Transient intracellular acidification regulates the core transcriptional heat shock response

Heat shock induces a conserved transcriptional program regulated by heat shock factor 1 (Hsf1) in eukaryotic cells. Activation of this heat shock response is triggered by heat-induced misfolding of newly synthesized polypeptides, and so has been thought to depend on ongoing protein synthesis. Here, using the budding yeast Saccharomyces cerevisiae, we report the discovery that Hsf1 can be robustly activated when protein synthesis is inhibited, so long as cells undergo cytosolic acidification. Heat shock has long been known to cause transient intracellular acidification which, for reasons which have remained unclear, is associated with increased stress resistance in eukaryotes. We demonstrate that acidification is required for heat shock response induction in translationally inhibited cells, and specifically affects Hsf1 activation. Physiological heat-triggered acidification also increases population fitness and promotes cell cycle reentry following heat shock. Our results uncover a previously unknown adaptive dimension of the well-studied eukaryotic heat shock response.


Introduction 1
To survive and thrive, organisms must rapidly respond when their environments turn harsh. 2 Cells across the tree of life possess the capacity to adaptively respond to primordial 3 stresses-heat, starvation, hypoxia, exposure to noxious compounds-in a conserved 4 program involving the production of so-called heat shock proteins, many of which act as 5 molecular chaperones. 1 Transcription of heat shock proteins surges at the onset of stress, 6 reaching as much as a thousand fold during thermal stress, with more modest induction 7 accompanying nutrient withdrawal and diverse other stresses. 1,2 In eukaryotes, the 8 transcriptional stress response is controlled by multiple factors, with the heat shock 9 transcription factor Hsf1 regulating the induction of a core group of chaperones. 3 Induced 10 chaperones, in turn, assist with protein folding, as well as preventing and dispersing 11 stress-induced molecular aggregates. 4, 5 12 The same diverse stresses which stimulate the transcriptional response are also 13 accompanied by intracellular acidification-a drop in cytosolic pH. [6][7][8][9] Like the 14 transcriptional response, stress-induced acidification is broadly conserved in eukaryotes, 15 including mammals, 7, 10-13 insects, 14,15 plants, 16 and fungi. 6,9 Although acidification has 16 sometimes been viewed as a toxic consequence of stress, particularly in studies of hypoxia 17 and ischemia-associated acidosis, 11,16 the cytoprotective effects of short-term acidification 18 were identified decades ago. 11 Recent work has shown that interfering with 19 energy-depletion-induced acidification in budding yeast and in fission yeast (which diverged 20 from budding yeast more than half a billion years ago 17 ) compromises the fitness of both 21 species, 8 indicating a cytoprotective effect of acidification by an unknown mechanism. 22 We became interested in intracellular pH during studies of the stress-triggered 23 aggregation of specific endogenous proteins and RNA into large clusters, termed stress 24 granules when observed microscopically, which occurs in all eukaryotes. 18 During heat stress 25 in budding yeast this aggregation is reversible, 19 and stress-induced molecular chaperones 26 facilitate cluster dispersal 5 which accompanies resumption of translation and the cell 27 cycle. 5,9 A recent cascade of studies has revealed that protein components of these 28 structures can undergo phase separation, demixing from solution into liquid and hydrogel 29 droplets in a process triggered by temperature and pH. [20][21][22][23][24] Importantly, physiological 30 stress-associated temperature and pH conditions suffice to trigger demixing of specific 31 proteins in vitro, 22,23 leading to the proposal that such proteins may act as primary sensors 32 of stress. 22 33 What role does stress-induced cellular acidification play in the stress response? Early 34 work in Drosophila produced mixed results: one study indicated that acidification had little 35 impact on the production of heat shock proteins, 14 while later work showed that Hsf1 36 trimerization, a key activation step, could be induced by acidification in vitro. 15 Recent 37 work has revealed that the starvation-stress-responsive transcription factor Snf1 senses and 38 2/47 is regulated by pH. 25 Despite discovery of cytoprotective effects associated with 39 acidification during stress, 8,26,27 how pH influences the transcriptional heat shock response, 40 and how intracellular acidification, chaperone production, and cellular growth are related, 41 remain longstanding open questions. 42 To answer these questions, we developed a single-cell system to both monitor and 43 manipulate cytosolic pH while tracking the induction of the heat shock response in budding 44 yeast. We find that heat stress without acidification leads to suppression of heat shock 45 protein synthesis and a substantial fitness defect, but that intracellular acidification alone 46 does not elicit a response except under extreme conditions. Single-cell data reveals that 47 only cells which restore intracellular pH to pre-stress levels mount a robust heat shock 48 response, with an associated acceleration in growth. Finally, global measurement of 49 transcript levels as a function of intracellular pH during stress reveals specific suppression of 50 core Hsf1 target genes when intracellular acidification is prevented. Building on previous 51 work positing a role for temperature-and pH-dependent phase separation in sensing 52 stress, 22 we propose a specific mechanism for induction of the heat-shock transcriptional 53 response in which elevated pH suppresses a stress-sensitive phase separation process. Our 54 results link cytosolic acidification to the regulation of the canonical transcriptional heat 55 shock response and subsequent stress adaptation in single cells, indicating that pH 56 regulation plays a central role in the Hsf1-mediated stress response. 57

