Cobalt induces heme oxygenase-1 expression by a hypoxia-inducible factor-independent mechanism in Chinese hamster ovary cells: regulation by Nrf2 and MafG transcription factors.

We have shown previously that activation of the heme oxygenase-1 (ho-1) gene by hypoxia in aortic smooth muscle cells is mediated by hypoxia-inducible factor-1 (HIF-1). In mutant (Ka13) Chinese hamster ovary cells lacking HIF activity, accumulation of ho-1 mRNA in response to hypoxia and the hypoxia-mimetic CoCl(2) was similar to that observed in wild type (K1) cells. These results support the existence of HIF-dependent and HIF-independent mechanisms for ho-1 gene activation by hypoxia and CoCl(2). In Ka13 cells, CoCl(2) stimulated expression of a luciferase reporter gene under the control of a 15-kilobase pair mouse ho-1 promoter (pHO15luc). Mutation analyses identified the cobalt-responsive sequences as the stress-response elements (StREs). In electrophoretic mobility shift assays, two specific StRE-protein complexes were observed using extracts from Ka13 cells. In response to cobalt, the level of the slower migrating complex X increased, whereas that of complex Y decreased, in a time-dependent manner. Members of the AP-1 superfamily of basic-leucine zipper factors bind to the StRE. Antibody supershift electrophoretic mobility shift assays did not detect Jun, Fos, or ATF/CREB proteins but identified Nrf2 and the small Maf protein, MafG, as components of complex X. Furthermore, dominant-negative mutants of Nrf2 and small Maf, but not of other bZIP factors, attenuated cobalt-mediated gene activation. Additional experiments demonstrated that induction by cobalt does not result from increased expression of MafG or regulated nuclear translocation of Nrf2 but is dependent on cellular oxidative stress. Unlike cobalt, hypoxia did not stimulate pHO15luc expression and did not increase StRE binding activity, indicating distinct mechanisms for ho-1 gene activation by cobalt and hypoxia in Chinese hamster ovary cells.

The adaptive response to low oxygen tension (hypoxia) includes increased production of a select set of proteins involved in oxygen homeostasis at the cellular and systemic levels. These proteins include: erythropoietin (EPO), 1 which stimulates erythropoiesis and the oxygen carrying capacity of the blood; vascular endothelial growth factor (VEGF) and one of its receptors Flt-1, which promote angiogenesis and the delivery of oxygen carrying blood to hypoxic sites; and glucose transporter-1 (Glut-1) and various glycolytic enzymes that provide for increased energy production through glycolysis during periods of reduced energy production via oxidative phosphorylation (reviewed in Ref. 1). Hypoxia-dependent expression of these and other proteins is regulated by both transcriptional and post-transcriptional mechanisms.
Transcriptional regulation of many hypoxia-responsive genes is critically dependent on a common cis-acting sequence, the hypoxia response element (HRE), and the trans-acting factor hypoxia-inducible factor-1 (HIF-1), a heterodimeric protein comprising ␣ and ␤ subunits. The ␣ subunit is unique to HIF-1, whereas the ␤ subunit is shared by other transcription factors, most notably the arylhydrocarbon receptor. In addition to hypoxia, certain transition metals such as cobalt, nickel, and manganese, and the iron chelator desferrioxamine (DFO) also stimulate expression of several hypoxia-responsive proteins including EPO, VEGF, phosphoglycerokinase 1, and Glut-1. These agents are considered to be hypoxia mimetics as they are proposed to exploit the cellular oxygen sensing mechanism or the hypoxia signal transduction pathway for gene activation. Consistent with this idea, cobalt and DFO stimulate HIF-1 activity and the HRE is required for target gene induction by these agents (reviewed in Refs. [1][2][3]. Another protein whose expression is stimulated by both hypoxia (4,5) and cobalt (6,7) is heme oxygenase-1 (HO-1). HO-1 catalyzes the first and rate-limiting reaction in heme catabolism, the oxidative cleavage of b-type heme molecules to yield equimolar quantities of biliverdin, carbon monoxide (CO), and iron. Biliverdin is subsequently converted to bilirubin by the action of biliverdin reductase. Both biliverdin and bilirubin are potent antioxidants, and CO has been shown to function as a neuronal messenger and a vasodilator (reviewed in Refs. 8 and 9). Recent studies indicate that CO modulates other cellular activities including signal transduction pathways (10) and apoptosis (11). In addition to hypoxia and cobalt, expression of HO-1 is widely induced by a variety of oxidative stress-associated agents including the substrate heme, inflammatory cytokines, heavy metals, hyperthermia, and UV irradiation (reviewed in Ref. 12). Because of its inducibility and the pleiotropic properties of the reaction products, HO-1 manifests potent antioxidant and anti-inflammatory activities and helps maintain cellular homeostasis in response to stress and injury.
