Bach1 Repression of Ferritin and Thioredoxin Reductase1 Is Heme-sensitive in Cells and in Vitro and Coordinates Expression with Heme Oxygenase1, β-Globin, and NADP(H) Quinone (Oxido) Reductase1*

Ferritin gene transcription is regulated by heme as is ferritin mRNA translation, which is mediated by the well studied mRNA·IRE/IRP protein complex. The heme-sensitive DNA sequence in ferritin genes is the maf recognition/antioxidant response element present in several other genes that are induced by heme and repressed by Bach1. We now report that chromatin immunoprecipitated with Bach1 antiserum contains ferritin DNA sequences. In addition, overexpression of Bach1 protein in the transfected cells decreased ferritin expression, indicating insufficient endogenous Bach1 for full repression; decreasing Bach1 with antisense RNA increased ferritin expression. Thioredoxin reductase1, a gene that also contains a maf recognition/antioxidant response element but is less studied, responded similarly to ferritin, as did the positive controls heme oxygenase1 and NADP(H) quinone (oxido) reductase1. Bach1-DNA promoter interactions in cells were confirmed in vitro with soluble, recombinant Bach1 protein and revealed a quantitative range of Bach1/DNA stabilities: ferritin L ∼ ferritin H ∼ β-globin, β-globin ∼ 2-fold >heme oxygenase1 = quinone reductase β-globin ∼ 4-fold >thioredoxin reductase1. Such results indicate the possibility that modulation of cellular Bach1 concentrations will have variable effects among the genes coordinately regulated by maf recognition/antioxidant response elements in iron/oxygen/antioxidant metabolism.

Genes encoding proteins that manage proteins of iron and oxygen traffic and metabolism are at the nexus of chemical reactions that are both critical and dangerous to life. The iron porphyrin complex, heme, has emerged as a key signal for iron and oxygen metabolism genes, including ferritin L (ftl) 3 (1) and ferritin H (fth) (2). Transcriptional regulation of NADP(H) quinone (oxido) reductase (qr) (3), heme oxygenase1 (ho1) (4), and ␤-globin (5) by heme requires the maf recognition/antioxidant response element (MARE/ARE), a conserved regulatory sequence found in the promoter or enhancer, and the heme binding transcriptional repressor Bach1.
Both ftl and fth contain heme-responsive canonical MARE/ ARE promoter sequences (1,2). The role of Bach1 in hemeregulated ferritin transcription is not known. By contrast, the role of IRP1 and IRP2 in heme-regulated ftl and fth mRNA translation is known (6 -9). The translational mechanism uses IRP1 and IRP2 to coordinate fth and ftl mRNA regulation with that of several other mRNAs important in iron and oxygen homeostasis by binding to iron-responsive elements (IRE) in each of the mRNAs. The IRE is a specific three-dimensional loop-helix-loop-helix structure (10) in the noncoding regions of the mRNAs (11)(12)(13)(14)(15)(16). Specific interactions between the iron regulatory proteins IRP1 and IRP2 and the different IREs in the mRNAs create a natural, combinatorial array of RNA⅐protein complexes (17). In the case of ferritin, when IRP1 or 2 are bound to the mRNA the ability of eukaryotic initiation factor 4F to recruit translational machinery to the mRNA is compromised and translation is repressed (18). Heme is known to repress IRP1 and IRP2 binding to the IRE, thus increasing ferritin expression (9,19,20). To our knowledge, ftl and fth are the only genes that have mechanistically distinct heme-responsive elements: the MARE/ARE in the DNA promoter and the IRE in the mRNA 5Ј-untranslated region (21). Together, they respond synergistically to cytoplasmic heme (1).
To determine if Bach1 is the heme target in MARE/ARE-dependent regulation of ferritin and thioredoxin reductase1 (trr), in analogy to ho1, ␤-globin, and qr promoters, we compared effects of Bach1 siRNA and Bach1 overexpression in HepG2 cells on heme-regulated fth, ftl, and trr expression and demonstrated that Bach1 was the heme-reversible repressor. In addition, using soluble, recombinant Bach1 and MARE/ARE-DNA duplexes (40 base pairs) in vitro, we showed heme-reversible binding of Bach1 to the ftl MARE/ARE sequence. Furthermore, we demonstrated that Bach1 binding to ftl, fth, trr, ho1, qr, and ␤-globin MARE/ARE sequences falls into three stability groups that may relate to differential regulation in vivo.

