Cellular memory of past experiences has been observed in several organisms and across a variety of experiences, including bacteria “remembering” prior nutritional status and amoeba “learning” to anticipate future environmental conditions. Here, we show that Saccharomyces cerevisiae maintains a multifaceted memory of prior stress exposure. We previously demonstrated that yeast cells exposed to a mild dose of salt acquire subsequent tolerance to severe doses of H2O2. We set out to characterize the retention of acquired tolerance and in the process uncovered two distinct aspects of cellular memory. First, we found that H2O2 resistance persisted for four to five generations after cells were removed from the prior salt treatment and was transmitted to daughter cells that never directly experienced the pretreatment. Maintenance of this memory did not require nascent protein synthesis after the initial salt pretreatment, but rather required long-lived cytosolic catalase Ctt1p that was synthesized during salt exposure and then distributed to daughter cells during subsequent cell divisions. In addition to and separable from the memory of H2O2 resistance, these cells also displayed a faster gene-expression response to subsequent stress at >1000 genes, representing transcriptional memory. The faster gene-expression response requires the nuclear pore component Nup42p and serves an important function by facilitating faster reacquisition of H2O2 tolerance after a second cycle of salt exposure. Memory of prior stress exposure likely provides a significant advantage to microbial populations living in ever-changing environments.
NATURAL environments are complex and often vary significantly in space and time, posing challenges for the organisms living within them. Single-cell organisms are particularly vulnerable, since variation in external conditions can directly impact internal homeostasis. Therefore, preparing for environmental change after early signs of fluctuation would present a significant advantage for cells growing in the wild. Indeed, many organisms can become tolerant to severe stress after an initial mild pretreatment with the same or a different stressor. This response, termed “acquired stress resistance,” has been observed in microbes such as bacteria and yeast as well as in multicellular organisms including worms, plants, mammals, and even humans (Lu et al. 1993; Davies et al. 1995; Lewis et al. 1995; Lou and Yousef 1997; Swan and Watson 1999; Chi and Arneborg 2000; Schenk et al. 2000; Durrant and Dong 2004; Kandror et al. 2004; Scholz et al. 2005; Hecker et al. 2007; Kensler et al. 2007; Matsumoto et al. 2007). The conservation of this response suggests that it plays an important role in surviving environmental stress in diverse species.
We previously conducted a systematic analysis of acquired stress resistance in Saccharomyces cerevisiae and found that the response is common, but not universal, for all pairs of stress treatments (Berry and Gasch 2008). The level and duration of mild-stress pretreatment varies according to the mild stressor used, but the timing of stress-tolerance acquisition correlates with the gene-expression dynamics during each pretreatment. Consistently, acquired stress resistance is dependent on nascent protein synthesis during the mild-stress exposure, but not during the severe-stress treatment (Berry and Gasch 2008; Lewis et al. 2010). In separate work, we identified genes important for acquisition of stress tolerance after different mild-stress pretreatments (Berry et al. 2011). Somewhat surprisingly, the mechanism of acquired stress resistance is condition-specific, rather than being dependent on the commonly activated environmental stress response. For example, acquisition of H2O2 tolerance is commonly observed after different mild-stress pretreatments but occurs through largely distinct gene sets in each case (Kelley and Ideker 2009; Berry et al. 2011). This result is explained in part by redundant functions served by different gene sets, but also emerges due to the condition-specific nature of stress defense for each pair of mild- and severe-stress treatments.
One open question is how long acquired stress tolerance persists after cells have been removed from the initial stressor, and whether the acquired stress tolerance can be transmitted to new daughter cells. Cellular memory of past life experiences has been observed in several organisms. For example, yeast and bacteria exposed to prior nutrient signals display altered metabolism and gene expression or chemotaxis patterns, respectively, upon later exposure to specific environmental cues (Koshland 1977; Casadesús and D’Ari 2002; Acar et al. 2005; Brickner et al. 2007). Adult Caenorhabditis elegans that emerge from starvation-induced “dauer” development grow up with higher fecundity and correspondingly altered gene expression (Hall et al. 2010). Even amoeba can “learn” to predict environmental fluctuations based on historical events, altering their behavior in anticipation of a recurring pulse of stimulus (Saigusa et al. 2008). In some cases, cellular memory can be transmitted across generations, in both single-celled organisms (Casadesús and D’ari 2002; Ng et al. 2003; Acar et al. 2005; Ajo-Franklin et al. 2007; Brickner et al. 2007; Burrill and Silver 2011) and multi-cellular offspring (Goh et al. 2003; Anway et al. 2005; Molinier et al. 2006; Sigal et al. 2006; Crews et al. 2007; Boyko et al. 2010; Burns and Mery 2010; Carone et al. 2010; Ng et al. 2010).
