Misfolded Proteins are Competent to Mediate a Subset of the Responses to Heat Shock in Saccharomyces cerevisiae 1

To this possibility, we the effect of misfolding on gene in the absence of temperature changes. The imino acid analogue azetidine 2-carboxylic acid (AZC) is incorporated into protein competitively with proline and causes reduced thermal stability or misfolding. We find that adding AZC to yeast, at sublethal concentrations sufficient to arrest proliferation, selectively induces expression of heat shock factor-regulated genes to a maximum of 27 fold and that these inductions are dependent on heat shock factor. AZC treatment also selectively represses expression of the ribosomal protein genes, another heat shock factor-dependent process, to a maximum of 20 fold. AZC treatment thus strongly and selectively activates heat shock factor. AZC treatment causes this activation by misfolding proteins: induction of HSP42 by AZC treatment required protein synthesis; treatment with ethanol, which can also misfold proteins, activates heat shock factor, but treatment with canavanine, an arginine analogue less potent than AZC at misfolding proteins, does not. However, misfolded proteins do not strongly induce the STRE regulon. We conclude that misfolded proteins are competent to specifically trigger activation of heat shock factor in response to heat shock.

Eukaryotic cells respond to heat shock by the induction of a conserved set of proteins, the heat shock proteins, via transcriptional activation of the corresponding genes (1). In the budding yeast, Saccharomyces cerevisiae, two distinct promoter elements mediate transcriptional activation in response to heat shock (reviewed in 2). Heat shock elements (HSEs) are found upstream of many heat-induced genes, e.g., HSP42 and SSA4. Heat-shock factor binds to HSEs and is required for the induction of HSE-driven genes in response to heat shock. Stress response elements (STREs) are also found upstream of many heatinducible genes, e.g., CTT1 and DDR1, and bind the transcription factors Msn2 and Msn4 (2). Loss of both transcription factors compromises heat shock-induced expression of STREcontaining genes. Some heat shock-inducible genes contain both HSEs and STREs in their promoters, e.g., HSP12, HSP30 and HSP104. The HSE and STRE regulons constitute the majority, if not all, of the genes that are specifically induced by heat shock (3).
Rapid upshifts in temperature within the permissive growth range of yeast (so-called "temperature upshifts"), e.g., 23 to 36 degrees Celsius, result in the transient and selective induction of the heat shock genes (both HSE-and STRE-containing) and in the transient and selective repression of the 137 ribosomal protein genes (4 and references therein). Heat shocks to non-permissive temperature (e.g., to 42 degrees Celsius) also cause a global repression of gene expression not seen upon temperature upshift (4). Repression of the ribosomal protein genes by heat shock is dependent on activation of heat shock factor,.
Because these genes do not contain HSEs, the repression is thought to be indirect (4).
The current model for how cells sense heat shock is as follows (6, and references therein). Heat shock is proposed to cause the thermal misfolding of a fraction of cell protein.
Because activation of heat shock factor requires protein synthesis, it is thought that nascent proteins are the most susceptible to thermal denaturation. Misfolded proteins then bind to cytoplasmic Hsp70 chaperones. Prior to heat shock, these chaperones are believed to equilibrate between being bound to heat shock factor (and inactivating it) and being free in the cytoplasm. Because misfolded proteins bind Hsp70s very tightly, their accumulation upon heat shock is proposed to titrate Hsp70 chaperones resulting in liberated and active heat shock factor. Consistent with this model, misfolded proteins have been detected in by guest on March 23, 2020 http://www.jbc.org/ Downloaded from

Materials and chemicals
All chemicals were from Sigma-Aldrich Company (Dorset, UK) Components of growth media were from Becton Dickinson (Sparks, Maryland, USA) and Fisher Scientific UK (Leicester, UK). D/L-AZC was used throughout this work but L-AZC is the active agent in this racemic mixture 3 . Any given concentration of AZC herein refers to the concentration of the racemic mixture.

Plasmids, yeast strains and manipulations
Liquid media, both rich and minimal, were prepared as described previously (16).
Solid media contained 2% agar. Strains were routinely grown on YPD agar plates at 25 o C.
Liquid cultures were grown in YPD broth.

