Generation of a Stable Antioxidant Response Element–Driven Reporter Gene Cell Line and Its Use to Show Redox-Dependent Activation of Nrf2 by Cancer Chemotherapeutic Agents

Drug-metabolizing enzymes, such as NAD(P)H:quinone oxidoreductase 1 (NQO1), aldo-keto reductase (AKR), and glutathione S-transferases (GST), along with reduced glutathione (GSH) and its biosynthetic enzymes glutamate cysteine ligase (GCL, comprising GCLC and GCLM subunits) and GSH synthase, protect cells against carcinogenic electrophiles and reactive oxygen species (ROS; refs. 1, 2). This defense can be induced in response to redox stressors, thereby allowing cells to adapt to the presence of pro-oxidants and electrophiles. It is also overexpressed in certain tumors. Induction of these genes is controlled primarily by NF-E2 p45-related factor 2 (Nrf2; refs. 3, 4), a transcription factor belonging to the family of cap “n” collar (CNC) basic-region leucine zipper (bZIP) proteins (5). Nrf2 mediates induction of detoxication and antioxidant genes that contain an antioxidant response element [ARE, 5′-(A/G)TGACNNNGC(A/G)-3′] in their promoters (2, 5); the ARE has also been referred to as the electrophile response element (EpRE). A key role for Nrf2 in controlling the ability of cells to withstand harmful environmental agents has been revealed by studies in which Nrf2 knockout mice have been shown to exhibit sensitivity to hyperoxia-induced lung injury and increased susceptibility to toxic xenobiotics, including carcinogens (6, 7).

The activity of Nrf2 is repressed by the inhibitory factor Kelch-like ECH associated protein 1 (Keap1) that facilitates degradation of Nrf2 through acting as a Cul3 substrate adaptor (8, 9). Electrophilic agents modify Keap1 and prevent it from targeting Nrf2 for degradation (10). Such inactivation of Keap1 allows Nrf2 to accumulate in the nucleus where it forms a heterodimer with other bZIP proteins and transactivates target genes, including NQO1, AKR, GST, GCLC, and GCLM (3, 11). Genetic knockout or knockdown of Keap1 also increases expression of the ARE-gene battery (8, 11).

A number of the genes that are regulated by Nrf2 have been linked to drug resistance. For example, the antioxidant GSH, the synthesis of which is primarily dictated by the activity of GCLC and GCLM, has been implicated in resistance of tumor cells to several chemotherapeutic agents, including cisplatin and melphalan (12, 13). Similarly, overexpression of GST isoenzymes, which catalyze the conjugation of GSH with electrophilic compounds (14), has been reported in a large number of tumor types (1) and implicated in resistance towards chemotherapeutic agents (12, 13, 15).

Because the levels of drug-metabolizing enzymes help determine the sensitivity of tumor cells to anticancer agents, it is important to understand how their genes are regulated. Based on the evidence that Nrf2 regulates the expression of ARE-driven antioxidant and detoxication genes in response to Michael reaction acceptors, redox-cycling hydroquinones, quinones, and other electrophiles (16), we postulated that its activity might be modulated by anticancer drugs. For example, we expected that the activity of Nrf2 could be increased by alkylating agents and also possibly by redox-cycling compounds. Based on the facts that chlorambucil, melphalan, carmustine [1,3-bis(2-chloroethyl)-1-nitrosourea (BCNU)], cyclophosphamide metabolites, and thiotepa are electrophilic agents that can be conjugated with GSH (15), and that many agonists of Nrf2 are thiol-active electrophiles (17), it might be anticipated that chemotherapeutic alkylating agents would activate the ARE-gene battery. In the present article, we have tested this hypothesis. We have also examined whether doxorubicin and etoposide can induce ARE-driven gene expression because they are capable of redox-cycling, thereby producing ROS that might activate Nrf2. Furthermore, the anticancer agents cisplatin, methotrexate, mitoxantrone, and paclitaxel, which are neither alkylating agents nor redox-cycling compounds, were also included in the study.

We now describe the generation of a sensitive, stable ARE reporter cell line AREc32, which contains a luciferase gene under the transcriptional control of multiple tandemly arrayed copies of the minimal functional cis-element found in the promoters of rat GSTA2 (18) and mouse gsta1 (19). To evaluate the usefulness of the reporter lines, we challenged it with the redox-cycling agent tert-butylhydroquinone (t-BHQ), produced as a cytochrome P450 metabolite of the phenolic antioxidant butylated hydroxyanisole (BHA; ref. 20), and the isothiocyanate sulforaphane (SUL), as both of these compounds induce ARE-driven gene expression (16). t-BHQ was used in the original study to define monofunctional inducers (21) and also to define the ARE (18). Induction of the luciferase reporter gene in the stable AREc32 reporter cells by t-BHQ and SUL was found to be sensitive to changes in cellular redox status and is regulated specifically by Nrf2. Importantly, we also report that anticancer drugs are capable of inducing reporter gene expression in AREc32 cells, and that depletion of cellular GSH with l-buthionine-S,R-sulfoximine (BSO) enhances induction of ARE-driven luciferase by certain anticancer agents. Our study provides evidence that Nrf2 is involved in the responsiveness of mammary tumor cells to chemotherapeutic agents in a redox-dependent fashion.

