Proteomics analysis of cellular response to oxidative stress. Evidence for in vivo overoxidation of peroxiredoxins at their active site.

The proteomics analysis reported here shows that a major cellular response to oxidative stress is the modification of several peroxiredoxins. An acidic form of the peroxiredoxins appeared to be systematically increased under oxidative stress conditions. Peroxiredoxins are enzymes catalyzing the destruction of peroxides. In doing so, a reactive cysteine in the peroxiredoxin active site is weakly oxidized (disulfide or sulfenic acid) by the destroyed peroxides. Cellular thiols (e.g. thioredoxin) are used to regenerate the peroxiredoxins to their active state. Tandem mass spectrometry was carried out to characterize the modified form of the protein produced in vivo by oxidative stress. The cysteine present in the active site was shown to be oxidized into cysteic acid, leading to an inactivated form of peroxiredoxin. This strongly suggested that peroxiredoxins behave as a dam upon oxidative stress, being both important peroxide-destroying enzymes and peroxide targets. Results obtained in a primary culture of Leydig cells challenged with tumor necrosis factor alpha suggested that this oxidized/native balance of peroxiredoxin 2 may play an active role in resistance or susceptibility to tumor necrosis factor alpha-induced apoptosis.

Organisms living under aerobic conditions need to protect themselves against the damage caused by reactive oxygen species (O 2 . , H 2 O 2 , and OH ⅐ ), arising from either the incomplete reduction of oxygen during cellular respiration or exposure to external agents such as light, ionizing radiation, or some redox drugs (1,2). These reactive oxygen species can damage various components of living cells such as unsaturated lipids (giving rise to deleterious organic peroxides), proteins, or nucleic acids.
To counter these deleterious processes, cells use several protective systems that either repair the various types of damage (e.g. DNA repair enzymes) or destroy the reactive oxygen species. One of the latter systems depends upon superoxide dismutases, which destroy O 2 . , but in turn produce hydrogen per-oxide. Hydrogen peroxide can be destroyed by catalase, but this requires the hydrogen peroxide to reach the peroxisomes, where catalase is present. For destruction of hydrogen peroxide and organic peroxides without transport to the peroxisomes, various peroxidases exist that are present in many cellular compartments. Whereas catalase destroys H 2 O 2 to produce water and molecular oxygen, peroxidases destroy peroxides by their reduction to the corresponding alcohol (or water) with the simultaneous oxidation of a specific cosubstrate. A typical example is glutathione peroxidase, which in destroying peroxides oxidizes glutathione from its thiol form to its disulfide form. The oxidized glutathione is then reduced back to its thiol form by glutathione reductase, using NADPH as the reducing agent. The final overall reaction is thus the destruction of peroxides to the corresponding alcohols (and/or water) and consumption of NADPH. Among the cellular enzymes using a peroxidase-like mechanism, peroxiredoxins represent a special case. These proteins constitute both the peroxidase and the cosubstrate because the enzyme itself is oxidized upon reaction with the peroxide. Whereas many peroxidases use either heme or selenocysteine in their active site, peroxiredoxins have a cysteine at their active site. The presence of additional conserved cysteines in the sequence is variable and provides the basis for the classification of the peroxiredoxins into two peroxiredoxin subfamilies that are differentiated by the presence of one or two conserved cysteine residues in their sequence (1-Cys and 2-Cys forms) (3). The active site cysteine can be oxidized by the peroxide to either one of two forms: cysteine sulfenic acid in 1-Cys peroxiredoxins (4) or disulfide in the 2-Cys peroxiredoxins (5). To complete the enzymatic catalytic cycle, the peroxiredoxins are then reduced back to their active thiol form, for example by the thioredoxin-thioredoxin reductase system for 2-Cys peroxiredoxins (5,6).
Although they were described rather recently, the list of identified peroxiredoxins is growing rapidly, and their ubiquitous nature is apparent. In addition to the classical cytosolic 2-Cys peroxiredoxins, named Prx1 1 and Prx2 (however, a variety of other names are also encountered), a third isoform (Prx3, also named AOP or SP22) is present in mitochondria (5). Other peroxiredoxins have been described more recently, and microsomal-secreted (7), peroxisomal (8), and chloroplastic isoforms (9) are now known, in addition to 1-Cys peroxiredoxin (10).