Here we show that yeast cells previously exposed to mild stress retain a memory of acquired stress tolerance for many generations after the initial stressor has been removed. We found that the memory exists on two separate levels: persistence of acquired H2O2 tolerance and transcriptional memory that promotes faster reacquisition of stress tolerance after a second round of treatments. We present mechanisms of this cellular memory and discuss its functional relevance in natural environments.
Materials and Methods
Strain and growth conditions
All experiments were done in YPD (1% yeast extract, 2% peptone and 2% dextrose) or in synthetic drop-out media lacking leucine and uracil where indicated. Unless otherwise noted, strains used were of the BY4741 background (Mata his3Δ1 leu2Δ0 met15Δ0 ura3Δ0). Deletion strains were purchased from Open Biosystems, and gene deletions and correct integration of the KANMX cassette were validated by PCR amplification. An independent ctt1Δ strain was created in BY4741 by replacing CTT1 with the URA3 gene (AGY639). This strain was used to introduce FLAG-tagged CTT1, generated by PCR mutagenesis to introduce one copy of the FLAG epitope (DYKDDDDK) between the fifth and sixth amino acid of Ctt1p. The cassette was then integrated into the ctt1::URA3 locus by homologous recombination and selection on 5-fluoro-orotic acid and verified by sequencing. The strain showed roughly wild-type sensitivity to NaCl and H2O2 stress (data not shown). Inducible FLAG-CTT1 expression was accomplished by cloning the FLAG-CTT1 fusion described above, along with 629 bp downstream of the CTT1 coding sequence under control of the GAL1 promoter in plasmid pRS316-GAL1. This plasmid was cotransformed, along with pGEV-LEU (Gao and Pinkham 2000), kindly provided by David Eide, into a ura− revertant of AGY714 in which GAL4 had also been deleted for maximal estradiol-dependent induction.
Memory of acquired stress resistance experiments
Cultures were grown in shaker flasks at 30° for at least eight generations to an optical density (OD600) of 0.3–0.4. Each culture was split into two cultures: one was treated with 0.7 M NaCl for 60 min. The other culture served as a mock control that was handled identically but received no stress. Cells were then washed once with medium and resuspended in prewarmed, stress-free medium and grown for at least 6 hr.
H2O2 tolerance was measured by transferring cells to a 96-well plate and exposing for 2 hr to 11 doses of H2O2 plus a no-stress control, ranging from 0 to 6 mM for experiments done in YPD or from 0 to 12 mM for cells grown in synthetic medium (since cells grown in synthetic medium display slightly higher H2O2 tolerance). After a 2-hr incubation with shaking, the cells were either plated on YPD plates to assess colony-forming units [using a four-point scale as previously described (Berry and Gasch 2008)], or exposed to LIVE/DEAD stain (Invitrogen, Carlsbad, CA) to measure percentage viability on a Guava 96-well flow cytometer (Millipore, Billerica, MA) and then normalized to viability measured in the unstressed cells. To capture the breadth of H2O2 tolerance, we calculated a single acquired-stress survival score for each time point, based on the sum of the viability scores at each dose of H2O2 minus the comparable sum for mock-treated cells. Unless otherwise noted, the percentage maximum score relative to the time point immediately after pretreatment is shown.
To induce FLAG-CTT1 expression in the inducible strain, 1 μM β-estradiol was added to the culture for 3 hr, after which time cells were removed from the culture, washed with fresh medium, and grown in stress-free medium for denoted times.
