Constitutive and beta-naphthoflavone-induced expression of the human gamma-glutamylcysteine synthetase heavy subunit gene is regulated by a distal antioxidant response element/TRE sequence.

Glutathione (GSH) is an abundant cellular non-protein sulfhydryl that functions as an important protectant against reactive oxygen species and electrophiles, is involved in the detoxification of xenobiotics, and contributes to the maintenance of cellular redox balance. The rate-limiting enzyme in the de novo synthesis of glutathione is γ-glutamylcysteine synthetase (GCS), a heterodimer consisting of heavy and light subunits expressing catalytic and regulatory functions, respectively. Exposure of HepG2 cells to β-naphthoflavone (β-NF) resulted in a time- and dose-dependent increase in the steady-state mRNA levels for both subunits. In order to identify sequences mediating the constitutive and induced expression of the heavy subunit gene, a series of deletion mutants created from the 5′-flanking region (−3802 to +465) were cloned into a luciferase reporter vector (pGL3-Basic) and transfected into HepG2 cells. Constitutive expression was maximally directed by sequences between −202 and +22 as well as by elements between −3802 to −2752. The former sequence contains a consensus TATA box. Increased luciferase expression following exposure to 10 μM β-NF was only detected in cells transfected with a reporter vector containing the full-length −3802:+465 fragment. Hence, elements directing constitutive and induced expression of the GCS heavy subunit are present in the distal portion of the 5′-flanking region, between positions −3802 and −2752. Sequence analysis revealed the presence of several putative consensus response elements in this region, including two potential antioxidant response elements (ARE3 and ARE4), separated by 34 base pairs. When cloned into the thymidine kinase-luciferase vector, pT81-luciferase, and transfected into HepG2 cells, both ARE3 and ARE4 increased basal luciferase expression approximately 20-fold. When cloned in tandem in their native arrangement the increase in luciferase activity was in excess of 100-fold, suggesting a strong interaction between the two sequences. Luciferase expression was elevated in β-NF-treated cells transfected with the ARE4-tk-luciferase vector and all DNA fragments containing ARE4. In contrast, ARE3 did not direct increased luciferase expression in response to β-NF nor did it significantly modify the magnitude of induction directed by ARE4. The influence of the ARE4 oligonucleotide on constitutive and induced expression was eliminated by introduction of a single base mutation, converting the core ARE sequence in ARE4 from 5′-GGACTCAGCG-3′ to 5′-GGACTCAGCG-3′. When introduced into the full-length −3802:+465 segment, the same single base mutation also eliminated both functions. Collectively the data indicate that the constitutive and β-NF-induced expression of the human GCS heavy subunit gene is mediated by a distal ARE sequence containing an embedded tetradecanoylphorbol-13-acetate-responsive element.

Glutathione (GSH) is an abundant cellular non-protein sulfhydryl that functions as an important protectant against reactive oxygen species and electrophiles, is involved in the detoxification of xenobiotics, and contributes to the maintenance of cellular redox balance. The rate-limiting enzyme in the de novo synthesis of glutathione is ␥-glutamylcysteine synthetase (GCS), a heterodimer consisting of heavy and light subunits expressing catalytic and regulatory functions, respectively. Exposure of HepG2 cells to ␤-naphthoflavone (␤-NF) resulted in a time-and dose-dependent increase in the steady-state mRNA levels for both subunits. In order to identify sequences mediating the constitutive and induced expression of the heavy subunit gene, a series of deletion mutants created from the 5-flanking region (؊3802 to ؉465) were cloned into a luciferase reporter vector (pGL3-Basic) and transfected into HepG2 cells. Constitutive expression was maximally directed by sequences between ؊202 and ؉22 as well as by elements between ؊3802 to ؊2752. The former sequence contains a consensus TATA box. Increased luciferase expression following exposure to 10 M ␤-NF was only detected in cells transfected with a reporter vector containing the full-length ؊3802:؉465 fragment. Hence, elements directing constitutive and induced expression of the GCS heavy subunit are present in the distal portion of the 5-flanking region, between positions ؊3802 and ؊2752. Sequence analysis revealed the presence of several putative consensus response elements in this region, including two potential antioxidant response elements (ARE3 and ARE4), separated by 34 base pairs. When cloned into the thymidine kinase-luciferase vector, pT81-luciferase, and transfected into HepG2 cells, both ARE3 and ARE4 increased basal luciferase expression approximately 20-fold. When cloned in tandem in their native arrangement the increase in luciferase activity was in excess of 100-fold, suggesting a strong interaction between the two sequences. Luciferase expression was elevated in ␤-NF-treated cells transfected with the ARE4-tk-luciferase vector and all DNA fragments containing ARE4. In contrast, ARE3 did not direct increased luciferase expression in response to ␤-NF nor did it significantly modify the magnitude of induction directed by ARE4. The influence of the ARE4 oligonucleotide on constitutive and induced expression was eliminated by introduction of a single base mutation, converting the core ARE sequence in ARE4 from 5-GTGACTCAGCG-3 to 5-GGGACTCAGCG-3. When introduced into the fulllength ؊3802:؉465 segment, the same single base mutation also eliminated both functions. Collectively the data indicate that the constitutive and ␤-NF-induced expression of the human GCS heavy subunit gene is mediated by a distal ARE sequence containing an embedded tetradecanoylphorbol-13-acetate-responsive element.
