Global Translational Responses to Oxidative Stress Impact upon Multiple Levels of Protein Synthesis*

Global inhibition of protein synthesis is a common response to stress conditions. We have analyzed the regulation of protein synthesis in response to oxidative stress induced by exposure to H2O2 in the yeast Saccharomyces cerevisiae. Our data show that H2O2 causes an inhibition of translation initiation dependent on the Gcn2 protein kinase, which phosphorylates the α-subunit of eukaryotic initiation factor-2. Additionally, our data indicate that translation is regulated in a Gcn2-independent manner because protein synthesis was still inhibited in response to H2O2 in a gcn2 mutant. Polysome analysis indicated that H2O2 causes a slower rate of ribosomal runoff, consistent with an inhibitory effect on translation elongation or termination. Furthermore, analysis of ribosomal transit times indicated that oxidative stress increases the average mRNA transit time, confirming a post-initiation inhibition of translation. Using microarray analysis of polysome- and monosome-associated mRNA pools, we demonstrate that certain mRNAs, including mRNAs encoding stress protective molecules, increase in association with ribosomes following H2O2 stress. For some candidate mRNAs, we show that a low concentration of H2O2 results in increased protein production. In contrast, a high concentration of H2O2 promotes polyribosome association but does not necessarily lead to increased protein production. We suggest that these mRNAs may represent an mRNA store that could become rapidly activated following relief of the stress condition. In summary, oxidative stress elicits complex translational reprogramming that is fundamental for adaptation to the stress.

Cells must be able to maintain their intracellular homeostasis in the face of changing conditions. Typically, they respond by invoking complex regulatory mechanisms, including global inhibition of translation (1,2). This reduction in protein synthesis may prevent continued gene expression during potentially error-prone conditions as well as allow for the turnover of existing mRNAs and proteins, whilst gene expression is reprogrammed to deal with the stress.
Four mammalian protein kinases that inhibit translation initiation by phosphorylating eukaryotic initiation factor-2 (eIF2) 2 have been identified. GCN2 (the amino acid control kinase), PKR (the double-stranded protein kinase activated by RNA), HRI (the heme-regulated inhibitor), and PERK/PEK (the PKRlike endoplasmic reticulum eIF2␣ kinase) are regulated independently in response to various different cellular stresses (2,3). For example, PERK has been found in all multicellular eukaryotes and is a component of the unfolded protein response. Consistent with its central role in the endoplasmic reticulum (ER) stress response, cells lacking PERK fail to phosphorylate eIF2␣ and do not down-regulate protein synthesis during ER stress conditions (4,5). Attenuating protein synthesis may act to reduce the burden of newly synthesized ER client proteins on the ER folding machinery. Additionally, eIF2 phosphorylation induces translation of specific mRNAs, such as that encoding the metazoan activating transcription factor-4 (6,7). Activating transcription factor-4 mediates the integrated stress response, the targets of which include genes encoding proteins involved in amino acid metabolism and resistance to oxidative stress, ultimately protecting against the deleterious consequences of ER oxidation (8).
In yeast, Gcn2 is the sole eIF2 kinase and phosphorylates eIF2␣ in response to nutrient starvation and sodium or rapamycin exposure. eIF2 is a guanine nucleotide-binding factor and, in the GTP-bound form, interacts with the initiator methionyl-tRNA (Met-tRNA i Met ) to form a ternary complex (eIF2⅐GTP⅐Met-tRNA i Met ) that is competent for translation initiation. Following each round of initiation, eIF2 is released from the ribosome as a binary complex with GDP. GDP is replaced by GTP in a guanine nucleotide exchange reaction promoted by eIF2B. Phosphorylation of eIF2␣ by Gcn2 converts eIF2 from a substrate to an inhibitor of the guanine nucleotide exchange factor eIF2B (9). The resulting decrease in eIF2B activity leads to reduced ternary complex levels. Paradoxically, translation of the GCN4 mRNA is activated in response to low ternary complex levels in a mechanism involving short upstream open reading frames (10). Gcn4 is a transcription factor that activates gene expression of many targets, including amino acid biosynthetic genes (11). Thus, analogous to the mammalian integrated stress response, activation of Gcn4 serves to overcome the imposed starvation, which initially led to the translational control. More recently, we used microarray analysis combined with polysome analysis to demonstrate that lowering ternary complex levels results in widespread translational reprogramming, identifying a fundamental role for translational control in the adaptation to nutrient limitation (12).
