Nrf2 Mediates the Induction of Ferritin H in Response to Xenobiotics and Cancer Chemopreventive Dithiolethiones*

,

Ferritin is a 480-kDa intracellular protein that can store up to 4500 atoms of iron (1). The protein consists of 24 subunits of the H and L chain type (2). The ratio of subunits within the ferritin protein varies widely by tissue type; the ratio can be further modulated by environmental signals, including cytokine stimulation, stress signals, and disease state (3,4). The H chain has ferroxidase activity (5), whereas the L subunit is responsible for iron nucleation and protein stabilization (6). Because iron functions as a catalyst in the formation of oxygen free radicals, storage of iron by ferritin may serve a cytoprotective function (2). Several laboratories have demonstrated an induction of ferritin by oxidants and pro-oxidant xenobiotics, including agents such as hydrogen peroxide (7), t-BHQ 1 (7), sodium arsenite (8), phorone (9), carbon tetrachloride (10), and aqueous extracts of cigarette smoke (11). Activation of ferritin by these agents is thought to be a stress response mechanism that contributes to limiting cellular and organismal damage from these xenobiotics.
Regulation of transcription at the EpRE/ARE is incompletely understood. Early studies focused on members of the AP1 transcription factor family (38) as regulators of EpRE/ARE-dependent transcription (39 -42). More recently, compelling evidence for the involvement of members of the cap and collar family of transcription factors (43), particularly Nrf2, has been presented. For example, induction of cytoprotective enzymes, such as GST and NQO1, by the phenolic antioxidant butylated hydroxyanisole was lost in cells isolated from Nrf2 knockout mice (44). In addition, forced expression of cap and collar family members, such as Nrf1 and Nrf2, resulted in the induction of EpRE/ARE-dependent reporter gene expression (45)(46)(47)(48). Activation of Nrf2 following stimulation of cells with an inducer requires its dissociation from a cytosolic actin binding protein, Keap-1, and subsequent translocation to the nucleus (49,50). Release of Nrf2 from Keap-1 may be triggered by modification of reactive cysteine residues in Keap-1 (51) and/or post-translational modification of Nrf2 by protein kinases (52).
Our laboratory has identified an EpRE/ARE in the ferritin H gene that mediates the induction of ferritin H transcription in response to H 2 O 2 and t-BHQ (7). The ferritin H EpRE/ARE is 75 bp in length and is located ϳ4.1 kb from the transcription start site (7). It is comprised of the ferritin H basal enhancer (53), FER-1, and an AP1/NF-E2 consensus sequence located 8 bp 3Ј of FER-1. The basal enhancer, FER-1, is in turn composed of an element with close sequence similarity to both AP1 and NF-E2 consensus sequences (previously termed AP1-like element (53,54) and referred to in this report as AP1/NF-E2-like element), and a recognition sequence for the SP1/3 transcription factors (53,54). The AP1/NF-E2-like and the AP1/NF-E2 consensus sequence of the ferritin H EpRE/ARE are arranged in inverse repeat, and both of these elements are necessary for full induction of ferritin H by H 2 O 2 and t-BHQ (7). An EpRE/ ARE has also been identified in the murine ferritin L promoter (55). Ligation of this element to a luciferase reporter gene demonstrated that the ferritin L EpRE is functional as an enhancer element in HepG2 cells treated with t-BHQ (55).
Collectively, these results suggest that ferritin may constitute a component of the cytoprotective response induced by xenobiotics (electrophiles or polycyclic aromatic hydrocarbons) and candidate chemopreventive agents. However, the mechanism of ferritin induction by these agents is unknown. Here, we demonstrate that ferritins H and L are induced by oltipraz, D3T, and ␤-NF in fibroblasts and hepatic cells. Furthermore, we show that induction of ferritin occurs via an EpRE/ARE-dependent mechanism that requires Nrf2. These results link ferritin induction mechanistically to the chemopreventive response.
Chemicals-Oltipraz and D3T were provided by the Division of Cancer Prevention Repository of NCI, National Institutes of Health, and ␤-NF was obtained from Sigma. All compounds were dissolved in Me 2 SO and final Me 2 SO concentration in all treatment conditions was 0.4%.