Northern analysis
Extraction of total RNA was performed as previously described (Ogas et al, 1991

RESULTS
In this study, we set out to determine if AZC treatment mimics the effect of heat shock on genome-wide gene expression and, if so, to explore the mechanism by which the analogue causes these effects.
Laboratory strain backgrounds differ in their sensitivity to AZC, as judged by colony formation on plates (15) 2 . The concentrations used herein are those just sufficient to permanently inhibit proliferation of each strain background under study, unless stated otherwise. All experiments were conducted in the presence of normal amounts of proline in the growth medium.
We determined the expression changes of the 6129 protein-encoding genes by microarray analysis (see experimental procedures) using wild-type W303-1A cells treated with a growth-inhibitory concentration of AZC (50 mM) for five hours. We have also performed an equivalent analysis after one hour of AZC treatment (50 mM). The result of the latter experiment was qualitatively similar to the five hour data set but with less potent expression changes 2 . Our subsequent analysis has therefore focused on the five hour data set.

AZC treatment does not cause starvation.
AZC is an analogue of proline and may thereby interfere with proline uptake or metabolism, resulting in starvation. Nutrient-starved cells or cells treated with rapamycin, an inhibitor of the TOR proteins, stop proliferation in the G1 phase of the cell cycle and enter quiescence-like, non-proliferating states (20) superficially reminiscent of AZC-arrested cells (15). We find by microarray analysis that AZC treatment ( We determined the correlations between the microarray-derived expression profiles of AZC-arrested cells (5 hr, this work) and temperature-upshifted cells (23 o C to 37 o C for 10 min, 20 min and 40 min) (18). This analysis focused on those genes (2470 in total) of known function because data for this subset upon temperature upshift are publicly available (18).
We find that the correlation between AZC (5 hr) and temperature upshift microarray data sets is highest for the 20 min timepoint of temperature upshift, peaking at 0.576. This correlation compares well with those between temperature upshift timepoints: 10 min versus 20 min -correlation of 0.793; 10 min versus 40 min -correlation of 0.582 (18) 3 . These data suggest that AZC treatment selectively causes the majority, but not all, of the gene expression changes characteristic of temperature upshift. We conclude that AZC treatment partially mimics temperature upshift.
AZC treatment selectively causes heat shock factor-dependent gene expression changes.
What genes are selectively induced by AZC treatment? Expression of a subset of genes was strongly induced by treatment with AZC (see Table 1  Activation of heat shock factor directly or indirectly causes repression of the ribosomal protein genes (4). If AZC treatment indeed activates heat shock factor, then we expect treatment with the analogue to also repress expression of the ribosomal protein genes.
Expression of a subset of genes (293 in total) was repressed by a factor of three or more (to a maximum of 20 fold) by AZC treatment (see Table 1 for a partial list and supplementary data for the complete list). Of the 50 genes of known function that are repressed most strongly by AZC treatment, 47 are also strongly repressed by temperature upshift (18). The majority of these (42 out of 47) encode ribosomal proteins and repression of these genes by severe heat shock is dependent on heat shock factor (4). Our data suggest that AZC treatment, like temperature upshift, selectively activates heat shock factor and thereby causes increased expression of HSE-containing genes and repression of the ribosomal protein genes.
Of the 50 genes of known function whose expression is repressed most strongly by temperature upshift (18), 46 are also strongly repressed by AZC treatment. Again, the majority of these genes encode ribosomal proteins. Thus AZC treatment selectively causes the vast majority of the repressions caused by temperature upshift.

AZC treatment fails to strongly induce expression of the STRE regulon.
Almost all of the genes whose expression is induced by temperature upshift are (3). We conclude that the STRE regulon is at best weakly activated by AZC treatment. This possibility is supported by our previous finding that AZC treatment fails to cause accumulation of glycogen and trehalose, phenomena dependent on activation of the STRE pathway (15). We conclude that the STRE regulon is much less sensitive to AZC treatment than is the HSE regulon.
To a first approximation, temperature upshift selectively induces the STRE regulon and the HSE regulon (and thereby represses expression of the ribosomal protein genes), whereas AZC treatment selectively induces only the latter.

