Sodium-induced GCN4 Expression Controls the Accumulation of the 5′ to 3′ RNA Degradation Inhibitor, 3′-Phosphoadenosine 5′-Phosphate*

Most cytoplasmic mRNAs are decapped and digested by the 5′–3′-exonuclease Xrn1p in Saccharomyces cerevisiae. The activity of Xrn1p is naturally inhibited in the presence of 3′-phosphoadenosine 5′-phosphate (pAp), a metabolite produced during sulfate assimilation that is quickly metabolized to AMP by the enzymatic activity of Hal2p. However, pAp accumulates and 5′–3′ degradation decreases in the presence of ions known to inhibit Hal2p activity, such as sodium or lithium. We have shown that yeast cells can better adapt to the presence of sodium than lithium because of their ability to reduce pAp accumulation by activating HAL2 expression in a Gcn4p-dependent response, a regulatory loop that is likely to be conserved in different yeast species. We have thus identified a new role for the transcriptional activity of Gcn4p in maintaining an active mRNA degradation pathway under conditions of sodium stress. Since deregulation of proteins involved in different metabolic pathways is observed in xrn1Δ mutants, the maintenance of mRNA degradation capacity is likely to be important for the accurate and rapid adaptation of gene expression to salt stress.

Although the study of gene regulation has traditionally focused on transcription as a major regulator of gene expression, it has recently become apparent that the post-transcriptional regulation of gene expression may play an equally important role (1). Xrn1p is a major player in this post-transcriptional control in yeast because it is responsible for the turnover of the majority of cytoplasmic mRNAs (2). This role is evident in both the general deadenylation-dependent mRNA decay (3,4) and nonsense-mediated mRNA decay pathways (5). Xrn1p is not essential, but XRN1 deletion mutants grow slowly, and mutations in XRN1 are known to have pleiotropic effects (6). Since Xrn1p acts at the last step in mRNA turnover, after decapping, xrn1 mutants accumulate decapped intermediates (2) that can persist in polysomes (7,8). This defect is potentially detrimental to cellular fitness, through an inability to remove decapped transcripts from the translation machinery in a timely fashion. An analysis of global protein expression in xrn1⌬ mutant strains revealed both up-and down-regulation of different sets of proteins involved in several metabolic pathways (9). Such alterations in gene expression would certainly be problematic under conditions of environmental stress, when cells are required to establish a particular gene expression program for adaptation.
It has been demonstrated that inhibition of 5Ј-3Ј-exoribonuclease activity by the metabolite, 3Ј-phosphoadenosine 5Ј-phosphate (pAp), 3 occurs in the presence of toxic ions such as sodium and lithium (10) (Fig.  1). pAp is produced during the assimilation of sulfate, an essential process in all living organisms. Sulfate assimilation in Saccharomyces cerevisiae is initiated by the production of adenosine 5Ј-phosphosulfate (ApS) from ATP and sulfate (11). Phosphorylation of ApS yields 3Ј-phosphoadenosine 5Ј-phosphosulfate (pApS), which is then reduced to sulfite and pAp. Sulfite is reduced further to sulfide, which is incorporated into homocysteine, which in turn can be metabolized to methionine (12) (Fig. 1). The product of the HAL2/MET22 gene is required for the dephosphorylation of pAp to AMP, but the presence of sodium or lithium in the growth medium inhibits the enzymatic activity of Hal2p and causes accumulation of pAp (13) (Fig. 1). Part of the salt toxicity is known to be due to the accumulation of pAp since sodium or lithium toxicity in cells can be efficiently relieved by overexpression of Hal2p, which limits pAp accumulation (10,14,15). Similarly, the presence of extracellular methionine, which limits sulfate assimilation and consequently pAp accumulation, is known to increase salt tolerance (10,16).
