The transcriptional response to environmental changes is a major topic in both basic and applied research. From a basic point of view, to understand this response includes unravelling how the stress signal is sensed and transduced to the nucleus, to identify which genes are induced under each stress condition and, finally, to establish the phenotypic consequences of this induction in stress tolerance. The possibility of using genetic approaches has made the yeast Saccharomyces cerevisiae a compelling model to study stress response at a molecular level. Moreover, this information can be used to isolate and characterise stress-related proteins in higher eukaryotes and to design strategies to increase stress resistance in organisms of industrial interest. In this review the progress made in recent years is discussed.
Different types of environmental and physiological stress conditions constantly challenge all living organisms. To cope with the deleterious effects of stress, cells have developed rapid molecular responses to repair the damage and protect against further exposure to the same and other forms of stress. The best-characterised response is the synthesis of a limited number of proteins, the stress proteins. Although post-transcriptional mechanisms can play an important role in the regulation of the stress response, research on this topic has been mainly focused at the level of transcription, and several stress-induced transduction pathways and stress-controlled transcription factors have been identified and characterised (for review see, for instance, ). However, little is know about the way through which the stress conditions activate these pathways and, in some cases, the connections between pathways and transcription factors remain elusive.
It is commonly assumed that the synthesis of stress proteins is intended for the cell's survival and adaptation to adverse conditions. The existence of so-called induced tolerance would seem to support this; short-term treatment with mild stress results in an increase in tolerance against a subsequent, normally lethal, dose of the same stress. The most direct interpretation of these results is that the presence of stress proteins, previously induced by mild stress, increases cell tolerance. However, the relationship between the induction of individual stress proteins and the acquisition of stress tolerance is not always evident and it has frequently been the object of controversy.
Stress responses are of particular importance in microorganisms, whose environment is highly variable, and conditions such as temperature, osmolarity or nutrient availability are far from constant. Among them, the yeast Saccharomyces cerevisiae has become a model organism to study how eukaryotic cells respond to stress and define the specific role of the stress-induced proteins. Moreover, the high degree of evolutionary conservation of the stress pathways between yeast and higher eukaryotes means that yeast can serve as a suitable model system for the characterisation of the stress response in more complex organisms.
In this review I analyse the signal transduction pathways triggered by the most common types of stress in yeast, paying special attention to the transcription factors involved in stress-induced gene expression. Among them, the general stress response will be discussed in detail. I will also describe the genes induced by the activation of these pathways and the role of their products in protecting cells against the damage caused by one or several stress conditions.
2The general stress response
Yeast cells exposed to mild stress develop tolerance not only to higher doses of the same stress, but also to stress caused by other agents. This phenomenon, known as cross-protection, suggests the existence of an integrating mechanism that senses and responds to different forms of stress. Beside these physiological observations, the integrated response is supported by molecular results. Several heat shock proteins were identified as induced by other unfavourable conditions besides thermal stress, such as nutrient starvation [2,3], and, in a similar way, the catalase-encoding gene, CTT1, proved to be induced by heat shock and other types of stress in a heat shock factor-independent manner . In this context, the idea of a general stress response emerges as a system that regulates the coordinated induction of many stress genes through a common cis element in their promoter (known as stress response element, STRE, see below). The STRE would be induced by different environmental or metabolic stresses and the transcriptional activation mediated through this element would result in the acquisition of a tolerant state towards any stress condition. This hypothesis has been corroborated in recent years.
2.2The cAMP–PKA pathway and the stress response
The idea that there would be a group of genes activated through a common pathway in response to very different stress conditions was strengthened by the finding that many stress-induced genes were negatively regulated by the cAMP–protein kinase A (PKA) pathway. The cAMP–PKA pathway has been involved in many cellular processes, including nutrient sensing, regulation of yeast cell proliferation, carbon storage and stress response (reviewed in [1,5,6]). The components of the cAMP–PKA pathway (Fig. 1) have been studied extensively (reviewed in ). Adenylate cyclase, encoded by the CYR1/CDC35 gene, can be activated by Ras1p and Ras2p or by the G protein-coupled receptor system Gpr1p-Gpa2p . The cAMP produced activates the cAMP-dependent protein kinase (PKA) by binding to the regulatory subunit (encoded by the BCY1 gene) and dissociating the catalytic subunits (encoded by TPK1, TPK2 and TPK3). Recent results point to the G protein-coupled receptor system as responsible for activation of adenylate cyclase in response to glucose, whereas Ras proteins would be responsible for the signal transmission in conditions of stress (, see below). The activation of PKA causes transient changes in several systems including trehalose and glycogen metabolism, glycolysis and gluconeogenesis, which contain components controlled by PKA-mediated phosphorylation . Moreover, basal and induced expression of some HSP genes was found to be enhanced in ras2 mutants with low PKA activity and dramatically reduced in bcy1 mutants with high constitutive PKA levels [1,7].
It has been suggested (Fig. 1) that the function of the Ras proteins might be related to the sensing and transmission of stress signals on the basis that the activity of the guanine nucleotide exchange factor for Ras, Cdc25p, is positively regulated by the cytosolic Hsp70 Ssa1p, through a direct interaction . Therefore, the accumulation of denatured proteins upon stress conditions would reduce the activity of the cAMP–PKA pathway by recruiting Hsps and reducing their interaction (and thereby its positive effect) with Cdc25p. Intracellular acidification can also control the function of Ras proteins . The activation of Ras by acidification requires the Ira proteins, which are activators of Ras–GTPase activity. However, the increase in Ras-bound GTP observed after intracellular acidification was not dependent on Cdc25p. The function of this intracellular acidification could be related to the maintenance of a proper intracellular pH and ATP level during carbon source starvation . Thus, the ATP shortage produced under these circumstances would result in a drop of intracellular pH and stimulation of cAMP synthesis, which leads to the production of ATP through the mobilisation of storage carbohydrates. On the other hand, although low external pH induces STRE-dependent transcription, recent results exclude internal acidification as an obligatory signal triggering the general stress response .
2.3The stress response element
A cis-regulatory element able to mediate transcriptional induction by different forms of stress was identified in the promoter of CTT1 and DDR2, a DNA damage-responsive gene [11,12]. This element was designated STRE stress responsive element), and has the core consensus sequence AGGGG . STRE sequences have been identified in many stress-induced genes and a computer search of the entire yeast genome predicts as many as 186 potential STRE-regulated genes [13,14]. However, the presence of a STRE-like sequence in the promoter does not imply the functionality of this element (see for instance ). A single copy of this element is sufficient to activate a reporter gene by different types of stress, although two or more copies of this sequence provide a greater than additive effect on stress-induced gene expression . The STRE is functional in both orientations (CCCCT or AGGGG) and the stress-induced expression conferred by the STRE is negatively regulated by the cAMP–PKA pathway [11,16]. Analysis of mutational variants of the STRE shows that base changes within the STRE completely abolish its function, while its efficacy as a regulatory element was only slightly reduced by changes in neighbouring nucleotide sequences and interelement spacing . A similar cis-regulatory element has been identified in the promoter of the SSA3 gene. This element, whose sequence resembles the STRE (T/AAGGGA) and is also regulated by the cAMP–PKA pathway, was called PDS (post-diauxic shift) and is responsible for the induction of SSA3 late during growth . However, both elements seem to be regulated in a different way since PDS is only able to cause a marginal increase in expression in response to heat shock , it cannot be induced by osmotic stress under conditions in which induction of STRE-containing genes is observed , and it is not affected by mutations in transcription factors involved in STRE-driven gene expression (, M. Amorós and F. Estruch, unpublished results).
