The transcriptional response of Saccharomyces cerevisiae to osmotic shock. Hot1p and Msn2p/Msn4p are required for the induction of subsets of high osmolarity glycerol pathway-dependent genes.

We have analyzed the transcriptional response to osmotic shock in the yeast Saccharomyces cerevisiae. The mRNA level of 186 genes increased at least 3-fold after a shift to NaCl or sorbitol, whereas that of more than 100 genes was at least 1.5-fold diminished. Many induced genes encode proteins that presumably contribute to protection against different types of damage or encode enzymes in glycerol, trehalose, and glycogen metabolism. Several genes, which encode poorly expressed isoforms of enzymes in carbohydrate metabolism, were induced. The high osmolarity glycerol (HOG) pathway is required for full induction of many but not all genes. The recently characterized Hot1p transcription factor is required for normal expression of a subset of the HOG pathway-dependent responses. Stimulated expression of the genes that required the general stress-response transcription factors Msn2p and Msn4p was also reduced in a hog1 mutant, suggesting that Msn2p/Msn4p might be regulated by the HOG pathway. The expression of genes that are known to be controlled by the mating pheromone response pathway was stimulated by osmotic shock specifically in a hog1 mutant. Inappropriate activation of the mating response may contribute to the growth defect of a hog1 mutant in high osmolarity medium.

The ability to adapt to changes in the osmolarity of the surrounding medium is fundamental to life, and the accumulation of compatible solutes to decrease intracellular water potential is an adaptation strategy employed by all cell types (1). The production of compatible solutes is controlled at the level of gene expression in yeasts (2,3), plants (4), and mammals (5). To properly control gene expression, the cell has to sense osmotic changes and transmit the signal to the nucleus.
It appears that many if not all eukaryotic cells employ mitogenactivated protein (MAP) 1 kinase pathways for this purpose (3,6,7).
The best characterized osmosensing MAP kinase pathway is the high osmolarity glycerol (HOG) pathway in the yeast Saccharomyces cerevisiae (3,6). Two different transmembrane proteins, Sho1p and Sln1p, appear to serve as independent osmosensors (8,9). Sho1p transmits its signal via Ste50p and Ste20p to the MAPKKK Ste11p and further to the MAPKK Pbs2p (9 -12). Sln1p together with Ypd1p and Ssk1p forms a phosphorelay system that transmits the signal to a redundant pair of MAPKKK, Ssk2p and Ssk22p, and further to the MAPKK Pbs2p (6,8,13,14). The two branches of the pathway converge at the level of Pbs2p, which also seems to play a role as a scaffold protein (6,9,10). Pbs2p phosphorylates the MAPK, Hog1p, leading to translocation of Hog1p to the nucleus (8,(15)(16)(17)(18). Protein interaction and phosphotransfer in the HOG pathway and the nuclear translocation process of Hog1p have been studied in great detail (3,6,18).
The stimulation by osmotic shock of Hog1p phosphorylation is rapid and transient (8). 2 Within about 1 min after osmotic upshift (depending on severity of the shock) tyrosine phosphorylation of Hog1p and nuclear translocation are apparent (8). 2 After a mild osmotic shock (0.5 M NaCl), Hog1p remains phoshorylated and located in the nucleus for about 10 min (8,16,17,19). At this time point, an increase in the mRNA level of osmo-induced genes becomes apparent (19,20). The induction by osmotic shock of many genes studied so far is characterized by a sharp peak, which is observed for instance 45 min after an osmotic shock of 0.7 M NaCl (19,20). The transient nuclear localization of Hog1p and detailed analysis of the time courses of osmotic induction of certain genes in a hog1 mutant suggest that Hog1p may be specifically involved in the early induction phase of gene expression (16,17,19,20).
Recently, transcription factors that are required for the stimulation of gene expression after an osmotic shock have been described. Msn2p/Msn4p are a pair of apparently redundant transcription factors that bind to stress response elements (STREs) (21)(22)(23). These elements are characterized by the core sequence CCCCT in either orientation and are usually found in two or more copies in front of Msn2p/Msn4p target genes (24,25). Msn2p/Msn4p appear to control a general stress response because target genes are induced by heat and osmotic shock, oxidative stress, and nutrient starvation (21, 22, 26 -28). Upon stress, Msn2p/Msn4p are phosphorylated and translocated to the nucleus, events that are influenced by protein kinase A (29,30). The induction by osmotic shock of an artificial promoter, whose only activating sequence was a repeat of STREs, has been shown to depend on the HOG pathway (31), suggesting that in addition to protein kinase A Hog1p may also control Msn2p/Msn4p activity. However, this possibility has not been directly addressed experimentally.