Chemicals and cell culture. Unless otherwise stated, all chemicals were purchased from Sigma-Aldrich Co. Ltd. (Dorset, United Kingdom). d,l-sulforaphane was obtained from LKT Laboratories, Inc. (St. Paul, MN). HepG2 (human hepatoblastoma), MCF7 (human breast carcinoma), Hepa1c1c7 (mouse hepatoma), and Chinese hamster ovarian carcinoma (CHO) cell lines were obtained from the cell services of Cancer Research UK (London, United Kingdom). The growth medium for MCF7 and HepG2 cells was DMEM with glutamax supplemented with 10% fetal bovine serum and antibiotics. For Hepa1c1c7 cells, the growth medium was supplemented with 1% nonessential amino acids and 2.5 μg/mL bovine insulin. For CHO cells, the growth medium was supplemented with 16 μmol/L thymidine and 100 μmol/L hypoxanthine. All cells were cultured at 37°C, in 95% air and 5% CO₂, and passaged every 3 to 4 days. All media supplements for cell culture were purchased from Life Technologies, Inc. Ltd. (Paisley, United Kingdom).

Reporter plasmids and expression constructs. The ARE-luciferase reporter plasmids were generated using the pGL3-promoter vector (Promega UK, Southampton, United Kingdom) containing an SV40 promoter upstream of the firefly luciferase gene. The sequences of the inserts used in the different plasmids are summarized in Table 1. They differ in the number of copies of ARE sequences that have been inserted, in head-to-tail orientation, through NheI and XhoI restriction sites upstream of the promoter-luc⁺ transcriptional unit. Five plasmids were made containing either one, two, four, six, or eight copies of the ARE (5′-GTGACAAAGCA-3′, with the minimal functional sequence underlined) present in both rat GSTA2 and mouse gsta1; these were called pGL-nxARE (where n is the number of concatenated elements). A linker with the sequence of 5′-CCC-3′ and 5′-GGG-3′ on the opposite strand was placed between individual cis-elements. In addition, a plasmid named pGL-GSTA2.41bp-ARE was generated that represents the 41-bp 5′-GAGCTTGGAAATGGCATTGCTAATGGTGACAAAGCAACTTT-3′ between nucleotides −682 and −722 in the rat GSTA2 gene promoter (with the minimal functional enhancer shown underlined), driving the luciferase reporter gene. In mouse gsta1, this sequence is 5′-TAGCTTGGAAATGACATTGCTAATGGTGACAAAGCAACTG-3′ (14, 19) and was originally called the EpRE. The oligonucleotides were synthesized by MWG-BIOTECH AG (Eberserg, Germany). After the plasmids were generated, the DNA sequence of the inserts was verified.

The pHyg-EF-hNrf2 expression vector encoding a green fluorescent protein (GFP)–tagged human Nrf2, was a gift from Prof. Masayuki Yamamoto (Institute of Basic Medical Sciences, University of Tsukuba, Tsukuba, Japan). The plasmid pEGFP-N1, a GFP expression vector employed as a negative control, was obtained from BD Clontech UK (Hampshire, United Kingdom).

Transient transfection and analysis of luciferase reporter gene activity. The Dual-luciferase Reporter Assay System (Promega) was used to examine reporter gene activity in transiently transfected cells. Briefly, cells were seeded at a density of 2 × 10⁵ per well in 24-well plates and grown in the appropriate medium. After incubation overnight, the cells were transiently transfected with various ARE-luciferase reporter plasmids. The plasmid pRL-TK, encoding Renilla luciferase, was used to control for transfection efficiency. Transfections were done using LipofectAMINE 2000 reagent (Life Technologies, Inc. Ltd., Coventry, United Kingdom) according to the manufacturer's instructions. Following transfection, the culture medium was replaced 24 hours later with fresh growth medium containing 50 μmol/L t-BHQ (in a solution giving a final concentration of 0.1% v/v DMSO), which was prepared immediately before each experiment. For control experiments, vehicle alone (0.1% v/v DMSO) was added to the growth medium. Cells were left for 24 hours to respond to xenobiotics before being harvested, and the firefly and Renilla luciferase activities in their lysates were measured using a luminometer (Turner Designs Model TD-20/20, Promega) following addition of Luciferase Assay Reagent II (Promega). After quenching the reaction, the Renilla luciferase reaction was initiated by adding Stop and Glo Reagent (Promega). The relative luciferase activity was calculated by normalizing firefly luciferase activity to that of Renilla luciferase.

Generation of stable ARE-driven reporter systems. The pGL-8xARE, along with the pcDNA3.1 plasmid containing the neomycin selectable marker, was stably transfected into MCF7 cells using the calcium phosphate method. Transfected cells were selected using 0.8 mg/mL G418 in the media for 3 to 4 weeks. The G418-resistant clones were isolated and screened by measuring their basal and inducible (obtained by treatment with 50 μmol/L t-BHQ) luciferase activities as described above. Positive clones, which showed low background and high inducible luciferase activity, were passaged and maintained in growth medium containing 0.8 mg/mL G418.

Treatment of stable ARE-luciferase reporter cells with anticancer drugs. BCNU and melphalan were dissolved in acidified ethanol as 1,000× concentrated solutions. Doxorubicin, epirubicin, cyclophosphamide, methotrexate, and paclitaxel were dissolved in PBS. The other chemotherapeutic agents were prepared as 1,000× concentrated stock solutions in DMSO and were stored at −20°C until use. For treatment with these drugs, cells were seeded at a density of 1.2 × 10⁴ per well in 96-well microtiter plates in growth medium. After overnight recovery, the culture medium was replaced with fresh DMEM supplemented with antibiotics along with the particular drug of interest. An equal volume of vehicle was added to the control wells. After treatment for 24 hours, firefly luciferase activity was determined as described above.