Mother–daughter cell experiments
S288C-derived yeast strain UCC8613 that carries the loxP-flanked ADE2 gene was transformed with the plasmid pDL24, which contains a fusion of the Cre gene and estradiol-binding domain driven by the DSE2 daughter-specific promoter (Lindstrom and Gottschling 2009) (both kindly provided by the Gottschling lab). The memory of acquired stress resistance was measured as described above, except that (i) the transformed cells were grown in YPD containing 100 μg/ml clonNAT (WERNER BioAgents) to maintain the plasmid and (ii) 1 μM estradiol (Sigma Aldrich, St. Louis) was added to the culture after treatment of primary stress to activate the Cre recombinase, which is expressed only in daughter cells. Cells were plated on SC−ade to score ade+ mothers or plated on YPD to count ade+ (white) mother cells and ade− (red) daughter cells.
Cells were grown for at least three doublings in YPD at 30° to early log phase. A sample of the culture was collected as the unstressed reference, and the remaining culture was split into three subcultures. The first subculture immediately received 0.5 mM H2O2, and cells were collected for microarray analysis at 10, 20, 30, or 40 min. The second subculture was treated with 0.7 M NaCl for 60 min, washed with medium, returned to 30° YPD medium for 240 min, and then treated with 0.5 mM H2O2 as described above. The third subculture served as mock control and was handled identically but received no NaCl treatment before exposure to 0.5 mM H2O2. All cells were collected by brief centrifugation and snap-frozen in liquid nitrogen. A similar experiment was performed for paired wild-type and nup42Δ cultures and analyzed before stress at 30 min after 0.7 M NaCl treatment and at 10 and 20 min after H2O2 treatment.
Total RNA extraction, cDNA synthesis, and labeling were performed as previously described (Gasch 2002) using amino-allyl-dUTP (Ambion, Austin, TX), Superscript RT III (Invitrogen, Carlsbad, CA), and NHS-ester cyanine dyes (Flownamics, Madison, WI). Spotted microarrays used for Figure 4A data were produced in-house using 70-mer oligonucleotides representing each of the yeast ORFs (Qiagen, Valencia, CA). Array hybridization, scanning, and data analysis were performed as described (Berry and Gasch 2008; Alejandro-Osorio et al. 2009). Each time-point sample was compared to the paired unstressed sample to measure fold changes in gene expression. Analysis of the paired wild-type and nup42Δ responses was done similarly but hybridized to tiled genomic arrays from Roche Nimblegen as previously described (Huebert et al. 2012). All microarray data are available in the National Institutes of Health Gene Expression Omnibus database under accession no. GSE32196.
Genes with significant expression differences in paired pretreated vs. naive samples were identified using the Bioconductor package limma (Smyth 2005), and q-values (Storey and Tibshirani 2003) were calculated to assess the false discovery rate (FDR). Data shown in Figure 4 were organized by hierarchical clustering (Eisen et al. 1998). Motif analysis was done using the program MEME (Bailey and Elkan 1994), searching 1000 bp upstream of genes from each cluster in Figure 4 using the zero-or-one per sequence model. Examples of the motif were scored in all 1000 bp upstream regions using the program FIMO (Grant et al. 2011), and enrichment in the group of genes from cluster 3 was scored using the hypergeometric distribution.
Quantitative RT-PCR and Westerns
Expression of selected genes was measured by real-time quantitative PCR (qPCR) using SYBR Green Jumpstart Taq (Sigma) and an Applied Biosystems 7500 detector (Foster City, CA). The average of technical replicates was normalized to the internal control transcript ERV25, which does not show stress-dependent changes in expression. The samples collected after stress addition were compared with the unstressed control. Three biological replicates were performed to assess statistical significance.
Protein was quantified from whole-cell extract prepared from 5 ml of cells collected by centrifugation and immediate freezing. Cell pellets were resuspended in sample buffer, normalizing by cell count. Westerns were performed using primary antibodies against the FLAG tag (Sigma monoclonal FLAG antibody F3165) or Sse3/4p (kindly provided by Betty Craig) and actin (Sigma Actin Antibody AC40) as an internal loading control and secondary goat anti-mouse antibody (Li-cor, Lincoln, NE; 926-32210). Bands were quantified on an Odyssey Infrared Imaging System (Li-cor), and protein was normalized to actin abundance within each lane.
