Mechanism of Heme Oxygenase-1 Gene Activation by Cadmium in MCF-7 Mammary Epithelial Cells . ROLE OF p 38 KINASE AND Nrf 2 TRANSCRIPTION FACTOR

The mouse heme oxygenase-1 (HO-1) gene, ho-1, contains two inducible enhancers, E1 and E2. Of several cell lines tested, induction of an E1/luciferase fusion construct, pE1-luc, by CdCl(2) is most pronounced in MCF-7 cells. In these cells, E1, but not E2, is necessary and sufficient for ho-1 gene activation. Exposure of MCF-7 cells to 10 micrometer CdCl(2) stimulates phosphorylation of ERK, JNK, and p38 mitogen-activated protein kinases, implicating one or more of these signaling pathways in ho-1 gene induction. SB203580, an inhibitor of p38, diminishes cadmium-stimulated pE1-luc expression and HO-1 mRNA levels by up to 70-80%. PD098059, an ERK pathway inhibitor, does not affect HO-1 mRNA induction at the highest concentration (40 micrometer) tested. Similarly, co-expression of a dominant-negative mutant of p38alpha, but not of ERK1, ERK2, JNK1, or JNK2, reduces basal and cadmium-induced pE1-luc activity. E1 contains binding sites for the activator protein-1 (Fos/Jun), Cap'n'Collar/basic leucine zipper (CNC-bZIP), and CCAAT/enhancer-binding protein (C/EBP) families of transcription factors. A dominant-negative mutant of Nrf2 (a CNC-bZIP member), but not of c-Jun or C/EBPbeta, inhibits pE1-luc activation by cadmium. Induction of the endogenous ho-1 gene is also inhibited by the Nrf2 mutant. Mutations of E1 that inhibit cadmium inducibility also suppress the trans-activation and DNA binding activities of Nrf2, and SB203580, but not PD098059, attenuates Nrf2-mediated trans-activation of pE1-luc. Taken together, these results indicate that cadmium induces ho-1 gene expression via sequential activation of the p38 kinase pathway and Nrf2.

ducer specificity or selectivity, however, can still be achieved via the StREs because these motifs are targets for multiple members of the basic-leucine zipper (bZIP) superfamily of sequence-specific DNA-binding proteins, including the AP-1, CREB/ATF, Maf, and Cap'n'Collar/bZIP (CNC-bZIP) classes of transcription factors.Functional bZIP proteins exist as obligate dimers generated by both intra-family homo-and heterodimerization and by inter-family heterodimerization.Although the identity of the specific dimeric species utilized by any of the HO-1 inducers that act via the StREs has not been conclusively established, our recent study in L929 fibroblasts (9) demonstrated potent trans-activation of the E1 enhancer by CNC-bZIP factors, in particular Nrf2.
Stress-related agents, like other extracellular signals, stimulate intracellular networks of signaling cascades that regulate various cellular functions, in large part by modification of transcription factor activities and target gene expression.Mitogenactivated protein kinases (MAPKs) are serine/threonine protein kinases that occupy a central position in the signaling cascades regulating cellular processes such as cell growth, differentiation, and apoptosis.Three major subfamilies of MAPKs have been described as follows: extracellular signal-regulated kinase (ERK), c-Jun N-terminal kinase (JNK), and p38.Each MAPK subfamily is recognized and activated by specific MAPK kinases, which in turn are phosphorylated by several different and overlapping sets of MAPK kinase kinase.In principle, each kinase module represents a parallel but independent signaling pathway.Depending on the cellular and stimulatory context, however, there is significant cross-talk between transduction modules as they can respond to common upstream activators and phosphorylate common downstream targets.In general, the ERK pathway mediates cellular responses to growth and differentiation factors, whereas the JNK and p38 enzymes are activated by distinct and overlapping sets of stress-related stimuli, including heat shock, inflammatory cytokines, ultraviolet and gamma irradiation, and hyperosmolarity.Exceptions to this paradigm, however, have been reported (reviewed in Refs. 10 and 11).
Despite the extensive characterization of stress-activated signal transduction pathways and the nearly universal induction of HO-1 expression by stress-related stimuli, surprisingly little is known about the signaling mechanisms responsible for ho-1 gene activation.Furthermore, the limited studies carried out so far have yielded conflicting results.For example, in human HeLa cells, tyrosine kinase inhibitors, but not inhibitors of the ERK and p38 MAPK pathways, attenuate induction of the ho-1 gene by cadmium, heme, and arsenite (12).In contrast, arsenite-mediated activation of the chicken ho-1 promoter requires both ERK and p38 kinase activities and is presumably mediated by AP-1 transcription factors (13).This discrepancy may reflect cell type-and/or species-specific differences in the regulatory mechanism.Additionally, these two studies examined the expression of different genes, the endogenous ho-1 gene versus a transfected promoter/reporter chimera, which may not exhibit equivalent regulation under all conditions (12).