Many metabolic activities are sensitive to salt inhibition, but cells counter the accumulation of toxic ions in the cytoplasm by homeostatic mechanisms, which maintain intracellular ion concentrations and are essential for living cells (17,18). For instance, many strains of the yeast S. cerevisiae contain a major sodium and lithium extrusion ATPase encoded by the ENA1-4/PMR2 gene cluster (19,20) (Fig. 1). Extrusion of toxic ions, such as sodium or lithium (lithium being a sodium analog with higher toxicity), occurs efficiently in cells and affords salt tolerance, although it does not efficiently prevent pAp accumulation under high salt conditions. Degradation of the body of mRNAs has already been extensively studied in terms of the effect of mutations in trans-acting factors or cisacting elements on mRNA stability (21), but only a few investigations into how external stimuli regulate global mRNA turnover have been pursued in S. cerevisiae. One example of specific effects comes from studies of the target of rapamycin pathway, which senses external nutrient availability and mediates changes in gene expression upon diauxic shift (22) by controlling the turnover of specific mRNAs (23). The beststudied mRNAs in terms of a context-specific regulation of RNA stability are mammalian transcripts containing AU-rich elements, also observed in a few yeast transcripts (24). We are only just beginning to learn how many RNA-binding proteins associated with these AU-rich elements modulate mRNA decay in response to environmental stimuli (1,25).
Gcn4p is a transcriptional activator of hundreds of genes in response to many different stress or starvation conditions (26,27). Adaptation to changes in the extracellular environment is a critical event for cell survival, and the activation of GCN4 expression in the presence of NaCl may contribute to the well known ability of yeast to adapt to extreme perturbations in culture conditions. The study we describe here has provided the first evidence that Gcn4p mediates the regulation of mRNA degradation in response to external stimuli. We demonstrated that cells take advantage of the derepression of GCN4 translation in the presence of sodium to maintain the degradation activity of 5Ј-3Ј-exonucleases. We have shown that Gcn4p participates directly in the activation of HAL2 expression, thus permitting the limitation of pAp accumulation. In cells deficient in the 3Ј-5Ј-mRNA decay pathway, we have shown that cells unable to maintain Gcn4p-dependent 5Ј-3Ј-mRNA degradation activity display severe growth defects in the presence of salt.

EXPERIMENTAL PROCEDURES
Strains and Media-Yeast were grown in media based on Hartwell's synthetic complete or synthetic minimum medium (SD) as described elsewhere (28). For cell cultures, minimal medium SD was completed with tryptophan, isoleucine, arginine, valine, lysine, leucine, histidine, adenine, and uracil in concentrations based on Hartwell's synthetic complete media to minimize GCN4 expression. Uracil was omitted to keep selection for plasmids p16, p238, p180, and p227 when present in cells. Histidine was omitted in liquid cultures treated with the antimetabolite 3-amino-1,2,4-triazole (3-AT) at a final concentration of 10 or 50 mM as indicated. The GCN4 and HAL1 null alleles were obtained by one-step gene replacement using PCR fragments of the KanMX gene, amplified with long primers containing 5Ј and 3Ј sequences of GCN4 and HAL1, respectively (29). In the same way, the sequence of the potential Gcn4p-binding site 5Ј-TGACTC (which we call GCRE for Gcn4presponse element) was modified by insertion of the KanMX gene upstream to create the new sequence 5Ј-TGTCGC.
Plasmids-In this study, we used plasmid p16 (URA3 CEN) and plasmid p238, which overproduces Gcn4p constitutively (GCN4c URA3 CEN) (30). Plasmids p180 and p227 contain the GCN4-lacZ fusion with or without the four upstream open reading frames, respectively, and have been described previously (31,32). These plasmids were generous gifts from A. Hinnebusch. Yeast transformations were performed by the LiAc procedure (28).
Preparation of Yeast Nucleotide Extracts and HPLC Analysis-Yeast nucleotide extracts and HPLC analysis were done as described previously with minor modifications (13). Yeast cells were grown to an absorbance of 0.2 at 600 nm and then treated with LiCl or NaCl at the indicated concentrations. At the indicated times, samples corresponding to 20 OD 600 nm (2.10 8 cells) were harvested by filtration through Millipore filters (0.45 m) and extracted with 1 ml of 2 N perchloric acid at 0°C for 15 min. Extracts were clarified by centrifugation at 12,000 rpm for 3 min. 0.8 ml of supernatant were neutralized with 0.36 ml of 2 N KOH and 4 M H 2 PO 4 and centrifuged as above. Supernatants were filtered through Millipore HV filters (0.45 m) and stored at Ϫ80°C. 100 l of each extract were analyzed by HPLC (Waters). Samples were applied to a reversed phase C18 column (Lichrosphere 100, 4 ϫ 250 mm, 5-m particle size, Merck), eluted, and detected as described (33). Nucleotide peaks were identified by co-injection with standards (AMP, ADP, ATP, and pAp from Sigma).
