Mice with a Homozygous Null Mutation for the Most Abundant Glutathione Peroxidase, Gpx1, Show Increased Susceptibility to the Oxidative Stress-inducing Agents Paraquat and Hydrogen Peroxide*

Glutathione peroxidases have been thought to function in cellular antioxidant defense. However, some recent studies on Gpx1 knockout (−/−) mice have failed to show a role for Gpx1 under conditions of oxidative stress such as hyperbaric oxygen and the exposure of eye lenses to high levels of H2O2. These findings have, unexpectedly, raised the issue of the role of Gpx1, especially under conditions of oxidative stress. Here we demonstrate a role for Gpx1 in protection against oxidative stress by showing that Gpx1 (−/−) mice are highly sensitive to the oxidant paraquat. Lethality was already detected within 24 h in mice exposed to paraquat at 10 mg·kg−1 (approximately 1 7 the LD50of wild-type controls). The effects of paraquat were dose-related. In the 30 mg·kg−1-treated group, 100% of mice died within 5 h, whereas the controls showed no evidence of toxicity. We further demonstrate that paraquat transcriptionally up-regulatesGpx1 in normal cells, reinforcing a role forGpx1 in protection against paraquat toxicity. Finally, we show that cortical neurons from Gpx1 (−/−) mice are more susceptible to H2O2; 30% of neurons fromGpx1 (−/−) mice were killed when exposed to 65 μm H2O2, whereas the wild-type controls were unaffected. These data establish a function for Gpx1 in protection against some oxidative stressors and in protection of neurons against H2O2. Further, they emphasize the need to elucidate the role of Gpx1 in protection against different oxidative stressors and in different disease states and suggest thatGpx1 (−/−) mice may be valuable for studying the role of H2O2 in neurodegenerative disorders.

Reactive oxygen species (such as singlet oxygen, hydrogen peroxide, and hydroxyl radicals) are highly reactive molecules produced during the course of normal cellular processes involving oxygen (1,2). These strong oxidants have the potential to create situations of oxidative stress within cells by reacting with macromolecules and causing damage such as mutations in DNA, destruction of protein function and structure, and peroxidation of lipids (3,4). In order to protect against such harmful reactive intermediates, aerobic organisms have developed a number of cellular defenses. These include the antioxidant enzymes, namely superoxide dismutases, glutathione peroxidases and catalase (1,5,6).
Glutathione peroxidases (Gpxs) 1 catalyze the reduction of hydrogen peroxide as well as a large variety of hydroperoxides (such as DNA peroxides and lipid peroxides) into water and alcohols, respectively. There are at least five Gpx isoenzymes found in mammals. Although their expression is ubiquitous, the levels of each isoform vary depending on the tissue type. The cytosolic and mitochondrial Gpx1 (7,8) and the phospholipid hydroperoxide Gpx4 (or PHGpx) (9,10) are found in most tissues. Gpx1 is particularly abundant in erythrocytes, kidney, and liver, and Gpx4 is highly expressed in renal epithelial cells and testes. Cytosolic Gpx2 (or Gpx-GI) (11,12) and extracellular Gpx3 (or Gpx-P) (13)(14)(15) are poorly detected in most tissues except for the gastrointestinal tract and kidney, respectively. Recently, a new member, Gpx5, has been added to the Gpx family. This Gpx, which is specifically expressed in mouse epididymis, is interestingly selenium-independent (16,17).
Despite the well known ability of glutathione peroxidases to remove hydrogen peroxide and hydroperoxides, the exact role of these enzymes under physiological and oxidative stress conditions is still not clearly defined. For many years, there has been controversy regarding the primary role of these enzymes. While several authors supported the notion that these enzymes play a major role in the protection against oxidative damage under normal physiological conditions (18 -20), others believed in a protective role for these enzymes only in situations of oxidative stress (21)(22)(23). Recent studies using knockout mice as a model have provided unexpected data about the function of the most abundant glutathione peroxidase, Gpx1. Mice deficient in Gpx1 are healthy and fertile and do not show any histopathologies up to 15 months of age, thus suggesting a limited role for Gpx1 during normal development and under physiological conditions (24,25). More intriguingly, however, Gpx1 knockout mice display no abnormalities when exposed to an oxidative stressor such as hyperbaric oxygen (24). Furthermore, even though Reddy et al. (26) were able to show in vitro that lenses from Gpx1 knockout mice exposed to low concentrations of H 2 O 2 displayed evidence of abnormalities, Spector et al. (27,28) found that lenses from Gpx1 (Ϫ/Ϫ) mice exposed to high levels of H 2 O 2 (also in vitro) were not significantly different from those of wild-type mice. These latter data were interpreted as providing evidence that Gpx1 was dispensable even under conditions of oxidative stress, since catalase and glutathione were the functional molecules that detoxified H 2 O 2 . Therefore, the aforementioned studies (24,27,28) suggest, somewhat surprisingly, a limited role for this enzyme under situations of oxidative stress. Cheng et al. (25) have also shown that the knockout of the Gpx1 gene does not render the mice more susceptible to dietary, selenium, and vitamin E deficiencies and concluded that further studies on delineating the role of Gpx1 should employ more sensitive assays and more direct oxidative stresses.
In an attempt to further clarify the antioxidant role of Gpx1, we have investigated the susceptibility of our Gpx1 knockout mice to other oxidative challenges. Our data show that Gpx1 functions in the cellular protection against oxidative stress induced by paraquat as well as in the protection of neurons against hydrogen peroxide. Moreover, our results suggest that the protective role conferred by Gpx1 against paraquat toxicity is mediated, partly at least, via an increase in transcriptional activity induced by this agent.