Glutathione (L-␥-glutamyl-L-cysteinyl-glycine, GSH), 1 a nonprotein sulfhydryl compound present in millimolar concentrations in virtually all cells, serves a myriad of cellular functions and plays a prominent role as an intracellular protectant (1,2). GSH is an effective oxygen radical scavenger and serves as a critical co-factor in peroxide detoxification via a reaction catalyzed by glutathione peroxidase. Furthermore, conjugation with GSH is an integral step in the detoxification and elimination of diverse classes of toxic chemical compounds. The formation of hydrophilic glutathionyl conjugates is catalyzed by glutathione S-transferases, a family of isozymes that mediate the conjugation reaction in a substrate-dependent fashion (3). Long the object of interest from a toxicology perspective, the protective properties of GSH have assumed even further significance since GSH not only plays a critical role in protection of normal cells, but it has recently been implicated in protection of neoplastic cells from a number of chemotherapeutic agents that exert their cytotoxic effects via generation of reactive oxygen species or production of electrophilic intermediates (4,5). The augmentation of GSH and GSH-related detoxification systems has also engendered considerable interest as a possible approach for the chemoprevention of cancer. Many chemical chemopreventive agents have been shown to exert an effect on GSH homeostasis or on other elements of GSH detoxification pathways (6 -8).
Exposure of cells to a number of xenobiotic agents has been demonstrated to result in an increase in the total intracellular GSH content. In several cases (9 -16) where it has been examined, the increase in GSH has been attributed to an increase in the activity of ␥-glutamylcysteine synthetase (GCS, EC 6.3.2.2), the rate-limiting enzyme in its de novo synthesis (17). The GCS holoenzyme is a heterodimer that can be dissociated under nondenaturing conditions into light (GCS l ) and heavy (GCS h ) subunits of 28,000 Da and 73,000 Da, respectively (18). Catalytic activity resides entirely with the heavy subunit, but it has recently been demonstrated that the kinetic properties of the heavy subunit can be profoundly influenced by association with the light, or regulatory, subunit and the redox state of a single disulfide linkage between the two subunits (19,20). The heavy and light subunits are encoded for by two distinct genes located on chromosomes 1 and 6, respectively (21)(22)(23). The cDNAs for each subunit of human GCS has been cloned and sequenced (24,25), as has the 5Ј-flanking sequence of the heavy subunit gene (26). Northern analyses reveal that xenobiotic-induced GCS enzyme activity is frequently accompanied by an increase in the steady-state levels of GCS h -specific mRNA transcripts (10,(12)(13)(14)16). Similar analyses for the light subunit have not yet been reported as cloning of the corresponding cDNA has only recently been completed. Exposure to the antioxidant butylated hydroxyanisole (16), tert-butyl hydroquinone (14), and methyl mercury (10) results in transcriptional up-regulation of the heavy subunit gene, suggesting that GCS h gene expression in response to these toxic insults is regulated via as yet unidentified specific cis-and transacting factors.
Since several treatments that induce expression of key phase II detoxifying enzymes also result in elevated GCS activity as well as increased intracellular GSH levels, we hypothesized that the transcriptional regulation of the GCS h subunit gene and genes in the phase II battery are mediated by common regulatory elements. Regulation of several phase II enzymes in response to a wide array of chemically distinct classes of agents has been demonstrated to be mediated, at least in part, by the presence of an antioxidant response element (ARE) within the 5Ј-flanking region of the gene (27)(28)(29)(30)(31). Recent evidence suggests that other cis-acting elements might also participate in concert with the ARE in transcriptional regulation of these genes (30). We identified numerous potential response elements and/or enhancer sequences, including a putative ARE, in the 5Ј-flanking sequence of the human GCS h gene (26). However, the involvement of any specific elements in constitutive or induced expression of the GCS h gene has not been established by functional analyses.
In the present study, we utilized a deletion mutagenesis analysis strategy to identify those sequences that modulate constitutive expression of the human GCS h gene and to begin to decipher mechanisms involved in the induced expression of the gene. To this end, HepG2 cells transfected with a series of deletion mutant/reporter fusion genes were exposed to ␤-naphthoflavone, an inducing agent which activates phase II gene expression via AREs (31,32) and which we demonstrate also induces expression of both the endogenous heavy and light subunit genes. The results indicate that both constitutive and induced expression of the heavy subunit gene is mediated via a distal consensus ARE sequence.

