Opposite Roles of Selenium-dependent Glutathione Peroxidase-1 in Superoxide Generator Diquat- and Peroxynitrite-induced Apoptosis and Signaling*

Oxidative injuries including apoptosis can be induced by reactive oxygen species (ROS) and reactive nitrogen species (RNS) in aerobic metabolism. We determined impacts of a selenium-dependent glutathione peroxidase-1 (GPX1) on apoptosis induced by diquat (DQ), a ROS (superoxide) generator, and peroxynitrite (PN), a potent RNS. Hepatocytes were isolated from GPX1 knockout (GPX1−/−) or wild-type (WT) mice, and treated with 0.5 mm DQ or 0.1–0.8 mm PN for up to 12 h. Loss of cell viability, high levels of apoptotic cells, and severe DNA fragmentation were produced by DQ in only GPX1−/− cells and by PN in only WT cells. These two groups of cells shared similar cytochromec release, caspase-3 activation, and p21 WAF1/CIP1 cleavage. Higher levels of protein nitration were induced by PN in WT than GPX1−/− cells. Much less and/or slower cellular GSH depletion was caused by DQ or PN in GPX1−/− than in WT cells, and corresponding GSSG accumulation occurred only in the latter. In conclusion, it is most striking that, although GPX1 protects against apoptosis induced by superoxide-generator DQ, the enzyme actually promotes apoptosis induced by PN in murine hepatocytes. Indeed, GSH is a physiological substrate for GPX1 in coping with ROS in these cells.

Reactive oxygen species (ROS) 1 and reactive nitrogen species (RNS) are constantly generated in aerobic metabolism and involved in pathogenesis of many diseases (1,2). Pro-oxidants such as diquat (DQ) also induce cellular production of ROS including superoxide anion (O 2 . ), hydrogen peroxide (H 2 O 2 ), and hydroxyl radical (OH ⅐ ) (3). Peroxynitrite (PN), a potent RNS, may be formed by O 2 . and nitric oxide (NO) at a diffusionlimited rate (4). As PN nitrates a variety of biomolecules (5), formation of nitrotyrosine in proteins is often used to assess its cellular activity (6). Selenium is an essential antioxidant nutrient (7) that has potential in preventing cancer (8), viral infection (9), and chronic disease (10). Among the 18 identified mammalian selenoproteins (11,12), glutathione peroxidase-1 (EC 1.11.19, GPX1) was the first discovered (13,14) and the most abundant (15). Using GPX1 knockout mice (GPX1Ϫ/Ϫ) (16), we have demonstrated that GPX1 is the metabolic mediator of body selenium to protect mice against pro-oxidantinduced death and oxidative injuries (17,18). In contrast to such strong evidence for the long-assumed role of GPX1 in coping with ROS in vivo (19), the impact of GPX1 on RNSmediated oxidative stress in various organisms is virtually unknown. Earlier, Sies et al. (20) showed that adding GPX1 in human fibroblast lysates was able to reduce PN to nitrite and thus attenuated the PN-mediated protein nitration in the presence of adequate glutathione (GSH). Because nitration of protein tyrosine residues may impair the tyrosine phosphorylationrelated signaling and function (21), their finding has physiological relevance. However, the metabolic role of GPX1 in intact cells in coping with PN might be different from that in cell lysates, because of a strong reactivity between PN and CO 2 to form more active intermediates such as ⅐ NO 2 or CO 3 . (22), a possible inactivation of GPX1 by PN in oxidative state (23), and modulations of cellular ROS on PN cytotoxicity (24). Because apoptosis is induced by moderate levels of ROS in many types of cells (25), and by PN in HL-60 (26), PC12 (27), and human endothelial cells (28), it can be used to assess oxidative injury. Two key events in the induced apoptosis include cytochrome c release from mitochondria and activation of caspase-3 (29). During the early stage of apoptosis, the activated caspase-3 cleaves p21 WAF1/CIP1 , a cyclin-dependent kinase inhibitor that protects cells from apoptosis (30), at a specific aspartate residue (Asp-112) and causes the loss of its localization and function in nuclei (31). c-Jun NH 2 -terminal protein kinase (JNK) and p38 kinase, two mitogen-activated protein kinases (MAPK), are also activated in apoptosis induced by diverse stimuli (32). It is unknown how GPX1 affects the DQ-and PN-induced apoptosis and related signaling.