Generation of Gpx1 Knockout Mice
A mouse 129/SVJ bacteriophage FIX II genomic library (Stratagene) was screened by hybridization with a Gpx1 genomic probe (29) to obtain a 5.2-kb SacI fragment that contained the entire mouse Gpx1 genomic sequence (confirmed by hybridization and restriction mapping analysis; data not shown). A neomycin resistance gene cassette (neo, 1.1-kb fragment, from pMC1neoPoly(A), Stratagene) was then blunt end-ligated into the unique SacII site within the first exon of Gpx1. Finally, a herpes simplex virus thymidine kinase cassette (Tk, 1.7 kb, from pMC1neoPoly(A)) was inserted into the SalI site, within the multicloning site, of pGEM 11Zfϩ (see Fig. 1A).
Mouse D3 embryonic stem cells (approximately 10 7 cells) were electroporated with 20 g of linear targeting sequence, as per Hwang et al. (30) and selected with 175 g⅐ml Ϫ1 G418 (Life Technologies, Inc.) and 2 M Cymevene™ (ganciclovir sodium) (Syntex, Australia). Resistant colonies were isolated and screened by Southern blotting. DNA was digested with BamHI and hybridized with a 770-bp Gpx1 EcoRI fragment (probe 1; see Fig. 1A). BamHI-restricted DNA was also hybridized with a 1.1-kb neomycin resistance gene (probe 2). Targeted clones were further confirmed by a second restriction with NcoI and hybridization with Gpx1 (probe 1) and neo (probe 3; a 850-bp neomycin resistance gene fragment excised from plasmid pMC1neoPoly(A) by NcoI and XhoI digestion) (see Fig. 1A). Clones containing the Gpx1 targeted sequence were microinjected into Balb/c blastocysts and transferred into pseudopregnant females. Chimaeras were mated with CF1 females, and progeny was analyzed by Southern blotting and/or PCR. 5Ј and 3Ј primers were designed to detect targeted and non-targeted Gpx1 alleles. The 5Ј primer is 5Ј-AAG GAG GTG CAG GCG GCT GTG AGC G-3Ј, and the 3Ј primer is 5Ј-GCG CGG AGA AGG CAT ACA CGG TGG-3Ј. PCR conditions were as follows: one 7-min denaturation step at 94°C followed by 30 cycles of (i) 1 min at 94°C, (ii) 1 min at 60°C, and (iii) 1 min at 72°C.

Gpx1 mRNA Expression
Total RNA was extracted from frozen tissue according to the method of Chomczynski and Sacchi (31). Twenty g of total RNA was loaded on each track of a 17.5% formaldehyde, 1% agarose gel, electrophoresed, transferred to Genescreen Plus membranes (NEN Life Science Products), and probed with a 770-bp EcoRI genomic [ 32 P]dCTP-labeled Gpx1 fragment (29). A 1.1-kb rat glyceraldehyde-3-phosphate dehydrogenase cDNA probe was used as a control for RNA loading. Filters were prehybridized for 2 h at 42°C in 50% formamide, 25% SSC, 1% Denhardt's solution, and 300 g⅐ml Ϫ1 salmon sperm DNA. Hybridizations were performed at 42°C in 10% dextran sulfate, 50% formamide, 1% SDS, 1 M NaCl, and 0.4 g⅐ml Ϫ1 salmon sperm DNA. Filters were washed in 2ϫ, 1ϫ, and 0.1ϫ SSC, 0.1% SDS for 15 min, and signals for Gpx1 were detected on a Fujix BAS 1000 phosphor image analyzer.

Tissue Preparation for Gpx Assays
Two-month-old mice were anaesthetized with an intraperitoneal injection of 1% avertin solution and cardiac perfused with phosphatebuffered saline (PBS). Brain, heart, liver, lung, and kidney were immediately removed and snap-frozen. Grinding was done in liquid nitrogen, and homogenates were obtained by adding 50 mM Tris-HCl, pH 7.8, containing 0.5% Triton X-100 and 0.1 mM EDTA and kept on ice. The samples were then centrifuged at 100 000 ϫ g for 1 h at 4°C, and the supernatants were collected and stored at Ϫ70°C until assays were performed.

Gpx and Protein Assays
Gpx activity was assayed using a modified method of Lawrence and Burk (32). Briefly, reduced glutathione and hydrogen peroxide were used as substrates, and oxidized glutathione produced by Gpxs was monitored through NADPH oxidation with glutathione reductase. The final concentrations of the reagents in the reaction mixture were 50 mM potassium phosphate, pH 7.0, 1 mM EDTA, 1 mM NaN 3 , 0.2 mM NADPH, 1 unit⅐ml Ϫ1 GSSG reductase, 4 mM GSH, and 0.25 mM H 2 O 2 . The change in absorbance at 340 nm was recorded. Gpx activity was calculated using an extinction coefficient of 6.22 ⅐ 10 Ϫ3 M Ϫ1 ⅐cm Ϫ1 for NADPH at 340 nm. Specific activities are expressed as nmol of NADPH⅐min Ϫ1 ⅐mg Ϫ1 protein. Protein content was determined using the Bio-Rad protein assay kit. Bovine serum albumin (Promega) was used as a protein standard.

