Identification of the NF-E2-related Factor-2-dependent Genes Conferring Protection against Oxidative Stress in Primary Cortical Astrocytes Using Oligonucleotide Microarray Analysis*

The antioxidant responsive element (ARE) mediates transcriptional regulation of phase II detoxification enzymes and antioxidant proteins such as NAD(P)H:quinone oxidoreductase (NQO1), glutathione S-transferases, and glutamate-cysteine ligase. In this study, we demonstrate that NF-E2-related factor-2 (Nrf2) plays a major role in transcriptional activation of ARE-driven genes and identify Nrf2-dependent genes by oligonucleotide microarray analysis using primary cortical astrocytes from Nrf2+/+ and Nrf2−/− mice. Nrf2−/−astrocytes had decreased basal NQO1 activity and no induction bytert-butylhydroquinone compared with Nrf2+/+ astrocytes. Similarly, both basal and induced levels of human NQO1-ARE-luciferase expression in Nrf2−/− astrocytes were significantly lower than in Nrf2+/+ astrocytes. Furthermore, human NQO1-ARE-luciferase expression in Nrf2−/−astrocytes was restored by overexpression of Nrf2, whereas ARE activation in Nrf2+/+ astrocytes was completely blocked by dominant-negative Nrf2. In addition, we observed that Nrf2-dependent genes protected primary astrocytes from H2O2- or platelet-activating factor-induced apoptosis. In support of these observations, we identified Nrf2-dependent genes encoding detoxification enzymes, glutathione-related proteins, antioxidant proteins, NADPH-producing enzymes, and anti-inflammatory genes using oligonucleotide microarrays. Proteins within these functional categories are vital to the maintenance and responsiveness of a cell defense system, suggesting that an orchestrated change in gene expression via Nrf2 and the ARE gives a synergistic protective effect against oxidative stress.

NF-E2-related factor-2 (Nrf2) is a basic leucine zipper transcription factor that can bind to an NF-E2/AP-1 repeat sequence in the promoter of the ␤-globin gene (13). Nrf2 has two kinds of binding partners, a cytoplasmic repressor and multiple nuclear binding partners. Itoh et al. (14,15) have demonstrated that Nrf2 is sequestered in the cytoplasm by its repressor Keap1 (mouse), released under conditions of oxidative stress, and translocated into the nucleus. This cytoplasmic repressor of Nrf2 was also identified in human and rat (14,16,17). The suggested binding partners that have been demonstrated to bind with Nrf2 consist of other basic leucine zipper proteins such as small Maf (18,19), Jun (20), activating transcription factor-4 (21), and cAMP-responsive element-binding proteinbinding protein (22).
The function of Nrf2 and its downstream proteins has been shown to be important for protection against oxidative stressor chemical-induced cellular damage in liver (36,37) and lung (38) as well as for prevention of cancer formation in the gas-trointestinal tract (39,40) and promotion of the wound-healing process (41). In addition, many chronic neurodegenerative diseases (i.e. Parkinson's disease and Alzheimer's disease) are thought to involve oxidative stress as a component contributing to the progression of the disease. The regulation and cellspecific expression of these genes in cells derived from brain could therefore be important for understanding how to protect neural cells from oxidative stress. One of the Nrf2-dependent ARE-driven genes, NQO1, has been demonstrated to play an important role in protecting cells against oxidative stress (42)(43)(44). Interestingly, overexpression of NQO1 and one GST isoenzyme does not protect N18-RE-105 rodent neuroblastoma cells from free radical-mediated toxicity (45), although tert-butylhydroquinone (tBHQ) treatment, which up-regulates a battery of ARE-driven genes, protects N18-RE-105 cells from glutamate toxicity (43). These observations imply that the coordinate upregulation of ARE-driven genes, not one or two genes, is more efficient in protecting cells from oxidative damage. A recent study identified the ARE-driven genes including NQO1 that are responsible for protecting IMR-32 human neuroblastoma cells from H 2 O 2 -induced apoptosis (32,33). Therefore, Nrf2, which mediates transcription of ARE-driven genes, is presumably the driving force behind increasing a cluster of protective genes that play an important role in cellular defense against oxidative stress.
