BRCA1 Induces Antioxidant Gene Expression and Resistance to Oxidative Stress

Inherited mutations of the breast cancer susceptibility gene breast cancer susceptibility gene 1 (BRCA1) confer an increased risk for breast, ovarian, and prostate cancers (1, 2). In addition, BRCA1 expression is often decreased or absent in sporadic breast and ovarian cancers due, in part, to promoter methylation or other causes, suggesting a role for BRCA1 in nonhereditary tumors (3, 4). The specific functions of the BRCA1 gene that contribute to tumor suppression are unclear. However, established functional roles for BRCA1 include the regulation of cell cycle progression, DNA damage signaling and repair, maintenance of genomic integrity, and the regulation of various transcriptional pathways [reviewed by Rosen et al. (5)].

A role for BRCA1 in transcriptional regulation was first suggested by the finding that BRCA1 has a conserved acidic COOH-terminal transcriptional activation domain (6). Although BRCA1 is not known to bind to specific DNA sequences, it may regulate transcription through protein:protein interactions with components of the basal transcription factor (e.g., RNA helicase A and RNA pol II), transcriptional coactivators and corepressors [e.g., p300 and its functional homologue CBP (the cAMP-responsive element binding proteinbinding protein), retinoblastoma 1, retinoblastoma 1-associated proteins (RbAp46/48), and several histone deaceylases (HDAC-1/2)], and/or sequence-specific DNA-binding transcription factors (e.g., p53, c-Myc, estrogen receptor, and other proteins; refs. 7, 8, 9, 10, 11, 12).

Some of the functions of BRCA1 cited above may be due, in part, to regulation of specific transcriptional pathways by BRCA1, but the linkage of these functions to BRCA1-regulated transcription is not well understood. We used cell culture models of BRCA1 overexpression, underexpression, and mutational inactivation to identify patterns of BRCA1-regulated gene expression. The identification of antioxidant genes as transcriptional targets of BRCA1 led to the findings that BRCA1 regulates the activity of the antioxidant response transcription factor nuclear factor erythroid-derived 2 like 2 [also called NRF2 (NFE2L2)] and protects cells against oxidative stress.

Human prostate (DU-145, LNCaP) and breast (MCF-7, T47D) cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA) and cultured as described before (12, 13). The stable wild-type BRCA1 (wtBRCA1) and control (Neo) DU-145 cell clones were isolated and characterized earlier (14, 15). A mouse embryonic fibroblast (MEF) cell line homozygous for a deletion of Brca1 exon 11 and the control wild-type (Brca1^+/+) MEFs (16) were provided by Dr. Chuxia Deng (National Institutes of Diabetes, Digestive and Kidney Diseases, Bethesda, MD). All of the above cell types were grown in DMEM supplemented with 5% (DU-145) or 10% (all other cell types) v/v fetal calf serum, l-glutamine (5 mmol/L), nonessential amino acids (5 mmol/L), penicillin (100 units/mL), and streptomycin (100 μg/mL; all obtained from BioWhittaker, Walkersville, MD).

EBV-immortalized peripheral blood lymphocyte cell lines R794 and R1041 were derived from a female BRCA1 (185delAG) and BRCA2 (6174delT) mutation carrier, respectively. These lymphoblastoid cell lines were provided by the Tissue Culture Shared Resource of the Lombardi Comprehensive Cancer Center. The genotypes of the cells were confirmed by the Familial Cancer Registry of the Lombardi Comprehensive Cancer Center.

For transient expression experiments, cells were transfected with a wild-type BRCA1 expression vector (wtBRCA1) consisting of the full-length BRCA1 cDNA within the pcDNA3 mammalian expression vector (Invitrogen, Carlsbad, CA) or within the pCMV-Tag2B vector (Stratagene, La Jolla, CA), which allows expression of the full-length protein containing a NH₂-terminal FLAG epitope tag. Both the untagged and the FLAG-tagged proteins are expressed well and exhibit identical biological activities (13). Methodologies used for transient transfections have been reported previously (13, 15) and are also briefly described below.

The BRCA1 and control (scrambled-sequence) siRNAs were described earlier (15). All siRNAs were chemically synthesized by Dharmacon, Inc. (Lafayette, CO). For siRNA treatments, subconfluent proliferating cells were treated with each siRNA (50 nmol/L), with siPORT Amine reagent (Ambion, Austin, TX). The cells were incubated with siRNA for 72 hours (to reduce BRCA1 protein levels to <25% of control) before the start of the experiment. The control siRNA has no effect on BRCA1 levels (15), and neither siRNA is toxic to the cells under these experimental conditions, as determined by the use of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays.