EXPERIMENTAL PROCEDURES
GCS h Genomic DNA and Sequencing-Genomic DNA containing the GCS heavy subunit gene was isolated by screening a human foreskin fibroblast P1 library (Genome Systems) as described previously (26). Sequencing was performed using Sequenase (U. S. Biochemical Corp.) and synthetic oligonucleotide (20-mers) primers corresponding to internal sequences. All sequence information was verified by multiple bidirectional sequencing reactions.
Recombinant Plasmids-Recombinant expression vectors were created by cloning restriction fragments isolated from the 5Ј-flanking sequence of the GCS h gene into pGL3-Basic (Promega) for determination of promoter activity or by introducing GCS h DNA or synthetic oligonucleotides into pT81 (ATCC 37584) for determination of enhancer activity. A 4.2-kb genomic DNA fragment was isolated from the 5Јflanking region of GCS h by HindIII/NcoI restriction digestion and cloned into the HindIII/NcoI sites of pGL3-Basic creating the recombinant plasmid Ϫ3802/GCS h 5Ј-luc (Fig. 3). This plasmid was subjected to digestion with additional restriction enzymes to generate a series of deletion mutants as detailed in Fig. 3. The enhancer/reporter transgene, Ϫ3491:Ϫ2438/GCS h 5Ј-luc, was created by cloning an ϳ1-kb PpuMI fragment (Ϫ3491:Ϫ2438) containing AREs 2, 3, and 4 into pT81. Digestion of Ϫ3491:Ϫ2438/GCS h 5Ј-luc with SacI generated two additional constructs, containing either ARE2 (Ϫ2754:Ϫ2438) or ARE3 and ARE4 (Ϫ3491:Ϫ2755). In addition, a series of enhancer/reporter transgenes was created by cloning specific oligonucleotides (Fig. 5) corresponding to ARE1, ARE3, ARE4, ARE3,4, NF-B, and NQO1hARE sequences into pT81. To prepare the oligonucleotides, complementary oligonucleotides were synthesized, annealed, phosphorylated, and then cloned into pT81. Oligonucleotides were dissolved in 1 ϫ Sequenase reaction buffer and annealed by heating to 90 -100°C, cooled slowly to 35°C, extracted once with phenol/chloroform/isoamyl alcohol and once with chloroform/isoamyl alcohol, ethanol-precipitated, and finally resuspended in water. These double-stranded fragments were then phosphorylated, extracted, and precipitated as before and resuspended in water and ligated into pT81.
Cell Culture and Transfection-HepG2 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 50 g/ml gentamicin (complete medium). Cells were transfected with recombinant plasmids using a standard calcium phosphate-glycerol shock procedure. HepG2 cells were plated at 1 ϫ 10 6 cells/35-mm dish on day 0. On day 1, medium was replaced with fresh complete medium. Two to four hours later the cells were transfected by the addition of appropriate DNA expression vectors. Equimolar concentrations of plasmid DNA were used to compensate for variations in plasmid size. To correct for transfection efficiency, 1.25 g of the reporter plasmid pCMV␤ (33) containing the lacZ gene encoding ␤-galactosidase under the control of the human cytomegalovirus immediate early promoter/enhancer was cotransfected with each recombinant plasmid. Four hours after addition of DNA, cells were shocked with media containing 10% glycerol for 3 min at room temperature and then maintained at 37°C for an additional 24 h. At the conclusion of this incubation period, the medium was once again replaced with fresh complete medium containing Me 2 SO (0.1%) or 10 M ␤-NF (Sigma) dissolved at 1000 ϫ in Me 2 SO. Sixteen hours later cells were harvested and prepared for determination of luciferase, ␤-galactosidase activity, and protein content. For cell harvest, transfectants were washed twice with phosphate-buffered saline (Mg 2ϩ -and Ca 2ϩ -free) and incubated at room temperature for 15 min in 250 l of reporter/lysis buffer (Promega). Cells were then scraped from the plates and the resulting lysates spun at top speed in a microcentrifuge for 2 min at 4°C. The resultant supernatants were transferred to Eppendorf tubes and stored on ice pending assay.
Biochemical Assays-Total (oxidized and reduced) glutathione content was determined using the technique described by Tietze (34) as modified by Bump et al. (35).
GCS activity was determined as described previously using a modification (36) of a high performance liquid chromatography technique described by Nardi (37). The rate of ␥-glutamylcysteine formation was quantitated by comparison to ␥-glutamylcysteine standards and normalized on the basis of protein concentration.
␤-Galactosidase activity was quantified as described by Rosenthal (38). Briefly, this assay monitors cleavage of o-nitrophenyl-␤-D-galactopyranoside and yields ␤-galactosidase units as (A 420 ϫ 380)/t where t ϭ time in min at 37°C and 380 is a conversion factor to convert absorbance to mol of o-nitrophenyl-␤-D-galactopyranoside.