Stimulation of HO-1 expression by most inducers is controlled primarily at the level of gene transcription, and we have previously identified two 5Ј distal enhancer regions, E1 and E2, within the mouse ho-1 promoter that mediate gene activation by a variety of pro-oxidants (12). Both E1 and E2 contain multiple copies of a motif, termed the stress response element (StRE), that are sufficient and necessary for gene activation by multiple agents including heme, cadmium, and various xenobiotics (13,14). Subsequent analysis of the mouse ho-1 promoter identified two sequence elements distinct from the StREs that resembled the consensus HRE. These elements bound HIF-1 and mediated hypoxia-dependent activation of a reporter gene in rat aortic vascular smooth muscle cells, indicating that HREs and HIF-1 regulate HO-1 expression in response to hypoxia (15). In contrast to these findings, Wood et al. (16) have developed mutants of Chinese hamster ovary (CHO) cells lacking the HIF-1␣ subunit and demonstrated that, although activation of certain hypoxia-responsive genes such as glut-1 is attenuated in these cells, induction of HO-1 mRNA by hypoxia and cobalt is not impaired. One explanation for the inducibility of HO-1 in these cells is the possibility of the more recently characterized HIF-2␣ (17) and HIF-3␣ (18) subunits to compensate for the HIF-1␣ deficiency. The mutant CHO cells, however, do not express HIF-2␣ and otherwise do not exhibit inducible HRE binding activity or HRE-dependent transcription activity (16,17). Taken together, these results imply the existence of both HIF-dependent and HIF-independent mechanisms for ho-1 activation by hypoxia and hypoxia mimetics. We have used the mutant CHO cells to characterize the HIF-1-independent mechanism. Here we show that induction by cobalt, but not by hypoxia, is mediated by the StREs and the heterodimeric transcription factor Nrf2⅐MafG. Indirect evidence suggests that induction by hypoxia in CHO cells may not occur at the level of gene transcription.

EXPERIMENTAL PROCEDURES
Materials-Tissue culture media were from Life Technologies, Inc., and fetal bovine serum was obtained from Mediatech (Herndon, VA). Restriction endonucleases and other DNA modifying enzymes were purchased from either Life Technologies, Inc. or New England Biolabs (Beverly, MA). Oligonucleotides were synthesized by IDT, Inc. (Coralville, IA). Radiolabeled nucleotides were obtained from PerkinElmer Life Sciences. Reagents for luciferase assays were purchased from Sigma-Aldrich. All antibodies except anti-mouse Nrf2 (kindly provided by M. Yamamoto, University of Tsukuba, Tsukuba, Japan) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). All other chemicals were reagent-grade.