All aerobic organisms are exposed to reactive oxygen species (ROS), such as H 2 O 2 , the superoxide anion, and the hydroxyl radical, during the course of normal aerobic metabolism or following exposure to radical-generating compounds. These ROS cause wide-ranging damage to macromolecules, eventually leading to cell death (13,14). To protect against oxidant damage, cells contain effective defense mechanisms, including antioxidant enzymes and free radical scavengers (15). Yeast cells can adapt to oxidative stress by altering global transcription, including genes encoding antioxidants and other metabolic enzymes (16,17). However, we have shown that oxidative stress caused by exposure to H 2 O 2 results in a rapid and reversible inhibition of protein synthesis (18). Thus, it is unclear how changes in the gene expression program are translated into the cellular proteome. In this study, we have analyzed the regulation of protein synthesis in response to oxidative stress induced by exposure to H 2 O 2 . Our data show that H 2 O 2 causes a dosedependent inhibition of protein synthesis mediated in part by Gcn2-dependent phosphorylation of eIF2␣. In addition, we provide evidence that protein synthesis is repressed by a Gcn2independent inhibition of ribosomal transit. We used microarray analysis to demonstrate that certain mRNAs are translationally maintained following oxidative stress, indicating that translational control is a key component of the cellular response to oxidative stress.
Assays of GCN4-lacZ Reporter Expression-Standard methods for measuring the ␤-galactosidase activity for strains bearing GCN4-lacZ fusions were used (21). ␤-Galactosidase levels are expressed as nanomoles of o-nitrophenol ␤-Dgalactopyranoside hydrolyzed per min/g of total protein.
Western Blot Analysis-Protein extracts were electrophoresed under reducing conditions on SDS-polyacrylamide minigels and electroblotted onto polyvinylidene difluoride membrane (Amersham Biosciences). Blots were probed using anti-eIF2␣ and anti-phospho-eIF2␣ antibodies as described previously (22). TAP-tagged proteins were detected by West-ern blot analysis using peroxidase⅐anti-peroxidase complex (Sigma).
Analysis of Protein Synthesis-The rate of protein synthesis was measured in exponential phase cells treated with H 2 O 2 for 15 min and pulse-labeled for the last 5 min with 85 M L-[ 35 S]cysteine/methionine (18). TAP-tagged proteins were immunoprecipitated using IgG-Sepharose beads as described previously (23).
For the analysis of ribosome distribution on sucrose density gradients, yeast cultures were grown to exponential phase and treated with H 2 O 2 as appropriate. Extracts were prepared and analyzed as described previously (24). Monosome and polysome peaks were quantified using NIH Image J software.
For transit time measurements, cells were pulsed-labeled with 85 M L-[ 35 S]cysteine/methionine, and aliquots were taken at regular intervals. Extracts were prepared in 200 g/ml cycloheximide, and radioactive incorporation was measured in the total (completed and unfinished proteins) and completed protein fractions by liquid scintillation counting (25). Transit times were determined by comparing the incorporation of radioactive amino acids into total proteins and completed proteins released from ribosomes.
Polysome Analysis and RNA Preparation-Cell extracts were prepared as described previously (12). Briefly, 60 A 260 units were layered onto 35-ml 15-50% sucrose gradients and sedimented via centrifugation at 16,900 rpm for 13 h in a Beckman ultracentrifuge. Gradients were collected, and RNA quality was assessed using a 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). For comparison of RNA levels, fractions 4 -8 and 11-15 from the gradient were pooled together for monosomal and polysomal RNA samples, respectively (see Fig. 4B). Thus, we focused on only the most strongly and poorly translated mRNAs, omitting some mRNAs that are located in the dimer region of the gradient. Total RNA samples were prepared according standard protocols (www.cogeme.man.ac.uk/Facilities/TRF%20Protocols.htm). RNA analysis by real-time reverse transcription (RT)-PCR was carried out using the MyiQ singlecolor real-time PCR detection system and iQ SYBR Green Supermix (Bio-Rad).
Affymetrix GeneChip Expression Microarray Analysis-Microarray experiments were performed using the GeneChip yeast genome S98 oligonucleotide array (Affymetrix) according to the manufacturer's instructions (www.affymetrix.com/support/technical/manuals.affx). Approximately 10 g of polysomal, monosomal, or total RNA was processed into biotinylated cRNA according to the Affymetrix protocols. 15 g of biotinylated cRNA targets was fragmented and hybridized to the arrays at 45°C for 16 h. The arrays were then processed using the Affymetrix EukGE-WS2 fluidics protocol (Version 4_450) and stained with R-phycoerythrin conjugated to streptavidin (Molecular Probes). Microarray images were acquired using GeneChip Scanner 2500 (Affymetrix) and Microarray Suite Version 5.0 software. For consistency with previous studies (12), we used robust multi-array analysis normalization to normalize the array data as well as further analysis using the Affymetrix library of procedures (Affy Version 1.5.8) in Bioconductor (Version 1.5, www.bioconductor.org) within R (Version 2.0.1, www.r-project.com). Robust multi-array analysis nor-malization was observed to produce more robust treatment of replicates as seen from standard MA plots used to compare the biological replicate samples (data not shown). Pooled polysomal and monosomal fractions generated very modest dispersal, demonstrating that the procedure we have developed is highly reproducible and fulfills all of the standard criteria generally applicable to microarray data (26). The data sets are publicly available at ArrayExpress (accession number E-MEXP-526).  (Fig. 1A).