Memory of acquired H2O2 resistance after mild NaCl treatment
We first recapitulated that yeast cells acquire H2O2 tolerance after a mild NaCl pretreatment. Actively growing cells were treated with a viable dose of 0.7 M NaCl (referred to as the “primary” stress treatment), and at various times an aliquot of culture was removed, exposed to 11 doses of H2O2 (referred to as the “secondary” stress) for 2 h, and scored for viability. To capture the full spectrum of acquired H2O2 tolerance relative to unstressed cells, a single tolerance score was computed and normalized to the maximum score to represent percentage maximal tolerance (see Materials and Methods for details). In agreement with previous results (Berry and Gasch 2008), we found that H2O2 tolerance increased during the NaCl pretreatment and reached maximum levels by 60 min (Figure 1B and data not shown).
We next investigated how long the acquired H2O2 tolerance persisted after cells were removed from salt. To do this, cells were exposed to 0.7 M NaCl for 60 min, but then removed from the primary stress and allowed to grow in stress-free YPD medium for many generations (tracked by changes in cell density or colony-forming units over time). Remarkably, H2O2 tolerance persisted for over 360 min (more than three generations, Figure 1) and remained statistically above background for over four generations after cells were removed from the primary NaCl stress (Supporting Information, Figure S1, and data not shown). Resistance slowly decayed over time, revealing that the increase in H2O2 tolerance was due to an epigenetic mechanism rather than to accumulated genetic mutations. Thus, cells retain a memory of prior salt treatment in the form of elevated H2O2 tolerance.
Memory of acquired stress resistance is independent of de novo protein synthesis but decays with cell division
To test if active transcription or translation is required to maintain the memory of the acquired H2O2 resistance, we added the protein-synthesis inhibitor thiolutin after NaCl treatment, when cells were returned to stress-free medium. A final concentration of 10 μg/ml thiolutin—a dose at which both transcription and translation are halted (Jimenez et al. 1973)—was added to the culture, either immediately after or 120 min after removal of the primary NaCl stress (Figure S1A). In both cases, thiolutin halted the decay of H2O2 resistance, indicating that active transcription/protein synthesis is not required to maintain the memory of H2O2 tolerance.
Because thiolutin also immediately inhibits cell growth, the above result suggests that the decay of H2O2 resistances requires cell division. To further test this, we measured H2O2 tolerance in NaCl-treated cells that were removed from stress but subsequently exposed to mating pheromone to arrest cell division. Indeed, the decay of the acquired H2O2 resistance was dramatically slowed when cell division was arrested (Figure S1B). Thus, cell division is required for the disappearance of H2O2 tolerance within the culture.
Memory of acquired stress resistance is inherited by daughter cells
Two models could explain the decay in H2O2 resistance in the cell culture. The first is a “marked” model, in which cells originally exposed to the mild primary stress become marked with H2O2 resistance and are slowly diluted out of the dividing culture. In this model, the original cells would retain maximal H2O2 resistance but disappear exponentially at the rate of cell division. However, several lines of evidence argue against this model. First, the decay of H2O2 resistance in the culture was slower than the measured rate of cell division (Figure 1B), indicating that exponential dilution of marked cells cannot account for the decay. Second, we observed that cells in the culture lost resistance to the highest doses of H2O2 earlier than they lost resistance to lower doses (see Figure 1A). This revealed quantitative loss of H2O2 resistance over time in individual cells, rather than dilution of cells permanently marked with high H2O2 tolerance.
Instead, we found that the H2O2 tolerance was distributed between mother and daughter cells upon cell division, such that new cells inherited H2O2 tolerance. We used a reporter construct developed and kindly provided by Lindstrom and Gottschling (2009) to track mother vs. daughter cells in the culture. Using this system, we found that the level and persistence of H2O2 resistance in daughter cells was indistinguishable from that in mother cells (R = 0.99, Figure S2). Together, these results show that cells inherit H2O2 resistance but that resistance diminishes with each cell division.
Mechanism of cellular memory of stress resistance
There are several possible mechanisms of cellular memory in yeast, including propagation of chromatin marks (Brickner 2009; Kundu and Peterson 2009; Kaufman and Rando 2010), inheritance of long-lived memory factors (Acar et al. 2005; Zacharioudakis et al. 2007; Kundu and Peterson 2010), and feedback in signaling pathways that can maintain a response once cellular memory is activated (Xiong and Ferrell 2003; Acar et al. 2005; Mettetal et al. 2008). We found no requirement for several chromatin factors implicated in expression memory, including the deacetylase SIR2, the Pho23p subunit of the Rpd3-large histone deacetylase complex, or the histone variant H2A.z (Q. Guan and A. P. Gasch, unpublished data). We also found no significant differences in transcript abundance after cells had re-acclimated to stress-free medium, nor evidence of a sequestered pool of nontranslated, stored mRNAs (Aragon et al. 2008) (data not shown). These negative results are consistent with the dispensability of nascent protein synthesis in maintaining the memory of H2O2 tolerance.