Cell Culture, Transfection, and Enzyme Assays-CHO cells were cultured in Ham's F-12 medium supplemented with 10% fetal bovine serum and 50 g/ml gentamicin in a humidified atmosphere of 5% CO 2 and 95% air at 37°C. Cells were subjected to hypoxia in a controlled atmosphere chamber (Billups-Rothenburg; Del Mar, CA) supplied with a constant flow of a hydrated 1% O 2 , 5%CO 2 , balanced N 2 gas mixture. Hypoxic exposures were performed as cells approached 75-90% confluence. Transient transfections of luciferase reporter gene constructs were carried out by the calcium phosphate precipitation technique as described previously (26) or with Fugene6 transfection reagent (Roche Molecular Biochemicals) according to the manufacturer's recommendation. For calcium phosphate transfections, cells were seeded (4 ϫ 10 5 / well) in six-well plates and cells in each well were transfected with a DNA mixture consisting of 1 g of the luciferase plasmid, 0.5 g of pCMV␤-gal, and 3.5 g of pBluescriptII SK-(Stratagene). For Fugene6 transfections, cells were seeded (2 ϫ 10 5 /well) in 12-well plates and cells in each well were transfected with a DNA mixture consisting of 50 ng of the luciferase plasmid, 10 ng of pCMV␤-gal, and, where indicated (see Figs. 6 and 7D), 50 ng of the dominant-negative mutant or Gdbd fusion construct. Additional details are provided in the figure legends. Preparation of cell extract and measurement of luciferase and ␤-galactosidase activities were carried out as described previously (26).
RNA Isolation and Analysis-Total RNA was isolated by the procedure of Chomczynski and Sacchi (27). For RNA dot blot analysis, 5 g of total RNA was transferred to Zeta-Probe (Bio-Rad) nylon membrane according to the manufacturer's instructions. ␣-32 P-Radiolabeled hybridization probes were generated by random priming of cDNA fragments encoding rat HO-1 or ribosomal protein S3. Hybridization and washing conditions were identical to those described previously for Northern blots (28). HO-1 hybridization signals were quantified using a Storm PhosphorImager (Molecular Dynamics, Sunnyvale, CA). After signal quantification, the membranes were stripped and re-hybridized to the S3 probe. Relative mRNA levels were calculated after correcting for RNA loading by normalizing the primary hybridization signal with the S3 signal.
Unlabeled, double-stranded oligonucleotides, either wild-type or those containing specific mutations within the core StRE, were used as competitors. In antibody supershift assays, 1 l (2 g) of pre-immune IgG or specific rabbit polyclonal antibodies were added to the reaction mixture and incubated for 20 min at room temperature prior to electrophoresis. In the case of MafG, 1 l of whole serum (pre-immune and anti-MafG) was used.
Western Blot Analysis-Cell culture and treatment, preparation of whole cell extracts, and target protein detection were carried out as described previously (19,21). Cytoplasmic and nuclear fractions were prepared as described (31). Anti-mouse Nrf2 and anti-MafG antibodies were used at 1:1000 and 1:2000 dilutions, respectively. Anti-␣ tubulin and anti-histone H1 antibodies were used at concentrations recommended by the manufacturer.

RESULTS
Induction of ho-1 mRNA by Hypoxia or Cobalt Is Not Impaired in HIF-1␣-deficient Cells-Wild-type (K1), clonal control (C4.5), and clonal HIF-1␣-deficient (Ka13) CHO cells were exposed to normoxia, hypoxia (1% O 2 ), or 100 M CoCl 2 , NiCl 2 , or DFO for 18 h, and the steady-state level of ho-1 mRNA was measured by RNA dot blot analysis. In part confirming the results of Wood et al. (16), HIF-1-positive (K1, C4.5) and HIF-1-deficient (Ka13) cells exhibited similar levels of ho-1 mRNA accumulation in response to hypoxia or cobalt (Fig. 1). In addition, DFO, a compound known to activate HIF-1␣ in CHO cells and NiCl 2 , another hypoxia mimetic, did not increase ho-1 mRNA levels even in HIF-1-positive cells. Taken together these results imply that HIF-1 activity is neither sufficient nor necessary for ho-1 gene activation by hypoxia or cobalt in CHO cells.