In the central nervous system, astrocytes have been shown to express many of these protective ARE-driven genes and AREdriven human placental alkaline phosphatase in primary cortical neuronal cultures derived from transgenic reporter mice (34). To further understand how Nrf2 contributes to the regulation of ARE-driven genes in astrocytes and how expression of these genes affects the sensitivity of astrocytes to oxidative stress, we compared primary cortical astrocyte cultures derived from Nrf2 ϩ/ϩ and Nrf2 Ϫ/Ϫ mice. Astrocytes were treated with tBHQ to induce nuclear translocation of Nrf2 leading to ARE activation and H 2 O 2 or platelet-activating factor (PAF) (46) to determine differential sensitivity. To understand how Nrf2-dependent genes are associated with this differential sensitivity, we performed oligonucleotide microarray analysis.

EXPERIMENTAL PROCEDURES
Nrf2 Knockout Mice-Nrf2 knockout mice were generated by replacing the basic leucine zipper domain with the lacZ reporter construct as described previously (47).
NQO1 Activity-Endogenous NQO1 enzymatic activity was determined by a colorimetric method for whole cell extracts (with menadione as a substrate) (48) and histochemistry for fixed cultures (LY 83583 as a substrate) (34) as described previously.
Western Blotting-For glutamate-cysteine ligase modifier subunit (GCLM) and glutamate-cysteine ligase catalytic subunit (GCLC) Western blotting, 50 g of whole cell extracts (2, 3) were used. Representative Western blots are shown in the figures.
Nrf2-dependent ARE Activation-hNQO1-ARE-luciferase gene expression and endogenous NQO1 activity were determined in tBHQ-treated Nrf2 Ϫ/Ϫ and Nrf2 ϩ/ϩ astrocytes. In Nrf2 Ϫ/Ϫ astrocytes, basal ARE-luciferase reporter gene expression was markedly decreased, and there was no induction of reporter gene expression by tBHQ compared with Nrf2 ϩ/Ϫ and Nrf2 ϩ/ϩ astrocytes ( Fig. 2A). Similarly, both basal and induced levels of endogenous NQO1 activity in Nrf2 Ϫ/Ϫ astrocytes were significantly lower than in Nrf2 ϩ/Ϫ and Nrf2 ϩ/ϩ astrocytes (Fig. 2B), implying that Nrf2 plays an important role in both basal and induced ARE-driven gene expression in mouse primary cortical astrocytes. In addition, histochemical detection of NQO1 activity confirmed the Nrf2-dependent NQO1 gene expression. The NQO1 staining of vehicle-treated Nrf2 ϩ/ϩ astrocytes was significantly higher than that of vehicle-treated Nrf2 Ϫ/Ϫ cells (Fig. 2C, upper left panel versus lower left panel), and tBHQ increased NQO1 staining intensity only in Nrf2 ϩ/ϩ astrocytes (lower left panel versus lower right panel). To further investigate the role of Nrf2 in ARE activation, we transfected Nrf2 Ϫ/Ϫ astrocytes with an Nrf2 overexpression vector to restore ARE activation and Nrf2 ϩ/ϩ astrocytes with dominantnegative Nrf2 to inhibit ARE activation. Dominant-negative Nrf2 (N-terminally truncated Nrf2) inhibits endogenous Nrf2 function by occupying and limiting its binding partners and DNA-binding sites (5). Indeed, overexpression of Nrf2 led to dramatic ARE activation in Nrf2 Ϫ/Ϫ astrocytes (Fig. 3A). tBHQ did not activate the ARE in pEF-transfected Nrf2 Ϫ/Ϫ astrocytes. However, tBHQ did activate the ARE in Nrf2-overexpressing Nrf2 Ϫ/Ϫ astrocytes in a dose-dependent manner (Fig.  3A). Finally, dominant-negative Nrf2 blocked both basal and induced ARE activation by tBHQ in Nrf2 ϩ/ϩ astrocytes (Fig.  3B).