The presence of STRE on the promoter does not imply that all STRE-containing genes show a uniform pattern of expression. Thus, in spite of the presence of STRE elements in their promoters, expression of the trehalose biosynthetic genes (TPS1, TPS2, TPS3 and TSL1) differs from the expression of other STRE driven genes such as CTT1. The complexity on STRE-regulation is also illustrated by the results of Inoue et al.  on the expression of the GLO1 gene, encoding a glyoxalase involved in the detoxification of methylglyoxal. This gene has two STREs in its 5′ flanking region, but unlike other STRE-containing genes is only induced by osmotic stress and not by heat, H2O2 or high ethanol concentration, stresses that can induce the expression of an artificial STRE-LEU2-lacZ gene fusion [16,20].
The different patterns of expression among STRE-containing genes could be due to the promoter context in which the STREs are found. Thus, some of these genes include in their promoter other regulatory sequences involved in stress-induced transcription, such as heat shock elements (HSE). The relationship between HSE and STRE depends on the gene; for some genes the regulatory systems act redundantly but for others the contribution of each system in the activation is additive (; M. Amorós and F. Estruch, unpublished results). The redundancy of regulatory elements on the promoter of several STRE-containing genes can be functionally important. Thus, redundancy between HSE and STRE/PDS is observed in genes such as SSA3 and HSP104 that, as will be discussed, play an important role in thermotolerance. In a similar way, genes involved in the adaptive response to oxidative stress, such as SOD2, contain both STRE and binding sites for the transcription factor Yap1p. Another example of transcriptional redundancy is GPD1, a key gene in the tolerance to high osmolarity, whose osmostress-regulated expression is only abolished by the simultaneous deletion of three different transcription factors (, see below).
2.4The STRE-binding factors Msn2p and Msn4p
Two trans-acting factors, Msn2p and Msn4p, have been involved in STRE-mediated gene expression [20,23]. The MSN2 gene, encoding a C2H2 zinc finger protein, was isolated as a multicopy suppressor of the raffinose utilisation defect shown by mutants with a thermosensitive allele of the SNF1 gene. A second similar zinc finger gene, MSN4, was isolated by sequence homology and, like MSN2, is able to suppress the thermosensitive snf1 mutation when overexpressed . However, Snf1p is not required for transcriptional induction mediated by STRE . Msn2p and Msn4p recognise and bind STREs both in vitro and in vivo [20,25], and are required for the induction of a STRE-LEU2-lacZ reporter gene in response to stresses such as heat shock, low pH, sorbic acid and high ethanol concentrations . Thus, one can imagine a scenario where Msn2p and Msn4p are activated by very different forms of stress and induce the transcription of a large battery of STRE-containing genes. However, the situation is not so simple. For osmotic and oxidative stresses the STRE-LEU2-lacZ is expressed at low levels in the msn2msn4 mutant strain, although the factor of induction was comparable to the wild-type strain . This induction could be explained by the existence of other STRE-binding factors that would be specifically activated under these stress conditions. On the other hand, the significant reduction in the level of expression suggests the participation of Msn2p and Msn4p in the osmotic and oxidative response mediated by STRE . Moreover, recent results show that the contribution of Msn2p and Msn4p can be essential for the osmotic activation of some STRE-containing genes (; M. Amorós and F. Estruch, unpublished results).
Genetic evidence suggests that Msn2p and Msn4p are functionally redundant , although different results show that these factors could have different roles in the regulation of the stress response. Thus, Treger et al.  have analysed the requirement for Msn2p or Msn4p in the induction by heat shock of several STRE-containing genes. They have found that for some genes, such as PGM2, heat induction is completely eliminated in an msn2 single mutant. For most of them, the induction is reduced in the msn2 single mutant and abolished in the msn2msn4 double mutant. They even found a gene, PDE2, whose heat induction is not affected by the absence of Msn2p and is only defective in the double mutant strain. The distinctive contribution of each factor would depend on the promoter context, and could act as a mechanism to differently modulate gene expression through a unique cis-regulatory sequence. Another level of complexity emerges when the individual contribution of Msn2p and Msn4p is analysed under different stress conditions. Our results suggest that the activity of Msn2p and Msn4p can vary depending on the stress condition (M. Amorós and F. Estruch, unpublished results). The particular role of these factors is also suggested by the different patterns of mRNA accumulation shown by the MSN2 and MSN4 genes at the diauxic transition .
2.5Regulation of Msn2p and Msn4p function by stress conditions
Gel shift assay experiments performed with extracts prepared from unstressed and stressed cells suggested that the DNA-binding affinity of Msn2p is independent of the stress condition . However, in vivo footprinting analysis shows that the binding is enhanced by stress . Two non-exclusive models of regulation could explain these results: regulated nuclear localisation or regulated DNA binding controlled inside the nucleus. Using epitope-tagged proteins as well as GPF fusions has shown that Msn2p and Msn4p are distributed throughout the cytoplasm and partly excluded from the nucleus in unstressed cells. After stress treatment the proteins accumulate in the nucleus . Two regulatory pathways are involved in the localisation of Msn2p and Msn4p [25,28]. An inverse correlation was found between PKA activity and nuclear localisation of these factors. In addition, mutations in the potential PKA modification sites of Msn2p lead to constitutive nuclear localisation of this protein . Recently, Beck and Hall  have reported that rapamycin, acting through the TOR pathway, induces the nuclear accumulation of Msn2p and Msn4p. The effect of the rapamycin treatment is to release Msn2p and Msn4p from Bmh2p that retains these factors in the cytoplasm. Release of Msn2p and Msn4p from Bmh2p can also be induced in response to glucose withdrawal, although the effect of other stress conditions on this interaction has not been investigated. Bmh2p, and its homologue Bmh1p, are 14-3-3 proteins required for Ras–mitogen-activated protein kinase (MAPK) cascade signalling during pseudohyphal development . The relationship between the TOR and the Ras–PKA pathway in controlling the nuclear localisation of Msn2p and Msn4p is an important question that deserves further analysis.