The Stress Response Elements Mediate Induction by Cobalt-Induction of the ho-1 gene by most agents is mediated at the level of gene transcription. Induction by cobalt appears to occur by the same mechanism, as CoCl 2 activated a 15-kbp ho-1 promoter/luciferase reporter fusion gene, pHO15luc (Fig. 2), in a dose-dependent manner. Maximum activation, 7.3-fold, was observed at a metal concentration of 100 M (Fig. 3A). Higher concentrations of CoCl 2 resulted in cellular toxicity and reduced luciferase expression. Consistent with the RNA analysis, no significant induction was observed with NiCl 2 or DFO even up to a concentration of 300 M. 5Ј deletion mutants of pHO15luc were used in initial experiments to localize the cobalt-responsive cis-element(s). Although relatively large regions were deleted, partial or complete loss of cobalt responsiveness correlated with removal of the E2 or both E2 and E1 regions, respectively (Figs. 2 and 3B). The role of the E1 and E2 enhancers in cobalt responsiveness was confirmed using mutants of pHO15luc specifically lacking E1, E2, or both enhancers (Fig. 3C) and by directly testing the activity of E1 and E2 in isolation (Fig. 3D). Both E1 and E2 alone are activated by cobalt and appear to contribute equally to the overall activity of the 15-kbp promoter. cis-Elements common to E1 and E2 include one or more binding sites for specificity protein 1, C/EBP, and members of the AP-1 superfamily of bZIP proteins (12). In either enhancer, mutation of the latter, also referred to as the StREs, completely abolished cobalt responsiveness (pE1luc-M739 and pE2lucM45, Fig. 3D). The StRE alone (p3xStREluc), but not the HIF response sequences (p3xHifREluc), conferred cobalt responsiveness. From these results we conclude that the StRE is sufficient and necessary for cobalt-mediated ho-1 gene activation in Ka13 cells.
Hypoxia Does Not Activate the ho-1 Promoter or Induce StRE Binding Activity in CHO Cells-Reporter gene transfection assays were carried out to determine if ho-1 gene activation by hypoxia and cobalt in CHO cells occurred by the same mechanism. Surprisingly, in contrast to that observed with CoCl 2 , hypoxia did not stimulate expression of the luciferase gene under the control of the StRE, the E1 enhancer, or the 15-kbp ho-1 promoter in either K1 or Ka13 cells (Fig. 4). This result was not due to a general lack of responsiveness of CHO cells to hypoxia, as expression of the luciferase gene, under the control of multiple HIF-1 binding sites (p3XHifREluc), was clearly stimulated in K1 cells. As expected, HIF-1-dependent gene regulation was inoperative in Ka13 cells and mutation of the HIF-1 binding sites (p3XHifREMluc) abolished reporter gene activation in K1 cells.
StRE-binding Proteins in Ka13 Cells-EMSA reactions using whole cell extracts from Ka13 cells were carried out in order to identify DNA-binding proteins responsible for cobalt-mediated ho-1 gene induction. In vehicle-treated cells, three major StRE-protein complexes were observed (Fig. 5A). Competition experiments using wild-type and mutant StRE sequences indicated that the fastest migrating complex resulted from nonspecific DNA-protein interaction (data not shown). The specific complexes, X and Y, exhibited opposing behavior upon treatment of CHO cells with cobalt. Although the amount of complex X increased in a time-dependent manner, the level of complex Y progressively decreased over a period of 24 h. In contrast to cobalt and consistent with the promoter analysis, hypoxia did not increase StRE binding activity in Ka13 or K1 cells (Fig. 5B,   FIG. 1. Cobalt and hypoxia stimulate ho-1 mRNA accumulation in CHO cells with and without HIF activity. The indicated CHO cells (ϳ5 ϫ 10 5 /well) were seeded in six-well plates and cultured for 40 h. The culture medium was replaced with fresh, complete medium, and the cells were exposed to vehicle/normoxia, hypoxia (1% O 2 ), CoCl 2 (100 M), NiCl 2 (100 M), or DFO (100 M) for 18 h and then harvested for RNA isolation. RNA dot blot analysis, quantization, and signal normalization were carried out as described under "Experimental Procedures." -Fold induction relative to control is presented, and each bar represents the average Ϯ S.E. of three or four independent experiments. data not shown). Together with the reporter gene transfection data, these results clearly point to divergent mechanisms for ho-1 gene activation by cobalt and hypoxia.