In the present study, we investigated in detail the mechanism of ho-1 gene induction by cadmium in MCF-7 mammary epithelial cells.By using pharmacological inhibitors and dominant-negative mutants of MAPKs and by comparing the expression of the endogenous gene and transfected ho-1 promoter constructs, we provide evidence supporting a role for p38, but not for ERK or JNK, in the induction process.Additional experiments indicate that Nrf2 is a target of the p38 pathway and mediates cadmium-dependent ho-1 gene activation.

EXPERIMENTAL PROCEDURES
Materials-Tissue culture media were from Life Technologies, Inc., and fetal bovine serum was obtained from Mediatech.Restriction endonucleases and other DNA-modifying enzymes were purchased from either Life Technologies, Inc., or New England Biolabs.Oligonucleotides were synthesized by IDT, Inc. Radiolabeled nucleotides were obtained from NEN Life Science Products.Heme (as hemin chloride) was purchased from Porphyrin Products.Kinase inhibitors were from Calbiochem.Reagents for luciferase assays were purchased from Sigma.All other chemicals were reagent grade.
Cell Culture, Transfection, and Enzyme Assays-RAW 264.7 macrophage, F9 embryo carcinoma, L929 fibroblasts, and HeLa cells were purchased from ATCC.MCF-7 cells were kindly provided by Dr. Louise Nutter.Cells were cultured in Dulbecco's modified Eagle's medium with 0.45% glucose, or Eagle's minimal essential medium (HeLa), containing 10% fetal bovine serum and 10 ng/ml insulin (MCF-7).Unless otherwise indicated, transient and stable transfections were carried out by the calcium phosphate precipitation technique as described previously (15).Briefly, for transient assays, cells were seeded (ϳ4 ϫ 10 5 /well of a 6-well plate) 16 h prior to transfection.Cells were exposed to the DNA-CaPO 4 precipitate for 6 h, shocked by a 1-min treatment with 10% glycerol in phosphate-buffered saline, and cultured for 24 (induction experiments) or 48 h (trans-activation experiments) in complete medium.In induction experiments, the cells were cultured for an additional 18 h in serum-free medium and treated with vehicle or inducing agents for 5 h in serum-free medium.Where indicated, kinase inhibitors were added 1 h prior to the addition of CdCl 2 and maintained during the remainder of the incubation period.Preparation of cell extract and measurement of luciferase activity were carried out as described previously (18).␤-Galactosidase activity was measured using the Galacto-Light chemiluminescent assay kit (Tropix, Inc.) according to the manufacturer's protocol.To generate stable transfectants, MCF-7 cells were plated (1 ϫ 10 6 /10-cm plate) and transfected as described above with 10 g of pEF/Nrf2M or pEF/c-JunM.Transfectants were selected over a 3-week period in the presence of G418 (up to 800 g/ml), and individual clones were isolated by limited dilution.
RNA Isolation and Analysis-Total RNA was isolated by the procedure of Sacchi and Chozymski (19).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 hy- bridization probes were generated by random priming of the human HO-1 (obtained from Genome Systems, Inc.) or GAPDH cDNA fragments.Hybridization and washing conditions were identical to those described previously for Northern blots (20).HO-1 hybridization signals were quantified using a Storm PhosphorImager (Molecular Dynamics).After signal quantitation, the membranes were stripped and re-hybrid-ized to the GAPDH probe.Relative mRNA levels were calculated after correcting for RNA loading by normalizing the primary hybridization signal with the GAPDH signal.
Western Blot Analysis-For detection of Nrf2M and c-JunM, confluent cells from one 60-mm plate were harvested in cold phosphatebuffered saline and pelleted by centrifugation at 8,000 rpm for 1 min at 4 °C.Cells were resuspended in 100 -200 l of lysis buffer (50 mM Hepes (pH 7.5), 1.5 mM NaCl, 1.5 mM MgCl 2 , 1 mM EGTA, 10% glycerol, 1% Triton X-100, 100 M phenylmethylsulfonyl fluoride, and 1 g/ml of antipain, chymostatin, leupeptin, and pepstatin A) and left on ice for 10 min.Cytoplasmic extracts were separated from the nuclei by centrifugation.Nuclear and whole cell extracts were prepared by direct lysis in 1ϫ electrophoresis sample buffer (62.5 mM Tris-HCl (pH 6.8), 2% SDS, 10% glycerol).For detection of MAPKs, cells were lysed directly in 1ϫ electrophoresis sample buffer containing 2 mM EGTA and 50 mM NaF.Protein concentration was determined using the Bicinchoninic Acid Protein Assay Kit (Sigma).Twenty-microgram samples were size-fractionated on 10 (MAPK) or 15% (Nrf2M, c-JunM) denaturing polyacrylamide gels, and protein blot analysis was carried out using the ECL Western blotting system (Amersham Pharmacia Biotech) according to the manufacturer's recommendation.Antibodies to Nrf2 and c-Jun were obtained from Santa Cruz Biotechnology, and those for non-phosphorylated and phosphorylated MAPKs were obtained from New England Biolabs.All antibodies were used at dilutions recommended by the manufacturer.
