Transcriptional Factor Mutations Reveal Regulatory Complexities of Heat Shock and Newly Identified Stress Genes in Saccharomyces cerevisiae *

A computer-aided pattern search of the entire yeast genome was designed and used to identify 186 putative stress response element-regulated genes in Saccharomyces cerevisiae. Transcript levels of eight of these candidate genes were examined, and three (37%) were shown to be heat shock- and DNA damage-inducible and to require the Msn2p and Msn4p transcriptional activators for stress regulation. Significantly, several heat shock protein (HSP) genes were identified in this computer search. Using a series of single and multiple regulatory mutants, we demonstrate unexpected regulatory complexities among the HSP genes from S. cerevisiae following heat shock.

We have used this structural and functional information as the basis for a computer search of the entire yeast genome and have generated a list of 186 potential STRE-regulated genes. Using Northern hybridization to analyze the regulation of eight of these genes, we show that three of these loci are stressregulated and require Msn2p/Msn4p for their activation. Among the genes identified in the STRE pattern search were several yeast heat shock protein (HSP) genes assumed to be regulated exclusively via the heat shock factor/heat shock element (HSF/HSE) pathway. Using a series of multiple regulatory mutants, we demonstrate that several HSP genes are redundantly controlled by both the Msn2p (Msn4p)/STRE and HSF/HSE pathways following heat shock.

MATERIALS AND METHODS
Strains, Growth Medium, and Chemicals-Yeast strains used in this study are listed in Table I. Phenotypes were determined by growing cells on SC drop mix minimal plates (10). Cells were grown in YPD (10) for stress treatments.
Construction of Double and Triple Mutant Strains-Strains MCY2144 (msn2⌬) and MCY2146 (msn4⌬) were mated to create the diploid strain JT615, which was sporulated as described previously (11). The sporulated culture was treated exhaustively with glusulase to disrupt the asci and vortexed with glass beads to separate the spores (10). The spores were plated onto SC his Ϫ , SC ura Ϫ , and SC his Ϫ ura Ϫ plates to recover msn2⌬, msn4⌬, and msn2⌬ msn4⌬ haploid strains, respectively. From each mating and subsequent sporulation, 20 red ade Ϫ haploid colonies were isolated, restreaked to isolate individual colonies, and patched to SC plates lacking one of several nutrients to determine their phenotypes. The resulting strains, JT616 (msn2⌬), JT620 (msn4⌬), and JT624 (msn2⌬, msn4⌬), were subsequently mated to strain MYY385 (mas3) to generate diploid strains JT628, JT629, and JT632, respectively. These three strains were sporulated and haploid ade Ϫ colonies containing the appropriate markers were identified as described above. The presence of the disrupted msn2⌬ and msn4⌬ alleles was confirmed by Southern hybridization analysis of genomic DNA. Temperature-sensitive growth at 37°C and the loss of heat shock induction of the SSA3 transcript were scored to confirm the presence of the mas3 mutation.
Stress Treatments-Cells were subjected to stress treatment as described previously (4). Briefly, cells were exposed to 0.07% methyl methanesulfonate (MMS) for 60 min. For post-diauxic shift (PDS), cells were harvested 1-2 h after the diauxic shift from fermentative to respiratory growth. The diauxic shift was determined by monitoring optical density of the culture and was reached approximately 12-14 h after the control sample was harvested at a cell density of ϳ7 ϫ 10 6 cells/ml. For heat shock (HS) treatment, cultures were grown at 23°C, and an aliquot was shifted to 37°C for 20 min.
RNA Preparation and Northern Hybridization-Total RNA was prepared as described previously (11). Denatured RNA (25 g) was electrophoresed in formaldehyde-agarose gels and transferred to nylon membranes. RNA blots were UV-cross-linked and hybridized with [␣-32 P]dCTP-radiolabeled probes and washed as described (4). The Northern blots were exposed to x-ray film and quantitated using an Ambis radio-imager system (Scanalytics). The probes used for hybridization were the 1.45-kb HindIII fragment of DDR2 from plasmid pBRA2 (12), the 1.3-kb PstI-BglII fragment of HSP26 from plasmid pS26-1 (S. Lindquist); the 3.2-kb HindIII fragment of UBI4 from plasmid pUC-UBI4, the 1.3-kb HindIII fragment of HSP78 from pLS12 (T. Mason), and the 1.4-kb HindIII-SstI fragment of HSP104 from plasmid pYS104 (13). The remaining probes were polymerase chain reactionamplified from selected coding regions of appropriate genes using genomic DNA from individual M12B colonies as templates. The regions amplified were: nucleotides ϩ1 to ϩ714 of ACT1, ϩ142 to ϩ768 of PGM2, ϩ599 to ϩ1120 of PDE2, ϩ249 to ϩ694 of YKL151C, ϩ44 to ϩ501 of RPB4, ϩ1558 to ϩ1930 of SSA3, ϩ1 to ϩ326 of HSP12, and ϩ1 to ϩ424 of HSP82.