We have recently identified Hot1p, another putative transcriptional regulator involved in osmo-induced gene expression. Hot1p interacts in the two-hybrid system with the Hog1p kinase, and further genetic evidence suggests that Hot1p is part of the HOG pathway (19). The induction after osmotic shock of genes encoding enzymes in glycerol production, GPD1 and GPP2, is dramatically diminished in a hot1 mutant (19). Binding of Hot1p to DNA has not yet been demonstrated. Stimulated expression of some HOG pathway-dependent genes after osmotic shock appears to be the result of derepression (32). Sko1p has recently been described as a repressor of such genes, and genetic evidence places Sko1p downstream of Hog1p (33). 3 In this work we have used genome-wide transcriptional analysis to identify the genes whose induction after an osmotic shock was dependent on the HOG pathway, Hot1p and Msn2p/ Msn4p, respectively.
Cells were grown as batch cultures in YPD (1% yeast extract, 2% peptone, 2% glucose) to an A 600 of 1.0 (Ϯ 0.2). At this point cells were collected for determination of expression profiles under basal conditions. Osmotic stress was applied by addition of 5 M NaCl to a final concentration of 0.5 or 0.7 M NaCl or, alternatively, by collection and subsequent resuspension of cells in YPD containing 0.95 M sorbitol. Samples were taken at time points at which mRNA levels of representative osmoresponsive genes are maximal as previously revealed by Northern blotting. For the wild type these time points were 45 min for 0.7 M NaCl and 30 min for 0.5 M NaCl or 0.95 M sorbitol; for the hot1 mutant 45 min for 0.7 M NaCl; for the msn2 msn4 mutant 30 min for 0.5 M NaCl; for the hog1 mutant 45 min for 0.5 M NaCl (19,20).
mRNA Preparation and Synthesis of cDNA-Total RNA was isolated as described previously (38). [ 33 P]CTP-labeled cDNA was synthesized in the following way. 5 g of total RNA and 2 g of oligo(dT) (10 -20-mer mixture, Research Genetics) were mixed in 8 l of water, heated for 10 min at 70°C, and then chilled on ice. The following components were added: first strand buffer (Life Technologies, Inc.), dithiothreitol (3.3 mM), dATP, dGTP, and dTTP (1 mM each), Superscript II reverse transcriptase (300 units; Life Technologies, Inc.), and [ 33 P]CTP (100 Ci, 3000 Ci/mmol). The mixture (30-l volume) was incubated at 37°C for 90 min. The probe was then purified by passage through a Sephadex G-50 column. Approximately 60 -70% of the label was incorporated in high molecular weight products.
Genefilter® Hybridization-Genefilters® (Research Genetics) were prehybridized for 3-6 h with 5 ml of MicroHyb solution (Research Genetics) at 42°C in a roller oven (Hybaid). The purified cDNA probe was denatured for 5 min at 95°C and added to the prehybridization mixture. After overnight hybridization at 42°C filters were rinsed with 2 ϫ SSC, 0.1% SDS and incubated in the same buffer at 50°C for 45 min. Filters were then rinsed with 0.5 ϫ SSC, 0.1% SDS, transferred to a plastic box, and washed with 0.5 ϫ SSC, 0.1% SDS at room temperature for 15 min. A PhosphorImager (Fuji, BAS-1000) was used to obtain a digital image of the filters. Filters were kept moist to facilitate stripping between hybridizations, and stripping was done with 0.5% SDS, which was first heated to 100°C. Filters were rinsed with this solution and submerged in it for 15 min on a shaker while the solution was allowed to cool. More than 95% of the signal was consistently removed by this procedure.
Data Analysis-Images produced by MacBas® (Fuji) were converted to TIFF and imported into the Pathways® software (Research Genetics). Pathways® was used for a preliminary quantification of spot intensities to compare gene filter images pairwise. In this way the genes that were differentially expressed were identified. Spot intensities were then quantified using the original image files with the MacBas® program. We noted that information was lost during conversion to TIFF and that therefore the differences between conditions/strains calculated by Pathways® greatly underestimated the actual differences.
To determine the extent of induction or repression of gene expression, all spot intensities were normalized against IPP1 (which encodes inorganic pyrophosphatase), whose expression is not affected by osmotic stress (20). The mRNA level of IPP1 is very similar to that of ACT1 (encoding actin), which is commonly used as a control but fluctuates after osmotic shock (20). A detection threshold was set at 0.02fold the IPP1 mRNA level to avoid distortion of calculated induction or repression ratios by very small and unreliable values. The relative mRNA levels calculated in this way indicate whether a gene is strongly or poorly expressed. It should be noted, however, that hybridization intensity also depends on the length of the ORF and on that of the cDNA used for hybridization. This means that comparison of relative mRNA levels between different genes should be done with caution.