AZC treatment causes few gene expression changes not caused by temperature upshift.
Although incorporation of AZC into any given protein molecule is likely to cause dysfunction of that protein, the efficiency of incorporation of AZC in place of proline in our experiments is likely to be low (15) (see discussion). Hence, most molecules of any particular protein in AZC-treated cells are unlikely to contain AZC and thus are functional.
However, proteins that are large, proline-rich and short-lived are most likely to incorporate at least one residue of AZC in place of a proline. Thus AZC treatment may preferentially inactivate a particular subset of proteins within the cell and thereby directly cause gene expression changes not caused by temperature upshift. Surprisingly, we find only a handful of genes in this set: CUP1-1, CUP1-2 and ICS3 are induced by AZC treatment and not temperature upshift; ACS2, RPN8, CTS1 and BUD2 are repressed by AZC treatment but not by temperature upshift (18, see Table 1 for a partial list and supplementary data for the complete list of genes whose expression is affected by a factor of 3 or more by AZC treatment). Therefore, the majority of the gene expression changes caused by AZC treatment are not due to selective inactivation of any particular subset of proteins by the analogue. If AZC treatment and temperature upshift cause widespread misfolding of cell proteins, then we would expect these treatments to activate the unfolded protein response pathway of the endoplasmic reticulum (the ER-UPR pathway). This signalling pathway is activated by the accumulation of misfolded proteins in the endoplasmic reticulum and promotes transcription (and thus expression) of genes containing unfolded protein response elements (UPREs) in their promoters (19). Significantly, we find that genes that are strongly induced by the ER-UPR pathway, e.g., EUG1, PDI1, and LHS1, are not strongly induced by AZC treatment (1.15-fold, 1.7-fold and 1.63-fold respectively) 2 . Expression of these genes is also not strongly induced by temperature upshift (18). These findings are consistent with AZC treatment and temperature upshift causing only a low level of protein misfolding in the endoplasmic reticulum (and presumably throughout the cell).

AZC treatment strongly and robustly causes heat shock factor-dependent gene expression changes.
We wished to confirm the salient features of the gene expression changes detected from microarray profiling by Northern blot analysis. We also wished to compare the magnitudes of these expression changes with those caused by temperature upshift. For this analysis, we used a wild-type haploid strain of the S288C background whose proliferation in YPD is inhibited by 10 mM AZC (15). We determined the expression level of the following genes as a function of time after addition of AZC (10 mM) or upon temperature upshift However, the promoters of these genes fall into two classes, those containing potential binding sites for the Rap1 transcription factor and those containing potential binding sites for the Abf1 transcription factor. We included representatives of both subclasses for completeness. Equal amounts of total RNA were loaded in each lane, and blots were hybridised with the same probe at the same time, and exposed to the same film for equal amounts of time for each probing.
The results of the Northern analysis are shown in Fig. 1a Figure 2a,b, we find that these HSE-containing genes are indeed transiently induced in this strain background in response to AZC treatment.
Furthermore, we find that expression of HSP42 is induced more strongly in the EXA3-1 mutant compared to the wild-type strain in response to both AZC treatment and temperature upshift. Expression of HSP42 is also more persistent in the mutant in response to both treatments. Although there may be subtle effects of the EXA3-1 mutation on the extent of SSA4 induction in response to both AZC treatment and temperature upshift, expression of this gene is clearly more persistent in the mutant in response to both treatments. The EXA3-1 mutation in HSF1 thus affects the extent or persistence, or both, of induction of HSP42 and SSA4 in response to AZC treatment and temperature upshift. These data indicate that AZC treatment and temperature upshift cause induction of HSP42 and SSA4, and by inference all HSE-containing genes, by the same mechanism, namely activation of heat shock factor.
The induction of HSP42 and SSA4 by AZC treatment is clearly transient for the wildtype strain used in Figure 2, in contrast to the equivalent data for the S288c wild-type strain shown in Figure 1. The persistence of heat shock factor activation in response to AZC treatment in these strain backgrounds clearly parallels the persistence of proliferation arrest caused by the analogue, which gratifyingly is also dependent on heat shock factor (15).
Although the reason for the different responses of these two strain backgrounds to the analogue are not known, it is likely to be due to some combination of differences in the rate of uptake or efflux of the compound, to different efficiencies of incorporation of the analogue into protein or to different capacities to degrade analogue-containing peptides.
From our data in Figures 1 and 2, we cannot compare the relative expression levels of HSP42 and SSA4 between these two strain backgrounds since different batches of labelled probes were used for each experiment.