Electrophoretic Mobility Shift Assays (EMSA)-cDNA fragments encoding p18 and Nrf2M were cloned into plasmid pGEM-2, and in vitro transcription and translation reactions were carried out using the TNTcoupled Wheat Germ Extract System (Promega Biotech).Protein synthesis was carried out in duplicate reactions, with unlabeled methionine or [ 35 S]methionine.Radiolabeled samples were analyzed by SDS-PAGE and fluorography to monitor the integrity of the reaction products and to estimate relative levels of protein synthesis.Similar amounts of unlabeled protein were used for EMSA reactions that were carried out as described previously (15).A double stranded oligonucleotide containing the sequence 5Ј-GATCTTTTATGCTGTGTCATG-GTTT-3Ј (core StRE underlined) was used as probe in EMSA reactions.Analogous, unlabeled oligonucleotides, containing specific mutations within the core StRE (see Fig. 9), were used as competitors.

The E1 Enhancer Is Activated by Cadmium in MCF-7 Cells-
The mouse ho-1 gene contains two inducible enhancers, E1 and E2 (previously labeled SX2 and AB1, respectively; Fig. 1A), identified by their ability to activate the chloramphenicol acetyltransferase reporter gene in stably transected cells in response to multiple HO-1 inducers (14 -16).Induction of the enhancer/chloramphenicol acetyltransferase constructs was not observed in transient transfection assays (14).In such assays, an analogous luciferase reporter gene construct, pE1luc, was minimally responsive (ϳ2-3-fold induction) to 10 M CdCl 2 in several cell lines but exhibited significant induction (between 15-and 20-fold) in MCF-7 cells (Fig. 1B).In the latter cells, CdCl 2 activated the E1 enhancer to a greater degree than all the other agents tested at their optimal concentrations and, in general, El exhibited higher inducible transcription activity than E2 (Fig. 1C).In the context of a 15-kilobase pair promoter fragment, deletion of E1, but not of E2, abolished luciferase induction (Fig. 1D), indicating that E1 is necessary and sufficient for ho-1 gene activation by CdCl 2 , at least in MCF cells.These observations, and the conservation of the E1 enhancer in the human ho-1 gene (7), provided reasonable justifications for the use of mouse ho-1 gene sequences in human cells.
Cadmium induced pE1-luc expression in a dose-dependent manner, and maximum induction (ϳ27-fold) was observed at 20 M CdCl 2 (data not shown) further highlighting the sensi- tivity of ho-1 gene transcription to this particular agent in MCF-7 cells (compare with other inducers, Fig. 1C and below).For the remainder of the studies, CdCl 2 was used at a concentration of 10 M.
Cadmium Activates ERK, JNK, and p38 MAPKs-To determine the role, if any, of MAPKs in cadmium-mediated ho-1 gene activation, we first examined the effect of CdCl 2 on MAPK activities.MAPKs are activated by dual phosphorylation of threonine and tyrosine residues located in the "activation lip" of the conserved core kinase sequence (10), and the activated species can be detected by antibodies directed against phosphorylated peptides encompassing these residues.MCF-7 cells were untreated or treated with 10 M CdCl 2 for up to 4 h, and cell extracts were analyzed for phosphorylated and total MAPKs by Western blotting.Phosphorylated ERK1 (p44) and ERK2 (p42) were detected in untreated cells (Fig. 2, lane 1), Preparation of cell extracts, gel electrophoresis, and Western blot analysis were carried out as described under "Experimental Procedures."Filters were initially used to detect phosphorylated (P) MAPKs and then stripped and probed with antibodies that detect total MAPK proteins.Individual MAPKs are identified by their size (kDa).Similar results were obtained in 3-4 independent experiments.and an increase in the levels of these species was readily observed within 15 min of exposure to CdCl 2 (lane 2).Maximum phosphorylation of ERK1 and ERK2, approximately 4 -5fold above control values, was attained within 1 h of treatment and was maintained at this level for the duration of the experiment.Phosphorylated p38 was also detected in untreated cells, and its level increased in a time-dependent manner to greater than 40-fold above the control value at the last time point examined (4 h).In contrast to the ERK and p38 enzymes, phosphorylated JNK was not detected in unstimulated cells, and activation was not observed until 2 h after exposure to CdCl 2 .The phosphorylation status of the MAPKs was not altered if the cells were exposed to vehicle (distilled H 2 O) for up to 4 h (data not shown).The increase in the levels of phosphorylated MAPKs was not due to a concomitant elevation in the amount of the respective enzymes; indeed, cadmium appeared to cause a gradual decrease in the level of JNK proteins.