Translocation of Msn2p and Msn4p is not the only regulatory mechanism involved in STRE-mediated activation. Nuclear export of Msn2p and Msn4p depends on the product of the MSN5 gene (C. Schüller and H. Ruis, personal communication), encoding an export receptor that interacts with the small GTPase Ran and is involved in different signalling pathways [30–32]. In an msn5 mutant Msn2p and Msn4p are localised in the nucleus even in non-stress conditions, although regulation of STRE-dependent genes is not affected in this mutant (P.M. Alepuz and F. Estruch, unpublished results), suggesting additional mechanism(s) to regulate Msn2p and Msn4p activity. The situation may be similar to that found for the transcription factor Pho4p, in which multiple phosphorylation sites play distinct and separable roles in regulating the activity of Pho4p . The existence of redundant regulatory mechanisms, independently controlled by external and internal factors, could be important in ensuring the correct activation of the general stress response. Thus, PKA could act, through the induction of Msn2/4p export by Msn5p, as a modulator of the activity of these factors. This role would be more closely related to the level of response to cell growth conditions than to transducing stress signals. In this model, other pathways would perform transduction of stress signals by the regulation of Msn2/4p (or other STRE-binding factors) activity through a mechanism that differs from the control of subcellular localisation. Supporting this, a signal pathway transducing the stress effects on the plasma membrane to Msn2p, which would not have cAMP as an obligatory second messenger, has been proposed .
2.6Genes regulated through STRE and physiological role of the general stress response
As mentioned, with respect to the presence of STRE sequences in their promoter, many yeast genes are candidates for regulation through the general stress response pathway [13,14]. The list includes genes involved in carbon metabolism, transporters, proteases and genes with protective functions against different types of stress, including HSP104, CTT1 and the trehalose metabolism genes TPS1, TPS2, TPS3 and TSL1[13,14,34]. In agreement with their role as main STRE transcription factors, msn2msn4 double mutants show a higher sensitivity to different stresses, including carbon source starvation, heat shock, osmotic and oxidative stresses [20,35]. On the other hand, overexpression of MSN2 and MSN4 genes improves the resistance to starvation and thermal stresses . However, the phenotypes shown by the msn2msn4 mutants are only observed under severe stress conditions, suggesting the existence of an additional mechanism involved in the protection to milder unfavourable conditions. As has been mentioned, there is a redundancy of regulatory elements on the promoter of genes with important roles in stress response. It is possible that these genes could be activated differently depending on the severity of the stress condition. Thus, the activation of regulatory systems such as HSE/Hsf1p or ARE/Yap1 (see below) would confer protection under mild stress, but the activity of these systems would be impaired or would become insufficient under severe conditions, making the STRE/Msn2/4p system essential for cell survival. On the other hand, although most of the proteins induced at the diauxic transition depend on Msn2/4p , pleiotropic sensitivity is not observed in stationary msn2msn4 mutant cells, suggesting the activation of Msn2/4-independent protective mechanisms in the stationary phase (Fig. 1). A candidate to activate these mechanisms is the protein kinase Rim15p that has recently been involved in the physiological adaptations necessary for proper entry into the stationary phase . rim15 mutants show a defective induction of thermotolerance and starvation resistance on entering the stationary phase and are unable to induce the post-diauxic transcription of genes such as HSP12, HSP26 and SSA3. Genetic analysis has placed Rim15p immediately downstream of PKA and under the negative control of cAMP , although the possibility that Rim15p is involved in the regulation of Msn2/4p function is yet to be investigated.
3.1The heat shock element and the heat shock factor
When shifted to higher temperatures cells respond by inducing the synthesis of a set of proteins called the heat shock proteins or Hsps. Many Hsps are also expressed at normal temperatures and play an essential role in promoting the folding and unfolding of other proteins, the assembly and disassembly of proteins in oligomeric structures, and the degradation of proteins that are improperly assembled or denatured (reviewed in [37,38]). In S. cerevisiae at least 52 different proteins are induced by a temperature shift from 25°C to 38°C . Two distinct regulatory elements are involved in the induction of genes by heat shock, the STRE, discussed above, and the heat shock element (HSE). HSEs are the binding sites for the heat shock factor (HSF) and are composed of at least three copies of the repeating sequence nGAAn, arranged in alternating orientation (reviewed in ). The number of 5-bp units in a functional HSE can vary but usually ranges from three to six [40,41]. Deviations from the consensus (both in sequence and/or in the distance between modules) may be tolerated in vivo, but they can influence the affinity of the Hsf1p and, thereby, the level of transcriptional activation [42–44].
In S. cerevisiae the HSF is encoded by a single-copy essential gene, HSF1. Like metazoan HSFs, Hsf1p contains a helix-turn-helix DNA-binding domain and coiled-coil hydrophobic repeat domain, which mediates the trimerisation of Hsf1p [46,47]. However, it differs from mammalian, fly and plant HSFs in the presence of two transactivation domains, which respond differentially to heat shock [48,49]. The amino-terminal activation domain mediates a transient response to elevated temperatures, while the carboxy-terminal activation domain (CTA) is required to regulate both a transient and a sustained response . The CTA is required for the expression of the metallothionein CUP1 gene in response to heat and oxidative stress, but it is largely dispensable for the heat activation of the Hsp70-encoding genes SSA1 and SSA3. The requirement for the CTA has been related to the architecture of the HSE, since the activation of genes containing gapped HSE was revealed to be highly dependent on CTA . It has been suggested that the presence of two activation domains, together with the effect of the HSE architecture on their function, may provide additional levels of regulation or selectivity in gene activation, making up for the absence of several HSF isoforms .
S. cerevisiae, together with Kluyveromyces lactis, seems to be the only eukaryotic organism in which HSF is associated with HSE in the absence or presence of heat shock [51,52]. However, although the in vivo DNA-binding activity of Hsf1p is enough to bind HSE prior to heat shock, temperature stress increases this DNA-binding activity . The increased binding after heat shock is observed to a greater extent in weak HSEs, showing deviations from the consensus HSE sequence . In higher eukaryotes, the activation of DNA-binding activity depends on the trimerisation of monomeric HSF subunits upon heat shock . However, yeast Hsf1p appears as a trimer at relatively high concentration before heat shock, suggesting that other mechanisms could be involved in the regulation of DNA-binding activity in vivo . The regulation of Hsf1p binding might occur through the alteration of the chromatin structure in response to heat shock. Thus, the binding to a low-affinity site could be partially dependent on a change in chromatin structure that would facilitate the binding of the HSF to the non-consensus HSEs. In those promoters that contain both HSE and STRE, the binding of Msn2p and Msn4p could trigger this change. Results obtained in the author's laboratory on the regulation of HSP26 genes are compatible with this model (M. Amorós and F. Estruch, unpublished results). On the other hand, the possibility of Hsf1p binding without gene induction indicates that the binding of Hsf1p to HSE itself would not be sufficient for transcriptional activation: a second modification step would also be required . It has been suggested that phosphorylation is this modification, since Hsf1p is rapidly phosphorylated in response to both heat and oxidative stresses . However, this phosphorylation seems not to be essential for its activation, but instead could serve to deactivate Hsf1p, allowing a transient response .