58
A high-throughput assay allows quantification of pH-dependent, 59 single-cell responses to stress. 60 In yeast, intracellular acidification during stress is thought to be dictated by the cellular 61 exterior: yeast live in acidic environments while maintaining a slightly basic intracellular pH 62 through the activity of membrane-localized proton pumps, 28 and protons accumulate in the 63 cell during stress. While the mechanism of proton influx remains somewhat murky, elevated 64 temperature increases membrane permeability 29 and other stresses have been shown to 65 reduce proton pump activity. 28,30,31 We first sought to measure the pH changes associated 66 with heat stress in our system. 67 To track intracellular pH during stress and recovery, we engineered yeast cells to 68 constitutively express pHluorin, a pH-sensitive green fluorescent protein derivative used to 69 measure intracellular pH, 32 in the cytosol. We used this strain to characterize the 70 intracellular pH changes that occur during heat stress and recovery with millisecond 71 resolution. These results, for a 42 C, 10-minute heat stress in acidic media (pH 4), are 72 shown in Figure 1a. In agreement with previous results, 6 we find that cells exposed to 73 elevated temperature rapidly acidify from a resting pH of approximately 7.5 down to a 74 3/47 range of 6.8 to 7.0, and that this pH change is highly reproducible ( Figure S2a). In all of 75 these experiments, we consistently observe two phases of pH change following the onset of 76 heat stress: a short, sharp drop and partial recovery within the first two minutes, followed 77 by slower acidification over the next several minutes. When returned to ambient growth 78 temperature, cells return to the resting pH in approximately 10 minutes. The minimum pH 79 reached and the dynamics of recovery are similar for cells stressed at the same temperature 80 for 20 minutes ( Figure S2b). 81

Figure 1.
Yeast cells respond to stress with intracellular pH changes and production of heat shock proteins which can be tracked at the single-cell level. a) Intracellular pH changes during stress measured with continuous flow cytometry; each point is an individual cell, and the gray rectangle is the period during which cells were exposed to elevated temperature. A solid line shows a sliding-window average over all data; for visual clarity, only 2% of points are shown. Dashed lines represent the range we subsequently use as representative of the physiological pH drop. b) Production of molecular chaperones in response to temperature stress. Each plot is a timepoint during recovery from 42 C, 20-minute heat stress showing forward scatter pulse area, which correlates roughly with size, versus red fluorescence. Gray points are wild-type/unlabeled cells. Red points are cells expressing Ssa4-mCherry from the native locus. c) Induction curves showing the fold change in median ratio of fluorescence to size (forward scatter pulse area) as a function of time. Fold change is relative to unstressed cells and is common to all timepoints. Thin gray lines are individual experiments, thick red curve is the smoothed conditional mean.
The hallmark of the heat shock response is the production of molecular chaperones. 1, 4, 33 82

4/47
To assess the effects of intracellular pH on this response, we needed to track chaperone 83 induction and intracellular pH simultaneously. We further engineered a pHluorin-labeled 84 yeast strain to express Ssa4, a strongly stress-responsive Hsp70 heat shock protein, 33, 34 85 from its endogenous locus, tagged with the red fluorescent protein mCherry. 35 This 86 two-color reporter strain allowed us to simultaneously track intracellular pH and the stress 87 response at the single-cell level. 88 We stressed cells at 42 C for 20 minutes and then returned them to 30 C to recover.

