Tumor Necrosis Factor Increases Hepatocellular Glutathione by Transcriptional Regulation of the Heavy Subunit Chain of γ-Glutamylcysteine Synthetase*

Tumor necrosis factor (TNF) is an inflammatory cytokine that causes cell injury by generation of oxidative stress. Since glutathione (GSH) is a key cellular antioxidant that detoxifies reactive oxygen species, the purpose of our work was to examine the regulation of cellular GSH, the expression of heavy subunit chain of γ-glutamylcysteine synthetase (γ-GCS-HS), and control of intracellular generation of reactive oxygen species in cultured rat hepatocytes treated with TNF. Exposure of cells to TNF (10,000 units/ml) resulted in depletion of cellular GSH levels (50–70%) and overproduction of hydrogen peroxide (2–3-fold) and lipid peroxidation. However, cells treated with lower doses of TNF (250–500 units/ml) exhibited increased levels of GSH (60–80% over control). TNF treatment increased (70–100%) the levels of γ-GCS-HS mRNA, the catalytic subunit of the regulating enzyme in GSH biosynthesis. Furthermore, intact nuclei isolated from hepatocytes treated with TNF transcribed the γ-GCS-HS gene to a greater extent than control cells, indicating that TNF regulates γ-GCS-HS at the transcriptional level. The capacity to synthesize GSH de novo determined in cell-free extracts incubated with GSH precursors was greater (50–70%) in hepatocytes that were treated with TNF; however, the activity of GSH synthetase remained unaltered by TNF treatment indicating that TNF selectively increased the activity of γ-GCS. Despite activation of nuclear factor-κB (NF-κB) by TNF, this transcription factor was not required for TNF-induced transcription of γ-GCS-HS as revealed by deletion constructs of the γ-GCS-HS promoter subcloned in a chloramphenicol acetyltransferase reporter vector and transfected into HepG2 cells. In contrast, a construct containing AP-1 like/metal response regulatory elements increased chloramphenicol acetyltransferase activity upon exposure to TNF. Thus, TNF increases hepatocellular GSH levels by transcriptional regulation of γ-GCS-HS gene, probably through AP-1/metal response element-like binding site(s) in its promoter, which may constitute a protective mechanism in the control of oxidative stress induced by inflammatory cytokines.

Tumor necrosis factor-␣ (TNF) 1 is a polypeptide that elicits a diversity of cellular reactions, depending upon its concentration and the type of cell where it acts (1)(2)(3). Thus, it has been shown that TNF exerts an important physiological role as a modulator of immune responses by regulating specific genes needed for the host defense against a varied repertoire of agents. TNF appears to play a role in the control of cell cycle as DNA synthesis and cell proliferation increase in cells exposed to TNF, indicating that this cytokine acts as a mitogenic stimuli (4). Such regulation of gene expression by TNF is mediated by induction of early responsive genes including c-jun and transcription factors, i.e. NF-B (5,6). Yet, as a proinflammatory cytokine, whose production is increased in a number of stressful and pathological states, TNF promotes cell injury through several mechanisms including the overproduction of ROS (3,(7)(8)(9)(10).
The ability of TNF to kill cells appears to be restricted to tumor and virally infected cells since normal cells are generally insensitive to the toxic effects of TNF. Moreover, cells that normally are insensitive to TNF cytotoxicity can be sensitized by pretreatment with inhibitors of protein and RNA synthesis (11)(12)(13). Conversely, it has been shown that sensitive cells can be made resistant to TNF challenge by prior exposure to a sublethal dose of TNF. These findings imply that TNF leads to the induction of genes that confer protective effects on cells. Several protective genes induced by TNF have been reported including plasminogen activator inhibitor type 2, the zinc finger protein A20, and the Bcl-2 related family member A1 (14 -17).
The molecular basis of the cytotoxic action of TNF is not fully understood at present; however, one of the possible mechanisms involved in the toxicity elicited by TNF includes the generation of ROS that may damage critical cellular compo-nents such as proteins, lipids, and DNA causing cell injury (7)(8)(9). Consistent with the involvement of ROS in mediating the TNF-induced injury, cellular antioxidants attenuate the damaging effects of free radicals, and hence, these compounds may modulate the sensitivity of cells against TNF toxicity (18 -20). For instance, previous studies indicated that the antioxidant enzyme Mn-SOD determined sensitivity to TNF-induced cell death as TNF treatment up-regulated Mn-SOD gene expression; furthermore, overexpression of Mn-SOD conferred resistance to TNF toxicity in a human kidney embryonal cell line (18 -19). However, such a protective role of Mn-SOD may be restricted to specific cell types since studies in hepatocytes revealed that the TNF-induced expression of Mn-SOD was detected only at the mRNA level without increase in the protein level or enzyme activity (21).