Cloning and Constructs
The ftl promoter and MARE/ARE mutation constructs, cell culture experiments, and luminescence assay have been described previously (1). trr and qr promoter constructs known to contain a functional ARE (27) were used as positive controls.

Cell Culture
Hepa1c1c7, a mouse hepatoma cell line, was used for chromatin immunoprecipitation experiments and was maintained in Dulbecco's modified Eagle's medium (4500 mg/liter glucose; Sigma) supplemented with 10% fetal bovine serum and 100 unit/ml penicillin/streptomycin (Invitrogen). HepG2 cells, a human hepatoma line, were used for overexpression and siRNA studies and were seeded in collagen-coated, 12-well plates (160,000 cells/well) using minimal essential medium with 10% fetal bovine serum (Atlanta Biologicals) and 100 unit/ml penicillin/streptomycin (Invitrogen) and cultured in 95% ambient air, 5% CO 2 as previously described (1). The concentration of heme (80 M) was selected for maximum induction of ftl expression in HepG2 cells (1); preliminary data on ftl induction showed similar effects with either hemin or rabbit hemoglobin. Microscopic observation of the HepG2 cells showed no evidence of toxicity under any of the conditions used.

RNA Analysis
Total RNA was isolated (RNeasy; Qiagen) 24 h later. RNA concentrations of all genes were determined by reverse transcription PCR using TaqMan Gene Expression Assay kits (Applied Bioscience) and an ABI 7900 Sequence Detection System (Applied Bioscience). Differences in RNA concentrations were quantified by the cycles to fluorescence midpoint cycle threshold calculation (2 Ϫ[ ⌬Ct experimental gene Ϫ ⌬Ct housekeeping gene] ), using ␤-actin as the housekeeping gene.

Chromatin Immunoprecipitation
The procedures were used as we have previously described them (28).
Immunoprecipitation with Bach1 Antiserum-48 h after transfection with Bach1 siRNA, as described above, nuclear proteins were extracted from the cells using the NE-PER nuclear extraction kit (Pierce) and Halt Protease Inhibitor Mixture (Pierce). Equal amounts of nuclear protein (80 g) from each treatment were run on a Tris-HCl, 12% SDS-PAGE gel (Bio-Rad). Proteins were transferred to a polyvinylidene difluoride membrane (Bio-Rad) using Towbin's buffer (25 mM Tris, 192 mM glycine, 0.5% SDS, 20% v/v methanol) containing 20% methanol and a Mini Trans-Blot cell (Bio-Rad) run at 350 mA for 1 h. Membranes were incubated overnight at 4°C in a solution of T-TBS containing 1% nonfat milk and 1,000-fold dilution of anti-Bach1 antiserum A1-6 raised in rabbit (24). Blots were washed in T-TBS (Tween 20-Tris-buffered saline, pH 7.6, 0.1% Tween 20, 20 mM Tris⅐Cl, 137 mM NaCl) and then incubated for 1 h with a 30,000-fold dilution of alkaline phosphatase-conjugated, anti-rabbit IgG antibody (Sigma), followed by washing in T-TBS. Blots were developed with incubation with 0.56 mM 5-bromo-4-chloro-3indoyl phosphate (Sigma) and 0.48 mM nitroblue tetrazolium (Sigma) in 10 mM Tris/HCl and quantified by densitometry.

Bach1 Overexpression
The Bach1 overexpression plasmid was obtained from Origene (catalogue number SC107917); cells were transfected with 1 g of the plasmid DNA 24 h after seeding, using FuGENE 6 transfection reagent (Roche Applied Science). Control cells were transfected with empty vector pCMV6XL-6. RNA was extracted and analyzed as described above.