89
Samples were taken at 15-to 30-minute intervals during recovery and analyzed by flow 90 cytometry to monitor Ssa4-mCherry production. An example of the raw data, showing an 91 increase in fluorescence in the mCherry channel as a function of time, is shown in Figure 1b To determine how acidification influences the stress response, we chemically manipulated 100 intracellular pH independently of heat stress using an ionophore, nigericin, modifying a 101 published protocol. 36 Ionophores allow ions to penetrate cell membranes, temporarily 102 destroying the electrochemical gradient. Nigericin is a K + / H + antiporter 37 which has 103 been used in a variety of biological systems to equilibrate intracellular and extracellular 104 pH. [38][39][40][41] Exposing ionophore-treated cells to heat stress (42 C for 20 minutes; Figure 2a) at 105 a range of buffer-controlled pH levels permitted us to monitor the effect of intracellular pH 106 during stress on the subsequent stress response. We verified that measured intracellular pH 107 during stress matched the buffer pH, and that the efficacy of the ionophore was not affected 108 by temperature (Figure 2b). After stress, we returned cells to ionophore-free media at 30 C 109 and monitored Ssa4 induction by flow cytometry. Treatment with buffer and ionophore 110 delayed the stress response in all samples relative to untreated cells, but did so consistently 111 and did not affect the ultimate induction level (Figure 2c).

112
The range of pH values studied, from 7.5 to 5.0, reflected three main pH regimes. Cells 113 held at or near pH 7.5, their resting pH, experienced little or no acidification during stress. 114 Cells moved to pH 6.8 to 7.0 experienced an approximately physiological level of 115 acidification (cf. Fig. 1a). Cells moved to pH 5.0 experienced a larger-than-expected pH  . Acidic intracellular pH during stress is necessary for rapid production of heat shock proteins. a) Schematic of intracellular pH manipulation and stress. Colored stars correspond to the measurements shown in b). b) Measured intracellular pH distribution during pH manipulation before (green), during (red), and after (purple) 42 C heat stress. Dashed lines indicate buffer pH, black distribution shows unmanipulated cells. Intracellular pH is accurately manipulated during stress. c) Induction of Ssa4 during recovery from normal (red) or pH-manipulated (gray, pH 6.8) stress. Thin curves are individual experiments and thick curves are smoothed conditional means (see Methods for details). The red curve is the same data from Figure 1c for comparison. Although pH manipulation causes a delay in Ssa4 production, it does not affect the ultimate level of induction. d) Fold change in Ssa4 expression during recovery following stress at many different intracellular pHs. Points represent the median of individual flow-cytometric measurements, and at least three biological replicates were performed for each condition (see Methods). Lines are sigmoid fits (see Methods for fitting details). Preventing the pH change during stress causes a marked delay in the production of Ssa4. e) Fold change in Ssa4 expression during recovery in media buffered to pH 7.4 after stress with manipulated intracellular pH. f) Median Ssa4 fold change two hours after pH manipulation with (left) and without (right) heat stress. 6/47 optimal growth conditions. Intracellular pH values between the physiological post-stress 120 value (7.0) and the physiological resting value (7.5) caused intermediate delays as monitored 121 by Ssa4 expression after two hours (Figure 2f, left). To ensure that this was not an artifact 122 of fluorescent protein tagging, we also tagged Ssa4 with a FLAG tag and confirmed the 123 delay by Western blot ( Figure S3a). We also blotted against an untagged native small heat 124 shock protein, Hsp26, under the same conditions and observed the same pH-dependent 125 reduction in protein production, confirming the effect in another heat-shock protein (Figure 126 S3a). Given these results, we adopt the working model that the pH-dependence of heat 127 shock proteins likely generalizes to the full response; genome-scale results presented later 128 further support this extrapolation.

129
To further determine whether acidification or ionophore treatment alone induced the 130 response, we performed control experiments with pH manipulation at ambient temperature. 131 shock (cf. Figure 1a). These results indicate that ionophore treatment does not trigger the 135 heat shock response, and suggest that the physiological pH drop in heat-shocked cells is not 136 solely responsible for chaperone induction.

137
Because intracellular acidification results in large part from the influx of environmental 138 protons, we reasoned that intracellular pH, and its downstream effects on the heat-shock 139 response, might also depend on the pH of the medium during recovery. To fully deprive 140 cells of acidification, we allowed stressed cells to recover in media which was buffered to the 141 resting pH of 7.5 and, as before, free of ionophore. Cells stressed at pH 6.5 or pH 5.0 showed 142 similar chaperone induction to that observed in unbuffered media. In contrast, maintaining 143 cells at pH 7.5 during stress and recovery abolished chaperone production ( Figure 2e). 144 We draw several conclusions from these data. The physiologically observed acidification 145 of the cytosol is required for, but does not cause, rapid heat shock protein production under 146 these conditions-a remarkable result. Depriving cells of the opportunity to acidify silences 147 chaperone production. Cells permitted the chance to acidify after heat shock were still 148 capable of mounting a response albeit with a substantial delay, indicating that heat and 149 acidification do not need to co-occur to elicit a response. This suggested that intracellular 150 pH during recovery played a significant role in the production of heat shock proteins, and 151 we turned our attention to that possibility.