GSH, the most abundant antioxidant in cells, plays a prominent role in the defense against oxidative stress-induced cell injury. Thus, the GSH redox cycle, where reduced GSH is cofactor of GSH peroxidase, and GST downplay the consequences of a broad range of reactive species (22)(23)(24). GSH is synthesized from its constituent amino acids in two sequential enzymatic reactions catalyzed by ␥-glutamylcysteine synthetase (␥-GCS) and GSH synthetase. The reaction catalyzed by ␥-GCS is the rate-limiting step in de novo GSH synthesis; ␥-GCS is inhibited by GSH through a feedback mechanism (22). Previous studies have shown that manipulation of GSH levels prior to exposure to TNF modulates the cytotoxicity of the cytokine in different cell types (7,8). GSH depletion induced by inhibition of GSH reductase with 1,3-bis(chloroethyl)-1-nitrosourea or by incubation with BSO, a specific inhibitor of ␥-GCS, prior to the exposure of cells to TNF, results in an increased susceptibility of hepatocytes and fibrosarcoma cells to the cytotoxic effects of TNF (7,8). Furthermore, its lethality can be ameliorated by N-acetylcysteine, a GSH precursor, which replenishes cellular GSH by providing intracellular cysteine (20).
Given the importance of GSH in protecting against the oxidative stress elicited by TNF, the purpose of our work was to examine the regulation of cellular GSH, expression of ␥-GCS-HS, and control of intracellular generation of ROS in cultured rat hepatocytes treated with TNF. Our results demonstrate that TNF increases cellular GSH levels, mediated by transcriptional regulation of the ␥-GCS-HS gene, which attenuates the generation of hydrogen peroxide and lipid peroxidation. Our findings imply that the up-regulation of cellular GSH may represent an additional protective mechanism to control the consequences of oxidative stress induced by inflammatory cytokines.
Isolation and Culture of Rat Hepatocytes-Rat hepatocytes were isolated by collagenase digestion; 2 ϫ 10 6 cells were plated on rat tail collagen and cultured routinely in Dulbecco's modified Eagle's medium/ F12 medium. Cell attachment averaged 50%, and cell numbers were determined using a Coulter counter, model multisizer II (Coulter Electronics, UK), and verified by hemocytometry. Cell viability was determined by trypan blue exclusion and by measurement in the medium of GST by conjugation of GSH with DTNB. HepG2 cells were obtained from the American type culture collection (ATCC, Bethesda, MD) and seeded in Dulbecco's modified Eagle's medium supplemented with high glucose.
Determination of Total GSH Equivalents and Synthetic Rate of GSH-The molecular forms of GSH equivalents, mainly GSH and GSSG, both in cells extracts and extracellular medium were determined by HPLC (26,27). The dynamic rate of GSH synthesis was determined in cell-free extracts prepared by fractionation of hepatocytes. Cell extracts were dialyzed overnight at 4°C to deplete cytosol GSH content to minimize feedback inhibition of GSH on ␥-GCS. The GSH synthetic capacity was determined using GSH precursors, glutamate, glycine, cysteine, and monochlorobimane as described previously in detail (28). GSH synthetase activity was assayed using glycine and ␥-glutamylcysteine instead of cysteine and glutamate. The rate of GSH formation was monitored as the net rate of fluorescence increase of GSH-monochlorobimane adduct catalyzed by GST over time after subtracting the BSOinhibitable fluorescence signal (28).
Measurement of Hydrogen Peroxide and Lipid Peroxidation-Production of reactive oxygen species, mainly hydrogen peroxide and other organic peroxides (29,30), and lipid peroxidation were monitored spectrofluorometrically by DCFDA or DHR and cis-parinaric acid, respectively. One h prior to harvesting of cells, DCFDA or DHR (2 g/ml) or cis-parinaric acid (5 g/ml) was added to culture plates followed by washing to remove excess fluorochrome, and fluorescence of these probes was determined as described previously (24,29).
Preparation of Nuclear Extracts and Electrophoretic Mobility Shift Assay of NF-B and AP-1-Preparation of nuclear extracts and assay of NF-B activation using a 32 P-labeled B oligonucleotide (5Ј-AGTT-GAGGGGACTTTCCCAGGC-3Ј) were as described previously (29). To confirm binding specificity a mutated B oligonucleotide probe (5Ј-AGTTGAATTCACTTTCCCAGGC-3Ј) was used as a negative control. Probes were labeled at the 5Ј end with T4 kinase and [␥-32 P]ATP (3000 Ci/mmol). Proteins were separated by native 6% polyacrylamide gel electrophoresis and visualized by autoradiography. In some cases, supershift assays were performed by incubating nuclear extracts with antibodies to different subunits of the NF-B/Rel protein family generously provided by Dr. N. Rice (31). Similar conditions were used for the binding activity of AP-1 using oligonucleotide (5Ј-CGCTTGATGAGT-CAGCCGGAA-3Ј) that was labeled as described above for NF-B (29).