Histological Analysis
Mice of approximately 6 weeks of age were anaesthetized with an intraperitoneal injection of 1% avertin solution. The entire body was cardiac perfused with PBS and then with Bouin's fixative. The organs were then embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Sections were examined under a light microscope.

Oxidative Stress Induced by Paraquat
Mice of approximately 6 weeks of age were injected intraperitoneally with paraquat at doses of 10, 20, or 30 mg⅐kg Ϫ1 body weight (methyl viologen (1,1Ј-dimethyl-4,4Ј-bipyridinium dichloride), catalog no. 1910-42-5; Sigma) prepared in PBS, and their health status was observed several times per day for 10 days. In some experiments, Gpx1 wild-type and knockout mice of approximately 6 weeks of age were injected intraperitoneally with paraquat at a dose of 30 mg⅐kg Ϫ1 body weight (prepared in PBS) or PBS and killed 4 h later, and lung Gpx activity was assayed as described above.

Analysis of Transcriptional Activity of Gpx1 Promoter after Paraquat Treatment
Vector Construction-A 1013-bp fragment containing the 5Ј-flanking or upstream promoter region of human GPX1 (GenBank TM accession no. AF029317) was isolated from a human genomic library and sequenced using Sanger's dideoxynucleotide chain termination method (33). A KpnI site was attached to the 5Ј or upstream terminus. The 1-kb GPX1 fragment was directionally cloned (using the KpnI site and a naturally occurring MunI site of GPX1) into a pGEM3 vector (Promega; previously modified to contain a MunI site). The naturally occurring GPX1 MunI site occurs downstream of the mRNA initiation site but upstream of the start (ATG) codon. The GPX1 fragment was then cloned into the pGL2-enhancer (Promega) in the correct orientation using the SacI and HindIII sites to produce the reporter construct pGL2(2A)-5ЈGPX1-luciferase.
Cell Culture and Transfection Procedure-HepG2 and CHO cells (ATCC) were maintained in DMEM, 10% fetal calf serum, 2 mM glutamine, and supplemented with 50 ng⅐ml Ϫ1 of selenium to eliminate variation in endogenous cellular Gpx1 activity (34) due to post-transcriptional regulation of Gpx1 by selenium content (35,36). Cells were plated at 10 5 cells/35-mm well and incubated overnight at 37°C and 5% humidified CO 2 . Transfection mixture was freshly prepared by mixing equal amounts of solution A (2 g of pGL2-derived plasmid, 3 g of SV40-galactosidase vector (Life Technologies, Inc.), and 100 l of DMEM) and solution B (10 l of Lipofectin in 100 l of DMEM), allowing this to incubate for 15 min at room temperature, and then adding 800 l of DMEM. The pGL2-derived plasmids were always co-transfected with SV40-galactosidase vector to correct for variations in transfection efficiency. Cells were rinsed twice with PBS and once with DMEM, and finally lipofection solution (1 ml) was overlaid onto the cells. Cells were incubated for 5 h. The transfection solution was rinsed away by washing with PBS, and the cells were removed by trypsinization, centrifuged, and then replated to allow for uniform distri-bution of transfected cells. Two ml of normal growth medium containing increasing concentrations of paraquat (0 -2 g⅐ml Ϫ1 or 10 g⅐ml Ϫ1 depending on cell line used) were then added, and cells were incubated for 18 h before luciferase and ␤-galactosidase assays were performed.
Luciferase Assay-After removal of the medium, cells were washed three times with ice-cold PBS, and 400 l of Analytical Luminescence Laboratory lysis buffer (diluted 1:3) was added. This was followed by agitation for 15 min at 4°C. Buffer containing lysed cells was transferred to a microcentrifuge tube and centrifuged at maximum speed for 5 min at 4°C, and supernatant was removed for luciferase analysis.
␤-Galactosidase Assay-After removal of the medium, cells were washed three times with ice-cold PBS, and 400 l of diluted reporter lysis buffer was added. Cellular debris was removed by centrifugation, and supernatant was collected. One hundred fifty l of 2ϫ galactosidase assay buffer was added to 150 l of supernatant, and the mixture was allowed to incubate for 2 h at 37°C. The reaction was stopped with 50 l of sodium carbonate, and the absorbance was determined at 420 nm. Within each sample, endogenous galactosidase cellular activity was corrected for by assaying a plasmid-negative transfected cell extract.

