Glutathione Supplementation Potentiates Hypoxic Apoptosis by S-Glutathionylation of p65-NFκB*

In murine embryonic fibroblasts, N-acetyl-l-cysteine (NAC), a GSH generating agent, enhances hypoxic apoptosis by blocking the NFκB survival pathway (Qanungo, S., Wang, M., and Nieminen, A. L. (2004) J. Biol. Chem. 279, 50455-50464). Here, we examined sulfhydryl modifications of the p65 subunit of NFκB that are responsible for NFκB inactivation. In MIA PaCa-2 pancreatic cancer cells, hypoxia increased p65-NFκB DNA binding and NFκB transactivation by 2.6- and 2.8-fold, respectively. NAC blocked these events without having an effect on p65-NFκB protein levels and p65-NFκB nuclear translocation during hypoxia. Pharmacological inhibition of the NFκB pathway also induced hypoxic apoptosis, indicating that the NFκB signaling pathway is a major protective mechanism against hypoxic apoptosis. In cell lysates after hypoxia and treatment with N-ethylmaleimide (thiol alkylating agent), dithiothreitol (disulfide reducing agent) was not able to increase binding of p65-NFκB to DNA, suggesting that most sulfhydryls in p65-NFκB protein were in reduced and activated forms after hypoxia, thereby being blocked by N-ethylmaleimide. In contrast, with hypoxic cells that were also treated with NAC, dithiothreitol increased p65-NFκB DNA binding. Glutaredoxin (GRx), which specifically catalyzes reduction of protein-SSG mixed disulfides, reversed inhibition of p65-NFκB DNA binding in extracts from cells treated with hypoxia plus NAC and restored NFκB activity. This finding indicated that p65-NFκB-SSG was formed in situ under hypoxia plus NAC conditions. In cells, knock-down of endogenous GRx1, which also promotes protein glutathionylation under hypoxic radical generating conditions, prevented NAC-induced NFκB inactivation and hypoxic apoptosis. The results indicate that GRx-dependent S-glutathionylation of p65-NFκB is most likely responsible for NAC-mediated NFκB inactivation and enhanced hypoxic apoptosis.

NFB family proteins have a conserved domain of ϳ300 amino acids in the amino-terminal region known as the Rel homology region. The Rel homology region consists of a DNA binding domain, a dimerization domain, and a nuclear localization signal (8). The carboxyl-terminal domains of NFB proteins are divided into two classes depending on the presence (p65, RelB, and c-Rel) or absence (p50 and p52) of the transactivation domain (9). NFB has cysteine residues close to or within its DNA-binding loops (9), e.g. Cys 59 of p50 and Cys 38 of p65, that contact the DNA backbone (10). Cys 62 of p50 is required for DNA binding activity (11). Previously, changes in cellular redox state and disulfiram (thiocarbamate) treatment were shown to induce S-glutathionylation and subsequent inactivation of the p50 subunit of NFB (11,12). Because only p65-NFB is responsible for transcriptional activation (13), it follows that glutathionylation of the transcriptionally active subunit p65 should have a more profound effect on NFB activation than glutathionylation of the p50 subunit.
The GSH/GSSG equilibrium is the major redox buffer in cells. Apart from providing cells with a reducing environment ([GSH] Ͼ Ͼ [GSSG]) and maintaining proteins in a reduced state, the glutathione redox couple dynamically regulates protein function via reversible disulfide bond formation (14,15).
An oxidative environment with high GSSG can promote protein-SSG formation. However, protein-SSG formation typically occurs intracellularly without substantial changes in the GSH/ GSSG ratio, indicating alternative mechanisms of protein-SSG formation that are incompletely understood (18,19). The enzyme glutaredoxin (GRx), also called thioltransferase, reverses protein glutathionylation in a specific and efficient manner. Accordingly, GRx is used to characterize oxidant-induced protein modification as protein-SSG (17). Although GRx is normally regarded as a deglutathionylating enzyme, GRx also catalyzes the reverse reaction, namely GSSG-dependent protein-SSG formation, and remarkably GRx can also enhance the rate of S-glutathionylation of proteins, e.g. actin, glyceraldehyde-3-phosphate dehydrogenase, and protein-tyrosine phosphatase 1B, by an alternative mechanism in the presence of glutathione thiol radical (GS ⅐ ), despite high GSH content (20). The p50 subunit of NFB is reported to be inactivated by S-glutathionylation, but thus far no report has described an oxidative modification of the p65 subunit.
Here we report that glutathione supplementation with N-acetyl-L-cysteine (NAC) decreases resistance to hypoxia in pancreatic cancer cells by inhibition of p65-NFB binding to DNA. The inhibition of NFB is reversible by GRx1 in vitro, and the inhibition is diminished in cells where GRx1 is knocked down, both characteristics indicative of S-glutathionylation of p65-NFB as the mechanism of inhibition. Thus, S-glutathionylation and inactivation of NFB leading to hypoxic cancer cell death may provide a useful adjunct to chemotherapy against hypoxic tumors.