Analysis of ␥-GCS-HS mRNA-A cDNA probe for ␥-GCS-HS was generated by reverse transcriptase-PCR. An 804-base pair partial ␥-GCS-HS cDNA was prepared using rat kidney RNA as template and primers based upon the published rat kidney ␥-GCS-HS sequence (32,33). A 5Ј-primer (5Ј-AGACACGGCATCCTCCAGTT-3Ј sense strand of the ␥-GCS-HS 5Ј sequence at positions 105-124) and a 3Ј-primer (5Ј-CTGACACGTAGCCTCGGTAA-3Ј antisense strand of the 3Ј-sequence at positions 890 -909) were synthesized to span 804 nucleotides of the ␥-GCS-HS mRNA. The size of the cDNA generated was confirmed by agarose electrophoresis. The reverse transcriptase-PCR fragment was cloned into pTARGET (Promega), and several clones were sequenced using fmol DNA sequencing system (Promega) to verify its identity; those with PCR-induced mutations were discarded. Total cellular RNA (30 g/lane) was extracted, denatured, filtered under vacuum on nylon membrane, and fixed with UV light. The membranes were prehybridized at 65°C, and hydridizations were performed using the 32 P-labeled ␥-GCS-HS probe. The same blots were stripped of the hydridized ␥-GCS-HS and rehybridized with a cDNA fragment of 18 S rRNA used as an internal reference control to normalize for RNA loading. In some cases, total RNA was denatured in 50% formamide containing 7% formaldehyde and electrophoresed through a 1% agarose, 7% formaldehyde gel, transferred, and fixed with UV light. Hybridization with 32 P-labeled ␥-GCS-HS probe identified a ␥-GCS-HS mRNA of 3.7 kilobases. The levels of ␥-GCS-HS mRNA were calculated relative to the 18 S band and expressed as percentage of control. Densitometry quantitation of autoradiographs was performed with a densitometer Preference (Seba).
Nuclear Run-on Assay of ␥-GCS-HS-Cultured rat hepatocytes (5 ϫ 10 7 cells) were treated with TNF and then washed with ice-cold phosphate-buffered saline and nuclei prepared by lysis with 0.5% Nonidet P-40 lysis buffer and differential centrifugation as described before (33). The integrity of isolated nuclei was verified by examination on a hemocytometer with a phase contrast microscope to ensure that cells had been lysed and nuclei appeared free of cytoplasmic material. Collected nuclei were stored in 50 (33). Proteinase K (100 g) was then added and incubated overnight at 42°C. Samples were then extracted with the Trizol reagent. Labeled RNA synthesized by intact nuclei was hybridized with ␥-GCS-HS cDNA and cross-linked on strips of nylon membranes. Hybridization of radiolabeled RNA with 18 S cDNA was used as internal reference control.
Transient Transfection and CAT Assay-HepG2 cells were seeded in 6-well tissue cultures plates and cultured until cell density reached 65-75% confluency. pCAT 3 Enhancer and pCAT 3 Control plasmids were transfected using the LipofectAMINE reagent (Life Technologies, Inc.), according to the manufacturer's instructions. Twenty four h after transfection, cells were exposed to TNF (500 and 10,000 units/ml) overnight. After TNF treatment, the cell extracts were isolated, and the protein content was determined by the color change of Coomassie Brilliant Blue G-250 dye (Bio-Rad protein assay). Chloramphenicol acetyltransferase (CAT) activity was quantitated by the CAT enzyme-linked immunosorbent assay (Boehringer Mannheim). ␤-Galactosidase expression plasmid (PSVgal, Promega) was cotransfected as an internal control to normalize the transfection efficiency.
Statistical Analyses-Statistical analyses for multiple comparisons of mean values between cell preparations were made by one-way analysis of variance followed by Fisher's test.