Hydrogen Peroxide Inhibits
The inhibition of protein synthesis caused by H 2 O 2 prompted us to analyze translational activity by examining the distribution of polysomes. Polysomes are ribosomes that are actively translating mRNAs. They can be separated on sucrose density gradients and quantified by measuring absorbance at 254 nm. Extracts prepared from the untreated strain exhibited normal profiles, including peaks corresponding to 40 S and 60 S ribosomal subunits, monosomes (80 S ribosomes), and polysomes (Fig. 1B). There was a dramatic shift of ribosomes from the polysomal region into the monosome or 80 S peak following treatment with H 2 O 2 . The accumulation of ribosomes in the 80 S peak of a sucrose gradient is indicative of decreased translation initiation. Polysome profiles from cells treated with cycloheximide during the peroxide treatment were analyzed to ensure that this effect did not arise because of peroxide activating global mRNA degradation in yeast. Incubation with cycloheximide was found to preserve polysomes during peroxide treatment, ruling out any effect on mRNA degradation (Fig. 1C).
Oxidative Stress Induces Gcn2-dependent eIF2␣ Phosphorylation-One of the best characterized translational regulatory pathways in yeast involves phosphorylation at Ser 51 of the eIF2␣ subunit by the Gcn2 kinase. Gcn2 is activated in response to amino acid starvation and ultimately brings about an inhibition of translation initiation (10). To test whether the H 2 O 2 -dependent translation inhibition relies upon this pathway, we examined eIF2␣ phosphorylation by immunoblot analysis using anti-phosphorylated eIF2␣ antibody. Phosphorylation of eIF2␣ was observed in response to all concentrations and was maximal at 0.5 mM H 2 O 2 ( Fig. 2A), corresponding with maximal polysome decrease (Fig. 1B).
Deletion of GCN2 abrogates phosphorylation of eIF2␣ and results in translational resistance to amino acid starvation (10). Similarly, no inhibition of translation initiation was observed in a gcn2 mutant in response to oxidative stress because loss of GCN2 prevented the accumulation of monosomes in response to H 2 O 2 (Fig. 2B). However, polysomes were maintained in the gcn2 mutant following H 2 O 2 treatment compared with the wild-type strain, indicating an effect at the post-initiation phase of translation (see below). Inhibition of translation initiation was also abrogated in a strain containing a mutant form of eIF2␣ lacking Ser 51 that is the target of phosphorylation (data not shown). The Gcn1⅐Gcn20 complex binds to Gcn2 and is thought to mediate the activation of Gcn2 by uncharged tRNA. The inhibition of translation initiation promoted by H 2 O 2 stress was reduced by loss of GCN1 or GCN20, indicating that the activation of Gcn2 by H 2 O 2 is liable to operate via a mechanism similar to amino acid starvation (Fig. 2B).
Phosphorylation of eIF2␣ causes a global inhibition of protein synthesis as well as gene-specific translational activation of GCN4, which encodes a transcriptional activator protein (27). GCN4 expression has been widely measured using a GCN4-lacZ reporter containing the GCN4 promoter and 5Ј-untranslated region driving expression of the lacZ gene. This reporter provides a convenient means to assess Gcn2-dependent inhibition of eIF2B activity mediated by eIF2␣ phosphorylation. When expression from this GCN4-lacZ reporter was measured following H 2 O 2 treatment for 2 h, a modest 2-fold induction was seen in response to 0.2 mM H 2 O 2 (Fig. 2C). This induction was dependent on Gcn2 because there was no induction in a gcn2 mutant. Surprisingly, GCN4 expression was inhibited at higher concentrations of H 2 O 2 ( Fig. 2C) despite the finding that maximal phosphorylation of eIF2␣ was at 0.5 mM H 2 O 2 ( Fig.  2A). To investigate the reason for this inhibition, we measured the rate of protein synthesis in a gcn2 mutant that could not inhibit translation initiation in response to H 2 O 2 ( Fig. 2B) and found that protein synthesis was still inhibited in the gcn2 mutant, albeit to a lesser extent (Fig. 2D). Taken together, these data indicate that H 2 O 2 exposure causes a global inhibition of protein synthesis via Gcn2-dependent inhibition of translation initiation as well as via an additional mechanism that does not affect initiation.