Decreasing Bach1 mRNA
Targeting Bach1 RNA-A complex of four unique siRNA duplexes (obtained from a commercial vendor) (Dharmacon) that were developed and described by Shan et al. (26) was used to decrease Bach1 RNA. A titration showed that 50 nM siRNA decreased Bach1 RNA more than either 25 or 100 nM siRNA. The cells were transfected using Dharmafect 2 transfection reagent according to the manufacturer's protocol (Dharmacon) 24 h after seeding. Control cells were transfected with 50 nM siControl Non-targeting Duplex 1 (Dharmacon), consisting of four siRNA duplexes designed to avoid inhibiting any known eukaryotic genes. Bach1 siRNA-treated cells and controls were treated with Me 2 SO (0.2%) or Me 2 SO (0.2%) with hemin (80 M) 24 h after transfection. The final concentration of Me 2 SO was 0.2% for all treatments.
RNA Analysis-Total RNA was analyzed by reverse transcription PCR as described above.
Luciferase Reporter Experiments-4 h after siRNA transfection, the cells were co-transfected with pRLSV40 Renilla luciferase, a control for transfection, and the experimental luciferase constructs (test promoter ϩ firefly luciferase). siRNA-treated cells and controls were treated with Me 2 SO (0.2%) or Me 2 SO (0.2%) ϩ hemin (80 M) 24 h after transfection, harvested 24 h later, and assayed for luciferase using the Dual Luciferase Assay kit (Promega). Data from three independent experiments are expressed as the ratio of firefly luciferase (test promoter)/ Renilla luciferase (transfection control).
RNA Concentration-Real-time PCR was carried out as described above with samples from three independent experiments. Data are averaged from three independent experiments, and the errors are expressed as the standard deviation.

Expression of Soluble, Recombinant Bach1 Protein
Bach1 was expressed in Escherichia coli BL21(DE3) cells grown in Luria Bertani medium containing 1 M sorbitol and 250 M betaine to enhance production of native, folded Bach1 protein and to increase the yield of soluble protein (31). The cells were transformed with human, His-tagged Bach1 clone previously described (22). Induction of Bach1 expression after adding isopropyl-1-thio-␤-D-galactopyranoside (1 mM) was for 12 h at room temperature. Washed cells were sonicated and fractionated by centrifugation for 10 min, 3,000 ϫ g, 4°C. The supernatant fraction was further clarified by centrifugation for 2 h at 20,000 ϫ g at 4°C, and the resulting supernatant was incubated overnight at 4°C in Ni-NTA resin (Novagen), followed by elution with Novagen Ni-NTA buffer. Column eluates were analyzed by SDS-PAGE on pre-cast 12% gels (Invitrogen) and stained with Coomassie Blue. Pooled fractions of the protein isolated from E. coli were analyzed by both Coomassie Blue staining and Western blotting. Horseradish peroxidase conjugated to an anti-His tag mouse monoclonal antibody (Invitrogen) was used in the immunoblot with ECL ϩ chemiluminescence detection (Amersham Biosciences), and Coomassie Blue was used to stain the protein and markers (Fig. 1). Protein yields of the soluble, recombinant Bach1 protein from E. coli were ϳ15 mg/ml.

MARE/ARE DNA Gel Retardation Assay
Recombinant, soluble Bach1 from E. coli was used with double-stranded DNA (40 bp) labeled with 32 P on one strand. Bach1/DNA binding used the method described by Carey for protein-DNA interactions (32) and used by us for protein-RNA interactions (15,33,34). Binding occurred by mixing soluble, recombinant Bach1 protein and [ 32 P]DNA duplex (1.6 pmol) at protein:DNA stoichiometries of 0 -72:1 in buffer (20 l of 60 mM KCl, 24 mM Hepes⅐Na, 4 mM MgCl 2 , 5% glycerol, and 2% 2-mercaptoethanol, pH 7.2) that was incubated for 15 min at 4°C. One strand of the DNA (Integrated DNA Technologies) was labeled with [␥-32 P]ATP (Amersham Biosciences) using T4 polynucleotide kinase (New England Biolabs), purified with MicroSpin G-50 columns (Amersham Biosciences), followed by annealing to the complementary strand to make 32 P-labeled, double-stranded probes. Mouse ␤-globin containing the tandem MARE of HS2 (22) and human ho1 (35) MARE sequences were used as positive controls because they are known to be regulated by Bach1. Gene sequences and accession numbers are shown in Table 1. DNA and DNA⅐protein complexes were resolved on a 5% polyacrylamide gel, dried, and analyzed using phosphoimagery. To quantify the results, the intensities of bound or unbound [ 32 P]DNA were determined using Image-Quant software. The percent shift was calculated as follows: ([ 32 P]DNA:protein intensity/[ 32 P]DNA intensity ϩ [ 32 P]DNA: protein intensity) ϫ 100. To compare Bach1-DNA interactions among the different MARE/ARE sequences, the % bound DNA (1.6 pM) was plotted versus the protein:DNA ratio over the range zero, 9:1, 18:1, 36:1, and 72:1. The apparent K d (protein concentration required for 50% DNA binding) for each DNA duplex was calculated as previously described (36). The data ( Table 1) are presented as the mean (n ϭ 3) with the error as the standard deviation.