The fact that a variety of point mutations and deletions in different regions within HSF can cause an increase in the constitutive transcriptional activity suggests a negative regulatory control of HSF activity in the absence of stress ( and references therein). However, the mechanism by which the HSF is able to sense and respond to heat shock is unknown. It has been suggested that HSF may be able to sense the heat stress directly through thermally induced changes in the structure or dynamics within the DNA-binding domain . Other observations have led to the model of heat shock response auto-regulation, with Hsp70 serving as the cellular thermometer that regulates, through the modulation of the HSF activity, the expression of the HSP genes . This model would have Hsp70 interacting directly with HSF under non-stress conditions, maintaining the transcription factor in a low-active form. Upon heat shock, there would be an increase in the concentration of misfolded proteins, the Hsp70 substrates, leading to a depletion of ‘free’ Hsp70. As the pool of Hsp70 is reduced, HSF would be released, resulting in formation of a high-active HSF form. The response would be self-limiting because the ensuing overproduction of Hsps would restore the free pool of Hsp70 . Consistent with this model is the fact that human Hsp70 associates in vivo and in vitro with HSF1 and this association represses HSF1 . Human Hsp70 specifically interacts with the transactivation domain of HSF1 and negatively regulates transcriptional activity without having an effect on DNA binding . In yeast, different genetic data support the participation of Hsp70s, in particular those encoded by the SSA1–4 genes, in the regulation of Hsf1p function. Thus, ssa1ssa2 mutant strains show a high level of expression of the HSP genes under optimal growth conditions as a consequence of the activation of Hsf1p .
3.2On the role of Hsps on thermotolerance
Characterisation of the phenotypes associated with mutations in specific Hsps reveals that some of them are required at any temperature. This fact may explain why HSF1 is an essential gene, whose basal activity would be required for the expression of the Hsps that are necessary for normal cell growth. On the other hand, Hsp induction would be required for growth at temperatures beyond the normal range, as suggested by the temperature-sensitive phenotype caused by mutations in the HSF1 gene . However, the hypothesis that Hsps provide tolerance to heat and other stresses has frequently been the object of controversy [62–64]. One of the most intriguing reports on this subject was provided by Smith and Yaffe . These authors described a yeast strain containing a mutant allele of the heat shock factor, hsf1-m3, in which the acquisition of thermotolerance is not affected, although this mutant is defective in the induction of Hsps. However, the finding that this mutant constitutively expresses high levels of Hsp104, a protein directly involved in thermotolerance (, see also below), could be the reason for the strain's ability to acquire wild-type levels of thermotolerance . On the other hand, De Virgilio et al.  reported that a mutant strain unable to synthesise proteins during a preconditioning heat treatment acquires thermotolerance, albeit to a lesser degree than the corresponding wild-type strain, suggesting that, beside Hsps, other mechanisms are involved in thermotolerance. One of these mechanisms is provided by trehalose (see below). Elliot et al.  have found that Hsp104 and trehalose are synergistic with each other, and together account for most of the heat shock resistance of stationary phase cells.
3.3Hsp70 and Hsp104: a coordinated action
The specific functions of individual Hsps have been studied in detail (for review, see ). Special attention has been paid to two of the most conserved Hsp families: Hsp70 and Hsp100. Eukaryotic Hsp70s are encoded by a multigene family whose members are expressed under a variety of physiological conditions . In S. cerevisiae there are at least 10 genes related to HSP70 of higher eukaryotes (Table 1). Among them, the SSA subfamily, with four members, encodes the cytosolic Hsp70 . The Ssa proteins form an essential group; at least one of them must be present at high levels for cell viability. Three (Ssa1p, Ssa3p and Ssa4p) are heat-inducible (although the SSA1 gene is also expressed under normal growth conditions), whereas Ssa2p is constitutive. Besides their essential roles for normal growth, cytosolic Hsp70s have been involved in two processes that are directly related to thermotolerance: regulation of the heat shock response (discussed above) and prevention of protein aggregation and posterior refolding of proteins that have been damaged during heat shock . The latter role is directly related to their function as a chaperone. The Ssa proteins have ATPase activity that is stimulated by peptide binding  and are able to prevent the aggregation of heat-damaged proteins . Besides this preventive role, Hsp70s are important in repairing the damage done. Together with Hsp40, they are required for the Hsp104-dependent refolding activity over aggregates of denatured proteins (, see below).
Table 1. Major members of the Hsp70 family
Open reading frame
protein folding and translocation, heat shock regulation
protein folding and translocation, heat shock regulation
protein folding and translocation, heat shock regulation
protein folding and translocation, heat shock regulation
assembly and maturation of mitochondrial iron–sulfur proteins
GRP78, SSD1, BIP
endoplasmic reticulum lumen
protein import and assembly
endoplasmic reticulum lumen
protein transport and folding
The S. cerevisiae HSP104 gene is a member of the highly conserved Hsp100/Clp family . Cells carrying an hsp104 mutation grow at the same rate as wild-type at 25°C and 37°C and die at the same rate when exposed directly to high temperature . The hsp104 mutant is able to display induced thermotolerance, but this is very transient. Within a few minutes of exposure to 50°C cells begin to die at 100–1000 times the rate of wild-type . Hsp104 is less important for survival at lower killing temperatures, suggesting that at moderate stress conditions damage can be prevented by other mechanisms . In these conditions, both the Hsp70 chaperone function and trehalose accumulation could act as primary stress protectants (see below). However, these protective mechanisms would be insufficient under extreme stress, resulting in cell damage and the requirement for Hsp104 to repair the damage. The repairing function of Hsp104 is indicated by its ability to rescue heat-inactivated proteins directly from insoluble aggregates, a capability that is not shown by other chaperones [72,73]. Repair is performed in conjunction with other chaperones. The reaction catalysed by Hsp104 produces the disaggregated protein, substrate for the refolding activity of Hsp70 and Hsp40 . In summary, yeast has developed protective mechanisms to avoid denaturation of cellular proteins by heat shock. If damage occurs, the heat-damaged proteins can be reactivated by a chaperone system that allows even aggregated proteins to be rescued. Finally, those proteins that cannot be repaired will be targets for degradation (for a review of stress-induced proteolysis see ).
4.1Transcription factors involved in the oxidative stress response
When growing aerobically, yeast has to handle the generation of reactive oxygen species (ROS) such as hydrogen peroxide, hydroxyl radicals or superoxide anions. ROS are generated by normal metabolic processes (respiration or β-oxidation of fatty acids) as well as by exposure to pro-oxidants such as H2O2 or heavy metals. Using two-dimensional gel electrophoresis, Gordon and co-workers  have identified 167 proteins whose expression changes with H2O2 treatment, including many of the yeast antioxidant activities (see below). The possibility that different ROS induce different sets of genes has been investigated by comparing the pattern of proteins induced in response to H2O2 or to the superoxide-generating agent menadione. The conclusion is that, although there are proteins that are specifically induced by H2O2 or menadione, a significant overlap between both responses exists .