SB203580 Inhibits Cadmium-stimulated E1 Activity and HO-1 mRNA Levels-To address the role of individual MAPK pathways in ho-1 gene regulation by cadmium, we examined the effects of PD098059, an ERK pathway inhibitor, and SB203580, an inhibitor of p38, on pE1-luc expression.Treatment of cells with up to 40 M PD098059 had no effect on basal pE1-luc expression, whereas SB203580 reduced this activity by approximately 20% (Fig. 3, ϪCadmium).Both PD098059 and SB203580 diminished cadmium-stimulated pE1-luc expression in a dose-dependent manner by 50 and 70%, respectively (Fig. 3, ϩCadmium).Interestingly, up to the highest concentration tested (40 M), only SB203580 inhibited the level of HO-1 mRNA accumulation in response to cadmium (Fig. 4).The decreases in cadmium-stimulated pE-1luc activity and HO-1 mRNA levels by SB203580 were quantitatively similar.The discrepancy between the effects of PD098059 on pE1-luc expression and HO-1 mRNA accumulation can be explained by the observation that this compound, but not SB203580, inhibited luciferase enzyme activity, most noticeably in the presence of CdCl 2 (Fig. 5).Taken together, these results implicate the p38, but not the ERK, pathway in cadmium-mediated ho-1 gene induction.
A Dominant-negative Mutant of p38 Inhibits pE1-luc Expression-Because of the potential for nonspecific effects of pharmacological inhibitors, the role of MAPKs in cadmium-dependent ho-1 gene regulation was further examined using kinasedeficient, dominant-negative mutants of these enzymes.Consistent with the results obtained using SB203580, co-expression of a p38␣ mutant reduced basal and cadmium-stimulated pE1-luc activity by 60 -70% (Fig. 6).In contrast, ectopic expression of the ERK1 mutant enhanced both cadmium-independent and cadmium-dependent expression of pE-1luc.No significant effect was observed with mutants of ERK2, JNK1, or JNK2.
A Dominant-negative Mutant of Nrf2 Inhibits Induction of pE1-luc and the Endogenous ho-1 Gene by Cadmium-The E1 enhancer contains binding sites for the AP-1, C/EBP, and CNC-bZIP families of transcription factors (9,15,18).To determine the role, if any, of these factors in cadmium-dependent ho-1 gene regulation in MCF-7 cells, we examined the effects of dominant-negative mutants of individual family members on pE1-luc expression.Mutants of c-Jun (an AP-1 factor) and NF-IL6 (C/EBP␤) had no significant influence on luciferase activity, but the Nrf2 mutant, Nrf2M, dramatically diminished (Ͼ90%) both basal expression (Fig. 7, ϪCadmium) and cadmium inducibility (Fig. 7, ϩCadmium).To determine if Nrf2M also affects induction of the endogenous ho-1 gene, MCF-7 cells stably expressing Nrf2M were cloned and identified by Western blotting (Fig. 8A).Fig. 8B shows the steady-state level of HO-1 mRNA in control MCF-7 cells (pEF) and in an Nrf2Mtransfected clone (corresponding to lanes l and e, respectively,

FIG. 4. SB203580 (SB) inhibits induction of the ho-1 gene by
CdCl 2 .MCF-7 cells were plated as for transfections and cultured in complete medium for 48 h and subsequently in serum-free medium for 24 h.After this period, the indicated concentration of kinase inhibitor was added, followed 1 h later by the addition of vehicle or CdCl 2 (10 M).Cells were incubated for an additional 3 h and then harvested for RNA isolation.RNA dot blot analysis, quantitation, and signal normalization were carried out as described under "Experimental Procedures."A PhosphorImager scan from one experiment and normalized HO-1 mRNA levels in the presence of cadmium (Cd) (average of two independent experiments) are presented.PD, PD098059.in Fig. 8A) after treatment with vehicle or several HO-1 inducers.Expression of Nrf2M inhibited ho-1 gene induction in response to cadmium, heme, sodium arsenite, and TBHQ by greater than 75%.Basal ho-1 gene expression and TPA inducibility were not affected.Basal or induced HO-1 expression was not altered in MCF-7 cells expressing the dominant-negative mutant of c-Jun.Similar results were recently obtained with analogous stable transfectants of L929 fibroblasts (9).