Bach1, FTL, and FTH MARE/ARE DNA Sequences Are Coprecipitated from Chromatin of Cultured Cells-
The similarity of the MARE/ARE promoter sequences and regulation of ferritin genes (1,2) to that of other MARE/ARE genes that are regulated by Bach1 (3,4,30,37) suggests that Bach1 will be present in chromatin complexes that contain fth and ftl DNA sequences. To test the hypothesis, we examined the DNA immunoprecipitated from Hepa1c1c7 cell chromatin with antibodies to Bach1. The data show (Fig. 2) that both ftl and fth DNA sequences were contained in the Bach1 antiserum precipitate. ␤-globin and ho1 sequences were also precipitated by the same antiserum as previously observed; qr sequences have also been found in Bach1 immunoprecipitates of chromatin (3, 4,   30,37). Thus, based on chromatin precipitation, ftl and fth are members of the Bach1-regulated gene family.
Effects of Changing Bach1 Concentrations on MARE/ARE Gene Expression-We reasoned that the turnover of Bach1-DNA interactions indicated in earlier studies (23,30) could mean that MARE/ARE DNA were not saturated even under steady state conditions. If so, increased expression of Bach1 might further repress MARE/ARE genes. To test whether Bach1 regulates fth and ftl genes, we examined effects of Bach1 overexpression. The concentrations of ftl, ho1, trr, and qr RNA decreased when HepG2 cells were transfected with a Bach1 cDNA expression plasmid (Fig. 3); Bach1 mRNA increased significantly. Hemin reversed the repression caused by increased Bach1 expression (Fig. 3), confirming earlier results for ho1 (4). The results also indicate that constitutive levels of Bach1 are insufficient to repress fully all the MARE/ARE genes. In addition, endogenous heme may decrease Bach1 concentrations in the HepG2, as recently indicated in murine embryonic fibroblasts when heme synthesis was inhibited (25). The similar responses of ftl to ho1, qr, and trr to Bach1 overexpression and/or hemin treatment emphasize the coordinated regulation of multiple MARE/ARE genes.
Decreasing Bach1 RNA to ϳ20% of normal with siRNA increased expression of fth, ftl, trr, ho1, and qr significantly (p Ͻ 0.05) (see supplemental Fig. S1), confirming and extending results of Shan et al. for ho1 (26). The concentration of Bach1 protein decreased ϳ50% with siRNA when nuclear extracts of the siRNA-treated cells were analyzed by Western blotting with Bach1 antiserum, as previously described (22). The combination of the observations that Bach1 protein concentrations are influenced by endogenous heme and that MARE/ARE genes appear to be incompletely repressed in the transfected HepG2 cells can explain why decreases of only ϳ50% in the endogenous level of repressor protein induced by siRNA changed MARE/ARE gene expression levels.

