Decoupling yeast cell division and stress defense implicates mRNA repression in translational reallocation during stress

Summary

Stress tolerance and rapid growth are often competing interests in cells. Upon severe environmental stress, many organisms activate defense systems concurrent with growth arrest. There has been debate whether aspects of the stress-activated transcriptome are regulated by stress or an indirect byproduct of reduced proliferation. For example, stressed Saccharomyces cerevisiae cells mount a common gene expression program called the environmental stress response (ESR) [1], comprised of ~300 induced (iESR) transcripts involved in stress defense and ~600 reduced (rESR) mRNAs encoding ribosomal proteins (RPs) and ribosome biogenesis factors (RiBi) important for division. Since ESR activation also correlates with reduced growth rate in nutrient-restricted chemostats and prolonged G1 in slow-growing mutants, an alternate proposal is that the ESR is simply a consequence of reduced division [2–5]. A major challenge is that past studies did not separate effects of division arrest and stress defense; thus, the true responsiveness of the ESR – and the purpose of stress-dependent rESR repression in particular – remains unclear. Here we decoupled cell division from the stress response by following transcriptome, proteome, and polysome changes in arrested cells responding to acute stress. We show that the ESR cannot be explained by changes in growth rate or cell-cycle phase during stress acclimation. Instead, failure to repress rESR transcripts reduces polysome association of induced transcripts, delaying production of their proteins. Our results suggest that stressed cells alleviate competition for translation factors by removing mRNAs and ribosomes from the translating pool, directing translational capacity toward induced transcripts to accelerate protein production.

ETOC Blurb

The Environmental Stress Response (ESR) is a yeast transcriptomic response to diverse stresses. Ho et al. show that the ESR is not due to changes in division rate during stress acclimation. Rather, they propose that repressed transcripts help redirect translational capacity to induced transcripts to accelerate protein production.

RESULTS

ESR activation does not correlate with cell-cycle phase or division rate

We decoupled the response to acute stress from changes in proliferation by investigating cells arrested at different stages of the cell cycle. Cells were treated for >2 hours with either nocodazole or mating pheromone, which stably arrested 93% and 99% of cells at G2/M and G1 phases, respectively (Table S1). Although cell number plateaued over time, biomass and optical density (influenced by cell number and cell size/shape) increased but at a significantly reduced rate (Figures 1A–C). Much of biomass production is through protein accumulation [6]. Indeed, arrested cells showed reduced (but still active) global translation indicated by polysome profiles, with 2–3 fold higher monosome (‘M’) versus polysome (‘P’) abundance (Figure 1D) indicating reduced translation [7]. Thus, both nocodazole and alpha factor arrest division at G2/M or G1 and significantly reduce but do not fully arrest translation and biomass production.

Figure 1. ESR activation does not correlate with division or cell cycle phase.

Log-phase cells were arrested for 120 min and subsequent changes (versus the 120 time point) in (A) cell number, (B) optical density (OD) and (C) dry-cell weight were followed for asynchronously dividing cells (Asy, pink) or cells arrested with nocodazole (Noc, blue) or alpha factor (Alpha, orange). Average ± standard deviation is shown for biological triplicates. Biomass measurements were slightly compressed (see STAR Methods) but show retarded changes in arrested cultures. (D) Polysome profiles indicating 40S, 60S and 80S monosome (M) and polysome (P) peaks and area-under-curve M/P ratio is shown for cells arrested for denoted times. (E,F) Distribution of log₂(fold-change) in iESR, RiBi and RP transcripts in (E) arrested cells, dividing cells treated with 0.7M NaCl for 30 min, or (F) cells treated with 0.1 or 0.2 ug/mL cycloheximide (CHX) for 120 min, compared to unstressed, dividing cells (Asy). See also Figure S1 and Table S1.

If the ESR is simply a consequence of division arrest or prolonged G1 phase [2–5], then non-dividing cells should show strong but phase-specific ESR activation – however, this is not the case. Both nocodazole- and pheromone-arrested cells displayed mild ESR activation independent of arrest phase that was weaker than dividing cells exposed to salt stress (Figure 1E). Furthermore, reanalysis of cell-cycle transcription revealed no relationship between the ESR and G1 phase (Figure S1A), consistent with other recent work from our lab [8]. As another test, we measured ESR activation in slow-growing cells treated with low doses of cycloheximide, which reduces translation, division rate, and biomass production (Figures S1B and S1C). Although iESR transcripts were slightly higher in cycloheximide-treated cells, so too were many rESR transcripts, especially RPs (Figure 1F). While the increase in RP mRNAs may be a compensation for cycloheximide-induced translation defects [9], together these results show that rESR transcripts are not strictly correlated with growth rate and not related to cell cycle phase.