Intracellular GSH may play three roles in metabolism: as an independent antioxidant, as a presumed physiological substrate of GPX1 to be oxidized to GSSG and regenerated by NADPH-dependent glutathione reductase (EC 1.6.4.2) reaction (33,34), and as a regulator of apoptosis (35). It is fascinating to find out how GPX1 knockout affects the responses of cellular GSH/GSSG to ROS and RNS. Therefore, our objective was to dissect the metabolic role of GPX1 in cell death, apoptotic signaling, protein nitration, and GSH/GSSG responses induced by the ROS generator DQ and RNS donor PN in primary hepatocytes isolated from the GPX1Ϫ/Ϫ and the WT mice. Most strikingly, we found that GPX1 knockout did not attenuate, but enhanced hepatocyte resistance to the PN-mediated apoptosis, which was completely opposite to its impact on the DQ-mediated apoptosis or our expectation.
Culture of Primary Hepatocytes and ROS/RNS Generation-Hepatocytes were prepared from 8-week old GPX1Ϫ/Ϫ and WT mice (16) by collagenase D perfusion (36), and plated in 6-or 12-well collagen-coated plates (at the density of 6 or 3 ϫ10 5 ). In all experiments, viability of the isolated cells, as determined by trypan blue exclusion, was Ͼ85%. Cells were grown at 37°C in 5% CO 2 in William's medium E supplemented with 5% fetal bovine serum, 100 g of gentamycin/ml, 5 g of insulin/ ml, 1 g of glucagon/ml, 0.5 g of hydrocortisone/ml, and 10 mM HEPES, pH 7.0. After 20 h of culture, cells were incubated with superoxide generator DQ (diquat dibromide monohydrate, Chem Service, West Chester, PA; 0.5 mM dissolved in saline) or PN (0.1-0.8 mM in 4.7% NaOH, Calbiochem, La Jolla, CA) for different lengths of time. Both DQ and PN were added as a bolus into the media and mixed thoroughly for 30 s.
Cell Viability, DNA Fragmentation, and Apoptosis-Cell viability was assessed by the reduction of 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide to formazan. Actual values were read at 570 nm in a Microplate Reader (Elx15, Bio-Tek, Winooski, VT) and expressed as percentage of the untreated controls. DNA fragmentation was detected by ethidium bromide staining after the cellular DNA was extracted with phenol/chloroform and separated in 1.8% agarose gel. Apoptosis was quantified by terminal deoxynucleotidyltransferase-mediated dUTP nick-end labeling (TUNEL) assay kit (Roche Molecular Biochemicals) according to the manufacturer's instruction. Positive stained nuclei were counted in ϳ80 cells from each of four random fields using a fluorescent microscope (Olympus, Seattle, WA).
Glutathione and GPX Activity-Total (GSH ϩ GSSG) and oxidized (GSSG) glutathione were measured as described by Anderson (38) and expressed as nanomoles/mg of protein. Total GPX activity was measured by the coupled assay of reduced NADPH oxidation using H 2 O 2 as substrate (39). The enzyme unit was defined as 1 nmol of GSH oxidized/min.
Statistics-Data were analyzed using the GLM procedure of SAS (release 6.11, SAS Institute, Cary, NC). The Bonferroni t test was used for mean comparisons.

GPX1 Knockout Renders Mouse Hepatocytes Susceptible to DQ-induced, but Resistant to PN-induced, Apoptotic Death-
Compared with the untreated controls, viability of GPX1Ϫ/Ϫ hepatocytes was decreased by 0.5 mM DQ from 82% at 3 h to 4.9% at 12 h, whereas that of WT remained Ͼ84% throughout (Fig. 1A). In contrast, over 85% of GPX1Ϫ/Ϫ cells were viable after being treated with 0.2-0.8 mM PN for 12 h, whereas viability of WT cells was reduced to 30 and 10% by 0.4 and 0.8 mM PN, respectively (Fig. 1B). Similarly, 0.4 mM PN caused only Ͻ16% reduction in viability of GPX1Ϫ/Ϫ cells at various time points, but it decreased viability of WT cells to 50% at 3 h and further to 12% at 12 h (Fig. 1C). The PN vehicle alone, 4.7% NaOH, did not affect viability of either type of cells (data not shown). DNA fragmentation was produced by 0.5 mM DQ in only GPX1Ϫ/Ϫ cells and by 0.4 mM PN in only WT cells at 9 h, and the DNA ladder became pronounced at 12 h in these two groups ( Fig. 2A). TUNEL assay showed 53.8 and 43.6% apoptotic cells at 9 h in the DQ-treated GPX1Ϫ/Ϫ and the PNtreated WT hepatocytes, respectively (Fig. 2B). However, there was no detectable DNA fragmentation and only Ͻ3.5% apoptotic cells in the untreated, the DQ-treated WT, or the PNtreated GPX1Ϫ/Ϫ hepatocytes.