Oxidative Stress Induced by Hydrogen Peroxide
Neuronal Culture-Primary cultures of murine cerebral cortical neurons were prepared using a modified method of Larm et al. (37). Briefly, cerebral cortices were obtained from fetal Gpx1 wild-type and knockout mice (15 or 16 days of gestation) under sterile conditions and maintained on ice in Hanks' balanced salt solution (Life Technologies). Cortices were digested in medium containing 0.2 mg⅐ml Ϫ1 trypsin and 20 g⅐ml Ϫ1 DNase for 6 min at 37°C and dissociated by trituration (15 strokes with a 24-gauge needle). Cells were resuspended in Neurobasal Medium™ (Life Technologies) containing B27 (38), penicillin (100 units⅐ml Ϫ1 ), streptomycin (100 g⅐ml Ϫ1 ), L-glutamine (0.5 mM), and 10% dialyzed fetal calf serum (Sigma). Cells were seeded in poly-D-lysine (50 g⅐ml Ϫ1 )-coated plates at a density of 1.6 ϫ 10 5 cells⅐cm Ϫ2 and maintained in a humidified CO 2 incubator (5% CO 2 , 8.5% O 2 ; 37°C). After 24 h, the culture medium was replaced with serum-free growth medium. Test cultures were stained at day 8 for neuronal (MAP2) and glial (GFAP) markers using the method of Larm et al. (37).
Cell Viability-Cell viability was determined 24 h after H 2 O 2 treatment by the addition of MTT (Sigma) (37). MTT was dissolved in RPMI 1640 growth medium (Sigma) and 100 l of 5 g⅐ml Ϫ1 MTT was added to the cells in 1 ml of medium. Following a 30-min incubation at 37°C, cells were lysed by the addition of 500 l of 20% SDS in 40% dimethylformamide, pH 4.4. After overnight incubation at room temperature, absorbance at 570 nm was determined using a microplate reader.
Cultured Neuron Preparation for Gpx Assay-Cultured neurons were scraped from plates in PBS at day 8 of culture, pelleted by centrifugation, and resuspended in 50 mM Tris-HCl, pH 7.8, containing 0.5% Triton X-100 and 0.1 mM EDTA. The samples were then pelleted again by centrifugation, and the supernatants were collected and stored at Ϫ70°C until Gpx assays were performed as described above.

Generation of a Null Mutation in the Gpx1 Gene in Mice
The neomycin resistance and the thymidine kinase genes allowed both positive and negative selection of embryonic stem cells in culture (see Fig. 1A). One hundred forty-four resistant clones were isolated, and these were screened by Southern blotting. Two clones gave the expected BamHI restriction patterns (a 4.5-kb targeted allele and a 10.5-kb wild-type allele; Fig. 1B) after hybridization with a 770-bp Gpx1 genomic fragment (probe 1, see Fig. 1A). Southern blots hybridized with a neomycin probe (probe 2; see Fig. 1A) also gave a 7.1-kb fragment of a correctly targeted event (data not shown). A second restriction digest was performed to further validate the targeted event. NcoI-restricted DNA was hybridized with both the Gpx1 probe (probe 1) and a neomycin probe that lacked sequences 3Ј of the NcoI site (probe 3; see Fig. 1A). Probe 1 detected two fragments of approximately 15.0 and 11.2 kb (Fig.  1B) corresponding to the wild-type and targeted alleles, respectively, while probe 3 detected a fragment of approximately 5.0 kb (data not shown), further confirming a correctly targeted event. Heterozygous mice (ϩ/Ϫ) were identified by screening agouti pups by Southern blot analysis (Fig. 1B) and/or by PCR using primers that spanned regions of the Gpx1 gene adjacent to the neomycin resistance gene insertion site (5Јp and 3Јp; see Fig. 1A). A targeted allele gave the expected PCR product of approximately 1.3 kb, while the non-targeted allele gave a fragment of approximately 0.1 kb (Fig. 1C). Heterozygous mice were used in breeding experiments to generate mice that were homozygous (Ϫ/Ϫ) for the disruption of Gpx1 (Fig. 1, B and C).