Arrested cells respond to stress with transient ESR activation

We next monitored transcriptome changes in arrested cells immediately after exposure to 0.7M NaCl or 37°C heat shock. At peak response times, E SR activation was highly correlated in dividing cells and cells arrested (Figure 2A) or slow growing (Figure S1D) (R² 0.7–0.9), with reduced slopes indicating a weaker response, likely due to mild ESR activation before stress. Furthermore, cells showed a dose-dependent response (Figure S1E). Thus, arrested cells can still mount a stress response, consistent with a prior study of slow-growing cultures [10].

Figure 2. Arrested cells mount an ESR response to stress.

(A) Plots show the correlation in log₂(fold-change) in ESR transcripts 30 min after 0.7M NaCl treatment or 10 min after 25°C to 37°C heat shock, in arrested cells versus dividing cells responding to stress. Cells were arrested for 2h before stress treatment, and fold-change was compared to each unstressed culture immediately before stress treatment. (B–C) Average log₂(fold change) in transcript abundance for iESR, RiBi and RP genes in dividing and arrested cells relative to cells collected immediately before exposure to (B) 0.7 M NaCl or (C) 25°C to 37°C heat shock. Data of asynchronously growing cells from [1, 11]. See also Figure S1.

Like many stress, NaCl triggers a largely transient transcriptome response whose peak changes coincide with the period of reduced division rate [11]. Surprisingly, ESR activation in arrested cells treated with stress was also transient in most cases (aside of salt-induced iESR after nocodazole arrest, Figures 2B and 2C). In particular, rESR transcripts dropped transiently before returning nearer to pre-stress levels for all stresses and arrest methods – even though cells had already arrested division and reduced biomass production before stress. As the cells clearly remained arrested throughout the experiment (Table S1), these results point to another purpose for rESR repression during stress acclimation.

Minor role for mRNA-specific translational regulation during arrest or stress treatment

We previously proposed that rESR repression alleviates competition for ribosomes during stress acclimation, thereby directing ribosomes to newly made transcripts [11]. We therefore investigated protein production under these conditions. We first used quantitative mass spectrometry to measure protein abundance and generated a mass-action kinetic model of translation, for dividing and arrested cells (see STAR Methods). The model allowed us to calculate translation rates, k_s,p, for each mRNA in the two conditions. Despite significant differences in both mRNA and protein abundance between these conditions (Figure S2A), there was high correlation between transcript-specific k_s,p values calculated for arrested and dividing cells (R² ~0.8, Figure S2B). The reduced slope (~0.85) is consistent with the higher M/P ratio (Figure 1D) and suggests globally reduced translation in arrested cells, whereas the high correlation in k_s,p values in arrested and dividing cells suggests only minor differences in transcript-specific translational rates. This is consistent with the model that alteration of protein abundance during steady-state growth is mainly determined by changes in transcript levels [12, 13], in accordance with the global translational capacity. Thus, while changes in proliferation rate cannot explain ESR activation during the dynamic response to acute stress, ESR transcript abundance during continuous, acclimated growth is likely set according to the global translation capacity and the demand for new protein during division (see Discussion).

To test these predictions, we sequenced mRNAs associated with polysome fractions (‘polysome-seq’) [14]. We sequenced 9 polysome fractions per profile, done in biological duplicate with high reproducibility (Figure S3A), for alpha factor-arrested cells and dividing cells before and 30 min after NaCl treatment (see STAR Methods). We compared the polysome distribution for each mRNA in arrested versus dividing cells, using spline-based analysis (see STAR Methods) or by comparing the transcript’s summed reads across polysome (‘P’) fractions normalized to its total reads across all fractions (‘T’) to account for mRNA abundance differences. The correlation between transcript-specific P/T values was as correlated for arrested and dividing cells as for replicate P/T profiles (R²= 0.86 versus R² ~ 0.88), and only 189 transcripts (3.4%) displayed significantly different polysome distribution (FDR < 0.05, Table S2). This is consistent with the modeling that suggests only subtle transcript-specific translational control in arrested versus proliferating cells.