Similar Apoptotic Signaling Occurs in the DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT Cells-Cytochrome c release (Fig. 3A) and cleavage of caspase-3 ( Fig. 3B) was initially detected in the DQ-treated GPX1Ϫ/Ϫ at 6 h and in the PNtreated WT hepatocytes at 3 h. At the following time points, cytochrome c was accumulated in the cytosolic fraction and cleavage of caspase-3 progressed further in these cells. There was no cytochrome c release or caspase-3 cleavage in the DQtreated WT or the PN-treated GPX1Ϫ/Ϫ hepatocytes at any time point. After an initial increase at 3 h over the base line, p21 WAF1/CIP1 protein in the nucleic fraction of the DQ-treated WT and the PN-treated GPX1Ϫ/Ϫ cells continued to rise or was maintained at a fairly constant high level at 9 and 12 h (Fig.  4A). In contrast, it was decreased to approximately the base line at 6 h and remained at a low level at 9 and 12 h in the DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT cells. The upregulation of p21 WAF1/CIP1 protein expression was also observed in the whole cell extracts of the DQ-treated WT and the PN-treated GPX1Ϫ/Ϫ hepatocytes (Fig. 4B). However, there was no such initial increase in p21 WAF1/CIP1 in the DQ-treated GPX1Ϫ/Ϫ or the PN-treated WT hepatocytes. Instead, the protein showed significant decreases in these two groups at 6 and 3 h, respectively, and remained low thereafter. Although total p38 MAPK or JNK protein was unaffected by GPX1 knockout, DQ, or PN, both kinases were activated at 30 min in both types of cell by DQ and PN (Fig. 5). However, PN seemed to be a stronger stimulus than DQ and produced a greater level of p38 MAPK phosphorylation in WT than in GPX1Ϫ/Ϫ cells.
Intracellular GSH and GSSG in GPX1Ϫ/Ϫ Hepatocytes Respond to DQ and PN Differently from That in WT Cells-A much more abrupt time-dependent decline in intracellular GSH was produced by 0.5 mM DQ in WT than in GPX1Ϫ/Ϫ hepatocytes (Fig. 6A). Compared with the untreated controls, the decrease was 31 and 88% in WT cells, but only 7.5 and 46% in GPX1Ϫ/Ϫ cells at 30 min and 3 h, respectively. Thus, intracellular GSH was higher (p Ͻ 0.05) in GPX1Ϫ/Ϫ than in WT cells at these time points. Although DQ produced no change in intracellular GSSG in GPX1Ϫ/Ϫ hepatocytes at all, it resulted in a 15.5-fold increase over the untreated controls at 30 min in WT cells (Fig. 6B). That increase peaked at 1 h (18.2-fold), and declined to 9.3-fold at 6 h. In the PN-treated WT hepatocytes, intracellular GSH was decreased by 40.2% at only 5 min of the treatment (Fig. 6C). The decrease progressed linearly to 66.7% at 30 min and reached 82.3% at 6 h with a slight rise at 1 h. In contrast, the only significant decrease of GSH (46.2%, p Ͻ 0.05) caused by PN in GPX1Ϫ/Ϫ cells was seen at 3 h, along with a nearly complete restoration to the untreated cell level at 6 h. Likewise, PN did not affect intracellular GSSG in GPX1Ϫ/Ϫ cells, but caused a 5.4-fold increase over the untreated controls at 5 min in WT hepatocytes (Fig. 6D). That increase was progressively attenuated to 4.8-, 4.4-, and 1.9-fold at 10, 20, and 30 min, respectively, with a total disappearance at 1 h. In both DQ-and PN-treated GPX1Ϫ/Ϫ cells, ratios of intracellular GSH/GSSG were much higher than those in WT cells throughout or at most of the time points.