Oxidative Stress Inhibits Ribosomal Transit-Polysome analysis showed that H 2 O 2 caused an inhibition of translation initiation (Fig. 1A). The extracts for these profiles were prepared in the presence of the translation elongation inhibitor cycloheximide to prevent continued elongation and ribosomal runoff. An inhibition of ribosomal transit (elongation plus termination of polypeptide chains) should be seen as a preservation of polysomes when extracts are prepared in the absence of cycloheximide. However, even a diminished rate of ribosomal runoff is extremely rapid and difficult to detect (28). We therefore took the approach of inhibiting translation initiation by shifting cells to medium lacking glucose following the treatment with H 2 O 2 . Glucose withdrawal from the growth medium results in a rapid (Ͻ1 min) inhibition of translation initiation (24), and we reasoned that this would allow us to better detect any effect on ribosomal transit without the added complication of de novo translation initiation. Cells were either left untreated or treated with 2 mM H 2 O 2 prior to the switch to medium lacking glucose. Glucose withdrawal resulted in a rapid inhibition of translation initiation as expected (Fig. 3A), and ribosomal runoff occurred A, shown are the results from Western blot analysis of eIF2␣ and phospho-eIF2␣ (eIF2␣-P). The wild-type strain was grown to exponential phase in minimal synthetic dextrose medium and treated with the indicated concentrations of H 2 O 2 for 15 min. B, polysome traces are shown for the wild-type, gcn2, gcn1, and gcn20 mutant strains following treatment with 0.5 mM H 2 O 2 for 15 min. C, GCN4-lacZ expression was measured in the wild-type (wt) and gcn2 mutant strains following treatment with the indicated concentrations (millimolar) of H 2 O 2 for 2 h. The values shown are the means of at least three independent determinations. D, wild-type and gcn2 mutant cells were grown to exponential phase, and protein synthesis was measured as described in the legend to Fig. 1A. in control experiments, resulting in a loss of ϳ60% of the polysomes within 2 min (Fig. 3B). In contrast, polysomes were maintained following the H 2 O 2 treatment, and ϳ80% of the polysomes were maintained for 2 min.
Ribsosomal transit times can be measured by comparing the incorporation of [ 35 S]cysteine/methionine into total protein and completed polypeptide chains released from ribosomes (25). Using this technique, we determined that the average mRNA transit time in a gcn2 mutant is ϳ86 s (Fig. 3C). H 2 O 2 treatment (0.5 mM) increased the transit time by ϳ50% to ϳ131 s (Fig. 3D). Because a gcn2 mutant could not inhibit translation initiation (Fig. 2), these data confirm that there is an inhibition of translation elongation or termination in response to H 2 O 2 .
Global Gene Expression Profiling Reveals That Certain mRNAs Are Translationally Maintained following Oxidative Stress-The use of expression profiling techniques such as microarray analyses can be used to analyze protein synthesis (29). We used this technology to identify mRNAs that are translationally regulated in response to oxidative stress conditions. Cell extracts were prepared from yeast cells treated with 0.2 or 2 mM H 2 O 2 for 15 min. These concentrations were chosen because they had a similar effect on translation initiation ( Fig.  1B) but inhibited protein synthesis to different extents (Fig. 1A). The 15-min treatment with 0.2 mM H 2 O 2 reduced cell viability by ϳ20% compared with 65% loss of viability upon treatment with 2.0 mM H 2 O 2 . The 2.0 mM treatment also significantly slowed the growth of cells compared with the 0.2 mM treatment (Fig. 4A). Polysomal gradients from the selected stress conditions were separated into fractions (as described below), and in addition, total RNA samples (total stressed or total control) were prepared from stressed and control yeast cells to quantify standard transcript level changes. The resulting RNA samples were processed into cRNA and hybridized to Affymetrix microarrays. The analysis was performed in duplicate, and the data were processed and compared using the bioinformatics analyses described under "Experimental Procedures." Expression profiling studies generally compare the ratio of mRNAs in polysomal (P) and monosomal (M) fractions between stressed (S) and control (C) yeast cells. For example, when translation is inhibited at the level of initiation, the association of mRNAs with the polysomal versus non-polysomal fraction can be used as a marker that represents the translational activity of specific mRNAs (29). We recently defined the ratio PS/MS:PC/MC as the "translational state" of an individual mRNA and used it to identify mRNAs that are translationally maintained in response to the eIF2B-targeting stresses, amino acid starvation and fusel alcohol addition (12). However, when translation is inhibited by attenuating elongating ribosomes, as is the case for H 2 O 2 , increased polysome association cannot be used as a marker of the translational state of mRNAs. It was therefore necessary to develop a novel strategy to identify candidate mRNAs that are translationally