Northern Blot Analysis-Total RNA was isolated from cells treated for 24 h with vehicle, oltipraz, D3T, and/or ␤-NF as described by Chirgwin et al. (58) or utilizing the TRIzol reagent (Invitrogen) according to the manufacturer's protocol. 10 -15 g of RNA were size-fractionated on 1.1% agarose/6.6% formaldehyde gels and transferred to an Immobilon Nyϩ nylon membrane (Millipore) by capillary transfer. DNA probes for both ferritin H (59) and L (60) were generated by random prime labeling and subsequently hybridized to the UV-crosslinked RNA blot in Quick Hyb solution (Stratagene) according to the manufacturer's protocol. Membranes were subjected to autoradiography; quantitation was performed using a PhosphorImager analyzer (model 445SI, Amersham Biosciences).
Western Blotting of Ferritin Induction-To assess ferritin H and ␤-actin protein levels, cytosolic extracts were prepared as previously described by Schreiber et al. (61). 50 g of protein was fractionated on 12%SDS-polyacrylamide gels, transferred to nitrocellulose, blocked with 5% nonfat dry milk in phosphate-buffered saline, washed, and incubated with a 1:1000 dilution of polyclonal rabbit anti-ferritin H peptide antibody (BIOSOURCE International) followed by a 1:200 dilution of goat anti-rabbit IgG conjugated to horseradish peroxidase (Bio-Rad). The blots were developed using the Enhanced Chemiluminescence System (Amersham Biosciences). To demonstrate equivalent protein loading, blots were washed and re-blotted using a 1:20,000 dilution of anti-␤ actin antibody (Sigma) followed by a 1:5,000 dilution of goat anti-mouse IgG conjugated to horseradish peroxidase (Calbiochem).
Transfection of Ferritin H-human Growth Hormone Reporter Gene Constructs and RNase Protection Assay-NIH3T3 cells were transfected in duplicate with 2 or 3 g of FH-hGH reporter gene constructs using LipofectAMINE Reagent (Invitrogen) according to the manufacturer's protocol. Cells were allowed to recover for 20 -24 h and treated with 70 M Oltipraz, 70 M D3T, or vehicle (Me 2 SO). RNA was isolated after 24 h, and RNase protection analysis (RPA) was performed as described previously (3). The -fold induction was calculated based on means and standard errors of three to eight independent experiments.
Transfection of Ferritin H-luciferase Reporter Gene Constructs and Luciferase Assay-Hepa1-6 cells were transfected for 4 h with a total of 500 ng of DNA (FH-Luc reporter gene constructs, ␤-galactosidase trans-fection control plasmid, and pUC18 as detailed in the respective figure legends) using the LipofectAMINE reagent (Invitrogen) according to the manufacturer's procedures. Cells were allowed to recover for 20 -24 h and subsequently were treated with 25 M ␤-NF or vehicle (Me 2 SO) for 24 h. Cells were harvested, lysed in 1ϫ reporter lysis buffer (Promega), and assayed for luciferase activity. Luciferase activity was assessed using the Luciferase Assay kit (Promega) according to the manufacturer's protocol. Expression of the ␤-galactosidase transfection control was measured as previously described (62).

Xenobiotics and Chemopreventive Agents Induce
Ferritin mRNA-␤-NF is a polycyclic aromatic hydrocarbon that has been used to study activation of phase 2 enzymes by both XREand EpRE/ARE-dependent mechanisms (23,24,32,63). Oltipraz is an electrophilic dithiolethione that represents a widely studied class of candidate chemopreventive agents. To create a model system in which to explore the mechanism of ferritin induction by these agents, we treated cultured liver cells for varying periods of time with ␤-NF or oltipraz. As shown in Fig.  1, both ␤-NF and oltipraz induced ferritin H and L mRNA in Hepa1-6 cells in a time-dependent manner. We also performed Northern blot analysis of NIH3T3 fibroblasts treated with oltipraz and a second dithiolethione, D3T. As shown in Fig. 2, both these agents induced ferritin H and L mRNA in NIH3T3 cells. Induction of ferritin H as well as ferritin L mRNA by oltipraz in NIH3T3 cells was time-dependent and occurred as early as 3 h after treatment, with induction peaking at 24 h. mRNA induction was accompanied by an increase in ferritin protein of similar magnitude (Fig. 3). Induction of ferritin mRNA by oltipraz, D3T, and ␤-NF was also seen in HepG2 cells (data not shown). Taken together, these results demonstrate that ␤-NF, oltipraz, and D3T induce both ferritin H and L in a variety of cells, including murine and human hepatocytes and fibroblasts.