EMSA Complex X Contains Nrf2 and MafG Transcription Factors-The consensus StRE (13) resembles the binding sites for AP-1 (Fos/Jun), ATF/CREB, Maf, and CNC-bZIP families of transcription factors. In order to determine which, if any, member(s) of these families of proteins could function as the cobalt-responsive transcription factor(s), antibody "supershift" EMSA reactions were carried out to identify specific StRE-binding proteins in Ka13 cells. As shown in Fig. 6, commercially available antibodies directed against individual ATF/CREB, Jun, Fos, and Maf proteins did not significantly or consistently alter formation or migration of either complex X or Y. In contrast, incubation with anti-Nrf2 completely supershifted (and also possibly inhibited formation of) DNA-protein complex X. We have recently generated anti-bodies to MafG that do not, or only poorly, cross-react with MafF and MafK, respectively (Ref. 32; data not shown). Reduced migration, or supershifting, of complex X was also observed with anti-MafG. The nearly complete elimination or supershifting of complex X by antibodies against Nrf2 and MafG strongly suggest that Nrf2⅐MafG heterodimers are the primary protein components of complex X. Although extracts from Ka13 cells treated with CoCl 2 for 2 h were utilized in the experiments depicted in Fig. 6, similar results were obtained with untreated Ka13 cells and cells treated for 24 h (data not shown). None of the antibodies tested significantly or consistently altered the formation or migration of complex Y. However, because a nonspecific complex that migrates in the same vicinity as complex Y was generated with MafG antiserum, we cannot exclude the possibility that MafG is present in complex Y. N-Acetylcysteine Inhibits Cobalt-dependent Gene Activation-Recent studies have implicated Nrf2 (21,(33)(34)(35)(36) and, in some cases, Nrf2-small Maf dimers in induction of various genes by xenobiotics. Models resulting from some of these studies have suggested that 1) gene induction results as a consequence of cellular oxidative stress and Nrf2 functions as a sensor for oxidative stress, 2) oxidants stimulate Nrf2 activity at least in part by promoting translocation of Nrf2 from the cytoplasm to the nucleus, and 3) inducer-dependent increase in Nrf2⅐MafG DNA binding activity results in part from increased expression of MafG. We have examined these and other parameters to better characterize the mechanism of ho-1 gene activation by cobalt. As shown in Fig. 8A, induction of the StRE-regulated luciferase gene by cobalt is dependent on oxidative stress as this stimulation was completely abrogated by N-acetylcysteine, a glutathione precursor, and an antioxidant, in a dose-dependent manner. Treatment of K13 cells with 100 M CoCl 2 from 0 to 24 h (Fig.  8B, lanes a-f), however, did not enhance expression of either Nrf2 or MafG protein as judged by Western blot analysis. In a similar analysis, Nrf2 was detected in both cytoplasmic and nuclear (Fig. 8C, C and N) fractions, but nuclear accumulation did not increase upon exposure of K13 cells to cobalt. The integrity of the cytoplasmic and nuclear fractions were examined by monitoring the cytosolic protein ␣-tubulin (Tub) and the nuclear protein histone H1. Nrf2 function may also be affected by modulation of its trans-activation capacity. This possibility was tested using a yeast Gal4-based system to assay for Nrf2 trans-activation potential. As shown in Fig.   FIG. 4. Hypoxia does not stimulate StRE-dependent transcription activity. K1 or Ka13 cells were transfected with the indicated luciferase plasmids, exposed to normoxia or hypoxia (1% O 2 ), and analyzed as described in the legend to Fig. 3D and under "Experimental Procedures." Each bar represents the average Ϯ S.E. of three independent experiments.