We next investigated translational regulation 30 min after NaCl treatment, when dividing cells resume growth at a reduced rate [11, 14]. Although cells show a large increase in monosomes at this time point [7, 15, 16], we did not see a substantial increase in monosome-bound mRNA (Figure 3A). This indicates that stressed cells remove some ribosomes from the translating pool, as suggested in other recent studies [17–19]. At this time point after stress, there was substantially more evidence of translational regulation: 1855 transcripts displayed significantly different polysome distributions and/or P/T values (FDR < 0.05).

Figure 3. Translational changes in arrested and dividing cells responding to stress.

(A) Bulk polysome profiles of asynchronously growing cells before and 30 min after NaCl stress in blue. Total normalized read count (Millions of Reads) for each fraction indicated in pink. (B) Log₂(fold change) of transcript abundance (x-axis) and translation propensity (y-axis) in dividing cells responding to 0.7M NaCl for 30 min. Genes are classified and colored based on significant changes in mRNA abundance (FDR <0.05), translational propensity (q <0.05), or both. Enriched categories (Bonferroni-corrected p<0.05, hypergeometric distribution) are indicated. See also Figure S3, Table S3 and S4.

To quantify the relative contribution of changes in mRNA abundance versus translational regulation after stress, we classified mRNAs subject to altered abundance, translational propensity (P/T), or both (FDR <0.05, see STAR Methods). In general, there were larger changes in mRNA abundance (x-axis, Figure 3B) versus translational propensity (y-axis, Figure 3B), consistent with other studies [16, 20] and most altered transcripts showed little detectable change in translational propensity at this dose and timing. Nonetheless, there were discrete classes that deviated from the trend: transcripts encoding kinases and some iESR proteins increased in both mRNA and polysome association (Figure 3B, purple points), such that translational regulation reinforces mRNA increases as known for specific transcripts [15, 16, 20]. A corresponding group of decreasing mRNAs was reinforced by reduced polysome association and enriched for RiBi transcripts. But other mRNAs showed antagonistic control: many transcripts encoding cell-cycle regulators, transcription factors, and chromatin proteins were either unchanged or reduced in abundance but increased translational propensity. It is intriguing that many of these proteins affect transcription, yet may be regulated post-transcriptionally during stress.

Failure to repress rESR transcripts delays production of induced proteins

We previously proposed that the drop in rESR transcripts helps to redirect limiting ribosomes toward induced mRNAs as cells acclimate [11]. However, the large increase in apparently empty ribosomes (Figure 3A) indicates that ribosomes are not limiting cell growth or translation under these conditions. To further probe the role of rESR repression, we deleted RiBi repressors Dot6 and Tod6 [21, 22] and followed the mutant’s response to NaCl stress. Without stress, the dot6Δtod6Δ mutant grows normally with near-normal transcriptome [11] and bulk translation profiles (Figure S3B), but after NaCl it fails to repress many RiBi transcripts and does not acclimate to NaCl properly [11]. We performed polysome-seq in dividing dot6Δtod6Δ cells, before and 30 min after NaCl. The dot6Δtod6Δ cells failed to fully repress many RiBi transcripts after NaCl stress. Somewhat surprisingly, these and other transcripts actually increased in polysome association after stress in the mutant versus wild type (Figures S4A and S4B). For analysis, we scaled each mRNA’s abundance in each fraction to the total reads in that fraction, so that fold-change values before and after stress reflect the change in the proportion of that fraction comprised of each transcript.

The results show a striking effect at RiBi transcripts regulated by Dot6/Tod6. Whereas in wild-type cells these mRNAs are depleted from post-stress polysomes, in dot6Δtod6Δ cells the transcripts remain associated and comprise a greater proportion of polyribosome-associated mRNAs (Figure 4A, S4C). This was clear at individual transcripts including UTP15, although interestingly abundance of Utp15 (Figure 4C) or other proteins (Figure S4D) did not increase. In contrast to the RiBi transcripts, many iESR mRNAs were induced at near wild-type levels in the mutant but were less associated with polysomes, as seen for CTT1 mRNA (Figure 4B).

Figure 4. The dot6Δtod6Δ mutant shows aberrant translational regulation upon stress.

(A) The log₂(fold change) in mRNA in each of 9 polysome fractions from replicate wild type (WT) and dot6Δtod6Δ (dt) cells, for RiBi, RP, and iESR gene groups. Blue/yellow represents log₂ (fold change) in mRNA abundance normalized to the mean mRNA abundance in that fraction after NaCl treatment, whereas orange/purple indicates the difference (‘dif’) between WT and mutant. 80S is indicated by an arrow and p-values indicate significant differences in P/T for respective gene groups, see STAR Methods. (B) Median change in mRNA abundance for groups shown in (A). (C,D) Polysome profiles (left), changes in mRNA (qPCR, middle), and changes in protein (Western, right) for (C) UTP15 and (D) CTT1 (n=3). An asterisk indicates p<0.05, + indicates p<0.08, paired T-test. See also Figure S4.