Induction of Ferritin Is Mediated by the EpRE/ARE-To assess the mechanism of ferritin induction, we transiently transfected NIH3T3 cells with a chimeric ferritin H-human growth hormone (FH-hGH) gene construct that spans 4.8 kb of the murine ferritin H promoter region fused to the human growth hormone reporter gene (Ϫ4.8kbFH-hGH). Subsequently, cells were treated with 70 M oltipraz or 70 M D3T. After 24 h, RNA was isolated and RNase protection analysis was performed to assess the induction of the reporter gene as well as the endogenous ferritin H gene.  gene, confirming the Northern blot analysis shown in Fig. 2. In addition, induction of the Ϫ4.8kbFH-hGH reporter gene construct by oltipraz and D3T was observed. These results indicate that induction of ferritin H by oltipraz and D3T is mediated by a transcriptional mechanism.
To delineate the element responsible for transcriptional activation of ferritin H, deletion analysis was performed. FH-hGH 5Ј deletion constructs containing 4.8, 4.13, or 4.0 kb of the murine ferritin H 5Ј flanking region were transiently transfected into NIH3T3 cells, treated with 70 M oltipraz, and analyzed by RNase protection assay. As shown in Fig. 5, a region located between Ϫ4.13 and Ϫ4.0 kb of the ferritin H promoter is responsible for activation of ferritin H transcription. Thus, both the Ϫ4.8kbFH-hGH and Ϫ4.13kbFH-hGH construct were induced by oltipraz, whereas the Ϫ4.0kbFH-hGH construct was not. Because the 75-bp ferritin H EpRE/ARE is located between Ϫ4.13 kb and Ϫ4.0 kb of the 5Јferritin H promoter region, this result suggested that the ferritin H EpRE/ARE mediates induction of ferritin H in response to oltipraz. To test this, we used a chimeric gene in which a 107-bp region containing the ferritin H EpRE/ARE was inserted into a minimal ferritin H promoter construct (Ϫ0.32kbFH-hGH). As shown in Fig. 6, activity of this promoter was enhanced by oltipraz, demonstrating that the ferritin H EpRE/ARE is sufficient to mediate induction of ferritin H in response to oltipraz. Thus, ferritin H contains a functional EpRE/ARE that mediates responsiveness to oltipraz.
To determine whether the induction of ferritin H by ␤-NF utilized a similar mechanism, we first considered the potential role of XRE sequences. ␤-NF can activate transcription via both XRE-and EpRE/ARE-dependent mechanisms (16). The XRE consensus sequence has been defined as 5Ј-T(A/T)GCGTG-3Ј (18), and functional XRE sequences have previously been genetically identified in the cytochromes P450 1A1 and 1A2 (18,19,64), GST Ya (20), UDP-glucoronyl transferase 1A1 and 1A6 (21,22), NQO1 (23), and Cu/Zn-SOD gene (24,25). Inspection of the ferritin H 5Ј promoter sequence revealed five potential consensus XRE sequences, one on the sense strand and four on the antisense strand. Their specific locations and sequences are as follows. The XRE sequence present in the sense strand is located between Ϫ3089 and Ϫ3083 (5Ј-CTGCGTG-3Ј). The XRE sequences in the antisense strand are located between Ϫ4579 and Ϫ4573 (5Ј-CACGCAC-3Ј), Ϫ2817 and Ϫ2811 (5Ј-CACGCCC-3Ј), Ϫ413 and Ϫ407 (5Ј-CACGCTT-3Ј), and Ϫ351 and Ϫ345 (5Ј-CACGCAC-3Ј). To determine if these elements were functional, we prepared ferritin H-luciferase constructs (luciferase was used to simplify assessment of reporter gene expression) (Fig. 7). These constructs spanned Ϫ4.8, Ϫ4.2, and Ϫ4.0 kb of the murine