FIG. 5. Cobalt, but not hypoxia, alters StRE binding activities in Ka13 cells.
A, EMSA reactions were carried out as described under "Experimental Procedures" using whole cell extracts from Ka13 cells treated with 100 M CoCl 2 for the indicated time (t). Specific StREprotein complexes (X, Y) and the nonspecific complex (NS) are indicated. Lane a, EMSA reaction without cell extract. B, EMSA reactions were carried out using whole cell extracts from Ka13 cells exposed to normoxia (N) or hypoxia (H, 1% O 2 ) for 20 h. The gels were exposed to x-ray film for 16 (A) or 28 (B) h. 8D, a Gal4 DNA binding domain/Nrf2 fusion (Gdbd-Nrf2) potently induced expression of the luciferase reporter gene under the control of Gal4 binding sites. Cobalt, however, did not stimulate Nrf2 transcription activity. Because small Maf proteins lack transcription activation domains, as expected, the Gdbd-MafG fusion did not increase luciferase activity beyond that observed with Gdbd alone (data not shown for Gdbd). DISCUSSION With respect to mammalian gene regulation, cobalt has been most commonly studied as a non-physiological inducer of the heavy metal-responsive metallothionein genes and of several hypoxia-responsive genes. In this capacity, cobalt is able to activate the metal-responsive element/metal transcription factor-1 (37) and HRE/HIF-1 pathways. In this report we show that cobalt utilizes a different mechanism, activation of the StRE/Nrf2 pathway, for induction of HO-1 expression in CHO cells.
Accumulating data implicate Nrf2 as a key regulator of the adaptive response to oxidative stress (21, 33, 35, 36, 38 -40). In response to pro-oxidants and xenobiotics, Nrf2 coordinately activates transcription of a select set of target genes by binding to distinct but very similar DNA elements, individually or alternatively referred to as the NF-E2 binding site, the Maf recognition element, the stress-response element, or the antioxidant response element (ARE). Many of the Nrf2 target genes (21,33,35,36,(41)(42)(43) encode proteins that individually and collectively manifest anti-oxidant activity by one of several mechanisms: 1) metabolism or detoxification of xenobiotics and oxidants, 2) reduction of oxidized proteins, or 3) production of antioxidants. Proteins in each of these categories include 1) NAD(P)H:quinone oxidoreductase, which catalyzes two-electron reduction of quinones, preventing the participation of such compounds in redox cycling and oxidative stress and glutathione S-transferase, which conjugates hydrophobic electrophiles and reactive oxygen species with glutathione; 2) thioredoxin, an ubiquitous dithiol hydrogen donor for a variety of proteins including transcription factors; and 3) HO-1 and ␥-glutamylcysteine synthase (␥-GCS), which catalyzes the rate-limiting reaction in biosynthesis of glutathione, the primary non-protein thiol in cells.
Nrf2 is one of several members of the CNC-bZIP subfamily of basic region-leucine zipper (bZIP) transcription factors that function as obligate homo-or heterodimers and are characterized by a bipartite structure: a region enriched in basic residues necessary for DNA binding and an adjacent protein dimerization domain in which leucine (or equivalent) residues are present at every 7th position (i.e. the "leucine zipper"). The CNC-bZIP subfamily, along with the Fos, Jun, ATF/CREB, and Maf subfamilies, can be categorized into the larger "AP-1" superfamily of bZIP proteins because of similarities between the FIG. 6. StRE-protein complex A contains Nrf2 and MafG. Antibody "supershift" EMSA reactions were carried out as described under "Experimental Procedures" using whole cell extracts from Ka13 cells treated with 100 M CoCl 2 for 2 h, a time point at which complexes A and B are present at roughly equivalent levels. Antibodies (Ab) to individual transcription factors are indicated. "Supershifted" complexes are marked by asterisks. Lane 1 (left) corresponds to EMSA reaction without cell extract. CIgG, control IgG; CS, control serum. consensus binding sites, sequence conservation within the bZIP domains, and the tendency for individual members to form cross-family heterodimers.
