Yeast Metabolic and Signaling Genes Are Required for Heat-Shock Survival

Home , Chemostat

Yeast metabolic and signaling genes are required for PNAS PLUS heat-shock survival and have little overlap with the heat-induced genes

Patrick A. Gibneya,b,1, Charles Lub,1, Amy A. Caudya,2,3, David C. Hessc,3, and David Botsteina,b,4

aLewis-Sigler Institute for Integrative Genomics and bDepartment of Molecular Biology, Princeton University, Princeton, NJ 08544; and cDepartment of Biology, Santa Clara University, Santa Clara, CA 95053

Contributed by David Botstein, September 30, 2013 (sent for review August 25, 2013) Genome-wide gene-expression studies have shown that hundreds or proteins that increase in abundance as part of the heat-shock of yeast genes are induced or repressed transiently by changes in response (1, 11–14). Classically, this set of genes/proteins includes temperature; many are annotated to stress response on this basis. the molecular chaperones, which often are induced in response to To obtain a genome-scale assessment of which genes are function- stress (14). However, by using the deletion collection to assay ally important for innate and/or acquired thermotolerance, we thermotolerance, we also can identify genes important for stress combined the use of a barcoded pool of ∼4,800 nonessential, pro- resistance that are either constitutively present or posttransla- totrophic Saccharomyces cerevisiae deletion strains with Illumina- tionally modified. We therefore carried out two kinds of exper- based deep-sequencing technology. As reported in other recent iment to examine heat-stress resistance. First we studied the 50 °C studies that have used deletion mutants to study stress responses, survival time-course of samples from a steady-state chemostat cul- we observed that gene deletions resulting in the highest thermo- ture of the entire prototrophic barcoded deletion collection. Sec- sensitivity generally are not the same as those transcriptionally ond, we tested for the acquisition of thermotolerance by shifting fi induced in response to heat stress. Functional analysis of identi ed the temperature of the chemostat to 37 °C, a temperature known genes revealed that metabolism, cellular signaling, and chromatin to suffice for induction of the heat-stress–associated gene set, regulation play roles in regulating thermotolerance and in acquired before doing another 50 °C survival time-course. Results of these thermotolerance. However, for most of the genes identified, the experiments are summarized in Table 1. molecular mechanism behind this action remains unclear. In fact, a In the first experiment, we identified a surprisingly small num- large fraction of identified genes are annotated as having unknown ber of deletion strains (55 gene deletions) that differ significantly functions, further underscoring our incomplete understanding of the response to heat shock. We suggest that survival after heat from the rest of the population in their heat sensitivity (Tables 2 shock depends on a small number of genes that function in assess- and 3). Most of these strains are much more sensitive to heat ing the metabolic health of the cell and/or regulate its growth in shock than the rest of the population; only a few were more a changing environment. resistant. We found that there was very little overlap between

this set of genes and the much larger number of genes identified SYSTEMS BIOLOGY in other recent studies (5–8) as being induced as part of the heat- eat stress has been the subject of hundreds of studies, many Hof which have sought to understand the role of single genes fi or gene families in mediating the response to heat shock, whereas Signi cance others have attempted to examine system-level responses to heat (1, 2). Two basic approaches have been used to study the global We used the model eukaryote Saccharomyces cerevisiae to cellular response to heat or other stresses. One approach follows investigate which genes are important for survival of heat the response to temperature shift at a genomic scale using gene- stress. Previously, this question was addressed by examining expression microarrays to define the set of genes whose expres- which genes are turned on by mild heat stress; in this study, we sion is altered (3, 4). This approach necessarily limits the stress examined gene-deletion mutants for increased sensitivity to to temperatures at which the cells remain viable and express their lethal heat stress. This approach reveals that these two sets genes. The other approach uses single- or pooled-deletion strains of genes are largely nonoverlapping, demonstrating that mu- to assess the relative fitness of individuals alone or within a popu- tant analysis is a powerful complementary approach to gene- lation undergoing heat stress; this approach allows the use of lethal expression analysis. In addition, many of the genes identified heat to assess the ability to survive (5–8). as important for heat survival are involved in metabolism or We studied heat-stress resistance at a genomic scale further signaling, or their function is completely uncharacterized, sug- by introducing a number of experimental refinements. First, we gesting that our understanding of the systems-level response to used cultures growing at steady state before the heat stress to heat stress is incomplete. minimize effects of growth conditions. We had shown previously that the growth rate of cells has a profound effect on their ability Author contributions: P.A.G., C.L., A.A.C., D.C.H., and D.B. designed research; P.A.G., C.L., A.A.C., and D.C.H. performed research; P.A.G., C.L., A.A.C., and D.C.H. contributed new to survive a lethal heat shock (9): Slow-growing cells are much reagents/analytical tools; P.A.G., C.L., A.A.C., D.C.H., and D.B. analyzed data; and P.A.G., less sensitive than cells growing rapidly. By using the chemostat, C.L., A.A.C., D.C.H., and D.B. wrote the paper. variation in growth rate was minimized. Second, we used a newly The authors declare no conflict of interest. constructed prototrophic, barcoded nonessential deletion col- Freely available online through the PNAS open access option. lection to avoid the unwanted survival behaviors of auxotrophic 1P.A.G. and C.L. contributed equally to this work. mutations present in the standard, commercially available de- 2Present address: Banting and Best Department of Medical Research, Terrence Donnelly letion collection (7). Third, we used next-generation sequencing Centre for Cellular and Biomolecular Research, University of Toronto, Toronto, ON, and our previously developed analysis algorithms to follow in- Canada M5S 3E1. dividual barcodes and thereby quantitatively assess relative sur- 3A.A.C. and D.C.H. contributed equally to this work. vival during heat stress (7, 10). 4To whom correspondence should be addressed. E-mail: [email protected]. Both thermotolerance and acquired thermotolerance have This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. been studied in detail, with a general focus on the roles of transcripts 1073/pnas.1318100110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1318100110 PNAS | Published online October 28, 2013 | E4393–E4402 Downloaded by guest on October 1, 2021 Table 1. Phenotypic categories of heat-sensitive mutants 30–50 °C phenotype 37–50 °C phenotype Description

Sensitive (P < 0.01) Sensitive (P < 0.01) Strains are hypersensitive to heat when shifted from 30–50 °C and do not acquire thermotolerance after pretreatment at 37 °C (10 genes; Table 2) Sensitive (P < 0.01) Resistant (P value for Strains are hypersensitive to heat when shifted from 30–50 °C and do acquire sensitivity is > 0.01) thermotolerance after pretreatment at 37 °C (49 genes; Table 3) Wild-type (P value for Sensitive (P < 0.05) Strains exhibit sensitivity to heat similar to wild type when shifted from 30–50 °C but exhibit sensitivity is > 0.05) hypersensitivity to lethal heat shock after pretreatment at 37 °C (nine genes; Table 4)