RESULTS AND DISCUSSION
Regulation of Hepatocellular GSH Levels by TNF and Control of ROS-One of the consequences of the oxidative stress induced by TNF is the generation of ROS that mediates the injury of cells exposed to TNF (7,8,20). Since many of the cellular effects elicited by TNF appear to be dependent on the concentration of the cytokine and in view of the observation that TNF exerts an important physiological role in various cellular functions, we investigated the effect of TNF on the regulation of total cellular GSH equivalents (GSH ϩ GSSG) and generation of hydrogen peroxide as indicator of ROS in cultured rat hepatocytes exposed to a wide range of TNF concentrations (250 -10,000 units/ml; 9 -370 ng/ml). Hydrogen peroxide was monitored by quantitation of fluorescence of DCF, a highly fluorescent probe sensitive to peroxides, which is formed from DCFDA (29,30). As seen in Fig. 1, cells labeled with DCFDA displayed a significant increase in DCF fluorescence when incubated with a high dose of TNF (10,000 units/ ml) (Fig. 1B). To confirm further the burst of ROS induced by TNF, cells were labeled with DHR, another probe that assesses generation of peroxides. Compared with control hepatocytes, TNF (10,000 units/ml) led to a significant increase in DHR FIG. 1. Dose-dependent effect of TNF on the oxidative stress response of hepatocytes. Cultured rat hepatocytes were incubated for 20 h with various concentrations of TNF as indicated. A, total cellular GSH equivalents (GSH ϩ GSSG) were determined by HPLC. B, hydrogen peroxide measurement following fluorescence of DCF by staining of cells with DCFDA was determined as described under "Experimental Procedures." C, the estimation of lipid peroxidation of cells treated with 10,000 units/ml TNF was performed by labeling cells with cis-parinaric acid followed by determination of fluorescence as described (23). D, parallel cultures were labeled with DHR followed by washing to remove excess probe to determine the effect of treatment of TNF at 500 units/ml ( fluorescence indicating generation of hydrogen peroxide (Fig.  1D, panel c). Since the levels of hydrogen peroxide increased as result of TNF exposure, it was important to determine the total cellular GSH equivalents under these circumstances. A depletion of cellular GSH was observed at the highest concentration of TNF, an effect that was significant compared with control cells or cells incubated with lower doses of TNF ( Fig. 1A and Table I). This effect was reflected mainly as a decrease in reduced GSH, which was translated into a lower GSH/GSSG ratio, suggesting that TNF induced an oxidative stress in hepatocytes. One of the characteristic effects of oxidative stress is peroxidation of membrane lipids. We labeled hepatocytes with cis-parinaric acid, a fluorescent fatty acid analogue, to determine the extent of lipid peroxidation. Cells treated with TNF (10,000 units/ml) displayed a greater loss of cis-parinaric acid fluorescence (Fig. 1C). Despite the degree of lipid peroxidation seen in hepatocytes treated with such a level of TNF, cytoxicity was minimal as viability of TNF-treated hepatocytes was similar to control cells (85-90% after 20 h of treatment).
To gain a better understanding of the regulation of cellular GSH by TNF and its consequences on ROS generation, we examined the effect of lower doses of the cytokine on these parameters. In contrast to the results seen with 10,000 units/ml TNF, TNF at 250 -500 units/ml did not increase the level of hydrogen peroxide determined as fluorescence of DCF or DHR (Fig. 1, B and D). Interestingly, however, total GSH equivalents increased (60 -80%) above control levels, mainly reflected as an increase of reduced GSH levels ( Fig. 1 and Table  I). TNF at 2,500 units/ml did not influence the level of either ROS or GSH, probably reflecting an equilibrium between ROS generation and their detoxification by GSH. Kinetic analyses of the effects of TNF on GSH levels revealed that the increase of reduced cellular GSH required 3-5 h of incubation with TNF. During this course of incubation, TNF did not affect the levels of reduced GSH nor GSSG in cells or medium.
The steady-state concentration of cellular GSH is regulated by its transport through specific plasma membrane GSH carriers (36). Recent observations indicated that apoptosis of human Jurkat T lymphocytes induced with anti-FAS/APO-1 antibody was preceded by a rapid and specific efflux of reduced GSH (37). Such a stimulated efflux process led to a quantitative loss of cellular reduced GSH that was recovered in the extracellular medium without change in the levels of GSSG. In contrast to these observations, our data indicated that the depletion of hepatocellular GSH induced by a high dose of TNF was not accounted for by an increased efflux of GSH into the extracellular medium (Table I). Our results imply that TNF leads to an overproduction of ROS that exhausts cellular GSH stores, probably by formation of disulfides with proteins favored by the oxidized environment. Although the proportion of GSSG released into the medium was greater in cells treated with TNF (10,000 units/ml), the magnitude of this increase was insufficient to account for the observed cellular depletion of GSH levels induced by TNF (Table I). It is noteworthy that in addition to GSH and GSSG found in the medium, another molecular form, a cysteine-GSH mixed disulfide, formed by transpeptidation of GSH in the presence of cystine, was detected in the extracellular medium, although its magnitude was similar for control and TNF-treated hepatocytes ( Table I).
Regardless of the mechanism, a lower availability of reduced GSH may be a critical factor that sensitizes cells to the chain of reactions leading to cell death induced by either Fas ligand or TNF (7,37). In the case of the hepatotoxicity induced by TNF, previous studies revealed that pretreatment of murine hepatocytes with 1,3-bis(chloroethyl)-1-nitrosourea, which inhibits the GSH reductase and disrupts the GSH redox cycle, depletes hepatocellular GSH resulting in cytotoxicity (7). In our case, in agreement with these findings, we observed that depletion of GSH levels by pretreatment with diethyl maleate or BSO sensitized cultured rat hepatocytes to TNF or acidic sphingomyelinase treatment. 2 Taken together, these results regarding the regulation of cell GSH and ROS reflect a divergent response of hepatocytes to TNF exposure, which seems to be dependent on the dose of TNF. The up-regulation of cellular GSH levels evoked by a low dose of TNF (250 -500 units/ml) attenuated the generation of ROS induced by TNF. However, at higher levels (10,000 units/ml), TNF resulted in depletion of reduced GSH, an effect that was accompanied by an increase of ROS, indicating that the former was overwhelmed by the increased generation of reactive species.