Cell Culture-Human MIA PaCa-2 pancreatic carcinoma cells were obtained from the American Type Culture Collection (Rockville, MD). Cells were grown in complete culture medium containing Dulbecco's modified Eagle's medium supplemented with 25 mM glucose, 10% fetal bovine serum and antibiotics (50 units/ml of penicillin and 50 g/ml of streptomycin) in a humidified atmosphere of 5% CO 2 and 95% air. Cells were grown to a 60 -70% confluence before treatments.
Generation of the GRx1 Knock-down Cell Line-Phoenix amphotropic cells were transfected with a puromycin-selectable pSUPER.retro.puro retroviral vector (OligoEngine, Seattle, WA) containing a specific short hairpin oligonucleotide sequence against GRx1 (GRx1-shRNA) or non-silencing sequence (ns-shRNA) using Lipofectamine PLUS (Invitrogen). Viral supernatants were collected and used to infect MIA PaCa-2 cells. Cells were selected and grown in puromycin (0.5 g/ml) containing medium and screened for GRx1 protein expression and activity.
Hypoxia Experiments-MIA PaCa-2 cells were cultured on 60-mm Petri dishes (5 ϫ 10 5 cells/dish) for 42 h prior to exposure to hypoxia. Subsequently, cells were preincubated in the presence and absence of NAC for 3 h in complete culture medium supplemented with 10 mM glucose and 25 mM HEPES buffer. Petri dishes were placed in hypoxic Plexiglass chambers (Billups-Rottenberg, Del Mar, CA) and sparged with 95% N 2 , 5% CO 2 ; 0.5% O 2 , 94.5% N 2 , 5% CO 2 ; or 1% O 2 , 94% N 2 , 5% CO 2 at 2 p.s.i. for 4 min. Chambers were kept at 37°C throughout the course of the experiment. The oxygen concentration of the medium inside chambers was monitored continuously for 24 h using a FOXY-18G Fiber Optic Oxygen Sensor (Ocean Optics Inc., Dunedin, FL). After an initial equilibration of oxygen between gas phase and the medium, medium oxygen concentration inside the chamber remained constant during Ͼ24 h, indicating no oxygen back diffusion into the chambers. Nevertheless, some residual oxygen did remain inside the chambers (see "Results").
Transfection and Reporter Assay for NFB Activity-MIA PaCa-2 cells (5 ϫ 10 5 cells/dish) were plated on 60-mm Petri dishes for 30 h. Subsequently, cells were transiently transfected with 2 g of plasmid containing 5ϫ NFB promoter fused to the firefly luciferase gene (Stratagene) or 2 g of empty vector luciferase construct using Lipofectamine 2000 (Invitrogen). To normalize the transfection efficiency, cells were co-transfected with 2 g of ␤-galactosidase control vector. Sixteen hours posttransfection, cells were pretreated for 3 h with NAC (10 mM), exposed to hypoxia or normoxia for 18 h, and then harvested. Relative luciferase activity was measured with a luciferase assay system (Promega, Madison, WI) using a luminometer (Turner Design, Sunnyvale, CA). Luciferase activity was normalized for transfection efficiency using ␤-galactosidase activity, which was measured using a FluoReporter LacZ/␤-galactosidase quantitation kit (Molecular Probes) according to the manufacturer's instructions. Luciferase activity was assessed as relative light units and used as an indicator of transcriptional induction of NFB.
NFB Activation and Binding-Specific binding of activated NFB was measured with a Trans-AM TM NFB kit (Active Motif, Carlsbad, CA) according to the manufacturer's instructions. The oligonucleotide containing the NFB consensus sequence (5Ј-GGGACTTTCC-3Ј) is immobilized onto a 96-well plate. Only the active form of NFB in the cell extract specifically binds to the oligonucleotide. Epitopes on p50, p52, p65, c-Rel, and Rel-B are accessible only when NFB is bound to its target DNA. Nuclear extracts (5-10 g) were added to the wells followed by primary antibody against p65 and the horseradish peroxidase-conjugated secondary antibody. The optical density was measured at 450 nm with an absorbance plate reader.
Preparation of BSA-[ 35 S]SSG-BSA-[ 35 S]SSG was prepared as described earlier with the following modifications (22). After the reaction of S-carboxymethyl-BSA with N-succinimidylpyridyl bis(3,3Ј-dithiopropionate) and quenching with glycine, BSA was separated from small molecules either by gel filtration chromatography or by dialysis against two changes of 100 mM sodium phosphate buffer, pH 7.0. The modified BSA was then treated with excess [ 35 S]GSH for 1 h at room temperature. The resulting BSA-[ 35 S]SSG product was separated from [ 35 S]GSH as described previously and typically had ϳ0.9 GS eq/mol of BSA (22).
Radiolabel Assay for Glutaredoxin Activity-Whole cell lysates were analyzed for GRx activity with a radiolabel assay, which monitors the time-dependent release of radioactivity from BSA-[ 35 S]SSG (23). Aliquots of lysate and assay buffer (0.1 M Na ϩ /K ϩ phosphate buffer, pH 7.4, containing 0.5 mM GSH and 0.2 mM NADPH) were pre-warmed to 30°C and mixed with an aliquot of BSA-[ 35 S]SSG (final concentration of 0.1 mM) to initiate the reaction (total volume 0.5 ml). Aliquots of the reaction mixtures were precipitated with ice-cold trichloroacetic acid (final concentration of 10%) at 12 s, 1, 2, and 3 min, respectively. After centrifugation, the supernatants were analyzed for 35 S by scintillation counting. Total rates of deglutathionylation (slopes of [ 35 S]GSH released versus time) were corrected for non-enzymatic deglutathionylation by subtracting the rate of [ 35 S]GSH released by GSH in the absence of cell lysate. Enzymatic rates were expressed as nanomole of product/ min/mg of protein.