shock response. Interestingly, many of the 55 genes we identified resource without this drawback, we generated a prototrophic as conferring hypersensitivity to heat are functionally involved deletion collection consisting of 4,783 viable deletion strains (see with cellular signaling, chromatin regulation, or carbohydrate Materials and Methods for details). This prototrophic deletion metabolism, suggesting unknown and potentially interesting collection has a number of useful features. First, each deletion is connections between heat shock and these processes. In the marked with the same barcode used in the original yeast deletion second experiment (pretreatment at 37 °C), we found only 10 collection. Thus, array-based methods for measuring deletion deletion strains that show significantly lower rates of survival frequency in pooled samples of the deletion collection will work than the pooled population as a whole (Table 2). These strains with this set. Second, the presence of the magic marker (can1Δ:: generally appear to be a subset of the 55 genes found in the first STE2pr-SpHIS5) allows the addition of another reporter (e.g., experiment, although the data for a few of them fell below our CLONAT-resistance) using the synthetic genetic array strategy (16, stringent threshold. We also found a third set of deletions (nine 17). Third, each deletion is marked with the G418-resistance cas- strains) whose survival in the first experiment (30–50 °C) was sette, as in the original deletion collection, allowing the user to select typical of the rest of the population but whose thermotolerance for G418-resistance when manipulating the deletion collection and was reduced significantly (by the same criteria used above) in the thus reducing potential contaminants. Fourth, unlike the original second experiment (Table 4). deletion collection, we included 12 isogenic wild-type controls on In this study we found only a few genes whose absence had a every 96-well plate as controls for position-related variance. major effect on survival, and many of these genes have metabolic and/or signaling functions. We suppose that survival of lethal Experimental Design of the Heat-Shock–Survival Screens. To identify heat shock depends on the products of these genes to signal the genes required for survival of lethal heat stress, we screened metabolic health of cells and/or regulate growth processes so that nonessential, prototrophic deletion strains after they had been the cell is protected against future lethal events and/or to allow grown separately and pooled together. Phenotypic fitness for cells that have experienced a severe stress to resume growth when each strain was assessed using barcode sequences to measure the environment once again is stable. changes in strain abundance during heat treatment at 50 °C over a 15-min period (Fig. 1). Many studies have quantified barcode Results abundance in mixed populations using DNA array-based meth- Construction of a Prototrophic, Nonessential Gene-Deletion Collection. ods, and more recent studies, including ours, have used deep- Yeasts exist in the wild as prototrophic organisms, capable of de sequencing technology (5–8, 10, 18, 19). novo synthesis of all their amino acids and nucleotides. Auxo- It is important to have an estimate of how well this protocol trophic markers were included in the original yeast deletion preserved the abundances of the deletion mutants at the begin- collection to allow further genetic manipulation of the deletion ning of the heat shock, especially because we know that slow collection strains (5). These auxotrophic mutations sometimes growth rate induces thermotolerance (9). To this end, we mea- have unexpected phenotypes that can interfere with experi- sured the relative frequency of all of the barcodes before the heat ments, notably those involving nutrient starvation and potentially shock for both experiments (30–50 °C and 37–50 °C). It is clear other stresses also (7, 15). To produce a prototrophic deletion that there was an excellent correlation between the rank abundance

Table 2. Strains with signiﬁcant* death rates in response to heat shock with or without pretreatment at 37 °C Calculated death Calculated death rate for the 30 rate for the 37 ORF Name Description –50 °C experiment –50 °C experiment

YGR240C PFK1 Alpha subunit of phosphofructokinase −28.98 −29.39 YDR074W TPS2 Trehalose-6-phosphate phosphatase −24.72 −24.29 YKR029C SET3 Member of the SET3 histone deacetylase complex −8.05 −11.07 YJL175W YJL175W Dubious ORF (overlaps SWI3) −15.68 −9.61 YLL026W HSP104 Hsp100-type molecular chaperone −14.82 −9.21 YLR285W NNT1 Putative nicotinamide N-methyltransferase, has −2.94† −7.83 a role in rDNA silencing and in lifespan determination † YOR043W WHI2 Regulates general stress response with Psr1 −3.54 −7.02 YLR296W YLR296W Dubious ORF −14.66‡ −7.01 YBR077C SLM4 Component of the EGO complex, which is involved in −6.97‡ −6.7 the regulation of macroautophagy YKR017C YKR017C Putative protein of unknown function; contains RING ﬁnger motif −11.19 −6.24

*P < 0.01. †Death rates are very near the P < 0.05 cutoff. ‡Death rates are negative, but cfu at t = 0 are not sufﬁcient for conﬁdence about the calculated changes.

E4394 | www.pnas.org/cgi/doi/10.1073/pnas.1318100110 Gibney et al. Downloaded by guest on October 1, 2021 Table 3. Strains with signiﬁcant* death rates in response to heat shock (30–50 °C) that acquire thermotolerance after pretreatment PNAS PLUS at 37 °C Calculated death rate for Calculated death rate for ORF Name Description the 30 –50 °C experiment the 37 –50 °C experiment

† YDR017C KCS1 Inositol hexakisphosphate (IP6) and inositol −18.26 heptakisphosphate (IP7) kinase YBL064C PRX1 Mitochondrial peroxiredoxin −17.23 1.62 YKR055W RHO4 Nonessential small GTPase of the Rho/Rac −12.66 −4.57 subfamily of Ras-like proteins YGL229C SAP4 Required for Sit4 protein phosphatase function −11.88 −4.28 YNL307C MCK1 Serine/threonine/tyrosine kinase −11.73 −4.09 YER039C HVG1 Protein of unknown function; has homology to Vrg4 −11.5 1.25 YGL062W PYC1 Pyruvate carboxylase −11.41 −4.18 YOR235W IRC13 Dubious ORF (overlaps SNR17A) −11.19 −2.24 YPR154W PIN3 Induces appearance of [PIN+] prion when overexpressed −10.88 0.92 YCL022C YCL022C Dubious ORF (overlaps KCC4) −10.84 −4.75 YGR022C YGR022C Dubious ORF (overlaps MTL1) −10.77 −1.4 YPL197C YPL197C Dubious ORF (overlaps RPB7B) −10.62 −1.55 YKR018C YKR018C Putative protein of unknown function −10.44 0.88 YOR039W CKB2 Beta subunit of casein kinase 2 −9.97 −2.09 YJL162C JJJ2 Protein of unknown function; contains a J-domain −9.72 −0.34 YLL001W DNM1 Dynamin-related GTPase that regulates −9.71 −0.78 peroxisomal abundance YHR204W MNL1 Exomannosidase involved in ubiquitin −9.66 0.88 proteasome degradation YLR057W YLR057W Putative protein of unknown function −9.52 −1.28 YKL076C PSY1 Dubious ORF (overlaps YKL075C) −9.51 −1.29 YCR005C CIT2 Citrate synthase −9.26 −1.43 YGR033C TIM21 Translocase of the inner mitochondrial membranes −8.48 −3.63 YGL194C HOS2 Histone deacetylase −8.43 −4.95 YLL044W YLL044W Dubious ORF (overlaps RPL8B) −8.29 0.63 YDR139C RUB1 Ubiquitin-like protein −8.19 0.24 YJR070C LIA1 Deoxyhypusine hydroxylase −7.86 −1.93 YPR012W YPR012W Dubious ORF −7.81 −4.73 YNL067W RPL9B Protein component of the 60S ribosome −7.69 −2.11 SYSTEMS BIOLOGY YJL135W YJL135W Dubious ORF (overlaps LCB3) −7.56 −2.98 YDR269C YDR269C Dubious ORF (overlaps CCC2) −7.51 −0.22 YPL064C CWC27 Component of a pre-mRNA splicing complex −7.47 −3.03 YER182W FMP10 Putative protein of unknown function −7.45 −3.47 YNL056W OCA2 Putative tyrosine phosphatase; similar to Oca1 −7.33 −3.14 YIL064W SEE1 Protein with role in intracellular transport −7.28 −5.89 YFL033C RIM15 Protein kinase involved in signal transduction −6.92 −2.5 during stationary phase establishment YCL024W KCC4 Protein kinase involved in septin checkpoint −6.91 0.96 YEL010W YEL010W Dubious ORF −6.89 −2.98 YPR195C YPR195C Dubious ORF −6.75 −1.19 YLR176C RFX1 Transcription repressor involved in DNA damage −6.57 −1.69 YDR340W YDR340W Putative protein of unknown function −6.46 −0.18 YBR031W RPL4A Component of the 60S ribosome −6.36 −1.16 YER177W BMH1 14–3-3 protein; controls proteome at −6.3 −4.08 posttranscriptional level YIR027C DAL1 Allantoinase −6.29 −1.89 YML107C PML39 Protein required for nuclear retention of −6.25 −0.11 unspliced pre-mRNAs YER072W VTC1 Vacuolar transporter chaperone −6.23 −0.66 YOL153C YOL153C Pseudogene −6.19 0.24 YFR010W UBP6 Ubiquitin protease in 26S proteasome −6.18 −5.26 YBL081W YBL081W Protein of unknown function −6.18 0.13 YOR170W YOR170W Dubious ORF (overlaps LCB4) −6.18 0.83 YLR392C ART10 Protein of unknown function −6.13 0.96