As is the case for Fos proteins, the sequence of the leucine zipper domain of Nrf2 precludes self-dimerization (44) and thus Nrf2 functions as an obligate heterodimer. In accordance with the paradigm established by NF-E2 (45), the first CNC-bZIP containing mammalian transcription factor isolated, the most prominent dimerization partners of Nrf2 are the small Maf proteins, MafF, MafG, and MafK (46). The precise function of such Nrf2⅐Maf dimers, however, is somewhat controversial. For instance, Nrf2⅐small Maf dimers have been proposed to function as positive regulators of genes encoding NAD(P)H:quinone oxidoreductase and various glutathione S-transferase subunits in response to butylated hydroxyanisole (33). Similarly, the Nrf2⅐Maf G complex (and possibly other Nrf2⅐bZIP dimers) may mediate ARE-dependent induction of the ␥-GCS subunit genes by ␤-naphthoflavone and pyrrolidine dithiocarbamate (36). On the other hand, based in part on transfection studies that consistently show attenuation of Nrf2-mediated trans-activation of the ARE by co-expression of small Maf proteins, others have suggested that Nrf2⅐small Maf dimers do not function as positive regulators of the ARE and may indeed have repressor activity (47,48). In addition to Nrf2 and small Maf factors, Jun and Fos subfamily members have also been identified as ARE-binding proteins (36,41,49). Indeed, Nrf2⅐Jun complexes have been implicated as positive effectors of ARE-dependent genes (41). The supershift EMSA data presented here support a role for Nrf2⅐MafG but not of Nrf2⅐Jun dimers in ho-1 gene regulation by cobalt. The increase in the level of the Nrf2⅐MafG/StRE complex in response to cobalt and inhibition of gene activation by MafK and Nrf2 dominant-negative mutants further suggest that the Nrf2⅐MafG heterodimer functions as a positive regulator in this process.
Our results indicate that activation of the ho-1 gene by cobalt occurs, at least in part, as a consequence of increased binding of Nrf2⅐MafG to the StREs, a stimulation that could result from increased expression of MafG, a stress-responsive protein in multiple cell types. Agents known to stimulate MafG mRNA accumulation include hydrogen peroxide (50); arsenite, heavy metals, and hyperthermia (51); and ␤-naphthoflavone and pyrrolidine dithiocarbamate (36). In the latter case, increased MafG production may contribute to the ARE-dependent induction and even subsequent repression of the ␥-GCS subunit genes. Our data show that cobalt does not stimulate MafG synthesis in CHO cells; thus, such an induction is not necessary for increased StRE binding activity or ho-1 gene activation.