Forced expression of hNrf2 in stable ARE-luciferase reporter cells. For transfection, AREc32 cells were seeded at 1.5 × 10⁴ per well in 100 μL growth medium in 96-well plates. After overnight recovery, the cells were transfected with between 25 and 100 ng/well of the vectors pHyg-EF-hNrf2 or pEGFP-N1 using the LipofectAMINE 2000 reagent. Following a 4-hour recovery period after transfection, the culture medium was replaced with fresh DMEM containing glutamax and 10 μmol/L t-BHQ (or DMSO alone) supplemented with antibiotics. An equal volume of DMSO was added to the control wells. Finally, firefly luciferase activity was measured after treatment with t-BHQ for 24 hours.

Knockdown of Nrf2. pRS hNrf2, a pSUPER RNA interference (RNAi) vector targeting human Nrf2, was recovered from the glycerol stocks of the SUPER RNAi library (Netherlands Cancer Institute, Amsterdam, The Netherlands). The sequence of the oligo insert in the pRS-hNrf2 used in this study was 5′-GCATTGGAGTGTCAGTATG-3′, corresponding to nucleotides 1921 to 1939 of hNrf2 cDNA (where the nucleotides are numbered with the A in the ATG initiation codon designated as +1). A pSUPER RNAi vector targeting GFP, pRS-GFP, was also obtained from the SUPER RNAi library and used as a negative control.

For transfection with pSUPER RNAi, AREc32 cells were seeded at 1.5 × 10⁴ per well in 100 μL growth medium in 96-well plates. After overnight recovery, between 25 and 100 ng/well of the pRS-hNrf2 or pRS-GFP pSUPER vectors were transfected into the cells using the LipofectAMINE 2000 reagent. Following recovery from transfection (24 hours), the culture medium was replaced with fresh DMEM containing glutamax and 10 μmol/L t-BHQ (or DMSO alone) supplemented with antibiotics. After 24 hours of treatment, firefly luciferase activity was measured. The efficacy of the RNAi to knockdown Nrf2 was confirmed by measurement of mRNA using Taqman analysis as described below.

Real-time quantitative PCR (reverse transcription-PCR). Total RNA was isolated with TRIzol and further purified with the RNeasy Mini kit (Qiagen Ltd., Germany) in accordance with the manufacturer's instructions. The A₂₆₀/A₂₈₀ ratio of total RNA used was typically ≥1.9, and its quality was assessed using the Agilent 2100 Bioanalyzer. Reverse transcription-PCR (RT-PCR) was done as described previously (22). The primers were synthesized by MWG-BIOTECH. The probes, which were labeled with a 5′ fluorescent reporter dye (6-carboxyfluorescein) and a 3′ quenching dye (6-carboxytetramethylrhodamine), were synthesized by Qiagen. Each assay was done in triplicate. The specificity of PCR amplifications from the various sets of oligonucleotide primers was examined routinely by agarose-gel electrophoresis. The results were analyzed by using AB7700 system software. The level of 18S rRNA was used as an internal standard. For measurement of cDNA corresponding to human Nrf2 mRNA, the forward primer was 5′-ACTCCCTGCAGCAAACAAGAG-3′; the reverse primer was 5′-TTTTTCTTAACATCTGGCTTCTTACTTTT-3′; and the probe was 5′-TGGCAATGTTTTCCTTGTTCC-3′. For measurement of cDNA corresponding to human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA, the forward primer was 5′-GAAGGTGAAGGTCGGAGTC-3′; the reverse primer was 5′-GAAGATGGTGATGGGATTTC-3′; and the probe was 5′-CAAGCTTCCCGTTCTCAGCC-3′. The sequences for the primers and probes for measuring cDNA corresponding to human NQO1, AKR1C, GCLC, and GCLM mRNAs have been described previously (11).

Western blot analysis. For the whole-cell extracts, cells were lysed in an extraction buffer containing 0.1 mol/L HEPES (pH 7.4), 0.5 mol/L KCl, 5 mmol/L MgCl₂, 0.5 mmol/L EDTA, and 20% glycerol supplemented with protease inhibitor mixture (Roche Diagnostics, Indianapolis, IN). Nuclear extracts were prepared as described previously (23). Protein samples were separated in SDS-PAGE gels using a standard protocol. Immunoblottings were carried out using antiserum raised against AKR1C1 or Nrf2 as described previously (24, 25). To confirm equal loading, immunoblotting with antibody against actin (Sigma, St. Louis, MO) was also done.

Statistical analysis. Statistical comparisons were done by unpaired Student's t tests. P < 0.05 was considered statistically significant.

Generation of a stable cell line expressing a functional ARE-driven reporter transgene. In this study, a series of ARE-luciferase reporter plasmids containing either one, two, four, six, or eight copies of the cis-element common to the rat GSTA2 and mouse gsta1 gene promoters were made; the sequences introduced into the pGL3 promoter vector are listed in Table 1. MCF7 cells were used to test these reporter constructs by transient transfection because Taqman analysis showed they contained substantial levels of mRNA for Nrf2 and Keap1 (data not shown). When MCF7 cells were treated with the monofunctional inducer t-BHQ, a marked increase in the amount of Nrf2 protein was detected in whole-cell lysates and nuclei obtained by subcellular fractionation (Fig. 1A). Previous work has shown that Nrf2 accumulates in HEK293 cells and HepG2 cells treated with t-BHQ (10, 26), findings that are consistent with the discovery that activation of ARE-driven gene expression by monofunctional inducing agents involves stabilization of Nrf2 protein through inhibition of Keap1 (9, 25). Our Western blotting data for Nrf2 protein suggest that the Keap1-Cul3 degradation pathway is functional in MCF7 cells and can be inhibited by t-BHQ. The increase in Nrf2 protein we observed following t-BHQ treatment of MCF7 cells is comparable with that observed in HEK293 cells (10) but seems to be greater than that reported in HepG2 cells (26).