Northern Blot Analysis-Total RNA was extracted as described (34) from 10 ml of cell cultures grown at 30°C to mid-log phase and after NaCl treatment when indicated. 20 g of RNA were loaded onto a 1.25% agarose-formaldehyde gel. The size-fractionated RNA was blotted onto a Hybond-N membrane (Amersham Biosciences), and hybridization was performed as recommended by the supplier. Probes against scR1 RNAs were generated by the random priming method, using PCR products as templates. Amplification of genomic scR1 was performed using oligonucleotides obc13 (5Ј-GGTGGGATGGGATACGTTG) and obc14 (5Ј-CGGCCACAATGTGCGAG). The probe AJO130, specific for ITS1 (35), was end-labeled by polynucleotide kinase and [␥-32 P]ATP.
Western Blot Analysis-Whole cell protein extracts were prepared from 20 ml of cell cultures grown at 30°C to mid-log phase and after NaCl treatment for an additional 2 h as indicated. Western blot experiments were performed using rabbit anti-Hal2p (14). 40 mg of total proteins were loaded and separated by electrophoresis on 12% acrylamide gels as described (36). Transfer to nitrocellulose membranes and Western blotting with antibody (diluted 5000ϫ) were performed as described (37) using a peroxidase-coupled secondary antibody (diluted 5000ϫ), SuperSignal chemiluminescent substrate (Pierce), and Kodak X-Omat AR films.
␤-Galactosidase Assays-Cells were grown at 30°C to mid-log phase and treated with 3-AT, NaCl, LiCl, or KCl treatment for an additional 2 h as indicated. Whole-cell extracts were prepared as described elsewhere (38), except that Z buffer without ␤-mercaptoethanol was used as the extraction buffer. Protein concentration was determined by the Bio-Rad assay, and ␤-galactosidase activity was measured as described (39). ␤-Galactosidase units are expressed in nanomoles of 2-nitrophenyl-␤d-galactopyranoside hydrolyzed per minute per nanogram of protein.

Sodium Stimulates GCN4 Expression More Efficiently than Lithium-
It has been demonstrated previously that lithium treatment of wild type yeast strains results in strong inhibition of both Xrn1p and Rat1p 5Ј-3Јexonuclease activities (10). Lithium inhibition of Hal2p, a specific phosphatase that breaks down pAp, causes increased cellular levels of pAp, a direct inhibitor of Xrn1p and Rat1p. Although both lithium and sodium are inhibitors of Hal2p, lithium shows higher levels of toxicity and can be used as a growth inhibitor at lower concentrations than sodium (19). Because NaCl stress was also shown to induce the general amino acid FIGURE 1. Sodium and lithium lead to accumulation of pAp, a byproduct of the overlapping methionine biosynthesis and sulfate assimilation pathways (11). Sulfate assimilation is derepressed in the absence of extracellular methionine. Sulfate enters the cell via the Sul1p and Sul2p transporters. Activities of Met3p, Met14p, and Met16p contribute to sulfate assimilation by producing sulfite, which is used in the production of methionine. Production of ApS, 3Ј-phosphoadenosine 5Ј-phosphosulfate (pApS), and pAp, which is metabolized into AMP and inorganic phosphate by Hal2p, is greater in cells starved for methionine. In the absence of salt, Hal2p activity renders pAp almost undetectable. However, pAp accumulates in the presence of LiCl (indicated by Li ϩ ) or NaCl (indicated by Na ϩ ) because these ions inhibit Hal2p (13). Expression of MET3, MET14, and MET16 is enhanced by the transcriptional activator Met4p under conditions of extracellular methionine limitation (dashed arrows) (11,54). Ena1p is a (Na ϩ ,Li ϩ ) extrusion ATPase (20,57) activated in the presence of salt (58). FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 control response (40,41), we asked whether LiCl