ferritin H 5Ј promoter region. In addition, to determine the involvement of the XRE sequence located from Ϫ4579 to Ϫ4573 independent of the ferritin H EpRE/ARE sequence (Ϫ4117 to Ϫ4043), an internal deletion construct was made in which 3.5 kb of the ferritin H promoter were removed from the full-length Ϫ4.8kbFH-Luc construct spanning a region from Ϫ4477 to Ϫ941. The ferritin H-luciferase 5Ј deletion and internal deletion constructs were transiently transfected into Hepa1-6 cells. Subsequently, cells were treated with 25 M ␤-NF for 24 h, and luciferase activity was determined. The constructs Ϫ4.8kbFH-Luc and Ϫ4.2kbFH-Luc demonstrated luciferase induction in response to ␤-NF, whereas Ϫ4.8kb⌬3.5kbFH-Luc and Ϫ4.0kbFH-Luc did not (Fig. 7). Hence, induction of ferritin H by ␤-NF is mediated by an enhancer element located between Ϫ4.2 and Ϫ4.0 kb, a region that contains the ferritin H EpRE/ARE but none of the XRE sequences. To confirm that activation of ferritin H by ␤-NF occurred via the EpRE/ARE, the 75-bp EpRE/ARE element was inserted into a minimal promoter construct, which contains 225 bp of the ferritin H promoter (Ϫ0.225kbFH-Luc). As shown in Fig. 7, 75bpEpREFH-Luc mediated induction of luciferase activity, whereas Ϫ0.225kbFH-Luc did not. Thus, ferritin H is transcriptionally activated by ␤-NF via a mechanism that depends on the EpRE/ARE and not the XRE sequences.
Elements in the Ferritin H EpRE/ARE Bind Nrf2-Nrf2 is a member of the NF-E2 family of transcription factors that has been shown to mediate EpRE/ARE-dependent transcription of a variety of cytoprotective genes (36,47,56,65). The ferritin H EpRE/ARE contains two elements with sequence similarity to the NF-E2 consensus sequence. The AP1 consensus sequence of the ferritin H EpRE/ARE is embedded in a canonical NF-E2 site; the AP1-like sequence in the ferritin H EpRE/ARE also possesses considerable sequence similarity to the NF-E2 consensus (9/11 residues) (Fig. 8). To examine the ability of Nrf2 to bind to the ferritin H EpRE/ARE, we performed electrophoretic mobility shift assays. We assessed binding to both the AP1/NF-E2-like element and the AP1/NF-E2 element, both of which are contained in the ferritin H EpRE/ARE and show high homology to the consensus EpRE/ARE sequence (Fig. 8). These experiments were performed using nuclear extracts from HepG2 cells, because in our hands commercially available Nrf2 antibodies did not reliably bind to mouse Nrf2. As shown in Fig. 9, nuclear extracts isolated from ⌯epG2 cells treated with 70 M oltipraz or 25 M ␤-NF for 6 h bound these components of the EpRE/ARE. Fig. 9 also demonstrates that addition of an antibody to Nrf2 results in a supershift at both elements; in contrast, addition of normal rabbit serum did not result in a supershift. Thus, Nrf2 binds to the ferritin H EpRE/ARE.
Role for Nrf2 in Induction of Ferritin by Oltipraz, D3T, and ␤-NF-To test involvement of Nrf2 in ferritin induction, we compared the ability of Nrf2ϩ/ϩ and Nrf2Ϫ/Ϫ primary mouse embryo fibroblasts to induce ferritin H and L in response to oltipraz, D3T, and ␤-NF. As shown in Fig. 10, induction of ferritin H and L mRNA was blocked in Nrf2Ϫ/Ϫ cells, whereas both ferritin H and L were induced in the Nrf2ϩ/ϩ cells. Basal levels of ferritin H and L mRNA were also reduced in Nrf2 knockout cells, suggesting a role for Nrf2 in both basal and induced ferritin transcription.