Transcriptional Regulation of ␥-GCS-HS Gene by TNF-In view of these results, we investigated the mechanism accounting for the up-regulation of cellular GSH levels. GSH is synthesized by the sequential action of two enzymes, ␥-GCS and GSH synthetase. ␥-GCS is the regulatory and rate-limiting enzyme that is composed of two subunits, heavy and light chains, that are discoordinately synthesized. The catalytic activity resides in the heavy subunit, whereas the light chain lowers the K m for glutamate and increases feedback inhibition by GSH on the heavy subunit (32). Thus, we determined the effect of TNF on the level of ␥-GCS-HS mRNA. As seen in Fig.  2, TNF increased the ␥-GCS-HS mRNA levels compared with control hepatocytes. Northern blot analyses confirmed that TNF treatment increased the 3.7-kilobase size of ␥-GCS-HS mRNA (not shown). Interestingly, the effect of TNF on the steady-state mRNA level of ␥-GCS-HS was significant at both low (500 units/ml) and high (10,000 units/ml) concentrations of TNF, although the level of induction by the low dose was greater than the induction seen by the high dose ( Fig. 2A). Such an increase in the level of mRNA of ␥-GCS-HS may reflect either a stabilization of its mRNA or an increased transcription rate of ␥-GCS-HS gene. To discern between these possibilities, we isolated intact nuclei from cells treated with TNF to exam-TABLE I Cellular and medium GSH molecular forms of cultured rat hepatocytes exposed to TNF Cultured hepatocytes were treated with TNF at the dose indicated for 20 h. Thereafter, medium was removed and cells scraped and treated with 10% trichloroacetic acid. The molecular forms of GSH equivalents from cells and medium were determined by HPLC as described under "Experimental Procedures." Results are expressed as mean Ϯ S.D. of n ϭ 7 cell preparations. ine the transcription rate of ␥-GCS-HS gene by nuclear run-on assay. Intact nuclei isolated from cells treated with TNF revealed an increased capacity to transcribe the ␥-GCS-HS gene (Fig. 2B). The next question we examined was if the greater levels of ␥-GCS-HS mRNA were paralleled by higher enzymatic activity. We determined the capacity to synthesize GSH de novo in the presence of unlimited GSH precursors in cell-free extracts isolated from hepatocytes that were pretreated with TNF. The dynamic synthetic rate of GSH was determined using monochlorobimane, a probe that forms a highly fluorescent adduct with GSH catalyzed by GST (28). GSH synthesis in cytosol extracts incubated with GSH substrates (glutamate, glycine, and ATP) and cysteine as the sulfur amino acid reflects the activities of both ␥-GCS and GSH synthetase. The GSH synthetic rate increased significantly (40 -60%) in cell extracts isolated from hepatocytes treated with TNF (500 units/ml) compared with control cells (Fig. 3A). Similar results were obtained with higher doses of TNF (10,000 units/ml) (data not shown). ␥-Glutamylcysteine, the product of ␥-GCS, is the substrate for GSH synthetase to which glycine is added forming the final product, GSH. When ␥-glutamylcysteine was used in the in vitro synthetic assay instead of cysteine, the formation of GSH reflected the activity of GSH synthetase. In contrast to the results obtained using cysteine as sulfur amino acid GSH precursor, the formation of GSH from ␥-glutamylcysteine was similar in hepatocytes with or without TNF treatment, demonstrating that the cytokine did not affect the specific activity of GSH synthetase (Fig. 3B). These findings indicate that the increased capacity of hepatocytes to synthesize GSH conferred by TNF was due to a greater activity of ␥-GCS that paralleled the increase in mRNA levels induced by TNF exposure. Similar findings were recently reported where interleukin-6 treatment of tumor-bearing mice displayed a significant increase in the hepatic ␥-GCS activity that was associated with a decrease in the hepatocellular sulfate level and lower glutamine to glutamate ratio (38).