could have a similar effect. Stimulation of GCN4 translation occurs via a mechanism involving four short upstream open reading frames located in the 5Ј-noncoding region of the GCN4 mRNA. Since Gcn2p-dependent phosphorylation of eIF2␣ is a prerequisite for the induction of GCN4 translation in response to histidine starvation or NaCl stress, we used a GCN4-lacZ translational fusion containing all four upstream open reading frames in cells expressing wild type or a mutant form of GCN2 to measure and compare GCN4 translation in the presence and absence of LiCl (Table 1, p180 plasmid). As controls for GCN4 derepression, Gcn4-lacZ enzyme activity was also measured upon histidine starvation, mimicked by the addition of 3-AT, and in the presence of sodium (Table 1, 0.2 M NaCl or 0.4 M NaCl). Although NaCl and 3-AT caused a significant increase in ␤-galactosidase levels, as expected (32,41), LiCl had a much smaller effect ( Table 1). All increases were essentially abolished in cells deficient for Gcn2p activity (gcn2-1 mutant). As shown previously, the activation by NaCl was not observed in the presence of KCl (Table 1, 0.5 M KCl), suggesting that this effect is not due to a general osmotic stress response (41). We also analyzed the expression of a mutant GCN4-lacZ fusion under the same conditions. In this fusion, the four upstream open reading frames are non-functional, thus alleviating translational control ( Table 1, p227 plasmid). With this construct, Gcn4-lacZ enzyme activity was elevated in all conditions and independent of Gcn2p. Thus, exposure of cells to LiCl causes only a weak translational activation of GCN4, slightly greater than that observed in the presence of KCl but significantly lower than the induction by histidine starvation or sodium stress. This implies that, although lithium has been shown to be a more potent inhibitor of Hal2p activity, sodium is more efficient at activating GCN4 expression.

Gcn4p Controls Xrn1p Activity under Stress
GCN4 Expression Limits pAp Accumulation under Conditions of NaCl or LiCl Stress-Since lithium has more toxic effects on cell growth than sodium and leads to inhibition of both sulfate assimilation and RNA degradation (10), we asked whether the sodium-induced expression of GCN4 might somehow limit pAp accumulation. We performed HPLC analysis of nucleotide extracts from control and LiCl-treated or NaCl-treated wild type cells. As observed previously, a nucleotide peak corresponding to pAp accumulated after treatment of cells with LiCl (WT, 0.2 M LiCl) for 1 h when compared with untreated cells (WT, SD) (Fig. 2). Incubation of cells in the presence of 0.4 M NaCl for the same period did not increase the intracellular pAp concentrations to detectable levels (WT, 0.4 M NaCl). This is due in part to the much greater inhibition of Hal2p by lithium than sodium (15). Detectable levels of pAp were visible, however, in NaCl-treated cells lacking GCN4, suggesting that GCN4 indeed limits pAp accumulation (Fig. 2, gcn4⌬ 0.4 M NaCl). The effect of the gcn4⌬ mutation was only visible in the presence of salt since no pAp accumulated in this mutant in the absence of NaCl (gcn4⌬, SD). To confirm that GCN4 expression inhibits the accumulation of pAp, we compared pAp levels in LiCl-treated wild type cells, in which pAp levels are high, with cells bearing a GCN4c constitutive allele that overproduces Gcn4p (GCN4c, 0.2 M LiCl). As predicted, a significant decrease in pAp levels occurred under these conditions. Similarly, treatment of cells with both LiCl and 3-AT, which causes histidine starvation and strong derepression of GCN4 expression, also led to a decrease in pAp levels, in this case to near baseline levels (WT, 0.2 M LiCl ϩ3AT).