To determine the -fold induction or repression, the relative mRNA level after osmotic shock was divided by that before the shock. If basal expression was below the threshold, induction ratios were reported as "Ͼ(induced value/0.02)." We have listed all genes whose mRNA level after addition of salt and/or sorbitol was at least 3-fold higher or 1.5-fold lower than before the osmotic shock because those values were found to be reproducible and allowed inclusion of almost all genes known to be osmoregulated.
To give a quantitative impression of the effect on osmotic induction conferred by a certain mutation an "effect factor" was calculated. For this, the relative mRNA level after osmotic shock in the mutant strain was divided by that of the wild type (i.e. an effect factor of 0.25 is equivalent to 25%). Hence, the lower the effect factor the less the gene was expressed in the mutant relative to the wild type. It should be noted, however, that in many instances the mutations also affect the basal expression level. Hence, even if the effect factor was close to 0.00 it is possible that the gene was still induced in the mutant, although its expression level was more than 100-fold diminished. Usually, however, diminished expression in a mutant was also accompanied by diminished induction. We found that this representation of data gave a better reflection of the effects conferred by a mutation than comparison of induction factors.
Hybridization Specificity-We have evaluated the behavior of a number of highly similar isogenes to estimate the threshold sequence identity above which cross-hybridization makes data unreliable. GPD1- , and TDH1-TDH2 (90%) gave identical hybridization patterns. Therefore, we concluded that genes with a DNA sequence identity above 90% should not be considered in the analysis, genes below 83% identity can reliably be distinguished, and genes between 83 and 90% identity may show variable specificity. ALD2 (YMR170C, also previously called ALD5) and ALD3 (YMR169C, also previously called ALD4) are 92% identical (39). Both gave similar results in gene filter experiments and are also known to be both stressand osmo-induced (40,41). Hence the data concerning these two genes may be reliable. The three genes of the ENA/PMR2 gene cluster, ENA1, ENA2, and ENA5, are more than 98% identical (42) and are all induced by osmotic stress. Hence the data given for ENA1 are in fact cumulative for these three genes.
Reproducibility-The great majority of changes in expression found with the first set of filters was confirmed by repeating experiments with a second set of filters. Some cases where there were small changes in expression with only one set of filters were discarded. Comparison of expression profiles of independent cultures of wild type yeast grown under the same conditions (midlogarithmic growth phase in YPD) revealed no significant differences.

RESULTS
The Use of Genefilters® for Transcriptome Analysis-Data obtained from the Genefilter® analysis agreed very well with previous Northern blot data (19,20)    and simple way to perform genome-wide transcriptional analysis, especially where glass or chip-based microarrays are not yet available. A recent report using the same technology came to a similar conclusion (43). However, there are certain limitations. The gene set is incomplete, i.e. genes YPR131C through YPR204W (the "last" 74 genes on chromosome 16) were omitted and a number of smaller ORFs, which are now known to be expressed, are not represented. Overall, approximately 100 of the 6,200 genes are missing. None of these genes is known to be osmo-induced. A further limitation concerns the cross-hybridization of highly homologous genes (see "Materials and Methods"), which can largely be avoided by using oligonucleotide-  Transcriptional Response to Osmotic Shock 8293 based microarrays. Finally, the present analysis software did not allow reliable quantification in our hands, which was done manually on the original PhosphorImager data with the corresponding software (see "Materials and Methods"). The Expression of More than 300 Genes Is Rapidly Altered after Osmotic Shock-Our previous Northern blot analysis of osmo-induced genes such as GPD1, GPP2, CTT1, and HSP12 had shown that the mRNA level is highest 45 min after an osmotic shock with 0.7 M NaCl and 30 min after addition of sorbitol to a final concentration of 0.95 M (19,20). mRNA was extracted from wild type cells at these time points after a shift to high osmolarity medium and used for gene filter experiments. The mRNA level of a total of 186 genes was at least 3-fold higher after addition of NaCl and/or sorbitol and that of more than 100 genes was at least 1.5-fold lower (Tables I and  II). The 186 induced genes included all those previously reported to be induced after osmotic shock with the exceptions of HAL1 (32), GAC1 (44), and TPS3 (45). Perhaps induction of these genes follows a different time course and the peak of induction occurs later. In addition, induction of the genes UBI4 (46), ENO1, and TDH1 (47) could not be demonstrated because of the presence of identical or highly similar isogenes that cross-hybridized.