Several transcription factors have been involved in the regulation of gene expression under oxidative stress conditions. The best characterised are a family of b-ZIP transcription factors whose archetype is Yap1p. The YAP1 gene was originally cloned by its ability to bind to an SV40-derived AP-1 site . When overexpressed, YAP1 is able to confer resistance to several toxic agents [80–82] and, on the other hand, yap1 mutants were found to be hypersensitive to oxidants . A second gene, YAP2, shares some of these phenotypes , and six other members of the Yap family, with similar b-ZIP domains, were identified once the S. cerevisiae genome had been sequenced . Phenotypic analysis shows that members of the Yap family carry out overlapping, but distinct, biological functions .
Regulation of Yap1p takes place at the level of subcellular localisation; under normal conditions Yap1p is restricted to the cytoplasm, but it becomes nuclear in response to oxidative stress (reviewed in ). Yap1p contains a C-terminal region with three conserved cysteine residues called the CRD (cysteine-rich domain) and it has been proposed that the oxidation of these cysteines acts as a sensor of the redox state . The CRD is required for maintaining Yap1p in the cytoplasm , since it includes a nuclear export sequence (NES) through which the export factor Crm1p works to maintain Yap1 in the cytoplasm under non-stress conditions . The binding of Yap1p to Crm1p is sensitive to oxidation and in an oxidative environment the interaction between Yap1p and Crm1p is inhibited, resulting in the nuclear localisation of Yap1p [88,89]. Mutation in the three conserved cysteines of the CRD makes the interaction between Yap1p and Crm1p insensitive to oxidising conditions . This mutation renders the NES constitutive and Yap1p located at the cytoplasm even under oxidative conditions . These observations suggest a model in which under normal conditions Crm1p would continually export Yap1p from the nucleus. During oxidative stress, the oxidation of the cysteines contained in the CRD of Yap1p leads to a reduction in the interaction between Yap1p and Crm1p, resulting in a reduced export and the accumulation of the transcription factor into the nucleus . Inside the nucleus Yap1p binds to the AP-1 response element (ARE; TGACTCA). These elements have been located in the promoter of a number of yeast genes encoding antioxidant defences (see below). Transcriptional activation requires two different segments of Yap1p that function differentially depending on the agent used to generate oxidative stress, suggesting that the mechanisms underlying the response of Yap1p to different oxidants may not be the same .
In a genetic screen designed to identify elements involved in oxidative response, one of the complementation groups obtained, pos9, showed some phenotypes resembling the yap1 mutant in relation to the sensitivity to oxidative stress . Moreover, this sensitivity did not further increase by combining the single mutations in a yap1pos9 double mutant strain, raising the possibility that both functions belong to the same regulating circuit . Pos9p (also known as Skn7p) contains a receiver motif homologous to that found in bacterial two-component signal transduction systems , suggesting that its activity could be regulated by an upstream histidine kinase. In this connection, it has been reported that the activity of Pos9p can be regulated through the same sensor (Sln1p-Ypd1p) that controls a downstream MAPK in response to hyperosmotic stress (see below). However, the role of Pos9p in oxidative stress is independent of Sln1p and Ypd1p. Although the nature of the Sln1p-independent effect of oxidative stress on Pos9p is unknown, it has been found that the activity of this factor is negatively regulated by the Ras–PKA pathway . In a recent screen, the ISM1 and CCP1 genes were identified as being involved in the oxygen-dependent activation of Pos9p . ISM1 encodes mitochondrial isoleucyl tRNA synthetase, whereas CCP1 encodes the mitochondrial cytochrome c peroxidase, suggesting that mitochondria could be important for oxidative stress signalling through Pos9p.
Pos9p also contains a region similar to the DNA-binding domain of Hsf1p [95,96], suggesting a direct role in gene induction by oxidative stress. Moreover, a Gal4-Pos9p hybrid transcription factor, consisting of the Gal4p DNA-binding domain fused to Pos9p response regulator and activation domains, was sufficient to induce a GAL1-lacZ reporter gene upon oxidative stress . This transcriptional activation is independent of Yap1p and is specifically induced by oxidative stress conditions . In agreement with its role as a transcription factor, Pos9p has been found to be required, in conjunction with Yap1p, for the oxidative induction of genes such as TRX2 and TRR1. However, Pos9p is not necessary for the induction of other Yap1p-dependent genes (, see below).
4.2Role of Yap1p and Pos9/Skn7p in the oxidative response
Both yap1 and pos9/skn7 mutants show an increased sensitivity to H2O2 and tert-butyl hydroperoxide, a likely consequence of their requirement for the induction of many genes encoding defences against ROS (see below). However, Yap1p and Pos9p do not always act together in the stress response. Thus, only Yap1p is important for cadmium resistance, whereas Pos9p is not only dispensable but appears to negatively affect this response . Lee and co-workers  have analysed the roles of Yap1p and Pos9p in the induction of genes by H2O2. They found that both factors are required for the induction of some genes, such as TSA1, TRR1 or TRX2. In contrast, induction of others, such as GLR1 and GSH1, depends on Yap1p but was stronger in the pos9 mutant than in wild-type cells . A large-scale analysis by two-dimensional gel electrophoresis delineates two gene subsets in relation to their dependence on Yap1p and/or Pos9p. The Yap1p- and Pos9p-dependent group includes most of the known ROS-scavenging activities whereas many of the genes encoding components of metabolic pathways regenerating reducing power are Pos9p-independent .
4.3Genes induced by oxidative stress: metabolic enzymes and ROS scavengers
The analysis of the induced proteins by oxidative stress [77,78] points to proteins with ROS-scavenging activities and previously identified stress proteins, including Hsps. The list also includes metabolic enzymes as H2O2-responsive targets, indicating a redistribution of metabolic fluxes in response to the oxidative agent. Remarkable changes occur in carbohydrate metabolism, which appears to be diverted to the regeneration of NADPH . Metabolic changes would include a slowdown of glycolysis as suggested by the repression of TDH2 and TDH3 (glyceraldehyde-3-phosphate dehydrogenase). The hexose phosphate pool would be redirected to the pentose phosphate pathway and the trehalose synthesis through the induction of PGM2 (phosphoglucomutase), ZWF1 (glucose-6-phosphate dehydrogenase), TKL1 and TKL2 (transketolases), TAL1 (transaldolase), UGP1 (UDP-glucose pyrophosphorylase) and TPS1 (trehalose-6-phosphate synthase).
Among the induced antioxidant activities to protect cells against ROS there are both enzymatic and non-enzymatic defences (Table 2). In the following the most important systems of both types are discussed. A complete overview of the oxidative responses in yeast has recently been published .