Computer-aided Pattern Matching-The PatScan pattern matcher, maintained at Argonne National Laboratory (http://www-c.mcs.anl.gov/ home/overbeek/PatScan/HTML/patscan.html), was used to search the yeast genome for candidate STRE-regulated genes. The search was performed against both strands of the entire genomic sequence of S. cerevisiae accessed from the EBI nucleotide sequence data base (Release 49). The following pattern was used for the search: CCCCT 1 . . . 200 CCCCT 1 . . . 300 TATA in which letters represent direct matches to nucleotides and numerals represent an interval between x and y nucleotides.

Identification of Novel STRE-regulated Yeast Genes by Computer-aided Pattern
Matching-We demonstrated previously that C 4 T pentanucleotides function as stress response elements independently of nucleotide sequence context or spacing (4). This observation suggested that we could identify stress-controlled genes based upon the presence of multiple upstream STREs. Using a computer-aided pattern matching strategy, we searched the fully sequenced S. cerevisiae genome for candidate genes containing two or more STREs separated by up to 200 bp and lying within 300 bp of a TATA sequence. The search returned approximately 350 matches and of these, 186 matches were found to define groups of STREs located immediately upstream of characterized genes or predicted open reading frames (within 500 bp of the translation start). Eighty-three of these candidates corresponded to known yeast genes, while the remaining 103 sequences represented uncharacterized open reading frames. A diverse group of genes was identified, including those encoding mitochondrially localized gene products (CIT1, MCR1, and MDH1), known or predicted transporter proteins (GAL2, MEP1, PUT4, and YGR224W), known or predicted proteases (PAI3, PRC1, STE23, and YBR139W), homologs of DnaJ (MDJ1, YNL077W, and YPR061C) and several HSP genes (HSP12, HSP26, HSP42, HSP78, and HSP104).
Eight candidates (PDE2, PGM2, RAS2, RPA14, RPB4, SMC2, YGR067C, and YKL151C; see Table II) were selected and tested for STRE control by Northern hybridization analysis. Genes under STRE control exhibit the following characteristics: (i) transcript levels are elevated after heat shock, DNA damage, or other stresses; (ii) stress induction is curtailed or greatly reduced in strains that lack the STRE-binding transcription factors, Msn2p and Msn4p; and (iii) basal transcript levels are significantly elevated in a strain overexpressing Msn2p (2). Based on these criteria, three of the eight candidates tested (PDE2, PGM2, and YKL151C) were found to be STRE-regulated (Fig. 1).
Recently, using two-dimensional protein gels, Boy-Marcotte et al. (14) have shown that 39 proteins induced during diauxic shift are dependent upon Msn2p/Msn4p for their expression.
Among the genes identified are several that were found in our pattern search, including HSP104, HXK1, PGM2, TPS1, and YBR149W. These results reinforce the notion of a large family of yeast stress genes that are co-regulated via conserved STREs. Transcription of the PDE2, PGM2, and YKL151C genes was stress-inducible in the wild-type strain, but no transcripts accumulated in the msn2⌬ msn4⌬ double mutant strain following stress. However, each STRE-controlled gene exhibited different characteristics with respect to stress induction in the various yeast mutants examined. Stress induction of PGM2 transcripts was completely abolished, even in the msn2⌬ single mutant strain. This result contrasted with the results observed for other genes, which showed residual levels of induction in the msn2⌬ mutant strain. DDR2, the positive control, exhibited residual heat shock induction in the msn2⌬ mutant strain, a result that agrees with earlier studies (2). YKL151C transcript induction appeared more complex. Stress induction was clearly reduced in the msn2⌬ strain, but was not abolished unless both MSN2 and MSN4 genes were disrupted. The PDE2 gene exhibited a behavior that was unique among all STRE-controlled genes studied. Stress induction of its transcripts appeared completely refractory to the msn2⌬ disruption, while accumulation was abolished in the msn2⌬ msn4⌬ double mutant strain. Previous studies showed that overexpression of Msn4p partially suppressed the defect in transcript induction occurring in an msn2⌬ strain (2). However, the dominance of Msn4p in the regulation of PDE2 demonstrates the role of the MSN4 gene in STRE-regulated gene expression when it is present as a single chromosomal copy. Thus, there appears to be substantial variation in the relative contribution of Msn2p versus Msn4p in the transcriptional control of different STRE-regulated genes. Perhaps with the characterization of more STRE-controlled genes from the pool of candidates, regulatory patterns will emerge that will allow prediction of MSN2 versus MSN4 con-trol based on sequence characteristics among and around the STRE elements.