*P < 0.01. †There were too few colonies at 0 min to quantify death rate accurately.

of all of the barcodes between experiments (SI Appendix, Fig. S1). deletions we identiﬁed as more sensitive to heat appear to be The pooled strains grew about 10 generations before this mea- spread throughout the distribution of strain abundances, sug- surement was made, and SI Appendix, Fig. S1 shows that the gesting that they did not grow particularly faster or slower. We

Gibney et al. PNAS | Published online October 28, 2013 | E4395 Downloaded by guest on October 1, 2021 Table 4. Nine strains with normal sensitivity to heat shock (30–50 °C) that develop signiﬁcant sensitivity* after pretreatment at 37 °C Calculated death rate Calculated death rate for the 30 –50 °C for the 37 –50 °C ORF Name Description experiment experiment

YMR126C DLT1 Protein of unknown function, mutant sensitive to 6-azauracil (6AU) 0.47 −6.11 and mycophenolic acid (MPA) YDL020C RPN4 Transcription factor that stimulates expression of proteasome genes, 2.44 −4.14 transcriptionally regulated by various stress responses YIL090W ICE2 Integral ER membrane protein with type-III transmembrane domains; −0.54 −4.97 required for maintenance of ER zinc homeostasis YNL122C YNL122C Putative protein of unknown function 0.64 −3.79 YER056C FCY2 Purine-cytosine permease, mediates purine (adenine, guanine, −1.55 −5.76 and hypoxanthine) and cytosine accumulation YLR328W NMA1 Nicotinic acid mononucleotide adenylyltransferase −0.14 −4.31 YIL071C PCI8 Possible shared subunit of Cop9 signalosome (CSN) and eIF3 −2.51 −5.69 YOR328W PDR10 ATP-binding cassette (ABC) transporter, multidrug transporter involved −1.76 −4.81 in the pleiotropic drug resistance network YPL026C SKS1 Putative serine/threonine protein kinase; involved in the adaptation to −0.57 −3.59 low concentrations of glucose independent of the SNF3 regulated pathway

*P < 0.05.

confirmed directly that the growth rates of some of these strains To assess relative fitness, we used measurements of each do not differ significantly from that of the population, and thus barcode’s fractional frequency within the population over time differences in fitness are not caused by differences in growth rate (referred hereafter to as “barcode frequency”) rather than ab- (SI Appendix, Fig. S2). solute barcode counts (which vary depending on the number of reads produced by the instrument at each time point). Although we were able to detect ∼3,900 strains in our population, we limited our analysis to strains that had a significant number of A t = 0 minutes t = n minutes cfus before heat shock to detect changes in barcode frequency reliably. Therefore, our analysis of this dataset was restricted to Grow cells at desired 2,954 deletion strains. Among strains filtered out are some temperature, shiŌ to strains that potentially are highly sensitive (78 strains) or re- 50 C, collect samples sistant (37 strains) to heat shock, although we lack the statistical and plate 1 x 107 CFUs power to confirm this possibility using our data because of low Pooled collecƟon colony numbers at time 0 (SI Appendix, Table S1). of ~4,800 barcoded, Extract genomic DNA, amplify/sequence A Surprisingly Small Number of Genes Appear to Protect Cells from prototrophic barcodes, idenƟfy viable strains, and infer Lethal Heat Stress. Fig. 2A shows plots of each strain’s log - deleƟon strains populaƟon dynamics 2 transformed changes in barcode frequency over time relative to B its own pre–heat-shock value. It is clear that most of the population 100 changes very little in frequency over time. Linear regression was performed for each strain to calculate a slope associated with strain abundance, hereafter referred to as “death rate.” To estimate significance, a null distribution was calculated using 10 a bootstrapping method (Materials and Methods). The distribution of all death rates centers near zero, suggesting that most

Survival (percent) Survival = no 37°C pre-treatment genes do not play a role in modulating sensitivity to lethal heat =37°C pre-treatment (1 hour) stress (Fig. 2B). For P < 0.05, 175 strains exhibit significantly 1 higher heat sensitivity than the rest of the population, and 33 024681012 14 16 strains exhibit significantly higher heat resistance. For P < 0.01, Time at 50°C (minutes) only 55 strains exhibit significantly higher heat sensitivity than the rest of the population, and only three strains exhibit signifi- Fig. 1. Experimental approach. (A) A pooled collection of prototrophic deletion strains was inoculated into a chemostat, allowed to grow in batch cant higher heat resistance. The 55 heat-sensitive strains are culture for ∼8 h at 30 °C, and then grown in continuous culture for approxi- listed in Tables 2 and 3. mately three generations to reach a steady state. Then cells either were pre- Among the genes we identified whose deletion confers high treated at 37 °C for 1 h or not. In either case, cells were divided into 1-mL levels of sensitivity to 50 °C heat shock were tps2Δ and hsp104Δ. aliquots and were heat shocked at 50 °C for 3, 6, 9, 12, or 15 min. For each time These genes encode the only copies of yeast Tps2 (trehalose-6- point, cells were plated onto YPD, and at least 1 × 107 cells were collected (SI phosphate phosphatase) and the cytosolic Hsp100-type chaper- Appendix,TableS6). Genomic DNA was extracted; then barcode regions were one protein, Hsp104 (heat-shock protein 104), respectively. Both fi ampli ed by PCR and sequenced using an Illumina Genome Analyzer II. Bar- genes have already been well documented as important for sur- code reads were mapped back to individual genes to calculate changes in Δ vival of high temperature stress, providing validation of this strain abundances throughout the time-course. (B)Wild-type(ho ::kanMX) – fi cells were grown to early log phase in minimal medium and were pretreated technique (20 23). In addition, a number of identi ed strains at 37 °C for 1 h or not. Cells then were heat shocked at 50 °C for 3, 6, 9, 12, or contain gene deletions that were not expected. These include 15 min, and dilutions were plated to calculate cfus at each time point. pfk1Δ, kcs1Δ, prx1Δ, yjl175wΔ, and pyc1Δ. These genes have diverse

E4396 | www.pnas.org/cgi/doi/10.1073/pnas.1318100110 Gibney et al. Downloaded by guest on October 1, 2021 PNAS PLUS ) A 2 2 used this data for comparison (9). Because we analyzed only a subset of the genome in our survivability experiment, our gene- 0 expression comparison comprises only the 2,954 genes for which we have expression and survival data. The gene-expression data for mild heat shock were organized by hierarchical clustering using -2 Pearson correlation as the distance metric, and the data for

on fold change (log change Ɵ on fold changes in barcode frequency were organized in the same gene -4 order (SI Appendix,Fig.S4). The maximum fold induction of gene- expression values also was plotted against calculated death rates -6 (Fig. 3). It is clear that most of the genes whose expression is sig- Barcode frac Barcode 0 3 6 9 12 15 nificantly induced by heat stress display little or no change in death Time at 50°C (minutes) B rate, and vice versa (5–7, 24). 900 To illustrate the striking lack of correspondence, we also limited our analysis to the genes identified as induced or repressed in the 800 response to environmental stress (3). Heat-map representations 700 demonstrate that a strong gene-expression response does not correlate with a strong death-rate response for either induced or 600 repressed genes (Fig. 4). We also compared death rates with gene- expression patterns of many known heat-shock proteins and again 500 find no strong correlation (SI Appendix, Fig. S5 and Table S2). It is 400 important to note that many chaperone proteins are encoded in the genome in two or more redundant copies, and in these cases Number of Genes Number of 300 strains with single-gene deletions may not suffer functional loss of 200 these protein activities.