Differential Sensitivity to H 2 O 2 -and PAF-induced Cytotoxicity-Nrf2 regulates ARE-driven genes involved in detoxification and antioxidant potential. Therefore, we hypothesized that FIG. 4. Protective role of Nrf2-dependent genes. Nrf2 Ϫ/Ϫ (knockout (KO)) and Nrf2 ϩ/ϩ (wild-type (WT)) primary astrocytes were pretreated with vehicle (V; 0.01% Me 2 SO) or tBHQ (T; 50 M, 48 h), followed by H 2 O 2 (4 h) or PAF (24 h). To measure cell viability, 3-(4,5dimethylthiazol-2-yl)-5-3-carboxymethoxyphenyl)tetrazolium salt assay was used (A and B), and to measure apoptotic nuclei, TUNEL staining was performed (C) according to the manufacturers' protocols. Scale bars ϭ 20 m. Treatment with tBHQ alone (tBHQ Ϫ vehicle) did not induce cytotoxicity in either Nrf2 ϩ/ϩ or Nrf2 Ϫ/Ϫ astrocytes (data not shown).  Nrf2 Ϫ/Ϫ astrocytes would be more sensitive to oxidative stress compared with Nrf2 ϩ/ϩ astrocytes due to reduced levels of detoxification and antioxidant potential. To investigate this differential sensitivity, we pretreated Nrf2 Ϫ/Ϫ and Nrf2 ϩ/ϩ astrocytes with tBHQ (50 M, 48 h) to increase ARE-driven gene expression and then with H 2 O 2 to investigate differential sensitivity. Also, we used the potent inflammatory agent PAF (46) to investigate the anti-inflammatory effect of Nrf2. As shown in Fig. 4A, vehicle-pretreated Nrf2 Ϫ/Ϫ astrocytes were more sensitive to H 2 O 2 -induced cytotoxicity compared with vehicle-pretreated Nrf2 ϩ/ϩ astrocytes. Furthermore, tBHQ pretreatment significantly increased cell viability in Nrf2 ϩ/ϩ (but not Nrf2 Ϫ/Ϫ ) astrocytes (Fig. 4A). Similarly, Nrf2 Ϫ/Ϫ astrocytes were more sensitive to PAF compared with Nrf2 ϩ/ϩ astrocytes, and tBHQ pretreatment protected only Nrf2 ϩ/ϩ astrocytes (Fig. 4B). TUNEL staining and the corresponding phase-contrast microscope pictures confirmed this differential sensitivity. As shown in Fig. 4C, the numbers of TUNEL-positive cells in H 2 O 2 -or PAF-treated Nrf2 Ϫ/Ϫ astrocytes were greater than in the corresponding Nrf2 ϩ/ϩ astrocytes. Although tBHQ did not decrease the number of TUNEL-positive cells in Nrf2 Ϫ/Ϫ astrocytes, tBHQ pretreatment decreased TUNEL-positive cells in both H 2 O 2 -and PAF-treated Nrf2 ϩ/ϩ astrocytes (data not shown). Consistent with the TUNEL data, H 2 O 2 and PAF induced more caspase-3 activation in Nrf2 Ϫ/Ϫ astrocytes than in Nrf2 ϩ/ϩ astrocytes (data not shown). These observations suggest that Nrf2 Ϫ/Ϫ astrocytes are more sensitive to oxidative stress and inflammation compared with Nrf2 ϩ/ϩ astrocytes and that coordinate up-regulation of ARE-driven genes by tBHQ further protects Nrf2 ϩ/ϩ cells from H 2 O 2 -and PAFinduced cytotoxicity. Identification of the Nrf2-dependent Genes-To identify the Nrf2-dependent genes that play an important role in protecting astrocytes from H 2 O 2 -and PAF-induced apoptosis, we performed oligonucleotide microarray analysis. The genes changed by Nrf2 and/or tBHQ were identified