The Cadmium Induction Profile of E1 Mutants Correlates with Nrf2 trans-Activation and DNA Binding Activities-Because Nrf2M could potentially inhibit ho-1 gene activation indirectly, by interfering with the action of other factors as a consequence of DNA occupancy, we sought additional support for a positive role for Nrf2 in cadmium-mediated induction.Initial studies showed trans-activation of pE1-luc by Nrf2 in MCF-7 cells, as previously observed with L929 cells (9).Therefore, we compared the trans-activation profile of wild-type and mutant E1 constructs to their cadmium responsiveness.The E1 enhancer contains three StREs (Fig. 9A) that resemble the consensus binding site, (T/C)GCTGA(G/C)TCA(C/T), for the CNC-bZIP/NF-E2 class of transcription factors (21).As the consensus NF-E2-binding site encompasses the consensus AP-1 heptad, TGA(G/C)TCA, such motifs are also targets for AP-1 proteins.The results and conclusions from the mutation analysis (Fig. 9, B-D) are summarized as follows.1) AP-1 proteins are not responsible for cadmium activation of E1.This conclusion is based on the observations that AP-1 heptads are insufficient for cadmium responsiveness (M789), and mutations that abolish AP-1 binding (M2, Ref. 15) do not alter cadmium inducibility (Fig. 9C).(It should be noted that StREa does not bind recombinant c-Jun homodimers (18), and more extensive mutations within the AP-1 heptads (M239 and M739) would be expected to abrogate both AP-1 and CNC-bZIP pro-FIG. 5. PD098059 (PD) inhibits luciferase enzyme activity in the presence of CdCl 2 .MCF-7 cells were mock-transfected or transfected with pE1-luc (10 g) as described under "Experimental Procedures."Individual wells of mock-transfected cells were treated with vehicle or kinase inhibitors at a final concentration of 40 M. A constant amount of extract from pE1-luc transfected cells was mixed with varying amounts of extracts from mock-transfected (vehicle-and inhibitor-treated) cells, and the mixtures were used for measurement of luciferase activity.Experiments were carried out in which all cultures were untreated (Ϫ Cadmium) or exposed to 10 M CdCl 2 for 5 h (ϩ Cadmium).Each data point, presented as % of luciferase activity in extracts derived from cells not exposed to kinase inhibitors, represents the average Ϯ S.D. from three independent experiments.The apparent kinase inhibitor concentration reflects the ratio of cell extract derived from inhibitor-treated cells to the total amount of cell extract in the luciferase assay reaction.SB, SB203580.FIG. 6. Co-expression of a kinase-deficient mutant of p38␣ inhibits basal and cadmium-dependent pE1-luc expression.Each well of MCF-7 cells was transfected with a plasmid mixture containing 3 g of pE1-luc and 0, 3, 6, or 9 g of the indicated MAPK dominant-negative mutant (dnm).Total DNA was equalized with an empty mammalian expression vector.Transfection and cell treatment (vehicle or 10 M CdCl 2 ) were carried out as described under "Experimental Procedures."Cell extract equivalent to 5 g of protein was used for measurement of luciferase activity.Each data point, presented as % of luciferase activity in the absence of MAPK mutants, represents the average value from three to six independent experiments.The range of standard deviation was between 10 and 25%.

FIG. 7. Co-expression of a dominant-negative mutant of Nrf2 inhibits basal and cadmium-dependent pE1-luc expression.
Transfection and cell treatment (vehicle or 10 M CdCl 2 ) were carried out as described in the legend to Fig. 6.Cell extract equivalent to 5 g of protein was used for measurement of luciferase activity.Each data point, presented as % of luciferase activity in the absence of transcription factor mutant, represents the average value from three to six independent experiments.tein binding (21) as evidenced by the lack of trans-activation of these mutants by Nrf2.) 2) Individually, StREb appears to be more important than either StREa or StREc for E1 activity (compare M700, M080, and M009), and two intact StREs are sufficient (compare E1 with M700 or M009) and necessary (data not shown) for at least partial cadmium inducibility and Nrf2 trans-activation.3) There is a 100% correspondence between the cadmium inducibility and Nrf2 trans-activation potentials of E1 mutants.
If Nrf2 mediates induction of E1 via the StREs, mutants of StRE that are not responsive to cadmium should also not bind to Nrf2.Because Nrf2 is not expected to homodimerize (22), as expected, in vitro translated Nrf2M (which still retains the DNA-binding and the leucine zipper dimerization domains) did not bind to an oligonucleotide encompassing StREb in EMSA reactions (Fig. 10, lanes c and d).CNC-bZIP proteins heterodimerize most prominently with small Maf proteins such as p18 (23,24).p18 can homodimerize but exhibited weak and variable binding (lane e), whereas Nrf2⅐p18 heterodimers bound avidly to StREb (lanes f and g).Competition experiments demonstrated that the M2 mutant of StREb bound to Nrf2⅐p18 with an affinity similar to that of wild-type StREb even though this mutation, in the context of E1, did not affect cadmium inducibility.In contrast, mutations that affect inducibility of E1 exhibited significantly reduced (M080) or no (M239) binding to Nrf2⅐p18.Similar results were observed when StREc and StREa (along with their respective mutants) were used in Nrf2⅐p18 binding assays, although StREa exhib-ited reduced (ϳ50%) affinity compared with StREb and StREc (data not shown).Together, the functional and DNA binding studies provide compelling evidence for the role of Nrf2 in ho-1 gene induction by cadmium.
SB203580 Inhibits Nrf2-mediated trans-Activation of pE1luc-Because cadmium-mediated ho-1 gene activation is dependent on p38 kinase and Nrf2 activities, we reasoned that the p38 pathway should modulate Nrf2 function.As predicted, treatment of MCF-7 cells with SB203580 attenuated Nrf2mediated trans-activation of pE1-luc by greater than 80% in a dose-dependent manner (Fig. 11).In contrast, treatment of cells with 5 M PD098059 stimulates Nrf2 trans-activation by approximately 2-fold.This enhancement, which gradually diminished to control levels at higher concentrations of inhibitor, is reminiscent of the effect of the ERK1 dominant-negative mutant on pE1-luc expression (Fig. 6).