Nrf2 Mediates Transcription of Ferritin H at the EpRE/ ARE-To demonstrate that Nrf2 mediates transcriptional activation of ferritin H via an EpRE/ARE-dependent mechanism, we transiently cotransfected a Nrf2 dominant negative mutant expression plasmid and the 75bpEpREFH-Luc reporter gene construct into Hepa1-6 cells. As seen in Fig. 11, cotransfection of the 75bpEpREFH-Luc reporter gene construct with an empty expression vector (pEF) and subsequent treatment with 25 M ␤-NF for 24 h resulted in induction of luciferase activity. However, cotransfection of a dominant negative mutant of Nrf2 (pEF/Nrf2dnm) and the 75bpEpREFH-Luc reporter gene construct suppressed ␤-NF-induced activation of luciferase activity. In addition, pEF/Nrf2dnm decreased basal activity of the 75bpEpREFH-Luc reporter gene construct, supporting the results in Fig. 10 indicating that Nrf2 affects basal as well as inducible ferritin H expression. Neither pEF/Nrf2dnm nor ␤-NF had any effect on basal or inducible expression of a ferritin H minimal promoter construct lacking the EpRE/ARE (Ϫ0.225kbFH-Luc) (Fig. 11). To confirm these results, 75bpEp-REFH-Luc was cotransfected with increasing amounts of a Nrf2 expression plasmid (pEF/Nrf2). As shown in Fig. 12, Nrf2 activates luciferase activity in a dose-dependent manner. This response requires the EpRE/ARE, because Ϫ0.225kbFH-Luc was unaffected by increasing amounts of pEF/Nrf2 (Fig. 12). DISCUSSION Several laboratories have demonstrated that ferritin H and L are induced in response to oxidants and pro-oxidant xenobiotics (7-11), and a transcriptional mechanism has been impli- FIG. 9. Nrf2 binds to the AP1/NF-E2like and AP1/NF-E2 consensus elements of the ferritin H EpRE/ARE. Nuclear extracts were isolated from HepG2 cells that had been treated with vehicle (Me 2 SO), 70 M oltipraz, or 25 M ␤-NF for 6 h. 10 -20 g of extract was incubated with 32 P-labeled AP1/NF-E2like or AP1/NF-E2 consensus oligonucleotide (100,000 -150,000 cpm), and 100-fold molar excess of specific (SC) and nonspecific (NC) competitors where indicated. A NFB binding element (Promega) was used as a nonspecific competitor. Following a 20-min incubation at room temperature, an antibody specific for Nrf2 or normal rabbit serum (c) was added to the indicated samples for a 30-min room temperature incubation. The DNA-protein complexes were separated by 4% polyacrylamide gel electrophoresis. Specific bands are pointed out by the large arrows, and supershifted Nrf2 is marked by the small arrows.

FIG. 10. Nrf2 is required for activation of ferritin H and L by oltipraz,
D3T, and ␤-NF. Total RNA was isolated from Nrf2ϩ/ϩ and Nrf2Ϫ/Ϫ cells that had been treated with vehicle (Me 2 SO), oltipraz, D3T, and ␤-NF for 24 h. RNA was size-fractionated on denaturing agarose gels, transferred to nylon membranes, and allowed to hybridize with cDNA probes for ferritin H and L. Ethidium bromide staining was done to assure equal RNA loading. The -fold induction was calculated with Nrf2ϩ/ϩ cells treated with vehicle defined as 1. Shown are the average Ϯ S.E. values for two (␤-NF) or four (oltipraz and D3T) independent experiments. cated (9). Activation of ferritin in response to these agents is thought to serve a cytoprotective function from iron-catalyzed oxidative damage. Previously, we have genetically defined the DNA element responsible for inducing ferritin H transcription in response to oxidants (Fig. 8). This element contains a cisacting DNA-enhancer located ϳ4.1 kb from the transcription start site and possesses sequence similarity to the consensus electrophile/antioxidant-responsive element (EpRE/ARE) (7). Sequence searches and reporter assays have also identified a functional EpRE/ARE in the ferritin L gene (55).
In this report we demonstrate that ferritin H and L mRNA are induced by ␤-NF and by chemopreventive dithiolethiones in fibroblasts and hepatic cells. Elevation of ferritin H mRNA by the dithiolethiones oltipraz and D3T is the result of a transcriptional mechanism. Similarly, Primiano et al. (15) provided evidence for transcriptional activation of ferritin H and L in the livers of rats treated with D3T. Using ferritin H deletion constructs fused to the human growth hormone gene, the results presented in this study identify the EpRE/ARE as the mediator of transcriptional activation of ferritin H in response to ␤-NF, oltipraz, and D3T. These results implicate ferritin in the chemopreventive response and suggest that ferritin, together with other EpRE/ARE-mediated cytoprotective proteins, contributes to the chemoprotected phenotype in cells in which these proteins have been induced. Classes of proteins induced by chemopreventive agents include both conjugating enzymes (glutathione S-transferases and UDP-glucuronosyltransferase) involved in detoxification and export and antioxidant enzymes (superoxide dismutase and NAD(P)H:quinone oxidoreductase 1). Ferritin, with its ability to sequester iron and prevent the formation of oxygen free radicals, has some characteristics of this latter class. In fact, we and others (66,67) have shown that repression of ferritin synthesis is correlated with enhanced sensitivity to oxidative stress; conversely, induction/overexpression of ferritin has been linked to enhanced cellular protection against oxidant-induced cytotoxicity (68,69).