Nuclear and Cytoplasmic Extracts-Nuclear and cytoplasmic extracts were prepared according to a modification of the method of Wang et al. (24). Briefly, cells cultured on 150-mm Petri dishes were scraped into 1 ml of ice-cold phosphate-buffered saline. Cells were centrifuged at 1,600 ϫ g for 5 min at 4°C. Cell pellets were lysed in 300 l of low salt buffer (20 mM HEPES, 20% glycerol, 10 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, and 0.1% Triton X-100, pH 7.6) on ice. After a 20-min incubation, supernatants were collected as cytoplasmic extracts after centrifugation at 4,000 ϫ g for 5 min at 4°C. Nuclear pellets were then resuspended in high salt buffer (20 mM HEPES, 20% glycerol, 500 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, and 0.1% Triton X-100, pH 7.6) and tubes were rotated for 30 min at 4°C, centrifuged at 16,000 ϫ g, and supernatants were saved as nuclear extracts.
In Vitro Deglutathionylation Assay-Recombinant GRx1 was isolated as described previously (25). Nuclear extracts were pre-pared as described above except 10 mM N-ethylmaleimide (NEM) was included in lysis buffer to block all free thiols. NEMcontaining nuclear extracts were processed through spin columns (cut off 600 Da) to rid excess free NEM that potentially would inactivate the subsequent enzymatic reaction with GRx. Extracts were divided into different parts. One part was untreated and the other part was subdivided and aliquots were incubated with one of the following: (a) 4 mM dithiotheitol (DTT), (b) GRx1 plus GRx reducing system/recycling buffer (containing 0.2 mM NADPH, 0.5 mM GSH and 2 units/ml GSSG reductase), (c) recycling buffer alone. The assay was done with varying GRx1 concentrations using 1 M (0.023 units), 100 nM, and 10 nM GRx1. The incubation times for DTT and GRx1 were 10 and 5 min, respectively. GRx1, along with its recycling system, efficiently converts glutathione-protein disulfides to free protein thiols and GSH. NEM binds irreversibly to free sulfhydryl groups but does not react with cysteine residues that are already oxidized. Subsequently, p65-NFB DNA binding activity was measured by ELISA.
Western Blot Analysis-At the indicated time points, total cell extracts were prepared in ice-cold RIPA lysis buffer (150 mM NaCl, 1 mM EGTA, 1% sodium deoxycholate, 1% Triton X-100, 50 mM Tris-Cl, pH 7.4) supplemented with a mixture of protease inhibitors (Roche Diagnostics). The lysates were centrifuged, and the resulting supernatants were assayed for total protein content (Bio-Rad). Equivalent amounts of protein were diluted in sample buffer (200 mM Tris-Cl, 15% glycerol, 10% SDS, 5% ␤-mercaptoethanol, 0.01% bromphenol blue, pH 6.8) and resolved on SDS-polyacrylamide gel. The proteins were then transferred and immobilized on polyvinylidene difluoride membranes (Millipore, Bedford, MA) and probed with appropriate primary and secondary antibodies. Immunodetection was accomplished by an enhanced chemiluminescence detection system (Pierce).
Immunocytochemistry-MIA PaCa-2 cells were cultured on 35-mm glass bottom Petri dishes (MatTek Corp., Ashland, MA). Cells were fixed with 1% paraformaldehyde for 30 min at room temperature and permeabilized with 0.25% Triton X-100 for 5 min at room temperature. After washes using Tris-buffered saline with 0.1% Tween 20 (TBST), cells were incubated with blocking solution (1% BSA in TBST) for 1 h followed by primary monoclonal antibody against p65 NFB (1:100) overnight at 4°C. Cells were then washed three times with TBST and incubated with anti-mouse secondary Alexa Fluor 488 (1:500) for 1 h at room temperature, followed by three washes with phosphate-buffered saline. DRAQ-5 (1 M) was included in the final wash to stain the nuclei. Images of Alexa Fluor 488 fluorescence were collected using 488-nm excitation and Ͼ505-nm emission filters. DRAQ-5 fluorescence was imaged with a 633-nm excitation light and Ͼ640-nm emission filter using a Zeiss 510 NLO laser scanning confocal microscope.
Caspase-3 Activation-Caspase-3 activity was measured using a Caspase-Glo TM 3/7 kit (Promega) according to the manufacturer's instructions. At an indicated time point, cultured MIA PaCa-2 cells were collected into a test tube followed by centrifugation. The pellet was resuspended and lysed with RIPA buffer. Caspase-Glo TM 3/7 reagent and the lysate were mixed in 1:1 ratio, and luminescence was measured with a lumi-nometer. The resulting luminescence is proportional to caspase activity.