Upon transfection of the various reporter constructs into MCF7 cells, it was found that increasing the number of copies of the ARE in the promoter of pGL3 had no significant effect on the basal level of luciferase activity observed under normal homeostatic conditions. However, there was a good correlation between the number of ARE copies in the pGL3 promoter vector and the level of induction of luciferase activity by t-BHQ in the MCF7 cells (Fig. 1B). A similar result was also observed in HepG2 cells transfected with the same reporter constructs (data not shown). These results confirm the findings of Nguyen et al. (27) in which they showed that transfection of reporter plasmids containing multiple copies of the rat GSTA2-ARE increased the sensitivity of chloramphenicol acetyltransferase reporter gene activity to induction by t-BHQ treatment.

To choose an appropriate cell system for the generation of a stable reporter cell line, pGL-GSTA2.41bp-ARE was transfected into HepG2, MCF7, CHO, and Hepa1c1c7 cells. Following transient transfection of this construct into MCF7 cells, luciferase activity was induced up to 50-fold after treatment for 18 hours with 50 μmol/L t-BHQ (Table 2). By contrast, the reporter gene was only induced between 2- and 4-fold following similar transfection experiments in HepG2, CHO, or Hepa1c1c7 cells. These results, therefore, show that MCF7 cells provide a particularly sensitive means of measuring ARE-driven transcription.

We decided to employ pGL-8xARE as the plasmid to generate a stable reporter cell line because this construct gave a reasonably high level of inducible luciferase production following treatment with t-BHQ. To this end, pGL-8xARE and pcDNA3.1, which contained a neomycin selectable marker, were stably cotransfected into MCF7 cells and selected in the presence of G418. In total, 153 G418-resistant clones were isolated. After the first passage, 32 clones were kept for further monitoring according to their basal and inducible luciferase activities. Among them, one clone, called AREc32, showed low basal and high inducible luciferase activity; an ∼10- to 13-fold increase in reporter gene activity was observed following treatment with 10 μmol/L t-BHQ. This clone also showed a stable phenotype after >20 passages. The rest of the clones were discarded because they showed either a lower level of induction (2- to 6-fold) by 10 μmol/L t-BHQ, or an unstable phenotype upon further passage. Therefore, AREc32 cells were retained for further study.

Induction of ARE-driven luciferase activity in AREc32 cells is mediated by Nrf2. To confirm that the luciferase activity in AREc32 cells was responsive to Nrf2, this CNC bZIP protein was overexpressed in AREc32 cells by transient transfection with the expression construct pHyg-EF-hNrf2. As shown in Fig. 2A, the control cells where no DNA was included in the transfection mix gave 13-fold induction of luciferase activity when treated with 10 μmol/L t-BHQ. When 25 ng of pHyg-EF-hNrf2 plasmid DNA was used per well, the basal luciferase activity did not change significantly, and the inducible luciferase activity increased to 18-fold. Following transfection with 50 ng of pHyg-EF-hNrf2 per well, the basal level of luciferase activity increased to 2.6-fold, and the inducible level increased to 22-fold. Finally, following transfection with 100 ng of pHyg-EF-hNrf2, the basal reporter gene activity increased to 4-fold and the inducible level to 28-fold. In different wells, the same amount of pEGFP-N1, an expression vector for enhanced GFP (EGFP), was transfected into AREc32 cells as a negative control. Neither the basal nor the inducible luciferase activities were altered by overexpression of EGFP.

To determine whether Nrf2 mediates induction of luciferase activity by t-BHQ in AREc32 cells, an RNAi vector was used to knockdown its expression. Figure 2B shows that 24 hours after transfection of AREc32 cells with pRS-Nrf2, the endogenous level of mRNA for Nrf2 was reduced to ∼40% of the control level, but its abundance was not affected by transfection with the control pRS-GFP vector. Transfection of these cells with either pRS-hNrf2 or pRS-GFP did not affect the level of GAPDH mRNA (Fig. 2B). This finding indicates that transfection of AREc32 cells with pRS-hNrf2 specifically suppressed expression of the CNC bZIP factor. AREc32 cells were treated with variable amounts of RNAi against Nrf2 and the effect on both basal and inducible luciferase reporter gene activity examined. Transfection of these cells with 25 to 100 ng pRS-hNrf2 per well reduced the basal level of ARE-driven luciferase activity to 60% of the control level (Fig. 2C). When the effect of pRS-hNrf2 on induction of reporter gene activity by 10 μmol/L t-BHQ was tested, a dose-dependent inhibition was observed. The inducibility of luciferase activity was not affected significantly following transfection with 25 ng of pRS-hNrf2 per well, when compared with the mock-transfected control cells (9-fold induction). When 50 ng of pRS-hNrf2 DNA was used to transfect cells in each well, induction of luciferase activity by 10 μmol/L t-BHQ was reduced to 8-fold. When 100 ng of pRS-hNrf2 DNA was used per well, only 5-fold induction by 10 μmol/L t-BHQ was detected. In different wells, the basal and inducible luciferase activities were not affected when AREc32 cells were transfected with the same amount of pRS-GFP DNA, which targeted GFP mRNA (Fig. 2C). These data indicate both basal and inducible ARE-driven gene expression is mediated by Nrf2 in the AREc32 cells.