100 Screen for Deletion Mutants Unable to Acquire Thermotolerance After Pretreatment with a Sublethal Heat Shock. The phenome- 0 -30 -20 -10 010non of acquired thermotolerance, in which cells are protected Death Rate from a lethal heat stress via pretreatment by sublethal heat stress, has been well characterized (11, 25). We sought to identify Fig. 2. A small fraction of gene-deletion strains exhibits extreme thermosensitivity (30–50 °C). (A) Barcode fraction fold change for each deletion

mutant is plotted as a log2-transformed change compared with t = 0 in gray. Wild type (hoΔ::kanMX) is shown in red (HO is the site-speciﬁc endonuclease 10 used in mating type switching and is nonfunctional in most laboratory wild-

type strains). (B) Death rates for 2,954 deletion strains were calculated using 5 SYSTEMS BIOLOGY SPG4 ALD3 linear regression of the curves shown in A and are represented in this his- YER067W YBR116C ﬁ togram. Dashed green lines indicate rates that are signi cantly different from 0 HSP42 CTT1 SSA4 BTN2 HSP30 the mean based on the boot-strapping method (rate cut-offs ± 6.13, P < 0.01). -5 roles, including glucose metabolism, inositol-phosphate phos-

on death rate on death -10 phorylation (with known signaling roles), mitochondrial oxygen Ɵ

radical reduction, and a dubious ORF whose deletion overlaps -15 YJL175W HSP104 SWI3 PRX1 with the chromatin remodeling factor . Furthermore, 23 of KCS1 these 55 genes are annotated to the Gene Ontology (GO) term Gene dele -20 “biological process unknown.” It is not yet clear why these and other identiﬁed genes affect survival at high temperatures. A -25 TPS2

sample of gene deletions exhibiting increased heat sensitivity PFK1 fi -30 was reconstructed and tested again to con rm the screen-based -20 0 20 40 60 80 100 120 SI Appendix phenotype ( , Fig. S3). Gene expression maximal fold change We observed 33 resistant strains using a P value cutoff of 0.05 but only three resistant strains with the more stringent 0.01 Fig. 3. Deletion strains most sensitive to lethal heat shock (30–50 °C) do not cutoff. No GO terms were enriched in either set of genes, al- correlate with specific gene-expression responses to a mild heat shock (shift – though, again, many genes were annotated to “biological process from 28 36 °C). Comparison of maximal fold change in gene expression and ” death rates for cognate gene-deletion strains. Some of the most severely unknown (9 of 33 strains). affected genes are labeled. Dashed green lines indicate death rates significantly different from the mean based on a boot-strapping method (rate cut- Genes Important for Protection Against Lethal Heat Stress Do Not offs ± 6.13, P < 0.01). Death rate was calculated as described in Materials and Correlate with Genes Induced in Response to Heat Stress. Surpris- Methods. The following genes are highlighted as strongly induced by heat: ingly, our screen to identify genes that modulate survival of lethal YBR116C (dubious ORF, overlaps transketolase TKL2), HSP30 (molecular heat stress identified a relatively small number of genes. In con- chaperone), BTN2 (v-SNARE binding protein), ALD3 (cytoplasmic aldehyde trast, gene-expression studies have identified hundreds of genes dehydrogenase), SPG4 (uncharacterized stationary phase gene), SSA4 that are activated in response to heat stress (3, 4, 9). Given these (molecular chaperone), CTT1 (cytosolic catalase), YER067W (uncharacterized results, we were interested in comparing the set of genes important protein induced by respiratory conditions, also called RGI1), and HSP42 (molecular chaperone). The following genes are highlighted as exhibiting for survival of lethal heat stress with those genes whose expression extreme sensitivity to lethal heat stress: PFK1 (phosphofructokinase), TPS2 changes in response to heat. We previously had measured the gene- (trehalose biosynthesis), KCS1 (inositol hexakisphosphate kinase), PRX1 expression response to a mild heat pulse (28–36 °C) in a continuous (mitochondrial peroxiredoxin), and YJL175W (dubious ORF, overlaps chro- culture system (phosphate-limited, growth rate = 0.15/h), and we matin remodeler SWI3), HSP104 (molecular chaperone).

Gibney et al. PNAS | Published online October 28, 2013 | E4397 Downloaded by guest on October 1, 2021 A ESR Induced (202 genes) B ESR Repressed (135 genes) We also observed a set of nine genes that exhibited typical heat sensitivity when shifted from 30–50 °C but were hypersen- 28 C - 36 C 30 C - 50 C 28 C - 36 C 30 C - 50 C Expression Frequency RaƟo Expression Frequency RaƟo sitive to heat stress after pretreatment at 37 °C (Table 4). The genes in this set may be important for sensing and/or signaling of heat stress. Among this set of genes, a number have described roles in gene expression and gene-expression–related signaling, and one of the uncharacterized genes is sensitive to mycophenolic acid and 6-azauracil, hallmarks of a role in gene transcription. ) ) 2 2 Additionally, this set contains three genes involved in membrane transport, one gene involved in metabolism, and one completely uncharacterized gene.

Carbohydrate Metabolism and Thermotolerance. Prominent among Fold Change Change (log Fold Change (log Fold the genes identified as important for surviving lethal heat stress are PYC1 (pyruvate carboxylase), PFK1 (phosphofructokinse), and CIT2 (citrate synthase), all of which are involved in central carbon metabolism. The precise role of these genes in modulating sensitivity to heat is unclear currently and will be the subject of future analysis. TPS2 (trehalose-6-phosphate phosphatase) also is involved Color Scale Color Scale -3 3 -3 3 in sugar metabolism. Trehalose metabolism has been well described as a determinant for stress survival, including heat (20, Fig. 4. Deletion strains most sensitive to lethal heat shock (30–50 °C) do not 26–28). However, multiple roles have been ascribed to trehalose cluster with changes in gene expression induced or repressed by the envi- in mediating heat protection, and a detailed mechanistic role has ronmental stress response. (A) Comparison of environmental stress response not been completely determined. Other deletion strains in the (ESR)-induced changes in gene expression in response to a mild heat shock trehalose biosynthesis pathway (TPS1, TPS3, and TSL1) did not with cognate death rates. (B) Comparison of ESR-repressed changes in gene exhibit significantly negative death rates, although there may be expression in response to a mild heat shock with cognate death rates. trivial explanations for this observation. First, TPS3 and TPS1 are paralogs deriving from the yeast ancestral whole-genome duplication (29). Only the double deletion has a strong reported deletion strains unable to acquire thermotolerance because of phenotype (30–32). Further, deletion of TPS1 results in death on either a lack of upstream sensing/signaling defects or elimination glucose-containing medium in the W303 background and se- of genes that are primarily responsible for mediating thermo- verely impaired growth in the S288C background (30). Therefore tolerance. The experimental procedure was identical to that used we suspected that the TPS1-deletion strain present in the de- when screening for mutants that fail to withstand lethal heat letion collection, which has been passaged in glucose-based me- stress, except in this case cells were pretreated for 1 h at 37 °C dium, has acquired one or more suppressor mutations. We have before shifting to 50 °C (Fig. 1). A comparison of the distribution confirmed this assertion by reconstructing a TPS1 deletion and of death rates in both screens reveals that pretreatment at 37 °C comparing its growth with that of the deletion collection strain results in a narrower distribution (SD of 30–50 °C death rate (Fig. 5A). Further examination of the tps1Δ in a clean back- distribution = 2.26; SD of 37–50 °C death rate distribution = ground demonstrates that it does exhibit thermosensitivity sim- 1.38). This result suggests that the great majority of the pop- ilar to tps2Δ (Fig. 5 B and C). This thermosensitivity appears to ulation benefits from the pretreatment and experiences similar be unrelated to cellular trehalose levels, however, because in- levels of thermotolerance. creased trehalose levels in a neutral trehalase deletion (nth1Δ) After filtering strains with low numbers of colony forming units do not result in increased thermotolerance, as shown and pre- before the 50 °C heat shock, we analyzed data for 3,160 strains viously suggested for different types of stress (Fig. 5 B and C) with a single-gene deletion. Again, some of the strains that were (33). At this point none of the standard explanations sufficiently removed may truly exhibit heat sensitivity (20 strains) or re- explains the phenotypes of these mutants. sistance (17 strains), but our data lack the statistical power to Discussion determine this possibility (SI Appendix, Table S3). We again fi calculated death rates for each of the examined strains and found The rst and most surprising result of our genomic screen for deletion mutants hypersensitive to lethal heat exposure (50 °C) is that a relatively small number of the examined genes promote fl acquired thermotolerance. We also again used the boot-strap- that only about 50 genes have a substantial in uence on survival. fi fi Among the most sensitive mutants are deletions of genes with ping method to de ne a null distribution and statistical signi - PFK1 KCS1 P < known metabolic and/or signaling functions [e.g., , cance for each death rate. For 0.05, 65 strains exhibited PYC1 TPS2 CIT2 CKB2 fi (inositol hexakisphosphate kinase), , , , signi cantly higher heat sensitivity than the rest of the pop- RHO4 fi (casein kinase 2 subunit), and (small GTPase of rho/rac ulation, but no strains exhibited signi cant higher heat re- PFK1 P < fi subfamily of ras-like proteins)]. Some of these, such as , are sistance. For 0.01, 10 strains exhibited signi cantly higher likely to have both metabolic and signaling functions. Although heat sensitivity than the rest of the population (Table 2). Of the preheating of cultures to a nonlethal high temperature before 10 strains that have the highest death rates, six also have sta- the lethal heat stress induces thermotolerance in wild type and fi tistically signi cant death rates without pretreatment. Among most of the deletion mutants, 10 of the most hypersensitive the remaining genes, two did not have enough colonies at time deletions (including deletions of PFK1, TPS2, and HSP104)do 0 to assess death rate confidently (although both appear more not show this induced thermotolerance. sensitive than wild type); the other two were just below the cutoff The second interesting result is that most of these heat-sensitive to be considered highly significant (P < 0.01). This result suggests gene deletions are not among the several hundreds of genes that that the highly significant genes identified in the acquired are induced by nonlethal heat stress. This very large set of yeast thermotolerance screen are simply a subset of genes from the genes has, because of its heat inducibility, often been annotated thermotolerance screen. as being involved in heat-stress response and, by implication, in