The total cellular RNA was extracted with TRIzol Reagent (Life Technologies, Inc., Rockville, MD), according to the manufacturer’s instructions, additionally purified with chloroform and precipitated with 95% etomidate before cDNA synthesis. The quality of isolated RNA was verified by electrophoresis through 1.0% agarose-formaldehyde gels, and its quantity was determined from absorbance measurements at 260 and 280 nm.

cDNA-spotted slides corresponding to 9216 human genes (including expressed sequence tags) and 9568 mouse genes (including expressed sequence tags) were prepared at the Albert Einstein College of Medicine microarray facility (Bronx, NY). cDNA synthesis, hybridizations, scanning, gridding, and analysis have been described earlier (ref. 17; also see web site).⁴ On the basis of our experience suggesting that cDNA spotted microarrays often underestimate differences in gene expression (17), ratios of gene expression were considered to be significant if they were ≥1.5 or ≤ 0.7 in at least two independent experiments.

For DU-145 cells, we compared gene expression in two different wtBRCA1 versus Neo clone pairs, with two independent experiments per clone pair, for a total of n = 4 independent experiments. For MCF-7 cells, subconfluent proliferating cells were transiently transfected with wtBRCA1 or empty pcDNA3 vector (15) and postincubated for 24 hours to allow gene expression. Three independent experiments comparing wtBRCA1- versus pcDNA3-transfected cells were made after confirming that the wtBRCA1 gene was expressed in each experiment. For MEFs, we performed three independent comparisons of Brca1-deficient (Δ exon 11) versus wild-type MEFs. In each case, the ratios of gene expression were considered to be significant if they were ≥1.5 or ≤0.7 in at least two independent experiments.

Affymetrix microarray analyses were performed at the North Shore-Long Island Jewish Research Institute core facility. RNA isolation, cRNA synthesis, gene chip hybridizations, and data analysis were performed as described earlier (18). We performed one experiment each comparing a DU-145 wtBRCA1 versus Neo clone pair and comparing Brca1-deficient MEFs versus wild-type MEFs. The gene chips used for these experiments were HG-U133A (which contains ∼16,000 human probe sets) and MG_U74Av2 (which contains ∼12,000 mouse genes plus expressed sequence tags). Differences in gene expression were considered to be significant if the log signal ratios were ≥ +1 or ≤ −1 and the P values were significant according to the Affymetrix algorithm. These log signal ratio cutoffs correspond to ratios of ≥2.0 or ≤0.5, respectively.

Rigorously controlled semiquantitative reverse transcription-PCR assays were performed as described before (15, 17). The PCR primers, reaction conditions, and cycle numbers are shown in Tables 1 and 2. The PCR reactions were individually optimized so that each reaction fell within the linear range of product amplification. The first-strand cDNA template was generated from 1 μg of total RNA in a final volume of 20 μL, with SuperScript II reverse transcriptase (Life Technologies, Inc.) and oligo(dT) primers. One microliter (of 20 μL) of 1:2.5-diluted cDNA template was amplified in a total volume of 50 μL, containing 200 μmol/L each of all four deoxynucleoside triphosphates, 2 μmol/L each of specific primers, and 1 unit of Tag DNA polymerase (Perkin-Elmer, Norwalk, CT). β-Actin, whose expression is unaffected by BRCA1, was used as a control for loading. The PCR products were analyzed by electrophoresis through 1.0% agarose gels containing ethidium bromide (0.1 mg/mL) and photographed under UV illumination.

Subconfluent proliferating cells in 96-well dishes were treated with different doses of H₂O₂ or paraquat (Sigma Chemical Co., St. Louis, MO) for 24 hours (or for different time intervals) and then assayed for MTT dye reduction, a measure of mitochondrial viability (14, 19). Cell viability was normalized to 0 dose control cells. Cell viability values were calculated as means ± SE of n = 10 replicate wells or as means ± SE for three independent experiments, each of which used n = 10 replicate wells per cell type per assay condition.

This assay measures the ability of intact cell membranes of viable cells to exclude trypan blue dye. Subconfluent proliferating DU-145 cells in 100-mm plastic Petri dishes were transfected overnight using Lipofectamine with a FLAG-wtBRCA1 expression vector or the empty pCMV-Tag2B vector (15 μg of plasmid DNA per dish), washed, and allowed to recover and express the transgene for 24 hours. wtBRCA1-transfected, empty vector-transfected, and untransfected control cells were harvested, plated into 2-cm² wells (8 × 10⁴ cells per well in quadruplicate wells), allowed to attach, and exposed to different doses of H₂O₂ for T = 24 hours at 37°C. The cells were then collected, suspended in a solution containing 0.4% trypan blue, and counted with a hemocytometer. For each experiment and dose of H₂O₂, at least 200 cells were counted per well. Three independent experiments were performed; and the cell viability values were expressed as means ± SE.

Subconfluent proliferating cells in 24-well dishes were transfected overnight with wtBRCA1 or empty pcDNA3 vector (0.25 μg per well) with Lipofectamine, washed, and postincubated in fresh culture medium for 24 hours to allow gene expression. The cells were then treated with different doses of H₂O₂ for T = 24 hours and assayed for reduced (GSH) or oxidized (GSSG) forms of glutathione using a kit from Oxis, Inc. (Portland, OR).