NaCl is known to induce a specific sodium toxicity response (48,49). Surprisingly, the expression of only 10 genes was stimulated more than 2-fold more strongly by salt than by sorbitol. Because the sodium-specific response of gene expression is stimulated by signaling pathways different from those controlling osmotic responses (48,49), it is possible that saltspecific genes follow a different stimulation time course. Unexpectedly, there were 26 genes whose expression was at least 2-fold more strongly stimulated by sorbitol than by salt, and 4 genes, CAR1, YDL222C, ALD6, and CDC55, were induced more than 4-fold more strongly by sorbitol. This observation indicates that either sorbitol induces specific responses or that NaCl has a negative effect on the expression of some osmoinduced genes.
We grouped the induced as well as the repressed genes into functional categories to identify biochemical processes possibly controlled by osmotic shock (Tables I and II) (see "Discussion"). Ribosomal protein genes were not listed individually (Table II). Some induced genes were placed in more than one functional category. Eighty of the induced and 15 of the repressed genes have not been studied so far. Among those uncharacterized ORFs are some of the most strongly induced genes, such as for instance YDL223C, YGL037C, and YML128C. The majority, however, of the uncharacterized osmo-induced genes are neither very strongly regulated nor highly expressed.
A Subset of the Osmo-induced Genes Depends on Hog1p-To identify the genes whose rapid induction by osmotic shock was dependent on the HOG pathway, we performed gene filter experiments with mRNA isolated from the hog1 mutant 45 min after the addition of salt to a final concentration of 0.5 M NaCl. These conditions were chosen because we previously observed that osmotic induction of genes such as GPD1 in a hog1 mutant was (i) diminished, (ii) only observed at relatively low NaCl concentrations, and (iii) the peak of induction was shifted to a later time point (20). Table III lists the 48 genes whose mRNA level in the hog1 mutant was 25% or less than that of the wild type (i.e. the effect factor was 0.25 or less). The genes identified in this way include the known HOG pathway targets GPD1 (36), GPP2 (50), ENA1 (51), CTT1 (31), GLO1 (52), and HSP12 (53). In most instances expression was not completely abolished in a hog1 mutant, as observed previously for GPD1 (20). The group of HOG-dependent genes contains many of the most strongly induced and most highly expressed osmoregulated genes, and the majority of the genes have been characterized previously.
Conversely, there were 51 genes with a Hog1 effect factor of 0.75 or higher, and those were classified as moderately or not affected by deletion of HOG1 (Table III). This class contains many relatively weakly induced and poorly expressed genes, and several of them are more strongly induced by sorbitol than by salt. This group contains only very few genes that have previously been reported to be induced by osmotic shock, such as GSY2 (44). Promising target genes to study HOG-independent osmotically regulated gene expression include CAR1, PRY2, YNL208W, NCE3, and YER079W, which are all relatively strongly expressed.
Expression of a Small Subset of HOG-dependent Genes Is Affected by Deletion of HOT1-We next determined those genes whose induction by salt was dependent on Hot1p. We had observed previously that deletion of HOT1 did not change the time of peak induction but only the -fold induction (19) (see also Fig. 1). Hence, we employed the same conditions for mRNA isolation as in the wild type, i.e. 45 min after shift to 0.7 M

Transcriptional Response to Osmotic Shock
NaCl. The expression of only 9 genes was affected by deletion of HOT1 (Table III), and the induction of just one gene, STL1, was completely abolished in the hot1 mutant. STL1 encodes a homolog of unknown function of the sugar transporter family (54). Expression of this gene is undetectable under normal conditions but strongly induced after osmotic shock, as confirmed by Northern blot analysis (Fig. 1). We confirmed by Northern blot analysis the expression pattern for two further novel Hot1p dependent genes, YGR043C and YGR052C (Fig. 1). The expression of all nine Hot1p-dependent genes was strongly affected by deletion of HOG1, in fact in all instances more strongly than by deletion of HOT1 (Table III, Fig. 1), as observed previously for GPD1 and GPP2 (19). Induction of all genes, with the exception of YGR043C, was only moderately affected by deletion of MSN2 and MSN4, also in line with previous reports on GPD1 and GPP2 (19,20).