Expression of Glutathione Peroxidase in Gpx1
Homozygous Null Mutant Mice Northern Blot Analysis-Northern blot analysis of mice homozygous for the targeted Gpx1 mutation (Ϫ/Ϫ) showed that Gpx1 mRNA expression was not detected in all of the five organs examined (liver, brain, heart, kidney, and lung) (Fig. 2). In wild-type mice, Gpx1 mRNA was abundantly expressed in the five organs with particularly high levels in the liver and kidney (Fig. 2).
Enzymatic Assays-Gpx activity was measured in tissues of mice homozygous for the targeted mutation and age-matched wild-type controls. The relative amount of Gpx activity in liver, kidney, lung, and brain of knockout mice was not detectable and/or was negligible as compared with controls (Ͻ4% of controls; Table I). A 32% residual Gpx activity was found in the heart of knockout mice. Such residual activity is most likely due to other Gpx isoenzymes, since the Gpx assay used in this study is not specific for Gpx1 but measures other Se-dependent Gpxs as well.
After Oxidative Stress Induced by Paraquat-To examine the importance of Gpx1 in protecting against oxidative stress, we challenged Gpx1 wild-type and knockout mice with paraquat, since it is a well known generator of free radicals within cells (40 -43). Homozygous Gpx1 knockout mice challenged with paraquat at 30 mg⅐kg Ϫ1 body weight all died within 5 h. No lethality occurred in wild-type controls even up to 10 days (240 h) of regular surveillance (Fig. 3). The 30 mg⅐kg Ϫ1 dose of paraquat that induced a 100% lethality within 5 h in the Gpx1 (Ϫ/Ϫ) mice is well below the documented LD 50 value of paraquat in saline or PBS (approximately 60 -70 mg⅐kg Ϫ1 (44)). Thus, mice deficient in Gpx1 display enhanced susceptibility to paraquat.
The susceptibility of Gpx1 (Ϫ/Ϫ) mice to paraquat appeared to be dose-related. Seventy-six percent of Gpx1 (Ϫ/Ϫ) mice succumbed to a dose of 20 mg⅐kg Ϫ1 body weight (Fig. 3). In this instance, lethality occurred between 6 and 14 h after paraquat Both the wild-type and targeted fragments are shown after restriction with BamHI or NcoI and hybridization with probe 1, 2, and 3, respectively. The 5Ј and 3Ј primers (5Јp and 3Јp, respectively) used in the polymerase chain reaction are indicated by the arrows. B, BamHI; N, NcoI; S, SacII. B, Southern blot analysis of a targeted mutation in the Gpx1 gene of embryonic stem (ES) cells and mice. Probe 1 was used to detect both wild-type (ϩ/ϩ) and heterozygous (ϩ/Ϫ) ES cell clones and wild-type (ϩ/ϩ), heterozygous (ϩ/Ϫ) and homozygous (Ϫ/Ϫ) mice after restriction with either BamHI or NcoI. Only one band of approximately 10.5 and 15.0 kb was detected in wild-type DNA of both ES cells and mice after BamHI or NcoI digestion, respectively. In (ϩ/Ϫ) ES cells and mice, a second hybridizing band of 4.5 and 11.2 kb was detected after hybridization with probe 1 after BamHI or NcoI digestion, respectively. Only one band of 4.5 and 11.2 kb was detected in (Ϫ/Ϫ) mice after BamHI or NcoI digestion, respectively. C, PCR to identify a targeted mutation in the mouse Gpx1 gene. Mouse tail DNA was used to establish the genotype by PCR. Wild-type DNA (ϩ/ϩ) gave only one PCR fragment of 0.1 kb. Heterozygous DNA (ϩ/Ϫ) detected one wild-type band (0.1 kb) and one targeted allele of 1.3 kb. Homozygous DNA (Ϫ/Ϫ) showed only one band of 1.3 kb. Each set of PCR reactions included a control (CONT) that was identical to the experimental reaction except that it did not contain DNA. treatment. One of 11 mice (9%) challenged with paraquat at 10 mg⅐kg Ϫ1 body weight died, and here lethality occurred at 15 h post-paraquat treatment. Essentially mice challenged with paraquat that succumbed did so within the first 24 h. However, animals were kept under regular surveillance for up to 10 days post-paraquat treatment to monitor whether those that survived displayed any adverse effects, and none were detected.

Gpx Activity Increases in Lungs after Paraquat Treatment
Having demonstrated that mice with a homozygous null mutation of the Gpx1 gene are particularly susceptible to paraquat, we investigated the effects of paraquat exposure on Gpx activity in organs of wild-type and knockout mice (in this instance, the lung). The rationale for these experiments was underpinned by our previous findings that paraquat treatment of cell lines induces Gpx activity (45). However, in previous studies, we were unable to differentiate whether this increase in Gpx activity involved an increase in Gpx1 activity, since the Gpx assay measured total Gpx activity.
Intraperitoneal treatment of wild-type mice with paraquat at a dose of 30 mg⅐kg Ϫ1 body weight significantly induced an approximately 2-fold elevation of Gpx activity (p Ͻ 0.01) as compared with controls injected with PBS alone (Fig. 4). On the contrary, no Gpx activity was detected in the lungs of knockout mice both in the absence of paraquat and after challenge with paraquat (Fig. 4). These results therefore indicate that Gpx1 up-regulation is part of the in vivo antioxidant defense against paraquat toxicity.

Transcriptional Activity of the GPX1 Promoter Is Induced after Paraquat Treatment
On the basis of our findings that paraquat treatment of mice induces Gpx activity in the lungs (at least), we investigated whether paraquat induces the transcription of the glutathione peroxidase-1 gene in order to ascertain whether such an induction was transcriptionally or post-transcriptionally mediated. A GPX1 promoter-luciferase-reporter construct (pGL2(2A)-5ЈGPX1-luciferase) was cotransfected with the SV40 promoter-␤-galactosidase-reporter construct into two different cell lines, rodent CHO and human HepG2 cells; the cells were treated with varying doses of paraquat; and luciferase and ␤-galactosidase activity were measured. All data were corrected relative to ␤-galactosidase activity in order to control for transfection efficiencies. Prior to commencing, we determined that exposure of cells up to 10 g⅐ml Ϫ1 paraquat had no effect on SV40 promoter activity and that the GPX1 promoter was approximately 4-fold stronger than the SV40 promoter at driving the luciferase reporter (3.6 Ϯ 0.3, n ϭ 3; data not shown).
Paraquat treatment of cells induced the transcription of the GPX1 promoter-luciferase-reporter construct in both CHO and HepG2 cells (Fig. 5, A and B, respectively). The transcriptional induction of the GPX1 promoter by paraquat was statistically significant at all concentrations tested in both cell lines. Furthermore, the transcriptional induction appeared to be dose- FIG. 2. Northern blots of Gpx1 expression in control and Gpx1 homozygous null mutant mice. Total RNA was extracted from liver, brain, heart, kidney, and lung of Gpx1 wild-type (ϩ/ϩ) and homozygous (Ϫ/Ϫ) mice, and 20 g was electrophoresed on 1% gels. After Northern transfer and hybridization with a Gpx1 probe (probe 1), one hybridizing band of 0.9 kb was detected in wild-type mice, as described previously (29). The Gpx1 probe failed to detect any Gpx1 expression in the homozygous mice (Ϫ/Ϫ). To control for RNA loading, the filter was also hybridized with a glyceraldehyde-3-phosphate dehydrogenase probe (Gapd).