The wild-type NRF2 vector, dominant negative NRF2 vector (DN-NRF2), NQO1-ARE-Luc reporter, and mutant or truncated BRCA1 expression vectors have been described earlier (12, 20). The NQO1-ARE-Luc reporter contains the antioxidant response element (ARE) of NAD(P)H dehydrogenase quinone 1 (NQO1), driving a minimal promoter upstream of the luciferase gene. Transient transfection assays were performed to measure transcriptional activity, as described earlier (12, 15). Briefly, subconfluent proliferating cells in 24-well dishes were transfected overnight with the indicated expression vector(s) (0.25 μg per well) and luciferase reporter (0.25 μg per well), with Lipofectamine. The cells were washed and postincubated for 24 hours to allow luciferase expression. Luciferase values (minus background) were normalized to the control (reporter only) and expressed as means ± SE of quadruplicate wells. Transfection efficiency was monitored using the control plasmid pRSV-β-gal (15).

Whole cell lysates were prepared and subjected to Western blotting, as described earlier (14, 15). Briefly, equal aliquots of total cellular protein (50 μg per lane) were electrophoresed on a 4 to 13% SDS-polyacrylamide gradient gel, transferred to nitrocellulose membranes (Millipore, Bedford, MA), and blotted with primary antibodies directed against human BRCA1 (C-20, rabbit polyclonal, 1:200; Santa Cruz Biotechnology, Santa Cruz, CA) and α-actin (I-19, goat polyclonal, 1:500; Santa Cruz Biotechnology). After incubation with the appropriate horseradish peroxidase conjugated secondary antibody (Amersham Lifescience), immune complexes were visualized by using an enhanced chemiluminescence detection system (Amersham Lifescience, Buckinghamshire, UK), with colored markers (Bio-Rad, Hercules, CA) as molecular size standards.

Where appropriate, statistical comparisons were made using the two-tailed Student’s t test.

To determine the effect of BRCA1 overexpression on the transcriptosome, we compared the gene expression profiles of DU-145 human prostate cancer cell clones stably expressing a wild-type BRCA1 gene (wtBRCA1) with control (Neo) clones with cDNA spotted microarrays. These wtBRCA1 and Neo cell clones have been described and extensively characterized in previous studies (13, 14, 15). A partial list of genes up-regulated in the wtBRCA1 clones, categorized by function, is shown in Table 3 (see Supplemental Material for complete list). The wtBRCA1 cells showed up-regulation of various types of genes, including those involved in transcription, stress responses, DNA replication and repair, signal transduction, metabolism, differentiation, and RNA and protein processing. As a check on the methodology, a separate experiment revealed that a large number of genes identified by the cDNA-spotted arrays were concordantly up-regulated with an Affymetrix oligonucleotide microarray (Table 3).

Noticeably, the wtBRCA1 cell lines overexpressed many genes that protect against oxidative stress, including microsomal glutathione S-transferases (GSTs; MGST1 and MGST2), cytoplasmic GSTs (GSTT1 and GSTZ1), a glutathione peroxidase (GPX3), and various oxidoreductases [e.g., NQO1, alcohol dehydrogenase 5 (ADH5), and malic enzyme (ME2)]. wtBRCA1 also up-regulated other potential antioxidant genes, including paraoxonase 2 (PON2), an enzyme that hydrolyzes toxic organophosphates (e.g., pesticides) and oxidized lipids (e.g., oxidized low density lipoprotein; ref. 21), the Klotho gene (KL), a deficiency of which causes oxidative brain damage and a shortened life span in mice (22), and ubiquitin carboxyl-terminal esterase L1 (UCHL1), an oxidation-sensitive ubiquitin recycling enzyme that has been implicated in Parkinson’s disease (23). A number of these genes are involved in xenobiotic and drug metabolism: e.g., GSTs, NQO1, PON2, and member of PAS protein 2 (MOP2, also called HIF2α), an aryl hydrocarbon receptor (AhR) family gene.

Somewhat fewer genes were down-regulated by wtBRCA1 than were up-regulated (see Table 4 for a partial list and see Supplemental Material for the complete list). Again, an Affymetrix microarray experiment identified many of the same genes found with cDNA spotted arrays (Table 4). Only one GST, GSTP1, was decreased in wtBRCA1 clones. Interestingly, the overexpression of this particular isoform of GST in cancer cell lines is associated with cellular chemoresistance (24). Various genes involved in cell cycle regulation and DNA repair were down-regulated, including the retinoblastoma susceptibility gene retinoblastoma 1, which is known to be down-regulated by wtBRCA1 (13).