E4398 | www.pnas.org/cgi/doi/10.1073/pnas.1318100110 Gibney et al. Downloaded by guest on October 1, 2021 A significant increases in death rate for these deletions in our PNAS PLUS screen. It is interesting that we did find WHI2, a gene reported to regulate MSN2 and MSN4, among the heat-hypersensitive deletion mutants (38). We also observed a number of other genes with reported roles in transcriptional and chromatin regulation, further underscoring a connection to cellular signaling and regulation of gene expression. The lack of correlation between genes involved in survival and B 100 25 DBY12000 those induced by stress may mean simply that the changes in tps2Δ ) 20 gene expression caused by exposure to sublethal stress exposure 10 nth1Δ 15 serve a preparative role rather than an immediate protective 1 role, as recently proposed by Berry and Gasch (24, 39). In our 10 experiment, we screened in a way that specifically targeted the

0.1 (mM Trehalose immediate protective genes.

Survival (percent) Survival 5 Only three gene deletions resulted in highly significant re- 0.01 0 sistance to heat stress (P < 0.01). These genes are ALG8 (glu- 0246 8101214 16 Log phase VPH1 YNB+Dex cosyl transferase), (subunit of vacuolar ATPase), and minutes at 50°C YNL140C (a dubious ORF that overlaps THO2, which is involved with transcriptional elongation). At the surface, these 35 C 100 genes appear to be completely unrelated, and it is unclear why 30 any would result in increased thermotolerance. It is worth noting 10 ) 25 that in all three cases the death rates were very close to the 20 defined threshold, suggesting that this resistance is not as ex- 1 15 treme as is the sensitivity of the gene deletions we identified. fi DBY12000 10 Interestingly, a number of genes that were ltered out of the data

0.1 (mM Trehalose

Survival (percent) Survival tps1Δ 5 because of low numbers of colony forming units before heat nth1Δ shock had large positive death-rate values (SI Appendix, Tables 0.01 0 0 246 8101214 16 Log phase S1 and S3). Some of these genes have been annotated as re- minutes at 50°C YNB+Gal sistant to heat stress (EAP1 and ALG6, for example), but because these strains were poorly represented at time zero, we are Fig. 5. Trehalose metabolism gene deletions exhibit thermosensitivity un- unable to quantify their changes throughout the time-course Δ related to trehalose levels. (A) Wild-type (DBY12000) and a number of tps1 reliably. Further work is required to validate and understand the mutants were spotted onto YPD or YP galactose and were incubated at 30 °C basis of heat resistance in these mutants. for 3 d. tps1Δ is the strain constructed de novo as part of this study; PG Zakrzewska et al. (8) also recently described similar experi- tps1Δproto is a prototrophic derivative (described in Materials and Methods) of the OpenBiosystems collection strain (tps1ΔOpenBio). (B) The indicated ments using barcode sequencing to identify genes important for

strains were grown to early logarithmic phase in yeast nitrogen base (YNB) + thermotolerance and acquired thermotolerance, among a num- SYSTEMS BIOLOGY glucose medium at 30 °C and then were submitted to a 50 °C heat shock. ber of other stressors. However, they did not discuss the specific Survival was monitored over time (Left), and intracellular trehalose content roles of the genes they found, nor did they emphasize the lack of was assessed before heat shock (Right). Note that the same color scheme is correlation with the set of genes that is induced by heat stress. used for strain identification in the left and right panels. (C) The indicated Zakrzewska et al. confirmed our previous observation that growth strains were grown to early logarithmic phase in YNB + galactose medium at rate is inversely proportional to stress tolerance, although we show 30 °C and then were submitted to a 50 °C heat shock. Survival was monitored here that growth rate is not the main determinant of mutant over time (Left), and intracellular trehalose content was assessed before heat SI Appendix shock (Right). Note that the same color scheme is used for strain identifi- survival ( , Figs. S1 and S2) (9). Although some cate- cation in the left and right panels. Genes highlighted in this figure are TPS1 gories of genes important for stress survival and acquisition of (trehalose-6-phosphate synthase), TPS2 (trehalose-6-phosphate phospha- thermotolerance are shared between their data and ours, some are tase), and NTH1 (neutral trehalase). different. Variations in experimental design, including different genetic backgrounds or different media conditions used for the screen, could explain these differences. More importantly, heat-stress survival. Our results suggest that this attribution may our death-rate data were based on a series of samples taken not be correct or, at least, that the number of genes important throughout a time-course, whereas Zakrzewska et al. examine for survival of heat stress has been underestimated. This dis- differences between time zero and a single end point, with cordance between gene induction and genes required for survival consequent degradation of the ability to assess statistical signif- has been seen a number of times (5–7, 24). icance. Finally, we reconstructed and individually tested several The remaining interesting result is the detection of a very deletions identified in this screen and in each case confirmed small subset of deletion mutations that result in exacerbation of their heat-sensitivity phenotypes (Fig. 5 and SI Appendix,Fig.S3). heat sensitivity when given a prior mild heat stress, completely opposite to what happens to wild-type cells and most of the other Metabolism, Growth Rate, and Heat-Stress Survival. The promi- deletion mutations we were able to follow. nence of genes with known metabolic and/or signaling functions We also found some expected results, the most important of requires an explanation, especially if one accepts the textbook which is that deletions of HSP104 and TPS2, both of which are view that heat sensitivity is substantially the result of protein implicated in cellular protein chaperone activity, result in hy- stability and refolding issues. Without entirely discounting this persensitivity to heat stress, as well documented previously (21, view, an alternative view would be that the response to heat 22, 34, 35). It should be recalled in this context that many other stress (or to other stresses) may be related even more strongly to chaperone proteins have many paralogs and overlapping func- the way in which cells assess and control their growth rates, tions, so our experiment does not speak directly to the need for a possibility that is more compatible with the spectrum of de- chaperones for heat-stress survival (14, 36, 37). Similarly, the letion mutants we found. In this broader view, one might suppose realities that HSF1 is essential and that MSN2 and MSN4 have that the genes we found act as sensors or regulators of cellular overlapping functions may well account for our failure to see health and readiness for growth. In this view, Hsp104 would