regulated following H 2 O 2 stress. Specifically, we compared the mRNAs in monosomal and polysomal fractions during stress conditions (PSϩMS) with the mRNAs in monosomal and polysomal fractions during control conditions (PCϩMC) (Fig. 4B). We reasoned that any mRNAs that can overcome the initiation block would have more ribosomes bound during stress conditions, resulting in an increase in PSϩMS compared with PCϩMC. These values were plotted on a scatter plot for stressed ( y axis) and unstressed (x axis) cells (Fig. 5, A and B), where the intersection generates the change in translational activity for the mRNAs (PSϩMS:PCϩMC). This analysis facilitates the identification of those mRNAs that can overcome the block in translation initiation because they contain more ribosomes (both monosomal and polysomal) following the stress treatment. Within this population of mRNAs, we reasoned that it is possible to identify those mRNAs that are also somewhat resistant to the ribosomal transit block because they will contain more polysomes bound during stress conditions compared with control conditions, resulting in an increase in the PS:PC ratio (Fig. 5, A and B, red). Thus, we were able to identify those mRNAs that can overcome the inhibition of translation initiation (using the PSϩMS: PCϩMC ratio) as well as the mRNAs within this population that can overcome the inhibition of ribosomal transit to accumulate an increase in polyribosomes (using the PS:PC ratio).
Interestingly, the two stress conditions gave very different translational profiles. Using a 2-fold cutoff value for the change in translational activity (PSϩMS:PCϩMC), 231 mRNAs (130 up and 101 down) and 230 mRNAs (98 up and 132 down) were significantly altered following treatment with 0.2 or 2.0 mM H 2 O 2 , respectively (Fig. 5, A and B). The majority of these mRNAs were also somewhat resistant to the ribosomal transit block because they contained more polysomes bound during stress conditions compared with control conditions (0.2 mM, 199 mRNAs; and 2.0 mM, 198 mRNAs) (Fig. 5, A and B). To assess whether a similar set of mRNAs are altered following each stress condition, we highlighted those mRNAs that changed following the 2.0 mM treatment on the 0.2 mM plot (Fig. 5C), and conversely, we highlighted the 0.2 mM regulated mRNAs on the 2.0 mM plot (Fig. 5D). These plots show that the mRNAs up-or down-regulated were different for each stress condition.
To assess whether the response to these stress conditions is coordinated in terms of transcript level and translational activity, we plotted the change in translational activity (PSϩMS: PCϩMC) against the change in transcript level for each stress condition (Fig. 6). mRNAs that were translationally up-regulated and also changed at the transcript level and mRNAs that were translationally down-regulated and also changed at the transcript level are depicted, along with other translationally regulated mRNAs (Fig. 6, red, blue, and yellow, respectively). This analysis did not reveal any strong correlation between changes in transcript level and the translational response to H 2 O 2 . This may be because of the relatively short exposure times used in these experiments. Of the genes that were translationally up-regulated, just 10% (13 of 130) and 11% (11 of 98) were also induced at the transcriptional level by the 0.2 and 2 mM H 2 O 2 treatments, respectively (supplemental Tables I and  II). A co-regulation of the transcript level and translation has been noted previously following heat shock, rapamycin treatment, and amino acid starvation of yeast and has been termed "potentiation" (12,30). In contrast, no potentiation was observed here or for butanol stress, suggesting that co-regulation of the transcript level and translational activity is a stressspecific phenomenon (12).
We confirmed the microarray results using real-time RT-PCR analysis for a range of mRNAs in terms of both the overall abundance of the mRNA and the change in abundance across polysomal gradients (Fig. 7A). The Affymetrix system used for the microarray analysis requires that equal cRNA concentrations are used for each RNA hybridization. This effectively normalizes any changes in monosomal or polysomal fractions between stressed and unstressed samples for most mRNAs even though global translation has been inhibited by the stress. This is best illustrated by ACT1 mRNAs, which were unaffected by the 0.2 or 2.0 mM treatment according to the microarray analysis, whereas the RT-PCR analysis showed that less ACT1 mRNA was associated with monosomes and polysomes following stress conditions consistent with the global inhibition of translation (Fig. 7A). Nevertheless, RT-PCR analysis could be used to confirm the general trends that were detected from the microarray data. For example, RT-PCR analysis confirmed that the HSP30 mRNA was increased in abundance in monosomal and polysomal fractions following both stress conditions, that OYE3 and SRX1 mRNAs were increased in monosomal and polysomal fractions following the 0.2 mM H 2 O 2 treatment, and that the YCF1 mRNA was increased in the polysomal fraction following the 0.2 mM H 2 O 2 treatment.
