise to the high prevalence of specific mutations in the starving population. This leads to the conclusion, as is the situation in bacterial GASP (Avrani et al., 2017), that a mutant outgrows the others in the culture and takes over the vast majority of the population, only to wait for the presentation of a secondary mutant and its selection. From our movies, we devised a mathematical model (see Methods) for the calculation of the number of mutations in the population during the starvation period. Our estimation amounts to the accumulation of the 170,262 mutations for Sc404 and 79,456 mutations for Sc424 in 2 years of starvation. This amount of mutations is largely sufficient for the accumulation of approximately 30,000 changed nucleotides in each of the survivors' genomes (after selection). These results also show that dynamic processes and the accumulation of many mutations govern the behavior of cells during starvation, even after years without nutrients. This is indeed the situation in most ecological environments on earth, where microorganisms live with a poor supply of nutrients, competing for available resources and adapting to the situation in a dynamic manner.

We have shown here that GASP can occur in unicellular eukaryotic organisms, but in principle, this adaptation could also operate in multicellular organisms. One intriguing example occurs in tumors. The interior of tumors is characterized by being an extremely stressful environment (Cairns and Mak, 2017). Hypoxia, low pH, and nutrient unavailability all restrict the ability of cells to proliferate. In the same way that yeast cells adapt to a stressful environment in a beer bottle, cancer cells may adapt to harsh conditions in the interior of a tumor. This adaptation may explain the more aggressive phenotype of tumors when they re-grow after surgical excision (Goldfarb and Ben-Eliyahu, 2006). The principle of adaptation through GASP may therefore have broad implications, and the principles shown here in yeast may be conserved in other species and systems.

Limitations of the Study

Our study is limited to the start and end time points of the adaptation process. In the future, a careful time course for the progress of the adaptation needs to be performed. Another limitation comes from the fact that we characterized (albeit to a great degree of detail) only two clones from the adaptation process. In the future, more clones should be analyzed to assess the reproducibility of the course of genetic adaptation.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.