Table 2. Major antioxidant defences
Phenotypes of mutants
Induction by H2O2
hypersensitive to H2O2 and methylglyoxal
insensitive to H2O2 and other oxidative stresses
defects in cell cycle
defects in Met/Cys metabolism
sensitivity to H2O2
sensitivity to alkyl hydroperoxide
thioredoxin reductase (cytoplasm)
sensitivity to H2O2
sensitivity to alkyl hydroperoxide
insensitive to H2O2
thioredoxin reductase (mitochondria)
sensitive to H2O2
sensitive to oxidative stress by superoxide anion
sensitive to H2O2
moderately sensitive to oxidative stress
moderately sensitive to oxidative stress
highly sensitive to oxidative stress
sensitive to osmotic stress
sensitive to oxygen and several oxidant agents
sensitive to heat shock
sensitive to oxygen
reduction in respiratory competence
sensitive to heat shock
sensitive to lethal heat shock
Glutathione is the most abundant low-molecular-mass intracellular thiol compound. The important role of glutathione in the adaptive response of S. cerevisiae to oxidative damage is suggested by the enhanced H2O2 sensitivity and by the suppression of the H2O2-adaptive response caused by depletion of cellular glutathione . Glutathione biosynthesis requires the products of the GSH1 and GSH2 genes, encoding, respectively, γ-glutamylcysteine synthetase and glutathione synthase. gsh1 mutants are hypersensitive to hydrogen peroxide whereas gsh2 are insensitive to this and other oxidative agents . These phenotypes suggest that γ-glutamylcysteine can be a substitute antioxidant for glutathione. According to its main role in oxidative response, only GSH1 is induced by oxidants . Yap1p mediates both basal and hydrogen peroxide induction of GSH1, but is not required for superoxide anion-mediated induction (; Table 2). In addition, gsh1 mutants are unable to grow in non-fermentable carbon sources, suggesting an essential role for glutathione in protecting mitochondria from ROS produced during respiration .
Thioredoxin is a small protein with a conserved sequence (Trp-Cys-Gly-Pro-Cys) in its active site. When reduced, the dithiol group of the active site is able to catalyse the reduction of disulfites in a number of proteins. The thioredoxin system is composed by thioredoxin (Trx), thioredoxin reductase (Trr) and NADPH and in yeast there are two of such systems, one in the cytoplasm with two thioredoxins (Trx1p and Trx2p) and one thioredoxin reductase (Trr1p) and a second in the mitochondria with one thioredoxin (Trx3p) and one thioredoxin reductase (Trr2p) . Deletion of TRX1 and TRX2, besides defects in cell cycle and methionine/cysteine metabolism , causes sensitivity to hydrogen peroxide  and alkyl hydroperoxide  as occurs in trr1 mutants [107,108]. As mentioned, both TRX2 and TRR1 are induced by hydrogen peroxide in a Yap1p- and Skn7p-dependent way (Table 2). In relation to the mitochondrial system, the trr2 mutant is twice as sensitive to H2O2 as the wild-type; however, the behaviour of the trx3 mutants in this oxidative condition is similar to wild-type. These results suggest a role for Trr2p as an antioxidant independent of its partner Trx3p in a situation similar to that found in mammalian cells .
Acting in a similar way to thioredoxins, yeast contains glutaredoxins with two subfamilies that differ in the number of cysteine residues at the active site. The first subfamily protects cells against H2O2 (Grx2p) and superoxide anions (Grx1p) . The second family includes three additional members (Glx3p, Glx4p and Glx5p) and, among them, Grx5p seems to play an important role in protection against oxidative stress, both during ordinary growth conditions and after exposure to oxidants such as H2O2 and menadione .
Superoxide dismutase (SOD) acts to disproportionate two molecules of superoxide anion to hydrogen peroxide and water, playing an important role in protecting aerobic organisms against oxidative damage. S. cerevisiae contains two forms of SOD. Sod1p is a copper- and zinc-bound SOD (Cu/Zn-SOD) with a cytoplasmic localisation and Sod2p, a mitochondrial, manganese-bound SOD (Mn-SOD). Both SOD1 and especially SOD2 require Yap1p and Skn7p for hydrogen peroxide-mediated induction (; Table 2). Sod2p is the primary defence against the toxicity of the superoxide generated in the mitochondria under physiological conditions. According to this, the sod2 mutants show hypersensitivity towards oxygen that can be suppressed by mutations that block respiration . Mn-SOD has been found to protect certain mitochondrial enzymes against mitochondrially generated superoxide. Thus, in sod2 mutants the activity of aconitase and succinate dehydrogenase, but not cytochrome oxidase and ATPase, decreases greatly as cells enter the stationary phase . The role of Sod1p seems to be more complex. sod1 mutants are also sensitive to oxygen , to the superoxide-generating agent plumbagin  and paraquat . Besides, these mutants are auxotrophic for some amino acids, such as cysteine and methionine, when grown aerobically in the absence of additional environmental oxidants, suggesting that the metabolic defects shown by the sod1 mutants arise as a product of metabolism-derived oxygen radicals [114,115]. Another role of Cu/Zn-SOD is to buffer the intracellular copper concentration and, in accordance, sod1 mutants show increased sensitivity to copper . The ability to bind Cu by Sod1p depends on the presence of the LYS7 gene, which besides its role in the lysine biosynthetic pathway has been involved in Cu metabolism and oxidative stress protection .
The enzyme catalase acts by removing the hydrogen peroxide. In yeast there are two different enzymes, catalase A and catalase T, located in the peroxisome and cytoplasm, respectively. Catalase A is encoded by the CTA1 gene and its main function seems to be to remove the H2O2 resulting from fatty acid β-oxidation . Similarly to other antioxidant-scavenging agents, full induction of CTT1 requires both Yap1p and Skn7p (; Table 2). The role of catalase T, encoded by the CTT1 gene, is not as clear, although it has been involved in stress response [118,119]. The importance of catalases in protecting against hydrogen peroxide is controversial. In yeast, only stationary phase cta1ctt1 double mutant cells are more sensitive than wild-type to H2O2 stress, but both mutant and wild-type showed a similar susceptibility in the exponential growth phase .
5.1The high osmolarity glycerol response (HOG) pathway
The control of water content is essential for all types of cells. When yeast is exposed to a hyperosmotic shock a loss of cytoplasmic water occurs and several mechanisms are initiated to counteract cell dehydration and protect the cellular structures. A MAPK pathway has been involved in the transduction of the signal generated by an increase in osmolarity (reviewed in Fig. 2) contains a MAPK module, which consists of a cascade of kinases which, when activated, culminates in the phosphorylation and activation of Hog1p (high osmolarity glycerol) protein kinase. In this cascade the dual-specificity serine/threonine tyrosine MAPKK that phosphorylates Hog1p is Pbs2p. Pbs2p is in turn phosphorylated and activated by the serine/threonine MAPKKKs Ssk2p and Ssk22p. The sensor of external osmolarity is the transmembrane protein kinase Sln1p that connects the cell membrane with the MAPK module by using a phospho-relay system (Ssk1p+Ypd1p) similar to the bacterial two-component system. There is a second membrane osmosensor called Sho1p, which transmits a signal to Pbs2p via the MAPKKK Ste11p . Phosphorylation of Hog1p by Pbs2p induces its nuclear accumulation [121,122]. This accumulation is transient  and coincides with the osmotic induction of a variety of genes, suggesting a role for Hog1p in gene expression, although the identification of the Hog1p kinase targets has remained elusive.