Two of the Hot1p-affected genes showed unusual behavior. The expression of PHO84, a high affinity carrier for inorganic phosphate (55), was strongly diminished in both a HOT1 and a HOG1 mutant, but the expression of this gene was not stimulated by osmotic shock. This suggests that the Hog1p-Hot1p module may also affect gene expression independently of osmotic shock for some genes either directly or indirectly. The expression of PUT4, a proline permease (56), was induced by osmotic shock in a partly Hog1p-dependent fashion, but induction by osmotic shock was in fact stronger in a HOT1 mutant than in the wild type. Stimulated expression in a hot1 mutant of some genes such as CTT1 and HSP12 has been observed previously but only in cells fully adapted to high osmolarity (19). The reason for this is not known.
Msn2p and Msn4p Control a Different Subset of HOG Targets-In a similar way we identified the genes whose induction by a shock with 0.5 M NaCl was affected by deletion of the genes MSN2 and MSN4. The expression of 46 genes had an effect factor of 0.25 or lower (Table III). Virtually all these genes were also strongly, or at least moderately, affected by deletion of HOG1. The few exceptions, GSY1, HSP26, YER067W, YDL222C, and GRX1, are all genes which are only weakly induced, and hence the effect factors may be less reliable. Hence it appears that Msn2p/Msn4p control a subset of the Hog1p-dependent genes. In virtually all cases the effect of deletion of MSN2 and MSN4 was stronger than that of deletion of HOG1.
Msn2p and Msn4p bind to STRE elements in front of their target genes. Most of these genes contain several STREs, many of them in close vicinity, a criterion used in searches for STREregulated genes (24,25). However, some of the genes whose osmotic induction was strongly affected by deletion of MSN2 and MSN4 do not appear to contain STRE elements at all, such as YLL023C, DAK1, and UGP1.
HOG-dependent Genes That Were Neither Affected by Deletion of HOT1 Nor by Deletion of MSN2 Plus MSN4 -Several of the Hog1p-dependent genes do not appear to be dependent on Hot1p or Msn2p/Msn4p. Hence, expression of these genes is likely to be controlled by other transcription factors. Those genes include GRE2, ENA1, and YML131W, which have recently been shown to be controlled by the Hog1p-dependent transcriptional repressor, Sko1p (33). 3 Genes whose osmotic induction or derepression does apparently not depend on either Hot1p, Msn2p/Msn4p, or Sko1p include ARO9, YGR243W, CWP1, YGR086C, FAA1, and YJL108C (Hog1 effect factor of less than 0.25), as well as PUT4, YLR270W, and HXT1 (Hog1 effect factor of less than 0.35).
Genes Whose Expression Is Stimulated in a hog1 Mutant-In salt medium a set of 20 genes was at least 3-fold more strongly expressed in the hog1 mutant than in the wild type, and many of those genes were also salt-induced specifically in the hog1 mutant (Table III). The majority of these genes are known to be involved in the mating of yeast cells, their expression is stimulated by mating pheromone, and they contain pheromone response elements in the promoter. Stimulated expression of these genes is mediated by the pheromone response pathway, a MAP kinase pathway that shares some components with the HOG pathway (3). Recently it has been reported that salt addition stimulates the mating pheromone response pathway in a hog1 mutant (11,57). As demonstrated here, inappropriate stimulation of the pheromone response pathway in the hog1 mutant by salt also leads to the corresponding transcriptional response.

DISCUSSION
The Transcriptional Response to Osmotic Stress Comprises About 5% of All Yeast Genes-The purpose of this genome-wide transcriptional analysis was the identification and classification of genes whose rapid induction by osmotic shock is dependent on the Hog1p kinase and the transcription factors Hot1p and Msn2p/Msn4p. With this approach we also obtained an overview of the set of genes whose expression responded rapidly to osmotic shock. However, the picture of osmoregulated gene expression is likely to be incomplete. Production of many proteins may be stimulated or diminished only during specific phases of the adaptation process. For instance, in cells growing actively in high osmolarity medium we would not expect to see diminished expression of genes encoding ribosomal proteins (58 -60). Furthermore, induction of all genes tested so far is delayed when higher amounts of salt or sorbitol are added to yeast cells (20,61,62) whereas certain genes only become induced upon severe osmotic stress (40). Finally, the induction (or repression) of certain genes may be compound-specific (48,49). Despite these reservations, our analysis revealed almost all genes whose expression has previously been reported to be induced by osmotic stress.
Although probably not comprehensive, our analysis is likely to be specifically sensitive. For instance, the GPD1 mRNA and protein level in cells growing in the presence of 0.7 M NaCl is only about 3-fold higher than in cells growing under normal conditions (19,20,36,60,63). Hence, in fully adapted cells this strongly induced gene (40-fold) would be just at the threshold level according to our criteria. Many weakly and/or transiently induced genes may therefore escape detection in adapted cells.