On the basis of previous experience with cDNA-spotted microarrays, we expected a low rate of false positivity with the selected filtering criteria (ref. 17; see Materials and Methods). To confirm this expectation, we tested n = 15 genes with rigorously controlled semiquantitated reverse transcription-PCR assays (14, 15, 17) of RNA samples from parental DU-145 cells and three clones each of Neo and wtBRCA1 cells. β-Actin, which is unaffected by BRCA1, was used as a control gene. The expression of BRCA1 in the wtBRCA1 relative to control cell lines is shown in Fig. 1,A. For all 15 genes, the expected increases (Fig. 1,B) or decreases (Fig. 1,C) in gene expression were confirmed. In some cases, the fold changes (determined by densitometry and expressed relative to β-actin) were greater by reverse transcription-PCR than by microarray assays, consistent with our impression that cDNA spotted arrays often underestimate gene expression changes. We tested the effect of BRCA1 knockdown with a previously validated siRNA (15) on the expression of three antioxidant response genes that were up-regulated by wtBRCA1 (MGST1, NQO1, and GSTZ1). In each case, the mRNA levels were decreased by BRCA1-siRNA but not by control-siRNA (Fig. 1 D). These findings suggest that BRCA1 regulates the expression of some antioxidant response genes over a very wide range of intracellular BRCA1 protein levels.

We also examined the effect of overexpression of wtBRCA1 on gene expression in MCF-7 human breast cancer cells. In these studies, gene expression was compared in MCF-7 cells transiently transfected with wtBRCA1 versus empty pcDNA3 vector. Gene expression was compared in wtBRCA1 versus control (pcDNA3)-transfected cells in the absence (−E2) or presence (+E2) of exogenous estrogen (17β-estradiol, 1 μmol/L × 24 hours). Although the cell types and duration of BRCA1 expression differed, we identified >40 genes concordantly regulated by wtBRCA1 in DU-145 versus MCF-7 cells (Table 5). These include MGST1, ANX1, ADH5, DSS1, MOP2, IGFBP3, GNG10, NOV, G6PD, KRT19, IFRD2, and others. In the absence and/or presence of E2, wtBRCA1 up-regulated expression of genes involved in the oxidative stress response or the detoxification of xenobiotics and drugs in MCF-7 cells, including MGST1, MOP2, ADH5, ALDH8, an epoxide hydrolase (EPHX2), several selenoproteins (SEPHS1 and SEPW1), and several other antioxidant proteins (PRDX4 and TSA).

Next, we determined the effect of loss of the endogenous full-length Brca1 protein on gene expression. Thus, we compared gene expression in Brca1-deficient (exon 11-deleted) versus wild-type (Brca1^+/+) MEFs. The Brca1 Δ exon 11 MEFs express a M_r 92,000 Brca1 protein that is defective in DNA repair function (25). Examples of genes down-regulated in Brca1-deficient MEFs are listed in Table 6 (see Supplemental Material for the full list.) Categories of genes underexpressed in the Brca1-deficient cells included those involved in transcription, stress responses, cell cycle regulation, DNA replication and repair, signal transduction, and other processes. Stress response genes up-regulated in Brca1-deficient cells included a GST (Gsta2), a glutathione peroxidase (Gpx3), the antioxidant response transcription factor Nfe2l2 (also called Nrf2), a Nrf2 binding partner [activating transcription factor 2 (Atf2)], a superoxide dismutase (Sod1), a selenoprotein (Sepp1), the aryl hydrocarbon receptor (Ahr), several etomidate-responsive genes, and several heat shock proteins. In contrast, very few genes were up-regulated, and the magnitude of the increases was small (see Supplemental Material). Examples of genes concordantly increased in BRCA1-overexpressing (DU-145 or MCF-7) cells and decreased in Brca1-deficient MEFs include PROS1, GPX3, BNIP3, and TGFB2 (see Table 7 A).

We compared our findings with published microarray studies of gene regulation by BRCA1 in 293T cells (26), colon cancer cells (27), and Brca1-deficient mouse embryonic stem cells (28). Examples of genes concordantly regulated in our study versus published studies are provided in Table 7 B. These include (a) genes commonly induced by wtBRCA1 in 293T and in DU-145 and/or MCF-7 cells (MGAT2, CCNG2, FSTL1, LAMA3, and KCNK1), (b) genes induced by wtBRCA1 in 293T cells and decreased in Brca1-deficient MEFs (SEPP1, ZNF148, and ENPP2), (c) genes induced by wtBRCA1 in colon cancer cells and decreased in Brca1-deficient MEFs (TOP1 and SOD1), (d) genes decreased by wtBRCA1 in colon cancer and DU-145 cells (CD59), (e) genes decreased in Brca1-deficient embryonic stem cells and MEFs (Rock2, Qk, and Nfl), and (f) genes decreased in Brca1-deficient embryonic stem cells and up-regulated in DU-145 wtBRCA1 cells (CKB, SYHUQT, and PSMC2).