152
Reversal of stress-induced acidification during recovery promotes heat 153 shock protein production in single cells 154 How does intracellular pH during recovery influence chaperone production? In the absence 155 of ionophore treatment, stress-associated intracellular pH changes reverse rapidly after 156 Figure 3. Preventing acidification during stress dysregulates the return to resting pH during recovery, suppressing heat shock protein production. a) Intracellular pH during recovery for cells stressed at various pH values show variation induced by pH during stress. Thin gray traces are replicates, thick colored lines are averages over replicates. b) Relationship between intracellular pH and Ssa4 fold change on the single cell level during recovery. Return to the resting pH, bounded by dotted lines, appears to precede Ssa4 induction, and is necessary but not sufficient for high expression levels. Color and label refers to the pH during stress. c) The fraction of the population in the resting pH range predicts Ssa4 expression, summarizing data in b; circles show the population median, and triangles show the median of only the subpopulation of cells within the resting pH range. Data show cells after three hours of recovery. d) During recovery from heat stress at intracellular pH levels between 7 and 7.5 a bimodal distribution of Ssa4 fold change was observed. A two-component mixture model was used to classify cells into two groups: low and high induction level (> 0.95 posterior probability cutoff used for assignment). Cells stressed at pH 7.2 are shown here as an example. e) Intracellular pH as a function of time for the low and high expression groups. Cells stressed at pH 7.2 are shown, see Figure S4 for all conditions. **p << 0.01, Wilcoxon rank sum test.

8/47
stress ends 8 (cf. Figure 1a), and, if heat shock is brief enough, precede detectable 157 accumulation of heat shock proteins. Most populations treated with ionophore during heat 158 shock also rapidly returned to the pre-stress pH upon return to ambient growth 159 temperature ( Figure 3a). However, cells stressed at pH values above 7.0 took longer on 160 average to recover and, counterintuitively, cells held at the pre-stress (resting) pH during 161 stress acidified during recovery and failed to recover their intracellular pH after two hours at 162 ambient growth temperature (Figure 3a, right hand side, blue trace). At the same time, 163 these cells fail to robustly produce heat shock proteins. Given these observations at the 164 population level, we sought to determine whether intracellular pH during recovery is 165 predictive of chaperone induction at the single-cell level.  These data demonstrate that although cells require acidification during stress to mount a 190 rapid response at the population level, the response further depends on subsequent reversal 191 of acidification. The return to the resting pH dictates the dynamics of chaperone 192 production. Acidification, either simultaneous with or following heat stress, followed by 193 return to the resting pH are required for robust induction of chaperones after heat stress. 194

9/47
Intracellular pH contributes to cellular fitness during stress and recovery 195 Given the highly conserved nature of the heat shock response, it is reasonable to expect 196 that the response improves cellular fitness after stress. In light of the connections we have 197 established between intracellular pH changes and the deployment of this response, we 198 sought to determine whether these pH changes affected fitness during recovery. In 199 single-celled organisms such as S. cerevisiae, fitness differences may be quantified by 200 measuring the growth rate relative to a wild-type competitor 43 (Figure 4a). We stressed 201 pHluorin/Ssa4-mCherry dual-labeled cells in different pH conditions and, just prior to 202 recovery, mixed them with exponentially growing, unlabeled wild-type cells. Samples were 203 taken during recovery and analyzed by flow cytometry; the number of cells belonging to 204 each type was counted, and the log ratio of that value was plotted as a function of time.