Induction of HSP12 by AZC treatment is partly dependent on Msn2 and Msn4.
From the microarray analysis above, STRE-containing genes are at best poorly induced by AZC treatment (  Figure 1a,b), we set out to determine if Msn2 and Msn4 contribute to induction of HSP12 upon AZC treatment. We find that the induction of HSP12 by AZC treatment and by temperature upshift is significantly reduced in a strain deleted for both MSN2 and MSN4 relative to its congenic wild-type strain (Figure 3a,b). We conclude that Msn2 and Msn4 (and by inference the STRE regulon) may indeed be activated by AZC treatment, but only partially.

Induction of HSP42 by AZC treatment is dependent on protein synthesis.
If AZC exerts its effects in vivo by misfolding proteins into which it is incorporated, then the expression changes caused by AZC treatment should require ongoing protein synthesis. We therefore determined the effect of cycloheximide addition on the ability of AZC treatment to induce expression of HSP42. Treatment of S288c wild-type cells with cycloheximide alone or vehicle alone does not alter expression of HSP42. In contrast, we find that the presence of cycloheximide prevents induction of HSP42 by AZC treatment (Figure 4a). We infer that ongoing protein synthesis is required for AZC treatment to induce the HSE regulon, consistent with the analogue functioning via misfolding nascent proteins into which it is incorporated.

Ethanol treatment mimics AZC treatment.
If AZC acts via misfolding cellular proteins, then we expect other treatments that misfold proteins in the cell to induce the same spectrum of gene expression changes as that caused by AZC treatment.
Ethanol can disrupt protein folding by a mechanism distinct from that of AZC (5).
The results of the Northern analysis of ethanol-treated cells are shown in Fig 4a. These blots, and those for canavanine-treated cells (see next section, Figure 4b) were prepared identically to, and probed with the same batch of labelled probe at the same time and exposed to the same film for the same length of time as, the blots shown in Figure 1a,b. Thus, the data shown in Figures 1 and 4 are directly comparable. The concentration of ethanol used in this experiment (8% v/v) is just sufficient to stop proliferation of the strain used (S288C wild type) 2 . We find that ethanol treatment mimics AZC treatment: 1) in strongly inducing the expression of genes regulated by heat shock factor, e.g., HSP42, SSA4, HSP12 and HSP30; 2) in failing to strongly induce CTT1 (a STRE-driven gene) in agreement with the very weak activation of STRE-driven genes by ethanol reported previously (26); 3) in failing to repress ACT1 (i.e., no global repression); 4) in strongly repressing the expression of the ribosomal protein genes tested. We conclude that ethanol and AZC treatments, both of which can misfold proteins but by very distinct mechanisms, cause similar gene expression changes attributable to activation of heat shock factor.