Apoptosis Assay-Apoptotic cell death was determined from nuclear morphology after propidium iodide staining in the presence of digitonin. At the indicated time points, floating and adherent cells were collected, centrifuged, and resuspended in phosphate-buffered saline containing 200 M digitonin (Calbiochem, La Jolla, CA) and 25 M propidium iodide (Molecular Probes). Digitonin permeabilizes the plasma membrane and allows propidium iodide to enter cells and stain all nuclei. Thus, propidium iodide staining in the presence of digitonin is equivalent to staining with Hoechst and 4Ј,6-diamidino-2-phenylindole, the two commonly used fluorescent dyes in the literature to assess apoptosis on a cell by cell basis. Apoptotic nuclei were scored as apoptotic based on nuclear condensation and fragmentation and counted by a ϫ40 microscope objective using a rhodamine filter set and expressed as a percentage of total cells. At least 200 cells were counted from three different microscopic fields for each sample. In addition, we previously showed that apoptotic death after hypoxia plus NAC treatment assessed by propidium iodide and digitonin correlates closely with apoptotic death assessed with a Cell Death Detection ELISA PLUS kit (Roche Applied Science), which measures mono-and oligonucleosomes of 180-bp multiples formed in the cytoplasmic fractions during apoptosis (3).
Statistical Analysis-Data are presented as mean Ϯ S.E. from at least three independent experiments. Differences were assessed by two-tailed paired Student's t test with Instat Software (GraphPAD, San Diego, CA). A p value Ͻ0.05 was considered statistically significant.

NAC-induced Apoptosis Is Oxygen-dependent in Pancreatic
Cancer Cells-Our previous study showed that NAC, a GSH generating agent, enhanced hypoxic apoptosis in transformed murine embryonic fibroblasts (MEFs) by a mechanism apparently involving GSH-dependent suppression of NFB transactivation (3). Because several hypoxia-dependent genes are activated within a narrow O 2 concentration range, we wanted to determine the window within which NAC is capable of enhancing hypoxic apoptosis in cancer cells. We exposed MIA PaCa-2 cells to different concentrations of O 2 in the presence and absence of NAC (10 mM). Based on direct measurements with an oxygen probe, O 2 concentrations in chambers flushed with 95% N 2 , 5% CO 2 ; 0.5% O 2 , 94.5% N 2 , 5% CO 2 ; and 1% O 2 , 94% N 2 , 5% CO 2 were 0.6, 1.6, and 2.1% O 2 , respectively. Thus, even in chambers with 95% N 2 , 5% CO 2 , medium O 2 content did not reach zero (anoxia), possibly due to residual O 2 in the gas mixture and outgassing of O 2 from plastic Petri dishes and other plastic parts inside the chamber. Hypoxia alone for 22 h in 2.1, 1.6, and 0.6% O 2 induced very little apoptosis (1.9% during normoxia versus 4.7, 3.7, and 3.1%, respectively). In contrast, NAC (10 mM) enhanced hypoxic apoptosis in a O 2 -dependent manner (2.4% during normoxia versus 10.3, 19.6, and 22.6%, respectively) (Fig. 1A).
NAC is rapidly taken up by the cells. Once inside the cells, endogenous esterases release free L-cysteine, which is a precursor of GSH synthesis (26). Like NAC, GSH ester (5 mM) enhanced apoptosis from 2.4% (normoxia) to 24% after 22 h exposure to hypoxia (1.6% O 2 ) (Fig. 1B). To confirm that NACmediated enhancement of apoptosis in hypoxic cells was due to an increase in GSH, we incubated cells with the D-isomer of NAC (D-NAC), which is not a precursor of GSH synthesis. Unlike the L-isomer of NAC, the D-isomer did not promote hypoxic apoptosis in MIA PaCa-2 cells (Fig. 1B).
NAC activated caspase-3 in hypoxic cells indicative of apoptotic death (Fig. 1C). PARP is a target of cysteine proteases, and caspase-3 cleaves 116-kDa PARP into a 85-kDa fragment during apoptosis (27). Hypoxia alone did not cause PARP cleavage assessed by Western blot analysis. In contrast, hypoxia in combination with NAC (10 mM) produced the 85-kDa cleavage product of PARP (Fig. 1D). Collectively, the results from Fig. 1 indicate that elevated GSH is responsible for enhancing hypoxic apoptosis in pancreatic cancer cells in agreement with our previous results in MEFs (3).
Inhibition of NFB Activation Is Oxygen-dependent in MIA PaCa-2 Cells-Hypoxia alone (1.6 and 0.6% O 2 ) increased NFB luciferase activity by 2.4-and 2.8-fold, respectively, compared with normoxia. The increase was blocked by NAC ( Fig.  2A). To determine whether blockade of hypoxia-induced transcriptional activation of NFB was due to decreased p65-NFB binding to DNA, NFB DNA binding was measured. Hypoxia (1.6% O 2 ) alone (18 h) increased p65-NFB DNA binding 2.1fold (compared with normoxia), which was abolished by NAC . Cells were then scored for apoptosis as in A. ‡, p Ͻ 0.001 compared with hypoxia alone. C, cells were treated with 10 mM NAC or no addition followed by hypoxia (0.6% O 2 ) for 18 h. Samples were prepared for caspase-3 measurements, as described under "Experimental Procedures." Z-VAD-fmk (50 M) was included as a negative control. Results are expressed as -fold increase over normoxia (N). #, p Ͻ 0.05 compared with hypoxia (H). Z, benzyloxycarbonyl; fmk, fluoromethyl ketone. D, cells were preincubated with 10 mM NAC for 3 h followed by hypoxia for 22 h. Whole cell lysates were subjected to Western blot analysis to detect PARP cleavage. Actin was used as a loading control. (Fig. 2B). NAC also affected NFB target genes. In our previous study, we identified XIAP as a target gene for NFB during hypoxia to MEFs (3). We also confirmed NFB-dependent activation of XIAP in MIA PaCa-2 cells during hypoxia, which was prevented by NAC (data not shown). The results indicated that NAC treatment inhibited transcriptional activation of NFB thereby inhibiting the survival of hypoxic cancer cells.