205
The fitness loss was quantified by fitting these data to a line and taking the slope (Figure 206 4a). We performed additional controls to correct for potential strain differences and for the 207 fitness effect of ionophore; see Methods and Figure S6. Both of these measures reflect population-level behavior, yet a causal link predicts 220 correlations between pH, induction, and growth in single cells. A priori, these correlations 221 could be in any direction in subpopulations, so long as the population average is preserved. 222 Moreover, the complex web of connections between the variables in question make many 223 outcomes possible. For example, pH influences protein folding and stability, 44 heat shock 224 causes protein misfolding, 4 and protein misfolding promotes chaperone production and 225 reduces fitness. 43 We therefore exploited cell-to-cell variation within populations to assess 226 the empirical relationships between intracellular pH, heat-shock protein production, and   . Populations forced to physiological range of pH values during stress have the smallest fitness deficit during recovery. a) A schematic of the relative growth rate experiment which measures population fitness. Stressed, labeled cells and exponentially-growing, unlabeled cells are mixed and allowed to grow at 30 C. 'Labeled' cells are Ssa4-mCherry/pHluorin diploids. Fitness is measured by fitting the log-ratio of the population sizes as a function of time to a line; the slope of the line is the difference in exponential growth rates (see Methods for full explanation). All values are expected to be  0 because stressed cells are being compared to exponentially growing cells. b) intracellular pH during stress vs. relative growth rate (each point is an independent experiment); points are the slope of the line illustrated in a). Values between 30 minutes and 2 hours of recovery were fit with a line. See Methods for details. Figure 5. Fitness, intracellular pH, and heat shock protein production during recovery is correlated in single cells. a) Classification of cells into large (red) and small (gray) populations. Classification was performed by fitting the forward-scatter pulse width to a two-component Gaussian mixture model and using the point of maximum overlap as a cutoff between the two categories. Classification of cells stressed at pH 6.8 is shown. Black labels are the total number of cells in each category. b) Ssa4 fold-change versus intracellular pH for budded (red) and unbudded (gray) cells during recovery; data for 3 hours post-stress are shown. pH during stress is shown on the right side of the plot. Black lines show summary statistics of the entire population (budded and unbudded) and span the middle 50% of the data, crossing at the median of each dimension. c) Proportion of cells budded as a function of time during recovery. The characteristic shape of the curve is represented in the left-most panel, with data from cells stressed without pH manipulation. There is a peak in the proportion budded at approximately 2 hours recovery after stress (vertical dashed line). Whether and when this peak occurs after stress varies between populations stressed at different pHs; cells stressed close to the normal stress pH are most similar to the native curve. d) Summary of c); the average proportion of cells budded between 90 and 120 minutes after stress.
12/47 and resuming growth) represents an important stage in recovery, and failure to do so on the 233 population level would appear as reduced population-level fitness. Delayed release from G1 234 arrest provides a potential explanation for the reduced growth rates we measured in Figure 235 4b. We tested the resulting prediction that single cells in these populations would show 236 signs of G1 arrest by examining the distribution of cellular morphologies in each population 237 during recovery.

238
Cells without a bud cannot be confidently assigned to a growth state, as they may be 239 actively growing and in G1, arrested in G1, or in G0. However, the presence of a bud 240 following stress indicates that the cell has re-entered the cell cycle and begun reproducing. 241 By classifying cells as either unbudded or budded and looking for differences between the 242 populations, we could determine whether budding cells that are actively growing during   stress-induced cell-cycle arrest, as others have observed. 9 We conclude that acidification 275 during stress is adaptive. where ionophore treatment appears to affect the timing of production rather than ultimate 290 levels of Ssa4 protein.