TABLE I Gpx activity in tissues and cultured neurons of
Gpx1 wild-type and knockout mice Gpx activity was measured in liver, brain, heart, kidney, lung, and cultured neurons (day 8 of culture) of Gpx1 wild-type (ϩ/ϩ) and homozygous (Ϫ/Ϫ) mice. Specific activities are shown as nmol of NADPH ⅐ min Ϫ1 ⅐ mg Ϫ1 protein. The relative amount of Gpx activity in liver, kidney, lung, brain, and cultured neurons of knockout mice was negligible as compared with controls. A 32% residual Gpx activity was found in hearts of knockout mice. Values are means Ϯ S.E. (n ϭ 4) for all tissues except neurons, where n ϭ 3.

Organs
Gpx activity Percentage of wild-type , not detectable; defines specific activities lower than 3.5 nmol of NADPH ⅐ min Ϫ1 ⅐ mg Ϫ1 protein.

FIG. 3. Survival curve of Gpx1 wild-type and homozygous null mutant mice after challenge with the oxidizing agent paraquat.
Gpx1 wild-type and knockout mice of approximately 6 weeks of age were injected intraperitoneally with paraquat at 10, 20, or 30 mg⅐kg Ϫ1 body weight, respectively, and observed several times per day for 10 days for health status. Gpx1 wild-type mice were unaffected by an intraperitoneal injection of paraquat at 30 mg⅐kg Ϫ1 body weight (the mice were constantly monitored for 10 days). In contrast, homozygous knockout mice died rapidly within 5 h. This response to paraquat was dose-dependent. related in both cell lines with a plateau being reached at concentrations of 0.7 g⅐ml Ϫ1 paraquat in CHO cells (Fig. 5A) and 0.3 g⅐ml Ϫ1 paraquat in HepG2 cells (Fig. 5B). Interestingly, the maximal transcriptional induction of the GPX1 promoter by paraquat was between 2-and 3-fold, depending on the cell line tested, which was analogous to the levels of Gpx induction we observed when exposing the lungs of wild-type mice to paraquat. These data demonstrate that paraquat up-regulates glutathione peroxidase activity, partly at least, by inducing the transcription of the glutathione peroxidase-1 gene.