In conclusion, a full picture of the transcriptional program of osmoadaptation will only be possible when genome-wide analyses are performed with mRNA samples taken at different time points and employing different osmotica at different concentrations.
Overlap between the Transcriptional Response to Osmotic Shock and to Other Stress Conditions-A significant proportion of the induced genes is also controlled by other stress conditions. In many, but clearly not all, instances this reflects a general stress response mediated by Msn2p and Msn4p. Principally, there are at least two explanations for the large overlap between stress responses. (a) Different stress conditions can often occur in nature simultaneously, and therefore any type of stress stimulates a broad response. In this case, a certain function may not be necessary to cope with all the stresses that induce its production. (b) The specific stress disturbs cellular functions leading to another type of stress. For instance, osmotic stress may interfere with electron transport leading to the production of reactive oxygen species. In this case osmotic shock would indirectly induce another specific stress response but probably with a distinct temporal pattern. Because the genes for proteins required for oxidative protection and for the heat shock response are apparently induced very rapidly upon osmotic shock, co-induction of defense mechanisms appears to be the strategy employed by the cell.
Cellular Systems Potentially Involved in Osmoadaptation-A large proportion of all genes whose mRNA level drops after osmotic shock encodes either ribosomal proteins or proteins involved in translation. Diminished expression of such genes upon stress is well documented and reflects the temporary growth arrest (58,59). It appears that the cell has mechanisms to ensure that the translational capacity is sufficient to  Table I for genes whose basal level was below detection threshold).
a Relative mRNA level in mutant divided by relative mRNA level in wild type.
b Fold induction in salt medium in the mutant divided by that in the wild type. appropriately stimulate production of the proteins needed for the adaptive responses.
Twenty-seven of the induced and four of the repressed genes encode proteins or enzymes involved in different aspects of carbohydrate metabolism (Fig. 2). It is well documented that the genes for enzymes in glycerol, trehalose, and glycogen metabolism are induced by osmotic as well as other stresses. Glycerol production is essential for growth in high osmolarity medium (36,64,65). S. cerevisiae possesses two isoforms each for glycerol-3-phosphate dehydrogenase and glycerol 3-phosphatase; one of each (GPD1 and GPP2) was already known to be induced by osmotic shock (19,20,36,50,66). Our analysis adds GPP1. GPD2 is known to be specifically induced under anaerobic conditions when glycerol production is needed for cellular redox regulation (65). GLO1 and DAK1 are listed here because (one of) their role(s) could be the detoxification of by-products of glycerol production (47,52). In addition, dihydroxyacetone kinase could be part of a glycerol degradation pathway to either fine-tune the cellular glycerol content or to prepare the cell for glycerol degradation after a hypo-osmotic shock (47). On the other hand, cellular glycerol levels are already very rapidly down-regulated by export from the cell upon hypo-osmotic shock (67)(68)(69).
The expression of the genes encoding enzymes in trehalose and glycogen production and the synthesis of their common precursor UDP-glucose have previously been reported to be stress-controlled (44,45,70). It should be noted, however (and in contrast to glycerol), that neither trehalose nor glycogen is accumulated by yeast cells under these conditions, and trehalose or glycogen production is not essential for growth at high osmolarity (44,70,71). Evidence has been reported that under different stress conditions a futile cycle of trehalose and glycogen production and degradation occurs (44,70). Indeed, the genes encoding enzymes in trehalose and glycogen degradation are also induced as shown here and reported elsewhere (44,70,72,73). Although it is not known why the cell performs such futile cycles, this example shows that gene expression data have to be interpreted with care when correlating them with actual physiological and metabolic processes.
S. cerevisiae has isogenes for many enzymes in sugar metabolism, and often one of those is much more strongly expressed than the other(s). Remarkably, many of the more poorly expressed isogenes appear to be induced after osmotic shock. Examples include genes for (putative) sugar transporters, glucokinase (GLK1) and an uncharacterized glucokinase homolog (YDR516C), as well as enzymes in the pentose phosphate pathway. In the case of transaldolase (TAL1, YGR043C) and 6-phosphogluconate dehydrogenase (GND1, GND2), the expression of the main isoform was even diminished, indicating a switch of the isoform expression pattern. Further examples include enolase and glyceraldehyde-3-phosphate dehydrogenase (60, 74 -76), which were not resolved in this analysis due to crosshybridization. These observations indicate that the isogene pattern of S. cerevisiae could reflect an adaptation to a variable environment. Perhaps a certain isoform has a higher stability under stress conditions whereas under normal growth conditions another isoform is more efficient. Careful analysis of the enzymatic properties of different isoforms is needed to address this interesting question.