Bach1 Binds MARE/ARE DNA Sequences Selectively with Sensitivity to Heme-The experiments with Bach1 in cells or in chromatin (Figs. 2 and 3) suggest direct DNA-protein interactions between Bach1 and with ftl MARE/ARE promoters that would be sensitive to hemin in vitro.
To test this hypothesis, soluble, naturally folded, recombinant Bach1 protein (Fig. 1) was mixed with wild-type or mutant DNA duplexes (40 base pairs) in buffer and the interactions analyzed by electrophoretic mobility shift in native acrylamide gels.
Selectivity of Bach1 bound to the native MARE/ARE DNA sequences was indicated by the absence of binding to a mutated ARE sequence and by competition for binding with unlabeled DNA. Addition of a 100ϫ excess of unlabeled, mutated ftl MARE/ARE DNA had no effect on the interaction between wild-type labeled MARE/DNA and Bach1, while 100ϫ excess of unlabeled wild-type DNA completely abrogated Bach1 binding to [ 32 P]-/MARE/ARE with the same sequence (Fig. 4).
The observation that the soluble, recombinant Bach1 protein bound MARE/ARE DNA in the absence of small Maf protein contrasts with some previous observations, e.g. Ref. 22. In one case where small Maf protein was required for Bach1 binding, high concentrations of urea used in the isolation may have led to misfolding of Bach1, which was reversed in the presence of small maf protein (21). In another case (3), the V5 conjugate of Bach1, and residual reticulocyte proteins that bound Bach1, may have altered Bach1-DNA interactions that were also rescued by small maf proteins (3). Determination of the stoichiometry of binding of Bach1 to each DNA sequence (32) is outside the scope of this report where the focus is on comparing and enlarging the size of the Bach1/MARE-ARE family of genes and providing the first insight to differences in the Bach1 inter-  actions among them. Sorbital and betaine, known to improve the solubility of recombinant proteins expressed in E. coli (31), were used in this study to obtain soluble Bach1 protein that could be isolated without urea. All or any of these factors, solubility, absence of urea, and/or a long V5 tag or competing protein may relate to the binding of Bach1 to MARE/ARE DNA in the absence of small Maf proteins.
Hemin reversed the Bach1-ftl MARE/ARE DNA interactions in vitro (Fig. 4, right), supporting conclusions that the effects of hemin on MARE/ARE genes observed in vivo reflect heme-Bach1 interactions (1)(2)(3)(4). To determine whether the heme effect on Bach1/DNA binding was iron-dependent, cobalt protoporphyrin IX was tested under the same conditions as hemin and also blocked Bach1-DNA interactions in vitro (Fig. 4, right). The in vitro results complement in vivo studies that have demonstrated ho1 induction by cobalt-protoporphyrin (38 -40). Protoporphyrin IX alone had no effect on ftl MARE/ARE-Bach1/DNA interactions in vitro (data not shown), contrasting with ftl studies in HepG2 cells (1). Thus, a metalloporphyrin appears to be required to reverse Bach1/DNA binding in vitro.
Bach1 Concentrations Required for Full Binding in Vitro Vary among Different MARE/ARE Gene Sequences-The limited endogenous concentrations of Bach1 suggested by the increased repression when Bach1 was overexpressed in HepG2 cells (Fig. 3) could mean that under conditions where Bach1 is limiting, some MARE/ARE DNA sequences form more stable complexes with Bach1 than others do. To determine if there are variations in the stability of Bach1 complexes with different MARE/ARE DNA sequences, we compared the percentage of DNA bound by equal concentrations of DNA with ftl, fth, qr, ho1, or trr MARE/ARE DNA sequences at different stoichiometries of soluble, recombinant Bach1 protein. The range of protein/DNA ratios was 0 -72:1, and the DNA concentrations were 1.6 pM (Fig. 5). A protein:DNA ratio of 72:1 caused a complete or nearly complete binding of ftl, fth, qr, ␤-globin, and ho1 MARE/ARE DNA (Fig. 5). However, at lower Bach1 concentrations, each DNA displayed quantitative differences in the fraction of DNA complexed to Bach1 (Table 1, Fig. 5).