To determine whether the ability of BRCA1 to stimulate the expression of antioxidant response genes has functional consequences, we measured the effects of BRCA1 on the cellular sensitivity to two different oxidizing agents, hydrogen peroxide (H₂O₂), and paraquat. DU-145 wtBRCA1 or Neo clones were exposed to different doses of H₂O₂ for 24 hours, after which, the cell viability was determined by using MTT assays. The wtBRCA1 cells were significantly more resistant to H₂O₂ over a wide range of doses (P < 0.001, two tailed t tests; Fig. 2,A). In concordance with these findings, pretreatment of parental DU-145 cells with a BRCA1-siRNA caused significant sensitization to H₂O₂ (P < 0.001; Fig. 2,B). Please note that the experiments shown in Fig. 2, A and B, are representative of two or three independent experiments of each type that showed similar results. These results suggest that both exogenous and endogenous BRCA1 protects DU-145 cells against oxidative stress due to H₂O₂.

We also tested the ability of endogenous Brca1 to protect MEFs against H₂O₂ induced cytotoxicity. Consistent with the results obtained with DU-145 cells, Brca1-deficient MEFs were more sensitive than control (Brca1^+/+) MEFs to H₂O₂ (P < 0.001 to 0.01; Fig. 2,C). The cell viability values shown in Fig. 2,C are means ± SE of three independent experiment, each of which used 10 replicate wells per dose of H₂O₂. The herbicide paraquat induces cytotoxicity by causing the generation superoxide ions (O₂⁻), which are detoxified by a mechanism distinct from H₂O₂. Brca1-deficient MEFs were more sensitive to paraquat than wild-type MEFs (P < 0.001 to 0.01; Fig. 2,D). A DU-145 wtBRCA1 clone was less sensitive than the Neo clone, whereas BRCA1-siRNA conferred increased sensitivity to paraquat (P < 0.001 to 0.05; Fig. 2 E). These findings suggest that exogenous and endogenous BRCA1 protects cells against several distinct forms of oxidative stress.

The assays shown in Fig. 2 A–E used a 24-hour exposure to H₂O₂ or paraquat. We performed additional studies to rule out the possibility that the effects of BRCA1 are limited to short-term assays. Thus, DU-145 wtBRCA1 or Neo cell clones were incubated with H₂O₂ for different time intervals from T = 16 to 96 hours and then tested for cell viability with MTT assays. Because of the prolonged exposure times, lower doses of H₂O₂ (either 10 or 25 nmol/L) were tested. These studies revealed persistent and significant increases in viability of the wtBRCA1 cell clones. Thus, at the lower dose of H₂O₂, cell viability increases of up to ∼20% were observed, whereas at the higher doses, increases of up to 30 to 35% were found (P < 0.001 to 0.05, two-tailed t tests).

Finally, we tested the effect of transient expression of wtBRCA1 on the response of DU-145 cells to H₂O₂ with a different end point, trypan blue dye exclusion. MTT assays assess the ability of mitochondria to reduce a tetrazolium salt to formazan, a measure of mitochondrial viability, whereas trypan blue exclusion assesses the ability of an intact plasma membrane to exclude the dye. At H₂O₂ doses of 300 to 500 nmol/L, we found significant increases (11 to 40%) in the proportion of wtBRCA1-transfected cells that excluded trypan blue dye, as compared with empty vector transfected cells or untransfected cells (P < 0.001; Fig. 2,G). Values in Fig. 2 G are means ± SE of three independent experiments. Although there is some variability from experiment to experiment, the cell viability values tended to be higher with the trypan blue assay than the MTT assay. This may reflect the fact that loss of membrane integrity is a late end point; thus, cells that have lost mitochondrial function (MTT end point) may not yet have lost their membrane integrity. Regardless, it seems clear that overexpression of BRCA1 (by either stable or transient transfection) protects and inactivation of BRCA1 (by either knockdown or gene deletion) sensitizes cells against oxidative stress.

The response to oxidative stress depends upon the ability of the cell to maintain its redox balance (i.e., the ratio of reduced to oxidized glutathione) in the setting of stress. We examined the effect of exogenous wtBRCA1 on the redox balance of prostate (DU-145 and LNCaP) and breast (MCF-7) cancer cell lines after treatment with different doses of H₂O₂ for 24 hours. The end point was the ratio of GSH to GSSG. BRCA1-transfected cells showed a mostly similar basal redox balance to vector-transfected and untransfected control cells (Fig. 3). H₂O₂ caused a dose-dependent shift in the redox state to increased GSSG and decreased GSH levels. However, wtBRCA1-transfected cells were able to maintain significantly higher ratios of GSH/GSSG than control cells, especially at high doses of H₂O₂ (P < 0.001, two tailed t tests). These findings suggest that BRCA1 enhances the production of GSH in response to oxidative stress.