The Xrn1p-sensitive RNA ITS1 Accumulates in NaCl-treated Cells Deficient in GCN4 Expression-We chose to study the effect of pAp accumulation on an RNA known to be sensitive to Xrn1p in wild type and gcn4⌬ strains by Northern blot analysis (Fig. 3). Xrn1p activity

Gcn4p Controls Xrn1p Activity under Stress
participates in the degradation of the internal transcribed spacer (ITS1), a product of rRNA processing (42). Strains defective for Xrn1p exonuclease activity show increased levels of ITS1 RNA due to its stabilization (Fig. 3, lanes 3 and 4), independently of the presence of NaCl (compare with lanes 7 and 8). Within the first hour of NaCl addition, cells deficient for Gcn4p accumulated higher levels of ITS1 RNAs than wild type cells (Fig. 3, compare lanes 5 and 6 and lanes 2 and 6). These results are in perfect correlation with the accumulation of pAp, the inhibitor of Xrn1p, observed in the gcn4⌬ mutant strain in the presence of NaCl (Fig. 2). We wondered whether the inhibition of the 5Ј-3Ј-mRNA degradation pathway, in gcn4⌬ cells in the presence of NaCl, would affect cell growth since xrn1⌬ mutant cells have a slow growth phenotype (6). This slow growth phenotype is independent of GCN4 expression (Fig. 4A, see the similar growth rate of xrn1⌬ or xrn1⌬gcn4⌬ strains). In the presence of salt, however, both the xrn1⌬ and the xrn1⌬ gcn4⌬ strains show similar growth rates to the wild type strain, suggesting that some adaptation has occurred, possibly through increased activity of the alternative 3Ј-5Ј degradation pathway (43). Inactivation of both the 5Ј-3Ј and the 3Ј-5Ј pathways has been shown to be lethal in yeast (44). We thus decided to measure the effect of salt on growth of a gcn4⌬ mutant in the absence of cytoplasmic 3Ј-5Ј degradation by also removing Ski2p. Ski2p is a putative RNA helicase necessary for cytoplasmic 3Ј-5Ј-mRNA degradation (45). We predicted that in the absence of Ski2p, the accumulation of pAp in the presence of salt in cells also lacking Gcn4p would show a slow growth phenotype due to inhibition of Xrn1p. Indeed, a significant decrease in growth rate was observed in the ski2⌬ gcn4⌬ strain grown in the presence of salt when compared with the control strains ski2⌬ or gcn4⌬, whereas these three strains grew similarly in the absence of salt (Fig. 4). On the other hand, the xrn1⌬ gcn4⌬ strain (in which only the cytoplasmic 3Ј-5Ј-mRNA degradation pathway is active), the wild type strain, and the xrn1⌬ and gcn4⌬ strains showed similar growth rates in the presence of sodium (Fig. 4B), in agreement with the hypothesis that the 3Ј-5Ј degradation pathway is not inhibited by the accumulation of pAp (43).

Inhibition of HAL2 Expression, and Not an Imbalance in Ion Homeostasis, Is Responsible for pAp Accumulation in NaCl-treated gcn4⌬
Cells-In the absence of extracellular methionine, the expression of the HAL2 gene, which contains a specific cis-acting element responding to methionine limitation upstream of its promoter sequence, is expected to be maximal (14). Under this condition, sulfate assimilation is efficiently triggered, Hal2p activity is sufficient to convert all pAp to AMP (11), and thus only traces of pAp are detectable (Fig. 2, WT in SD medium). We first asked whether the effect of Gcn4p on pAp accumulation was related to the fact that Gcn4p is an activator of HAL1 expres-sion (46). Hal1p plays an important role in maintaining Na ϩ /K ϩ ion homeostasis and was shown to confer salt tolerance when overexpressed in yeast cells, possibly by activating Ena1p, a sodium/lithium extrusion ATPase (Fig. 1) (47). It was therefore possible that the accumulation of pAp in gcn4⌬ mutants in the presence of NaCl was due to higher internal sodium concentrations in a gcn4⌬ genetic background, resulting in greater inhibition of Hal2p enzymatic activity. We thus performed HPLC analysis of nucleotide extracts from control and NaCl-treated wild type and hal1⌬ cells. We used 0.6 M NaCl for this experiment to allow significant accumulation of pAp in wild type cells, allowing a better comparison with hal1⌬ cells (Fig. 5A). Incubation of the hal1⌬ mutant in 0.6 M NaCl for 4 h did not result in a significant increase in the intracellular levels of pAp when compared with cells deficient for GCN4 (Fig. 5A, compare WT and hal1⌬ to gcn4⌬ in 0.6 M NaCl). Therefore, the absence of HAL1 expression and, a fortiori, the defect in its Gcn4p-dependent activation does not account for the accumulation of pAp in gcn4⌬ cells grown in the presence of sodium.