DISCUSSION
Cadmium is an environmental and occupational toxin with no known physiological function.It is, however, classified as a human carcinogen, and cadmium compounds induce tumors in several animal tissues including the lung and prostate.The carcinogenicity of cadmium may be attributed to the ability of this metal to enhance DNA mutation rates in response to various chemical mutagens, possibly secondary to the inhibition of DNA repair mechanisms, and to stimulate mitogenic signaling pathways and expression of oncoproteins that control cellular proliferation (reviewed in Ref. 25).
Cadmium activates the expression of several mammalian genes.The proteins encoded by many of these genes can be classified into three major categories (25) as follows: 1) detoxifying proteins such as metallothionein (MT) or enzymes, such as ␥-glutamyl-cysteine synthase, that generate protective molecules (e.g.glutathione); 2) immediate-early response gene products, such as c-Fos, c-Jun, c-Myc, or Egr-1, that play a prominent role in cellular proliferation; and 3) stress-responsive proteins, such as heat shock proteins and HO-1, that provide cytoprotective functions.Analysis of the mechanism(s) of target gene regulation indicates that activation is not mediated by a single, sequence-specific transcription factor (or family of factors).Indeed, cadmium, a non-physiological agent, appears to co-opt the principle mechanism by which the target genes are regulated in response to other physiological stimuli.For instance, induction of the hsp70 gene by hyperthermia and other suboptimal growth conditions is regulated by heat shock factors, primarily HSF1, which also mediates cadmium-dependent hsp70 gene activation (26).Similarly, the serum response element (and presumably its cognate transcription factor complex), which is in large part responsible for transcriptional activation of the c-fos gene by growth factors and other mitogens, is also responsive to cadmium (27).In the case of metallothionein, of the physiological transition metals, zinc is probably the most potent inducer of MT gene transcription.Induction of MT genes by zinc is mediated by the interaction between metal response elements and MTF-1, a zinc finger transcription factor in the Cys 2 His 2 family.Targeted deletion of the mtf-1 gene (28) or inhibition of MTF-1 expression (29) aborgates MT gene induction not only by zinc but also by other transition metals including cadmium.
Clearly, in mammalian cells, several different sequencesspecific DNA-binding proteins can function as "cadmium response" factors.The results presented here suggest that Nrf2 also functions in this capacity.Analogous to the transcription factor/target gene systems described above, Nrf2 appears to be a dominant regulator of ho-1 gene activation.This conclusion is based on the fact that overexpression of the Nrf2 dominantnegative mutant in both MCF-7 and L929 fibroblasts (9) atten-  c, e, g, i, and k), cytoplasmic (Nrf2M; lanes b, d, f, h, j, and l), or total cellular extracts (cJunM) were prepared from clones transfected with the dominantnegative mutant expression plasmids or empty vector (Nrf2M; lanes k and l; cJunM; lanes a and b).Western blot analysis was carried out as described under "Experimental Procedures," and the filter was exposed to film for 1-5 min.The size (kDa) and migration of the molecular weight standards are indicated.B, RNA analyses.Control (pEF), Nrf2M-and cJunM-expressing MCF cells were plated and treated with vehicle (Veh), CdCl 2 (Cd, 10 M), heme (10 M), sodium arsenite (Ars, 50 M), TBHQ (50 M), or TPA (100 ng/ml) for 3 h.RNA isolation, dot blot analyses and signal normalization were carried out as described under "Experimental Procedures."HO-1 mRNA levels, relative to the amount in vehicle-treated control cells, are presented and represent the average of two independent experiments.uates ho-1 gene induction by multiple agents including the physiologically relevant stimuli, heme.Such a role for Nrf2 in ho-1 gene regulation is also supported by the recent observation that peritoneal macrophages derived from nrf2 Ϫ/Ϫ mice exhibit impaired induction of HO-1 in response to various electrophiles and oxidants including arsenite, TBHQ, and cadmium (30).
Nrf2 is one of four CNC-bZIP proteins identified in mammalian cells that function as transcription activators.These proteins contain a conserved C-terminal bZIP domain necessary for protein dimerization and DNA binding and a conserved upstream CNC motif homologous to a region within the Drosophila homeotic selector protein encoded by the Cap'n'Collar gene (31).CNC-bZIP polypeptides do not homodimerize and can only form obligate heterodimers, most prominently with the small Maf factors.The N-terminal transcription activation domains of CNC-bZIP proteins are less well conserved, and the activity of individual members varies considerably, with Nrf2 exhibiting the highest level of trans-activation when compared directly in the same cellular context (9,32).As with other bZIP transcription activators, deletion of the Nrf2 N-terminal activation domain results in a truncated protein that retains the ability to form dimers (that can bind to target sequences) and, when expressed at high levels, can function in a dominantnegative fashion if the resulting dimer does not possess a transcription activation domain.A conceptual and practical drawback to such mutants is that it is difficult to determine whether inhibition of target gene expression occurs because of nonspecific interference resulting from DNA occupancy by the mutant-containing dimer or specific sequestration of the positive-acting endogenous factor(s) (33).Indeed, interference by DNA occupancy is one plausible explanation for the observation that Nrf2M inhibits ho-1 gene activation in response to multiple agents.