Time- and dose-dependent induction of luciferase in AREc32 cells. Luciferase activity in AREc32 cells could be induced in a time- and dose-dependent manner; after treatment for 24 hours, luciferase activity was increased 2-fold by 1 μmol/L t-BHQ and 5-fold by 5 μmol/L t-BHQ (see Fig. 3A; Table 3). A maximum luciferase activity (around 10-fold increase) was seen following treatment with 10 μmol/L t-BHQ. Induction of luciferase activity by t-BHQ was also time dependent; it increased 4-fold after 8 hours of treatment with 10 μmol/L t-BHQ and reached 10-fold after 18 hours of treatment with the same dose of t-BHQ (Fig. 3B). A similar magnitude of induction of luciferase activity in AREc32 cells was observed after 24 hours of exposure to 10 μmol/L SUL (see later), a potent NQO1 and AKR enzyme inducer in human HaCaT, LS174, and Caco2 cell lines (11, 28).

Expressions of endogenous AKR1C and luciferase reporter activity in AREc32 cells are both increased by monofunctional inducing agents. To test whether luciferase activity in AREc32 cells reflected expression of endogenous ARE-driven genes, its induction by t-BHQ was compared with that of NQO1, GCLC, GCLM, and AKR1C. After 24 hours of treatment with 10 μmol/L t-BHQ, the levels of mRNAs for NQO1, GCLC, GCLM, and AKR1C were increased 3.0-, 1.7-, 1.7-, and 15-fold, respectively (Fig. 4). By contrast, luciferase activity was induced 10-fold.

Luciferase activity in AREc32 cells was increased significantly by monofunctional and bifunctional inducing agents (for a review of such compounds, see ref. 2) and other Nrf2 activators (Table 3; Supplementary Fig. S1), including acrolein, SUL, BHA, ethoxyquin, pyrrolidine dithiocarbamate (PDTC), indole-3-carbinol, H₂O₂, and β-naphthoflavone (data not shown). In addition, we found that 7-ethoxycoumarin and methyl methanesulfonate (MMS) were effective inducers of luciferase activity in AREc32 cells (Table 3; Supplementary Fig. S1). In our stable reporter cell line, butylated hydroxytoluene (BHT) had no effect on the luciferase activity.

As endogenous AKR1C in AREc32 cells proved to be highly responsive to t-BHQ treatment, we tested whether it was also induced by the compounds that increased luciferase activity. As shown in Table 4, AKR1C mRNA was increased substantially by these compounds, with acrolein, ethoxyquin, and PDTC proving to be particularly effective.

Redox status influences reporter luciferase activity in AREc32 cells. It is unclear whether antioxidants can inhibit inducible ARE-driven gene expression because contradictory data have been published. For example, Pinkus et al. (29) reported that induction of transcription through the mouse gsta1-ARE in HepG2 cells by 100 μmol/L BHA was inhibited by inclusion of 30 mmol/L GSH in the medium, whereas Lee et al. (30) reported that induction of gene expression through the human NQO1-ARE in IMR-32 cells by 10 μmol/L t-BHQ was not inhibited by 2 hours of pretreatment with either 5 mmol/L GSH or 2.5 mmol/L N-acetylcysteine (NAC). In our study, either NAC or GSH was added simultaneously to AREc32 cells along with either t-BHQ or SUL, and their effects on reporter gene expression were examined. As shown in Fig. 5, neither NAC nor GSH influenced basal ARE activity in the stable reporter cells. The induction of luciferase activity by t-BHQ was, however, reduced to only 20% in the presence of 5 mmol/L NAC and to just 30% by 5 mmol/L GSH (Fig. 5A). A similar degree of inhibition of induction was also observed in AREc32 cells treated with SUL (Fig. 5B), where the presence of 5 mmol/L NAC or 5 mmol/L GSH diminished its ability to increase luciferase activity to only 30% of the level seen in the absence of antioxidant. By contrast, the free radical scavengers vitamin E and vitamin C did not markedly inhibit induction of luciferase activity by t-BHQ or SUL. Close examination of Fig. 5A and B suggests that vitamin E can modestly reduce induction of luciferase activity by t-BHQ, but it had no effect on induction by SUL.

To examine whether the intracellular level of GSH itself might negatively control induction of ARE-driven luciferase activity, the cellular GSH pool was depleted using BSO before treatment with t-BHQ or SUL. As shown in Fig. 5C, treatment with BSO alone did not affect the luciferase activity. However, when the cells were pretreated for 24 hours with 50 μmol/L BSO before challenge with t-BHQ alone, the induction of luciferase activity by 10 μmol/L t-BHQ increased from 10- to 25-fold. Furthermore, if the cells were not only pretreated for 24 hours with BSO, but this GSH synthesis inhibitor was also included in the media during challenge with t-BHQ, luciferase activity increased further to 40-fold. In the case of induction by 10 μmol/L SUL, BSO pretreatment resulted in an increase in luciferase activity from 13- to 21-fold, and when BSO treatment was continued in the presence of SUL, luciferase activity increased further to 34-fold.

Anticancer drugs can activate ARE-reporter gene expression. To find out whether cancer chemotherapeutic agents modulate the Nrf2-ARE system, a number of anticancer drugs were screened using AREc32 cells. Based on the IC₅₀ results (data not shown), AREc32 cells were treated for 24 hours with multiple sublethal doses of the therapeutic agents. According to their effect on luciferase activity, these drugs were divided in Table 5 into three groups: those with no significant effect on luciferase activity, modest activators, and strong activators. Thus, doxorubicin, epirubicin, paclitaxel (Taxol), methotrexate, and thiotepa treatment had no effect on the level of luciferase activity in AREc32 cells. By contrast, cisplatin, melphalan, and etoposide modestly increased luciferase activity. Treatment of these cells with the alkylating agents chlorambucil, BCNU, and the cytotoxic antibiotic mitoxantrone produced a stronger induction of luciferase activity of between 2- and 4-fold.