Gibney et al. PNAS | Published online October 28, 2013 | E4399 Downloaded by guest on October 1, 2021 serve to assess/regulate protein folding; Pfk1, Pyc1, and Tps2 inherit the auxotrophic markers (leu2Δ0 met15Δ0 ura3Δ0) from the original would serve to assess/regulate glycolytic flux; Kcs1 would assess/ deletion collection. Finally, to ensure that the resultant strains were pro- regulate the state of cellular lipids; and Prx1 would assess/regu- totrophic, the strains were selected on minimal medium with G418. After late the burden of reactive oxygen species. this step the strains were archived in rich medium glycerol stocks with G418. Speculating further, any or all of these sensors/regulators serve Because of the methodology used to construct the set (namely utilization of the magic marker strain and selection for prototrophic strains), a few to trigger a rapid cessation of growth in response to heat stress, strains, in addition to all auxotrophic mutations, are missing from this col- and the characteristic changes in gene expression (sometimes lection (SI Appendix, Table S4). The 39 deletions that cause catastrophic called the “environmental stress response”) associated with failures in mating and/or sporulation are not included in this set. Another 41 sublethal heat shock and a great variety of other stresses would auxotrophic deletions dropped out of our selection scheme. Twenty-one result from the decrease in growth rate, as proposed previously deletion strains showed sensitivity to the combination of drugs (G418, can- (40, 41). Exposure of growing cells to 50 °C is a potentially lethal avanine, and thialysine) used for selection after sporulation. Three strains heat shock, but such a shock is readily survivable by cells that are were removed because they had been found to be incorrect in the original growing slowly or cells that have been pretreated with heat stress deletion collection. Last, 100 deletion mutations failed to come through our at a lower temperature. The temperature of 50 °C is inimical to selections on multiple attempts for a variety of reasons, the most common being the deletion causing a severe fitness defect. de novo mRNA production or protein synthesis (SI Appendix, Fig. S6), so in this case survival would depend upon the ability of Yeast Strains and Culture Methods. The prototrophic yeast deletion collection these sensors to stop growth safely even in the absence of any strains described above were used for both mutant screens conducted as part − gene expression. For instance, it may turn out that glycolysis is of this study (MATa can1Δ::STE2pr-SpHIS5 lyp1 his3 ho ). As a surrogate for halted by Pfk1, which cannot happen if Pfk1 is absent. Similarly, wild-type behavior, the hoΔ::kanMX strain from the deletion collection was failure to be able to phosphorylate inositol hexakisphosphate used, although all strains have nonfunctional HO alleles containing the could result in the same type of defect. It remains to be de- standard amino acid replacements found in S288C-derived strains (T189A, termined whether these putative sensors act in concert. Such G223S, L405S, H475L) (44, 45). Data for the analysis of changes in gene ex- – a model accounts nicely for the many reports of cross-resistance pression as the result of mild heat shock (28 36 °C) were acquired using a prototrophic, CEN.PK derivative (DBY11092 MATa MAL2-8C) as part of to stresses by exposure to other stresses, the many observations a previous study (9, 46). For follow-up experiments, gene deletions were of all kinds of stress resistance in stationary-phase cells, and the reconstructed in a prototrophic, diploid background and then were sporu- many reports of low-growth-rate phenotypes in survivors (or lated and dissected to obtain haploid progeny. Strain genotypes were con- persisters) of culture under lethal stress (6, 9, 24, 39, 42, 43). firmed by PCR (for an example with tps1Δ,seeSI Appendix, Fig. S8). The Another aspect of heat-stress survival that deserves mention is haploid wild type of this strain used for growth comparisons was DBY12000 the resumption of growth after the stress has subsided or been (a prototrophic, HAP1-repaired version of FY4). Cell plating for the screen removed. All our assays were assays of viability on agar plates. was performed on rich YPD medium (20 g/L bacto peptone, 10 g/L yeast To form a colony, a heat-stressed cell that ceased growing extract, 20 g/L bacto agar, 20 g/L glucose). somehow must “reboot” its growth program. It could be that some of the genes we identified are involved in this aspect of Continuous Culture. Chemostat medium was prepared as detailed in SI Ap- pendix, Table S5, essentially as described in Brauer et al. (40). Continuous survival instead of growth cessation; potentially, any of them cultures were grown in 500-mL chemostat vessels at a final volume of 300 mL could be involved in both aspects. For example, glycolysis must in phosphate-limited medium (10 mg/L potassium phosphate). The dilution be restarted, which might be impossible in the absence of Pfk1 or rate, which is equivalent to the population growth rate at steady state, was Pyc1. One experiment was performed to test this possibility: set at 0.15/h (4.6 h per doubling). The dilution rate was estimated by direct allowing heat-shocked pfk1Δ cells to form colonies on yeast ex- measurement of the effluent volume over time. Cultures were stirred at 400 tract peptone dextrose (YPD) or yeast extract peptone glycerol rpm and sparged with 5 L/min of humidified, filtered air. Temperature ethanol (YPGE) medium (nonfermentable carbon sources of probes were calibrated with an electronic thermometer and were used to glycerol and ethanol) (SI Appendix, Fig. S7). Although pfk1Δ monitor the culture temperature throughout the experiment. Sixfors chemostat vessels and instrumentation were purchased from Infors. thermosensitivity comparable to that in wild type was not re- stored by growth on nonfermentable carbon sources, it is possi- Screen for Heat-Sensitive and Acquired Thermotolerance-Defective Mutants. To ble that Pfk1 plays some role in general resumption of growth, screen for heat-sensitive mutants, an aliquot of the pooled, prototrophic regardless of the carbon source available. deletion collection (∼1 × 108 cells) was inoculated into a chemostat vessel Taken together, these results provide a solid foundation for containing 300 mL of medium. This culture was grown in batch-growth the future elucidation of the mechanics of heat-shock resistance mode (no fresh medium was pumped into the system) at 30 °C for 8 h. Next, and indicate that central carbon metabolism and signaling pro- the pumps were turned on, and the culture was allowed to grow for 15 h teins have a fundamental role in this process. Finally, our results (approximately three doublings) to achieve steady-state growth. After emphasize the conclusions of other recent studies concerning the achieving steady state, 1-mL aliquots were removed from the culture into ∼ × 8 importance of using deletion screening as a complementary ap- microcentrifuge tubes (each containing 2 10 cells). For the heat shock, tubes were immersed in a 50 °C water bath for 0, 3, 6, 9, 12, or 15 min and proach to gene-expression analysis to determine the biological then were chilled on ice for the remaining duration of the time-course. (We functions of genes or the gene complement required for bi- found that ∼2 min are required for 1 mL of 30 °C culture to reach 50 °C in the ological phenomena (5–8). water bath.) To identify viable strains, cells were plated on rich medium (YPD) so that colonies were not confluent. Enough plates were prepared to Materials and Methods ensure collection of at least 1 million separate colonies from each time point Construction of a Barcoded, Nonessential Yeast Deletion Collection. The strains (SI Appendix, Table S6). To screen for mutants with acquired defective in the deletion collection (MATa yfgΔ0::kanMX his3Δ1 leu2Δ0 met15Δ0 thermotolerance, all steps were performed exactly as described as above, ura3Δ0) were mated to strain ACY742 (MATα can1Δ::STE2pr-SpHIS5 his3Δ1 with the addition of a 1-h incubation at 37 °C in the chemostat vessel before lyp1Δ0) creating diploids (because the STE2 gene promoter is MATa-specific, the 50 °C heat shock. Datasets for both experiments are available for the SpHIS5 gene will be expressed only in MATa cells). Two rounds of diploid download (Datasets S1 and S2). selection on minimal plates with histidine and G418 were performed with the resulting diploids pinned onto sporulation medium. Following sporula- Assessing Comparative Fitness of Barcoded Deletion Strains. Cells for each time tion, the first round of haploid selection was on synthetic complete medium point were collected, and genomic DNA was isolated using a standard glass without histidine, arginine, or lysine and with canavanine, thialysine, and bead lysis protocol. Upstream (uptag) and downstream (downtag) barcodes G418. This process selects for MATa haploids that inherited the kanMX- were amplified using universal primers that contain a common primer se- marked deletion (the magic marker) and lyp1Δ0. The second round of sequence (lowercase below), a 4-bp identifier for multiplexing (underlined lection was identical to the first but also had uracil, methionine, cysteine, below), and the sequences required for Illumina cluster formation (uppercase and leucine left out of the medium. This round selects for strains that did not below):