291
To isolate the pH-specific effects on transcript levels, we focused on the per-gene  (Figure 6b). These acidification-sensitive, Hsf1-regulated genes include the core 305 molecular chaperones long associated with the canonical heat-shock response: Hsp70s (the 306 cytosolic SSA family and ER-localized KAR2 ), Hsp90 and co-chaperones (HSC82 and 307 Figure 6. Failure to acidify during stress specifically represses Hsf1-activated genes. a) Transcript abundance (transcripts per million, tpm) in stressed versus unstressed samples. b) Cumulative distribution function (CDF) of per-gene transcript abundance in cells stressed at pH 6.8 relative to cells stressed at pH 7.4. The red line shows all heat shock proteins; this group is further divided into genes regulated by Msn2/4 (green) which show similar behavior to all detected transcripts (gray, (P = .402, Wilcoxon rank sum test)), and those regulated by Hsf1 (orange), which are significantly higher in acidified cells (P < 0.01).
HSP82 ; CPR6, STI1 ), Hsp40/J-proteins (SIS1, APJ1 ), and small heat-shock proteins 308 (HSP42, BTN2, HSP10 ), among others. With the caveat that we are measuring transcript 309 levels and not transcription rates, we conclude that the effect of pH is specific to Hsf1.  The ability to silence Hsf1 activation during a robust heat shock with a modest change 320 in pH is thoroughly unexpected, because heat-induced protein misfolding has long been 321 thought to provide the trigger for Hsf1 activation. 33,56 In the currently accepted model for 322 heat-triggered Hsf1 activation, here referred to as the 'misfolding' model, events proceed as 323 follows. Hsf1 is constitutively bound and repressed by the molecular chaperone Hsp70 324 before stress. 57,58 Heat stress destabilizes proteins, causing them to misfold and expose 325 hydrophobic regions 4 for which Hsp70 has high affinity. 59 Titration of Hsp70 away from 326 Hsf1 is sufficient for Hsf1 activation. 57 How simply maintaining the pre-stress pH would 327 prevent Hsf1 activation is not immediately obvious. We weigh three major possibilities to 328 explain our data: 1) that prevention of acidification prevents heat-induced misfolding; 2) Together, these studies indicate the existence of proteins that undergo stress-triggered, 379 pH-dependent demixing processes and produce assemblies that conditionally recruit Hsp70. 380 If robust Hsf1 activation requires such pH-sensitive protein demixing, then preventing the 381 17/47 drop in pH would suppress Hsf1 activation, exactly as we observe. We thus favor this last 382 possibility to explain our results. Figure 7. A model for the pH dependence of the Hsf1-mediated transcriptional response. a) Hsp70 represses Hsf1 before heat shock. One or more sensor proteins expose Hsp70 binding sites in response to stress, likely achieving high sensitivity by demixing. b) The relationship between the pH/temperature phase diagram of poly(A)-binding protein (solid line, demixing in the shaded area), the observed pH/temperature induction of the Hsp70 Ssa4 reported in the present study (circles show tenfold or greater induction, x's show less induction), and a shifted phase boundary consistent with a protein capable of serving as the sensor in a. The yeast cell sits at the pre-stress position, and arrows show environmental changes studied here: 1) physiological 42 C heat shock, with resulting acidification; 2) the same heat shock with acidification artificially suppressed; 3) acidification alone without a temperature change, as occurs during other stresses such as starvation. Crossing the hypothetical sensory phase boundary corresponds with empirical observation of robust Hsf1-mediated heat-shock protein expression.

383
Temperature sensing, intracellular pH, and the transcriptional heat shock 384 response. We have previously proposed that stress-triggered protein demixing can take 385 the place of misfolding-induced aggregation in the standard model for Hsf1 activation, with 386 demixing proteins acting as the primary sensors of temperature. 22 In the misfolding model 387 of Hsf1 activation, the temperature sensors are the thermally unfolded proteins. Because only what the misfolding model already requires, namely to recruit Hsp70 away from Hsf1 398 in a temperature-dependent fashion. Formation of toxic species is not required.

399
What might these sensors be? In previous work we identified dozens of proteins which 400 demix into reversible assemblies in response to heat stress in budding yeast. 19   This finding agrees with a recent study of pH changes during starvation stress; there, the 425 stress-triggered aggregation of a transcription factor, a pH-dependent process, is required 426 for production of glucose-starvation specific genes. 25 These demonstrations underscore the 427 contribution of intracellular pH to cellular regulation, as well as the potential utility of 428 integrators which 'read' combinations of signals and synthesize this information to initiate 429 an appropriate response.

430
One notable wrinkle apparent in our data is that the pH and temperature changes need 431 not coincide to elicit a response. This is most clearly demonstrated when heat stress at the 432 resting pH, followed by intracellular acidification at the ambient non-stress temperature, 433 19/47 suffices to produce a response, albeit one that is delayed (Figure 2d). The implication is 434 that whatever process is responsible for triggering the downstream signal, which we 435 postulate to be demixing, has at least two steps. Although many possibilities exist, one 436 with some empirical support is a nucleation and growth process in which nucleation is 437 strongly temperature-dependent, while growth depends strongly on pH. Both Pab1 and 438 Pub1 show evidence for this type of behavior. 22,24 Pab1 in vitro, for example, forms large 439 numbers of small, sticky droplets when subjected to elevated temperature, but 440 comparatively fewer and larger separate droplets when subjected to a pH drop at non-stress 441 temperatures. Pub1 shows changes in material properties as temperature increases. A 442 nucleation process dependent on thermally triggered local unfolding, followed by pH-and 443 temperature-dependent demixing (such as phase separation), would explain all of these 444 observations. In this model, heat-triggered formation of nuclei, followed by pH-triggered 445 growth at normal temperatures, would still produce demixing, along with the subsequent 446 response, but with a delay, just as we observe. We note that the timescales of the response 447 are different depending on whether the two signals, pH and temperature, are received 448 coherently or incoherently. Future work is necessary to determine whether a nucleation and 449 growth process can explain these dynamics, or whether other possible explanations, such as 450 activation of two arms of a signaling pathway or even two entirely separate pathways, must 451 be invoked.