Because of this potential limitation, it was important to obtain additional evidence for the role of Nrf2 in ho-1 gene activation.In the present study, at least with respect to induction by cadmium, such support is provided by the observation that mutants of E1 that are not responsive to cadmium are also not trans-activated by Nrf2 and do not bind Nrf2⅐p18 dimers in vitro.In a recent report (9), we suggested that Nrf2⅐p18 dimers do not mediate induction of the ho-1 gene.This conclusion was based mostly on data from Nrf2 trans-activation experiments using a mutant of p18 that we erroneously assumed did not homodimerize.In fact, the mutant protein can still form homoand heterodimers, but the resulting dimers do not bind to DNA (34).The data obtained with the mutant p18 are consistent with such characteristics, and reassessment of that data suggests that they provide inconclusive evidence regarding the role of Nrf2⅐p18 dimers in ho-1 gene regulation.Dimers of Nrf2 with non-Maf proteins have not been characterized, and thus it cannot be known if the DNA binding specificities of such proteins differ from that of Nrf2⅐p18.Although the dimerization partner(s) has not been identified, we have nonetheless provided compelling evidence for a role of Nrf2 in ho-1 gene activation by cadmium.
Activation of the three MAPK subfamilies by cadmium has been observed in other cells (26,35,36), although, to the best of our knowledge, this is the first demonstration of the activation of all three pathways in a single cell type.In principle, each activated pathway can lead to the expression of a different, but possibly overlapping, subset of cadmium-responsive genes.For instance, in rat mesangial cells, induction of the c-fos gene by cadmium is mediated primarily by the ERK pathway (36), whereas activation of either p38 or ERK enzymes can lead to induction of the hsp70 gene in rat brain tumor cells (26).Clearly, in MCF-7 cells, cadmium-dependent ho-1 gene activation is mediated by the p38, but not the ERK or JNK, pathway.This conclusion is supported not only by studies with kinase inhibitors and kinase mutants but, in retrospect, could be predicted from the MAPK activation profiles.For instance, whereas cadmium potently stimulates HO-1 mRNA accumulation (ϳ300 -400-fold above basal levels), it only weakly activates ERK1 and ERK2.Furthermore, in mouse hepatoma cells, nearly maximal rates of ho-1 gene transcription are attained within 1 h of exposure to CdCl 2 (20), a time point at which JNK activation is not detected in MCF-7 cells.Thus, the characteristics of ho-1 gene regulation are most consistent with the temporal and quantitative patterns of p38 activation.
Because MAPKs can directly phosphorylate and activate transcription factors, the signaling pathway(s) utilized for regulation of a specific cadmium-responsive gene may ultimately depend on the transcription factor responsible for gene induction and its ability to serve as a substrate for the MAPKs.Although SB203580 can inhibit Nrf2 activity, direct phosphorylation of Nrf2 by p38 remains to be tested.Of course, p38 may stimulate Nrf2 activity by indirect mechanisms.For example, Itoh et al. (37) have recently identified a protein termed Keap1 that inhibits Nrf2 activity, presumably by cytoplasmic retention of the transcription factor through physical interaction.Electrophilic agents antagonize Keap1 repression by potentiating translocation of Nrf2 to the nucleus.In analogy with the mechanism of activation of nuclear factor-B, cadmium-activated p38 may promote dissociation of the putative Nrf2⅐Keap1 complex by phosphorylation of Keap1, either directly or through intermediary kinases, thus permitting traversal of the liberated Nrf2 to the nucleus and activation of target genes.
Our conclusion that induction of the ho-1 gene by CdCl 2 is mediated largely by the p38 MAPK pathway contradicts that of Masuya et al. (12), who observed no effect of SB203580 on HO-1 mRNA accumulation in response to cadmium, heme, or arsenite in HeLa cells.The reason for this discrepancy is not presently clear.Whether this difference results from cell-specific variations in the induction mechanism cannot be properly gauged as the activation of p38 by cadmium or the other inducers was not examined in HeLa cells.The inability of SB203580 or the p38 dominant-negative mutant to completely abolish ho-1 gene and/or pE1-luc inductions in MCF-7 cells, respectively, suggests either that these effectors do not completely inhibit endogenous p38 activity under the conditions utilized or that other undefined signaling mechanisms contribute to the overall activation of the ho-1 gene by cadmium.