Using AREc32 cells, we found that cyclophosphamide treatment had no effect on ARE-luciferase activity. By contrast, acrolein, a major metabolite of cyclophosphamide was found to be a potent ARE activator; 10 μmol/L acrolein gave a 27-fold increase in luciferase activity.

Activation of ARE-driven gene expression by anticancer drugs is redox dependent. To examine whether cellular GSH level has any effect on the ability of anticancer drugs to activate luciferase activity, we pretreated AREc32 cells with 50 μmol/L BSO for 24 hours before challenging them with chemotherapeutic agents. As can be seen in Fig. 6A, pretreatment with BSO caused the induction of luciferase activity by cisplatin and melphalan to be increased to 3- and 5-fold, respectively. More remarkably, BSO caused the induction of luciferase activity by chlorambucil and BCNU to be increased to >10-fold. Such inductions were nearly completely repressed by the addition of 5 mmol/L NAC. For treatments with etoposide and mitoxantrone, we found that BSO pretreatment did not change luciferase activity significantly (data not shown).

To find out whether anticancer drugs similarly activate the expression of an endogenous Nrf2-regulated gene, we examined expression of AKR1C in AREc32 cells. Without pretreatment with BSO, the mRNA level of AKR1C was only slightly increased by the treatment of the reporter cell line with melphalan, cisplatin, or chlorambucil. However, when the cells were pretreated with 50 μmol/L BSO for 24 hours, melphalan and cisplatin increased the expression of AKR1C mRNA by 3- and 4-fold, respectively, and chlorambucil increased this mRNA 31-fold (Fig. 6B). Treatment with BCNU induced the expression of AKR1C mRNA 3-fold, and with BSO pretreatment, BCNU induced AKR1C mRNA 42-fold (Fig. 6B). Immunoblotting revealed that AKR1C protein was also increased by these anticancer drugs (Fig. 6C). BSO pretreatment did not further enhance the expression of AKR1C protein by t-BHQ treatment. This is, however, possibly because the induction of AKR1C by 10 μmol/L t-BHQ alone has already reached the maximum level.

We have generated a stable ARE-reporter human mammary cell line called AREc32, derived from MCF7 cells, in which only the minimal enhancer sequence (5′-GTGACAAAGCA-3′) is present to direct expression of the luciferase transgene. The ARE employed for this purpose was designed around that found in the promoters of both rat GSTA2 and mouse gsta1. In the latter gene, both its basal and inducible expressions have been shown to be regulated by Nrf2 in vivo (31). It should also be noted that our constructs did not contain the upstream 5′-T(^A/_C)A-3′ trinucleotide sequence that is found in the extended ARE (32, 33); in the rat GSTA2 and mouse gsta1 genes, this is 5′-TAATGGTGACAAAGCA-3′, with the relevant upstream trinucleotides underlined.

We have shown that in our AREc32 cells, expression of luciferase activity is mediated by Nrf2 and is sensitive to redox status. This cell line gave a ∼10-fold increase in reporter gene activity by treatment with relatively modest doses of typical monofunctional inducers and therefore provides a good model system that can be used to screen chemical libraries to identify agonists of Nrf2.

Previously, Zhu and Fahl (34) generated a stable ARE-GFP reporter HepG2 cell line. The reporter construct they employed contained four concatenated copies of the 41-bp ARE-containing promoter sequence from mouse gsta1 ligated to the thymidine kinase promoter driving GFP and gave just 3-fold induction following treatment with 90 μmol/L t-BHQ. Based on induction of reporter genes by t-BHQ, it is apparent that our AREc32 cells gave very substantially greater levels of induction than was found in HepG2 cells. A murine Hepa1c1c7 cell line stably transfected with the mouse 41-bp mouse gsta1-ARE driving the luciferase gene in pGL3-basic has been described recently (35), and it seems to give comparable levels of induction as we find in our AREc32 cells.

Mechanism of ARE-driven gene induction by t-BHQ. We have employed t-BHQ as our principal monofunctional inducer to challenge AREc32 cells because it was used in the original experiments to define both this type of agent (21) and the ARE consensus sequence (18). In this case, t-BHQ probably induces ARE-driven genes because it undergoes a two-electron oxidation to form an electrophilic quinone. Correlation of the relative abilities of t-BHQ and its analogue 2,5-di-tert-butylhydroquinone to induce rat GSTP1 with their relative redox-cycling activities, their abilities to produce superoxide, or their electrophilic properties has established that it is the last characteristic that is most important for activation of ARE-driven gene expression (35). In particular, Nakamura et al. (36) found that t-BHQ can form a quinone and is a good inducer of GSTP1, whereas 2,5-di-tert-butylhydroquinone cannot form a quinone and is not a good inducer of GSTP1. As these workers showed that both compounds could be oxidized to their respective benzoquinones and both could generate superoxide by redox-cycling, but only t-BHQ and not 2,5-di-tert-butylhydroquinone could form a quinone that would react with GSH, they considered that the ability to form a thiol-reactive electrophile is the essential characteristic required for induction of ARE genes (36).