The cytoprotective antioxidant response is mediated, in part, by the nuclear factor (erythroid-derived 2)-like factors NFE2L2 (NRF2) and NFE2L1 (NRF1) via the ARE (29). In this regard, BRCA1 appears to regulate a subset of genes that are known to be regulated by NRF2 [NQO1, MGST2, G6PD, malic enzyme (ME2), and Gsta1/2] and/or that are known to contain AREs in their promoters (NQO1, MGST1/2, and Gsta1/2; refs. 30, 31). This finding suggests that BRCA1 protection against oxidants may be mediated, in part, by NRF2. To test this hypothesis, we performed transient transfection assays using an NRF2-responsive reporter driven by the ARE of NQO1 (NQO1-ARE-Luc).

wtBRCA1 increased the basal activity of the NQO1-ARE-Luc reporter in DU-145, T47D, and MCF-7 cells by 1.6 to 6.6-fold, as compared with empty pcDNA3 vector or no vector (Fig. 4,A). In these assays, MCF-7 cells showed larger wtBRCA1-induced increases in ARE-Luc activity than DU-145 or T47D cells, but all cell types showed significant increases in ARE-Luc activity (P < 0.01). Co-expression of wtBRCA1 with NRF2 caused a modest but significant increase in NRF2-stimulated NQO1-ARE-Luc activity in DU-145 and T47D cells (36 to 50%; P < 0.01) but caused a much larger increase in NRF2-stimulated activity (4.3-fold) in MCF-7 cells (P < 0.001; Fig. 4,B). In plasmid dose-response studies of MCF-7 cells, increases in NQO1-ARE-Luc reporter activity were detectable at 10 to 50 ng per well of wtBRCA1 and were half maximal by 100 ng per well (Fig. 4 C). An 8-fold stimulation of reporter activity was achieved at our standard plasmid dose (0.25 μg per well), and the stimulation reached a maximum of 11-fold at 2.5 μg of plasmid per well.

In contrast to wtBRCA1, BRCA1-siRNA (but not a control-siRNA) significantly decreased basal and NRF2-stimulated NQO1-ARE-Luc activity (Fig. 4, D and E). Decreases in basal and NRF2-stimulated activity ranged from 60 to 100% (P < 0.001). These findings suggest that BRCA1 regulates NRF2 activity through the ARE over a wide range of intracellular BRCA1 protein levels. Next, we determined the BRCA1 structural requirements for stimulation of ARE activity with a series of previously described expression vectors encoding truncated or mutant BRCA1 proteins (Fig. 4,F; refs. 12, 14, 15). These studies revealed that COOH-terminal truncations (Δ BamHI, Δ KpnI, and Δ EcoRI) or mutations (5382insC and C5365G) of BRCA1 retained the ability to stimulate reporter activity, but a point mutation in the NH₂-terminal RING domain (T300G) or a NH₂-terminal truncation abrogated the ability of BRCA1 to stimulate activity (Fig. 4 F). These findings suggest that the NH₂ terminus of BRCA1 is both necessary and sufficient to stimulate NQO1-ARE-Luc activity.

Fig. 4,G shows BRCA1 protein levels in MCF-7 and DU-145 cells experimentally manipulated to over- or underexpress BRCA1. As noted earlier (14), DU-145 cells show low basal BRCA1 expression that is significantly increased by stable (Fig. 1,A) or transient (Fig. 4,G) expression of exogenous wtBRCA1. In this (Fig. 4,G) and a prior study (15), basal BRCA1 protein levels in MCF-7 cells were significantly higher than in DU-145 cells were similar to or slightly less than those observed in wtBRCA1-transfected DU-145 cells. By way of comparison, BRCA1 levels in lymphoblastoid cell lines derived from a BRCA1 (185delAG) [R794] and a BRCA2 (6174delT) [R1041] mutation carrier were generally similar to the BRCA1 levels observed in untransfected MCF-7 cells or in wtBRCA1-transfected DU-145 cells (Fig. 4 G). For both MCF-7 and DU-145 cells, BRCA1-siRNA abolished or nearly abolished BRCA1 protein expression, whereas the control-siRNA had little or no effect on protein expression. The physiologic significance of these findings is considered in the Discussion.

We used a dominant negative NRF2 expression vector (20) to determine whether the endogenous NRF2 protein is required for BRCA1 to stimulate the antioxidant response. Here, we found that co-expression of DN-NRF2 ablated basal NQO1-ARE-Luc activity (some of which is dependent upon endogenous BRCA1), as well as wtBRCA1-stimulated activity (P < 0.001; Fig. 5,A). Consistent with these findings, transient expression of the DN-NRF2 sensitized MCF-7 cells to H₂O₂ and abrogated the ability of wtBRCA1 to protect MCF-7 cells against to H₂O₂ (P < 0.01; Fig. 4 B).

DNA microarray analyses of BRCA1 overexpressing and Brca1-mutant cells identified various categories of genes positively regulated by BRCA1, including genes involved in transcription, stress responses, signal transduction, DNA replication and repair, cell proliferation, metabolism, and other processes. A number of these findings were confirmed by using independent mRNA assays. We identified potential BRCA1-regulated genes consistent with its known functions in DNA repair and cell cycle regulation, e.g., deleted in split hand/split foot syndrome 1 (DSS1) [a BRCA2-interacting protein required for homologous recombination (32)], a DNA cross-link repair gene (Dclre1a, also called SNM1), and several cell cycle regulatory genes [e.g., CDKN2C (p18), G0S2, and Gspt1]. Brca1-deficient cells, which exhibit a defect in centrosome function (16), showed decreased expression of a major centrosome protein, centrosomin A (Csma), three mitotic kinases (Stk2, Stk10, and Clk), and a chromosome segregation gene (Ttc3).