Our model is that the drop in rESR transcripts helps to direct translational capacity to induced mRNAs. A key prediction is that failure to repress rESR transcripts should delay production of induced proteins. We followed mRNA and protein production of tagged Ctt1 in the wild type and dot6Δtod6Δ cells responding to NaCl (Figure 4D). The mutant showed similar induction of CTT1 mRNA in early time points (albeit slightly lower at the first measurement), but then overcompensated with more CTT1 than wild type. Despite the increase in CTT1 mRNA, the dot6Δtod6Δ mutant showed 2–3× less CTT1 mRNA in the polysome fractions and exhibited a significant delay in Ctt1 protein production. We analyzed prior proteomic data from our lab [23] and found that other iESR proteins showed delayed accumulation in the dot6Δtod6Δ mutant responding to NaCl, despite normal underlying mRNA changes (Figure S4E). These results suggest a model in which failure to repress rESR transcription in the dot6Δtod6Δ mutant causes delayed production of stress-induced proteins.

DISCUSSION

Our results address an ongoing debate regarding the purpose and responsiveness of the yeast ESR. On the one hand, ESR transcript abundance correlates with steady-state growth in nutrient limited chemostats. This relationship and the fact that rESR transcripts support ribosome production fits nicely with bacterial growth laws in which growth rate is often proportional to ribosome availability [19, 24–28]. On the other hand, acutely stressed cells typically show dramatic but in many cases transient changes in ESR mRNA abundance. The question had been if the dynamic response during stress is regulated or merely an indirect byproduct of reduced proliferation.

Here, we show conclusively that ESR activation after acute salt or heat stress cannot be explained as an indirect response to arrest or changes in division. Instead, we propose a different model: that transient rESR repression during acute stress accelerates production of new proteins, by redirecting translational capacity to induced mRNAs. At the same time, our results confirm that steady-state protein levels are largely determined by mRNA abundance and global translational capacity, which fits with the growth-rate model of rESR expression. How can we reconcile these models? In fact, both models can explain ESR transcript abundances but under different scenarios.

Differences in protein production are largely determined by differences in mRNA abundance in steady-state conditions

Actively dividing cells must supply the right suite of proteins for the emerging daughter. Both our modeling and polysome profiling suggest that, at a given translational capacity, differences in steady-state protein production are largely set by differences in steady-state mRNA abundance, consistent with other reports [12, 13]. Thus, rESR transcripts will correlate with growth rate as long as translational capacity remains unchanged. This explains the discordance between growth rate and rESR expression during cycloheximide treatment, which elevates many rESR transcripts despite slower growth (Figure 1E) [9, 28]. Thus, although rESR mRNA levels predict growth rate under some situations, the true relationship between the two requires modeling both mRNA and global translational capacity [29, 30].

Transient rESR repression during an acute response temporarily redirects translational capacity

rESR reduction has a different role during an active stress response, because the transient drop of rESR transcripts still occurs in already arrested cells. During active division, rESR transcripts are highly transcribed and translated to supply of ribosomes to new cells [31, 32]. In this case, most ribosomes are actively engaged in translation and the production of rESR proteins themselves [11, 14, 31, 33]. Immediately after NaCl treatment – much before detectable mRNA changes in bulk samples – cells dramatically inhibit global translation [11, 34]. This causes a large but transient spike in apparently empty ribosomes [7, 11, 18] that are removed from the translating pool. Thus, the number of ribosomes is unlikely to be limiting during this response [19], but rather the drop in active ribosomes may alleviate competition for other limiting translational factors.