Response of AREc32 cells to monofunctional inducers. Initial characterization of our stable reporter cell line showed that ARE-driven luciferase activity could be induced up to 70-fold by treatment with 50 μmol/L t-BHQ (with serum). However, in most experiments, we used just 10 μmol/L t-BHQ (without serum), and this elicited an ∼10-fold increase in reporter activity. Treatment of AREc32 cells with 10 μmol/L SUL, another widely used monofunctional inducer, resulted in a similar increase in luciferase activity. We also established that the endogenous ARE gene battery is inducible in the stable cell line. Taqman RT-PCR and Western blotting showed that endogenous members of the ARE gene battery, such as AKR1C, NQO1, GCLC, and GCLM, could be induced by t-BHQ in our stable reporter cell line. The increase in expression of these genes was similar to the increases seen in Caco2 and LS174 cells treated with t-BHQ or SUL (11, 28).

Response of AREc32 cells to metabolizable phenolic antioxidants. Following confirmation that monofunctional inducers increase the expression of endogenous genes in AREc32 cells, we tested whether BHA and ethoxyquin were capable of inducing luciferase reporter activity in these cells because they are classic chemopreventive-blocking agents. In rodents, these antioxidants are each strong inducing agents. This property is not, however, thought to be due to their antioxidant properties. Rather, it is thought that it is necessary for them to be metabolized into compounds that contain quinone or α,β-unsaturated carbonyl groups in order for them to induce ARE-driven genes (20). For example, BHA is O-demethylated by cytochrome P450 to yield t-BHQ that, as described above, is susceptible to a two-electron oxidation to form tert-butylbenzoquinone (20). This latter metabolite is thiol-active and is postulated to be the ultimate inducing agent (36). In AREc32 cells, t-BHQ was 20-fold more potent than BHA at inducing luciferase activity (Table 3), an observation that is consistent with the hypothesis that BHA requires to be metabolized via t-BHQ to tert-butylbenzoquinone to induce gene expression. We also found that ethoxyquin is a relatively potent inducer of reporter gene activity in AREc32 cells, which suggests it is efficiently converted by O-de-ethylation and oxidation reactions to quinoline- and quinolone-containing metabolites, which are considered to be the ultimate inducers (20). In further experiments, we found that BHT did not induce gene activity in our stable reporter cell line despite it being a potent inducing agent in rats (14). Unfortunately, we do not know which cytochrome P450s are expressed in our stable reporter cell line, although in terms of explaining the different potency of induction affected by the antioxidants, it would be interesting to know if the AREc32 cells catalyze O-de-alkylation of BHA and ethoxyquin but not of BHT.

Besides BHA, ethoxyquin, and BHT, we have examined the effect of PDTC on luciferase activity in AREc32 cells. In this case, the dithiocarbamate was shown to be a robust inducer of ARE-driven reporter gene expression and endogenous AKR1C. This is not surprising as PDTC has previously been found to induce GCLM in HepG2 cells (37), and there is good evidence that following the conversion of PDTC to thioram disulfides, it is able to oxidize GSH and protein sulfhydryls (37). Taken together with previous studies (17, 38, 39), our findings support the hypothesis that synthetic antioxidants are not themselves inducers, but those that are metabolized to thiol-active quinones, quinolones, α,β-unsaturated carbonyls, or thioram disulfides can activate transcription of the ARE gene battery.

Modulation of reporter gene activity in AREc32 cells by redox status. The contribution that ROS may make to Nrf2-mediated gene transcription is difficult to determine because they can generate α,β-unsaturated carbonyls, such as acrolein and 4-hydroxynonenal, through modification of intracellular macromolecules that are potent inducing agents (14). In the present study, we found H₂O₂, which is included among ROS, to be a relatively poor inducer of luciferase reporter activity and AKR1C (Tables 3 and 4), but it is not known if this modest level of induction is a direct or indirect effect of the oxidant. To address the question of whether ROS contribute to homeostatic ARE-driven transcription, we treated AREc32 cells with the thiol antioxidants NAC and GSH and with the free radical scavengers vitamin C and vitamin E. None of these compounds influenced basal luciferase reporter activity in AREc32 cells. We also treated the reporter cells with BSO to deplete GSH and thereby increase the intracellular level of ROS, and this was also found to have no effect on basal luciferase activity (Fig. 5C). Whereas these experiments do not exclude ROS as agonists of ARE-driven gene expression, our data suggest that these agents are not potent inducers.

Although NAC and GSH did not have any effect on basal luciferase activity in AREc32 cells, they substantially attenuated induction of ARE-driven luciferase by either t-BHQ or SUL. These findings are consistent with the hypothesis that the intracellular GSH pool acts as a buffer to negatively regulate the ARE gene battery by preventing thiol-active inducing agents from interacting with critical cysteine residues in proteins including Keap1 (38, 40). By comparison, vitamin C and vitamin E only modestly inhibited induction of luciferase by t-BHQ. This small decrease in induction by t-BHQ is presumably because vitamins C and E inhibit its two-electron oxidation to tert-butylbenzoquinone, the ultimate inducing agent. Neither vitamin C nor vitamin E inhibited induction of luciferase by SUL presumably because the isothiocyanate itself is the inducer and does not require further oxidation to activate Nrf2.

Our finding that NAC can substantially repress induction of ARE-driven gene expression by t-BHQ is in marked contrast to the results of Lee et al. (30) who reported that NAC did not influence induction of NQO1 in IMR-32 cells by t-BHQ. A probable explanation for this discrepancy is that the human NQO1-ARE contains an embedded AP1 site (5′-GTGACTCAGCA-3′), whereas the ARE in rat GSTA2 and mouse gsta1 (5′-GTGACAAAGCA-3′) does not. It is distinctly possible that transcriptional activation of the human NQO1 gene via its ARE involves Jun family members as well as Nrf2 (2). The ARE in the promoter of the human ferritin H gene contains an embedded AP1 site and can recruit JunD (41). Significantly, c-Jun and c-Fos can be induced by t-BHQ (42), suggesting that temporal differences may occur in the complexes recruited to the human NQO1-ARE. Our interpretation of these apparently contradictory findings is that antioxidants can inhibit thiol-active inducers from activating Nrf2, but they do not have a similar effect on c-Jun and c-Fos. This hypothesis requires to be tested.