NAC Does Not Change NFB Protein Levels-NFB inactivation by NAC may also be due to decreased stability of NFB protein. To examine this possibility, we determined p65-NFB protein expression in whole cell lysates after hypoxia with Western blotting. During normoxia, p65-NFB protein levels were low and not changed by NAC (10 mM) (Fig. 3A). Hypoxia alone (1.6% O 2 , 16 h) increased p65-NFB protein levels, which were not affected further by NAC. Therefore, the capability of NAC to enhance hypoxia-induced apoptotic death did not appear to be due to decreased stability of the NFB protein (decreased expression or increased degradation) that would then lead to decreased expression of anti-apoptotic proteins. Rather, some other event was regulating NFB inhibition after NAC treatment of hypoxic MIA PaCa-2 cells. (3). Here, we further investigated whether NAC affects the nucleocytoplasmic shuttling process of NFB with Western blotting. Hypoxia increased nuclear translocation of p65-NFB, and NAC did not prevent the translocation (Fig. 3B). To corroborate these results by Western blotting, we performed immunocytochemical analysis of the p65-NFB after 18 h hypoxia (1.6% O 2 ). During normoxia, p65-NFB was primarily localized to the cytoplasm, and large dark voids of NFB immunofluorescence in confocal images corresponded to DRAQ-5-stained nuclei (Fig. 3C), indicating that NFB was localized to the cytoplasm and therefore was not activated. Hypoxia-induced p65-NFB translocation to the nucleus, which was not blocked by NAC (Fig. 3C). Thus, NAC did not affect NFB protein stability or nuclear translocation during hypoxia. Rather, the effect of NAC lays downstream of these events.

NAC Does Not Prevent Hypoxia-induced Nuclear Translocation of NFB-Previously, we showed that NAC does not prevent hypoxia-induced IB␣ degradation in MEFs
Our results so far indicated that NFB activation was critical for pancreatic cancer cell survival during hypoxia. To confirm independently the importance of NFB, we determined whether a NFB inhibitor would enhance hypoxic apoptosis in MIA PaCa-2 cells. Accordingly, we exposed cells to hypoxia (1.6% O 2 , 19 h) in the presence and absence of JSH-23 (10 M). JSH-23 inhibits NFB DNA binding and transcriptional activation without affecting IB␣ degradation (28). Even during normoxia, JSH-23 increased apoptosis from 2 to 5.2%, indicating a FIGURE 2. Inhibition of NFB transactivation is oxygen-dependent. A, MIA PaCa-2 cells were co-transfected with the NFB luciferase reporter construct or empty vector (EV) and ␤-galactosidase for 16 h. Cells were then exposed to different concentrations of O 2 (21, 1.6, and 0.6%) in the presence and absence of 10 mM NAC for 18 h. Cells were exposed to tumor necrosis factor ␣ (TNF) for 6 h as a positive control. Luciferase activity was normalized against ␤-galactosidase activity. Normoxic (21% O 2 ) value was taken as 1. #, p Ͻ 0.01 compared with normoxia. B, cells were incubated in the presence and absence of 10 mM NAC for 3 h followed by normoxia (N, 21% O 2 ) or hypoxia (H, 1.6% O 2 ) for 18 h. Binding of p65-NFB to its consensus DNA sequence in nuclear extracts was assessed by the enzyme-linked immunosorbent assay as described under "Experimental Procedures." *, p Ͻ 0.05 compared with normoxia. B, cells were exposed to normoxia (N) or hypoxia (H, 1.6% O 2 ) for 18 h in the presence and absence of 10 mM NAC. Cytoplasmic and nuclear fractions were prepared as described under "Experimental Procedures," and p65-NFB proteins were analyzed with Western blotting (15 and 30 g of protein/lane for cytoplasmic and nuclear extracts, respectively). YY-1 and actin were used as a nuclear marker and loading control, respectively. C, cells were pretreated with 10 mM NAC for 3 h followed by hypoxia (1.6% O 2 ) for 18 h. Samples were prepared for immunocytochemistry using anti-p65 antibody and Alexa 488conjugated anti-mouse IgG (green). DRAQ-5 dye (red) was included to stain the nuclei. Shown are representative confocal images. Scale bar is 10 m.
survival function of NFB that is constitutively activated in MIA PaCa-2 cells (Fig. 4). After hypoxia, JSH-23 and NAC increased apoptosis from 3.5 to 15 and 20%, respectively (Fig.  4). JSH-23, however, did not further enhance NAC-induced hypoxic apoptosis. Thus, NAC treatment during hypoxia is sufficient to inhibit NFB and induce apoptosis. Also, the results indicate that NFB pathway is crucial in protecting MIA PaCa-2 cells from hypoxia-induced cell death.