Together, this and past work from our lab suggest a systems-wide model for how yeast acclimate to stress. We previously showed that the large but transient over-shoot in iESR transcripts accelerates production of the encoded proteins [11], likely by increasing chance encounters with translation machinery [12]. Here we propose that the transient drop in rESR transcripts, along with the temporary removal of translating ribosomes, further enhances this goal in two ways. First, temporarily reducing the load of rESR transcripts increases the proportion of iESR mRNAs in the translatable pool. Because the rate of translation initiation is heavily influenced by relative mRNA abundance [12], the combined effect should enhance iESR translation initiation. Second, by reducing both rESR mRNAs and available ribosomes the cell may alleviate competition for other translation machinery (such as tRNAs or other factors [35, 36]). mRNAs can be sequestered from the translated pool in stress granules [37], but our results show that Dot6/Tod6 targets are not marked for translational silencing but instead likely removed from the translated pool by transcriptional repression (although Dot6/Tod6 could have other, as yet unknown functions). Failure to repress these transcripts correlates with continued and even increased association of the transcripts with polysomes (Figure S4A), a phenotype that can result from elongation defects [38]. Interestingly, Dai et al. recently proposed that bacterial cells remove ribosomes from the active pool upon nutrient stress to maintain elongation rates. A similar effect occurs during osmotic stress, where the number of ribosomes per cell does not correlate with growth rate [28, 36]. Our model predicts that failure to remove rESR transcripts from the cell should delay production from induced transcripts, a prediction we validated (Figures 4 and S4C). These results demonstrate that aberrant mRNA-polysome associations can have widespread effects on cellular translation and stress acclimation.

Coordination of the iESR and rESR alleviates competition for cellular resources

Induction of iESR transcripts is anti-correlated with reduction of rESR mRNAs across a wide range of conditions, stress doses, and temporal profiles. The opposing expression changes during an acute response alleviates competition for translational machinery, but it may also do so for transcriptional machinery that can also be limiting during stress [39]. Acute stress triggers relocalization of RNA Polymerase (Pol II) from the heavily transcribed rESR genes to the stressinduced iESR genes [40–42], mediated in part by direct phosphorylation of the Pol II carboxyl-terminal domain by stress-activated kinases [42, 43] (Nemec et al. unpublished). Thus, regulated coordination of iESR induction and rESR repression may help to alleviate competition for multiple types of cellular resources, including machinery. It will be important to dissect how the models presented here for yeast pertain to other organisms, and th e extent to which these trends extend from free-living microbes to multicellular organisms.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Audrey Gasch (agaschwisc.edu).

EXPERIMENTAL MODEL AND SUBJEST DETAILS

All yeast strains are listed in Key Resources Table. All S. cerevisiae strains used were of the BY4741 background, including the previously published dot6Δtod6Δ strain [11]. Unless noted, BY4741 was grown >7 generations to an optical density (OD₆₀₀) ~ 0.25 at 30°C in batch culture in YPD (1% yeast extract, 2% peptone and 2% dextrose) medium. Cells were arrested with either 10 µg/ml nocodazole (EMD Millipore, Bilerica, MA) or 40 µg/ml alpha factor (ZymoResearch, Irvine, CA) for 2 or 3.5 hours, where indicated, before stress treatment. Asynchronous cells were grown > 7 generations to OD₆₀₀ ~ 0.6. Nocodazole and alpha factor concentrations were maintained during stress experiments despite media additions in some cases. AGY1363 (BY dot6Δtod6Δ CTT1-GFP) and AGY1397 (BY dot6Δtod6Δ UTP15-GFP) strains were generated by using the GFP-HIS3 construct from the yeast GFP-tagged library (ThermoFisher) integrated into the native CTT1 or UTP15 loci in the dot6Δtod6Δ strain and verified by diagnostic PCR.