Several transcription factors have been involved in the stimulation of gene transcription after osmotic shock. Among them, Msn2p and Msn4p, the binding factors of STREs, are involved in the induction of many genes in response to a variety of stress conditions, including osmotic stress (see above). Interestingly, the osmotic induction of an artificial promoter containing STRE sequences depends on the HOG pathway, suggesting that Msn2/4p activity could be regulated by Hog1p . However, although osmotic stress induces nuclear accumulation of both Hog1p and Msn2/4p, the increase in nuclear concentration of these transcription factors can also be observed in a hog1 mutant and vice versa [25,122]. So far, the only direct relationship between Hog1p and Msn2/4p has been established at the level of duration of nuclear residence of Hog1p after osmotic stress . It has been suggested that Msn2/4p act by providing a nuclear anchorage for the kinase that might protect Hog1p against the action of specific protein phosphatases . Although the relative importance of this retention is difficult to estimate, it could be responsible, at least in part, for the reduced level of expression, but not for the induction factor, shown by a STRE-LEU2-lacZ gene fusion in the msn2msn4 mutant under osmotic stress. A rapid inactivation by dephosphorylation of Hog1p in the msn2msn4 mutant could lead to a less efficient activation of other putative STRE-binding factors under osmotic stress.
Recently, two transcription factors, Hot1p and Msn1p, have also been related to the osmotic transcriptional induction . Hot1p physically interacts with Hog1p and is required for full expression of several osmotically induced genes , although binding to DNA has not yet been demonstrated. MSN1, like MSN2 and MSN4, was isolated as a multicopy suppressor of snf1 mutants  and, besides osmotic induction, is involved in several pathways [124,125]. Both Hot1p and Msn1p are responsible for the bulk of the Msn2/4p-independent osmotic stress activation of several yeast genes . Finally, the Sko1p repressor, which binds to CRE-like sequence, mediated the osmostress signalling that, through the HOG pathway, regulates ENA1 expression . However, whether these factors are direct targets of the HOG1 pathway has yet to be proved.
5.2Genes induced in response to osmotic stress
In two recent papers [127,128], genome-wide approaches have been used to analyse the transcriptional response to osmotic stress. These analyses show that the mRNA level of 186 genes is at least threefold higher after the addition of 0.7 M NaCl and/or 0.95 M sorbitol  whereas more than 1300 genes are induced in a similar way after a short exposure (10 min) to a lower concentration of NaCl (0.4 M). Differences in the number and genes identified in each analysis are a consequence of the different conditions used. Thus, when cells are exposed to 0.4 M NaCl for a longer period (20 min) the number of responsive genes is reduced about 10-fold in relation to those induced after a 10-min exposure . The induced genes have been grouped in functional categories covering very different biochemical processes [127,128]. Among them, the family of genes involved in carbohydrate metabolism is well represented, including the enzymes required for glycerol synthesis (see below) and sugar transporters. Also the expression of genes encoding enzymes in trehalose and glycogen production was increased [127,128]. However, neither trehalose nor glycogen is accumulated by yeast cells under these conditions, and trehalose and glycogen production are not essential for growth at high osmolarity [127,129,130].
This same approach has been used to identify those genes whose osmotic induction was dependent on the HOG pathway [127,128]. The results obtained indicate that this pathway is involved in the regulation of most of the osmotic-induced genes. However, in many cases the induction was not completely abolished in the hog1 mutant, suggesting the involvement of other signalling pathways [127,128]. The list of genes whose expression is affected by the deletion of HOG1 includes many of the genes that are affected by the deletion of HOT1 or/and MSN2/MSN4. However, there are genes, such as ALD2 and ALD3, where transcriptional induction by osmotic stress depends on Msn2/4p but is independent of the HOG pathway .
In relation to the phenotypes of HOG pathway mutants, considering that the HOG pathway controls the osmotic induction of the glycerol-producing enzymes GPD1 and GPP2, and the role in yeast of glycerol as a compatible solute (see below), it is not surprising that both hog1 and pbs2 mutants fail to grow in high-osmolarity medium . However, the importance of other Hog1-dependent mechanisms in the protection against osmotic stress is revealed by hypersensitivity of the hog1 mutant compared to the gpd1 mutant .
5.3The accumulation of glycerol as a compatible solute
Among the yeast responses to an increase in external osmolarity, the accumulation of the compatible solute glycerol deserves special attention (reviewed in ). The accumulation of glycerol in response to a hyperosmotic challenge includes both the regulation of its production and its efflux. Glycerol is produced by a reduction of dihydroxyacetonephosphate to glycerol-3-phosphate (G-3-P) that is then dephosphorylated to yield glycerol (Fig. 2). The first reaction is catalysed by a cytosolic NAD+-dependent glycerol-3-phosphate dehydrogenase for which two isoenzymes have been described, encoded by GPD1 and GPD2. Also two different glycerol-3-phosphatases have been identified, encoded by the two highly homologous genes GPP1 and GPP2. Phenotypes shown by gpd1 and/or gpd2 mutants reveal that Gpd1p and Gpd2p, and thereby the glycerol production, have two different roles in yeast. Thus, null gpd1 mutants are sensitive to osmotic stress but not gpd2 mutants, which, in contrast, show a slow-growing phenotype under anaerobic conditions [132,134]. When the expression of the GPD1 and GPD2 genes is analysed a different pattern emerges: GPD1 is induced by osmotic stress  whereas GPD2 expression increases in the absence of oxygen . These results have been interpreted to imply that Gpd1p plays an important role in osmoregulation but Gpd2p acts mainly in redox balancing. A similar situation could occur for the glycerol-3-phosphatase-encoding genes, where the transcription of GPP2 but not of GPP1 is induced by high osmotic conditions. Induction of both GPD1 and GPP2 genes by osmotic stress partly depends on the HOG pathway [22,135,137]. Both genes contain STRE sequences in their promoters, although deletion of MSN2 and MSN4 has little effect on their osmotic induction. Only in cells lacking Msn1p, Msn2p, Msn4p and Hot1p is the induction of these genes almost completely abolished .
The role of glycerol as a compatible solute is important for its accumulation in the cell after a hyperosmotic shock. This accumulation is regulated in part by the export of glycerol through the plasma membrane protein Fps1p (reviewed in ). Fps1p is a member of the MIP family of channel proteins characterised by six transmembrane domains . This glycerol facilitator closes after hyperosmotic shock in 1–2 min, allowing glycerol accumulation. Glycerol accumulation is important for cell survival since a mutant lacking the N-terminal regulatory domain of Fps1p, which shows an uncontrolled glycerol extrusion through the channel, is highly sensitive to high osmolarity . The regulation of the Fps1p function seems to be independent of any of the known yeast osmosensing signalling pathways. Moreover, the down-regulation of Fps1p after osmotic shock occurs very rapidly and is almost complete within 15 s, raising the possibility that Fps1p itself senses changes in osmolarity controlling glycerol transport .