NAC Causes NFB Inactivation through Sulfhydryl Modification-p65-NFB has critical cysteines in its DNAbinding domain. To determine whether p65-NFB undergoes sulfhydryl modification, MIA PaCa-2 cells were exposed to hypoxia (1.6% O 2 ) for 18 h with and without 10 mM NAC. Subsequently, cells were treated with NEM (10 mM) to adduct free thiols and prevent alterations in thiol-disulfides during cell lysis and preparation of nuclear extracts. The extracts were then processed through spin columns (600 Da cut off) to rid excess NEM that would inactivate the later enzymatic reactions with GRx. Subsequently, extracts were divided into four parts. One part was untreated, and the other three parts were incubated with (a) DTT (4 mM), (b) GRx1 plus recycling buffer (GSH (0.5 mM), NADPH (0.2 mM), and GSSG reductase (2 units/ml)), and (c) recycling buffer alone. NFB DNA binding activity was then measured after each of the treatments.
NEM binds irreversibly to free sulfhydryl groups but does not react with oxidized sulfhydryl. DTT (a disulfide reducing agent) does not restore NEM-adducted thiols but does convert disulfides to free thiols. After hypoxia, DTT (4 mM) did not increase NFB DNA binding after NEM treatment, indicating that most sulfhydryls in the NFB protein were in the reduced and activated form after hypoxia (Fig. 5A). After hypoxia plus NAC, however, DTT increased NFB DNA binding by 3.2-fold compared with hypoxia alone (Fig. 5A). This result indicated that critical sulfhydryls in NFB required for DNA binding were converted to disulfides after hypoxia plus NAC, because reduction of these disulfides by DTT restored NFB DNA binding.
To determine whether S-glutathionylation contributes to NFB inactivation by NAC, NEM-treated cell lysates were incubated with the GRx system (GRx1, GSH, GSSG reductase, and NADPH) for 10 min before measurement of the NFB DNA binding capacity. GRx1 is a specific catalyst of protein-SSG reduction coupled to the GSH recycling system that maintains GRx and GSH in their reduced forms. Thus, effects of GRx in this system are diagnostic of functional modifications of proteins via S-glutathionylation (protein-SSG) (29 -31). The complete GRx system increased NFB DNA binding in hypoxia plus NAC-treated samples as compared with hypoxia (Fig. 5B). However, the GSH recycling system in the absence of GRx1 was without effect (Fig. 5C). As little as 10 nM GRx1 reactivated the NFB (Fig. 5D). Because GRx specifically catalyzes reduction of protein-SSG mixed disulfides, these results support the conclusion that NFB becomes S-glutathionylated during hypoxia plus NAC treatment to produce NFB inactivation. Moreover, ratios of NFB DNA binding for hypoxia plus NAC to hypoxia alone in DTT-treated and GRx-treated samples were indistinguishable (Fig. 5E). This indicates that GRx reverses the inhibition of NFB DNA binding to the same extent as DTT, which supports the conclusion that NAC-mediated thiol modification of NFB is mostly due to S-glutathionylation.
GRx Activity Is Required for p65-NFB-SSG Formation-Several proteins have been shown to undergo GRx-facilitated S-glutathionylation in the presence of GS ⅐ (20). First we determined whether hypoxia and/or NAC modulates cellular GRx activity. MIA PaCa-2 cells were incubated for 18 h under normoxic or hypoxic conditions in the presence and absence of NAC. None of the treatments, including hypoxia plus NAC, caused an irreversible change in the endogenous activity of GRx (Fig. 6).
To determine whether p65-NFB glutathionylation is affected by steady state GRx activity, we generated a stable GRx knock-down cell line by infecting Mia PaCa-2 cells with GRx1-shRNA or ns-shRNA. Cells infected with GRx1-shRNA showed a substantial decrease (60%) in GRx1 protein levels (Fig. 7A) and a corresponding decrease (77%) in GRx1 activity (Fig. 7B) compared with cells containing ns-shRNA. In ns-shRNA cells, NAC decreased hypoxia-induced p65-NFB DNA binding (Fig. 7C) and increased hypoxic apoptosis (Fig. 7D). By contrast, in GRx1-shRNA cells, NAC treatment did not decrease hypoxiainduced p65-NFB DNA binding (Fig. 7C), and increased hypoxic apoptosis only by 3% (Fig. 7D). Thus, knock-down of GRx1 protein was sufficient to prevent NAC-induced NFB inactivation and apoptosis during hypoxia consistent with a direct link between NFB S-glutathionylation and hypoxic apoptosis (see "Discussion").