Canavanine treatment does not mimic AZC treatment.
Canavanine is an arginine analogue that, like AZC, is incorporated into protein competitively with the corresponding natural amino acid (27). Canavanine differs from arginine in the structure of its sidechain and, as such, is not expected to significantly alter the conformation of the polypeptide backbone into which it is incorporated (relative to the same protein containing arginine at the equivalent position(s)). Thus, treatment with canavanine is not expected to cause protein misfolding, at least not to the same extent as does AZC treatment.
We examined the effect of canavanine at a sub-lethal concentration (10 mM) just sufficient to inhibit cell proliferation, on gene expression by Northern analysis. Our results are shown in Figure 4b. In contrast to ethanol and AZC treatments, canavanine treatment does not significantly induce expression of the HSE-containing genes HSP42 and SSA3.
Canavanine treatment also fails to significantly repress expression of the ribosomal protein genes. We infer that canavanine treatment does not strongly activate heat shock factor.
Canavanine treatment also fails to induce expression of CTT1 and is thus unlikely to strongly activate the STRE regulon. It should be noted that in our experiments, both AZC and canavanine are present in the growth medium at concentrations sufficient to inhibit proliferation 2 , yet only AZC treatment induces the HSE regulon. Thus, activation of heat shock factor by AZC treatment is not a consequence of growth inhibition per se. Rather, the ability of the analogues, AZC and canavanine, to activate heat shock factor correlates with their relative capacity to misfold proteins into which they are incorporated.  (HSP12 and HSP30) but too weakly to drive expression of genes dependent on one or other system.

AZC causes protein misfolding in vivo.
Multiple lines of evidence indicate that AZC exerts its effects on yeast cells via Finally, ubc4.ubc5 mutants, which are defective in the ubiquitin-dependent degradation of short lived and analogue-containing polypeptides, are hypersensitive to AZC treatment (15).
The simplest explanation for this observation is that, at a given concentration of AZC, the amount of analogue-containing protein in the cell is higher when these proteins are stable than when these proteins are unstable. Thus, a concentration of AZC that is insufficient to arrest cell proliferation in a wild-type cell would be sufficient to arrest a cell lacking Ubc4 and Ubc5. Taken together, these arguments strongly indicate that AZC acts via incorporation into cellular protein.
Given that incorporation of AZC into protein is known to cause reduced thermal stability or misfolding (12)(13)(14), the effects of the analogue on gene expression are most likely due to its misfolding proteins. Two lines of evidence support this notion. First, induction of the HSE regulon is also strongly and selectively caused by treatment with ethanol, another agent capable of misfolding proteins but by a mechanism different to that of AZC. Second, canavanine, an arginine analogue that is incorporated into protein competitively with arginine, does not induce the HSE regulon, whereas AZC does so efficiently. Canavanine incorporation is not expected to disrupt protein folding as efficiently as does incorporation of AZC (27). Thus, the relative of the analogues to induce the HSE regulon correlates with their capacity to misfold proteins into which they are incorporated.

AZC treatment selectively activates heat shock factor.
AZC treatment selectively causes the gene expression changes attributable to activation of heat shock factor. First, the expression level of only a small fraction of the protein-encoding genes are altered by a factor of three or more after five hours of treatment with an inhibitory concentration of AZC (8.2% affected in total: 3.5% induced; 4.7% repressed). Hence, AZC treatment does not cause any global changes in gene expression, but selectively affects expression of a discrete subset of genes. Second, HSE-containing transcripts predominate amongst those induced by AZC treatment. Third, the ribosomal protein genes (and co-regulated genes encoding components of the translation apparatus) comprise the vast majority of the genes that are strongly repressed by AZC treatment.
Repression of the ribosomal protein genes by heat shock is known to be dependent on activation of heat shock factor (4).
We have confirmed that AZC treatment activates the HSE regulon (induction of HSE-containing genes and consequent repression of the ribosomal protein genes) by Northern analysis. Furthermore, we find that AZC treatment activates the HSE regulon as strongly if not more strongly than does temperature upshift. We have also shown that a mutation in heat shock factor affects induction of the HSE-containing genes HSP42 and SSA4 in response to AZC treatment. Critically, the mutation alters the induction of these genes in the same way in response to either AZC treatment or temperature upshift. Thus, by guest on March 23, 2020 http://www.jbc.org/ Downloaded from AZC treatment strongly and selectively induces the HSE regulon by the same mechanism as does temperature upshift, namely by activating heat shock factor.