Superficially, some of the results presented here suggest that signaling via the ERK pathway negatively regulates HO-1 expression.For instance, low concentrations of PD098059 slightly enhance cadmium-mediated HO-1 mRNA accumulation (Fig. 4) and Nrf2 trans-activation of pE1-luc (Fig. 11).Similarly, overexpression of the ERK1 mutant stimulates basal and cadmium-dependent pE1-luc activity (Fig. 6).Without additional data, however, a definitive conclusion with regard to the negative regulation by this pathway is premature.Nonetheless, it is clear that the ERK pathway does not contribute positively to cadmium-dependent gene activation in MCF-7 cells.This result is in contrast to the mechanism of ho-1 gene induction by other agents.In human fibroblasts, phorone and diethyl maleate selectively activate ERK1 and ERK2, and inhibition of signaling through this pathway by PD098059 attenuates HO-1 mRNA accumulation in response to these agents (38).Furthermore, in chicken hepatoma cells, both the ERK and p38 pathways are utilized for ho-1 gene activation by arsenite (13).Interestingly, in contrast to our observations in MCF-7 cells, AP-1 factors (c-Fos and c-Jun, respectively) have been implicated in induction of the ho-1 genes by these agents, providing a plausible explanation for the requirement of different signaling pathways.The utilization of different signal transduction pathways and transcription factors for inducerdependent ho-1 gene induction is consistent with our earlier proposal (39) that such regulation is mediated by multiple subfamilies of bZIP proteins interacting with the StREs, possibly in an inducer-specific or inducer-selective manner.Of course, reconciliation of such a proposal with the nearly universal inhibition of inducer-dependent ho-1 gene activation by Nrf2M will necessitate further investigation.

FIG. 1 .
FIG. 1.The E1 enhancer is activated by cadmium in MCF-7 cells.A, the 5Ј-flanking region of the mouse ho-1 gene indicating the location of the E1 and E2 enhancers.The indicated cells (B) or MCF-7 cells (C and D) were transfected with a DNA mixture consisting of 3 g of pE1-luc (B) or the indicated luciferase plasmid (C and D) and 2 g of pCMV␤-gal and treated with CdCl 2 (B and D) or the indicated agent (C).Transfection, induction, and enzyme assays were carried out as described under "Experimental Procedures." Background luciferase activity (from mock-transfected cells) was subtracted from each experimental measurement, and the resulting value was corrected for variation in transfection efficiency by normalization with background-subtracted ␤-galactosidase activity in the same cell extract.Each data bar represents the average Ϯ S.D. from three to five independent experiments.Cd, CdCl 2 (10 M); heme (10 M); Ars, sodium arsenite (50 M); TBHQ (50 M); TPA (100 ng/ml).WT, wild type.

FIG. 3 .
FIG.3.Effect of PD098059 and SB203580 on pE1-luc expression.Each well of MCF-7 cells was transfected with a mixture of plasmids pE1-luc (3 g) and pBluescript II SKϪ (2 g) as described under "Experimental Procedures."The indicated concentration of kinase inhibitor was added to the culture media 1 h prior to the addition of vehicle or CdCl 2 (10 M).Cell extract equivalent to 5 g of protein was used for measurement of luciferase activity.Each data point, presented as % of luciferase activity in the absence of kinase inhibitors, represents the average Ϯ S.D. from four independent experiments.

FIG. 8 .
FIG. 8. Nrf2M, but not of c-JunM, inhibits induction of the ho-1 gene by cadmium and other agents.A, identification of MCF-7 stable transfectants expressing Nrf2M or c-JunM.Stable transfection, selection, and clonal isolation were carried out as described under "Experimental Procedures."Nuclear (Nrf2M; lanes a,c, e, g, i, and k), cytoplasmic (Nrf2M; lanes b, d, f, h, j, and l), or total cellular extracts (cJunM) were prepared from clones transfected with the dominantnegative mutant expression plasmids or empty vector (Nrf2M; lanes k and l; cJunM; lanes a and b).Western blot analysis was carried out as described under "Experimental Procedures," and the filter was exposed to film for 1-5 min.The size (kDa) and migration of the molecular weight standards are indicated.B, RNA analyses.Control (pEF), Nrf2M-and cJunM-expressing MCF cells were plated and treated with vehicle (Veh), CdCl 2 (Cd, 10 M), heme (10 M), sodium arsenite (Ars, 50 M), TBHQ (50 M), or TPA (100 ng/ml) for 3 h.RNA isolation, dot blot analyses and signal normalization were carried out as described under "Experimental Procedures."HO-1 mRNA levels, relative to the amount in vehicle-treated control cells, are presented and represent the average of two independent experiments.

FIG. 9 .
FIG. 9. Correlation between cadmium responsiveness and Nrf2 trans-activation of wild-type and mutant E1 enhancer constructs.A, sequence and position of the StREs in the 268-base pair E1 enhancer fragment.Mutations are highlighted in bold italics.Cadmium induction analyses (B and C) were carried out as described in the legend to Fig. 1B.Basal activities of E1 mutant constructs are normalized to that of wild-type pE1-luc.For trans-activation analysis (D), each well of MCF-7 cells was transfected with a DNA mixture consisting of 3 g of pE1-luc or the indicated mutant construct, 3 g of the Nrf2 expression plasmid or the empty vector (pEF), and 2 g of pCMV␤-gal.For each luciferase construct, fold trans-activation (normalized luciferase activity in the presence of Nrf2/activity in the absence of Nrf2) is presented.Each data bar represents the average Ϯ S.D. from three to five independent experiments.