To initiate the luciferase assay, cell lysate (1-5 l) was added to reaction buffer (14 mM MgCl 2 , 14 mM glycylglycine, 0.1 mg/ml bovine serum albumin, 18 mg/ml ATP, pH 7.8), vortexed, and placed in an Analytical Luminescence Laboratory luminometer (400-l total volume). Following injection of 100 l of luciferin (4 mg/ml in 10 mM Na 2 CO 3 , pH 6.0), luminescence was recorded as relative light units. Luciferase activity was normalized for transfection efficiency on the basis of ␤-galactosidase activity and protein content of the lysate yielding a final value of (relative light units/␤-galactosidase)/g protein.
Protein content was determined by the Bradford method (39) using bovine serum albumin as a standard.
Northern Analysis and RNase Protection-Total cellular RNA was isolated using TRI Reagent (Molecular Research Center, Inc.) according to the procedure recommended by the manufacturer. For Northern analysis 25 g of total RNA was size-fractionated in formaldehydeagarose gels, transferred to a charged nylon membrane (Hybond N ϩ , Amersham Corp.), and hybridized with a 764-bp or a 736-bp radiolabeled probe specific for GCS h or GCS l transcripts, respectively (24,25). A radiolabeled probe specific for 36B4 (40) was used to normalize for RNA loading. RNA quantitation was accomplished using the RPAII (Ambion) ribonuclease protection assay. To generate single-stranded radiolabeled probes, the 764-bp PstI fragment isolated from the GCS h cDNA was cloned into the pAlter vector (Promega) and digested with NcoI. The GCS l -specific probe was prepared by cloning a 305-bp HindII fragment from GCS l cDNA into pSKII Bluescript (Stratagene) followed by digestion with DdeI. Using T7 RNA polymerase and [ 32 P]UTP, these vectors yield 301-and 210-bp radiolabeled run-off transcripts and 250and 179-bp protected fragments for GCS h and GCS l , respectively. To normalize for RNA content glyceraldehyde-3-phosphate dehydrogenase was used as an internal standard. pTRI-GAPDH-Human (Ambion) yields a 404-bp transcript and a 316-bp protected fragment. Following hybridization with total cellular RNA and incubation with RNase, fragments were separated on a 5% acrylamide, 8 M urea gel, visualized, and quantitated by using the PhosphorImager system (Molecular Dynamics).

RESULTS
Induction of GCS Activity by Exposure to ␤-NF-To examine regulation of GCS h induction, we first elected to investigate the effect of exposure of HepG2 hepatocarcinoma cells to ␤-NF, a potent inducer of several phase II enzymes and an agent that exerts its effects, at least in part, through activation of AREs. The effect of ␤-NF exposure on GCS activity in exposed HepG2 cells is illustrated in Fig. 1A. There was a gradual decrease in GCS activity over the 24-h incubation period in untreated and Me 2 SO-treated HepG2 cells. In contrast, exposure to 10 M ␤-NF resulted in a progressive increase in GCS activity, culminating in a moderate, although significant (p Ͻ 0.05), increase in enzyme activity by 24 h after addition of ␤-NF to the culture medium (Fig. 1A). Induction of GCS expression occurred sooner and was greater following exposure to 25 M ␤-NF.
These changes in GCS expression were accompanied by a progressive increase in intracellular GSH levels at each ␤-NF dose over the 24-h observation period (Fig. 1B). By 24 h GSH levels had increased significantly, reaching in excess of 2-and 4-fold over controls for 10 and 25 M exposures, respectively. In the case of 25 M exposure, a significant increase was evident as early as 12 h after addition of ␤-NF.
Increase in Steady-state mRNA Levels for Heavy and Light Subunits-Exposure of HepG2 cells to increasing concentrations of ␤-NF also resulted in a dose-dependent increase in steady-state mRNA levels for both subunits of the GCS holoenzyme (Fig. 2, A and B). In all cases, the two heavy subunit transcripts increased proportionally ( Fig. 2A). Changes in the relative abundance of the transcripts for the light and heavy subunits as a function of time after addition of ␤-NF or Me 2 SO was quantitated by a ribonuclease protection assay engineered to permit quantitation of both transcripts simultaneously. A representative radiograph is shown in Fig. 2C. Quantitation by nuclease protection assay was particularly important in the case of the light subunit transcripts that are typically present at low abundance. Quantitative and temporal changes in transcript abundance in ␤-NF-treated cells are illustrated in Fig. 2, D and E. These data confirm the dose-dependent increase in mRNA levels observed by Northern analysis. The ␤-NF-induced increase in steady-state mRNA levels for the GCS h and GCS l subunits displayed similar kinetic patterns, reaching a maximum at approximately 12 h after initiation of exposure. Interestingly, the increase in GCS l transcripts levels was consistently higher than that observed in the case of the GCS h mRNA.