Although the functional categorization of genes is somewhat arbitrary (many genes fit into more than one category), it appeared that overexpression (mutation) of BRCA1 led to increased (decreased) expression of a sizeable group of genes involved in the response to stress, including the antioxidant response, detoxification of xenobiotics, and drug metabolism. Genes up-regulated in BRCA1 overexpressing cells include GSTs and peroxidases (e.g., MGST1/2, GSTT1, GSTZ1, and GPX3), oxidoreductases (e.g., NQO1 and ME2), alcohol and aldehyde dehydrogenases (e.g., ADH5 and ALDH7), a paraoxonase (PON2), an AhR-like protein (MOP2), and other antioxidant proteins.

Consistent with these findings, Brca1-deficient MEFs showed decreased expression of stress-response genes, including Gsta2, Gpx3, Nrf2, Sod1, Ahr, and Sepp1, a selenoprotein that mediates protection against oxidative stress (33). Mice deficient for the major antioxidant response transcription factor Nrf2 exhibited increased susceptibility to hyperoxic lung damage, a reduced expression of several ARE-dependent phase II drug-metabolizing enzymes, increased sensitivity to carcinogens, and decreased protection against carcinogenesis by chemoprevention agents (34, 35). Our findings suggest that BRCA1 regulates the expression of several genes that are known to be regulated by NRF2 and/or to contain AREs in their regulatory regions. A recent study identified Nrf2-regulated genes for which basal and/or inducible expression was increased in the small intestine of Nrf2^+/+ relative to Nrf2^−/− mice (31). BRCA1 increased the expression of some of these Nrf2-regulated genes (NQO1, MGST1/2, Gsta2, G6PD, and ME2). BRCA1 also induced (and Brca1 mutation inhibited) expression of a glutathione peroxidase (GPX3), other isoforms of which are down-regulated in Nrf2^−/− cells (31, 34). These results suggest an overlap in the genes regulated by BRCA1 versus NRF2.

Consistent with its ability to up-regulate antioxidant gene expression, BRCA1 overexpression conferred resistance, whereas BRCA1 mutation or underexpression conferred sensitivity to two different oxidizing agents (H₂O₂ and paraquat). Because peroxides and superoxide, which is generated by paraquat, are detoxified by distinct enzymatic pathways (e.g., those involving catalase versus superoxide dismutase, respectively), these findings suggest that BRCA1 may stimulate more than one antioxidant defense pathway. However, this remains to be demonstrated. BRCA1 is classified as a caretaker gene based on the findings that BRCA1 mutations lead to chromosomal instability and defects in DNA repair (reviewed in ref. 5). The ability of BRCA1 to protect against oxidative stress may contribute to its caretaker function because reactive oxygen species (e.g., H₂O₂, O₂⁻, and hydroxyl radicals) generated endogenously in mitochondria and other organelles can cause DNA damage (oxidation). In addition to endogenous reactive oxygen species, which contribute to carcinogenesis (36), many DNA-damaging agents and xenobiotics cause oxidative stress, resulting in DNA damage, protein oxidation, and lipid peroxidation. Some of these lesions are detoxified by BRCA1-regulated genes (e.g., GSTs, GPXs, oxidoreductases, and paroxonases).

Consistent its ability to up-regulate antioxidant genes and protect against oxidants, wtBRCA1 attenuated the loss of GSH due to H₂O₂, thus helping stressed cells to maintain their redox balance. It is not clear how BRCA1 stimulates GSH production under oxidizing conditions. GSH is produced via two processes: (a) conversion of GSSG to GSH by glutathione reductase, which requires NADPH; and (b) de novo synthesis via γ-glutamylcysteine synthetase (31). Both glucose-6-phosphate dehydrogenase (G6PD) and malic enzyme (ME2), which are up-regulated by wtBRCA1, stimulate NADPH formation (process 1). Although γ-glutamylcysteine synthetase was not on the list of BRCA1-regulated genes, γ-glutamylcysteine synthetase in an NRF2/ARE-regulated gene, and BRCA1 stimulates NRF2/ARE activity. Finally, we reported recently that BRCA1 up-regulates the expression of the small heat shock protein HSP27 (19), which functions to maintain the redox balance, possibly by helping to maintain the activity of cellular redox enzymes (37). Small heat shock proteins such as HSP27 protect cells against oxidative stress, in part, by enhancing G6PD activity (37), which helps to generate the reducing power for conversion of GSSG to GSH. In this regard, our findings suggest that G6PD may be a transcriptional target of BRCA1. The role of HSP27 and other small heat shock proteins in the BRCA1-mediated protection against oxidative and generation of GSH in stressed cells is a subject for additional investigation.