These results led us to consider the possibility that HAL2 expression itself might be limited in a gcn4⌬ mutant context. A dimer of Gcn4p binds to the consensus sequence 5Ј-TGACTC-3Ј located in the upstream control regions of many amino acid biosynthetic genes (48,49) and activates their transcription under conditions of amino acid starvation. This consensus sequence is also found within the HAL2 promoter (Fig. 5B, GCRE). We thus modified this site in the chromosomal HAL2 gene, creating a mutant called HAL2 GCRE⌬ . This mutation resulted in a strong accumulation of pAp in cells grown in the presence of sodium when compared with those grown without, similar to the accumulation of pAp in HAL2 cells with the gcn4⌬ FIGURE 3. Sodium treatment results in the accumulation of the Xrn1p-sensitive ITS1 RNA in gcn4⌬ cells. 20 g of total RNA were separated on agarose-formaldehyde gels, transferred to nylon membranes, and probed. The RNA component of the signal recognition particle, scR1, is used as a standard (59). All cultures were grown to mid-log phase and for an additional hour with or without salt as indicated. Wild type and gcn4⌬ strains were yRP840 and yRP840gcn4⌬, respectively, and xrn1⌬ and xrn1⌬gcn4⌬ strains were yRP884 and yRP884gcn4⌬, respectively. mutation in trans (Fig. 5A, compare HAL2 GCRE⌬ with or without NaCl and compare HAL2 GCRE⌬ with gcn4⌬ in 0.6 M NaCl). This suggested that the Gcn4p-dependent limitation of pAp accumulation in the presence of salt was due to a direct effect of Gcn4p on HAL2 expression.
Gcn4p Regulates HAL2 Expression Directly-To demonstrate a direct effect of Gcn4p on Hal2p expression, we compared Hal2p levels in wild type, gcn4⌬, or HAL2 GCRE⌬ cells by immunoblot analysis in the presence and absence of salt. Cells deficient for Gcn4p or lacking the Gcn4p-binding site upstream of HAL2 showed decreased levels of Hal2p both in the presence and in the absence of 0.6 M NaCl (Fig. 6, A and B, compare WT with gcn4⌬ or HAL2 GCRE⌬ ). An unidentified faster-migrating species that cross-reacts with the Hal2p antibody, and whose expression level varies under the conditions studied, was observed on the immunoblot. This protein is not a degradation product of Hal2p since it is also detectable in the hal2⌬, strain and its variation is not due to loading variations, as shown by the Coomassie Blue-stained gel underneath each blot. Thus, Gcn4p contributes strongly to basal levels of Hal2p expression. The inhibition of the low levels of Hal2p in the gcn4⌬ mutant by salt is probably sufficient to account for accumulation of pAp under these conditions (Fig. 5A). Like the gcn4⌬ mutation (Fig. 3), the GCRE⌬ mutation upstream of HAL2 resulted in greater stabilization of the Xrn1psensitive ITS1 RNAs in these growth conditions (Fig. 6C).

Differential Activation of GCN4 Expression by Sodium and Lithium
Has Consequences for pAp Accumulation-HAL2 nucleotidase is a well characterized physiological target of both sodium and lithium (13)(14)(15). Both of these toxic ions can displace loosely bound, but essential, magnesium ions from proteins to form inactive complexes (50). Lithium shows greater levels of toxicity than sodium (13,19), and it has been shown that the half-maximal inhibition of Hal2p activity in vitro is obtained at 0.1 and 20 mM with lithium and sodium, respectively (15). The work presented here has shown that yeast cells sense the presence of these toxic ions differently. We have shown that lithium is significantly less efficient than sodium in causing derepression of GCN4 translation and thus an induction of the general amino acid control response (Table 1). This has major consequences for cellular 5Ј-3Ј-mRNA degradation since induction of GCN4 expression under these conditions has a significant effect on the levels of pAp, a specific 5Ј-3Ј-exonuclease inhibitor (10). Cells deficient in the expression of GCN4 can accumulate high levels of pAp in the presence of sodium, and conversely, increased GCN4 expression, either from a plasmid constitutively overproducing GCN4 or by using the drug 3-AT, abolishes the elevated levels of pAp normally observed in the presence of lithium (Fig. 2). The fact that up to FIGURE 5. HAL2, rather than HAL1 expression, limits pAp accumulation during NaCl treatment. A, HPLC profiles of nucleotides from cells grown to exponential phase in minimal medium completed with amino acids (see "Experimental Procedures") and cells stressed with 0.6 M NaCl for an additional 4 h. Nucleotides were extracted and separated by reversed phase HPLC as described under "Experimental Procedures." Wild type strains and isogenic mutants are described in supplemental Table 1, and strain constructions are described under "Experimental Procedures." B, nucleotide sequence upstream of the translational start (ATG) of HAL2. The putative TATA box (TATAAA) is underlined, and the sequence for specific methionine regulation (CACGTG) and the consensus sequence for Gcn4p-binding site (TGACTC, indicated as GCRE) are shown in boxes. containing the Gcn4p-binding site (GCRE). Wild type RS16 and isogenic HAL2 GCRE⌬ strains were grown to mid-log phase in SD medium with or without 0.6 M of NaCl during 4 h. C, Xrn1p-sensitive ITS1 RNA accumulates in the HAL2 GCRE⌬ mutant in the presence of NaCl. 20 g of total RNA were separated on agarose-formaldehyde gels, transferred to nylon membranes, and probed. The RNA component of the signal recognition particle, scR1, is used as a standard (59). Cultures were grown to mid-log phase and for an additional 4 h with or without NaCl as indicated.