Contribution of STREs in Heat Shock Regulation of Yeast Heat Shock Proteins-Five genes encoding known heat shock proteins were identified in the pattern search for STRE-controlled genes (HSP12, HSP26, HSP42, HSP78, and HSP104). Significantly, the promoters of these genes contain HSEs, which have been shown to be required for heat shock activation of several genes in yeast and higher eukaryotes (15)(16)(17)(18). To evaluate the contribution of the STREs to HSP gene regulation, we constructed a series of regulatory mutant strains containing the msn2⌬, msn4⌬, and mas3 (hsf1) mutations, singly or in combination. As reported previously (2), heat shock induction of DDR2 transcription was solely regulated by the STRE pathway as judged by Northern blots of RNAs from these different regulatory mutants (Fig. 2; Table III). For the SSA3 gene, encoding an Hsp70 cognate, regulation by heat shock was mediated exclusively through a functional HSF gene and showed no requirement for Msn2p or Msn4p. This result is consistent with earlier studies (19,20). The two small HSPs, HSP12 and HSP26, showed differential requirements for the STRE and HSE pathways. HSP12 was predominantly regulated via Msn2p/STRE after heat shock with very little heat shock induction attributable to the HSF/HSE pathway. Conversely, regulation of HSP26 expression after heat shock occurred predominantly via HSF/HSE with little contribution to heat shock regulation through its STREs. In each case, however, transcript accumulation was completely abolished following heat shock exposure of the triple mutant strain (msn2⌬, msn4⌬, mas3; Fig. 2; Table III). Thus, these two pathways accounted for all the heat shock induction of transcripts from these two genes.
For two HSP genes, HSP78 and HSP104, and the stressregulated polyubiquitin gene, UBI4, we observed that heat shock induction was redundantly regulated by the two pathways. As shown in Fig. 2, each pathway alone was capable of activating transcription to nearly wild-type levels. When both pathways were inactivated in the triple mutant strain, however, there was a substantial decrease in the accumulation of these transcripts following heat shock. However, for each of these genes, unlike HSP12 and HSP26, there was residual transcript accumulation in the triple mutant suggesting that additional transcriptional activators or pathways are involved in stress activation of these genes ( Fig. 2; Table III).
Surprisingly, for the HSP82 gene, inactivation of both the Msn2p(Msn4p)/STRE and HSF/HSE pathways had relatively little effect on transcript accumulation following heat shock ( Fig. 2; Table III). There was negligible contribution to the heat shock regulation of the HSP82 gene by the Msn2p(Msn4p)/ STRE pathway, while the mas3 mutation reduced HSP82 heat shock-induced transcript levels by about 40%. Thus, it appears that the Msn2p(Msn4p)/STRE pathway plays no role in heat shock induction of the HSP82 gene and that the HSF/HSE pathway plays a less significant role in HSP82 regulation following heat shock than previously thought (21).
The mas3 mutation, originally isolated by Smith and Yaffe (22) based upon its temperature-sensitive defect in mitochondrial protein import, is an amber mutation located in the portion of the gene encoding the trimerization domain of HSF. This mutation is suppressed in yeast strain MYY385 by an uncharacterized nonsense suppressor, which renders the protein thermolabile. The temperature-sensitive phenotype can be recapitulated by site-directed mutagenesis of the position corresponding to the mas3 amber mutation within the HSF trimerization domain. Moreover, HSF protein levels remain unchanged after heat shock of the mas3 mutant. 2 These results indicate that the temperature sensitivity associated with the mas3 allele is due to an intrinsic thermolability of this mutant transcription factor.