Regulation of Constitutive Expression-We recently reported (26) the cloning and partial sequence (Ϫ1460:ϩ547) of a genomic clone, designated P1-GCS5Ј, containing approximately 4.7 kb of 5Ј-flanking sequence and 0.5 kb of exon I of the human GCS catalytic subunit gene. A 4.2-kb fragment from this clone, spanning the sequence Ϫ3802 to ϩ465, was cloned into the pGL3-luciferase reporter vector to create the GCS h / luciferase fusion gene, Ϫ3802 GCS h 5Ј-luc (Fig. 3). A series of deletion mutants was then generated from this parent chimeric gene and used to identify functional response elements directing constitutive and ␤-NF-induced expression of the GCS h gene. The various reporter constructs shown in Fig. 3 were transfected into HepG2 cells and luciferase expression examined 48 h later. Transfection of Ϫ3802/GCS h 5Ј-luc increased the luciferase expression approximately 35-fold relative to the expression detected in cells transfected with pGL3-luciferase alone. Progressive deletions from Ϫ3802 to Ϫ510 resulted in significant reductions in luciferase expression when compared with that produced by Ϫ3802/GCS h 5Ј-luc. However, the differences in luciferase expression among these three intermediate constructs were not statistically significant (p Ͼ 0.05) from each other. Transfection of Ϫ202/GCS h 5Ј-luc resulted in significantly increased luciferase activity, attaining levels comparable with those observed following transfection with the fulllength fusion gene. Deletion of sequences between Ϫ202 and ϩ22 resulted in a significant reduction in luciferase expression, whereas further deletion to ϩ358 completely eliminated promoter activity. Collectively these results suggest that the fragment between Ϫ202 and ϩ22 contains a positive regulatory sequence(s) capable of directing constitutive expression of the gene in HepG2 cells; negative regulatory sequences exist between Ϫ202 and Ϫ814, and one or more positive regulators exists between Ϫ2752 and Ϫ3802. Further analyses revealed that the enhanced activity associated with inclusion of sequences from Ϫ202 to ϩ22 corresponded to the presence of the TATA sequence previously identified in this region of the promoter (data not shown).
Regulation of ␤-NF-induced GCS h Expression-To determine whether the increase in steady-state mRNA levels observed following ␤-NF exposure was the result of increased transcription and to identify response elements that might mediate ␤-NF-induced GCS h gene expression, HepG2 cells were transiently transfected with the GCS h promoter/reporter constructs and then exposed to 10 M ␤-NF for 16 h. As shown in Fig. 3, ␤-NF failed to increase luciferase expression in cells transfected with any of the recombinant reporter genes other than the full-length Ϫ3802/GCS h 5Ј-luc. The magnitude of induction was approximately 2-3-fold, comparable with the magnitude of increase in endogenous GCS h mRNA expression following similar ␤-NF exposure. Hence, the response element(s) mediating GCS h gene induction in response to ␤-NF is present between Ϫ2752 and Ϫ3802.

Potential Response Elements Present in the Distal 5Ј-Flanking Region of the GCS h Gene-Transfection
experiments therefore suggest that the DNA sequence from Ϫ2752:Ϫ3802 contains elements that influence both constitutive and ␤-NFinduced expression of GCS h . This region is beyond the 5Ј-distal end of the sequence we had previously reported, and therefore we extended our original sequencing to include an additional 2342 bp of this 5Ј-distal fragment in order to identify potential enhancer sequences. The complete sequence of Ϫ3802:ϩ547/ GCS h 5Ј is shown in Fig. 4. In addition to the potential regulatory sequences identified previously, the distal genomic fragment includes a consensus NF-B binding site, a consensus AP-1 site, Sp-1 sites, and three additional potential AREs.
In order to approximate the position of elements or enhancers that contribute to the constitutive and induced expression of the heavy subunit gene, a series of enhancer/reporter constructs were developed by cloning various restriction fragments or synthetic oligomers (Fig. 5) into the reporter vector, pT81, in which luciferase expression is under the control of the herpes simplex thymidine kinase (tk) minimal promoter. Fig. 6A illustrates the structure of the individual enhancer vectors and their effect on constitutive luciferase expression. The fragment Ϫ3491:Ϫ2438, representing the majority of the distal region of GCS h 5Ј responsible for increased constitutive expression, increased luciferase expression 5-fold relative to that observed in pT81 transfected cells. When this 1-kb fragment was subdivided by SacI digestion, the fragment corresponding to the distal 0.7 kb (Ϫ3491:Ϫ2755) maintained enhanced luciferase expression, whereas the proximal 0.3-kb fragment (Ϫ2754:Ϫ2438) failed to do so.