Gcn4p Controls Xrn1p Activity under Stress
10-fold more extracellular sodium than lithium was required for equivalent levels of pAp accumulation in wild type cells was previously attributed solely to the greater inhibition of Hal2p activity by lithium (13). We have shown that the differential activation of GCN4 expression by sodium and lithium also contributes to these differences in pAp accumulation.
A Sodium-specific Process for GCN4 Activation-In the yeast S. cerevisiae, starvation for amino acids or the presence of sodium induce phosphorylation of eIf2␣ by Gcn2 protein kinase, which leads to elevated translation of GCN4 (40,41,51). Stress signals activating Gcn2p in response to NaCl exposure are not well understood, however. In the case of amino acid starvation, the elevated levels of uncharged tRNA associate with the histidyl-tRNA synthetase-related domain of Gcn2p, leading to an enhancement of kinase-substrate interaction and phosphorylation of eIF2␣ (52). Elevated levels of NaCl have been shown to reduce uptake of many different amino acids (46), and this is one explanation for the activation of Gcn2p. Consistent with this idea, lithium treatment does not significantly affect the transport rate of amino acids (41) and does not strongly enhance GCN4 derepression (Table 1). However, GCN4 expression can be induced in prototrophic strains in response to 1 M NaCl (41), arguing against a model in which limited amino acid uptake triggers activation of Gcn2p. Moreover, it has also been reported that high concentrations of sodium or potassium reduce amino acid uptake equally, without a noticeable induction of GCN4 expression in the case of potassium (40) ( Table 1). It has been proposed that dimerization of Gcn2p contributes to its own activation, and thus increased expression of GCN4, during NaCl-induced stress (41). All these results taken together support the idea that some sodium-specific cellular processes, not directly related to amino acid transport, potassium, or lithium, are involved in the effect on GCN4 expression observed here.
Gcn4p Does Not Regulate the pAp Pool through Hal1p under Conditions of NaCl Stress-Gcn4p has been referred to as the "master regulator" of many genes involved in remedying stress perturbations (26), and thus one possible reason for the elevated accumulation of pAp in gcn4 mutants during salt stress was that high intracellular concentrations of toxic ions persist and inhibit Hal2p enzyme activity. At first glance, HAL1 looked like a good candidate for a Gcn4p-dependent control of ion homeostasis for multiple reasons. Hal1p functions in the maintenance of cellular Na ϩ /K ϩ ion balance and confers salt tolerance when overexpressed in yeast cells, and loss of Gcn4p activity prevents HAL1 expression in response to high salt concentrations (46). It was therefore possible that a modest increase in GCN4 expression enhanced sodium efflux, through Hal1p-mediated activation of the sodium-pumping Ena1p ATPase (19,53). The fact that pAp accumulation is similar in hal1⌬ and wild type cells in the presence of salt (Fig. 5A) excludes the possibility that the lack of Gcn4p-dependent activation of HAL1 was responsible for pAp accumulation through inhibition of Hal2p activity. These results supports another hypothesis proposed by Pascual-Ahuir et al. (46) in which the loss of GCN4 expression does not affect factors implicated in ion homeostasis. This hypothesis was based on the observations that the intracellular accumulation of sodium and potassium was not significantly altered by loss of GCN2 function after NaCl shock and therefore should not be altered by loss of GCN4 expression.