Misfolded proteins are competent to be mediate selective activation of heat shock
factor in response to heat shock.
Based on the above arguments, we conclude that the misfolding of a fraction of cell protein in the absence of temperature change mimics heat shock in selectively and strongly activating heat shock factor. Therefore, misfolded proteins illicit the appropriate cellular response, and do so sufficiently strongly and selectively, for them to be intermediates in the cellular response to heat shock. Given that misfolded proteins are known accumulate in heat shocked cells (7), They are competent to mediate at least part of the cellular response to heat shock.
Unfortunately, it is not yet known if misfolded proteins are kinetically competent to be intermediates in the heat shock response, i.e., that misfolded proteins accumulate sufficiently rapidly upon heat shock and that misfolded proteins cause activation of heat shock factor sufficiently quickly. Although activation of heat shock factor is slow in response to AZC treatment, it is likely that equilibration of the analogue into the cellular proline pool prior to incorporation into protein is slow. The issue of kinetic competence remains unresolved.
If misfolded proteins are intermediates in the cellular response to heat shock, then heat shock factor must be very sensitive to protein misfolding in cytoplasmic space.
Temperature upshift is a very mild environmental change and is unlikely to cause extensive protein misfolding. In addition, AZC treatment at concentrations sufficient to activate heat shock factor does not appear to cause widespread protein dysfunction: 1) AZC treatment almost exclusively affects the expression of a small and discrete subset of genes that are also induced by temperature upshift; 2) AZC-arrested cells are viability (15); 3) AZC arrest is reversible (15); 4) AZC-arrested cells are responsive to subsequent treatments, e.g. heat shocks (15) and rapamycin 2 . Indeed, neither AZC treatment nor temperature upshift strongly activates the ER-UPR, even though both treatments should misfold proteins throughout the cell, including in the endoplasmic reticulum.
The fraction of protein containing AZC (when cells are treated with a concentration of the analogue just sufficient to activate heat shock factor) should constitute an upper limit for the fraction of cellular protein whose misfolding is just sufficient to activate heat shock factor. We are attempting to determine this number.
The STRE regulon is relatively insensitive to protein misfolding.
Although AZC treatment profoundly activates the HSE regulon and genes whose expression is dependent thereon (e.g., the ribosomal protein genes), it at best weakly induces the STRE regulon. This possibility is supported by our observation that AZC does not lead to the accumulation of glycogen and trehalose, a STRE regulon-dependent phenomenon (15), nor does it significantly activate expression of STRE-LacZ reporter constructs 3 .
However, AZC incorporation into cell protein does not appear to affect the activateability of the STRE regulon (15). Rather, AZC treatment simply fails to strongly activate this regulon.
The primary signal for activation of the STRE regulon by heat shock may be the misfolding of cellular protein, but with the STRE regulon requiring higher levels of protein misfolding than those sufficient to activate heat shock factor (and inhibit proliferation). Indeed, the timecourse of activation of CTT1 expression by temperature upshift parallels that of the heat shock factor-dependent transcripts, consistent with the notion of a common trigger.
Alternatively, the STRE pathway may primarily respond to heat-induced oxidative stress or some other stress that coincides with protein misfolding upon heat shock (2,5). It is clear that Msn2 and Msn4 contribute to the induction of HSP12 by AZC treatment. Given that the STRE regulon is activated by multiple stresses to the cell, it is possible that any partial activation of the regulon in response to analogue treatment is caused by an indirect mechanism, e.g., because of proliferation arrest. The mechanism by which the STRE regulon is activated by heat shocks remains elusive.

How do cells sense heat shocks?
Thermally-misfolded protein likely triggers activation of heat shock factor in response to heat shocks. The sensor for activation of the STRE regulon upon heat shock remains unclear. However, misfolded protein is not the sole sensor of heat shock in yeast.
The cell integrity pathway, which is required for acquired thermotolerance and for   Samples were prepared and analysed as for the AZC treatment above. Blots were probed with the same probes and exposed to the same film and for the same times for each probe.     Supplementary table legend   Table S1: Genes whose expression is altered upon AZC treatment by a factor of three or more. Microarray analysis was performed on wild-type cells (W303) grown to midlogarithmic phase in YPD at 30 o C and treated with AZC (50 mM) for 5 hours. Expression changes were determined by probing DNA microarrays as outlined in the main text.
Genes whose expression was induced or repressed by a factor of three or more (