E4400 | www.pnas.org/cgi/doi/10.1073/pnas.1318100110 Gibney et al. Downloaded by guest on October 1, 2021 Uptag_For:5′-CAAGCAGAAGACGGCATACGAGCTCTTCCnnnngatgtccacga- were generated, each consisting of a random draw (with replacement) from PNAS PLUS ggtctct-3′ each time point in the data series (10,000 replicates). Uptag_Rev: 5′-AATGATACGGCGACCACCGAGATCTCtccatgtcgctggccgg-3′ Analysis of Gene-Expression Data (28–36 °C). Data for the mild heat-shock Dntag_For: 5′-CAAGCAGAAGACGGCATACGAGCTCTTCCnnnncggtgtcggtctcg- gene-expression analysis (28–36 °C) were acquired using a prototrophic, CEN. tag-3′ PK derivative (DBY11092 MATa MAL2-8C) as part of a previous study (9, 46). ﬂ Dntag_Rev: 5′-AATGATACGGCGACCACCGAGATCTCagatgcgaagttaagtg-3′ Brie y, cells were grown in continuous culture (0.15/h dilution rate; 4.6-h doubling time) in phosphate-limited medium. Samples were isolated and PCR was performed using the Roche Expand High Fidelity PCR System in analyzed at 0, 5, 10, and 15 min after the temperature was shifted from a Thermo Hybaid PCR Express Thermal Cycler using the following conditions: 28 °C to 36 °C. RNA isolation, labeling, and microarray hybridization were one cycle at 94 °C for 2 min; six cycles at 94 °C for 15 s, 58 °C for 30 s, and 72 °C performed as described by Brauer et al. (40).

for 45 s; 20 cycles at 94 °C for 15 s, 64 °C for 30 s, and 72 °C for 45 s; and one The abundance of each mRNA is represented as the log2-transformed ratio cycle at 72 °C for 4 min. The product (∼200 bp) was gel purified using a of that gene at t = n min to t = 0 min. Gene-expression data were clustered Qiagen Gel Extraction Kit. The purified samples were used directly for cluster using hierarchical clustering (Pearson correlation metric) in Multiple Exper- formation in the Illumina flow cell. iment Viewer (47, 48). The average maximal change in gene expression We sequenced uptag and downtag barcodes from each sample in one flow occurs at 10 min; the ratio at this time point was used for comparisons of cell lane in an Illumina Genome Analyzer II according to the manufacturer’s maximum fold change of gene expression. instructions (Illumina). Each flow cell lane produced 8–9 million reads, resulting in ∼4 million reads for both uptag and downtag barcodes. Reads Functional Annotation. Gene names and functional annotation were down- were mapped to barcodes (and then genes) according to Smith et al. (10). On loaded from the Saccharomyces Genome Database (SGD; www.yeastgenome. average, we were able to map 70% of our reads to genes; ∼3,900 genes org). SGD GO Term Finder was used to assess functional enrichment among were detected. (See SI Appendix, Table S7 for barcode-mapping details.) significant gene sets (P value cutoff set to 0.1). Because this screening technique used only a subset of gene deletions, a background set of genes Analysis of Comparative Fitness Data. In general, upstream and downstream was defined for GO Term searches using the statistically significant gene barcode mapping correlate well, although in some instances one barcode is deletions detected for a given screen. For the most part, GO Term Finder poorly represented, likely because of differences in PCR efficiency resulting searches did not identify any significantly enriched terms, although many from point mutations in the primer-binding sequences or mutations in the sets of genes had a high percentage of “biological process unknown.” For barcode sequence. For each mutant strain, either the upstream or down- detailed results of the GO Term Finder searches, see SI Appendix, Table S8. stream barcode was used to assess strain abundance within the population, Results from GO term searches are available for download (Dataset S3). based on which barcode had the higher relative frequency within the population at time zero. To assess decreases in strain abundance accurately Measurement of Trehalose Levels. Cellular trehalose content was measured as throughout the time course, we limited our analysis to mutants that we were described (49). Briefly, cells were grown to early logarithmic phase in either able to count with confidence before the heat shock. We required that minimal dextrose- or galactose-containing medium (depending on the strains have at least 40 theoretical colonies on the plate at time 0 (strain growth requirements of the strains used). From this culture three OD600 barcode frequency multiplied by the number of colonies sampled). In general, units were collected, washed in water, and resuspended in 0.25 M Na2CO3. this method performed consistently, because the ranked barcode frequencies These mixtures then were incubated for 4 h at 95 °C, followed by the ad- from both screens were highly correlated (SI Appendix, Fig. S1). dition of 150 μL of 1 M acetic acid and 600 μL of 0.2 M sodium acetate. An To compare strain changes within the population, barcode counts first aliquot of this mixture then was treated with 1 U/mL of trehalase (Mega- were transformed into frequency within the population by comparing each zyme) overnight at 37 °C. The amount of glucose present in the resulting strain’s barcode counts with the total number of barcode counts at that time sample (liberated from trehalose) was assessed using a Glucose (GO) Assay SYSTEMS BIOLOGY point. Next, the abundance of each strain was represented relative to its Kit (Sigma-Aldrich).

frequency within the population at time zero by the log2-transformed ratio of strain (barcode) frequency at t = n min to strain (barcode) fre- ACKNOWLEDGMENTS. We thank Wei Wang, Donna Storton, and Jessica quency at t = 0 min. Linear regression was performed for each gene on the Buckles for technical assistance with Illumina Sequencing and microarray analysis; members of the D.B. laboratory for helpful discussions; and R. Scott log -transformed ratio of relative frequencies at each time point (the de- 2 McIsaac for technical assistance with ﬁgure preparation. This research was nominator in this regression was hours). The slope of the linear regression was supported by National Institute of General Medical Sciences Center for ’ used to approximate the change in each strain s abundance in the population, Quantitative Biology Grant GM071508, National Institutes of Health Grants which we refer to as “death rate.” Signiﬁcance values were calculated for GM046406 (to D.B.), GM097852 (to P.A.G.), and GM101091 (to D.C.H.), and death rates using a bootstrap technique in which many theoretical replicates National Science Foundation Grant 1122240 (to D.C.H.).