Key Resources Table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal to GFP - ChIP Grade	abcam	Cat#AB290, RRID:AB_303395
Polyclonal Hog1 (yC-20) antibody	Santa Cruz Biotechnology	Cat#sc-6815, RRID:AB_650113
Anti-Actin Antibody, clone C4	EMD Millipore	Cat# MAB1501, RRID:AB_2223041
Chemicals, Peptides, and Recombinant Proteins
Nocodazole	EMD Millipore	Cat# 487928, CAS 31430-18-9
Alpha-Factor Mating Pheromone	Zymo Research	Cat# Y1001
TURBO DNase	Life Technologies	AM2239
Phenol/Chloroform/Isoamyl Alcohol, 25:24:1, pH 6.7/8.0	ThermoFisher	Cat# BP1752I-400
FailSafe PCR Enzyme Mix	Epicentre	FSES1100
Trypsin gold, Mass Spectrometry Grade	Promega	Cat#V5280
Critical Commercial Assays
LightCycler SYBR Green I Master	Roche	Cat#04707516001
ScriptSeq Complete Kit, Human/Mouse/Rat	Epicentre	BHMR1224
RNeasy Minielute Cleanup Kit	Qiagen	Cat# 74204
RNA Clean & Concentrator kit	ZymoResearch	Cat# R1015
TruSeq Stranded Total RNA library Prep Kit High Throughput	Illumina	Cat# RS-122-2203
Quantitative Colorimetric peptide assay	Pierce	Cat#23275
Deposited Data
RNA sequencing counts	this paper	GEO: GSE89554: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE89554
Raw proteomics data	this paper	Chorus: 1236: https://chorusproject.org/pages/dashboard.html#/projects/my/1236/experiments
Experimental Models: Organisms/Strains
S. cerevisiae  Strain background: BY4741	lab stock	N/A
S. cerevisiae: dot6Δtod6Δ	lab stock	N/A
AGY1363 (BY dot6Δtod6Δ CTT1-GFP)	This paper	N/A
AGY1397 (BY dot6Δtod6Δ UTP15-GFP)	This paper	N/A
AGY1357 (BY CTT1-GFP)	lab stock	N/A
Sz. Pombe PR100	lab stock	N/A
Software and Algorithms
Trimmomatic version 0.32	[45]	http://www.usadellab.org/cms/?page=trimmomatic
Bowtie2	[46]	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
HTseq version 0.6.1	[47]	https://htseq.readthedocs.io/en/release_0.10.0/
splineTimeR	[48]	https://bioconductor.org/packages/release/bioc/html/splineTimeR.html
edgeR	[50]	https://bioconductor.org/packages/release/bioc/html/edgeR.html
limma	[51]	http://bioconductor.org/packages/release/bioc/html/limma.html
Q-value	[52]	https://bioconductor.org/packages/release/bioc/html/qvalue.html
MaxQuant (v. 1.5.0.25)	[57]	http://www.coxdocs.org/doku.php?id=maxquant:start
Other
0.2µM Whatman Nylon Membrane	GE Healthcare	Cat#:7402-004
AxyPrep MAG PCR Cleanup beads	Axygen Scientific	Cat#:MAGPCRCL50
Acrylic Desiccator Cabinets	ThermoFisher	Cat#08-642-23C

METHOD DETAILS

Cell measurements and RNA preparation

Cell density at each time point was counted on a Guava flow cytometer (Millipore). Dry-cell weight was measured at each time point by filtering 12 mL of culture for arrest experiments or 90 mL (1.8 × 10⁸ –1.8 × 10⁹ cells) for cycloheximide treatment onto an overnight-dehydrated and subsequently microwaved (10 min 10% power) 0.2µM Whatman Nylon Membrane (GE Healthcare catalog number 7402-004) that had been weighed. Cells plus filter were dehydrated in a microwave (10 min 10% power), dried overnight in an Acrylic Desiccator Cabinet (Fisher 08-642-23C), and weighed. Dry cell weight was the difference in weight before and after filtering cells. Accurate measurement of dry cell weight requires large cell volumes; values here are compressed somewhat, especially for arrest treatments which used smaller culture volumes due to reagent costs. The figures show that biomass continues to increase after arrest even though cell number does not, which is consistent with our observation of continued but reduced protein translation (see Figures 1D and S1C).

For RNA collections, cells were harvested before arrest, at various times after arrest treatment, and at denoted times after stress by brief centrifugation then flash frozen and stored at −80 °C. For heat shock experiments, cells were grown in log phase at 25°C, collected by brief centrifugation and resuspended in 37°C medium for d enoted times. For experiments normalized by Sz. pombe RNA spike-in, equal numbers of cells (estimated by manual counting and/or OD₆₀₀) were mixed with a defined number of a single batch of Sz. pombe cells before RNA extraction. Total RNA was extracted with hot phenol [44], DNase-treated at 37 °C for 30 min with TURBO DNase (Life Technologies), and precipitated with 2.5 M LiCl for 30 min. rRNA depletion was carried out with ScriptSeq Complete Kit H/M/R (Epicentre, Madison, WI), and samples were purified with a RNeasy MinElute Cleanup Kit (Qiagen).

Polysome profiling and RNA isolation

Polysome profiles were collected as described previously with minor modifications [14]. Briefly, 35 ml of yeast culture were harvested by spin at 3000 r.p.m. for 3 min at 4°C in the presence of 0.1 mg/ml cycloheximide. Cells were washed 2× with cold lysis buffer (20 mM Tris-HCl at pH 8, 140 mM KCl, 5 mM MgCl₂, 0.5 mM DTT, 0.1 mg/ml cycloheximide, 1 mg/ml heparin and 1% Triton X-100) and lysed by vortexing with glass beads in cold lysis buffer. Crude extract was centrifuged and 10 OD₂₆₀ U of resulting supernatants were applied to 4 ml linear 5–50% continuous sucrose gradients in 20 mM Tris-HCl at pH 8, 140 mM KCl, 5 mM MgCl₂, 0.5 mM DTT, 0.1 mg/ml cycloheximide and spun for 120 min at 45,000 r.p.m. in a SW50.1 rotor.