To examine further the role of ␥-GCS activity on the homeostasis of cellular GSH and to decipher its impact on TNFinduced generation of ROS, these parameters were determined in conditions where ␥-GCS activity was inhibited by BSO. BSO treatment for 4 h did not significantly affect the GSH levels, yet it led to a significant depletion of cellular GSH after 20 h of BSO exposure (Fig. 4, A and B); however, this effect was not accompanied by an increased fluorescence of DCF (Fig. 4), indicating that depletion of GSH is insufficient for ROS overproduction. Furthermore, addition of TNF (500 units/ml) to cells pretreated with BSO resulted in a further depletion of cellular GSH compared with control or TNF-treated cells that were associated with a significant increase in DCF fluorescence, indicating overproduction of hydrogen peroxide. Similar results were obtained when gene transcription was blocked with actinomycin D. The effect of the transcription blocker on cellular GSH was modest as GSH levels did not decrease compared with control cells (Fig. 5). However, as in the case with BSO treatment, TNF, at the dose that efficiently increased cellular GSH levels, led to a significant depletion of GSH levels with respect to either control or TNF-treated cells, an effect that was accompanied by a significant increase of hydrogen peroxide (Fig. 5). Although blocking of transcription generally sensitizes cells against TNF, at the dose of TNF used (500 units/ml), hepatocytes remained viable (80 -90%). However, exposure of cells to a greater dose of TNF (10,000 units/ml) resulted in apoptosis as revealed by Hoechst-labeled hepatocytes, a fluorochrome that monitors chromatin condensation and fragmentation (data not shown). These findings suggest that the up-regulation of cellular GSH induced by TNF can be viewed as an adaptive mechanism to control the extent of ROS generation. Preventing the increase of GSH levels by inhibition of the rate-limiting enzyme of GSH synthesis, ␥-GCS, or blocking its transcription unmasks the full capacity of TNF to overproduce ROS.
Conversely, it can be inferred that if the molecular mechanisms by which TNF leads to ROS generation were disrupted, cells exposed to a high dose of TNF could display increased GSH levels. Since mitochondria play a critical role in the TNFinduced overproduction of ROS (8,23,24), we examined the effect of rotenone and thenoyltrifluoroacetone, blockers of electron transfer at complex I and II of respiration, respectively, on the effects of high doses of TNF (10,000 units/ml) on ROS and GSH levels. As seen in Fig. 6, blocking electron transfer at theses sites resulted in attenuation of TNF-induced ROS generation, which was translated in greater cell GSH levels compared with control or TNF-treated cells. These results are consistent with recent findings in isolated rat liver mitochondria incubated with ceramide, one the mediators of the biological effects of TNF, which highlight mitochondria as critical players in the generation of ROS induced by TNF (24). Similar findings were observed when hepatocytes were incubated with desferrioxamine prior to TNF exposure. 3 Desferrioxamine pre-vented the TNF-induced ROS generation as DCF fluorescence was similar for control and TNF-treated hepatocytes; consequently, the levels of total reduced GSH of TNF-exposed cells were greater than control cells. The increased GSH levels observed under these circumstances are consistent with the findings shown above that TNF leads to increased levels of ␥-GCS-HS mRNA. Therefore, these findings support the view that TNF initiates signaling pathways that culminate in overproduction of ROS and induction of cell GSH levels. The predominance of these cellular responses depends on the level at which TNF acts.
Our results have provided evidence that TNF leads to sustained elevation of hepatocellular GSH, an effect that may be considered as a protective mechanism to withstand the TNFinduced oxidative stress. This adaptive mechanism of GSH induction has also been observed in cells exposed to sublethal concentrations of quinones and Michael reaction acceptors (33,39). The mechanism whereby structurally unrelated substances, such as TNF and compounds with electrophilic or electron-deficient centers, lead to induction of ␥-GCS-HS mRNA remains to be characterized. A common feature of these agents is their ability to generate ROS, and consequently oxidative stress, which could mediate the induction of ␥-GCS-HS gene by increasing its transcriptional rate. Oxidative stress can influence gene transcription by activation of transcription fac-  tors such as NF-B that senses the redox environment of the cell.

Role of NF-B in the TNF-induced Transcription of ␥-GCS-HS: Functional Analyses of ␥-GCS-HS
Promoter-We next explored the mechanism by which TNF regulates ␥-GCS-HS at the transcriptional level. Activation of the pleiotropic factor NF-B is a well-described event initiated by a variety of stimuli including TNF (23,40). Such a transcription factor become activated upon release and subsequent degradation of the inhibitor moiety IB-␣ that sequesters NF-B in an inactive form in cytosol of resting cells. Upon activation, NF-B translocates to the nuclei where it binds to specific binding sites in the promoter of responding genes. Recent evidence revealed an anti-apoptotic functional role of NF-B in determining the fate of cells in response to TNF (41)(42)(43). Thus transgenic mice lacking the p65/RelA gene die during development as the result of massive apoptosis in the liver (41). In addition, cells express-ing a repressor of IB-␣ or its dominant negative mutant displayed an increased susceptibility to apoptosis induced by TNF. The mechanisms whereby NF-B protects cells against TNFinduced cell death is currently unknown. Recent studies have partially characterized the 5Ј-flanking region of ␥-GCS-HS promoter identifying several regulatory cis-acting elements including B binding site(s) (34,35,44). These observations, together with studies showing the involvement of NF-B in the induction of detoxification enzymes such as NAD(P)H:quinone oxidoreductase (DT-diaphorase) (45), prompted us to examine whether transcription factor NF-B mediates the transcriptional activation of ␥-GCS-HS induced by TNF.