This ubiquitous distribution suggests that these enzymes * This work was supported in part by the National Science Foundation Science and Technology Center for Molecular Biotechnology. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ To whom correspondence should be addressed. play an important role in the antioxidant defense of the cell. This hypothesis has received support by inactivation of the PRX1 gene in yeast (11). Furthermore, a mutation in a murine peroxiredoxin correlates with predisposition to atherosclerosis (12). However, little is known at the protein level in mammalian cells about the response to challenges by oxidative stress because most studies are carried out at the RNA level (e.g. Ref. 13). Recently, a proteomic approach has been used quite successfully in yeast cells (14). It showed a major reorientation in metabolism and an increase in the synthesis of antioxidant proteins. We therefore decided to investigate the response to oxidative stress in mammalian cells using a proteomics approach, which also detects putative posttranslational responses. An example of such a study can be found recently (15). In this study, some changes in peroxiredoxins were shown to occur upon oxidative stress, but these changes were not characterized.

MATERIALS AND METHODS
Cell Culture and Oxidative Stress-Jurkat T-cell lymphoma cells were cultured in suspension in RPMI 1640 medium containing 1 mM pyruvate, 10 M mercaptoethanol, 10 mM Hepes-NaOH, pH 7.5, and 10% fetal calf serum. Cell viability was assessed by trypan blue exclusion.
Various oxidative stresses were applied before harvesting and cell lysis: (i) the cells (either attached or in suspension) were cultured for 0.5-6 h with 75-150 M tert-butyl hydroperoxide (BHP) or (ii) the cells were treated with 14 units/liter glucose oxidase for 18 h in fresh Dulbecco's modified Eagle's medium with the supplements described above (16). Genotoxic stress was carried out by culturing the cells in the presence of 1 M daunomycin for 18 h.
For recovery studies, the cells were stressed for 0.5 h with 75 M tert-butyl hydroperoxide. The cells were then washed twice in complete medium without BHP and re-cultured for the desired period of time in BHP-free medium.
Sample Preparation-Cells were harvested by centrifugation, rinsed in phosphate-buffered saline, and resuspended in homogenization buffer (0.25 M sucrose, 10 mM Tris-HCl, pH 7.5, and 1 mM EDTA). A buffer volume approximately equal to the packed cell volume was used. The suspension was transferred to a polyallomer ultracentrifuge tube, and the cells were lysed by the addition of 4 volumes (relative to the suspension volume) of 8.75 M urea, 2.5 M thiourea, 25 mM spermine base, and 50 mM dithiothreitol. After 1 h at room temperature, the nucleic acids were removed by ultracentrifugation (30 min at 200,000 ϫ g). The protein concentration in the supernatant fraction was determined by a Bradford assay, using bovine serum albumin as a standard. Carrier ampholytes (0.4% final concentration) were added, and the protein extracts were stored at Ϫ20°C.
Gel Electrophoresis and Analysis-Proteins were separated by twodimensional electrophoresis, using home-made immobilized pH 4 -8 gradients with pH plateaus at both ends (18). The sample was loaded on the immobilized pH gradient strip by in-gel sample rehydration (18), using a urea-thiourea mixture as solubilizing agent (19). 120 g of total extract were loaded on analytical gels (4-mm-wide immobilized pH gradient strips), and up to 2 mg of total extract were loaded on 6-mmwide strips for micropreparative purposes. The isoelectric focusing was performed over 24 h for a total of 55,000 V-h. After equilibration of the isoelectric focusing strips (20), SDS electrophoresis was performed on 10% gels. After two-dimensional gel electrophoresis, proteins were stained with silver (21) or with zinc imidazole for subsequent mass spectrometry (22). Silver-stained gels were analyzed with MELANIE software (Genebio). The analysis consisted of spot detection and quantification after noise and background removal. For each gel, the spots abundances were expressed in parts/million of the sum of the volumes of all the spots detected on the gel. This compensated for the variations in protein loading from one gel to another.