PN Induces More Protein Nitration in the WT than in GPX1Ϫ/Ϫ Hepatocytes-Despite a PN-dose dependent protein  4. Changes of p21 WAF1/CIP1 in hepatocytes induced by 0.5  mM DQ and 0.4 mM PN. A, initial up-regulation followed by significant cleavage of p21 WAF1/CIP1 in the nucleic fraction of the DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT cells. After cells were treated with DQ or PN for 0, 3, 6, 9, or 12 h, nuclei fraction was prepared to detect p21 WAF1/CIP1 protein by Western blot using anti-p21 WAF1/CIP1 antibody. B, changes of p21 WAF1/CIP1 in the whole cell lysate of the DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT hepatocytes. After treatment, whole cell lysate was prepared to detect p21 WAF1/CIP1 protein as described in A.
nitrotyrosine formation in both types of cells at 12 h, the total band intensity was 64 and 76% greater in WT than GPX1Ϫ/Ϫ cells treated with 0.2 and 0.4 mM PN, respectively (Fig. 7). Treating WT hepatocytes with 0.4 mM PN for 12 h decreased total GPX activity by 34% compared with the untreated controls (207 versus 316 units/mg of protein, p Ͻ 0.05). However, DQ alone did not induce protein nitration in either type of cells or significant reduction of GPX activity in WT cells (data not shown). DISCUSSION It is remarkable that GPX1 knockout exerted completely opposite impacts on susceptibility of mouse hepatocyte to DQ and PN-induced apoptotic death. Because high levels of H 2 O 2 could be produced by DQ (3), the substantial loss of cellular defense against DQ-induced apoptosis in GPX1Ϫ/Ϫ over WT cells is consistent with the whole body responses of the GPX1Ϫ/Ϫ mice challenged with ROS generators (18,40,41). However, the positive impact of GPX1 knockout on hepatocyte resistance to PN cytotoxicity is rather striking and does not agree with the notion that selenoproteins such as GPX1 (20) and selenoprotein P (42) may protect against PN-induced oxidative stress in vivo. Although PN is highly reactive with a short half-life, our results are reproducible and physiologically relevant. This is because we demonstrated a PN-dose dependent response of cell viability and nitrotyrosine formation, a reliable indicator of PN activity in the cell (6). Our selected PN dose (0.4 mM), similar to that used by others (20,43), was the minimal level that distinguished GPX1Ϫ/Ϫ from WT cells. Comparable results were obtained by using different sources or manipulations of PN treatment (data not shown). The enhanced nitrotyrosine formation in WT cells over that in GPX1Ϫ/Ϫ cells treated with 0.2 or 0.4 mM PN reflects a promoting role of GPX1, similar to that of other peroxidases (44), in the PN-mediated protein nitration.
A fundamental question is how GPX1 affects the PN-mediated apoptosis and protein nitration. In cell lysate, extrinsic GPX1 was able to reduce PN to nitrite using GSH in a twoelectron catalysis (20). However, we did not see a difference in PN reduction or medium nitrite level between GPX1Ϫ/Ϫ and WT cells treated with 0 -0.8 mM PN for 12 h (data not shown). Because of the strong reactivity between PN and CO 2 (22), GPX1 in the cultured cells, unlike in the cell lysates (20), might be encountered with not only authentic PN, but also more reactive PN intermediates such as ⅐ NO 2 or CO 3 . . As PN inacti- effective in competing against thiols or CO 2 as for direct reactions with PN (42) in these cells. However, this PN-mediated inactivation of GPX1 could not explain the enhanced sensitivity of WT cells to PN toxicity, because those cells still had much higher GPX activity than the GPX1Ϫ/Ϫhepatocytes (207-316 versus 4.2 milliunits/mg of protein). More likely, GPX1 exerted its role by affecting H 2 O 2 removal and thus cellular balances of ROS and RNS that could modulate PN toxicity (24,45,46).
The DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT cells clearly underwent apoptosis and exhibited similar apoptotic signaling, because cytochrome c release and pro-caspase-3 cleavages preceded the appearance of apoptotic cells and severe DNA fragmentation in these cells. As a critical step in stressinduced apoptosis, cytochrome c release from mitochondria enables it to bind to Apaf-1 and caspase-9, leading to the activation of caspase-9 that in turn activates caspase-3 (29). Caspase-3 is an executioner of apoptosis with many target proteins, including p21 WAF1/CIP1 (31) that protects against apoptosis (30). Cleavage of p21 WAF1/CIP1 mediated by caspase-3 and the consequent activation of cyclin A/Cdk2 have been shown as prerequisite for the execution of apoptosis in human hepatoma cells SK-HEP-1 induced by ginsenoside Rh2 (47). In the present study, the initial up-regulation of p21 WAF1/CIP1 protein expression at 3 h over the base line was maintained later only in the DQ-treated WT cells and the PN-treated GPX1Ϫ/Ϫ cells that showed no induced apoptosis. In contrast, the DQ-treated GPX1Ϫ/Ϫ and the PN-treated WT cells exhibited significant decreases of p21 WAF1/CIP1 at 6 and 9 h over the levels at 0 or 3 h. Both p38 MAPK and JNK, two kinases involved in stress-induced apoptosis (32,48), were activated by DQ or PN at 30 min, but their responses were not consistent with the changes of the three assayed apoptotic signal molecules. Seemingly, GPX1 exerted its role in the PN-or DQinduced apoptotic events downstream or independent of activation of these two kinases. Inhibition of the PN-induced activation of p38 MAPK and JNK by selenite in the cultured rat liver epithelial cells has been suggested to be through selenium-containing proteins, including GPX (49). In our study, activation of p38 MAPK was slightly stronger by DQ and much so by PN in WT than GPX1Ϫ/Ϫ cells. Thus, GPX1 promoted its activation mediated by ROS or RNS in mouse hepatocytes, indicating a possible cell-specific or GPX1-independent effect of selenite on these kinases.
Distinct differences in the DQ-induced cellular GSH/GSSG changes between the GPX1Ϫ/Ϫ and WT cells in the present study support the idea that GSH is a physiological substrate of GPX1 in metabolism (33,34). In the presence of GPX1, WT cells displayed a sharp decrease in GSH, along with an abrupt rise of GSSG, within 60 min after the DQ treatment. Although this GSH depletion attenuated after 60 min, probably because of the decrease in ROS production and(or) an accelerated regeneration of GSH from GSSG by glutathione reductase, GSH was indeed oxidized to GSSG by GPX1 to reduce the DQ-generated H 2 O 2 and other hydroperoxides at a very high rate initially. In contrast, GPX1Ϫ/Ϫ cells responded to DQ or PN with much less and slower depletion of GSH than WT cells, without any GSSG accumulation at all. Clearly, lack of GPX1 spared the oxidation of GSH to GSSG and left it for direct and/or GPX1independent protections (33,34,50). In the PN-treated WT cells, GSH seemed to act as a substrate of GPX1 initially and then became more like a GPX1-indpendent protector because a sharp rise in GSSG along with the GSH depletion was not seen after 60 min. In comparison with these distinct roles of GSH in functioning as a GPX1 substrate and a major antioxidant, the suggested necessity of certain amount of cellular GSH for cells to undergo apoptosis instead of necrosis (35) was not fully shown in our study. Although apoptotic events occurred in the DQ-treated GPX1Ϫ/Ϫ cells in which cellular GSH was indeed greater than in WT cells, these events were also exhibited in the PN-treated WT cells in which cellular GSH was depleted to a very low level initially. Thus, cellular GSH alteration alone may not be sufficient to regulate apoptosis.
Elucidating the opposite role of GPX1 in DQ-and PN-induced oxidative injury has broad implications. It teaches us that antioxidant protection for a given enzyme or protein such as GPX1 may not be a general property, but depends on the specific nature of oxidants. Although pro-oxidant properties of high levels of vitamin E or C (51, 52) and overexpression of Cu,Zn-superoxide dismutase (53) have been reported previously, our study provides the first evidence to show the "doubleedged sword" function of an "antioxidant" enzyme at its physiological expression level in metabolically normal primary cells. The potent role of GPX1 in turning off the DQ-induced and in switching on the PN-induced apoptosis will help us in elucidating mechanisms of ROS/RNS in regulating cell death and related signaling (54), and in developing novel therapeutic strategies for the ROS and RNS involved diseases (55). Our findings also caution the public that blind antioxidant supplementation in clinic or nutrition may not always be desirable. In line of our view, knockout of GPX1 enhanced mouse brain resistance to the kainic acid-induced epileptic seizure (56), whereas overexpressing GPX1 promoted acetaminophen toxicity to mice (57) and tumorigenesis (58). Likewise, vitamin C was able to induce decomposition of lipid hydroperoxides to endogenous geneotoxins (59).