The residual accumulation of transcripts following heat shock of the mas3 mutant strains is not likely to result from incomplete inactivation of HSF upon the temperature shift. HSF contains two regions responsible for transcription activation following heat shock stress (23). Thiele and co-workers (24) have shown that CUP1 transcription is induced by heat shock and requires a carboxyl-terminal activation domain in HSF, while heat shock activation of SSA3 gene expression requires an amino-terminal activation domain (23,24). Our data indicate that the mas3 mutation inactivates both domains as shown by the failure to induce SSA3 transcription following heat shock (Fig. 2) and our finding that the mas3 allele blocks CUP1 expression at 39°C (data not shown). Taken together, these results argue that the transcript accumulation observed in the mas3 mutants was not due to residual HSF activity but was due to a distinct regulatory mechanism(s).
The redundant nature of HSP gene regulation following heat shock was demonstrated using the multiple regulatory mutant strains we constructed. Indeed, there were several HSP genes where inactivation of all three transcription factors (HSF, Msn2p, and Msn4p) did not completely abolish transcript induction. Several explanations for the residual accumulation of transcripts in the msn2⌬ msn4⌬ mas3 triple mutant are possible. First, a third independent stress regulatory pathway might account for heat shock induced expression of these genes. However, sequence comparisons of the upstream region of several of these genes did not reveal any consensus elements, other than the HSEs and STREs, which might identify this hypothetical pathway (data not shown). Second, it remains possible that the observed transcript levels increased because of transcript stabilization rather than increased transcriptional initiation. Such a mechanism might be expected to produce kinetic differences in the rates of transcript accumulation between mutant and wild-type cells. Our preliminary results indicate that there were no meaningful differences in the kinetics of transcript accumulation between the mas3 mutant and wild-type (data not shown). Last, residual transcripts may result from the action of another transcription factor acting through the STRE. Consistent with this explanation, we have obtained preliminary evidence that all HSF-independent heat shock-induced transcription of the UBI4 gene requires at least one intact STRE (data not shown). One candidate for this putative transcription factor (YER169W) was identified during an expression library screening using an STRE-containing oligonucleotide (2). This factor bound STREs specifically but is structurally unrelated to Msn2p except for the presence of a zinc finger domain. However, disruption of YER169W, singly or in combination with mutations in the HSF1, MSN2, and MSN4 genes, had little or no effect on heat shock induction of several STRE-containing genes (data not shown). Nevertheless, we cannot rule out the possibility that one or more additional zinc finger transcription factors act to control residual-induced expression of these stress response genes.
STREs May Function for a Subset of Stresses for Some Genes-Transcription of five candidate genes identified from the computer pattern search was noninducible following heat shock or MMS and showed no dependence upon MSN2 or MSN4 for their expression (data not shown). Additionally, the 2 M. P. Yaffe, personal communication. FIG. 3. Effect of msn2⌬ and msn4⌬ mutations on PDS induction. RNA (25 g) from control and PDS grown cells was analyzed by Northern hybridization. After washing, the hybridization was quantified using an Ambis radio-imager as described in Table III Fig. 2 were quantified using an Ambis radioimager. The amount of message for each gene was normalized to the amount of actin message present in each lane. The normalized amount of message remaining in each lane was expressed as a percentage of the wild-type set as 100%.
STREs in the HSP26 promoter were not functional following heat shock. To determine whether the STREs might function in response to different stress conditions, we tested the requirement for Msn2p and Msn4p for transcriptional activation of the HSP26 gene during PDS growth. As seen in Fig. 3, the transcriptional induction of the HSP26 gene during PDS was dependent on the Msn2p and Msn4p transcriptional activators. Furthermore, carbon source starvation induction of HSP26 required the Msn2p (Msn4p)/STRE pathway (1). Similarly, Inoue et al. (25) have shown that GLO1-lacZ expression was controlled via the Msn2p/Msn4p pathway only following osmotic shock. There was no inducible expression of the same GLO1-lacZ fusion gene by other stresses such as heat shock, ethanol, or hydrogen peroxide, which have been shown to mediate their effects on the transcription of the CTT1 and DDR2 genes via upstream STREs (1, 2, 4). Thus, for some genes, such as DDR2, transcriptional control following all stress conditions tested is under the regulation of the STREs and Msn2p/Msn4p (4). However, function of STREs in the promoters of other genes such as HSP26 may be more limited. Presently, we do not understand the reason for the "inactivity" of STREs in the upstream regions of some genes following specific stresses. It is possible that the binding of additional regulatory factors to neighboring elements in the promoters of some genes, such as HSP26, may interfere with or modulate the binding of Msn2p or Msn4p to the STREs. Analysis of the STREs and their possible interactions with other promoter elements or regulatory factors should help us understand the relationship between STRE and dependence upon the Msn2p and Msn4p transcriptional regulators following stress.