Statistical analysis was carried out using a heteroscledastic t test. Mass Spectrometry Analysis-Excised, chopped, and dehydrated protein spots from gels were rehydrated on ice for 45 min in 50 mM ammonium bicarbonate containing 12.5 g/l sequencing grade pig trypsin (Promega, Madison, WI). Digestion was carried out at 37°C for 15 h. The resulting peptides were extracted by sequential extraction for 20 min each in 50-l aliquots of 20 mM ammonium bicarbonate followed by 1% trifluoroacetic acid, 0.1% trifluoroacetic acid in 50% acetonitrile, and, finally, 5% acetic acid in 50% acetonitrile. Combined extracts were concentrated in a Speed Vac to ϳ5 l, re-dissolved in 45 l of 0.1 M acetic acid, and centrifuged for 2 min at 14,000 ϫ g. The supernatant fractions were carefully transferred into a fresh tube and concentrated to ϳ5 l. Samples were loaded from a stainless steel chamber pressurized to 1000 p.s.i. onto fused silica capillary (75 m, inner diameter) slurry packed to 8 cm with Magic C18 beads (Michrom, Auburn, CA). The column was developed with a linear gradient of 5-50% acetonitrile in 0.4% acetic acid and 0.005% hepta-fluorobutyric acid over 25 min at 300 -400 nl/min. Electrospray ionization was conducted by applying 1.0 kV to a Valco stainless steel union. Capillary columns terminated inside the Valco union, and a fused silica capillary (5 cm ϫ 75 m ϫ 360 m) tapered to ϳ5-m inner diameter served as an emitter.
The peptides eluting from the column were analyzed directly on a Finnigan tsq7000 mass spectrometer equipped with an in-house built microspray device. Data-dependent MS/MS spectra were acquired automatically by an Instrument Control Language procedure. Acquired MS/MS spectra were searched with SEQUEST (23) against the OWL protein data base.
Identification of the Porcine Peroxiredoxin 2-Porcine peroxiredoxin 2 was identified in porcine Leydig cell maps by image matching with a porcine erythrocyte two-dimensional map. This assignment was confirmed by comigration of 100 g of total Leydig cell extract with 1 g of porcine peroxiredoxin purified from pig erythrocytes by standard methods (25). Further confirmation was obtained by MALDI-TOF mass fingerprinting (24) on the putative porcine Prx2 spots obtained from gels loaded with Leydig cell extracts.

Acidic Peroxiredoxin Spots Appear upon Oxidative
Stress-In a search for protein modification occurring after oxidative stress, we used a proteomic approach on Jurkat cells stressed mildly with glucose oxidase or strongly with butyl hydroperoxide. Typical results are shown in Fig. 1. In the complete two-dimensional gel shown in Fig. 1, one of the most striking phenomena was the marked induction of an acidic satellite spot to peroxiredoxins. This phenomenon was most prominent for peroxiredoxin 2, which is the major enzyme of this family in Jurkat cells, and occurred for various types of oxidative stress. This spot is called acidic because its pI (5.45) is significantly more acidic than the theoretical pI of peroxiredoxin 2 (5.7), whereas the pI of the major spot present in normal lymphocytes and in Jurkat cells under normal culture conditions (5.8) fits the theoretical pI within Ͻ0.1 pH unit. The identity of the proteins in the acidic and basic spots was ascertained by MS/MS and in both cases proved to be peroxiredoxin 2. As also shown in Fig. 1, the appearance of an acidic spot also occurred for the mitochondrial peroxiredoxin, peroxiredoxin 3. These acidic spots appeared upon BHP treatment, but cell viability was severely hampered under these conditions because cells died within 6 h of treatment. We therefore investigated whether the acidic spots were associated with oxidative stress or just with cell death. These acidic satellite spots did not appear when other, nonoxidative stresses were used, e.g. treatment with genotoxic agents (1 M daunomycin, which induced complete cell death in 48 h), as shown in Fig. 2. In addition, when a strong oxidative stress was applied (BHP), the normal spots disappeared at the profit of the acidic ones within 30 min, whereas cell viability was still Ͼ90%. Conversely, under moderate stress conditions (glucose oxidase), the normal spots were still present after 24 h of treatment (as shown in Fig. 2C), and cell viability was still around 85%.