Further changes appear to occur in the metabolism of amino acids and lipids. A number of genes involved in the catabolism of certain amino acids are induced while at the same time several genes encoding enzymes in amino acid and nucleotide biosynthesis are repressed. These changes may reflect the temporary growth arrest. Genes encoding enzymes in methionine biosynthesis are among the most strongly repressed ones, perhaps because of the need of sulfhydryl groups for the production of glutaredoxin and thioredoxin (47).
The observed changes in the expression of genes encoding enzymes in lipid metabolism appear to aim at enhanced production of phosphatidylethanolamine and phosphatidylcholine and at diminished production of sterols. This could indicate changes in the lipid composition of cellular membranes upon osmotic stress, perhaps in order to alter membrane fluidity or the permeability for water or glycerol. It should be noted, however, that an increased (rather than a decreased) sterol content has been reported for osmostressed Zygosaccharomyces rouxii cells (77,78). Careful biochemical analyses are needed to Eleven genes have been placed in the category entitled chaperones and protective functions. The production of these proteins under stress is well established; however, the biochemical functions of for instance Hsp12p, Hsp26p, and Ddr2p are completely unknown. Most of the genes listed in this category are strongly dependent on the general Msn2p/Msn4p-mediated stress response (see below).
The expression of 10 genes, whose products are known or suspected to be involved in the protection from oxidative damage, is induced. Stimulated expression of 16 genes, whose products catalyze (potentially) redox reactions, induction of genes encoding enzymes in the non-oxidative part of the pentose phosphate pathway, and altered expression of a number of genes encoding mitochondrial functions may also be related to a response to oxidative damage and/or to altered redox metabolism. The production of many of these proteins has been reported to be stimulated by oxidative (79) and osmotic stress (47,63). It is presently unclear how far osmotic stress leads to production of reactive oxygen species in yeast, but the gene expression pattern indicates a close correlation between these two types of stress. Careful analysis of the phenotype under different stress conditions of mutants lacking such genes is needed to understand the role of these proteins.
Seven genes, which encode or may encode functions related to cell surface and cell wall formation are induced by osmotic shock. The cell shrinks upon osmotic shock and also grows at a smaller size in high osmolarity medium (80). Because modeling of the cell surface is controlled by the tendency of the cytoplasm to expand (81), it is sensible to expect that osmotic stress affects processes related to cell surface assembly. In addition, cell surface modeling is controlled by the protein kinase C MAP kinase pathway (3,81), which is stimulated by hypo-osmotic shock (82). Because the HOG and protein kinase C pathways appear to control each other (82), stimulation of the HOG pathway may affect cell surface modeling (at least) indirectly via its effect on protein kinase C pathway signaling.
The expression of four genes whose products are involved in vacuolar biogenesis is stimulated by osmotic shock. It is known that mutations affecting vacuolar function lead to osmosensitivity (83), although the specific reasons have not been systematically investigated. The vacuole is likely to participate in cellular water and ion homeostasis.
Eighteen of the induced and six of the repressed genes have been classified under signaling and gene expression. Some proteins, such as those encoded by TPK1 (28), YAK1 (84), PPZ2 (85), ACA2, 3 and YAP4 (86) have been reported to play a role in stress responses or the mutation of these genes to cause stress sensitivity. Information on the regulation of expression of genes encoding signaling proteins in yeast is very scarce, although for instance in plant systems it is well known that stress also induces the expression of proteins involved in stress signaling (87).
Osmotic Induction of Gene Expression Requiring the HOG Pathway-The expression after osmotic shock of 48 genes is diminished by more than 75% in a hog1 mutant, and those genes include many of the most strongly expressed and induced genes. In addition, the expression of almost half of the osmoinduced genes is moderately affected in the hog1 mutant. This result illustrates the major importance of the HOG pathway in osmoregulated gene expression (Fig. 3). On the other hand, there are only a few genes whose expression completely depends on the HOG pathway, such as ALD3 and STL1. Hence, induction by osmotic stress seems to be the result of different signaling pathways converging at the level of the promoter of these genes, as suggested previously for GPD1, GPP2, CTT1, and HSP12 (19,20,31,53,88).
HOG-dependent genes occur in all functional categories suggesting that the HOG pathway does not control specific functional aspects of the osmotic response.