Induction of reporter gene activity in AREc32 cells by mutagens and anticancer drugs. Monofunctional inducers that activate ARE-driven transcription are all soft electrophiles as they react with soft nucleophiles, such as SH groups, rather than reacting with N or O atoms in macromolecules (43). We found that the strong mutagen MMS, that can be classed as a hard electrophile because it is capable of N-methylating basic amino acids in protein as well as both N- and O-methylating nucleic acid bases in DNA (44), was able to induce ARE-luciferase activity in AREc32 cells. Interestingly, MMS can also methylate thiol compounds and deplete cellular GSH (45). Based on our data on MMS, it seems that both hard and soft electrophiles can activate ARE-driven transcription.

During the present study, AREc32 cells were used to examine whether chemotherapeutic drugs that produce electrophilic metabolites, such as alkylating agents, can induce ARE-driven gene expression. We found that melphalan, chlorambucil, and BCNU were effective inducers of luciferase reporter gene activity. Induction of ARE-driven gene expression by these alkylating agents was redox sensitive, insofar as it was augmented by BSO pretreatment and suppressed by NAC. It therefore seems most probable that melphalan, chlorambucil, and BCNU induce the ARE gene battery through modifying cysteine residues in Keap1. Further studies are required to test this prediction. We also found that thiotepa and cyclophosphamide, which are alkylating agents, did not induce luciferase reporter gene activity. In both cases, the agents are prodrugs and require to be metabolized to be effective. It therefore seems probable that the cytochrome P450s required for their activation are not expressed in our stable reporter cell line.

In addition, to induction of ARE-driven gene expression by certain alkylating agents, we found that mitoxantrone (a cytotoxic antibiotic), etoposide (a topoisomerase 2 inhibitor), and cisplatin (a platin compound that cross-links DNA strands) were all able to induce luciferase reporter gene activity. The metabolic processes by which these agents act as inducers of ARE genes, and how they impinge on Keap1, is not known. It should, however, be noted that induction by cisplatin was increased by pretreatment of AREc32 cells with BSO, an observation that is consistent with the notion that thiol chemistry underpins the process.

As mentioned above, the role of ROS in inducing ARE genes is poorly understood. In the course of this study, we found that the cytotoxic antibiotic doxorubicin did not induce luciferase activity in AREc32 cells. The fact that doxorubicin is thought to act as a redox-cycling agent and can therefore produce free radicals suggests that ROS may not be particularly potent at inducing ARE-driven genes.

Contribution of Nrf2 to drug resistance. The older literature focused on the contributions of the P-glycoprotein (MDR1) and GST to acquired resistance of tumor cells to chemotherapeutic drugs (1). There is, however, increasing evidence that the redox status of cells is an important factor in determining whether tumor cells can withstand chemotherapy (12). The fact that Nrf2 controls the expression of GCLC and GCLM, which together catalyze the rate-limiting step in GSH biosynthesis, may strongly influence cell survival through inhibition of apoptosis. In particular, ROS have been found to inhibit the growth of malignant cells and/or to induce apoptosis (46, 47). Increases in GSH levels would be expected to reduce ROS levels and antagonize apoptotic signals. Studies in a variety of tumor cell types have suggested that cancer chemotherapy drugs induce apoptosis in part by generating endogenous oxidants (48). Previous work has shown that an association exists between a coordinated increase in the expression of antioxidant proteins and resistance to anticancer drugs. For example, high levels of GCLC and MRP have reported to be associated with increased resistance towards doxorubicin in mesothelioma cells (49), and an increase in GCLC has been associated with resistance to cisplatin in ovarian cells (50). Recently, it has been reported that acquired resistance of KCL22 human chronic myelogenous leukemia cells to imatinib was linked to increases in both GSH levels and the amounts of nuclear Nrf2 (51). Treatment with vitamin C diminished the amount of nuclear Nrf2 and restored sensitivity to imatinib (51). The molecular mechanism responsible for this finding is unknown, but it infers that Nrf2 and GSH can play a pivotal role in drug resistance.

Concluding comments. We have developed a stable human mammary reporter cell line to allow inducers of ARE-driven genes to be identified. These cells have been used to show that alkylating agents and some other anticancer drugs can induce the ARE gene battery. Furthermore, in some instances, this induction can be dramatically enhanced by depletion of intracellular GSH. An important inference from these observations is that suboptimal treatment of cancer by chemotherapeutic drugs may induce adaptive cytoprotective genes in tumors. It is also clear that the redox status of cells in the tumor will influence their ability to activate such defenses. Thus, suboptimal treatment may cause depletion of GSH that could in turn allow increased up-regulation of cytoprotective genes. Other types of therapeutic agent, such as the analgesic acetaminophen, can deplete GSH, and these may similarly allow augmented induction of cytoprotective genes in tumor cells by anticancer drugs.

Grant support: Imperial Cancer Research Fund Incubator Grant (C.R. Wolf and J.D. Hayes).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Prof. Masayuki Yamamoto and Dr. Ken Itoh (Institute of Basic Medical Sciences, University of Tsukuba, Japan) for providing the pHyg-EF-hNrf2 construct and Lesley McLaughlin for helpful comments about the characteristics of the chemotherapeutic drugs studied.