As shown in Figs. 3 and 4, Ϫ3491:Ϫ2755/GCS h 5Ј contains two putative AREs (3 and 4). A third (ARE2) is present in the Ϫ2754:Ϫ2438/GCS h 5Ј fragment. Individually, ARE3 and ARE4 each significantly enhance pT81 luciferase expression, attaining levels of expression (ϳ20-fold) comparable with that produced by the ARE-positive control vector, NQO1hARE-tk-luc (32), containing the ARE from the human NQO1 gene (Fig. 6A). When cloned into pT81 in their normal tandem arrangement (GCS h ARE3,4), ARE3 and ARE4 increase luciferase expression in excess of 100-fold. In contrast, ARE1, the putative ARE identified previously on the basis of sequence similarity, did not function as an enhancer in this system. Although an oligomer corresponding to ARE2 was not cloned into pT81, the 0.3-kb Ϫ2754:Ϫ2438/GCS h 5Ј fragment, containing ARE2, did FIG. 2. Increased GCS h and GCS l mRNA levels following treatment with ␤-NF. HepG2 cells were treated as in Fig. 1. Sixteen hours after the addition of Me 2 SO (DMSO) or the concentrations of ␤-NF indicated total RNA was harvested, size-fractionated on a formaldehyde agarose gel, transferred to charged nylon membranes (Hybond N ϩ ), and hybridized with radiolabeled probes specific for GCS h (A) or GCS l (B) transcripts. In order to quantitate temporal changes in GCS subunit mRNA expression, RNA was harvested at various times after addition of ␤-NF. Changes in GCS h and GCS l levels were monitored by a ribonuclease protection assay as described under "Experimental Procedures" (C). Quantitation of protected species corresponding to GCS h (D) and GCS l (E) mRNA at various times after ␤-NF addition was determined using a PhosphorImager and ImageQuant software. not increase luciferase activity, suggesting that ARE2 was likewise not functional as an enhancer of constitutive expression. The consensus NF-B sequence also failed to influence luciferase expression.
When a similar series of experiments was conducted with transfected HepG2 cells exposed to 10 M ␤-NF for 16 h, GCS h ARE4, GCS h ARE3,4 (containing the tandem combination of ARE3 and ARE4), and those larger fragments containing them directed increased luciferase expression in treated cells (Fig.  6B). Luciferase expression in cells transfected with NQO1hARE-tk-luc was increased approximately 3.5-fold relative to that in untreated cells. A similar level of induction was detected in cells transfected with ARE4-tk-luc. However, in the presence of ARE3, which did not itself respond to ␤-NF induction, the ␤-NF inducibility of ARE4 was diminished slightly. Nevertheless, the induction of luciferase expression in ␤-NFtreated cells transfected with any vector containing the combination of ARE3 and ARE4 was similar in magnitude to 1) the increase in endogenous GCS h transcripts in HepG2 cells exposed to the same dose of ␤-NF, and 2) to the ␤-NF-induced increase in luciferase expression in cells transfected with the full-length Ϫ3802/GCS h 5Ј-luc. Other potential elements, including NF-B, ARE1, and ARE2, failed to direct a ␤-NFinduced increase in luciferase expression.
Of the four putative ARE sequences identified in the 5Ј-flanking sequence of the GCS h gene, only ARE4 influenced both constitutive and induced gene expression in a fashion typical of other functional AREs. In contrast, ARE3 enhanced constitutive expression only, although it did modulate both the constitutive and inducible properties of ARE4. In order to confirm that ARE4 functions as a true ARE, a synthetic oligonucleotide containing a single base mutation in the consensus core ARE sequence of ARE4 (converting the sequence from 5Ј-GTGACT-CAGC-3Ј to 5Ј-GGGACTCAGC-3Ј; Fig. 5) was synthesized and cloned into pT81. The T in this position has been shown to be a required element for the maintenance of both the constitutive and the inducible properties of AREs. When this vector (GCS h ARE4m) was transfected into HepG2 cells, constitutive expression was significantly reduced (Fig. 6A), and ␤-NF induction was eliminated (Fig. 6B) in cells transfected with the mutated ARE4 fusion gene.
To determine whether ARE4 functions in the same capacity when present at its distal location in the genomic sequence of the GCS h gene 3.1 kb upstream of the transcription start site, the same single-base mutation was introduced into the fulllength fusion reporter gene Ϫ3802/GCS h 5Ј-luc by site-directed mutagenesis. The mutation of this single base within this 3.8-kb GCS h fragment likewise reduced constitutive expression and eliminated induction in response to ␤-NF (Ϫ3802mA4/ GCS h 5Ј-luc in Fig. 3). Ϫ3802mA4/GCS h 5Ј-luc, a single base mutation (T 3 G) was introduced into ARE4 of Ϫ3802/GCS h 5Ј-luc (see Fig. 5) by site-directed mutagenesis. These 5Ј-deletion mutants were co-transfected along with the plasmid pCMV-␤ into HepG2 cells by calcium phosphate precipitation. Twenty-four hours after transfection the culture medium was replaced with fresh medium containing either Me 2 SO (control) or 10 M ␤-NF. Cells were harvested 16 h later and supernatants prepared for luciferase, ␤-galactosidase, and protein assays. Values represent the mean Ϯ S.E. luciferase units normalized for ␤-galactosidase activity and protein content.