The Acidic Peroxiredoxin 2 Spot Is Oxidized at the Active Cysteine Site-To characterize the modifications taking place in the acidic spots, both the normal and acidic spots from peroxiredoxin 2 were analyzed by LC/MS/MS. A modified peptide was found at the LC/MS stage as a peak occurring only in the acidic spot and not in the basic one (Fig. 3). The mass of this peak could correspond to the 30 -61 peptide (i.e. with two missed trypsin cleavage sites at lysines 34 and 36) plus three oxygen atoms. To confirm this hypothesis, this peak was analyzed by collision-induced dissociation (Fig. 4). The y ion series identified the peptide as the active site region. A mass difference of 151 absolute mass units was detected between y11 and y10, indicative of the presence of a cysteinyl residue modified by three oxygen atoms. This allowed unequivocal assignment of the oxidation of Cys-51 to cysteic acid. The precision in the mass determination allowed us to exclude intermediate oxidation states of cysteine (namely cysteine sulfenic and sulfinic acids) and any other modification on this peptide. Thus, the acidic spot corresponded to the in vivo oxidation of peroxiredoxin 2 at Cys-51, which is the active site of the enzyme. This oxidation brought an extra negative charge to the protein, resulting in the lower pI observed on the gels. It must be noted that this extra negative charge made the analysis of the peptide by mass spectrometry much more difficult, as shown by the 50 pmol of modified protein required for this determination. These elevated levels prevented us from carrying out the same experiments on peroxiredoxin 3, which gave significantly lower yields than peroxiredoxin 2 in the required micropreparative two-dimensional gels (2 mg of total extract loaded on the strip; data not shown).
However, due to the high sequence conservation between peroxiredoxin 2 and 3, we speculate that the acidic peroxiredoxin 3 spot also corresponds to an oxidized form at the active site. This has also been suggested in previous work (26).
Recovery after Transient Oxidative Stress-Chemically speaking, oxidation of cysteine to cysteic acid is likely to be irreversible under biological conditions. To investigate cell recovery after oxidative stress, we first stressed the cells with BHP for 30 min and then let the cells recover in a BHP-free medium for various periods of time. The cellular extracts were then analyzed by two-dimensional gel electrophoresis.
The results are shown in Fig. 5 and Table I. It must be noted that the amount of the modified spots remained unchanged for at least 3 h, whereas that of the normal spots increased back to the original levels. After 24 h of recovery, the cells showed normal levels of normal peroxiredoxin 2 but still showed elevated levels of the oxidized form. Although we cannot formally exclude that the retroreduction of the oxidized form of Prx2 may play a role during recovery of the normal levels of the normal form, the fact that there is a significant increase in the total Prx2 (i.e. normal ϩ oxidized), at least in the early phases of the recovery process (p Ͻ 0.01 at 1 and 6 h, p Ͻ 0.001 at 3 h), strongly suggested that the recovery of the normal spots occurred mainly through de novo synthesis. This is further evidenced by the persistence of high levels of oxidized Prx2 during the early phases of the recovery process (the variation in oxidized Prx2 is not statistically significant during the first 3 h of recovery). However, when we tried to block de novo synthesis with cycloheximide or emetine during the recovery period to confirm this hypothesis, massive cell death occurred and precluded any analysis by two-dimensional gel electrophoresis. This was not the case when these protein synthesis inhibitors were used on unstressed cells for the same period of time.