Our data show that mRNAs were differentially associated with polyribosomes in response to oxidative stress conditions. To begin to address whether this results in increased protein production, we analyzed protein synthesis and levels for a selected group of proteins that are available with a TAP tag. To analyze protein synthesis, cells were treated with 0.2 or 2.0 mM H 2 O 2 for 15 min, and proteins were pulse-labeled with [ 35 S]cysteine/methionine during the final 5 min of treatment. SDS-PAGE analysis confirmed a global inhibition of protein synthesis (Fig.  7B). Decreased protein production was observed following both peroxide treatments but was more pronounced for the 2 mM H 2 O 2 treatment, with relatively few proteins detected.
The microarray analysis indicated that the HSP30 mRNA increased in abundance in monosomal and polysomal fractions following both stress conditions. No Hsp30 was detected by Western blot analysis during normal growth conditions (Fig.  7C). In contrast, Hsp30 was dramatically increased following the 0.2 mM treatment. Similarly, microarray analysis indicated an increase in the translational activity of TRR1 and SOD2 in response to 0.2 mM H 2 O 2 . In agreement with this observation, immunoprecipitation of pulse-labeled proteins revealed increased synthesis of both proteins at 0.2 mM H 2 O 2 . In the case of Sod2, increased protein levels were also detected by Western blot analysis, whereas no changes in the high basal levels of Trr1 were observed over this relatively short treatment period (Fig.  7C). From the microarray analysis, HSP30 and TRR1 belong to Translationally regulated mRNAs that were not transcriptionally regulated are colored yellow. Translationally up-regulated mRNAs that were also regulated at the transcript level are colored red, whereas translationally down-regulated mRNAs that were also regulated at the transcript level are colored blue.
the class of mRNAs that were somewhat resistant to the block in ribosomal transit (PS:PC ratio Ͼ2), whereas the SOD2 mRNA did not show any increase in polyribosomes following the H 2 O 2 treatment. Thus, both classes of mRNA result in increased protein production. Microarray analysis indicated that the translational activity of HSP30, SGD1, and COX7 mRNAs was increased in response to 2.0 mM H 2 O 2 . However, no increase in protein production was observed for these proteins as analyzed by immunoprecipitation of pulse-labeled proteins (Fig. 7B). Taken together, these data indicate that low peroxide (0.2 mM) results in increased protein production, whereas high peroxide (2.0 mM) increases polyribosome association with certain mRNAs but does not necessarily result in increased protein production.
Functional Classification of mRNAs Regulated by H 2 O 2 -The data were analyzed in terms of the precise function of the regulated genes and the likely impact that changes in translational activity would have on cellular physiology. Gene products were grouped into functional categories according to the MIPS Functional Catalogue Database (mips.gsf.de/genre/proj/yeast/ index.jsp) and the Saccharomyces Genome Database GO Term Mapper (db.yeastgenome.org/cgi-bin/SGD/GO/goTermMap-per). Genes that were regulated at the translational level relative to the global trend are included in these categories (Fig. 8 and supplemental  Tables I-IV).
The two stress conditions showed distinct patterns of regulation. The low peroxide (0.2 mM) condition induced several antioxidants, as might be expected (supplemental Table I). These included a cytosolic catalase (CTT1) and an atypical 2-Cys peroxiredoxin (GPX2), which can reduce H 2 O 2 directly (31-33); thioredoxin reductase (TRR1), which provides the reducing power for peroxiredoxins; sulfiredoxin (SRX1), which can reduce cysteinesulfinic acid residues that are formed in peroxiredoxins following oxidative stress (34); and a glutathione transferase (GTT2) and two GS-X pumps (YCF1 and YBT1), which form part of the glutathione conjugation/removal system of cells that is active against a broad range of toxic substrates (35). The most prominent classes of genes up-regulated by 0.2 mM H 2 O 2 included genes encoding various cellular transporters. Strikingly, nine ABC (ATP-binding cassette) transporters from a predicted 31 proteins belonging to the ubiquitous ABC superfamily were identified (36,37). These included seven "full-size" ABC transporters (from a predicted 16), which can confer considerable multidrug resistance. A number of iron regulon genes were translationally upregulated, including those involved in cell-surface iron uptake (FET3, FTR1, FRE1, and FRE2), iron transport across the vacuole membrane (FET5 and FTH1), and a mitochondrial iron transporter (MRS4). Iron is an essential nutrient that is largely thought to be regulated at the level of transcription via the ironresponsive transcriptional activators Aft1 and Aft2 (38,39). Increased iron uptake during oxidative stress is somewhat surprising given that it can potentially lead to the generation of the hydroxyl radical via the Fenton reaction (40). However, these data may indicate a requirement for restoration of iron homeostasis following oxidative stress. Cellular iron is found largely complexed in cells, e.g. in iron-sulfur clusters. Oxidation of these clusters causes release of the iron, resulting in enzyme inactivation (41). A number of metabolic genes were up-or down-regulated following treatment with 0.2 mM H 2 O 2 (supplemental Tables I and II). These included genes affecting carbon, amino acid, nitrogen, lipid, and energy metabolism, indicating that significant metabolic reconfiguration is required following oxidative stress.