Accumulating evidence (34,43,52,53) suggests that, under normal conditions, Nrf2 exists in an inactive, cytoplasm-localized state, in part or fully as a consequence of binding to the cytoskeleton-associated protein Keap1. After exposure of cells to electrophiles or oxidative-stress generating agents, such as diethylmaleate, 12-O-tetradecanoylphorbol-13-acetate, or tertbutylhydroquinone, the cytoplasmic retention mechanism is inactivated and Nrf2 is transported to the nucleus by an as yet uncharacterized mechanism(s) but one that may involve protein kinase C-mediated phosphorylation of Nrf2 (52). We have been unable to obtain any evidence for such a process in cobaltstimulated CHO cells, either by subcellular fractionation (Fig.  8) or by immunocytochemistry (data not shown). These results suggest that regulation of Nrf2 activity by nuclear transport is an inducer-specific and/or a cell-specific mechanism. Certainly, at least in the latter case, this specificity could easily arise because of cell-dependent variations in the ratio of Nrf2 and Keap1. Indeed, variations in the level of Keap1 have been noted and fibroblasts appear to express lower amounts of this protein than, for instance, hematopoietic cell lines (34). The inability of cobalt to affect nuclear translocation of Nrf2, MafG, and Nrf2 expression or Nrf2 transcription activity suggests that regulation of Nrf2 function in response to cobalt occurs primarily at the level of DNA binding affinity (and of necessity, protein dimerization). The post-translation processes responsible for modulation of Nrf2 DNA binding activity are not known but are likely to involve phosphorylation/dephosphorylation reactions that are known to regulate the various activities of most transcription factors.
One interesting result of this study is that the 15-kbp ho-1 promoter, which contains the previously characterized HIF-1 binding sites observed to be active in aortic smooth muscle cells (15), is unresponsive to hypoxia in wild-type CHO cells. That the HIF-1 pathway is operative in these cells is evident from the fact that the same elements, when placed out of context and in multiple copies, are capable of supporting hypoxia-dependent reporter gene induction. These observations certainly point to cell-specific variations in the mechanism of HO-1 induction by hypoxia in vascular smooth muscle and CHO cells but do not explain why the HIF-1 binding sites would be unresponsive in their normal environment. Perhaps, in CHO cells, these elements function only in cooperation with other sequences (and their cognate binding factors) located outside of the region tested. We note, however, that sequences necessary for gene activation by all HO-1 inducers thus far tested reside within the 15-kbp segment (12). 2 -f in B) or the indicated time (in hours, C). Twenty g of protein (B) or 10% of cytoplasmic or nuclear extract (C) were fractionated on a denaturing polyacrylamide gel and subjected to Western blotting using antibodies directed against MafG or mouse Nrf2. The blot in C was stripped and probed with anti-␣-tubulin (Tub) or anti-histone H1 antibodies. D, Ka13 cells were transfected with pFRluc and constructs encoding the indicated Gdbd fusion, treated with vehicle or 100 M CoCl 2 , and analyzed as described in the legend to Fig. 3D and under "Experimental Procedures." Each bar represents the average Ϯ S.E. of three independent experiments. the possibility that hypoxic induction of HO-1 in CHO cells occurs not because of gene activation but by a post-transcriptional mechanism such as mRNA stabilization. The expression of several hypoxia-responsive proteins, including EPO, tyrosine hydroxylase, and VEGF, is in part regulated in this manner (54 -58), and Panchenko et al. (59) have recently provided evidence for hypoxia-dependent stabilization of HO-1 mRNA in human dermal fibroblasts. Furthermore, the results presented here provide interesting parallels with the latter study. For instance, desferrioxamine does not stimulate HO-1 expression in either CHO or skin fibroblast cells. Similar to our findings with cobalt, hydrogen peroxide induction of HO-1 mRNA in dermal fibroblasts is sensitive to antioxidants and appears to be regulated at the level of gene transcription. In contrast, hypoxic skin fibroblasts do not exhibit increased free radical production and HO-1 mRNA accumulation is not inhibited by free radical scavengers. Although we have not directly tested the effect of hypoxia on free radical levels in CHO cells, the lack of stimulation of Nrf2 DNA binding activity by hypoxia indirectly suggests that such an effect would be minimal.
In summary, cobalt and hypoxia utilize divergent mechanisms for stimulation of HO-1 expression in CHO cells. Induction by cobalt is regulated primarily at the level of gene transcription, is oxidative stress-dependent, and is mediated by the StRE/Nrf2 transcription factor pathway. On the other hand, induction by hypoxia does not involve the StRE/Nrf2 pathway, is apparently independent of oxidative stress, and may be regulated by a post-transcriptional mechanism.