Interestingly enough, the recovery kinetics were quite different for peroxiredoxin 2 (cytosolic) and peroxiredoxin 3 (mitochondrial). Whereas peroxiredoxin 2 recovery was Ͼ60% complete in 3 h and Ͼ80% complete in 6 h (see Table I), peroxiredoxin 3 recovery was barely visible after 6 h and was not complete even after 24 h. In addition, the level of oxidized Prx3 decreased steadily during the recovery process. This decrease is significant as early as 1 h (p Ͻ 0.03 at 1 h, p Ͻ 0.01 at 3 h, and p Ͻ 0.001 at 6 h), whereas the recovery of the normal form is significant only at 24 h. This means that the level of the oxidized form decreased, whereas that of the normal form did not increase in parallel. These data argue strongly against regeneration of Prx3 by a retroreduction of the oxidized form. They are also in agreement with the previous observation that Prx3, probably in its oxidized form, is a substrate for the mitochondrial ATP-dependent protease (26).
Peroxiredoxin 2 in Normal Cells-Because the results described above were obtained in transformed cells undergoing an experimental oxidative stress in vitro, we decided to investigate whether the same phenomena could occur under more physiological conditions and in a system where cell viability issues would not bias the results. As a model, we chose porcine Leydig cells in primary culture, which have been shown to be completely resistant to TNF-induced cell death (27) and thus remain fully viable under these conditions. This provided us with a means to eliminate any interference that could result FIG. 4. Collision-induced dissociation spectrum of the modified peptide. The collision-induced dissociation spectrum of the m/z 1281 peak provided enough sequence data for unequivocal assignment to peptide 30 -61 (i.e. with two missed trypsin cleavage sites) in a triplecharged state and with a modification. A partial MS/MS spectrum of this peak is shown with the assignment of some ions, leading to sequence information. The number in parentheses below or above each amino acid is its position in the sequence of the protein. Some numbers have been omitted to limit the crowding of the figure. The roman series (b and y) corresponds to single-charged fragment ions, whereas the italic series corresponds to double-charged fragment ions. These assignments, together with the mass of the peptide, allowed unequivocal assignment of the modification as the oxidation of Cys-51 to cysteic acid. from cell death or mortal wounding without needing to strongly overexpress peroxiredoxins to restore viability (28,29) during the assays with oxidative stress-related challenges such as TNF-␣ (30,31). We used porcine peroxiredoxin 2 extracted from erythrocytes to carry out the assignment by comigration (Fig.  6A). This assignment was further confirmed by mass spectrometry (data not shown). The position of the oxidized form was further confirmed by treatment of the cells in culture with 0.15 mM BHP for 2 h (Fig. 5B). This allowed the identification of the normal and acidic peroxiredoxin 2 spots in control and TNF-␣treated cells (Fig. 6, C and D). Here again, an increase in the amount of the acidic, oxidized spot could be seen upon TNF-␣ treatment (the increase ranged from 680 to 1710 ppm). This showed that the TNF signal led to an increase of the modified, inactive form of peroxiredoxin 2. However, the level of the normal spot in TNF-␣-treated cells remained similar to that observed in the control cells, and cell death was not observed, as in the case of glucose oxidase-treated Jurkat cells. Here again, cell death was observed after BHP treatment, correlating with a massive decrease of the normal Prx2 spot ( Fig. 6, B versus C).

DISCUSSION
Whereas examination of the response to oxidative stress in yeast cells showed major changes, including several affecting core metabolism (14), we detected only very limited changes in the mammalian cell system, as have other authors (15). The most prominent change observed was a posttranslational modification of peroxiredoxins. Although they have only been described rather recently, the importance of peroxiredoxins for control of the oxidative status of cells is rapidly emerging. As an example, the peroxiredoxin-based system (peroxiredoxinthioredoxin-thioredoxin reductase) is the major mitochondrial antioxidant system (32,33) together with manganese superoxide dismutase. This enzyme has also been shown to be induced under mild oxidative stress conditions in bovine aortic cells (34). In addition, several transfection experiments have shown that the overexpression of various peroxiredoxins is able to counteract several proapoptotic signals (28,29), thereby also indicating the importance of the oxidative status of cells in the onset of apoptosis.