DISCUSSION
Diverse chemical compounds, including Michael reaction acceptors, quinones, diphenols, peroxides, isothiocyanates, mercaptans, heavy metals, arsenicals, and certain planar aromatic and metabolizable polycyclic aromatic hydrocarbons, induce the expression of phase II detoxifying enzymes (8,30). The induction of several genes in the phase II battery, including the rat (31) and mouse (28) glutathione S-transferase-Ya genes, rat (41) and human (42) NAD(P)H quinone oxidoreductase genes, and the rat GST-P gene (43,44), is mediated through an antioxidant responsive element (or its functional equivalent, an electrophile responsive element, EpRE), in the promoter region of the gene (27). It is hypothesized that these inducing agents share in common the ability to generate reactive oxygen intermediates directly or via redox cycling (28,31) and result in activation of transcription factors, perhaps through alterations in their redox status. While this mechanism is favored by some, Talalay and colleagues (8) point out that all of these inducers are capable of generating electrophilic intermediates capable of reacting with sulfhydryl groups. They therefore suggest that a better designation for these responsive elements would be EpREs to reflect the nature of the active intermediates.
Interestingly, exposure to several of the agents that induce expression of the ARE-containing phase II genes results in an elevation in intracellular GSH levels as well. These increases are frequently coupled to an increase in the activity of GCS, which led us to hypothesize that the increased GSH levels might result from transcriptional up-regulation of GCS expression mediated by ARE-like elements in the 5Ј-flanking region of the GCS catalytic subunit and/or light subunit genes. Previous cloning efforts identified an ARE-like sequence in the promoter of the GCS h gene (26), and three additional potential ARE Functional AREs display two cardinal properties; they enhance basal expression and are inducible by phase II enzyme inducers (27). Using these criteria, the current study provides the first direct evidence that a functional ARE is present in the 5Ј-flanking region of the human GCS h gene. Despite the fact that four ARE-like sequences were identified on the basis of sequence similarity to the ARE consensus core sequence, only two, ARE3 and ARE4, contributed to basal expression, each augmenting expression of luciferase in the presence of the tk minimal promoter by ϳ20-fold. Interestingly, when cloned into pT81 in tandem in their native configuration, separated only by 34 bp, these two sequences enhance basal expression greater than 100-fold. However, this potent enhancer activity is dampened by the inclusion of additional DNA, such that in the full-length GCS h 5Ј sequence, presence of the native context ARE3/ARE4 tandem only doubles basal expression (i.e. Ϫ3802/ GCS h 5Ј-luc versus Ϫ2752/GCS h 5Ј-luc).
Both ARE3 and ARE4 match the consensus 11-nucleotide ARE motif described by Pickett et al. (31,45), yet only the ARE4 oligonucleotide directed enhanced expression following exposure to ␤-NF. Since these two enhancers only differ by two nucleotides, in the region of the core (NNN) previously demonstrated to be of no consequence in influencing either basal or ␤-NF induction (31,45), it is not immediately obvious why ARE3 failed to respond to ␤-NF induction. Analyses of functional AREs in other genes have revealed that the organization of these enhancers may be more complex than the simple linear arrangement of the "core" nucleotides, often involving proximal nucleotides flanking either side of the core motif itself. Furthermore, functionality of specific AREs may be dependent on the spatial arrangement of these multiple partners. As illustrated in Fig. 7, AREs in the rat GSTP (43) and GST-Ya (31) genes, the murine GST-Ya (28) gene, and rat (41) and human (42) NQO1 genes include an adjacent pair of TRE or TRE-like elements. Favreau and Pickett (45) divided the 31-base pair rat NQO1 gene ARE into three regions as follows: proximal and distal half-sites of a 13-base pair palindromic sequence and a 3Ј-flanking region containing 4 adenines. Mutations in these various regions influenced basal and induced expression in different ways supporting the conclusion that the two half-sites are not functionally equivalent. While only the proximal halfsite was required for binding of nucleoproteins to the ARE, both half-sites contribute to full basal gene expression. However, only three nucleotides in the proximal site were absolutely FIG. 5. Nucleotide sequence of oligonucleotides used to examine enhancer activity. Synthetic oligomers corresponding to the complementary strands for the positive control ARE, NQO1hARE, and the potential response elements in Ϫ3802/GCS h 5Ј-luc were annealed, phosphorylated, and ligated into the tk-luciferase enhancer vector, pT81. A single base mutation (lowercase letters) was introduced into the 76-mer GCS h ARE3,4 creating a SmaI restriction site which was then used to generate ARE3-tk-luc and ARE4-tk-luc from ARE3,4-tk-luc. GCS h ARE4m was created by introducing a single base change (lowercase) in the corresponding oligomers, converting the consensus ARE sequence 5Ј-GTGACTCAGC-3Ј to 5Ј-GGGACTCAGC-3Ј. required for maximal basal and induced transcription (45). Considering this complexity, it is possible that selection of the 20 base pairs used to create the ARE3 oligonucleotide might have omitted additional neighboring elements that are required for inducibility in situ. However, no evidence for comparable elements around ARE3 was found. Even though ARE3 doesn't appear to function as an independent ARE and despite the fact that mutation of ARE4 within the ARE3/ARE4 tandem eliminated basal and ␤-NF-induced expression by Ϫ3802/ GCS h 5Ј, the current data are insufficient to preclude the possibility that an interaction between ARE3 and ARE4 is required for regulation of GCS h gene expression in situ. Further mutational analysis of the ARE3/ARE4 tandem is being conducted to clarify the relationship between these two sequences.