The high peroxide (2.0 mM) condition resulted in significant up-regulation of genes involved in ribosome biogenesis and rRNA processing (45 of 98 genes induced) (supplemental Table  III). This is in contrast to the large number of similar genes that are transcriptionally repressed as part of the environmental stress response (42). Transcriptional co-regulation of genes involved in ribosome biogenesis has been linked to cell proliferation, and ribosome synthesis may serve as a measure for cell cycle progression (43,44). However, very few of the genes in our data set showed changes in transcript levels following the short H 2 O 2 treatment time used in this study (5 of 45 genes) (supplemental Table III). Additionally, a set of 236 genes that are transcriptionally co-regulated by the Spf1 transcription factor has been defined and termed the Ribi regulon (44). There is only a partial overlap between the Ribi regulon and the mRNAs that were translationally induced by 2.0 mM H 2 O 2 (48 genes) (data not shown). Rather than indicating a link between ribosome production and cell proliferation, these data may indicate that there is a requirement to replace ribosomal proteins and rRNA that become damaged by oxidative stress. In contrast to the low peroxide treatment, there was no significant up-regulation of metabolic genes. However, many genes affecting several aspects of metabolism were down-regulated in response to 2.0 mM H 2 O 2 , in agreement with the idea that metabolic reconfiguration is required in response to oxidative stress (supplemental Table IV). Surprisingly, a significant number of the genes that were translationally down-regulated in response to 2.0 mM H 2 O 2 were increased at the transcriptional level (59 of 132 mRNAs) (Fig. 6 and supplemental Table IV). These data indicate that certain genes are increased at the transcriptional level in response to H 2 O 2 but remain poorly translated. These mRNAs could represent an mRNA store that would become rapidly activated following relief of the stress condition.

DISCUSSION
Global inhibition of protein synthesis is widely recognized as a response of biological systems to stress conditions. However, it is becoming increasingly recognized that not all translation is inhibited and that translational control of specific mRNAs is required for survival during growth under stress conditions (12,30). The complement of proteins expressed in individual cells (or the proteome) is fundamental to their nature and diversity because proteins catalyze most of the reactions in cells and also serve numerous structural and regulatory roles. Control of protein levels via translational regulation offers a significant advantage to the cell because of the immediacy of the regulatory effect. Regulation of protein levels is therefore one of the key aspects in determining the fate of cells during stress conditions. Eukaryotic translation initiation is a complex highly regulated process involving Ͼ30 polypeptide factors interacting with ribosomal subunits, Met-tRNA i Met , and mRNAs (45). The initiation phase of protein synthesis is rate-limiting and is a target of extensive regulation. In mammalian cells, eIF2␣ is phosphorylated by PERK in response to oxidative stress as part of the integrated stress response (8). Yeast does not contain PERK, and hence, H 2 O 2 stress has not previously been shown to inhibit protein synthesis.
Yeast Gcn2 phosphorylates eIF2␣ in response to nutrient starvation and sodium or rapamycin exposure. Depletion of amino acids leads to an accumulation of uncharged tRNA, which activates the Gcn2 protein kinase through its histidyl-tRNA synthase-related domain. The signals activating Gcn2 in response to rapamycin or NaCl are not well understood. Rapamycin appears to work by blocking Tor-mediated phosphorylation of Gcn2 at Ser 577 (46). However, activation of Gcn2 by rapamycin and NaCl still requires the histidyl-tRNA synthaserelated domain of Gcn2 as well as Gcn1 and Gcn20, which are thought to mediate the activation of Gcn2 by uncharged tRNA (47). Similarly, the inhibition of translation initiation in response to H 2 O 2 requires Gcn1 and Gcn20. Oxidative stress may conceivably cause an accumulation of uncharged tRNA through a variety of mechanisms. Free amino acids and amino acids in proteins are highly susceptible to oxidation by ROS (48). Oxidized amino acids can be detected in yeast cells, and for example, oxidized phenylalanine (m-and o-Tyr) is elevated following exposure to concentrations of H 2 O 2 (0.2-2 mM) (49) similar to those used in this study. Although the levels of oxidized amino acids detected are relatively low, representing about one modification/10 3 phenylalanine residues, the total oxidative load on the amino acid pool may trigger an amino acid starvation response. Alternatively, the proteins and nucleic acids that are required for tRNA aminoacylation may be susceptible to oxidation, resulting in an accumulation of uncharged tRNA and activation of Gcn2. For example, oxidative damage to RNAs, including tRNAs, has been implicated in the pathogenesis of Alzheimer disease (50, 51) and ROS can affect the stability and activity of aminoacyl-tRNA synthetases (52).