However, the precise response of the peroxiredoxin systems in mammalian cells under oxidative stress or in response to proapoptotic signals was not known. Using a proteomics approach, we detected an alteration of the peroxiredoxin pattern upon oxidative stress. Two-dimensional electrophoresis showed an increase in satellite, acidic spots of peroxiredoxins upon oxidative stress. Analysis of tryptic peptides generated from the basic and acidic peroxiredoxin protein spots by mass spectrometry and MS/MS showed that the pI shift was caused by oxidation of the active site cysteine into cysteic acid, thereby adding a negative charge to the protein. This charge shift was detected by a mobility shift of the protein to a more acidic pI in the two-dimensional gel. Because the cysteic acid corresponds to a strong overoxidation of the cysteine, this acidic form must be considered as an inactive form of the peroxiredoxin. Analysis of the recovery phase showed that the oxidized form persisted for several hours after the arrest of oxidative stress but seemed to be eventually degraded. This degradation has been described previously for peroxiredoxin 3 (26). This cysteic acid form has also been described previously, but after in vitro oxidation of the protein with massive amounts of hydrogen peroxide (35). Lower cysteine oxidation states have also been described for another peroxiredoxin (1-Cys peroxiredoxin), but here again, only in vitro (36). From our study, it appears that this form is encountered in vivo after even a moderate oxidative stress and is constitutively present in normal erythrocytes (37). It must be mentioned, however, that our analysis takes place under reducing conditions, so that lower oxidation states of peroxiredoxins (e.g. the disulfide bridge or sulfenic acid states) will not be analyzed by our method. Another interesting input of the two-dimensional gel analysis lies in its quantitative description of the deconvoluted normal and inactive peroxiredoxin forms. After SDS electrophoresis and blotting (34) or protein quantitation by antibodies, the peroxiredoxin signal represents the sum of the normal and altered spots. As such, it gives the impression that the peroxiredoxin amount is increased by a mild oxidative stress or that it remains almost constant during a short, intense oxidative stress. However, our data show that the situation is more complex. Under mild oxidative conditions, the amount of inactivation caused by peroxiredoxin oxidation can be compensated, most likely by de novo synthesis of the native, active enzyme. Thus, the cell is able to "fill the gap" and keep its antioxidant defense level constant. In contrast, under strong oxidative stress, the normal form of peroxiredoxins almost disappears due to rapid and uncompensated inactivation by oxidation. This effectively annihilates the peroxiredoxin-based antioxidant defense, and cell death occurs shortly thereafter. In fact, we have observed a very good correlation between the state of the peroxiredoxins and cell survival. These data, added to the previously described transfection data (28,29), strongly suggest that peroxiredoxins play a key role in the resistance to pro-oxidant signals. However, the data obtained by transfection actually describe the effect of a massive overexpression of peroxiredoxins in transformed cells. We therefore chose to investigate the peroxiredoxin system in normal, nontransformed cells and without forced overexpression of peroxiredoxins. We chose as an experimental model the TNF resistance of porcine Leydig cells in primary culture (27). This model has the important feature of being naturally totally resistant to TNF-␣, with absolutely no loss in cell viability after challenge with TNF-␣ (27). We obtained the same result with TNF that we had with mild oxidative stress (e.g. glucose oxidase). An increase in the oxidized peroxiredoxin spot was observed upon TNF treatment, but the level of the active form remained high, again probably by de novo synthesis. Interestingly enough, Leydig cells treated with TNF-␣ and cycloheximide died within 48 h, whereas cells survived when challenged with only one of the two drugs. Thus, de novo synthesis of Prx2 may explain, at least in part, the survival of the cells under TNF challenge.
In conclusion, a detailed examination of peroxiredoxins by a proteomics approach provided physiologically relevant information. The normal spot indicates the level of antioxidant defense by peroxiredoxins, whereas the oxidized spot level is more an indicator of the oxidative injury to the cells. This, coupled to the various subcellular localizations of peroxiredoxins, provides a means to investigate the intensity of oxidative stress in various cell compartments (e.g. glucose oxidase stress versus BHP stress). Thus, parallel, quantitative examination of both forms allows detailed study of the phenomena occurring during oxidative or other stress and subsequent cell recovery.