A sequence search of 4.8 kb of the murine ferritin H 5Јflanking sequence revealed the presence of five XRE sequences that were potential candidates for regulating ferritin H tran-scription in response to polycyclic aromatic hydrocarbons, such as ␤-NF. However, deletion analysis of the ferritin H 5Ј-promoter region using ferritin H-luciferase (FH-Luc) reporter gene constructs demonstrated that these elements are non-functional in ferritin induction by ␤-NF, because reporter constructs containing these putative XRE sequences but lacking the EpRE/ARE were not induced in response to ␤-NF (Fig. 7). Rather, the EpRE/ARE controls ferritin H transcription in response to ␤-NF (Fig. 7). Because the five XRE sequences closely resemble the consensus XRE sequence (5Ј-T(A/ T)GCGTG-3Ј) (18), it was surprising that none of these elements in the ferritin H promoter region conferred inducibility of the gene in response to ␤-NF. In particular, comparison of the ferritin H XRE sequences to XRE sequences identified previously in various genes showed that the ferritin H XRE sequences at nucleotides (nt) Ϫ4579 to Ϫ4573 and nt Ϫ351 to Ϫ345 are identical to the functional XRE identified in the rat UDP-glucoronyl transferase 1A1 (21) and human CYP 1A2 (19) gene. Likewise, the XRE sequence at nt Ϫ3089 to Ϫ3083 and nt Ϫ2817 to Ϫ2811 are identical to the XREs present in the rat and human Cu/Zn-SOD genes, respectively (24,25). It has been suggested that the XRE core sequence itself is insufficient to confer transcriptional induction, and that additional nucleotides flanking the core sequence exert an important influence on the ability of the Ah receptor to bind to the core XRE sequence (70). These contextual requirements may contribute to the inactivity of the XRE elements in the ferritin H promoter.
The electrophile/antioxidant-responsive element consensus sequence resembles the recognition sequence for transcription factors of the AP1 and NF-E2 family of DNA binding proteins, allowing for a wide variety of transcription factors to bind to this element and mediate basal as well as inducible transcription. Several laboratories have demonstrated the involvement of the transcription factor Nrf2 in inducing cytoprotective proteins in response to a variety of agents, including ␤-NF and dithiolethiones (47,65,71). Some genes require Nrf2 for basal as well as D3T-inducible transcription, whereas others showed induction by D3T independent of Nrf2 status (65). Results presented here demonstrate that treatment of Nrf2ϩ/ϩ and Nrf2Ϫ/Ϫ primary mouse embryo fibroblasts results in induction of ferritin H and L mRNA in wild type but not Nrf2 knockout cells, indicating that Nrf2 is necessary for dithiolethione-and ␤-NF-induced transcription of both ferritin H and L. Gel shift and transfection experiments indicate that Nrf2mediated induction of ferritin H targets the ferritin H EpRE/ ARE (Figs. 9, 11, and 12). Although we did not quantitate nuclear Nrf2 following treatment with oltipraz, D3T, or ␤-NF, the increasing intensity in band shift seen in Fig. 9 is consistent with increased binding of Nrf2 to the ferritin H EpRE/ARE following stimulation with xenobiotics. This would be concordant with demonstrations of nuclear translocation of Nrf2 following treatment with inducing agents (36,49).