DISCUSSION
Bach1 binding to the chromatin region of fth and ftl genes observed in this study indicates that Bach1 is the heme-sensitive repressor of these genes and can explain the molecular basis of the previously observed heme-mediated regulation of ftl and fth genes (1,2). Ferritin genes expand the family of Bach1-regulated genes from those involved in oxygen, heme, and quinone metabolism (3,4,26,41) to genes for iron metabolism. The increased repression observed when Bach1 was overexpressed (Fig. 3) indicates that endogenous concentrations of Bach1 were too low to fully repress either ftl, fth, trr, or the positive control MARE/ARE genes ho1, ␤-globin, and qr. The observation that hemin reversed the repression by high concentrations of Bach1 (Fig. 3) emphasizes that Bach1 repression reflects both Bach1 protein concentrations and other cellular factors such as heme. Down-regulation of Bach1 was also observed in cells expressing hepatitis C proteins (42).
Comparison of the different MARE/ARE gene sequences ( Table 1) that share Bach1 binding and heme regulation (4, 26,  the concentration of DNA was 80 nM. Free DNA and bound DNA were resolved by electrophoresis (nondenaturing, 5% acrylamide gels run at 4°C); dried gels were analyzed by phosphoimagery and quantified with Imagequant. Arrows show the origin of each gel. Three groups of DNA⅐protein complex with significantly different stabilities were observed (See Table 1 for K d values): (i) stability ϭ ␤-globin (ftl and fth);(ii) stability significantly different from ␤-globin (ho1, qr, trr); and (iii) significantly (p Ͻ 0.05) different from trr (␤-globin, fth, ftl, ho1, and qr). 41) and the different stabilities of the DNA⅐repressor complexes in vitro suggest the possibility that differential MARE/ ARE promoter binding to Bach1 in vivo will amplify effects of changes in Bach1 protein concentrations. Exploration of effects of combinations of small Maf and Bach1 protein that have been observed in vivo or with insoluble recombinant or V5-tagged Bach1 (3,22) are outside the scope of this report and are also technically daunting since, to date, small Maf without any tag cannot be isolated as a soluble protein. 4 Analysis of Bach1 binding stoichiometry for the different MARE/ARE is a study for the future.
The addition of ferritin H and L genes to the Bach1/MARE/ ARE transcription family emphasizes the roles of ferritins in minimizing oxidative stress. Ferritin traps both reactants in oxy/radical chemistry in the solid mineral of a protein cage (Fe 2ϩ /O 2 in eukaryotes or Fe 2ϩ /O 2 or H 2 O 2 in bacteria) (43), and this fact is sometimes overlooked. The role of ferritin in response to oxidative stress is illustrated by a number of studies demonstrating that ferritin is regulated similarly to protective phase II antioxidant enzymes such as qr and glutathione synthesis genes (44 -46).
Throughout the study we have used the term MARE/ARE sequences to group the six genes under study. The MARE sequence, (TGCTGAG/CTCAGCA) (47)(48)(49) present in ho1 and ␤-globin genes is similar to the consensus core sequence of the ARE (TGACnnnGC) in Phase II antioxidant response genes qr, trr, and ferritin (ftl and fth) ( Table 1) (1,3,27). For that reason, we combined the two terms. Moreover, the genes were all regulated similarly by Bach1 and heme when compared under the same condition as in this study. The shared regulatory mechanisms and sequences between the two cis-element families suggest that they could be merged. However, it remains to be experimentally determined if all members of the MARE/ ARE gene family are regulated by Bach1.
Heme, an important signaling molecule in oxidative stress pathways, links regulation of genes encoding the oxidant-protective proteins such as ho1, fth, ftl, qr, and trr (1,3,23,26,27) with the oxygen-carrying protein, hemoglobin. The proposed regulatory mechanism is heme-induced displacement of Bach1 followed by Nrf2 binding on the MARE/ARE DNA for both qr (3) and ho1 (23). However, the effect of increasing Bach1 on repression of the MARE/ARE genes illustrates the opportunities for changing Bach1/DNA binding by direct heme binding as in Fig. 4 or by changing nuclear transport and degradation of Bach1. The identification of a range of Bach1:MARE/ARE DNA stabilities for the different genes in the family may be useful for understanding quantitative variations in responses of Bach1regulated genes to signals such as heme.
Ferritin MARE/ARE-controlled transcription and IRE-controlled translation use a common pathway for heme-induced degradation of the DNA repressor Bach1 and the RNA repressor IRP2. Recent data show polyubiquitination of Bach1 by the ubiquitin E3 ligase HOIL-1 in response to heme (25), which is the same ligase that modifies IRP2 in response to heme (7,9). Thus, the observation of heme regulatory synergy by heme when both the DNA and RNA promoters were combined (1) can be explained by HOIL-1 modification of both Bach1 and IRP2. Sharing regulatory elements in DNA among ferritin and other antioxidant response genes and in mRNA among ferritin and other iron-trafficking genes is complemented by sharing the same regulatory molecule, heme, for protein repressors of the DNA and RNA promoters to coordinate ferritin expression with iron metabolism and oxidative stress protection.