Instead, our results suggested that a long-lived memory factor was induced during NaCl treatment and distributed between dividing mother and daughter cells. We recently identified genes required for acquisition of H2O2 tolerance after pretreatment with several mild stressors (Berry et al. 2011). Cytosolic catalase Ctt1p, which converts H2O2 to oxygen and water, is critical for acquisition of H2O2 resistance immediately after mild NaCl treatment (Berry et al. 2011). We therefore reasoned that long-lived Ctt1 protein may be responsible for the memory of H2O2 tolerance. We first followed CTT1 mRNA and Ctt1p protein in a strain expressing genomically integrated FLAG-tagged CTT1 during NaCl treatment and as cells resumed stress-free growth (Figure 2). CTT1 mRNA is present at low levels before stress (Lipson et al. 2009) and highly but transiently induced after NaCl treatment before returning to basal levels by the time cells have been returned to stress-free medium. In contrast, Ctt1p protein reached maximum abundance after cells were removed from NaCl and remained at elevated levels in the actively dividing culture for >6 h after cells were removed from NaCl. This time frame correlated with the period of elevated H2O2 tolerance in the culture (Figure 1).
To directly test the role of Ctt1p in the memory of H2O2 resistance, we induced CTT1 expression in the absence of stress. CTT1 was deleted from the genome, and, instead, FLAG-tagged CTT1 was placed under control of the GAL1 promoter in a strain capable of estradiol-induced transcription (Louvion et al. 1993; Gao and Pinkham 2000). CTT1 expression was transiently induced by 1 μM estradiol treatment for 3 hr before cells were returned to plain medium. H2O2 tolerance increased upon estradiol induction (Figure 3A, red curve), but not after a similar treatment in the empty-vector control (data not shown), confirming that CTT1 induction is sufficient to increase H2O2 tolerance. Importantly, elevated H2O2 tolerance was detectable for well over 360 min after cells were removed from estradiol, during which time FLAG-tagged Ctt1p remained higher than in unstressed cells (Figure 3B). Adding NaCl during the estradiol induction did not further increase or prolong H2O2 tolerance in the culture (although cells were initially slightly more sensitive to H2O2 stress) (Figure 3A, gray curve). Thus, induction of Ctt1p is sufficient to explain the persistence of H2O2 tolerance after NaCl treatment.
Cells with a memory of acquired stress resistance show faster gene-expression changes upon further stress
Although nascent transcription was not required to maintain acquired H2O2 tolerance, we wondered if cellular memory of prior NaCl treatment extended to the transcriptional response to subsequent stress. Several studies have demonstrated transcriptional memory in cells previously exposed to galactose or inositol starvation, if cells are re-exposed to the nutrient shifts at a later time (Acar et al. 2005; Brickner et al. 2007; Kundu et al. 2007; Zacharioudakis et al. 2007). We therefore measured genomic expression in cells responding over the course of 60 min to a viable dose of 0.5 mM H2O2. We followed naive cells that had never before been treated with stress and also a culture that had been previously treated with 0.7 M NaCl for 60 min and then grown in stress-free medium for 240 min (two generations) before H2O2 treatment. Importantly, we saw no significant expression differences (at an FDR of 0.05) between the two cultures just before the H2O2 treatment (Figure 4A).
In contrast, the response to H2O2 was significantly different in cells with prior stress exposure (Figure 4A). There were 4341 genes whose expression was altered in naive cells responding to H2O2 (FDR < 0.05). Of these, 449 genes showed a statistically significant expression difference (FDR < 0.01) between naive and prestressed cells; relaxing the cutoff to FDR < 0.05 identified 1593 genes with an altered response, amounting to 37% of the H2O2-responsive genes. The affected genes included both induced and repressed genes that fell into distinct clusters. Nearly 90% of these genes showed expression changes during the NaCl pretreatment; however, of the 449 genes affected at the higher confidence level, 51 (11%) showed no significant expression change during NaCl treatment, revealing that the effect on expression extended to H2O2-specific expression changes. This result indicates that prior induction of affected genes was not a prerequisite for the altered gene-expression response in pretreated cells.