Our experiments showed not only that Nrf2 is involved in activating transcription of ferritin H and L in response to dithiolethiones and ␤-NF, but also suggested the involvement of Nrf2 in basal transcription of these genes, since expression of ferritin H and L mRNA was decreased in Nrf2Ϫ/Ϫ cells when compared with the wild type cells (Fig. 10). Effects on ferritin H basal transcription were confirmed by cotransfecting 75bpEp-REFH-Luc with a dominant negative mutant of Nrf2, which both suppressed EpRE/ARE-dependent induction of ferritin H in response to an inducer and decreased basal expression. A similar involvement of Nrf2 in basal transcription has been reported for other genes, including glutathione S-transferases (72) and ␥-glutamylcysteine synthetase (57).
The ferritin H EpRE/ARE has two sites with which Nrf2 interacts, namely the AP1/NF-E2 consensus sequence and the AP1/NF-E2-like element of FER-1 (Fig. 9). Although we did not examine the interaction of the ferritin L EpRE/ARE with Nrf2, inspection of the ferritin L promoter reveals a NF-E2 consensus sequence embedded in the ferritin L EpRE/ARE, suggesting that Nrf2 may be involved in the coordinate regulation of both ferritin subunits through targeting of the EpRE/ARE.
Nrf2 has been demonstrated to be important in the regulation of several EpRE/ARE-dependent genes. However, Nrf2independent mechanisms of gene regulation at the EpRE/ARE also exist. For example, it was recently reported that the ability of the model chemopreventive agent sulforaphane to induce some but not all EpRE/ARE-dependent genes was abrogated in the intestine of Nrf2 knockout mice (73). Kwak et al. (65) reported that in the livers of Nrf2 knockout mice treated with D3T, induction of ferritin was enhanced rather than suppressed. These findings differ from results presented here that indicate that Nrf2 is required for induction of ferritin by oltipraz, D3T, and ␤-NF. However, our experiments were performed in knockout cells derived from a different genetic background (74) than those used by Kwak et al. Given the existence of alternative pathways of regulation at the EpRE/ARE, and the potential of AP1 and Maf proteins and other transcription factors to modulate Nrf2 activity both positively and negatively, it is possible that the relative abundance and/or activity of such factors may influence gene inducibility. Thus, cellular context and genotype may determine the contribution of Nrf2 to the regulation of target genes. For example, in the GCS h gene, higher concentrations of c-Jun repressed expression, presumably due to formation of c-Jun/c-Fos complexes that interfered with binding of the Nrf2/c-Jun complex to the EpRE/ARE (75). Genetic variation, including polymorphisms in Nrf2 itself, may add an additional level of complexity to such interactions (76). Collectively, these observations may point to the existence of additional regulatory pathways that permit fine-tuning of the cellular response to xenobiotics, perhaps allowing the coordinate induction of subsets of antioxidant and detoxification genes in different cell types dependent on cellular context, genotype, and xenobiotic challenge.
We have previously identified a number of transcription factors that assemble at the FER-1 component of the ferritin H EpRE/ARE. These include members of the AP1 family such as JunD and FosB, and the ATF/CREB family member ATF1 (54). Results presented here implicate Nrf2 as an additional participant in this transcription factor complex, conferring added potential for both positive and negative regulation at this element. Nrf2 has been shown to bind to DNA as a heterodimer with small Maf (MafG and MafK) proteins, an interaction that can be negatively regulated by large Maf proteins such as c-Maf (77). Nrf2 may also interact directly with other transcription factors, such as AP1 (45,47) and ATF4 (78). Because the DNA recognition sequence for AP1 and Nrf2 family members overlaps considerably, it has also been suggested that under selected circumstances Nrf2 function may be modulated indirectly through displacement of Nrf2 from DNA by AP1 family members (75). Nrf2 can also interact with global adaptor and chromatin remodeling factors such as p300/CBP (79). We have previously demonstrated that the p300/CBP transcriptional adaptor proteins are involved in mediating basal ferritin H transcription through the FER-1 element, possibly due to an interaction between p300/CBP and FER-1 binding proteins (80). Indirect evidence indicates that p300/CBP may be necessary for EpRE/ARE-mediated induction of ferritin H, because NIH3T3 cells stably transfected with E1A lose the ability to induce ferritin H mRNA in response to t-BHQ, a classic inducer of antioxidant and phase II enzymes (67). These results suggest that Nrf2 may mediate some of its effects on basal and inducible ferritin H expression through interaction with p300/CBP. Further studies will be required to test how the assembly of this complex array of transcription factors, cofactors, and chromatin remodeling factors is regulated at the ferritin EpRE/ARE.