Primary Neuronal Cultures from Gpx1 Knockout Mice (Ϫ/Ϫ) Are More Susceptible to Hydrogen Peroxide
In light of our findings that Gpx1 knockout mice are more susceptible to an oxidative stressor such as paraquat, we investigated the effects of another oxidative stress-inducing stimulus, namely H 2 O 2 . In these experiments, we focused on the effects of this agent on cultured cortical neurons to elucidate whether Gpx1 plays a role in the protection of these cells against H 2 O 2 . Indeed, there is evidence that suggests that H 2 O 2 adversely affects the viability of neurons derived from Down syndrome brains (46,47) and that it may be involved in the cytotoxic action of the ␤-amyloid protein in the brains of individuals with Alzheimer's disease (47)(48)(49).
Prior to commencing the experiments, neuronal cultures were stained with the neuronal marker MAP2 (50) and the glial cell marker GFAP (51). Staining with MAP2 showed a monolayer of neurons with a complex network of neurite outgrowths. Staining with GFAP showed a small number of glia present in the culture. Counts of the positively stained cells revealed that the cell cultures used in this study contained more than 95% neurons (data not shown).
Gpx activity was assayed in neuronal cultures from Gpx1 wild-type and knockout mice. The level of Gpx activity in wildtype cells was low (9 Ϯ 1 nmol of NADPH⅐min Ϫ1 ⅐mg Ϫ1 protein, n ϭ 3) but reproducibly higher than the Gpx activity in neuronal cultures derived from Gpx1 knockout mice (the values obtained were below those of the sensitivity of the assay) (Table I).
Wild-type and Gpx1 knockout neuronal cultures were exposed to increasing hydrogen peroxide concentrations, and H 2 O 2 -mediated cell death was assessed using the MTT assay (37). No difference in cell viability was observed in cultures derived from Gpx1 (Ϫ/Ϫ) mice and wild-type controls in the absence of hydrogen peroxide (data not shown). Cell death due to H 2 O 2 was seen in Gpx1 (Ϫ/Ϫ)-derived neuronal cultures at concentrations as low as 65 M H 2 O 2 , where approximately 30% of neurons were killed. This contrasted with neurons derived from wild-type mice, where the same concentration had no significant effect on viability (Figs. 6A and 7). Furthermore, although the exposure of cultures to 80 and 100 M H 2 O 2 led to cell death of both wild-type and knockout neurons, significantly more neurons died in cultures derived from knockout mice as FIG. 4. Gpx activity in lungs of Gpx1 wild-type and knockout mice 4 h after challenge with paraquat (PQ). Gpx1 wild-type and knockout mice of approximately 6 weeks of age were killed 4 h after intraperitoneal injection of paraquat at 30 mg⅐kg Ϫ1 body weight or PBS, and their lungs were dissected and prepared for Gpx assays. Specific activities are shown as nmol of NADPH⅐min Ϫ1 ⅐mg Ϫ1 protein. Gpx activity was significantly increased in wild-type mice after paraquat treatment as compared with lungs of mice injected with PBS. On the contrary, Gpx activity remains not detectable (ND) in the lungs of both untreated and paraquat-treated knockout mice. Values are means Ϯ S.E.; n ϭ 3 (**, p Ͻ 0.01; Student's t test).
FIG. 5. Transcriptional induction of GPX1 promoter after exposure to paraquat. CHO cells and HepG2 cells were co-transfected with pGL2(2A)5ЈGPX1-luciferase and SV40-galactosidase-reporter constructs (to correct for variations in transfection efficiency) and exposed to increasing concentrations of paraquat. Cells were allowed to incubate for 18 h before luciferase and galactosidase assays were performed. The diagrams show luciferase activity of CHO cells (A) and HepG2 cells (B) after exposure to increasing paraquat concentrations. In both cell lines, a statistically significant increase in promoter activity occurred after exposure to increasing concentrations of paraquat (*, p Ͻ 0.05; Mann-Whitney U test). The highest concentrations of paraquat used (namely 2 and 10 g⅐ml Ϫ1 for CHO and HepG2 cells, respectively) were mildly and severely toxic to the cells, respectively. Values are means S.E.; n ϭ 3. SV40-luc, SV40 promoter-luciferase-reporter construct; GPX1-luc, GPX1 promoter-luciferase-reporter construct. compared with that derived from wild-type mice. At 80 M H 2 O 2 treatment, the percentage of viable neurons in the Gpx1 (Ϫ/Ϫ) cultures was approximately half of those of the wild-type controls, and at 100 M H 2 O 2 treatment no viable neurons remained in the cultures derived from knockout mice as compared with approximately 25% viability in control cultures (Fig. 6, B and C). Thus, these data demonstrate an enhanced susceptibility of neurons from Gpx1 knockout mice to an oxidative stressor such as H 2 O 2 . DISCUSSION We have successfully generated Gpx1 homozygous null mutant mice that show no Gpx1 mRNA by Northern blot analysis in all five organs, including the liver and kidney, where abun-dant levels of Gpx1 mRNA are found in wild-type mice. In addition, very little Gpx activity was measured in liver, kidney, lung, and brain as compared with wild-type mice, thus confirming the absence of Gpx1 protein synthesis and indicating that Gpx1 contributes to most of the Gpx activity in these tissues. In the heart, however, a 32% residual Gpx activity was found, most likely reflecting the activity of other Gpx enzymes. Under physiological conditions, our Gpx1 knockout mice developed normally, were fertile, and showed no evidence of macroscopic or microscopic phenotypic alterations. Furthermore, our analysis of neuronal cultures derived from Gpx1 (Ϫ/Ϫ) mice and wild-type controls showed no difference in cell death in the absence of oxidative challenge. Thus, our results, in agreement with Ho et al. (24) and Cheng et al. (25), indicate a limited role for Gpx1 during development and in the detoxification of H 2 O 2 and hydroperoxides under physiological conditions. Unlike Ho et al. (24), who failed to demonstrate any significant phenotype of GPX1 knockout mice under conditions of oxidative stress (i.e. hyperbaric oxygen), we were able to demonstrate a role for GPX1 in the protection against oxidative stress. Gpx1 (Ϫ/Ϫ) mice started dying rapidly (within 24 h) after exposure to the oxidant paraquat, at doses ranging from 1 ⁄7 the LD 50 of wild-type controls (LD 50 ϭ approximately 60 -70 mg⅐kg Ϫ1 when paraquat is prepared in saline or PBS (44)). All wild-type mice survived a 30 mg⅐kg Ϫ1 intraperitoneal injection of paraquat, while all knockout mice died within 5 h of exposure to this dose. Furthermore, the effect of paraquat was dose-related, since the lethality of Gpx1 (Ϫ/Ϫ) mice ranged from 9% in the 10 mg⅐kg Ϫ1 -treated group and progressively increased to 100% in the 30 mg⅐kg Ϫ1 -treated group. In addition, with increasing doses of paraquat, the Gpx1 (Ϫ/Ϫ) mice also died more rapidly. Our results therefore demonstrate an important role in vivo for Gpx1 in the protection against oxidative damage induced by paraquat. These data are corroborated by in vitro studies that suggest a role for Gpxs (in general) in the protection against paraquat toxicity; e.g. cells in which the cellular content of Gpxs was increased as a consequence of the transfection with Cu/Zn-superoxide dismutase cDNA were rendered more resistant to paraquat (22), and HL-60 cells selected for resistance to paraquat (in which no increase in superoxide dismutases occurred) were found to be enriched for Gpxs (45). Our current findings using the Gpx1 knockout mouse model, however, show that the Gpx1 isoform, specifically, is necessary for protection against paraquat toxicity. The exact biochemical mechanism whereby Gpx1 protects against paraquat toxicity still needs to be elucidated. Our mice provide a useful model for further biochemical studies to address this issue, especially since the treatment of paraquat toxicity in humans remains empirical and is not very successful (52).