DISCUSSION
The main conclusion of the present study is that GSH supplementation promotes hypoxic apoptosis in pancreatic cancer cells by suppressing the NFB survival pathway. The checkpoint for NFB inhibition does not lie at the level of IB␣ degradation or NFB nuclear translocation, as suggested by some studies (32,33). Instead, NAC enhances hypoxic apoptosis by inhibiting the binding of p65-NFB to DNA. We interpret this event to be largely due to GRx-dependent S-glutathionylation FIGURE 4. NAC-induced hypoxic apoptosis is not enhanced by NFB inhibitor. MIA PaCa-2 cells were preincubated with JSH-23 (1 h, 10 M) and NAC (3 h, 10 mM) alone or together with JSH-23 and NAC. Cells were then exposed to hypoxia (1.6% O 2 ) or normoxia for 19 h. Subsequently, cells were harvested and scored for apoptosis with propidium iodide and digitonin as described under "Experimental Procedures." of p65-NFB. This is the first evidence that GSH supplementation can promote S-glutathionylation of p65-NFB and apoptosis in hypoxic cancer cells.
A previous report shows that depending upon the nature of the stimulus, NAC either promotes or suppresses apoptosis independently of changes in GSH levels and redox status (34). However, we have previously shown that NAC-induced hypoxic apoptosis was not a consequence of an antioxidant effect but rather due to an increase of GSH content per se in MEFs (3). Here we showed that NAC and GSH-ester, both agents increasing intracellular GSH, promoted hypoxic apoptosis in pancreatic cancer cells (Fig. 1A). In contrast, D-NAC, which increases total cellular D-cysteine content but does not support GSH biosynthesis, did not enhance hypoxic apoptosis (Fig.  1B), confirming our previous findings in MEFs that NAC-mediated enhancement of apoptosis occurs via elevated levels of intracellular GSH (3).
GSH is the major thiol-disulfide redox buffer of the cell (35). Under pro-oxidant conditions, GSH itself can bind covalently to reactive cysteine thiols of various proteins to form mixed disulfides, a process termed S-glutathionylation (17,36). Many transcription factors, including NFB, have sensitive thiols that respond to changes in redox environment. Reduction of these sulfhydryls enhances transcriptional activation, whereas oxidation inhibits activation. Hypoxia caused p65-NFB activation and DNA binding, which was blocked by NAC (Fig. 2). Because D-NAC and non-thiol antioxidants like trolox did not block p65-NFB DNA binding (data not shown), the results indicated that the NAC-mediated effect on p65-NFB was GSH-dependent. These results were also in agreement with recent reports showing that NAC and increased GSH down-regulate cigarette smoke-and tumor necrosis factor-induced NFB activation in A549 alveolar epithelial, U937 human histiocytic lymphoma, HeLa human epithelial, and MCF-7 human breast cancer cells (37,38). Our results also indicate that the NFB pathway is crucial in protecting MIA PaCa-2 cells from hypoxia-induced cell death, as confirmed by the pro-apoptotic effect of the NFB inhibitor (Fig. 4).
To explore the nature of the sulfhydryl modification of the p65-NFB protein, we determined DNA binding of p65-NFB in vitro. After treatment with NEM, p65-NFB DNA binding in nuclear extracts of hypoxic cells was not increased by further treatment with DTT, suggesting that most sulfhydryls affecting p65-NFB DNA binding were in a reduced (NFB-SH) form after hypoxia and therefore susceptible to alkylation by NEM. In contrast after hypoxia plus NAC followed by NEM, DTT treatment increased the p65-NFB DNA binding capacity (Fig. 5A) suggesting that NAC led to modification of the sulfhydryls of the p65-NFB that are responsible for DNA binding. DTT is a general reducing agent and does not provide information regarding the nature of the sulfhydryl modification. However, GRx1, which specifically FIGURE 5. NAC causes NFB inactivation through sulfhydryl modification. MIA PaCa-2 cells were exposed to normoxia (N) and hypoxia (H, 1.6% O 2 ) in the presence and absence of 10 mM NAC for 18 h. Cultures were treated with 10 mM NEM to prevent any thiol/disulfide changes during processing, nuclear extracts were prepared and spun through the spin columns (600 Da cut off) to rid excess free NEM. Extracts were divided into different parts as described under "Experimental Procedures." Samples were then incubated with none, 4 mM DTT (A), 1 M GRx plus recycling buffer (0.2 mM NADPH, 0.5 mM GSH, and 2 units/ml GSSG reductase) (B), or recycling buffer alone (C). p65-NFB DNA binding activity was measured with enzyme-linked immunosorbent assay (ELISA). Absorbance value of normoxia without DTT (A), GRx/recyclin buffer (B), or recyclin buffer (C) treatment was set to 1. *, p Ͻ 0.05 compared with H; **, p Ͻ 0.01 compared with H. D, cells were exposed to hypoxia (H, 1.6% O 2 ) in the presence and absence of 10 mM NAC for 18 h. NEM-treated nuclear extracts were incubated with the GRx system (GRx, GSH, GSSG reductase, and NADPH) for 10 min. Concentrations of GRx (0 nM, 10 nM, 100 nM, and 1 M) were used for the assay. p65-NFB DNA binding activity was measured with ELISA. Absorbance value without treatment was set to 1. E, the values for hypoxia plus NAC (H ϩ NAC) were divided by the values for hypoxia alone (H) after DTT and GRx/Recycling treatment and expressed as ratios of H ϩ NAC/H. deglutathionylates protein-SSG mixed disulfides (17,23), also restored the p65-NFB DNA binding capacity in nuclear extracts of cells treated with hypoxia plus NAC (Fig. 5, B and D).