We first verified if TNF efficiently activated NF-B by determining the pattern and dose-dependent activation of this transcription factor in response to TNF. Nuclear extracts of hepatocytes incubated with TNF and analyzed by gel retardation assay revealed the activation of NF-B in a dose-dependent fashion, being significant at 250 units/ml and maximal at 10,000 units/ml (Fig. 7, A and C). Such activation was specific for NF-B as the retarded band was displaced by competition with a fold excess of unlabeled probe, in contrast to the lack of displacement by a mutated B oligonucleotide (not shown). Several members of the NF-B/Rel family of transcription factors have been identified that can form dimers in several homo or hetero combinations, the p65(RelA)/p50 heterodimer being one of the best characterized examples (40). To identify the members of NF-B activated by TNF, we performed super shift assays where nuclear extracts were incubated with specific antibodies to different NF-B/Rel members before addition of labeled B probe. TNF resulted in a predominant activation of the heterodimer p65/p50 with minimal activation of homodimer p50 (Fig. 7B). The magnitude of NF-B activation at the highest dose of TNF used (10,000 units/ml) was 2-3-fold greater than the activation seen with 500 units/ml TNF; as the levels of ROS constitute one of the major triggers of NF-B activation, these results reflect the effect of oxidative stress induced by TNF on NF-B activation (Fig. 7) (9,29). In contrast, the dose of TNF that resulted in maximal induction of ␥-GCS-HS mRNA was 500 units/ml (Fig. 2), suggesting that activation of NF-B and induction of ␥-GCS-HS mRNA in response to TNF correlated inversely.
To determine the functional role of B in the transcriptional regulation of ␥-GCS-HS and to assess the contributory role of other factors and cis-regulatory elements contained in the 5Јflanking region of ␥-GCS-HS, we partially characterized the promoter of ␥-GCS-HS gene prepared by PCR from genomic Hepatocytes were treated with TNF at the dose indicated. A, nuclear extracts were isolated and incubated in the presence of labeled B oligonucleotide as described under "Experimental Procedures." As control, nuclear extracts were isolated from hepatocytes in the absence of TNF. DNA binding to B was analyzed by gel retardation. Specificity of binding to B was verified by displacement with excess (100-fold) unlabeled B oligonucleotide (not shown). B, parallel aliquots of nuclear extracts were incubated with antibodies to different members of B/Rel family before addition of labeled B oligonucleotide. Control (Ctrl) in this panel refers to nuclear extracts isolated from TNF-treated hepatocytes in the absence of antibodies. C, densitometric quantitation of heterodimer p65/p50 activated in A. *, p Ͻ 0.05 versus control; #, p Ͻ 0.05 versus TNF (250 -2000 units/ml). DNA of hepatocytes. Subsequently, hepatocytes were transiently transfected with plasmids containing the full-length or deletion constructs of the 5Ј-flanking region of the ␥-GCS promoter linked to the CAT reporter gene (34,35,44) (Fig. 8A). Cells that were transfected with the full-length construct, Ϫ1088/ϩ225, displayed a significant TNF-stimulated CAT activity. However, the shorter construct (227 bp), containing NF-B-like and xenobiotic response/antioxidant response-like elements, resulted in minimal CAT expression following TNF treatment compared with the full-length plasmid. High levels of transcriptional activation of ␥-GCS-HS promoter were induced by TNF when cells were transfected with the deletion plasmid, Ϫ756/ϩ210, which contained AP-1/AP-like transcription factor and MRE, as the magnitude of CAT activity induced by TNF was similar to that seen with the full-length plasmid (Fig. 8B). Futhermore, to establish whether TNF treatment induced DNA binding to the ␥-GCS-HS AP-1 site, we performed electrophoretic mobility shift assays with nuclear extracts prepared from hepatocytes that were treated with TNF and with a radiolabeled oligonucleotide encompassing the AP-1 binding site. Although hepatocytes displayed a constitutive AP-1 activation, after an initial decline in AP-1 activation, TNF led to a further activation of AP-1 being significant at 4 h post-treatment (Fig. 9).
These findings indicate a dispensable role of NF-B binding site for the TNF-induced transcriptional activation of ␥-GCS-HS gene and suggest the involvement of AP-1-like response elements. These findings are in agreement with previous reports where c-jun/c-jun homodimers (AP-1) mediate the regulation of ␥-GCS-HS gene induced by cisplatin and cadmium (34,46). Recent observations also indicated an induction of ␥-GCS-HS by cigarette smoking in human alveolar epithelial cells that was associated with AP-1 response element (47). At present we cannot completely discard the involvement of other cis-regulatory elements located further upstream in the 5Ј-flanking region of ␥-GCS-HS gene (44). Additional work will be needed to fully characterize the cis-regulatory elements responsible for induction of ␥-GCS-HS by TNF.