We have established the principle that BRCA1 stimulates ARE signaling and NRF2 transcriptional activity, although the extent of stimulation varied in different cell lines. The stimulation of NQO1-ARE-Luc activity and protection against oxidative stress by wtBRCA1 were ablated by a DN-NRF2, suggesting that NRF2 may be downstream of BRCA1 in an antioxidant response pathway. Although DN-NRF2 also abolished basal ARE-Luc activity and sensitized cells to oxidative stress in the absence of exogenous wtBRCA1, the siRNA experiments suggest that endogenous BRCA1 contributes to basal ARE-Luc activity and resistance to oxidative stress. Hence, some of the effects of DN-NRF2 could be due to pathways downstream of the endogenous BRCA1.

The NH₂ terminus of BRCA1, including the RING domain, was necessary and sufficient to stimulate ARE signaling. A similar pattern (i.e., requirement for the NH₂ terminus but not the COOH-terminus of BRCA1) was observed for stimulation of the HSP27 promoter activity and TERT promoter activity by BRCA1 (15, 19). The siRNA studies suggest the relevance of our findings to sporadic cancers in which BRCA1 expression is reduced, but the implications for BRCA1 mutant cancers are unclear at present because we do not know the extent to which BRCA1 mutant proteins are expressed in human cancers. Although most cancer-associated BRCA1 mutations are protein truncating mutations that should retain the ability to stimulate ARE signaling, the ability to stimulate ARE signaling would be compromised if the mutant BRCA1 proteins are underexpressed or rapidly degraded. Moreover, one cancer-associated BRCA1 mutation, T300G (which affects the NH₂-terminal RING domain), abrogated the ability of BRCA1 to stimulate ARE-Luc activity. Our previous work indicates that the BRCA1-T300G mutant protein is stable and is well expressed (12, 15).

The ability of BRCA1 to protect against oxidant toxicity may be due, in part, to stimulation of antioxidant defenses (e.g., increased expression of antioxidant genes, increased production of GSH, and stimulation of NRF2 transcriptional activity). However, because DNA is a major target of oxidizing agents, the ability of BRCA1 to stimulate DNA repair (5) could also contribute to its cytoprotective activity. The extent to which BRCA1 functions to prevent DNA damage by enhancing detoxification of peroxides and superoxides as opposed to repairing established DNA lesions remains to be determined.

Our studies used three different models to investigate BRCA1 function: (a) overexpression (via stable or transient expression of exogenous wtBRCA1); (b) underexpression (via RNA interference); and (c) inactivation via gene deletion (Δ exon 11, which removes most of the Brca1 protein). Relative to model 1 (overexpression), studies of mice indicate that Brca1 is particularly highly expressed in the mammary gland in proliferating cells undergoing differentiation during puberty and pregnancy (38, 39). It has been suggested that the BRCA1 may play a particularly important role in preventing tumors during specific windows of time (e.g., puberty and pregnancy) in which it is highly expressed. The BRCA1 overexpression model might reflect these time periods when BRCA1is normally overexpressed. This expression pattern may also be reflected in vitro because BRCA1 expression is greatly increased when cultured mammary epithelial cells are forced to undergo differentiation (e.g., by the use of a hormonal mixture; refs. 40, 41). It remains to be proved whether these periods in which BRCA1 is highly expressed are directly related to its tumor suppressor function.

The role of endogenous BRCA1 in mediating protection against oxidative stress and/or stimulating NRF2 activity was documented in two different models (deletion of exon 11 in MEFs and knockdown of BRCA1 protein levels with an siRNA). The exon 11 deletion model may reflect the situation in BRCA1 mutant cancers, where the wild-type BRCA1 is usually lost (5, 42). We also note that the BRCA1 Δ exon 11 protein corresponds to a naturally occurring splice variant of BRCA1 in humans and mice (25, 43). As noted earlier, BRCA1 expression is often decreased or absent in sporadic breast cancers that do not exhibit a BRCA1 mutation (3, 4). This loss of BRCA1 expression may be due, in part, to epigenetic causes (hypermethylation of the BRCA1 promoter) and/or haploinsufficiency (loss of one BRCA1 allele; refs. 4, 5, 44). Regardless of the etiology, model 3 (BRCA1-siRNA) may reflect the underexpression of BRCA1 commonly observed in sporadic breast and ovarian cancers. The finding that BRCA1 can modulate various aspects of antioxidant defense over a wide range of BRCA1 expression levels is consistent with a physiologic role for BRCA1 in this pathway.

Taken together, our findings suggest a novel mechanism by which BRCA1 may prevent cancer development by enhancing antioxidant defenses, thereby protecting cells against damage caused by exogenous and/or endogenous reactive oxygen species. However, a definitive linkage between BRCA1-mediated protection against oxidative stress and tumor suppression remains to be demonstrated. They also suggest that in addition to its established roles in the repair of DNA damage, BRCA1 may prevent DNA damage due to ionizing radiation and other sources through the detoxification of reactive oxygen species, although this needs to be proven. Finally, these studies suggest a collaboration between BRCA1 and a transcription factor (NRF2) that functions to mobilize the cell’s antioxidant machinery.