by four comparisons, as depicted in Fig. 5A. tBHQ increased 16 genes (stromal cellderived factor, Induced in fatty liver dystrophy-2, histones 1H2B and H2A, histone H1, TG-interacting factor, Thy-1.2 glycoprotein, Lumican, cysteine-and histidine-rich-1, ectonucleotide pyrophosphatase/phosphodiesterase-2, proteasome 26 S subunit, and six expressed sequence tags) and decreased 27 genes in Nrf2 Ϫ/Ϫ astrocytes (comparison I in Fig. 5A), suggesting that the changes in expression of these genes are Nrf2independent. Genes changed by Nrf2 in the absence of tBHQ (comparison II) are listed in Table I, and genes changed by tBHQ in the presence (comparison III) or absence (comparison I) of Nrf2 are listed in Table II. Interestingly, the majority of the genes increased by tBHQ in Nrf2 ϩ/ϩ astrocytes (97.6%) were not changed by tBHQ in Nrf2 Ϫ/Ϫ astrocytes ( Fig. 5B and Table II), suggesting that most of the tBHQ-induced genes are Nrf2-dependent. Only five genes (Induced in fatty liver dystrophy-2, ectonucleotide pyrophosphatase/phosphodiesterase-2, TG-interacting factor, Thy-1.2 glycoprotein, and expressed sequence tag AW124185) were increased by tBHQ in both Nrf2 Ϫ/Ϫ and Nrf2 ϩ/ϩ astrocytes. The gene list in comparison IV includes most of the genes changed in comparisons I and III.
The major functional categories of Nrf2-dependent genes are 1) detoxification, 2) antioxidant/reducing potential, 3) growth, and 4) defense/immune/inflammation (Tables I and II). Clearly, the oligonucleotide microarray data verify that Nrf2 is important for the expression of NQO1 and other ARE-driven genes such as GSTs. Interestingly, cytochrome P450 1B1 was the only member of the cytochrome P450 family that appeared to be Nrf2-dependent in primary astrocytes (Tables I and II). Another evident Nrf2-dependent gene category is antioxidant/ reducing potential. As shown in Tables I and II, many glutathione-related proteins (GCLM, GCLC, GSTs, glutathione reductase, and glutathione peroxidase), antioxidant proteins (TXNRD1, thioredoxin, peroxiredoxin, HO-1, ferritin, catalase, and superoxide dismutase), and genes involved in NADPH production (glucose-6-phosphate dehydrogenase, malic enzyme, transaldolase, and transketolase) were identified as Nrf2-dependent genes. Furthermore, oligonucleotide microarray analysis revealed that many defense/immune/inflammation-related genes (i.e. cathepsin, complements, lipopolysaccharide-binding protein, and PAF acetylhydrolase), metabolic enzymes (i.e. lipoprotein lipase and esterase), growth factors (i.e. platelet-derived growth factor and nerve growth factor), and signaling proteins (i.e. protein kinase C) were regulated in an Nrf2-dependent manner (Tables I and II). Clearly, oligonucleotide microarray data showed that Nrf2-dependant antioxidant and anti-inflammatory genes play an important role in protecting primary astrocytes from the H 2 O 2 -and PAF-induced apoptosis observed in this study.