One possible mechanism behind the faster genomic response is that prior activation of the signaling system couldpromote faster reactivation, as shown for the galactose response (Zacharioudakis et al. 2007; Acar et al. 2008). However, in its simplest form this model cannot explain our results. Although we found significant enrichment among the 449 affected genes for targets of the H2O2-activated transcription factor Yap1p [P = 1.3 × 10−6 (Gasch et al. 2000)], the “general-stress” factor Msn2p [P = 1.1 × 10−4 (Berry et al. 2011)], and the Hog1p kinase [P = 4.1 × 10−7 (Berry et al. 2011)] that responds to osmotic shock, only about one-third of each factor’s targets showed significant expression differences between naive and prestressed cells (even if we relax the cutoff to FDR 0.05). Therefore, the altered gene-expression response cannot be simply due to faster activation of these factors in a manner that affects all of their targets similarly (see Discussion). We therefore investigated alternative possibilities.
Nuclear pore component Nup42p is required for the faster gene-expression response
Several studies have shown that a functional nuclear pore is required for transcriptional memory at specific genes due to tethering genes to the nuclear periphery and/or promoting transcriptional looping (Brickner et al. 2007; Tan-Wong et al. 2009; Ahmed et al. 2010; Light et al. 2010; Hampsey et al. 2011). To investigate the role of nuclear pore components, we followed gene expression in cells lacking Nup42p. Nup42p is required to target the INO1 gene to the nuclear periphery upon transcriptional induction (Ahmed et al. 2010; Light et al. 2010); it also has a role in mRNA export after extreme stress (e.g., a 25°–42° heat shock) but not under mild stress (25°–35° heat shock) or unstressed conditions (Stutz et al. 1995, 1997; Saavedra et al. 1997; Vainberg et al. 2000). We found that cells lacking NUP42 behaved like wild-type cells in several assays: they displayed no defect in basal H2O2 tolerance and no defect in the acquisition or memory of H2O2 tolerance after NaCl pretreatment (Figure S3A). Genomic expression analysis showed that these cells also had little significant difference in basal gene expression before stress (with only nine genes reproducibly altered more than twofold), nor any difference in the genomic response to a single dose of NaCl or H2O2 compared to wild-type cells (Figure 4B).
In contrast, nup42Δ cells had a major defect in the faster gene-expression response to repeated stress. We followed genomic expression in naive and pretreated nup42Δ cells responding to H2O2. In contrast to the pretreated wild-type strain, the nup42Δ mutant showed little difference in the H2O2 response (∼1.1-fold on average for the 449 genes affected in wild type) whether or not cells previously experienced NaCl stress (Figure 4B). Instead, both responses looked very similar to the H2O2 response of naive wild-type cells. The defect in nup42Δ cells was specific to successive stress treatments, since there was no difference in the response to a single dose of H2O2 or to the NaCl pretreatment. We confirmed the result using quantitative PCR at three representative genes (CTT1, TSA2, and HSP12; Figure 5 and Figure S4), which showed that the response of pretreated nup42Δ cells was superimposable with and statistically indistinguishable from the naive wild-type response (P > 0.05, t-test).
Nup42p has been implicated in mRNA export under severe-stress exposure, raising the possibility that a general defect in mRNA export prevents a normal expression response. However, several lines of evidence argue against this. First, Western blots showed a defect in production of heat-shock protein Ssa3/4p after a severe shock of 25°–42°, when mRNA export is known to be defective, but not in response to a mild 25°–35° heat shock or 0.7 M NaCl (Figure S3B), arguing against a gross mRNA-export defect under these conditions. Second, the mutant had no defect in acquiring H2O2 tolerance after a mild heat shock or 0.7 M NaCl, whereas it did have a defect after a severe 25°–42° shock (Figure S3C). Third, as discussed above, the nup42Δ cells showed no major expression defect before stress or after single stress treatments, which is not expected from a gross defect in mRNA transport after the initial stress treatment.