452
Temperature as a physiological signal regulating growth. That extremes of 453 temperature cause protein misfolding which triggers the heat shock response is beyond 454 question: as noted above, misfolded proteins suffice to induce the heat shock response in the 455 absence of heat, and all proteins will thermally denature at some temperature. What is not 456 clear is whether all temperatures trigger the response by causing protein misfolding.

457
Extremes of temperature may not be physiologically relevant to the response, by which we 458 mean reflecting temperatures (and times) encountered by the organism sufficiently often to 459 evolutionarily shape the organism's response to those conditions. compartments will be activated. And finally, that the response is deployed during multiple 491 stress conditions, even ones seemingly irrelevant to protein folding, reflects its likely purpose 492 as part of a regulatory strategy to contend with transiently growth-limiting change rather 493 than to contend with protein-destabilizing change.

494
The natural question then arises: what are the physiological conditions which have 495 shaped the heat shock response of this organism-whatever organism one is studying?

496
What signal is the organism responding to?

497
The evolutionary importance of temperature as a physiological signal.

498
Temperature acts as a physiological signal in other ascomycete fungi. For example, some 499 dimorphic fungi live and grow in the environment as a mold, and convert into a yeast (a 500 single-celled, reproducing fungus) in response to entering a mammalian host and detecting 501 the resulting increase in temperature, the critical sensory cue. 68 The budding yeast and 502 occasional human pathogen Candida albicans similarly requires a temperature increase to 503 trigger the bud-to-hyphae transition critical for infection, 69 which also induces chaperones 504 in a classical Hsf1-mediated heat shock response. 70 While it is physically possible that these 505 species express proteins which misfold to form toxic aggregates during their pathogenic 506 transitions, such a situation seems unlikely to us, given the existence of many thermophilic 507 21/47 species indicating no fundamental barrier to evolving proteins with higher stability.  Broader considerations. Recognition that a rise in temperature may represent a signal 533 rather than merely a damaging agent alters how one thinks about the purpose of the 534 response to temperature, the response's mechanistic basis, and the conditions under which 535 the response would be deployed. Here, the suppression of the heat shock response by 536 elevated pH suggests that acidification-and the capacity to acidify, which appears to be 537 determined in large part by extracellular pH-is a key part of the physiological context in 538 which this thermal signal is received. This logic applies broadly. In humans, for example, a 539 key physiological heat shock-fever-triggers the Hsf1-mediated heat shock response. 81 540 Perhaps fever causes new problems for cells, new self-inflicted damage to be cleaned up.

541
Alternatively, however, fever may be acting as a systemic signal which activates a cellular 542 program with key roles in modulating immune and inflammatory responses. 81 Indeed, the 543 apoptotic response of human neutrophils to fever temperatures is sharply dependent on 544 22/47 intracellular pH, with acidification promoting survival; local acidification is a hallmark of 545 inflammatory conditions and promotes neutrophil activation. 13 546 We began by noting that the biological meaning of the longstanding association of 547 cellular stress with cytosolic acidification, observed from single cells to vertebrate neurons, 548 has remained unclear. Our results speak to a potentially broad effect: that this association 549 in part reflects the dependence of the core Hsf1-mediated transcriptional stress response on 550 pH. The analogous pH-dependence of multiple proteins undergoing thermally triggered 551 demixing, coupled with the known mechanism for regulation of Hsf1, suggests a general 552 mechanism for integrated stress sensing upstream of this universally conserved eukaryotic 553 transcription factor. Moreover, this putative mechanism provides a crucial and potentially 554 general connection between protein demixing and its transduction into adaptive cellular 555 action. experiments, cultures were started from the same frozen stock, and grown so that the cell 576 density was below optical density (OD) 0.1 for at least 12 hours before stress; a dilution of 577 no more than 20-fold was performed at least 4 hours prior to stress. Cells were grown to 578 between OD 0.05 and OD 0.1 (flow cytometry) or to OD 0.3-0.4 (mRNA-Seq) before being 579 stressed.