Verification of Microarray Data-To verify the oligonucleotide microarray data, we performed reverse transcription-PCR and Western blot analysis for selected genes. As shown in Fig.  6 (A and B), the expression levels of the selected genes observed by reverse transcription-PCR were consistent with the oligonucleotide microarray analysis results, verifying the change in Nrf2-dependent genes identified by the oligonucleotide microarray. Also, Western blot analysis (Fig. 6C) and GSH quantification data (Fig. 6D) showed that Nrf2 plays an important role in both GCLM/GCLC expression and GSH synthesis, as expected from the reverse transcription-PCR and oligonucleotide microarray data. DISCUSSION In this study, we demonstrated that the basic leucine zipper transcription factor Nrf2 plays a critical role in both basal and induced gene expression of NQO1 in primary cortical astrocytes. Overexpression of wild-type Nrf2 restored the basal expression and activation of ARE by tBHQ in Nrf2 Ϫ/Ϫ cells, and dominant-negative Nrf2 significantly decreased both basal expression and activation of ARE by tBHQ in Nrf2 ϩ/ϩ astrocytes. The reduced expression and lack of ARE activation in Nrf2 Ϫ/Ϫ astrocytes directly correlate with an increased sensitivity to H 2 O 2 -and PAF-induced cytotoxicity compared with Nrf2 ϩ/ϩ astrocytes. Finally, the genes responsible for conferring protection against oxidative stress or inflammation were identified by oligonucleotide microarray analysis. The major functional categories are detoxification enzymes, antioxidant proteins, NADPH-producing proteins, growth factors, defense/immune/  In the present study, tBHQ did not increase Nrf2 expression levels, but induced nuclear translocation of Nrf2 (data not shown), suggesting that ARE activation by tBHQ is mediated by nuclear translocation of Nrf2, not by induction of Nrf2 gene expression in primary astrocytes. tBHQ did, however, increase expression of binding partners of Nrf2 (i.e. MafG and activating transcription factor-4) in Nrf2 ϩ/ϩ astrocytes (Table II). Maf proteins have been shown to regulate ARE activation negatively or positively depending on cell types and genes (27,28,50,51), and activating transcription factor-4 has been demonstrated to bind to Nrf2, leading to HO-1 gene expression (21). In addition, CCAAT/enhancer-binding protein-␤ was increased by Nrf2 and tBHQ in Nrf2 ϩ/ϩ astrocytes. CCAAT/enhancerbinding protein-␤ has been shown to mediate negative regulation of rat GST-Ya expression (52). Finally, in contrast to the increased expression of KIAA0132 (human homolog of Keap1) by tBHQ in IMR-32 cells (32), Keap1 was not changed by either Nrf2 or tBHQ in mouse primary astrocytes (Tables I and II). These observations suggest a possible balancing mechanism between positive and negative regulation in Nrf2-mediated gene expression and that the role and regulation of other binding partners of Nrf2 are dependent on the cell type and/or genes being studied.
The function of a number of Nrf2-dependent genes is dependent on GSH. GSTs catalyze the nucleophilic addition of GSH to an electrophilic group of a broad spectrum of xenobiotic compounds (54). Other GSH-dependent enzymes (i.e. glutathione peroxidase, peroxiredoxin, and glutathione reductase) were also increased in an Nrf2-dependent manner. Glutathione per-oxidase and peroxiredoxin metabolize H 2 O 2 , generating H 2 O and oxidized GSH (GSSG), and glutathione reductase regenerates reduced GSH. Ideally, in association with an increased utilization of GSH, there would also be an increased production of GSH. The rate-limiting step in the GSH biosynthesis is mediated by GCLM/GCLC. In this study, solute carrier family-1, glycine transporter, GCLM, and GCLC were shown to be Nrf2-dependent genes. The coordinate regulation of these genes can have a synergistic effect in the maintenance of GSH levels as well as detoxification of reactive intermediates (Fig.  7A).
Another cluster of genes including superoxide dismutase and HO-1 are very important for cellular defense against oxidative stress. Superoxide dismutase detoxifies superoxide, resulting in H 2 O 2 , and HO-1 generates a potent radical scavenger, bilirubin. However, superoxide dismutase and HO-1 can induce more oxidative stress because they increase the cellular concentrations of H 2 O 2 and free iron, which together can generate hydroxyl radical through the Fenton reaction. For complete detoxification of superoxide, H 2 O 2 should be further metabolized to H 2 O by glutathione peroxidase, catalase, or peroxiredoxin. Catalase directly detoxifies H 2 O 2 , whereas peroxiredoxin uses GSH (Fig. 7A) and/or thioredoxin as an electron donor for peroxidation of H 2 O 2 , resulting in generation of GSSG and oxidized thioredoxin, respectively (Fig. 7B). GSSG and oxidized thioredoxin are converted to their reduced forms by glutathione reductase and TXNRD1. Oligonucleotide microarray data showed that superoxide dismutase, catalase, peroxiredoxin, thioredoxin, and TXNRD1 are transcriptionally regulated through an Nrf2-dependent mechanism. In addition, proper management of free iron is also important for minimizing oxidative stress, and this can be best achieved by ferritin. Ferritin converts Fe 2ϩ to Fe 3ϩ (ferroxidase activity) and sequesters it, thereby preventing Fe 2ϩ from participating in the Fenton reaction. Thus, up-regulation of HO-1 together with ferritin is a way to increase antioxidant potential while minimizing hydroxyl radical formation. Based on these observations, we speculate that increased expression of these genes can dramatically increase the efficiency of detoxification of reactive oxygen species. Also, the genes depicted in Fig. 7B provide a molecular mechanism by which tBHQ-treated Nrf2 ϩ/ϩ astrocytes are resistant to H 2 O 2 -induced apoptosis.