Hot1p Affects Expression of a Small Subset of HOG-dependent Genes-Hot1p has recently been identified in a search for proteins that interact in the two-hybrid system with the Hog1p kinase (19). Although genetic evidence indicates that Hot1p is part of the HOG pathway, careful analysis of the time courses of induction of GPD1 in hog1, hot1, and hog1 hot1 mutants suggested that the interplay between Hog1p and Hot1p may be complex (19) (see also Fig. 1). Analysis of Hot1p-dependent genes further ties Hot1p to the HOG pathway because all genes whose mRNA level is diminished by deletion of HOT1 after osmotic shock are also Hog1p-dependent. In all instances the mRNA level is diminished more strongly by HOG1 than by HOT1 deletion. Hence, Hog1p also appears to control the expression of these genes by Hot1p-independent mechanisms. These observations suggest that Hot1p acts together with other factors in the control of gene expression. On the other hand, this analysis has identified one gene, STL1, whose expression completely depends on Hog1p and Hot1p (Fig. 3). We are analyzing the promoter of STL1 in detail.
Msn2p/Msn4p Control a Subset of HOG-dependent Genes-Remarkably, osmoregulated expression of virtually all genes, which is strongly diminished by deletion of MSN2 and MSN4, is also dependent on HOG1. Hence, a direct link between these regulators may exist (Fig. 3). Msn2p (and Msn4p) are known to be regulated by stress and by protein kinase A (21,22,26,29,31). Upon stress, Msn2p is dephosphorylated and translocated to the nucleus (29), but whether this process is sufficient to stimulate gene expression has not yet been addressed. After an osmotic shock, Msn2p and Hog1p are both localized in the nucleus where they could interact (16,17,29). In addition, it has been shown previously that osmostress-stimulated expression of a STRE reporter gene is both Msn2p/Msn4p-and Hog1pdependent (31). Taken together, these data are compatible with a model in which protein kinase A mediates translocation of Msn2p (and Msn4p) to the nucleus under any type of stress conditions where further stress-specific processes, for instance mediated by Hog1p, then confer full transcriptional activity to Msn2p (Fig. 3).
Other Hog1p-dependent Transcription Factors-The expression of several Hog1p-dependent genes appears to be independent of Hot1p and Msn2p/Msn4p. ENA1 and GRE2 are known to be controlled by Sko1p (33). 3 Sko1p represses target genes, and activation of Hog1p counteracts this repression (33). 3 Whereas the promoter of ENA1 is controlled in a complex way via different signaling pathways (32,33,51), osmotic regulation of GRE2 appears to be completely dependent on Hog1p and Sko1p, 3 making it a useful model system to study the Hog1p-Sko1p regulatory module (Fig. 3).
The expression of 11 genes was diminished by at least 65% in a hog1 mutant but was unaffected by deletion of the gene encoding any of the known transcription factors (Fig. 3). Msn1p is a candidate for a factor controlling at least some of those genes. Msn1p is distantly related to Hot1p and has been shown to affect the expression of GPD1 and CTT1 after osmotic shock (19). However, the relationship of Msn1p to the HOG pathway has not yet been investigated.
HOG-independent Genes-Fifty genes were classified as Hog1p-independent. This group contains many weakly expressed genes with the exception of NCE3, YER079W, and YNL208W, whose induced mRNA level was at least 50% that of IPP1. Many of the Hog1p-independent genes were only moderately induced by salt or sorbitol. It should be noted, however, that the induction of the Hog1p-independent genes may follow a different time course from that of the Hog1p-dependent genes and that the maximal induction of these genes may hence be bigger. Many of the genes in this group are more strongly induced by sorbitol than by salt, an observation that further hints at the involvement of novel regulatory mechanisms. Future work will address the nature of the signaling pathways and transcription factors controlling these Hog1p-independent genes (Fig. 3).
Osmotic Stimulation of a Mating Response in a hog1 Mutant-Twenty genes were found to be induced by salt in the hog1 mutant but not in wild type cells. All previously characterized genes in this category are known to be involved in the mating response of yeast cells (89). Most of these genes also have a clearly recognizable pheromone response element, the binding site for the Ste12p transcription factor, in their promoter (90). Hence, hog1 mutant cells shifted to high salt show characteristics of cells responding to mating pheromone. In high osmolarity medium, the hog1 mutant displays an aberrant morphology that resembles that of cells responding to pheromone, although in contrast to the latter the development of mating projections was not directed (15,34). The mating pheromone response pathway is stimulated by salt in a hog1 mutant (11,57), and our analysis demonstrates that this also leads to the corresponding transcriptional response. This inap-propriate activation may be because of the complex connections of yeast MAP kinase pathways, which share components such as Ste20p, Ste50p, and the Ste11p MAPKKK (3,6). Because the response to mating pheromone leads to a proliferation arrest (89), the inappropriate stimulation of this pathway is likely to contribute to the growth defect of hog1 mutants in high osmolarity medium.