Samples were pooled based on A₂₆₀ peaks (splitting fractions at the trough of each peak) and equal mass of Sz. pombe total RNA was spiked into each pool for future normalization. The mass was chosen by calculating a 1:20 ratio of Sz. pombe : S. cerevisiae RNA used as input for the polysome gradient; that value was divided by nine since there were nine fractions per profile. Pooled and spiked RNA fractions were adjusted with 2.5× volume of water (to dilute sucrose in high polysome fractions), extracted twice with 25 : 24 : 1 phenol:chroloform:isoamyl alcohol (ThermoFisher), and ethanol precipitated. DNase-treated RNA was purified on RNA Clean & Concentrator kit (ZymoResearch, Irvine, CA) and rRNA depleted with the TruSeq Stranded Total RNA library Prep Kit High Throughput (Illumina).

RNA-Seq

Libraries were generated with Index PCR Primers (Illumina for polysome samples and Epicentre for others) and FailSafe PCR Enzyme Mix (Epicentre, Madison, WI), purified on AxyPrep MAG PCR Cleanup beads (Corning, Corning, NY), and sequenced on Illumina's HiSeq 2500 and MiSeq System (UW-Madison DNA Sequencing Facility), generating single-end 100 bp reads. Reads were processed with Trimmomatic version 0.32 [45] and mapped to the S288c v64 S. cerevisiae genome and ASM294v2.25 Sz. pombe genome using Bowtie2 [46]. Gene-level counts were taken from HTseq version 0.6.1 [47]. Data were normalized by RPKM unless noted, or spike-in normalization by setting the slope of Sz. pombe reads across samples to 1.0. All sequencing data are available at in the GEO database under accession number GSE89554.

Mass Spectrometry

Yeast peptides were prepared and analyzed according to the published methodology [54–56]. In brief, yeast pellets were resuspended in lysis buffer consisting of 8 M urea, 100 mM Tris (pH 8), 10 mM tris(2-carboxyethyl)phosphine (TCEP), and 40 mM chloroacetamide (CAA). Cells were lysed by the addition of methanol to 90%, followed by vortexing for ~30 seconds, and protein precipitation through centrifugation at 10,000 × g for 15 minutes. The resultant pellets were resuspended in the lysis buffer and digested overnight at room temperature with trypsin (Promega, Madison WI) at 1:50 enzyme to protein ratio. Samples were desalted using Strata-X Polymeric Reversed Phase 10 mg/1 mL columns (Phenomenex, Torrance CA). Peptide concentration was measured using a Quantitative Colorimetric peptide assay (Pierce, Rockford IL) according to the manufacturer’s instructions.

2 µg of tryptic yeast peptides in 0.2% formic acid (FA) were injected on to a reverse phase column prepared in-house and separated over a 120-min gradient at flow rate of 300 nl/min using Dionex UltiMate 3000 nanoLC (Thermo, Sunnyvale CA). Mobile phase buffer A was composed of 0.2% FA in water, mobile phase B was 70% acetonitrile, 0.2% FA, and 5% DMSO. Peptide cations were converted to gas-phase ions by electrospray ionization and analyzed on a Thermo Orbitrap Fusion (Thermo, San Jose CA). The raw data were processed using MaxQuant (v. 1.5.0.25) [57]. Searches were performed against a target-decoy database of reviewed yeast proteins plus isoforms (Uniprot, downloaded January 20, 2013) using the Andromeda search algorithm. All raw proteomics data files were deposited into Chorus (ID 1236).

Mathematical modeling

Mathematic modeling was done similarly to previously described [11] using a mass-action kinetic model in which: where

$$\frac{d}{\mathit{\text{dt}}}[P_{\rho }]=k_{s,p}[{\mathit{\text{mRNA}}}_{\rho }]-(k_{d,\rho }+\mu )[P_{\rho }]$$

where k_s,p is the translation rate for protein ρ and k_d,p is the protein degradation rate [58], μ represents the dilution rate due to cell division, and [P_ρ] represents the concentration of the protein at time t. We assumed there is no change in protein abundance over time under steady state, so that $\frac{\mathit{d}}{\mathit{\text{dt}}}[{\mathit{P}}_{\mathit{\rho }}]=\mathbf{0}$.