To further probe the role of NPC subunits, we measured the transcriptional response of TSA2 in several other NPC mutants. Nup59p is required for mRNA export after severe heat shock but is dispensable for peripheral localization of INO1 after inositol starvation (Thomsen et al. 2008; Ahmed et al. 2010; Light et al. 2010). We found that a nup59Δ had no defect in stress-responsive transcriptional memory at TSA2, unlike nup42Δ cells (Figure 5); a similar result was found for the related nup60Δ mutant (Thomsen et al. 2008) (not shown). This strongly argues against a general defect in mRNA export as the cause of the transcriptional difference. In contrast, cells lacking Nup100p, which is required for INO1 transcriptional memory and persistence of its peripheral maintenance after inositol starvation (Light et al. 2010), showed no difference in TSA2 induction in naive and pretreated cells. However, both the NaCl-treated and unstressed mutant cells showed kinetics similar to the pretreated wild-type cells with transcriptional memory, making the result difficult to interpret. It is intriguing to note that data from Ahmed et al. (2010) showed slightly increased INO1 peripheral localization in a nup100Δ mutant compared to wild-type cells immediately after the first exposure to inositol starvation (see Discussion).
Cells with memory show faster acquisition of a second round of stress resistance
The faster gene-expression response to subsequent stress could have a very important function in nature: faster reacquisition of stress tolerance upon repeated NaCl exposures. Indeed, we found this to be the case. Cells were treated with 0.7 M NaCl for 120 min, returned to stress-free medium for 240 min, and then exposed again to 0.7 M NaCl; H2O2 tolerance was assessed throughout the experiment (Figure 6A). Cells reached roughly the same level of H2O2 tolerance after both NaCl treatments (data not shown); however, during the second cycle of NaCl treatment, cells acquired H2O2 resistance faster: cells reached 30% maximum tolerance 10 min faster during the second cycle compared to the first NaCl exposure (Figure 6B). The acceleration of acquired H2O2 matched the acceleration in gene-expression response. Consistent with the defect in transcriptional memory, the nup42Δ mutant did not show faster reacquisition of stress tolerance upon repeated cycles of NaCl (P > 0.05, t-tests).
Many previous examples of cellular memory are associated with historical fluctuations in nutrient availability (Koshland 1977; Acar et al. 2005; Sigal et al. 2006; Brickner et al. 2007; Crews et al. 2007; Burns and Mery 2010; Carone et al. 2010; Hall et al. 2010; Ng et al. 2010). Our main goal here was to characterize yeast memory of osmotic shock. Cells transiently exposed to a viable dose of salt retain a memory of that treatment that extends to at least two physiological levels: increased H2O2 resistance and an altered expression response to subsequent stress. Both aspects of memory are likely inherited by daughter cells that have never directly experienced the NaCl stress. However, memory of H2O2 resistance and memory at the gene-expression level are dependent on distinct mechanisms.
In the case of H2O2 tolerance, the memory is explained by long-lived Ctt1 protein produced during the NaCl pretreatment and then transmitted to daughter cells with each cell division. The transmission of cellular memory via a long-lived protein is reminiscent of the cellular memory of galactose exposure, which is at least partly dependent on long-lived Gal1p inherited by daughter cells (Acar et al. 2005; Zacharioudakis et al. 2007; Halley et al. 2010; Kundu and Peterson 2010). Presumably, inheritance of Ctt1p allows cells to rapidly detoxify H2O2, thereby increasing H2O2 survival. The memory of acquired H2O2 tolerance is not specific to salt treatment because we observed increased H2O2 resistance after starvation (data not shown) and after a mild heat shock (Figure S5), although the mechanism is likely independent of CTT1 [since heat-induced acquisition of H2O2 tolerance does not involve CTT1 (Berry et al. 2011)]. Burrill and Silver (2011) recently showed that cells can also maintain a memory of DNA damage. Thus, memory of prior environmental stress may be a common phenomenon, but it is likely explained by different mechanisms under different situations.