1. Craig EA (1985) The heat shock response. CRC Crit Rev Biochem 18(3):239–280. 15. Boer VM, Amini S, Botstein D (2008) Influence of genotype and nutrition on survival 2. Morano KA, Grant CM, Moye-Rowley WS (2012) The response to heat shock and and metabolism of starving yeast. Proc Natl Acad Sci USA 105(19):6930–6935. oxidative stress in Saccharomyces cerevisiae. Genetics 190(4):1157–1195. 16. Pan X, et al. (2004) A robust toolkit for functional profiling of the yeast genome. Mol 3. Gasch AP, et al. (2000) Genomic expression programs in the response of yeast cells to Cell 16(3):487–496. environmental changes. Mol Biol Cell 11(12):4241–4257. 17. Tong AH, Boone C (2006) Synthetic genetic array analysis in Saccharomyces cerevisiae. 4. Causton HC, et al. (2001) Remodeling of yeast genome expression in response to Methods Mol Biol 313:171–192. – environmental changes. Mol Biol Cell 12(2):323 337. 18. Shoemaker DD, Lashkari DA, Morris D, Mittmann M, Davis RW (1996) Quantitative fi 5. Giaever G, et al. (2002) Functional pro ling of the Saccharomyces cerevisiae genome. phenotypic analysis of yeast deletion mutants using a highly parallel molecular bar- – Nature 418(6896):387 391. coding strategy. Nat Genet 14(4):450–456. fi 6. Klosinska MM, Crutch eld CA, Bradley PH, Rabinowitz JD, Broach JR (2011) Yeast cells 19. Winzeler EA, et al. (1999) Functional characterization of the S. cerevisiae genome by can access distinct quiescent states. Genes Dev 25(4):336–349. gene deletion and parallel analysis. Science 285(5429):901–906. 7. Gresham D, et al. (2011) System-level analysis of genes and functions affecting sur- 20. Elliott B, Haltiwanger RS, Futcher B (1996) Synergy between trehalose and Hsp104 for vival during nutrient starvation in Saccharomyces cerevisiae. Genetics 187(1):299–317. thermotolerance in Saccharomyces cerevisiae. Genetics 144(3):923–933. 8. Zakrzewska A, et al. (2011) Genome-wide analysis of yeast stress survival and toler- 21. Sanchez Y, Lindquist SL (1990) HSP104 required for induced thermotolerance. Science ance acquisition to analyze the central trade-off between growth rate and cellular 248(4959):1112–1115. robustness. Mol Biol Cell 22(22):4435–4446. 22. Singer MA, Lindquist S (1998) Multiple effects of trehalose on protein folding in vitro 9. Lu C, Brauer MJ, Botstein D (2009) Slow growth induces heat-shock resistance in – normal and respiratory-deficient yeast. Mol Biol Cell 20(3):891–903. and in vivo. Mol Cell 1(5):639 648. fi 10. Smith AM, et al. (2009) Quantitative phenotyping via deep barcode sequencing. 23. Lindquist S, Kim G (1996) Heat-shock protein 104 expression is suf cient for ther- – Genome Res 19(10):1836–1842. motolerance in yeast. Proc Natl Acad Sci USA 93(11):5301 5306. 11. Parsell DA Taulien J Lindquist S (1993) The role of heat-shock proteins in thermo- 24. Berry DB, Gasch AP (2008) Stress-activated genomic expression changes serve a pre- – tolerance. Philos Trans R Soc Lond B Biol Sci 339(1289):279-285; discussion 285-276. parative role for impending stress in yeast. Mol Biol Cell 19(11):4580 4587. 12. Lindquist S (1986) The heat-shock response. Annu Rev Biochem 55:1151–1191. 25. Estruch F (2000) Stress-controlled transcription factors, stress-induced genes and stress 13. Lindquist S, Craig EA (1988) The heat-shock proteins. Annu Rev Genet 22:631–677. tolerance in budding yeast. FEMS Microbiol Rev 24(4):469–486. 14. Verghese J, Abrams J, Wang Y, Morano KA (2012) Biology of the heat shock response 26. De Virgilio C, Hottiger T, Dominguez J, Boller T, Wiemken A (1994) The role of tre- and protein chaperones: Budding yeast (Saccharomyces cerevisiae) as a model system. halose synthesis for the acquisition of thermotolerance in yeast. I. Genetic evidence Microbiol Mol Biol Rev 76(2):115–158. that trehalose is a thermoprotectant. Eur J Biochem 219(1-2):179–186.

Gibney et al. PNAS | Published online October 28, 2013 | E4401 Downloaded by guest on October 1, 2021 27. Hottiger T, De Virgilio C, Hall MN, Boller T, Wiemken A (1994) The role of trehalose 38. Kaida D, Yashiroda H, Toh-e A, Kikuchi Y (2002) Yeast Whi2 and Psr1-phosphatase synthesis for the acquisition of thermotolerance in yeast. II. Physiological concen- form a complex and regulate STRE-mediated gene expression. Genes Cells 7(6): trations of trehalose increase the thermal stability of proteins in vitro. Eur J Biochem 543–552. 219(1-2):187–193. 39. Berry DB, et al. (2011) Multiple means to the same end: The genetic basis of acquired 28. Singer MA, Lindquist S (1998) Thermotolerance in Saccharomyces cerevisiae: The yin stress resistance in yeast. PLoS Genet 7(11):e1002353. and yang of trehalose. Trends Biotechnol 16(11):460–468. 40. Brauer MJ, et al. (2008) Coordination of growth rate, cell cycle, stress response, and 29. Byrne KP, Wolfe KH (2005) The Yeast Gene Order Browser: Combining curated ho- metabolic activity in yeast. Mol Biol Cell 19(1):352–367. mology and syntenic context reveals gene fate in polyploid species. Genome Res 41. Castrillo JI, et al. (2007) Growth control of the eukaryote cell: A systems biology study 15(10):1456–1461. in yeast. J Biol 6(2):4. 30. Bell W, et al. (1998) Composition and functional analysis of the Saccharomyces cer- 42. Levy SF, Ziv N, Siegal ML (2012) Bet hedging in yeast by heterogeneous, age-corre- evisiae trehalose synthase complex. J Biol Chem 273(50):33311–33319. lated expression of a stress protectant. PLoS Biol 10(5):e1001325. 31. Luyten K, et al. (1993) Disruption of the Kluyveromyces lactis GGS1 gene causes in- 43. Lewis K (2012) Persister cells: Molecular mechanisms related to antibiotic tolerance. ability to grow on glucose and fructose and is suppressed by mutations that reduce Handb Exp Pharmacol 211:121–133. sugar uptake. Eur J Biochem 217(2):701–713. 44. Meiron H, Nahon E, Raveh D (1995) Identiﬁcation of the heterothallic mutation in 32. Thevelein JM, Hohmann S (1995) Trehalose synthase: Guard to the gate of glycolysis HO-endonuclease of S. cerevisiae using HO/ho chimeric genes. Curr Genet 28(4): in yeast? Trends Biochem Sci 20(1):3–10. 367–373. 33. Mahmud SA, Hirasawa T, Shimizu H (2010) Differential importance of trehalose ac- 45. Ekino K, Kwon I, Goto M, Yoshino S, Furukawa K (1999) Functional analysis of HO cumulation in Saccharomyces cerevisiae in response to various environmental gene in delayed homothallism in Saccharomyces cerevisiae wy2. Yeast 15(6):451–458. stresses. J Biosci Bioeng 109(3):262–266. 46. van Dijken JP, et al. (2000) An interlaboratory comparison of physiological and ge- 34. Parsell DA, Kowal AS, Singer MA, Lindquist S (1994) Protein disaggregation mediated netic properties of four Saccharomyces cerevisiae strains. Enzyme Microb Technol 26 by heat-shock protein Hsp104. Nature 372(6505):475–478. (9-10):706–714. 35. Glover JR, Lindquist S (1998) Hsp104, Hsp70, and Hsp40: A novel chaperone system 47. Eisen MB, Spellman PT, Brown PO, Botstein D (1998) Cluster analysis and display of that rescues previously aggregated proteins. Cell 94(1):73–82. genome-wide expression patterns. Proc Natl Acad Sci USA 95(25):14863–14868. 36. Craig EA, Huang P, Aron R, Andrew A (2006) The diverse roles of J-proteins, the 48. Saeed AI, et al. (2003) TM4Aa free, open-source system for microarray data man- obligate Hsp70 co-chaperone. Rev Physiol Biochem Pharmacol 156:1–21. agement and analysis. Biotechniques 34(2):374–378. 37. Boorstein WR, Ziegelhoffer T, Craig EA (1994) Molecular evolution of the HSP70 49. Parrou JL, François J (1997) A simpliﬁed procedure for a rapid and reliable assay of multigene family. J Mol Evol 38(1):1–17. both glycogen and trehalose in whole yeast cells. Anal Biochem 248(1):186–188.

E4402 | www.pnas.org/cgi/doi/10.1073/pnas.1318100110 Gibney et al. Downloaded by guest on October 1, 2021