The finding of this study that paraquat up-regulates Gpx1 transcription provides further support for a role for Gpx1 in protection against paraquat toxicity and sheds light on how paraquat increases Gpx1 in cells. Our data show an induction in Gpx activity in lungs of wild-type mice 4 h post-paraquat challenge. These data do not discriminate whether the increase in Gpx activity occurs in lung cells per se or is due to other effector cells (e.g. macrophages in an inflammatory response), although the second possibility is less likely, since (i) organs were thoroughly perfused before assay and (ii) no evidence for an inflammatory response was detected in lung tissue on histopathological analysis. Nevertheless, this induction was observed in wild-type mice only, establishing that Gpx1 is the isoenzyme increased by paraquat. Interestingly, these results also indicate that no redundancy exists for Gpx1 in these particular cells. Our in vitro experiments show that paraquat up-regulates the expression of a luciferase-reporter gene driven by the glutathione peroxidase-1 promoter and therefore indicate that paraquat elicits a response via the promoter/regulating regions of this gene to up-regulate its expression. This finding also adds further weight to the argument that paraquat induces Gpx1 expression in cells per se rather than solely inducing an increase via an influx of effector cells within a tissue. Our results are also consistent with earlier studies that show the transcriptional up-regulation of glutathione peroxidase-1 by oxidative stressors (53,54). It may be, however, that Gpx1 does not function universally as a protector against oxidant stress, since it appears not to protect against hyperbaric oxygen (24).
We also in this study challenged cultured neurons from our Gpx1 knockout mice with a second oxidative stress-inducing stimulus, namely H 2 O 2 , to elucidate whether Gpx1 plays a role in the protection of these cells against H 2 O 2 . Our data demonstrate that cultured neurons deficient in Gpx1 are more susceptible to H 2 O 2 -mediated toxicity than neurons from wild-type mice. Previous studies have suggested a major role for Gpx1 in the protection of the brain against hydrogen peroxide due to the fact that catalase activity is very low in this organ (55,56) and that Gpx4, the only other Gpx identified in the brain, reduces H 2 O 2 at a slower rate than Gpx1 (57). Our present data support the hypothesis that Gpx1 is important in the protection of neurons against H 2 O 2 toxicity. In addition, there is growing evidence indicating that reactive oxygen species and H 2 O 2 , in particular, play a role in aging and in the pathogenesis of some neurodegenerative diseases. Bar-Peled and colleagues (58) have shown that overexpression of Cu/Zn-superoxide dismutase predisposes neurons to oxidative stress-mediated apoptosis. Busciglio and Yanker (46) have demonstrated apoptosis and increased generation of H 2 O 2 in Down syndrome neurons in vitro. We have also previously shown that an altered Cu/Zn-superoxide dismutase/(glutathione peroxidase plus catalase) ratio exists in the brain of aging mice and that this correlates with increased lipid damage (59). In addition, our previous data also indicate that an altered antioxidant ratio induces features of cellular senescence and/or cell death in cultured cells and that this effect is mediated by hydrogen peroxide (60,61). Our present data are consistent with these studies and further strengthen the idea that an imbalance in the antioxidant defense system, resulting in elevated levels of H 2 O 2 , may be an important determinant in the aging process and neurodegenerative diseases.
In addition to the protection of neurons, Gpx1 may also function in protecting other tissues, specifically the eye lens and the myocardium, against H 2 O 2 -induced damage. Reddy et al. (26) have shown that lenses from Gpx1 (Ϫ/Ϫ) mice were much more sensitive in vitro (had elevated single strand DNA breaks and altered morphology) to assault by 25 M H 2 O 2 for 30 min than the corresponding wild-type controls. Yoshida et al. (62) have recently shown that isolated perfused hearts from Gpx1 knockout mice were more susceptible to ischemia reperfusion injury, suggesting the importance of Gpx1 in myocardial protection. Thus, our results, together with the aforementioned, demonstrate that functional deficits arise in tissues of mice deficient in Gpx1 activity, especially under certain conditions of oxidative stress. These data are indicative of the possible diseases that may be linked to a relative deficiency in Gpx1 activity (viz. neurodegenerative disorders, cardiovascular diseases, and/or ophthalmic disorders). Consequently, these Gpx1 (Ϫ/Ϫ) mice may prove useful in defining the exact role and/or biochemical functions of Gpx1 in such disorders. Our study and those of Reddy et al. emphasize and reiterate the suggestion of Chen et al. (25) that studies aimed at testing the role of Gpx1 under different oxidative stresses and/or the utilization of sensitive end points in such studies are necessary to define the biological/biochemical functions of Gpx1 and its role in disease pathogenesis.
In summary, our data demonstrate a major role for Gpx1 in the protection against oxidative stress induced by paraquat in vivo as well as its importance in protecting neurons against H 2 O 2 in vitro. This protection is not absolute, but the animals/ cells are much more susceptible when they are deficient in Gpx1 activity. In addition, our results shed light on the mechanism(s) whereby Gpx1 protects against paraquat toxicity, showing that this is mediated, partly at least, via an increase in the activity of Gpx1 and that such an increase may be transcriptionally mediated. Gpx1 knockout mice provide a unique model system to elucidate the contribution of Gpx1 in the protection against situations of oxidative stress and will prove illuminating in future studies on the mechanisms whereby its deficiency may predispose to certain disease states.