6Trehalose: a non-protean protectant
Trehalose has been regarded historically as a reserve carbohydrate in yeast. Its pattern of accumulation and utilisation suggests that it could serve as an energy source during starvation. Thus, trehalose is barely detectable in exponentially growing cells on glucose, but in stationary phase cells, it makes up to 20% of the dry weight of the cell (cited in ). This sugar is synthesised by trehalose synthase, a complex of three subunits encoded by TPS1 (trehalose-6-phosphate synthase), TPS2 (trehalose-6-phosphate phosphatase) and TSL1/TPS3 (encoding a subunit with putative regulatory function) [141,142]. Trehalose can be mobilised by two different trehalases encoded by NTH1 (neutral trehalase) and ATH1 (vacuolar acid trehalase) . A third gene, NTH2, highly homologous to NTH1, has been identified during the sequencing of the S. cerevisiae genome although its disruption does not cause a decrease in trehalase activity . Regulation of trehalose metabolism includes both transcriptional and post-translational mechanisms. Neutral trehalase is a known substrate of PKA . The activity of the enzyme complex catalysing the synthesis of trehalose increases with heat, but it is not certain whether the activity is influenced by phosphorylation . At the transcriptional level, STREs have been identified at the promoters of NTH1, TPS1, TPS2, TPS3 and TSL1[13,145,146]. According to this, the expression of most of the genes encoding enzymes of trehalose metabolism is co-regulated. They are induced under stress conditions in a similar way to other genes containing STREs [130,146] and this induction is lost in an msn2msn4 double mutant strain (; M.T. Martínez-Pastor and F. Estruch, unpublished results). However, both synthetic and degradative enzymes are transcriptionally induced, suggesting a recycling of reserve carbohydrates upon stress conditions. The existence of trehalose turnover, at the expense of ATP, is supported by the increase in the accumulation of trehalose observed in the nth1Δ mutant compared to the wild-type during adaptation to salt stress . Blomberg  has proposed that this futile cycle (together with a similar glycerol turnover) acts as a glycolytic safety valve, to avoid substrate-accelerated death under stress.
Trehalose levels have been correlated with cell survival under adverse conditions . The protective role of trehalose is an obvious interpretation of its production in response to stress. This role is also supported by the phenotypes shown by mutants in trehalose metabolism. Thus, ath1 mutants accumulate higher levels of trehalose and show an increased tolerance to dehydration, freezing and toxic levels of ethanol . On the other hand, disruption of TPS1 prevents cells from synthesising trehalose and reduces their thermotolerance . However, the direct role of trehalose in stress protection has been questioned by other observations. The finding that Hsp synthesis is impaired in tps1 strains suggests that the thermosensitive phenotype of this mutant could result from a failure to induce the production of Hsps fully . The fact that nth1nth2 mutants are impaired in their ability to survive and recover from severe heat also argues against a role of trehalose in protecting cells [140,151,152]. These discrepancies have been resolved by recent results that suggest a direct function for trehalose in thermotolerance (for review see ). In vitro observations revealed that trehalose stabilises proteins during heat shock, and it is even able to suppress the aggregation of proteins that have already been denatured, a capacity previously ascribed exclusively to Hsps . Accumulation of trehalose when cells are exposed to high temperatures results in cytosolic trehalose concentration at approximately 0.5 M, a concentration that has been shown in vitro to extend the range of temperatures over which proteins retain their native state . However, these trehalose concentrations have been shown to interfere with the protein refolding by molecular chaperones, both in vitro and in vivo . Taken together, these results propose an explanation for the role of trehalose as a protectant and also for the rapid degradation of trehalose after heat shock. According to this, the phenotype shown by the nth1 mutants would be the consequence of a less efficient refolding of proteins by molecular chaperones at high trehalose concentrations .
As mentioned, the induction of reserve carbohydrate genes is defective in msn2msn4 double mutants, causing the non-accumulation of trehalose under heat shock and osmotic stress  and, probably, a reduction in carbohydrate recycling. How much the stress-sensitive phenotype shown by the msn2msn4 mutants is due to defects in reserve carbohydrate metabolism or in the induction of stress induced genes remains to be clarified.
Recently used global approaches in S. cerevisiae for the analysis of gene expression under stress conditions reveal that a large number of proteins are induced by one or several types of stress [34,35,77,98,127,128]. The list includes proteins whose function leads to protection and repair of damaged components but also metabolic enzymes, indicating that a re-organisation of metabolic fluxes is required, or at least is convenient, to ensure survival and adaptation to the stress condition. In spite of the number and variety of proteins that are induced by stress, the number of transcriptional regulatory systems is low. Three major systems have been identified: two of them, whose transcription factors are Hsf1p and Yap1p, are stress-specific, whereas the third, with Msn2p and Msn4p as downstream elements, is activated by very different stress conditions.
A major point is how these pathways sense the environmental stress and how the information is transmitted to activate the transcription factors. As discussed, sensing and activation are closely related for both Hsf1p and Yap1p. In relation to STRE, further work is required to establish whether the variety of stress conditions that induces transcription via this element converges in a unique stress signal or different signals are able to activate Msn2p and Msn4p. Besides, at least three signal pathways converge on the STRE: the MAPK HOG cascade, the PKA pathway and the TOR pathway, suggesting a more complex mechanism of signal transduction.
A second question addressed in this review has been the phenotypic consequences of impairing the stress regulatory systems or the genes that are regulated through them. These phenotypes are highly variable. Thus, mutants in HSF1 are not viable because some of the genes activated by Hsf1p are required for growth at normal temperatures, whereas yap1 mutants are only defective in their response to oxidants. The stress-related phenotype of the msn2msn4 mutants is paradoxical. Although this mutant shows pleiotropic stress sensitivity, the phenotype is milder than expected, considering the number of STRE-containing genes. A possible explanation for this is that STRE are redundant with other regulatory systems for those genes with essential roles in the stress response. A similar situation is found in mutants lacking individual stress-induced genes, ranging from the absolute requirement of HSP104 for induced tolerance to the apparent absence of stress-sensitive phenotypes in hsp26 and/or hsp12 mutants.
Finally, although it has not been dealt with in this review, post-transcriptional regulation can also play an important role in protecting cells against stress. A clear example is found in the trehalose metabolism, where simultaneous induction of genes encoding both synthesising and degradative enzymes suggests that the net production or mobilisation of this carbohydrate must depend on post-transcriptional events.
Thus, in recent years major progress has been made in improving our knowledge about the stress response in yeast. Open questions such as how the cell cycle is regulated in response to environmental stresses and how to apply this knowledge to improve the resistance of industrial yeast are challenges for the coming years.
The author is indebted to Juana María Gancedo and Carlos Gancedo for critically reviewing the manuscript. I am also grateful to E. Matallana, J.E. Pérez-Ortín and M. del Olmo for comments on this review and to H. Ruis and C. Schüller for permission to cite unpublished observations. I also thank the referees who reviewed the manuscript for their constructive criticisms. Our own experimental work reported in this article was supported by Spanish Ministry of Education Grants PB97-1468-C02-02 and ALI98-0848.