Separable from the memory of H2O2 tolerance, cells with a history of NaCl treatment displayed a faster gene-expression response to H2O2, long after cells were removed from the salt. Over 1500 genes showed a faster expression response upon repeated stress, significantly more than the handful of metabolic genes known to demonstrate transcriptional memory. Furthermore, our results show that transcriptional memory can be invoked across distinct stressors. The altered expression response affected both induced and repressed genes and extended to targets of several independently acting regulators, including Hog1p, Msn2p, and Yap1p, indicating that the effect is unlikely due to a single regulatory system. Feedback in previously activated regulatory systems, including the GAL and HOG networks, can produce faster responses to recurring stimuli (Zacharioudakis et al. 2007; Acar et al. 2008; Mettetal et al. 2008). However, this alone is insufficient to explain our results. First, Hog1p is activated during NaCl treatment, but does not regulate gene expression in response to H2O2 (J. Clarke and A. P. Gasch, unpublished results). Second, many of the genes with a faster response in cells with memory are not regulated by Hog1p (O’Rourke and Herskowitz 2004), including targets of the H2O2-responsive Yap1p. Finally, despite the enrichment of Hog1p and Yap1p targets, only about one-third of each factor’s targets were affected; we could find no obvious or known difference in regulation of the affected subsets. Thus, faster activation of these networks cannot fully explain the faster genomic expression response.
Instead, our results suggest a role for nuclear pore components, including Nup42p. Nup42p could perhaps accelerate the expression response by facilitating mRNA export or transport of some other molecule; if so, this function must be restricted to successive stress treatments, since the nup42Δ mutant had no obvious defects before or after a single dose of stress. Nup42p could also facilitate association of the NPC with target genes. Several yeast genes associate with the nuclear pore upon induction (Brickner and Walter 2004; Casolari et al. 2004, 2005; Dieppois et al. 2006; Schmid et al. 2006; Taddei et al. 2006; Brickner et al. 2007; Sarma et al. 2007; Tan-Wong et al. 2009), which may promote transcriptional looping and/or couple transcription and mRNA export from the nucleus (Brickner 2009; Hampsey et al. 2011). Upon initial inositol starvation, INO1 translocation to the nuclear periphery is dependent on Nup42p (Ahmed et al. 2010; Light et al. 2010); while initial translocation is independent of Nup100p [but perhaps amplified in its absence (Ahmed et al. 2010)], Nup100p is required for persistent peripheral localization and transcriptional memory of INO1 after inositol repletion (Light et al. 2010). In our system, nup42Δ cells pretreated with NaCl behave like unstressed, naive wild-type cells, whereas cells lacking Nup100p display TSA2 induction reminiscent of the wild-type cells’ transcriptional memory. While further experiments will be required to dissect this function, these results suggest that gene localization to the NPC could be involved.
In support of this possibility, we identified several sequence motifs previously linked to transcriptional memory upstream of genes with a faster expression response after recurring stresses. Several DNA sequences, known as “DNA zip codes,” have been implicated in peripheral gene targeting and transcriptional memory (Ahmed et al. 2010; Light et al. 2010). One such motif required for initial gene targeting has been characterized upstream of TSA2, and we observed a similar sequence upstream of CTT1 (Figure 7A), both of which show a Nup42p-dependent expression effect after multiple stresses. A different zip code sequence, known as the Memory Recruitment Sequence (TCCTTCTTTCC), is required to maintain INO1 at the pore after stimulus has been removed and is important for transcriptional memory (Light et al. 2010). Strikingly, we identified a related motif (Figure 7B) enriched in the group of induced genes with transcriptional memory (cluster 3 from Figure 4A, P = 6 × 10−4). Although future work will be required to dissect the function of this sequence, these results are consistent with the model that association with the nuclear pore is required for the transcriptional memory at a large number of genes.
Faster activation of stress-dependent expression changes has an important physiological outcome: faster acquisition of a second round of H2O2 tolerance. The memory of prior stress treatment therefore extends to multiple physiological levels and arises through distinct mechanisms, suggesting its importance in nature. Life in the real world is particularly challenging for single-celled organisms that must maintain internal homeostasis despite an ever-changing environment. As stressful environments likely occur in succession, memory of prior exposure may provide a potent survival strategy in the wild.
We thank Derek Lindstrom, Dan Gottschling, David Eide, and Betty Craig for providing reagents and members of the Gasch Lab for constructive comments. This work was supported by a Beckman Young Investigator award to A.P.G. and National Institutes of Health National Institute of General Medical Sciences grant R01GM083989-01. D.G.B. was supported through the Univeristy of Wisconsin-Madison Integrated Biological Sciences Summer Research Program through National Library of Medicine training grant T15LM007359.
Communicating editor: M. D. Rose
- Received June 15, 2012.
- Accepted July 19, 2012.
- Copyright © 2012 by the Genetics Society of America