ARE4 is the only element to fulfill the functional requirements of an ARE in terms of both basal and inducible expression. Identification of ARE4 as a functional element was confirmed by mutational analysis in which conversion of the sequence 5Ј-GTGACTCAGCA-3Ј to 5Ј-GGGACTCAGCA-3Ј in the ARE4 oligonucleotide and in Ϫ3802/GCS h 5Ј-luc effectively eliminated constitutive and ␤-NF-induced activity in both cases, as demonstrated in the functional characterization of other AREs (31,45). Abolishment of the basal and induced expression directed by Ϫ3802/GCS h -5Ј by introduction of this mutation provides strong evidence that this sequence is involved in the regulation of expression of the endogenous GCS h subunit gene in situ. Hence, all or part of ARE4 is necessary for basal and ␤-NF-induced expression of the gene.
Although ARE4 contains a consensus ARE core sequence, this element also includes a canonical TRE motif (5Ј-TGA(C/ G)T(C/A)A-3Ј), an arrangement also found in AREs from the human NQO1 (42) and murine heme oxygenase (46) genes (Fig.  7). Since the mutation we introduced in ARE4 corrupted both the ARE and TRE sequences, we have not as yet distinguished experimentally which of these responsive elements is involved in the expression of the GCS h gene. Evidence from other ARE analyses favors the involvement of the ARE core, however. In analyzing the elemental requirements of the ARE in the human NQO1 gene, Xie et al. (32) reported that the sequence 5Ј-GATGAGTCAGCC-3Ј, containing a single TRE motif (in boldface), but not an ARE failed to increase expression in transfected HepG2 cells in the response to ␤-NF. Similarly, Fig. 5 and the three GCS h 5Ј-restriction fragments Ϫ3491:Ϫ2438, Ϫ3491:Ϫ2755, and Ϫ2754:Ϫ2438 were cloned into pT81 and co-transfected into HepG2 cells along with the plasmid pCMV-␤. Sixteen hours after addition of medium containing either 0.1% Me 2 SO or 10 M ␤-NF, cells were harvested and supernatants prepared for luciferase, ␤-galactosidase, and protein determinations. Fold induction (B) was calculated as the ratio of luciferase expression in ␤-NF-treated cells to luciferase expression in control cells after correction for ␤-galactosidase activity and protein content. The results are the mean Ϯ S.E. of 3-8 determinations. RLU, relative light units. Prestera et al. (46) demonstrated that maintenance of the ARE consensus was required for the induction of murine HO-1 gene expression in response to a battery of phase II inducers. Mutation of the TRE sequences had no influence on inducibility, provided the ARE core sequence was retained. In a recent characterization of the ARE sequence in the rat NQO1 gene, Favreau and Pickett (46) mutated (underlined) the existing sequence, 5Ј-GTGACTTGGCA-3Ј, to one containing a TRE motif, 5Ј-GTGACTCAGCA-3Ј. When the mutant oligonucleotide was used in electrophoretic band shift assays, unique nuclear proteins recognizing either the ARE or the TRE sequence were detected in HepG2 nuclear extracts. Since this mutated sequence matches the core sequence of ARE4 with the exception of an A to G substitution at the 3Ј-terminal nucleotide, a substitution that does not influence binding or ARE function (45), these data provide evidence that proteins capable of binding the ARE4 sequence are present in HepG2 cells. In preliminary gel shift experiments, 2 binding activity to a probe corresponding to ARE4 is elevated in nuclear extracts from ␤-NF-treated HepG2 cells; however, an increase in TRE-binding activity is not detected. This ARE4 binding activity is effectively competed by unlabeled ARE4 probe but not by a probe containing a consensus TRE sequence, providing further evidence to support the importance of the ARE sequence in mediating basal and ␤-NF-induced expression of GCS h . Definitive identification requires completion of additional mutational analyses currently in progress.

FIG. 6. Constitutive expression (A) and ␤-NF induction (B) of GCS h -tk-luc recombinant plasmids transfected into HepG2 cells. The oligonucleotides shown in
Finally, the location of a functional ARE in a position 3.1 kb distal to the transcription start site is atypical by comparison to the position of most other AREs, which have been typically identified proximal to the transcription origin in phase II enzymes. However, Prestera et al. (46) recently identified a functional ARE 4.1 kb upstream in the human heme oxygenase-1 gene, capable of regulating constitutive and xenobiotic induced expression of this important stress response gene. Collectively these observations in two different genes indicate that AREs can exert an effect on gene transcription over a long expanse of intervening sequence.
In summary, constitutive and ␤-NF-induced expression of the GCS heavy subunit gene, at least in part, is mediated by a distal sequence including a consensus ARE motif containing an embedded TRE. Experiments designed to determine specific sequences involved in GCS h expression as well as to identify additional elements that may contribute to the composite regulation of this important gene are in progress.