Concluding Remarks-The ability of mammalian cells to maintain cellular functions during oxidative stress depends on the rapid induction of protective antioxidant systems. Overproduction of TNF in certain pathological states results in cytotoxicity mediated by a variety of mechanisms including activation of phospholipase A 2 , increased generation of lipid intermediates such as ceramide, and formation of ROS leading to oxidative stress (3,7,8,23,24). Our findings provided evidence for the first time that TNF up-regulates the levels of reduced GSH as an adaptive control mechanism that attenuates the extent of ROS generation implying that cellular GSH depletion prior to exposure to TNF would determine a greater vulnerability to the cytotoxicity of TNF in a variety of cells 2 (7,8). In particular, as mitochondria stand as the source of the FIG. 8. TNF induction of ␥-GCS-HS-CAT promoter constructs transfected in HepG2 cells. A, the 5Ј-flanking region of ␥-GCS-HS was generated by PCR using specific primers described under "Experimental Procedures." The PCR fragment was isolated from agarose gels and digested with restriction enzymes to generate distinct fragments of the indicated size. These fragments were subcloned in the pCAT reporter vector as described under "Experimental Procedures." ARE, antioxidant response element; XRE, xenobiotic response element. B, deletion constructs were transiently transfected in HepG2 cells. Following transfection, TNF (500 units/ml) was added and incubated for 20 h. Subsequently, cells were harvested, and extracts were isolated and used for CAT activity that was determined by enzyme-linked immunosorbent assay. The pCAT reporter vector did not result in measurable CAT activity in cell extracts. CAT activity was normalized for protein content. Results are means Ϯ S.D. of n ϭ 3 individual experiments. *, p Ͻ 0.05 versus control; #, p Ͻ 0.05 versus cells transfected with 227 pb plasmid.
FIG. 9. AP-1 binding in nuclear extracts of hepatocytes treated with TNF. Nuclear extracts were prepared from cultured rat hepatocytes treated with TNF (500 units/ml) for various periods as indicated. Electrophoretic mobility shift assay was performed by incubating nuclear extracts with AP-1 oligonucleotide as described under "Experimental Procedures." Quantification of DNA binding was performed by densitometry shown in the right panel. Results are the mean Ϯ SD of n ϭ 3 different experiments. *, p Ͻ 0.05 versus control. burst of ROS induced by TNF, depletion of GSH in this organelle would favor a greater generation of ROS contributing to the TNF-induced cytotoxicity (24). Our findings are consistent with recent observations that nitric oxide protected cultured rat hepatocytes from TNF-induced apoptosis by a mechanism that required expression of heat shock protein 70 (48). Although these studies did not examine the regulation of hepatocellular GSH, it is possible that the protective role of heat shock protein 70 may have been mediated by an increase in the hepatocellular levels of GSH. In this regard, Mehlen et al. (49) have provided evidence that overexpression of small heat shock proteins protected murine fibrosarcoma cells against TNF-induced cell death by a mechanism that required up-regulation of cellular levels of GSH. Although these authors did not study the regulation of ␥-GCS, it is likely that the increased cellular GSH levels seen after overexpression of heat shock protein 27 may have been mediated by increased activation of ␥-GCS, as the protection afforded by heat shock proteins was abrogated by BSO pretreatment (49). The possibility that heat shock proteins may play an important role in mediating the TNFinduced transcriptional regulation of ␥-GCS deserves further work, which is currently under investigation. In this regard, previous observations have demonstrated that TNF induces a rapid phosphorylation of heat shock proteins (50).
Our findings indicate a dispensable role of NF-B in mediating the transcriptional activation of ␥-GCS. Despite activation of this transcription factor by TNF that mediates the expression of a repertoire of genes in response to TNF, other transcription factors that are also activated by TNF may participate in the gene regulation of this cytokine. In line with this, recent advances in the identification of TNF intermediates have linked the binding of TNF to its receptor 55-kDa TNF receptor subtype with the activation of mitogen-activated protein kinase cascade leading to activation of transcription factor AP-1 (51,52). Futhermore, since TNF elicits independent signaling cascades to transmit its action to the cell interior (53), future studies will be required to determine which signaling route mediates the TNF regulation of the ␥-GCS gene. Elucidation of the signaling mechanisms involved in the increased transcription of ␥-GCS may be helpful in designing strategies to control the oxidative stress induced by inflammatory cytokines in pathological situations, including tumor-induced cachexia, as GSH levels would control the extent and consequences of overproduction of ROS.