580
All temperature stresses occurred at 42 C for 20 minutes, except for the data in Figure 581 1d, which was 42 C for 10 minutes.   running, the sample tube was briefly removed and 1mL of media at 44 C was added (to 608 account for heat loss in mixing); the tube was rapidly returned to the cytometer and held in 609 a 42 C water bath for 10 minutes, followed by 10 minutes at room temperature.  were taken at 10 and 20 minutes and analyzed by flow cytometry. The intracellular pH was 645 calculated using a calibration curve generated at 30 C using different buffers. The close 646 correspondence between the measured buffer pH and the calculated intracellular pH from 647 the calibration curve is shown in Figure 2b. 648

26/47
Manipulating intracellular pH during stress. Intracellular pH during stress was 649 manipulated using calibration curve buffer. The concentration of the ionophore was low 650 enough that any anti-microbial effects were negligible, as seen by the small fitness effect on 651 pH-manipulated, unstressed cells (see Figure Supplement S6a and c, 'RT (mock)'. 652 1.2mL of cells grown as described in above 'Growth and stress conditions' section were 653 spun out of media and resuspended in 60µL freshly prepared calibration curve buffer plus 654 ionophore at the desired pH, equilibrated at room temperature for 15-30 minutes, and then 655 either exposed to 42 C temperature ('heat shock') or room temperature ('mock') for 20 656 minutes. After stress, cells were recovered by removing the buffer and resuspending in 1.2 657 mL of fresh SC media and holding at 30 C with 250 rpm shaking. The fresh SC was either 658 not pH adjusted (with a pH of approximately 4, data shown in Figure 2d, or was buffered 659 to pH 7.4 using 0.1 M Na 2 HPO 4 : NaH 2 PO 4 buffer (data in Figure 2e). averaged to produce data shown in Figure 6. Correlation between the biological replicates is 685 shown in Figure Supplement S8a.

686
Library preparation. Total cellular RNA was extracted using hot acid-phenol 687 extraction and the resulting RNA was chemically fragmented. Samples were barcoded using 688 a 3' adaptor with a unique sequence corresponding to each sample, and then pooled for Where n x (t) is the number of cells of type x at time t, r x is the instantaneous growth 718 rate (in units of t 1 ), and nwt(0) n pH (0) is the initial mixing fraction. This equality is true 719 assuming constant exponential growth, which our data indicate is valid at least for the early 720 stages of recovery (see Figure 4a, right hand side and S6a). We can use this equation to 721 calculate the difference in growth rate, i.e. the fitness loss, for each population of cells 722 having experienced stress at a different intracellular pH. values are relative to the same reference. It also implies that the difference r pH r wt will 729 always be either 0 or negative, since the treatments being compared (pH manipulation 730 either with or without heat shock) can only decrease the growth rate from maximal. To 731 ensure that the pH manipulation itself was minimally stressful, the relative growth of 732 pH-manipulated cells, which experienced 35 min at room temperature in calibration curve 733 buffer with ionophore, was calculated and was found to be extremely close to 0 for all pH 734 values considered (see figure S6c, 'RT (mock)' row).

735
To control for possible additional, strain-specific differences, we also calculated the 736 relative growth rate when both the wild-type and yCGT028 cells were treated identically 737 ('mix' or 'mix-in'); this value was also found to be nearly zero in every condition examined 738 (see Figure S6c, 'Mix-in' column).

739
Determination of budded fraction. We first computationally isolated the labeled, pulse width into 95% ethanol to fix, and then visualized the fixed cells using light 751 microscopy, Figure S7a shows sorting parameters and representative microscopy images.

752
Cells from both populations were scored as either budded (containing an obvious bud that 753 is at least 1/4 the size of the mother cell) or unbudded (having no bud). Full quantification 754 is shown in Figure S7b. Fixed cells were then stained with Sytox to assess cell cycle position 755 following a published protocol, 84 and DNA content was analyzed by fluorescence intensity 756 using flow cytometry. The 'budded' population contained more cells in the 2x DNA peak, 757 indicating that they were doubling their DNA and were thus actively growing; see Figure 758 S7c.

759
Code and data analysis 760 All data analysis was performed with R 85 using packages from the tidyverse. 86 Plots were 761 made with ggplot2. 87 Custom packages can be found on GitHub. Raw data and scripts 762 processing it to produce all figures that appear in this work are available online.

763
In general, summary lines on plots are moving averages, with the exception of Figure 1c, 764 and Figure 2b (same data) which were fit with local smoothing using the 'loess' method in 765 the ggplot2 87 function geom smooth(). In Figure 2d, the log-transformed data were fit to 766 sigmoids with the form: where a, b, c, and d are fitting parameters, and d is constrained to be greater than or 768 equal to 1.

769
Code for generating all processed data and plots is available in the supplemental 770 information.