Finally, NQO1, glutathione reductase, and TXNRD1 are important in detoxifying quinones and maintaining the cellular redox balance. One common feature of these proteins is that they use NADPH as an electron donor. So, for efficient detoxification and maintenance of cellular redox status, it would be beneficial to up-regulate these proteins together with the appropriate reducing potential (NADPH) to support enzymatic reactions. Glucose-6-phosphate dehydrogenase/malic enzyme can directly generate NADPH, and transketolase/transaldolase can increase NADPH production by regenerating substrates for glucose-6-phosphate dehydrogenase. Oligonucleotide microarray data showed that NQO1, glutathione reductase, TXNRD1, glucose-6-phosphate dehydrogenase, malic enzyme, transketolase, and transaldolase are Nrf2-dependent genes (Fig. 7C). These Nrf2-dependent genes would also contribute significantly to a cell's detoxification potential and cellular redox balance. Together, these coordinately regulated gene clusters presented in Fig. 7 strongly support the hypothesis that Nrf2dependent gene expression is central to efficient detoxification of reactive metabolites and reactive oxygen species as well as a cell's ability to deal with stress such as inflammation.
Can these changes in astrocytes protect neurons from oxidative stress-induced apoptosis? Astrocytes have been suggested to interact with neurons and to confer neuronal protection. It has been demonstrated in numerous neuronal culture systems that the survival of neurons is significantly enhanced by astrocytes (55)(56)(57). They can promote neuronal survival by removing excitotoxins (i.e. glutamate) from the synapse, modulating antioxidant levels (i.e. GSH), and secreting trophic factors (i.e. glial-derived neurotrophic factor) (58 -60). In support of this idea, Nrf2-dependent detoxification and antioxidant proteins in astrocytes can play a role in protecting neurons. However, genetic changes in neurons associated with increased expression of ARE-driven genes in astrocytes could also contribute to an overall protective mechanism. The extent to which this  Fig. 5A). The genes shown here were not changed by tBHQ in the absence of Nrf2 (comparison I in Fig. 5A). Primary astrocytes were treated with vehicle (0.01% Me 2 SO) or tBHQ (50 M) for 48 h. C, whole cell extracts were prepared, and Western blot analysis for GCLC and GCLC was performed as described under "Experimental Procedures." D, total glutathione levels (GSH ϩ GSSG) were measured as described under "Experimental Procedures." Each data bar represents the mean Ϯ S.E. (n ϭ 6). intercellular communication is required and the specific genetic remodeling in the neurons versus the astrocytes in a co-culture system remain to be determined. Preliminary data from our laboratory suggest that there are unique changes in both astrocytes and neurons that, when combined, may be responsible for protecting neurons from oxidative stress. 3 In summary, oxidative stress and reactive metabolites can induce apoptosis or programmed cell death. Programmed cell death can be prevented in many ways, such as addition of external growth factors, antioxidant supplementation, and inhibition of apoptotic signaling pathways. Here we present an alternative in that the coordinate up-regulation of Nrf2-dependent genes provides a way to protect cells through genetic remodeling, a process referred to as programmed cell life. We hypothesize that increased activation of programmed cell life pathways can balance programmed cell death and that, in combination with other techniques known to prevent programmed cell death, may be a powerful tool in controlling progressive neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, and amyotrophic lateral sclerosis. Currently, we are focused on evaluating the neuroprotective role of Nrf2-dependent genes in vivo by crossing Nrf2 knockout mice with established transgenic models representing human neurological disorders.