HAL2/MET22 expression depends strongly on GCN4 expression, in contrast to other MET genes in S. cerevisiae. The accumulation of pAp during salt stress can be explained by a defect in HAL2 expression in the gcn4⌬ mutant, with the remaining Hal2p being a target for toxic lithium and sodium ions. We initially considered this unlikely because HAL2/ MET22 expression was expected to be optimal and independent of Gcn4p, as has been shown for other MET genes in these growth conditions ( Fig. 1) (11). In our experimental conditions, cells were grown in the absence of extracellular methionine and cysteine, leading to maximal activation of the sulfate assimilation pathway. Under these conditions, the enzymatic activity of the products of the MET3, MET14, and MET16 genes and other methionine biosynthetic MET genes is high because of their transcriptional activation by Met4p in response to methionine limitation (11,54). Although MET16 transcription is enhanced by MET4 when the level of extracellular methionine is low, a second mechanism, involving GCN4, can also trigger MET16 expression (55). However, GCN4 was shown not to be essential for the methionine-specific (MET4-dependent) regulation of MET16 since MET16 gene induction still occurs in a gcn4 mutant upon methionine withdrawal. Apparently, Gcn4p-dependent expression of HAL2 is different because we showed that the lack of GCN4 expression, or modification of a cis-element responding to Gcn4p in the UAS of HAL2, has a severe negative effect on Hal2p accumulation in the absence of methionine. This deficiency in HAL2 expression only becomes critical in the presence of sodium (Fig. 6A) since, in the absence of salt, even the low levels of Hal2p are sufficient to metabolize pAp (Fig. 5). Interestingly, this Gcn4p-dependent regulation of HAL2 expression is likely to be conserved in different yeast species (see the Supplemental Data).
Gcn4p-dependent Regulation of HAL2 Expression Governs RNA Stability during NaCl Stress-We propose the existence of a regulatory loop involving a Gcn4p-dependent activation of HAL2 expression, triggered in the presence of sodium, and dedicated to limiting the inhibition of Hal2p activity (Fig. 7). In the present study, we have shown that the activation of HAL2 expression by the stress response factor Gcn4p is correlated with the maintenance of Xrn1p activity, an effect that could clearly be seen by the increased stability of the Xrn1p-sensitive ITS1 RNA. Gcn4p-dependent maintenance of the 5Ј-3Ј-mRNA degradation pathway in the presence of sodium appears particularly critical for cell growth when the 3Ј-5Ј-mRNA degradation pathway is abolished (Fig.  4) since the activity of at least one of these pathways is required for cell survival (44). High levels of pAp can also be detrimental to cell growth because they inhibit sulfate assimilation (11). A role for Gcn4p in maintaining sulfate assimilation would appear to be secondary, however, since we did not observe an effect on growth of the gcn4⌬ mutant in the absence of methionine and in the presence of salt (Fig. 4). Since inhibition of the Xrn1p activity occurs in the gcn4⌬ mutant under these con- ditions, we propose that the primary role of Gcn4p-mediated regulation of hal2p is to limit the inhibition of the 5Ј-3Ј-mRNA degradation pathway.
Gcn4p-dependent Maintenance of 5Ј-3Ј-RNA Degradation Can Be Viewed as a Facilitator of the Stress Response-S. cerevisiae cells respond rapidly to conditions of salt stress by activating many sets of genes (56). Microarray analysis of yeast cells grown at high salinity showed an early up-regulation of transcripts related to nucleotide and amino acid metabolism in particular, as well as transcripts related to cellular energy production (56). Therefore, the maintenance of Xrn1p activity, which produces 5Ј-mononucleotides from RNA, can be viewed as a means of maintaining cellular nucleotide pools, in addition to a means saving cells from a highly energy-consuming process such as the translation of decapped mRNAs (7,8). In this way, cells are assured of Xrn1p activity during sodium stress via the Gcn4p-dependent regulatory mechanism.
We believe that maintenance of Xrn1p activity helps to accurately mount this stress response since deletion of XRN1 is known to produce significant changes in protein accumulation and expression patterns. Proteome analysis has revealed that many of the proteins up-regulated in a Xrn1p-deficient strain are involved in amino acid biosynthesis and amine and nitrogen metabolism and produce other perturbations in sucrose, purine, and pyrimidine metabolism (9). Proteome analysis also identified proteins that are significantly down-regulated in the xrn1⌬ strain such as different structural components of the ribosome and factors involved in protein synthesis. Thus, inhibition of Xrn1p by pAp under conditions of salt stress would be likely to produce an inaccurate stress response or, at the very least, modify its timing.