Thus,

$$k_{s,p}=\frac{(k_{d,\rho }+\mu )[P_{\rho }]}{[{\mathit{\text{mRNA}}}_{\rho }]}$$

where μ is assumed to be 0 for cell cycle arrested cells and is based on a 90 min doubling time for dividing cells. Absolute mRNA and protein abundance per cell was taken from [59] and [60], respectively. For arrested cells, absolute abundance was scaled based on the average of our triplicate measured differences in mRNA and protein abundance compared to dividing cells.

Western blots and qPCR

BY4741 CTT1-GFP and AGY1363 (BY dot6Δtod6Δ CTT1-GFP) were grown as described above and exposed to 0.7M NaCl, and samples were taken over time as cells acclimated. Ctt1-GFP was monitored by Western blot, loading OD-normalized cells in sample buffer and using anti-GFP (Abcam) and anti-Act1 (EMD Millipore) or anti-Hog1 (Santa Cruz Biotech.) as a loading control, using a Licor Odyssey Infrared Imager. Results were very similar whether data were normalized by cell count, Act1, or Hog1 abundance; because some of the Utp15-GFP blots showed inconsistent Act1 bands, Hog1 was used for normalization in this case (results were the same when normalized by collected cell number). mRNA abundance over time was measured by quantitative RT-PCR similar to previously described [11], using a Roche LightCycler 480 II and Roche LightCycler 480 SYBR Green I Master. C_T values for each mRNA were measured in technical triplicate for each of three biological replicates. Values were internally normalized to ERV25 quantification and the ratio of before versus after stress calculated.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data analysis of polysome samples

Polysome-seq was normalized by spiked-in RNA, setting the slope of Sz. pombe reads to 1.0. One fraction in each of two profiles (5R in the unstressed profiles) was clearly off compared to the other replicates; for these two samples, we scaled the reads such that the mean reads in that fraction equaled the mean of the reads from the two flanking fractions, which substantially improved agreement across replicates. To compare polysome profiles across samples, mRNA reads in each fraction were first divided by the total reads for that mRNA across all fractions in that profile and logged. These normalized profiles were compared using the package splineTimeR [48] with 5 degrees of freedom as recommended and FDR < 0.05 as significant (See Table S2) [49]. Changes in mRNA abundance for Figure 3 were assessed using edgeR [50] and changes in translation propensity were identified using limma [51] and Q-value [52] with q < 0.05 taken as significant. Functional enrichments and gene groups in Figure 3 are available in Table S3 and Table S4.. Trends and enrichment results were similar if analyzed by edgeR, in which summed reads across polysome fractions and free/monosome fractions were compared before and after stress after setting the library sizes to be equal for each pair of libraries [53]; however, limma results were used in the figure since some edgeR calls were dubious (not shown). Data for Figure 4 were normalized by dividing the reads for each transcript in a given fraction by the mean of all transcript reads in that fraction. Subsequently, the log₂ ratio comparing stressed versus unstressed values for that fraction was calculated. P-values in Figure 4 were calculated based on the sample-mean centered data, comparing transcripts in designated groups versus all other transcripts (RiBi transcripts were removed from the RP and iESR comparisons to ensure they were not driving the significance). P-values were similar if calculated on the read data without mean centering, revealing that the significance in Figure 4 was not an artifact of the normalization.

DATA AND SOFTWARE AVALIABILITY

All RNA sequencing data reported in this paper have been deposited in the GEO under ID codes GSE89554: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE89554 All raw proteomics data files have been deposited into Chorus under ID code 1236: https://chorusproject.org/pages/dashboard.html#/projects/my/1236/experiments

Supplementary Material

2

Table S2. Transcripts identified by spline analysis. Related to STAR Methods. Tabs show transcripts with significant (FDR < 0.05) changes in polysome association comparing arrested versus dividing cells, dividing cells before and after NaCl treatment, and mutant versus wild-type cells responding to NaCl treatment.

3

Table S3. GO functional enrichment of gene groups from Figure 3.

4

Table S4. Transcripts in each category shown in Figure 3.

Highlights.

The yeast ESR is not a response to growth or cell-cycle arrest during stress

Failure to repress mRNAs during stress delays translation of induced transcripts

Repression of rESR transcripts likely redirects translational capacity during stress

mRNA changes serve different roles during continuous growth versus an acute response