We have previously determined the nature and extent of mRNAs that are translationally regulated in response to two stresses that lower eIF2B activity (12). Amino acid starvation leads to an accumulation of non-aminoacylated tRNAs and subsequent activation of Gcn2, whereas exposure to the fusel alcohol butanol inhibits eIF2B activity. Surprisingly, even though the stresses impact upon the same translation initiation factor (eIF2B), they have quite different outcomes in terms of the specific mRNAs that are translationally controlled. This creates a highly specific stress response that facilitates adaptation to the particular stress condition (12). We have suggested that, although these stresses both act via eIF2B, there must be other stress-specific modulatory inputs on the translational pathway that ultimately change which mRNAs are translationally selected following stress. The impact of oxidative stress on ribosomal initiation and transit is therefore particularly interesting because it indicates that control of the translational machinery at a stage subsequent to initiation can be used to modulate a proteomic output.
In contrast to amino acid starvation, protein synthesis is still inhibited in response to H 2 O 2 in the absence of Gcn2. Decreased ribosomal runoff is observed following H 2 O 2 stress, consistent with an inhibition of translation elongation or termination. Similarly, the average mRNA transit time is increased by ϳ50% in a gcn2 mutant, confirming that H 2 O 2 causes a postinitiation inhibition of protein synthesis. Regulation of mRNA expression levels by modulating translation elongation or termination is relatively poorly understood. It is known that cells can alter the bulk rate of protein synthesis in response to different growth conditions or hormones by changing the overall rates of elongation and/or termination (reviewed in Ref. 53). Similarly, oxidative stress in mammalian cells elicits a marked increase in eIF2 phosphorylation and oxidative modification, which is thought to contribute to an inhibition of translation (54,55). Attenuating elongating ribosomes in response to stress conditions, as opposed to ribosomal initiation, offers the advantage that ribosomes remain bound to mRNAs and can rapidly resume protein synthesis once the stress is removed or detoxified. For an oxidative stress condition, it would also prevent continued protein synthesis during potentially error-prone conditions. There are very few characterized examples in which the expression of individual mRNAs is regulated via alteration of ribosomal transit times (reviewed in Ref. 56). One good example is provided by the tyrosine aminotransferase mRNA, in which the rate of ribosomal transit is increased 5-fold in response to dibutyryl cAMP (57). Our data indicate that the majority of mRNAs that are translationally up-regulated following H 2 O 2 stress contain more ribosomes in polysomes following the stress condition. An obvious focus for future experimentation will be an investigation of how these mRNAs overcome this translation block.
The environmental stress response cluster encompasses ϳ900 genes that are transcriptionally activated or repressed by a large number of stress conditions, including ROS such as H 2 O 2 (16). The genes that are transcriptionally induced as part of the environmental stress response encode products that are thought to protect against and/or detoxify the stress agent as well as repair the resulting cellular damage. However, our data indicate that relatively few (ϳ15%) of the mRNAs that are translationally up-regulated in response to H 2 O 2 show concomitant increases in transcript levels. Increased transcript levels in the absence of active translation may therefore provide a source of mRNAs that can become rapidly translated once the stress is removed. In this view, the mRNAs that are actively translated during exposure to H 2 O 2 may provide the crucial functions that are required to detoxify and remove H 2 O 2 and its metabolic products. It is therefore particularly interesting that the low peroxide treatment results in increased production of stress protective proteins, whereas the high peroxide treatment increases the number of ribosomes associated with certain mRNAs but does not result in increased protein production. The low peroxide treatment used in this study has previously been shown to promote an adaptive response whereby cells become resistant to a subsequent higher and more lethal treatment with H 2 O 2 (58). Our data indicate that this may be explained by increased production of stress protective proteins, including antioxidants, at the low H 2 O 2 (0.2 mM) treatment. In contrast, the high H 2 O 2 (2.0 mM) treatment increases ribosome association on key mRNAs, which can then become rapidly translated once the oxidative stress is removed. In summary, our data indicate that the response to oxidative stress is complicated, requiring both translational and transcriptional reprogramming.