Moreover, the reactivation ratio for the GRx system matched that for DTT (Fig. 5E), which suggested that most of the p65-NFB sulfhydryl modification was p65-NFB-SSG. Although NAC also increases intracellular L-cysteine concentrations, the possibility that formation of p65-NFB-SS cysteine is responsible for NFB inactivation can be ruled out because GRx does not catalyze removal of cysteine from the protein-SS cysteine (14,17,23). Furthermore GSH-ester, which does not increase cellular L-cysteine levels, also enhanced hypoxic apoptosis (Fig.  1B), speaking against the involvement of protein-SS cysteine.
In addition to protein-SSG formation, protein thiols can form intermolecular disulfide linkages between different proteins during oxidative stress (39). Such modifications are reduced by DTT, but they are not sensitive to GRx catalysis. Furthermore, in Western blots of p65, we did not detect higher molecular weight bands under non-reducing conditions (data not shown). Thus, formation of inter-molecular disulfides on p65 is unlikely to be responsible for NAC-mediated NFB inactivation during hypoxia.
Unless intracellular GSSG concentrations reach unusually high levels (i.e. high levels of oxidant stress where GSH/GSSG Х 1), GSSG is unlikely to react to form protein-SSG based on typical redox potentials of cysteine residues in proteins (40). Consequently, GS-nitric oxide and GS ⅐ have been proposed as alternative GS donors for protein-SSG formation under conditions where GSSG does not accumulate (11,20,41,42). Hypoxia is known to induce formation of reactive oxygen species, such as superoxide (O 2 . ) and H 2 O 2 , that can be converted to hydroxyl radicals (OH ⅐ ) through the iron-catalyzed Fenton reaction (43). OH ⅐ is also generated without iron through peroxynitrite from the interaction between nitric oxide and O 2 . , both of which may be produced in hypoxia (43). GSH scavenges OH ⅐ to form the glutathione thiol radical GS ⅐ (20). GS ⅐ can also be formed from GS-nitric oxide (44  sine phosphatase 1B (20). This hypothesis would be consistent with our previous findings that NAC increases intracellular GSH by ϳ45% during hypoxia (3) creating conditions conductive to GS ⅐ generation, and that NAC-induced NFB transcriptional inhibition and apoptosis were O 2 -dependent ( Figs. 1A and 2A). Further support for the involvement of GRx in NAC-induced NFB inhibition and apoptosis is provided by the GRx1 knockdown experiment. Although GRx1 activity by itself was not affected by NAC treatment (Fig. 6), GRx1 protein knock-down prevented NAC-induced inhibition of p65-NFB DNA binding (Fig. 7C) and hypoxic apoptosis (Fig. 7D). Taken together, these data support the conclusion that GRx-mediated p65-NFB S-glutathionylation is a crucial event in GSH-mediated hypoxic apoptosis in cancer cells.
GRx1 lacks known nuclear translocation motifs and is localized primarily to the cytosol (45). This finding suggests that GRx1-dependent p65-NFB-SSG formation occurs in the cytosol and that the p65-NFB-SSG is further transported to the nucleus where it is unable to bind to DNA. However, the exact site of the p65-NFB-SSG formation remains to be determined. In a recent study, oxidative stress with H 2 O 2 inhibited NFB nuclear translocation and activation in lung cells, and this was attributed to inactivation of IB kinase (IKK) via S-glutathionylation of Cys 179 of IKK, most likely resulting from the initial formation of IKK-sulfenic acid (46). However, NAC treatment during hypoxia in the present study did not prevent IB␣ degradation (3) or p65-NFB nuclear translocation (Fig.  3), which is inconsistent with IKK inhibition. These findings indicate that IKK-␤ glutathionylation most likely did not play an important role in NAC-mediated NFB inhibition during hypoxia in the pancreatic cancer cells.
In summary, we identified GRx-dependent S-glutathionylation of p65-NFB as a major mechanism underlying the inhibition of the NFB survival pathway and promotion of apoptosis after GSH supplementation in hypoxic pancreatic cancer cells. Accordingly, we propose the following model (Fig. 8). Hypoxia activates NFB providing a survival signal against hypoxia-in-duced cell death. With enhanced intracellular GSH, OH ⅐ generated during hypoxia is scavenged by GSH (NAC) to form GS ⅐ . GRx1 reacts with GS ⅐ to form a GRx-SSG . intermediate (20), and this intermediate transfers GS ⅐ to form p65-NFB-SSG, a reaction that would be enhanced under hypoxic conditions (20). S-Glutathionylation then inactivates p65-NFB leading to cell death. Although the contribution of GS ⅐ to p65-NFB-SSG formation requires further investigation, our model would explain the observations that GRx1 knock-down prevents NAC-induced NFB inhibition and hypoxic apoptosis. Finally, the data suggesting that GSH supplementation enhances only hypoxic apoptosis have therapeutic implications, and administration of NAC may be a strategy to circumvent resistance to chemotherapy and radiation therapy in solid tumors whose core is often hypoxic. NAC treatment is attractive because NAC is already shown to be safe and effective for treatment of patients with acetaminophen overdose.